Does a High-Protein Diet Damage Healthy Kidneys? What the Evidence Shows

Executive summary

• Glomerular hyperfiltration, a temporary rise in filtration rate after a protein meal, is a normal adaptive response in healthy kidneys, whose "renal reserve" shrinks once kidney disease, transplantation, or donation reduces nephron supply (Ref 2, Ref 3).

• In healthy, resistance-trained adults, controlled trials of high-protein diets (roughly 1.6-3.3 g/kg per day) lasting sixteen weeks found no significant harm to kidney or liver markers versus a lower-protein control (Ref 6, Ref 7).

• An 11-year cohort of 1,624 women found no meaningful link between protein intake and kidney function decline in those with normal kidney function at baseline, but a significant one in women already showing mild kidney impairment, strongest for non-dairy animal protein (Ref 9).

• Much of the theoretical basis for treating hyperfiltration as dangerous comes from an influential 1982 hypothesis paper built largely on rodent models of severely reduced nephron mass, not from trials of ordinary protein intake in people with two healthy kidneys (Ref 5).

• The landmark trial behind current low-protein guidance in chronic kidney disease (CKD) found no significant difference in the rate of kidney function decline between a normal-protein and a low-protein diet in its main analysis (Ref 14).

• The most recent Cochrane review of low-protein diets in CKD found the moderate restriction most patients are actually given made little or no difference to death or kidney failure; only a stricter, harder-to-sustain very-low-protein diet showed probable benefit, certainty graded low to moderate throughout (Ref 15).

• People living with one kidney after donation, carrying permanent compensatory hyperfiltration, show a small but real increase in lifetime kidney-failure risk versus equally healthy non-donors, though the trigger, losing half one's nephrons, differs from a dietary protein load on a full set (Ref 17, Ref 18).

• Obesity tracks with glomerular hyperfiltration in population data at least as strongly as protein intake does, a reminder of how easily this question is confounded once research moves from controlled trials to observational cohorts (Ref 10).

A high-protein or carnivore-style diet is one of the most common things patients are warned will damage their kidneys, and the warning given is usually identical to the advice for someone who already has diagnosed kidney disease: eat less protein. That collapses two different populations, and two different bodies of evidence, into one.

I should declare my starting point first: I run a private clinic, offer paid consultations, and generally eat and recommend an animal-based, higher-protein way of eating, which shapes how I read this literature. I have tried to cite conventional protein-restriction advice at its strongest, not its weakest. Every study mentioned is referenced below, limitations included.

Two populations, one blanket rule

Chronic kidney disease (CKD) is defined, in the guideline used across most of the world, as a reduction in kidney filtering capacity, or other evidence of kidney damage, persisting three months or more (Ref 1). That is a specific, diagnosed state, physiologically different from someone whose kidneys filter normally and simply eats more protein than average.

Both populations get the same soundbite: protein is hard on the kidneys. But that soundbite comes from two different bodies of research asking different questions: whether high protein harms kidneys that already filter normally, and whether reducing protein slows the decline of kidneys already damaged. Conflating the two is where most of the confusion, public and professional, comes from.

What actually happens to the kidneys when you eat protein

Filtration rate, whether measured directly or estimated as eGFR, is the standard measure of how much blood the kidneys clear per minute, and a protein meal reliably raises it, temporarily, in people with normal kidney function. This is repeatable enough to have its own clinical name, renal functional reserve, and its own test, which shows progressively smaller rises in people with existing kidney disease, kidney transplants, or a single donated kidney (Ref 3). Having spare capacity to call on is what healthy kidneys do; losing it is what compromised kidneys look like. The rise itself is not the injury, much as a muscle recruiting extra fibres under load is not damage; the problem is only if that workload becomes permanently and pathologically raised.

A widely cited review of the human and animal data on protein and renal function reached the same conclusion: hyperfiltration is a normal adaptive mechanism also seen in pregnancy and after a large meal, with no significant evidence that habitually high protein intake harms healthy kidneys, though restriction may still be appropriate once kidney disease is already present (Ref 2). The precise mechanics are still being worked out: in mice, the same hyperfiltration persisted even when the two leading proposed mechanisms were experimentally removed, an animal finding that is directional rather than proven for humans (Ref 4).

Where the idea that protein damages kidneys came from

The theoretical case for treating hyperfiltration as inherently dangerous rests substantially on one hugely influential 1982 paper, in which Barry Brenner and colleagues proposed that hyperfiltration itself causes haemodynamic injury to the glomerulus that accelerates scarring over time, an idea used to explain kidney disease progression in ageing, after surgical kidney-tissue removal, and in intrinsic kidney disease (Ref 5). It remains one of the most cited papers in nephrology, but it is a hypothesis, not a clinical trial: its central evidence came from rats with most of their kidney tissue surgically removed, extrapolated outward to explain disease more broadly. Applying that to someone with two intact kidneys who simply eats more protein than average is an inference the paper invites but does not test.

The evidence in people with healthy kidneys

So what happens when researchers actually test higher protein intakes in people who start out with normal kidney function?

Two controlled trials in resistance-trained men point the same way. In one, eleven men completed an eight-week crossover between their normal diet (2.6 g/kg/day of protein) and a deliberately higher one (3.3 g/kg/day), with no significant change in blood lipids or kidney and liver markers (Ref 6). In the other, 48 men were randomised to a moderate (1.6 g/kg/day) or high (3.2 g/kg/day) protein intake over 16 weeks; some markers rose modestly within the higher-protein groups over time, with no significant difference between arms (Ref 7). Both are small, short trials rather than long-term safety data, and the second is too weak a signal to call a clean "no effect."

For context, the International Society of Sports Nutrition's position stand puts 1.4 to 2.0 g/kg/day as sufficient for most people who exercise regularly, rising to 2.3 to 3.1 g/kg/day for preserving muscle in resistance-trained people eating below maintenance calories, with newer evidence for benefits above 3.0 g/kg/day (Ref 8). A carnivore or heavily animal-based diet often sits at or above that range, simply because meat is the only food on the plate and satiety, not a macro target, regulates intake.

The largest, longest relevant data come from the Nurses' Health Study, which followed 1,624 women for 11 years, relating measured protein intake to the change in estimated GFR. In women with normal kidney function at baseline, higher protein intake was not significantly associated with faster decline: an extra 10 g of daily protein tracked with a change of just 0.25 mL/min/1.73m2 over the following year, small and uncertain enough to be indistinguishable from no effect (Ref 9). This is an observational cohort, not a trial, so it cannot fully exclude confounding, relied on food-frequency questionnaires rather than measured intake, and the women were overwhelmingly white, limiting how far the finding generalises. But it is a genuinely large, long-run dataset whose normal-kidney-function finding lines up with the smaller trials above.

Two further data points complicate a simple story. In a Swiss survey of 1,339 adults, hyperfiltration tracked far more strongly with body mass index than with protein or sodium intake, with odds roughly four times higher in obesity than in lean participants (Ref 10). A small bariatric-surgery cohort found GFR fell substantially post-surgery, but a controlled low-protein diet phase did not explain it; the surgery and falling sodium intake did (Ref 11). Hyperfiltration in the real world is entangled with body weight at least as much as with protein, exactly the confounding that makes this topic easy to oversimplify in either direction.

Established chronic kidney disease is a different calculus

None of the above licenses the same reassurance once a diagnosed, staged loss of kidney function is present. Surviving nephrons then do more work each, and protein restriction's rationale is that it lowers glomerular pressure via the same feedback signal implicated above, reducing the load on nephrons with little reserve left (Ref 12, Ref 13).

The evidence for how much this changes outcomes is more equivocal than most patient advice suggests. The trial much of current practice is built on, the Modification of Diet in Renal Disease (MDRD) study, randomised 840 people with CKD across two sub-studies. In the larger, moderate-disease group, a normal-protein diet (1.3 g/kg/day) and a low-protein diet (0.58 g/kg/day) produced no significant difference in projected GFR decline at three years, though the low-protein group's decline was inconsistently faster at first, then slower, an anomaly its own authors could not explain. In the smaller, more advanced group, a very-low-protein diet (0.28 g/kg/day, with a keto-acid supplement) slowed GFR decline only marginally versus the low-protein diet, a difference falling just short of significance. Neither comparison delayed dialysis or reduced deaths (Ref 14).

The most recent Cochrane review pooled 17 trials and almost 3,000 participants. Comparing a moderate low-protein diet (0.5-0.6 g/kg/day) against a normal intake (0.8 g/kg/day or more), the comparison closest to what most patients are actually advised, it found little or no difference in death (moderate-certainty) or in kidney failure (low-certainty), with the effect on GFR itself rated very low certainty. Only a stricter comparison, a very-low-protein diet (0.3-0.4 g/kg/day, with a keto-acid supplement) against the low-protein diet, showed a clearer benefit: a probable reduction in kidney failure of roughly 165 fewer per 1,000 people, graded moderate certainty. Every included trial was also open-label, a real source of bias (Ref 15).

A separate review notes that recommended amounts still differ across guidelines; some evidence also suggests plant protein produces less hyperfiltration than animal protein (Ref 16), a question I would treat as open rather than settled.

Set against that trial evidence, the same Nurses' Health Study also included women with mild renal insufficiency at baseline, an estimated GFR between 55 and 80. Here the story reversed: each extra 10 g of daily protein tracked with a decline of 1.69 mL/min/1.73m2 over the following year, a real, statistically significant effect, stronger still for non-dairy animal protein specifically, a decline of 1.21 mL/min/1.73m2 per 10 g (Ref 9). This is, in my reading, the single cleanest piece of evidence in this entire topic: the same cohort, the same measurement method, showing no association in normal kidneys and a real one once function is mildly reduced. That is exactly the distinction this article argues for, and the strongest evidence here that protein amount matters more once reserve capacity is gone.

A natural experiment: living for decades with permanent hyperfiltration

Kidney donors offer an unusual, real-world test of sustained hyperfiltration: removing one kidney forces the other into permanent compensatory overwork. The trigger differs from a dietary protein load, since it comes from losing half of total nephron mass rather than processing more nitrogen through a full set. But donors are the closest thing to a decades-long experiment in this workload.

The best-designed comparison to date matched 96,217 US kidney donors against 20,024 equally healthy, equally screened non-donors. At 15 years, kidney failure had developed in about 31 per 10,000 donors against roughly 4 per 10,000 in the matched group, a difference small in absolute terms but statistically clear (Ref 17). An earlier, longer-follow-up study told an apparently opposite story, with donors developing kidney failure at a lower rate than the general population (Ref 18). But it compared donors against an unscreened population, while the newer study used a screened control. Together, they suggest hyperfiltration from a halved nephron supply carries a real, measurable cost over decades, small against a low baseline risk, and driven by a very different starting point: losing half one's nephrons outright.

Why this question is easy to get wrong from population data

Much of the public messaging that protein damages the kidneys draws on food-frequency-questionnaire epidemiology rather than controlled experiments. That kind of data struggles to separate reverse causation (people who feel unwell change their diet before diagnosis), confounding by adiposity and dietary pattern (Ref 10), and simple measurement error. A well-designed cohort study that grapples honestly with its confounders is worth more than a poorly controlled trial, and the reverse is also true.

What this means in practice

For someone with normally functioning kidneys, the evidence reviewed here does not show that a high-protein, carnivore-style diet damages healthy kidneys over the periods studied. Hyperfiltration here behaves like the adaptive, reversible response the physiology literature describes, not a slow injury accumulating unnoticed. But I want to be honest about the limits of that reassurance: nobody has run a genuinely long-term, multi-decade controlled trial of a strict carnivore pattern, as opposed to "high protein added to an otherwise ordinary diet," against hard kidney outcomes. The honest position is that the evidence which exists shows no signal of harm, not that lifetime safety has been proven, because that trial does not exist for any dietary pattern, in either direction.

For someone with an existing, diagnosed loss of kidney function, the calculation is different, and nothing above should be read as an argument against the protein target a treating clinician sets for that person. The mechanism behind restriction is real. The trial evidence for how much it changes outcomes is genuinely mixed, graded low to moderate certainty in the best current synthesis, and stronger for very-low-protein approaches needing medical supervision than for the moderate restriction most people are actually given. That is a conversation for the nephrologist or renal dietitian managing that person's actual numbers and trajectory, not something to reason through from a blog post, and nothing here should be read as advice to change a prescribed renal diet.

Disclosures

I run a private clinic and offer paid consultations, so I have a commercial interest in readers finding this kind of analysis useful. I eat and generally recommend an animal-based, higher-protein way of eating, which predisposes me to read the healthy-kidney evidence charitably. I have tried to counter that by citing the trial and guideline evidence behind conventional protein-restriction advice at its strongest, including the numbers that do not favour my own dietary lens, rather than only its weaknesses. Nothing in this article is individual medical advice, and it is not a substitute for the guidance of a nephrologist or renal dietitian for anyone with diagnosed kidney disease. If you would like help applying any of this to your own circumstances I offer consultations, though this is well-trodden ground for many good clinicians and you certainly do not have to see me.

References

1. Stevens PE, Levin A. Evaluation and management of chronic kidney disease: synopsis of the Kidney Disease: Improving Global Outcomes 2012 clinical practice guideline. Ann Intern Med. 2013;158(11):825-830. PMID: 23732715. https://doi.org/10.7326/0003-4819-158-11-201306040-00007

2. Martin WF, Armstrong LE, Rodriguez NR. Dietary protein intake and renal function. Nutr Metab (Lond). 2005;2:25. PMID: 16174292. https://doi.org/10.1186/1743-7075-2-25

3. Loo CS, Zaki M, Sulaiman AB, Sukanya AB, Voon YC, Kua SL. Acute protein loading in the assessment of renal reserve. Med J Malaysia. 1994;49(1):36-43. PMID: 8057988. https://pubmed.ncbi.nlm.nih.gov/8057988/

4. Sällström J, Carlström M, Olerud J, et al. High-protein-induced glomerular hyperfiltration is independent of the tubuloglomerular feedback mechanism and nitric oxide synthases. Am J Physiol Regul Integr Comp Physiol. 2010;299(5):R1263-R1268. PMID: 20739607. https://doi.org/10.1152/ajpregu.00649.2009

5. Brenner BM, Meyer TW, Hostetter TH. Dietary protein intake and the progressive nature of kidney disease: the role of hemodynamically mediated glomerular injury in the pathogenesis of progressive glomerular sclerosis in aging, renal ablation, and intrinsic renal disease. N Engl J Med. 1982;307(11):652-659. PMID: 7050706. https://doi.org/10.1056/NEJM198209093071104

6. Antonio J, Ellerbroek A, Silver T, Vargas L, Peacock C. The effects of a high protein diet on indices of health and body composition--a crossover trial in resistance-trained men. J Int Soc Sports Nutr. 2016;13:3. PMID: 26778925. https://doi.org/10.1186/s12970-016-0114-2

7. Bagheri R, Kargarfard M, Sadeghi R, Scott D, Camera DM. Effects of 16 weeks of two different high-protein diets with either resistance or concurrent training on body composition, muscular strength and performance, and markers of liver and kidney function in resistance-trained males. J Int Soc Sports Nutr. 2023;20(1):2236053. PMID: 37516903. https://doi.org/10.1080/15502783.2023.2236053

8. Jäger R, Kerksick CM, Campbell BI, et al. International Society of Sports Nutrition Position Stand: protein and exercise. J Int Soc Sports Nutr. 2017;14:20. PMID: 28642676. https://doi.org/10.1186/s12970-017-0177-8

9. Knight EL, Stampfer MJ, Hankinson SE, Spiegelman D, Curhan GC. The impact of protein intake on renal function decline in women with normal renal function or mild renal insufficiency. Ann Intern Med. 2003;138(6):460-467. PMID: 12639078. https://doi.org/10.7326/0003-4819-138-6-200303180-00009

10. Ogna A, Forni Ogna V, Bochud M, et al. Association between obesity and glomerular hyperfiltration: the confounding effect of smoking and sodium and protein intakes. Eur J Nutr. 2016;55(3):1089-1097. PMID: 25971845. https://doi.org/10.1007/s00394-015-0923-0

11. Friedman AN, Quinney SK, Inman M, Mattar SG, Shihabi Z, Moe S. Influence of dietary protein on glomerular filtration before and after bariatric surgery: a cohort study. Am J Kidney Dis. 2014;63(4):598-603. PMID: 24387796. https://doi.org/10.1053/j.ajkd.2013.11.012

12. Patel J, Kalantar-Zadeh K, Joshi S. Low-Protein Diets and Its Synergistic Role in the SGLT2 Inhibitor Era. Adv Kidney Dis Health. 2023;30(6):523-528. PMID: 38453269. https://doi.org/10.1053/j.akdh.2023.12.005

13. Koppe L, Fouque D. The Role for Protein Restriction in Addition to Renin-Angiotensin-Aldosterone System Inhibitors in the Management of CKD. Am J Kidney Dis. 2019;73(2):248-257. PMID: 30149957. https://doi.org/10.1053/j.ajkd.2018.06.016

14. Klahr S, Levey AS, Beck GJ, et al. The effects of dietary protein restriction and blood-pressure control on the progression of chronic renal disease. Modification of Diet in Renal Disease Study Group. N Engl J Med. 1994;330(13):877-884. PMID: 8114857. https://doi.org/10.1056/NEJM199403313301301

15. Hahn D, Hodson EM, Fouque D. Low protein diets for non-diabetic adults with chronic kidney disease. Cochrane Database Syst Rev. 2020;10(10):CD001892. PMID: 33118160. https://doi.org/10.1002/14651858.CD001892.pub5

16. Sakaguchi Y, Kaimori JY, Isaka Y. Plant-Dominant Low Protein Diet: A Potential Alternative Dietary Practice for Patients with Chronic Kidney Disease. Nutrients. 2023;15(4):1002. PMID: 36839360. https://doi.org/10.3390/nu15041002

17. Muzaale AD, Massie AB, Wang MC, et al. Risk of end-stage renal disease following live kidney donation. JAMA. 2014;311(6):579-586. PMID: 24519297. https://doi.org/10.1001/jama.2013.285141

18. Ibrahim HN, Foley R, Tan L, et al. Long-term consequences of kidney donation. N Engl J Med. 2009;360(5):459-469. PMID: 19179315. https://doi.org/10.1056/NEJMoa0804883

Grey Hair: The Real Biology Behind Greying, and What Reversal Evidence Shows

Executive summary

• Grey hair begins with melanocyte stem cells held in reserve inside the hair follicle; hair loses colour once that reserve is damaged, exhausted, or outpaced by demand (Ref 1, Ref 2).

• The clearest mechanism in the human literature is not vague "oxidative stress" but hydrogen peroxide building up as catalase collapses, chemically disabling tyrosinase, the pigment-making enzyme (Ref 3, Ref 4).

• A widely cited mouse study shows acute stress can rapidly deplete melanocyte stem cells via a burst of noradrenaline, not yet demonstrated the same way in people (Ref 5).

• Family history of early greying is the single strongest predictor across the largest surveys available; smoking, body weight, and metabolic risk markers are repeatedly associated too, though only in cross-sectional data that cannot establish causation, and one dataset found smoking running the opposite direction (Ref 6, Ref 7, Ref 8, Ref 9, Ref 10).

• Nutrient-deficiency associations, vitamin B12 particularly, recur across the literature, but individual studies disagree with each other; I have set out that disagreement rather than the tidiest version (Ref 11, Ref 12, Ref 13, Ref 14).

• Reversal is real but narrow: some hairs naturally regain colour tracking with falling life stress, and case reports describe similar repigmentation from cancer drugs. Neither supports a marketed fix (Ref 16, Ref 17, Ref 18).

• I have no commercial interest in any hair-pigment product, dye, serum, or supplement. Everything below was retrieved and checked through PubMed.

Introduction

Grey hair is one of the most searched, most monetised, and least honestly discussed topics in cosmetic medicine, sold with more confidence than the biology supports. My aim here is the opposite of a sales pitch: to lay out what the melanocyte biology and epidemiological studies actually show, including where they disagree, and to be explicit about the difference between mouse and human evidence. I have no financial interest in any product marketed to prevent, slow, or reverse grey hair, and nothing to sell you here.

I trained first in dentistry, then in dermatology and, separately, anti-aging and metabolic medicine, fields that overlap in hair pigmentation, which tracks with the same metabolic markers anti-aging medicine already cares about. That overlap is why I find the topic interesting, not any product I have to move.

What grey hair actually is

Hair colour comes from melanin, made by cells called melanocytes and loaded into the hair shaft as it grows. Human hair follicles produce two forms: eumelanin, giving black and brown tones, and pheomelanin, giving red and auburn tones (Ref 1).

The melanocytes doing day-to-day pigment production are not permanent; they are topped up from a smaller population of melanocyte stem cells held in a specific compartment of the follicle. A landmark study using melanocyte-tagged mice, cross-checked against ageing human hair follicles, showed hair turns grey when this stem cell population fails to maintain itself over repeated growth cycles, rather than simply running out of raw pigment (Ref 2). The same study found this is dramatically accelerated when a survival gene called BCL2 is lost, causing the melanocyte stem cells specifically, not the mature pigment cells, to die off within their niche.

A follicle whose reserve is fully gone behaves very differently from one where it is merely struggling, a distinction that recurs throughout this piece.

The mechanism worth taking seriously: free radical damage, not "oxidative stress"

You will see "oxidative stress" used constantly in relation to grey hair, but there is no single agreed clinical test for it. What the better studies actually measure is more concrete: specific reactive molecules, at specific concentrations, overwhelming specific, measurable defence enzymes. That is free radical damage, worth naming precisely because the precision is where the real biology lives.

Making melanin is an oxidative process in itself, throwing off hydrogen peroxide as a by-product that a young follicle's catalase clears efficiently. A review of the ageing literature describes how, at the melanocyte stem cell niche, oxidative defences such as BCL2 decline with age, catalase activity falls sharply, and hydrogen peroxide accumulates, contributing to melanocyte malfunction and death (Ref 3). That review synthesises mechanistic work across several follicle compartments rather than a single trial, a map of where the evidence points rather than a definitive experiment.

The more striking human data comes from a study using Raman spectroscopy on living scalp hair, finding grey and white hair shafts accumulate hydrogen peroxide at millimolar concentrations, alongside an almost complete absence of catalase and of the repair enzymes that normally fix oxidised proteins throughout the follicle. Critically, this damage lands on a specific site: a methionine residue in the active site of tyrosinase, the master enzyme of pigment production, measurably impairing its function (Ref 4). The pigment factory is not just short of raw material; its own key machine is being chemically jammed by a waste product it can no longer clear.

Two honest caveats: the damage could be prevented from forming in a test tube using methionine itself, but that was shown in vitro, not in living human scalp, so it is not evidence that eating or applying methionine reverses anything in a person; and this describes what differs between grey and pigmented hair, correlational at the tissue level, not proof of which change happens first.

Stress and grey hair: mouse evidence only

A widely cited study found acute stress triggers a burst of noradrenaline from the sympathetic nerves supplying the melanocyte stem cell niche, driving those stem cells to proliferate, differentiate, migrate, and permanently disappear; blocking that proliferation prevented the stress-induced greying entirely (Ref 5). This was done entirely in mice, with no equivalent controlled human experiment, so whether the same pathway drives stress-associated greying in people remains a plausible inference, not a demonstrated fact. Human studies linking self-reported stress to greying, discussed below, rely on subjective questionnaires rather than any objective measure of stress, inherently weaker evidence than the mouse mechanism itself.

Genetics: the strongest predictor of all

Whatever else is true about smoking, stress, or diet, the largest studies agree: family history dominates every other factor measured. A study of 6,390 men under 30 found 25.3% had premature greying; family history carried an odds ratio of 12.82, nearly thirteen times the risk, dwarfing obesity (2.61) and a smoking history of more than five pack-years (1.61). A separate severity analysis found family history (2.63) and obesity (2.22) also tracked with it (Ref 6). The main limitation is that this was a questionnaire-based study in men only, carrying the usual recall-bias risk, and telling us nothing about women.

A separate, smaller study of 467 people broke family history down by relative: a mother's history carried an odds ratio of 3.165, a father's 5.166, a maternal grandparent's 2.442, a paternal grandparent's 2.369, and a sibling's 3.125, all statistically significant. The same study found iron deficiency modestly associated (odds ratio 1.751), but, oddly, found smoking negatively associated, the opposite direction from most other studies (Ref 7), most plausibly because of a smaller, self-selected, mostly female sample rather than evidence that smoking protects against greying, but worth showing rather than smoothing over.

None of this is a licence to treat premature greying as inevitable if it runs in your family; genetics sets the odds and, to some extent, the timing window.

Smoking, body weight, and metabolic health

A systematic review of 32 studies found hair loss and premature greying more prevalent among smokers, consistent with nicotine and its by-products accumulating in the hair shaft (Ref 8); this cannot establish causation, and, as noted above, at least one dataset ran the other way.

A study of 2,644 medical students and resident physicians found premature greying more common among those obese or overweight, and found higher self-reported stress linked to earlier onset (95% versus 90% versus 86% before age 25) (Ref 9); this is a cross-sectional, self-reported dataset skewed toward a medically literate population, and, as above, "stress level" means a self-report, not a measured quantity.

A separate study of 1,929 young adults found the premature-greying group had a slightly larger waist circumference (76.3cm versus 74.3cm), higher blood pressure, higher fasting glucose (91.6mg/dl versus 90.8mg/dl), and lower HDL cholesterol (65.4mg/dl versus 68.1mg/dl), all statistically significant given the sample size, with two or more of these markers together independently associated with premature greying (odds ratio 1.725) (Ref 10) — small differences individually, more persuasive as a cluster.

None of these markers is insulin or a direct insulin-resistance index, but the cluster overlaps with what I would elsewhere call early physiological insulin resistance, a plausible extrapolation of my own that the study did not test.

Nutrient deficiencies: a literature that keeps disagreeing with itself

Vitamin B12 is the deficiency most persistently linked to premature greying, worth walking through rather than accepting as settled fact, because it is not settled.

One study of 35 school-age children, cases and controls combined, found serum calcium, ferritin, and vitamin D low in the premature-greying group; B12 was measured but not among the significant findings (Ref 11). Against that, a retrospective analysis of 71 cases of premature canities in India found a positive family history in 90.1% of cases, and low vitamin B12 and hypothyroidism significantly associated with the condition; anaemia, ferritin, and fasting glucose were not (Ref 12). A separate case-control study of 52 patients under 20 found serum B12, folic acid, and biotin all significantly lower in cases than matched controls (Ref 13). Three studies, three overlapping but not identical sets of significant findings: a genuine feature of this literature, not something to smooth over.

At least one clinical case report describes premature greying as a presenting feature that led clinicians to an underlying B12 deficiency (Ref 14); a single case report cannot tell you how often this happens, but it shows the association is not purely statistical noise.

Copper deserves a mention on mechanistic grounds, though I have not found controlled human trial data linking copper deficiency to premature greying the way B12 has been studied. Tyrosinase requires copper at its active site, and laboratory work shows copper ions accelerate a downstream step in eumelanin synthesis, the conversion and oxidation of an intermediate called dopachrome (Ref 15): biochemistry showing why copper plausibly matters, not a clinical trial showing that correcting a shortfall regrows pigment.

Given how inconsistent these studies are, the reasonable conclusion is not "take a B12 supplement to fix your grey hair," but that a small subset of people have an identifiable, correctable deficiency worth a blood test. A diet built around well-sourced animal foods, meat, organ meats, seafood, dairy, and eggs if tolerated, for B12; red meat and organ meats for heme iron, far more bioavailable than any plant source; shellfish, oysters especially, and liver for copper, makes a genuine shortfall considerably less likely. That is a statement about where these nutrients are most reliably found, not a promise that eating this way will turn a grey hair dark again.

Can grey hair actually be reversed?

A systematic review and meta-analysis of the premature-greying literature found that treating the underlying condition, a genuine deficiency, an endocrine disorder, and so on, reversed greying in a meaningful number of cases reviewed, though it rated the pharmacological agents it assessed, such as calcium pantothenate and para-aminobenzoic acid, as only low-grade evidence (Ref 16), supporting the idea of correcting a real, identified cause rather than taking a product marketed at the hair itself.

The most interesting human evidence on natural reversal comes from a study mapping pigment along individual hair shafts like a timeline, finding hairs that naturally regained pigment across sexes, ages, and body regions, with regaining hairs showing higher levels of proteins involved in energy metabolism and antioxidant defence, in line with the free-radical mechanism described earlier. The same group tracked this alongside self-reported life stress and found reversal moving in parallel with it, with modelling suggesting a threshold below which the stem cell population can recover (Ref 17). This remains observational, since life stress was self-reported, so it shows reversal happens and tracks stress, not that reducing yours will reverse your own grey hair on a predictable timeline.

A different line of evidence comes from oncology: case reports describe hair repigmenting, sometimes dramatically, in patients treated with tyrosine kinase inhibitors, a class of cancer drug, apparently as a side effect on melanogenesis pathways elsewhere. One well-documented example, in the New England Journal of Medicine, describes marked repigmentation in a patient treated with nilotinib (Ref 18); similar single-case reports exist for other drugs in the same class, each the lowest tier of clinical evidence, and none appropriate for cosmetic reasons given their serious side effects. I mention this only as confirmation that a greyed follicle has not always permanently lost the capacity to produce pigment.

Put together: reversal is real, documented through more than one independent line of evidence, and depends on whether the stem cell reserve is exhausted or merely suppressed. Nothing in the literature reviewed here supports a specific consumer product reliably reversing established grey hair, and I am not aware of controlled trial evidence for any supplement, serum, or shampoo achieving this.

What I would actually do: foundations, not a fix

If there is no reliable product-based reversal, the sensible response is the one that applies to most of ageing biology: get the foundations right, and treat anything beyond that as a genuine unknown rather than a marketed certainty.

Well-sourced animal foods, meat, organ meats, seafood, dairy, and eggs if tolerated, supply B12; red meat and organ meats supply heme iron; shellfish, oysters particularly, and liver supply copper, none of it requiring a supplement aisle if the diet is already built this way. Smoking cessation stands on its own, given how consistently it turns up across this literature.

If greying has arrived early and fast, a short panel with your doctor covering vitamin B12, iron studies, and thyroid function, since hypothyroidism recurs here too, is more useful than a serum. If those come back normal, it is fair to ask whether "normal" means optimal for you or simply inside a wide population range; this is a request for information, not a treatment plan. Beyond that, the metabolic-health thread running through this piece is worth taking seriously for reasons that have nothing to do with your hair colour.

Conclusion

Grey hair is not a single problem with a single fix. It is the visible endpoint of melanocyte stem cells running down through more than one route: free radical damage from hydrogen peroxide as catalase declines, a stress-linked mechanism shown clearly in mice and only inferred in humans, a genetic ceiling that matters more than any other factor measured, inconsistent but real nutrient-deficiency associations in a minority of cases, and a metabolic-health signal anti-aging medicine already cares about. Reversal is genuinely documented, but threshold-dependent rather than universally available, and nothing in the current literature supports a specific product delivering it reliably. If you take one thing from this piece, take the distinction between a follicle that is struggling and one that is empty: almost everything worth doing here is aimed at the first, not the second.

Disclosures

I have no commercial interest in any product, brand, or supplement marketed to prevent, slow, or reverse grey hair, and no affiliate arrangement or discount code tied to this piece. I offer private consultations for people who want to think through their own bloods and their own situation in more depth, though there are many excellent clinicians who can help with exactly this, and seeing one of them is just as valid a route as seeing me. Nothing in this piece is individual medical advice. Any decision about testing, supplementation, or an existing prescription should be made with your own doctor, who knows your history in a way that an article cannot.

References

Identified and verified via PubMed; DOIs included.

1. Triwongwaranat D, Thuangtong R, Arunkajohnsak S. A review of the etiologies, clinical characteristics, and treatment of canities. International Journal of Dermatology, 2019. https://doi.org/10.1111/ijd.14399

2. Nishimura EK, Granter SR, Fisher DE. Mechanisms of hair graying: incomplete melanocyte stem cell maintenance in the niche. Science, 2005. https://doi.org/10.1126/science.1099593

3. Seiberg M. Age-induced hair greying, the multiple effects of oxidative stress. International Journal of Cosmetic Science, 2013. https://doi.org/10.1111/ics.12090

4. Wood JM, Decker H, Hartmann H, et al. Senile hair graying: H2O2-mediated oxidative stress affects human hair color by blunting methionine sulfoxide repair. FASEB Journal, 2009. https://doi.org/10.1096/fj.08-125435

5. Zhang B, Ma S, Rachmin I, et al. Hyperactivation of sympathetic nerves drives depletion of melanocyte stem cells. Nature, 2020 (mouse study). https://doi.org/10.1038/s41586-020-1935-3

6. Shin H, Ryu HH, Yoon J, et al. Association of premature hair graying with family history, smoking, and obesity: a cross-sectional study. Journal of the American Academy of Dermatology, 2014 (n=6,390 men; questionnaire-based, recall bias possible, men only). https://doi.org/10.1016/j.jaad.2014.11.008

7. Thompson KG, Marchitto MC, Ly BCK, Chien AL. Evaluation of Physiological, Psychological, and Lifestyle Factors Associated with Premature Hair Graying. International Journal of Trichology, 2019 (n=467; survey-based, recall bias possible). https://doi.org/10.4103/ijt.ijt_43_19

8. Babadjouni A, Pouldar Foulad D, Hedayati B, Evron E, Mesinkovska N. The Effects of Smoking on Hair Health: A Systematic Review. Skin Appendage Disorders, 2021 (review of 32 studies). https://doi.org/10.1159/000512865

9. Aldamanhori IB, Alghamdi NJ, Alharbi SM, et al. Premature Hair Graying and Its Associated Factors Among Medical Students and Resident Physicians at Imam Abdulrahman Bin Faisal University. Healthcare (Basel), 2025 (n=2,644; cross-sectional, self-reported, medically literate sample). https://doi.org/10.3390/healthcare13101185

10. Paik SH, Jang S, Joh HK, et al. Association Between Premature Hair Greying and Metabolic Risk Factors: A Cross-sectional Study. Acta Dermato-Venereologica, 2018 (n=1,929; cross-sectional, associative not causal). https://doi.org/10.2340/00015555-2974

11. Bhat RM, Sharma R, Pinto AC, Dandekeri S, Martis J. Epidemiological and investigative study of premature graying of hair in higher secondary and pre-university school children. International Journal of Trichology, 2013 (n=35 total, cases and controls combined; small sample). https://doi.org/10.4103/0974-7753.114706

12. Sonthalia S, Priya A, Tobin DJ. Demographic Characteristics and Association of Serum Vitamin B12, Ferritin and Thyroid Function with Premature Canities in Indian Patients from an Urban Skin Clinic of North India: A Retrospective Analysis of 71 Cases. Indian Journal of Dermatology, 2017 (retrospective, single-clinic). https://doi.org/10.4103/ijd.IJD_221_17

13. Daulatabad D, Singal A, Grover C, Chhillar N. Prospective Analytical Controlled Study Evaluating Serum Biotin, Vitamin B, and Folic Acid in Patients with Premature Canities. International Journal of Trichology, 2017 (n=52 cases, under 20 years old). https://doi.org/10.4103/ijt.ijt_79_16

14. Yoshida H, Kanamori M. Reappraisal: Premature Graying of Hair. American Journal of Medicine, 2020 (single case report). https://doi.org/10.1016/j.amjmed.2020.09.044

15. Ito S, Suzuki N, Takebayashi S, Commo S, Wakamatsu K. Neutral pH and copper ions promote eumelanogenesis after the dopachrome stage. Pigment Cell & Melanoma Research, 2013 (in vitro biochemistry, not a clinical deficiency study). https://doi.org/10.1111/pcmr.12137

16. Mahendiratta S, Sarma P, Kaur H, et al. Premature graying of hair: Risk factors, co-morbid conditions, pharmacotherapy and reversal, a systematic review and meta-analysis. Dermatologic Therapy, 2020 (review; pharmacotherapy evidence graded low, 2A). https://doi.org/10.1111/dth.13990

17. Rosenberg AM, Rausser S, Ren J, et al. Quantitative mapping of human hair greying and reversal in relation to life stress. eLife, 2021 (observational; life stress self-reported). https://doi.org/10.7554/eLife.67437

18. Kockerols CCB, Westerweel PE. Hair Repigmentation Induced by Nilotinib. New England Journal of Medicine, 2022 (single case report). https://doi.org/10.1056/NEJMicm2119953

mTOR, Rapamycin and Longevity: What the Animal Data Do and Don’t Show

Executive summary

• mTOR (mechanistic target of rapamycin) is an ancient, highly conserved nutrient-sensing enzyme that works inside two different assemblies, mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2). It is essential machinery for growth, muscle building and metabolic control, not a fault to be switched off (Ref 1).

• Feeding, and in particular the amino acid leucine, activates mTORC1 to drive protein synthesis, while fasting lowers its activity and releases the cellular recycling programme called autophagy (Ref 1, Ref 2).

• Rapamycin, a drug that inhibits mTORC1, reproducibly extends lifespan in mice, including when treatment starts in later life, with a dose-dependent effect that is consistently larger in females (Ref 3, Ref 4). This is one of the most robust pharmacological longevity signals we have, and it is a rodent signal.

• Reducing a single downstream mTORC1 target, S6K1 (ribosomal protein S6 kinase 1), by genetic deletion also extends lifespan in mice, which argues that the pathway is genuinely wired into ageing biology rather than being an artefact of one drug (Ref 5).

• No randomized controlled trial has shown that rapamycin extends human lifespan or healthspan. The human-relevant data are short-term surrogate outcomes: better vaccine responses and fewer infections in older adults on low, intermittent dosing (Ref 6, Ref 7), a small dog pilot with echocardiographic changes (Ref 8), and a one-year safety trial in healthy adults whose primary outcome was neutral (Ref 9).

• The trade-offs are real. Continuous rapamycin also disrupts mTORC2 and can impair glucose tolerance and insulin action (Ref 10), it blocks the normal muscle-protein-synthesis response to resistance exercise in humans (Ref 11), and it is associated with reduced sperm counts and fertility in men (Ref 12), on top of recognised effects such as mouth ulcers and immune suppression.

• My read: the mechanism is genuinely fascinating and the mouse data are real, but the case that a healthy person should take rapamycin to live longer has not been made. Protecting muscle and metabolic health, and letting the normal feeding-and-fasting rhythm move mTOR up and down, is the higher-confidence approach. Nothing here is advice to take rapamycin.

Why I am writing this, and my bias

A popular claim runs roughly like this: protein is dangerous because it activates mTOR, mTOR activity shortens lifespan, therefore you should eat as little protein as possible and take rapamycin if you want to live longer. It is a tidy story, and tidy stories about biology are usually wrong.

I should be honest about where I stand before I make my case. I run a private clinic, I see patients through paid consultations, and longevity and metabolic medicine are areas I work in commercially. That is a bias, and you should weigh it. I have tried to do the opposite of what a biased writer usually does: I have gone looking for the strongest evidence against my instinct, cited it fairly, and only then explained where I think it stops short. The mouse lifespan data for rapamycin are real and I present them as real. My disagreement is not with the rodent findings, it is with the leap from a mouse cage to a human life.

A note on tone. This is a mechanism-heavy piece written to be readable by a curious member of the public and defensible to a specialist. I expand the jargon as I go. Nobody understands the mTOR network in full, myself included, so treat any confident-sounding claim about it, including mine, with appropriate suspicion.

What mTOR actually is

Picture an apple and you can see it instantly. Ask most people to picture mTOR and they cannot, which is a problem, because you cannot sensibly argue about switching something off if you cannot say what it is.

Physically, mTOR is a protein, specifically a serine/threonine kinase. A kinase is an enzyme that attaches a phosphate group to another protein, and that small chemical tag acts like a switch, turning the target protein on or off. Serine and threonine are simply the two amino acids on the target that receive the tag. So mTOR is a master switch-thrower.

It does not act alone. The same core enzyme is built into two different assemblies that do different jobs (Ref 1). Think of one engine dropped into two different machines:

• mTOR complex 1 (mTORC1) is defined by a partner protein called RAPTOR (regulatory-associated protein of mTOR). This is the build foreman. When it senses that nutrients are plentiful, it orders construction: make protein, grow, build muscle, and stop recycling old parts.

• mTOR complex 2 (mTORC2) is defined by a different partner, RICTOR (rapamycin-insensitive companion of mTOR). This one is closer to a logistics and maintenance manager. It helps regulate the cell's shape and survival signalling, and it sits in the pathway that handles insulin and glucose. Its name reflects the classic view that it is not blocked by rapamycin, though as we will see, continuous dosing can disrupt it too.

The reason both exist, and the reason both have been conserved from yeast through worms and flies to mammals across an enormous span of evolutionary time, is that a cell constantly has to make one decision: is now a time to build, or a time to conserve and maintain? mTOR is how it reads the room. An organism with no mTOR activity at all is not a long-lived organism, it is a dead one.

How feeding switches it on

One of the strongest activators of mTORC1 is the amino acid leucine. Leucine is like a tap on the shoulder that tells the foreman supplies have arrived. For years the puzzle was that mTORC1 cannot actually detect leucine by itself. The current understanding is that a separate sensor protein, Sestrin2, binds leucine and, in doing so, releases its brake on the pathway, so the signal converges on mTORC1 (Ref 2). This is a good example of how the textbook keeps changing: the tidy "RAPTOR recognises leucine" line I might have used a decade ago turned out to be more layered than that, and it may well be revised again.

Once mTORC1 is active, it drives protein manufacture through two well-mapped steps (Ref 1). It phosphorylates a protein called 4E-BP1 (eukaryotic translation initiation factor 4E-binding protein 1), which normally acts as a parking brake on protein translation, so tagging it releases the brake. It also activates S6K1 (ribosomal protein S6 kinase 1), which behaves like an accelerator, ramping up the ribosomes, the cellular factories that string amino acids into new proteins. The downstream result in muscle is more of the contractile proteins actin and myosin, which is literally what "more muscle" is built from.

At the same time, active mTORC1 suppresses autophagy, the cell's recycling and self-cleaning programme that breaks down worn-out components for reuse. This is the kernel of truth inside the anti-mTOR story: when you are fed, especially on protein, you are in build-and-store mode and autophagy is dialled down; when you fast, mTORC1 activity falls and autophagy is released. Feeding and fasting are, at the molecular level, largely a conversation about mTORC1 activity.

Here is the point the "switch it off" crowd miss. The interesting biology is not on or off, it is modulation, timing and location. Different tissues run mTOR at different settings in the same moment, sending build signals where growth is needed and conservation signals where it is not. A blanket instruction to inhibit it everywhere, all the time, is not a sophisticated longevity strategy. It is turning off the thermostat because you once felt too warm.

The animal evidence: strong, and mostly in rodents

Now to the evidence that launched the whole rapamycin-for-longevity movement, because it deserves to be taken seriously rather than waved away.

Inhibiting the TOR pathway extends lifespan in yeast, nematode worms and fruit flies. That much has been reproduced many times in invertebrates. The landmark mammalian result came in 2009 from the National Institute on Aging Interventions Testing Program, a deliberately rigorous design that runs the same experiment at three independent sites in genetically varied mice, precisely so that a fluke in one mouse strain at one laboratory cannot masquerade as a longevity drug. Rapamycin, started late in life at 600 days of age, extended both median and maximal lifespan in both sexes. Measured at the age by which 90% of the animals had died, the increase was about 14% in females and 9% in males (Ref 3). Disease patterns did not obviously change, and the authors noted the gain might come partly from postponing death from cancer, to which these mice are prone.

A follow-up raised the dose roughly threefold and saw a larger effect, extending median lifespan by about 23% in males and 26% in females, again with a bigger response in females (Ref 4). That study made a second important point: the metabolic and gene-expression fingerprint of rapamycin is not the same as that of dietary restriction, so rapamycin is not simply a dietary-restriction pill in disguise.

The drug is not the only line of evidence. Genetically deleting the single downstream target S6K1 also extended lifespan in mice and improved several healthspan measures, including bone, immune and motor function and insulin sensitivity, with a gene-expression pattern resembling dietary restriction (Ref 5). When a drug and an independent genetic manipulation of the same pathway both point the same way, the pathway is probably genuinely involved in ageing. I do not dispute that.

Now the disclaimer, and it is a large one that I will repeat rather than bury. Almost all of this is in laboratory rodents. These are animals with defined genetics, on controlled chow, in protected environments, frequently dying of cancers that rapamycin may specifically delay. Mice and humans differ in the pace of ageing, in cancer biology and in the entire lived context of a life. The history of geroscience is full of interventions that reliably extended rodent lifespan and then did nothing measurable, or worse, in people. "It works in mice" is the beginning of a hypothesis, not the end of an argument.

The gap to humans is the whole story

There is no randomized controlled trial showing that rapamycin extends human lifespan or healthspan, and it is worth being honest that a trial with a lifespan endpoint is close to impractical: it would need to run for decades on very large numbers of people. So the field leans on shorter surrogate outcomes, and those are the human data we actually have.

The most cited come from immune ageing. In a trial in older adults, the rapalog everolimus (RAD001, a close chemical relative of rapamycin) improved the antibody response to influenza vaccination by roughly 20% and reduced the proportion of PD-1-positive T cells, a receptor that dampens T-cell responses and becomes more common with age (Ref 6). A larger phase 2a trial in 264 elderly subjects used six weeks of low-dose, selective mTORC1 inhibition and reported a significant reduction in self-reported infections over the following year, along with an up-regulation of antiviral genes and a better vaccine response (Ref 7). This is a real and encouraging signal that briefly dialling down mTORC1 can rejuvenate an ageing immune system. It is also, and this matters, a surrogate: improved immune function over weeks is not demonstrated extra years of life, some endpoints were self-reported, and the work was industry-sponsored.

Companion dogs are a better model of human ageing than lab mice, because they share our homes, our air and our medical care. A small randomized trial gave 24 middle-aged healthy dogs a non-immunosuppressive dose of rapamycin for ten weeks and found no clinical side effects and echocardiographic hints of improved heart function (Ref 8). It is genuinely interesting, and it is a ten-week pilot in 24 animals measuring a surrogate. Larger and longer canine studies are now under way, which is the right next step.

The most directly relevant recent human data come from the PEARL trial, a 48-week, decentralized, double-blind, placebo-controlled study of intermittent low-dose rapamycin, 5 mg or 10 mg weekly, in healthy, normally ageing adults (Ref 9). The headline is worth sitting with: the primary outcome, visceral fat measured by DXA (dual-energy X-ray absorptiometry) body scanning, did not change significantly. Adverse events were similar to placebo, so the main positive finding is reassurance on short-term safety. Some secondary measures, such as lean tissue mass and self-reported pain, improved in subgroups (women on the higher dose), but these are secondary, partly self-reported, hypothesis-generating findings from a study whose primary endpoint was neutral. Read fairly, PEARL says low-dose intermittent rapamycin is relatively safe over a year in healthy people. It does not say it makes them live longer, and it did not hit its main target.

There is a deeper problem the sceptical case gets right. mTOR tone is set substantially by what and when we eat, and we cannot even agree on an optimal human diet. Layering a drug on top of a wildly variable, uncontrolled dietary background makes clean causal inference about longevity extremely hard. In a mouse on identical chow you can isolate the drug's effect. In free-living humans eating whatever they eat, you cannot, at least not yet.

The trade-offs enthusiasts tend to skip

If rapamycin were free of downsides the risk-benefit conversation would be easy. It is not.

Glucose and insulin. This is the elegant paradox of the field. Dietary restriction improves insulin sensitivity, and is thought to act partly by inhibiting mTORC1, yet chronic rapamycin does the opposite and impairs glucose tolerance. The resolution is that continuous dosing does not only hit mTORC1; over time it also disrupts mTORC2, and mTORC2 is required for insulin to suppress glucose production by the liver (Ref 10). Reassuringly for the longevity thesis, the same work showed the lifespan benefit (through mTORC1) can be separated from the glucose harm (through mTORC2), which is a large part of why the field has moved toward intermittent, lower dosing. But on standard continuous dosing the metabolic downside is real, and the long-term metabolic effect of intermittent dosing in healthy people is not yet established.

Muscle. This one goes to the heart of the original claim. In humans, taking rapamycin before a bout of resistance exercise blocked the normal roughly 40% rise in muscle protein synthesis that exercise should produce, and abolished the associated six-fold activation of S6K1 (Ref 11). Muscle mass and strength are among the better predictors of a long, functional life, and older adults are already fighting sarcopenia, the age-related loss of muscle. Deliberately and chronically blunting the exact signal that builds and defends muscle is a real cost, and it sits awkwardly with taking a drug in the name of healthspan.

Fertility. This is the one almost nobody selling the drug mentions. In an observational study of men who had received kidney transplants, those maintained on sirolimus (the pharmaceutical name for rapamycin) had markedly lower total sperm counts and motility than those on other regimens, and a much lower rate of fathering a pregnancy, with only partial recovery in some men after stopping the drug (Ref 12). That population is heavily confounded, they are transplant recipients on other immunosuppressants and with serious underlying illness, so this is not proof that low-dose rapamycin harms fertility in healthy men. But it aligns with animal findings on the testis, and it is exactly the kind of signal that ought to be disclosed rather than quietly omitted.

On top of these, rapamycin and its relatives, at the doses used in transplantation and oncology, are recognised to cause mouth ulcers, impaired wound healing, disturbed blood lipids, immune suppression with infection risk, and reductions in blood cell counts. Longevity enthusiasts use much lower, intermittent doses that plausibly cause fewer of these, and the PEARL safety data are consistent with that. But "plausibly safer at lower doses over one year" is not the same as "known to be safe taken for decades by healthy people," and honesty requires keeping those two statements apart.

What this means in practice

The question I ask of any intervention is simple: does this change what I would actually do? For a healthy person asking whether to take rapamycin to live longer, my answer today is no, and I do not take it myself.

That is not because the science is worthless. It is because the honest status of the evidence is this: a genuinely important mechanism, strong and reproducible lifespan data in rodents, no human longevity proof, encouraging but short-term human surrogate data confined largely to immune function, and a real set of trade-offs, some of which (muscle, glucose, fertility) cut directly against the outcomes most people are actually chasing.

Set against that, the higher-confidence levers are unglamorous, and mostly free. Eat enough good-quality protein to build and keep muscle. Train, especially against resistance. Protect your metabolic health, your sleep and your time in daylight. Do not smoke. Stay socially connected. These act on the very same nutrient-sensing biology, but physiologically and reversibly, letting mTOR rise when you eat and train and fall when you fast and rest. The goal is not to pin mTOR low with a daily drug, nor to hammer it high with leucine every waking hour. It is to let it cycle, which is what it evolved to do.

None of this applies to people taking rapamycin or a rapalog for a genuine medical indication. In organ transplantation and in certain cancers these are valuable, properly prescribed and monitored drugs. That is a completely different clinical situation from a healthy adult self-experimenting for longevity, and it is the difference between a supervised treatment and a bet.

A word of future-proofing. This field is moving quickly, toward intermittent dosing schedules, more mTORC1-selective molecules, and better human endpoints than the surrogates we have now. It is entirely possible that a fair review written a few years from now reads more favourably, or less. I will update my view when the human data do.

To be completely explicit: none of the above is medical advice, and in particular it is not advice to take rapamycin. Any decision about a prescription drug belongs with your own clinician, who can see your whole picture in a way that no article can.

Disclosures

I run a private clinic and see patients through paid consultations, and longevity and metabolic medicine are areas of professional and commercial interest for me. That is a bias, and I have tried to state the evidence that runs against my instincts as fairly as the evidence that supports them, including presenting the rodent lifespan findings as the real results they are. I have no commercial interest in rapamycin, in rapalogs, or in their manufacturers. If it would help to think through your own longevity approach, you are welcome to book a consultation, though there are many excellent clinicians who can do the same and you certainly do not need to see me.

References

1. Saxton RA, Sabatini DM. mTOR signaling in growth, metabolism, and disease. Cell, 2017. https://doi.org/10.1016/j.cell.2017.02.004

2. Wolfson RL, et al. Sestrin2 is a leucine sensor for the mTORC1 pathway. Science, 2016. https://doi.org/10.1126/science.aab2674

3. Harrison DE, et al. Rapamycin fed late in life extends lifespan in genetically heterogeneous mice. Nature, 2009. https://doi.org/10.1038/nature08221

4. Miller RA, et al. Rapamycin-mediated lifespan increase in mice is dose and sex dependent and metabolically distinct from dietary restriction. Aging Cell, 2014. https://doi.org/10.1111/acel.12194

5. Selman C, et al. Ribosomal protein S6 kinase 1 signaling regulates mammalian life span. Science, 2009. https://doi.org/10.1126/science.1177221

6. Mannick JB, et al. mTOR inhibition improves immune function in the elderly. Science Translational Medicine, 2014. https://doi.org/10.1126/scitranslmed.3009892

7. Mannick JB, et al. TORC1 inhibition enhances immune function and reduces infections in the elderly. Science Translational Medicine, 2018. https://doi.org/10.1126/scitranslmed.aaq1564

8. Urfer SR, et al. A randomized controlled trial to establish effects of short-term rapamycin treatment in 24 middle-aged companion dogs. GeroScience, 2017. https://doi.org/10.1007/s11357-017-9972-z

9. Moel M, et al. Influence of rapamycin on safety and healthspan metrics after one year: PEARL trial results. Aging, 2025. https://doi.org/10.18632/aging.206235

10. Lamming DW, et al. Rapamycin-induced insulin resistance is mediated by mTORC2 loss and uncoupled from longevity. Science, 2012. https://doi.org/10.1126/science.1215135

11. Drummond MJ, et al. Rapamycin administration in humans blocks the contraction-induced increase in skeletal muscle protein synthesis. The Journal of Physiology, 2009. https://doi.org/10.1113/jphysiol.2008.163816

12. Zuber J, et al. Sirolimus may reduce fertility in male renal transplant recipients. American Journal of Transplantation, 2008. https://doi.org/10.1111/j.1600-6143.2008.02267.x

The Coronary Calcium Score: What a “Zero” Really Means, and What It Can Miss

Executive summary

• The coronary artery calcium (CAC) score measures calcified atherosclerotic plaque in the heart's own arteries. A computed tomography (CT) scanner grades each deposit by its area and its density and sums the result into one number, the Agatston score (Ref 1). The radiation is modest: historically an effective dose of 0.8 to 10.5 millisieverts (mSv) depending on protocol, with a median around 2.3 mSv, and modern scanners sit at the low end (Ref 2).

• A score of zero is genuinely reassuring. In a systematic review of roughly 85,000 people, only 0.56% of asymptomatic individuals with a zero had a cardiovascular event over about four years (Ref 5), and a zero has been linked to a roughly fifteen-year "warranty period" of low mortality (Ref 6).

• CAC predicts events across ethnic groups and adds information on top of the standard risk factors, most usefully in people at intermediate risk where the decision to treat is genuinely on the fence (Ref 3, Ref 4). Among the widely studied newer markers, calcium scoring improved prediction the most (Ref 7, Ref 8).

• A zero is not a guarantee. The scan sees only calcified plaque; softer, non-calcified plaque is invisible to it and can still rupture. In a 2025 analysis, 23% of people presenting with a heart attack or unstable angina had a CAC of zero, and their arteries carried non-calcified plaque (Ref 12).

• Age matters. In adults aged 32 to 46, any calcium at all carried roughly a five-fold higher risk of coronary heart disease (Ref 13), and a zero earned at 40 says less about the following decades than a zero at 65.

• The score earns its keep when it changes a decision. Among people a guideline would make statin-eligible, about 44% had a calcium score of zero with a low event rate, which reclassified roughly half of them out of drug eligibility, while a high score marks those at materially higher near-term risk (Ref 10). Guidelines accordingly treat CAC as a decision aid in selected, borderline-to-intermediate-risk adults (Ref 11).

• Statins muddy repeat scans: they tend to increase the calcium density of plaque even while shrinking its total volume (Ref 14), so a number that rises on treatment does not simply mean "worse."

• The practical filter is a single question: would the result change what you and your clinician actually do? A CAC scan is most useful for the genuinely undecided and least useful for the very-low-risk young person, or for anyone who would be treated, or not, regardless of the number.

Introduction

A recurring claim in the online metabolic-health world is that people have driven their coronary artery calcium (CAC) score down to zero, and that a zero proves the arteries are clean. The "power of zero" is real, but it is narrower than the slogan suggests. This article sets out what a calcium score does and does not tell you, the cohort evidence behind it, and the one question I think should decide whether anyone has the scan at all.

I should declare my position. I run a private clinic (drabs.clinic) where consultations are paid, and my professional interest is in metabolic health and longevity. I am sympathetic to low-carbohydrate and metabolic approaches, and I am sceptical of treating a number rather than a person. Those leanings could bias how I read this literature, so I have tried to cite the mainstream evidence in full and then criticise it fairly. Every study below was checked against PubMed for its actual finding rather than the version I might have preferred.

What a calcium score actually measures

Coronary arteries are the small vessels that supply the heart muscle itself. Atherosclerosis, the disease process behind most heart attacks, deposits a mixture of fats, inflammatory cells and, over time, calcium into their walls. The CAC scan is a quick, contrast-free CT (computed tomography) of the chest that detects that calcium. The scoring method, described by Agatston and colleagues in 1990, multiplies the area of each calcified deposit by a factor for its peak density and adds the pieces together (Ref 1). A score of zero means no detectable calcified plaque; higher numbers mean more.

Two features of that definition matter for everything that follows. First, the scan measures calcium, which is a marker of established, healed plaque, not of the soft, early plaque that has not yet calcified. Second, the exposure is real but modest. A 2009 analysis put the effective dose between 0.8 and 10.5 millisieverts (mSv) depending on the scanner protocol, with a median around 2.3 mSv (Ref 2); contemporary prospectively gated protocols sit near the bottom of that range, comparable to a few months of natural background radiation. Neither the cost nor the radiation is large, but neither is quite zero, and that matters when we come to who should be scanned.

The power of zero: the evidence a zero is reassuring

The strongest thing that can be said for calcium scoring is what a zero predicts. A 2009 systematic review pooling 49 studies and more than 85,000 people found that, among asymptomatic individuals with a CAC of zero, just 0.56% had a cardiovascular event over a mean of about four years (Ref 5). A later prospective follow-up of 9,715 people reported that a baseline zero was associated with low mortality, under roughly 1% per year, for around fifteen years, which is where the phrase "warranty period" comes from (Ref 6). And in the Multi-Ethnic Study of Atherosclerosis (MESA), a population sample of 6,722 adults, calcium predicted coronary events in white, black, Hispanic and Chinese participants alike; compared with no calcium, a score above 300 carried close to a ten-fold higher adjusted risk (a hazard ratio of 9.67) (Ref 3).

Cite, then critique. These are observational cohorts, not trials, so they show association, not proof that the scan itself improves outcomes. The people who get scanned are a selected group, often the risk-aware and comparatively healthy, which can flatter a zero. Most importantly, a zero means no calcified plaque, not no atherosclerosis. It is a statement about a roughly fifteen-year horizon of low risk in populations, not a personal guarantee, and the warranty is shorter in younger people whose disease has had less time to calcify.

What CAC adds beyond the usual risk factors

Standard risk calculators use age, sex, blood pressure, smoking, diabetes and a lipid panel. The interesting question is whether a calcium scan adds anything to those. In an early prospective study of 1,312 asymptomatic adults, calcium refined risk among people whose Framingham risk was above 10% but not among those below 10%; in other words, it helped precisely in the intermediate band where the decision to treat is least clear (Ref 4). In MESA, adding the calcium score reclassified a meaningful share of people into more accurate risk categories, with a net reclassification improvement of 0.25 (Ref 7). When several fashionable markers were compared head to head in intermediate-risk adults, including C-reactive protein, ankle-brachial index and family history, the calcium score was the strongest and improved reclassification the most (Ref 8). Researchers have since folded CAC into a risk equation that lifts the model's discrimination, or C-statistic, from 0.75 to 0.80 (Ref 9).

The caveats are familiar. All of this is observational, and much of it comes from the same cohort, so the findings travel together and could share the same blind spots. Reclassification statistics reward a test for moving people between arbitrary risk bands and do not, by themselves, show that acting on the new category makes anyone live longer. What the data support is narrower but still useful: in the right person, the score sharpens an uncertain estimate.

Where a zero can mislead

Because the scan sees only calcium, it can miss the plaque that most often kills. Soft, non-calcified plaque can rupture and cause a heart attack while the calcium score reads zero. A 2025 case-control analysis nested in the CONFIRM registry looked at people who had presented with an acute coronary syndrome, meaning a heart attack or unstable angina, and found that 23% of them had a CAC of zero; when their arteries were imaged with CT angiography, the culprit lesions were made of non-calcified fibrous, fibrofatty and necrotic-core plaque that the calcium score never registered (Ref 12). That is the clearest illustration of the limit: a zero lowers the odds, it does not abolish them, and in someone with symptoms it can be falsely reassuring.

Age sharpens the point in the other direction. In adults aged 32 to 46 in the CARDIA study, finding any calcium at all was a strong signal, carrying roughly a five-fold higher risk of coronary heart disease (Ref 13). So a young person with a positive score has a genuinely concerning finding, while a young person with a zero has bought less time than an older person with a zero, because calcification simply takes years to appear.

The statin wrinkle, and the reversal stories

Two things are worth separating here, because the popular narrative tends to fuse them. The first is what statins do to the calcium score. A pooled analysis of eight trials that tracked plaque with intravascular ultrasound found that intensive statin therapy shrank the total volume of plaque yet increased its calcium content in every group studied (Ref 14). The usual interpretation is that the drugs push plaque toward a denser, more stable, more calcified form, though that reading is inferred rather than proven. The practical consequence is concrete: a calcium score can rise on a statin without the underlying disease getting worse, so serial scans in someone on treatment are hard to interpret, and chasing the number is a poor idea.

The second is the online genre of dramatic reversals, including reports of a near-complete blockage said to have cleared on a low-carbohydrate, animal-based diet. It is worth being honest about what such a story can and cannot establish. If an initial CT angiogram showed a stenosis made largely of soft, non-calcified plaque, then a later scan showing improvement is a statement about that soft plaque, not about the calcium score, and it is biologically plausible that non-calcified plaque can regress. Calcified plaque, by contrast, rarely disappears. Either way, a single self-reported case, with no verified baseline, no control group and a diet recalled from memory, is a hypothesis, not evidence. The same limitation applies to the reassuring cases: people who report a zero after years of a particular way of eating may well be telling the truth, but selection and recall make it impossible to attribute the outcome to the diet with confidence.

There is, however, a legitimate observation buried in these anecdotes. Lipid levels and calcium scores do not always agree. In the statin-eligibility analysis discussed in the next section, a substantial minority of people with a high enough LDL (low-density lipoprotein) or ApoB (apolipoprotein B) to qualify for medication, and these are the cholesterol-carrying lipoproteins rather than cholesterol itself, nonetheless had a calcium score of zero and a low near-term event rate (Ref 10). That discordance is real and clinically useful. It does not, on its own, prove that a high LDL or ApoB is harmless, and I would be overstating the evidence if I claimed it did.

What this means in practice

Put the strengths and the limits together and the calcium score stops being a slogan and becomes a tool with a job. The job is to break a genuine tie. The 2019 American College of Cardiology and American Heart Association prevention guideline reflects this, positioning CAC as a reasonable option to guide preventive treatment in selected adults, broadly those aged about 40 to 75 at borderline-to-intermediate risk in whom the decision to start a statin is uncertain (Ref 11). The supporting data are direct: in a MESA analysis of people a guideline would have made statin-eligible, about 44% had a calcium score of zero with an event rate of roughly 4.2 per 1,000 person-years, and absence of calcium reclassified around half of them as no longer clearly needing the drug, while a high score did the opposite (Ref 10).

That is exactly the situation in which I think a scan is worth doing: a person on the fence, weighing years or decades of medication, for whom a zero or a high number would genuinely change the plan. It is just as important to name who does not benefit. A very-low-risk young person gains little, because a zero was the expected result and changes nothing, while a small radiation dose, the chance of an incidental finding and the cost are real if minor downsides. A person already at high risk, who would be treated whatever the scan showed, does not need the number to make the decision. Anyone who has decided in advance to ignore an inconvenient result should save their money. And someone with actual symptoms is a different problem altogether: they need proper assessment, not a screening calcium score, precisely because a zero can miss the soft plaque behind an acute event (Ref 12).

A few practical cautions follow from the evidence rather than from opinion. Do not repeat the scan often in the hope of watching a number fall, especially on a statin, where the calcium content is expected to rise regardless (Ref 14). Do not read a zero as permission to stop attending to blood pressure, ApoB, glucose control, sleep, muscle and the other ordinary determinants of a long life. And treat the decision to start or stop any medication as a conversation with your own prescriber first; a calcium score is one input to that conversation, not a verdict, and clinicians who recommend statins are usually working faithfully within risk-score-driven guidelines rather than acting carelessly.

If it would help to think through whether a calcium score, or anything else, would actually change your plan, I offer consultations, though there are many excellent clinicians who can do the same.

Disclosures

I run a private clinic (drabs.clinic) where consultations are paid, and my professional focus is metabolic health and longevity. I am sympathetic to low-carbohydrate and metabolic approaches and sceptical of treating numbers rather than people, which may colour my reading of this evidence; readers should weigh it accordingly. I have no financial relationship with any manufacturer of CT scanners, statins or the supplements sometimes marketed alongside this topic. According to PubMed, from which every reference below was retrieved and verified, the studies are as cited, and each is linked by its DOI.

References

1. Agatston AS, et al. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol, 1990;15(4):827-832. https://doi.org/10.1016/0735-1097(90)90282-t

2. Kim KP, Einstein AJ, Berrington de González A. Coronary artery calcification screening: estimated radiation dose and cancer risk. Arch Intern Med, 2009;169(13):1188-1194. https://doi.org/10.1001/archinternmed.2009.162

3. Detrano R, et al. Coronary calcium as a predictor of coronary events in four racial or ethnic groups. N Engl J Med, 2008;358(13):1336-1345. https://doi.org/10.1056/NEJMoa072100

4. Greenland P, et al. Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA, 2004;291(2):210-215. https://doi.org/10.1001/jama.291.2.210

5. Sarwar A, et al. Diagnostic and prognostic value of absence of coronary artery calcification. JACC Cardiovasc Imaging, 2009;2(6):675-688. https://doi.org/10.1016/j.jcmg.2008.12.031

6. Valenti V, et al. A 15-year warranty period for asymptomatic individuals without coronary artery calcium: a prospective follow-up of 9,715 individuals. JACC Cardiovasc Imaging, 2015;8(8):900-909. https://doi.org/10.1016/j.jcmg.2015.01.025

7. Polonsky TS, et al. Coronary artery calcium score and risk classification for coronary heart disease prediction. JAMA, 2010;303(16):1610-1616. https://doi.org/10.1001/jama.2010.461

8. Yeboah J, et al. Comparison of novel risk markers for improvement in cardiovascular risk assessment in intermediate-risk individuals. JAMA, 2012;308(8):788-795. https://doi.org/10.1001/jama.2012.9624

9. McClelland RL, et al. 10-year coronary heart disease risk prediction using coronary artery calcium and traditional risk factors: derivation in the MESA with validation in the HNR study and the DHS. J Am Coll Cardiol, 2015;66(15):1643-1653. https://doi.org/10.1016/j.jacc.2015.08.035

10. Nasir K, et al. Implications of coronary artery calcium testing among statin candidates according to ACC/AHA cholesterol management guidelines: MESA. J Am Coll Cardiol, 2015;66(15):1657-1668. https://doi.org/10.1016/j.jacc.2015.07.066

11. Arnett DK, et al. 2019 ACC/AHA guideline on the primary prevention of cardiovascular disease. Circulation, 2019;140(11):e563-e595. https://doi.org/10.1161/CIR.0000000000000677

12. Jonas RA, et al. CTA-derived plaque characteristics and risk of acute coronary syndrome in patients with coronary artery calcium score of zero: insights from the ICONIC trial. AJR Am J Roentgenol, 2025;225(1):e2431476. https://doi.org/10.2214/AJR.24.31476

13. Carr JJ, et al. Association of coronary artery calcium in adults aged 32 to 46 years with incident coronary heart disease and death. JAMA Cardiol, 2017;2(4):391-399. https://doi.org/10.1001/jamacardio.2016.5493

14. Puri R, et al. Impact of statins on serial coronary calcification during atheroma progression and regression. J Am Coll Cardiol, 2015;65(13):1273-1282. https://doi.org/10.1016/j.jacc.2015.01.036

Cholesterol, LDL and Longevity: What the Evidence Does and Doesn’t Show

Executive summary

• In the single largest cohort study on this question, 12.8 million Korean adults, total cholesterol had a U-shaped relationship with all-cause mortality. The lowest death rates sat at a total cholesterol of roughly 210 to 249 mg/dL (about 5.4 to 6.4 mmol/L), which is higher than the level many patients are told to aim below, and the excess risk at the low end was larger than the excess risk at the high end (Ref 1). Association is not causation, but a simple "lower is always better" message does not survive this data.

• In adults over 60, most cohort data show that LDL, the lipoprotein that carries cholesterol, is not positively associated, and is frequently inversely associated, with all-cause mortality (Ref 2). This is association data, it cannot prove direction, and reverse causation is a genuine confounder that I take seriously.

• Among 136,905 US patients hospitalised with coronary artery disease, mean admission LDL-C was about 105 mg/dL (about 2.7 mmol/L) and nearly half were below 100 mg/dL (Ref 3). That complicates a simple dose-response reading of LDL, with the important caveat that admission lipids in acutely unwell patients have limitations.

• The strongest case for the lipid hypothesis is real and deserves engagement, not dismissal: lifelong low LDL from PCSK9 gene variants tracks with markedly lower coronary disease (Ref 5), and lowering LDL-C with ezetimibe or a PCSK9 inhibitor does reduce cardiovascular events (Refs 6, 7).

• The absolute benefit, however, is modest. Adding ezetimibe to a statin cut the seven-year composite event rate by about 2 percentage points (Ref 6). Across statin trials, the median gain in survival within the trial period was about 3 to 4 days (Ref 8), and these particular trials did not demonstrate a clear reduction in all-cause mortality.

• Where plaque forms is governed substantially by blood-flow mechanics. Atherosclerosis is focal, clustering at arterial branch points where shear stress is low and disturbed, which places endothelial biology at the centre of the process (Ref 4).

• What appears to track better with risk in day-to-day practice is the metabolic environment: the triglyceride-to-HDL-C ratio as a proxy for insulin resistance (Ref 14), fasting insulin, HOMA-IR, and HbA1c read alongside red-cell turnover (Ref 15). A coronary artery calcium score of zero is associated with very low mortality (Ref 16).

• None of this is a reason to stop a prescribed statin on your own initiative. It is a reason to have a more detailed, better-informed conversation with your prescriber, and there are many excellent clinicians who will have that conversation with you.

A note on my bias, before anything else

I run a private clinic and I am paid for one-to-one consultations, so I have a commercial interest and you should weigh what follows accordingly. I also read human physiology through a metabolic, low-carbohydrate lens, and I hold postgraduate qualifications in metabolic medicine and anti-ageing medicine, which are separate disciplines rather than one. That lens shapes how I interpret the evidence below. I have tried to state the mainstream case as strongly as its proponents would, then set out where I think it is well supported and where I think it is weak. I may be wrong on parts of this, and I would genuinely welcome evidence that points the other way.

Two points of precision run through the whole piece. First, LDL is a lipoprotein, a particle that ferries cholesterol and fats through the blood. What a standard blood test reports is usually LDL-C, the mass of cholesterol carried inside those particles, and ApoB, which counts the number of atherogenic particles. These are related but not the same thing, and the distinction matters. Second, several of the studies cited here were funded by manufacturers of the drugs being tested, and I have flagged that where it is relevant, in both directions.

What the population-level data actually show

The claim that low cholesterol is uniformly protective does not hold up against the largest datasets we have.

In a prospective cohort of 12,815,006 Korean adults followed for around a decade, with 694,423 deaths recorded, total cholesterol showed a U-shaped association with all-cause mortality in every age and sex band (Ref 1). The total cholesterol range associated with the lowest mortality was 210 to 249 mg/dL for most groups, and the inverse association below 200 mg/dL was actually stronger than the positive association above it. In other words, across this enormous population, moderately high cholesterol was not where mortality was highest, and low cholesterol carried its own excess risk.

Focusing specifically on LDL rather than total cholesterol, a 2016 systematic review of 19 studies covering 68,094 people aged 60 and over found that LDL-C was inversely associated with all-cause mortality in 16 of the cohorts where this was assessed, representing 92 per cent of the participants (Ref 2). Higher LDL in older people was associated with living at least as long, not shorter.

Now the honest critique, because this evidence has real limitations. All of it is observational, so it establishes association, not causation. Reverse causation is the central problem: serious illness, cancer, chronic infection and frailty can lower cholesterol on the way to death, which can manufacture an apparent "low cholesterol, high mortality" signal without cholesterol causing anything. The Korean data are general-population rather than hospitalised patients, which weakens but does not eliminate that concern. The elderly LDL review (Ref 2) is also authored by a group of well-known lipid-hypothesis sceptics and has been criticised for its inclusion criteria and for not fully accounting for statin use and confounding. I think those criticisms deserve to be stated plainly. What I do not think survives is the simple message that lower is always better. The size, consistency and direction of these datasets are not compatible with that claim, even after the caveats.

The heart-attack admission data

A frequently cited observation is that people having cardiac events often do not have high LDL. In an analysis of 136,905 hospitalisations for coronary artery disease across 541 US hospitals, the mean admission LDL-C was 104.9 mg/dL, nearly half of patients had an LDL-C below 100 mg/dL, and only 17.6 per cent were below 70 mg/dL (Ref 3). Notably, more than half had an HDL-C below 40 mg/dL, a metabolic marker rather than an LDL one.

The fair reading, and the fair critique, matter here. This was a cross-sectional snapshot of admission lipids in people who already had coronary disease, not a comparison against a matched population without disease, so it cannot by itself prove that LDL is irrelevant. LDL-C also falls during the acute phase of a myocardial infarction, which can bias admission values downward. And these were coronary-artery-disease admissions broadly, not exclusively confirmed heart attacks. What the data do undercut is the intuition that people who have cardiac events are overwhelmingly the people with the highest LDL. They are not. That should at least prompt humility about using LDL-C in isolation as the number that decides who is at risk.

The strongest case for the lipid hypothesis, taken seriously

I am not interested in a straw man, so here is the mainstream case at its strongest.

First, genetics. People who by chance inherit PCSK9 gene variants that lower their LDL for life have substantially less coronary disease. In the Atherosclerosis Risk in Communities study, nonsense mutations in PCSK9 were associated with a 28 per cent lower LDL-C and an 88 per cent lower risk of coronary heart disease over 15 years in one group, and a separate variant with a 15 per cent lower LDL-C and a 47 per cent lower risk in another (Ref 5). This is a genuinely important finding, and it is the backbone of the Mendelian-randomisation argument that lifelong LDL exposure is causal.

Second, drug trials that lower LDL by mechanisms other than statins. In IMPROVE-IT, adding ezetimibe to a statin after acute coronary syndrome lowered LDL-C further, from a median of 69.5 to 53.7 mg/dL, and reduced the seven-year composite event rate from 34.7 to 32.7 per cent (Ref 6). In FOURIER, the PCSK9 inhibitor evolocumab lowered LDL-C by 59 per cent, to a median of 30 mg/dL, and reduced the primary composite endpoint from 11.3 to 9.8 per cent over a median 2.2 years (Ref 7). Because ezetimibe and PCSK9 inhibitors are not statins and act largely through LDL, these trials are strong evidence that lowering LDL-C itself, not merely some side effect of statins, contributes to fewer events. This is the single best rebuttal to anyone, myself included, who wants to argue that LDL is purely a bystander, and intellectual honesty requires me to put it in front of you clearly.

Here is where I land after taking all of that seriously. The genetic argument is about lifelong exposure across populations and does not translate cleanly into what a specific pharmacological intervention will do for the individual in front of me later in life, and PCSK9 has biological roles beyond LDL. The drug-trial benefits are real but small in absolute terms and are built on composite endpoints that include revascularisation, not only death. IMPROVE-IT's absolute risk reduction was about 2 percentage points over seven years (Ref 6), and FOURIER, despite a striking relative risk reduction, was not shown to reduce cardiovascular or all-cause mortality over its short follow-up (Ref 7). Put most soberly, a systematic review of statin trials found the median postponement of death within the trials was 3.2 days for primary prevention and 4.1 days for secondary prevention (Ref 8). That does not mean the drugs do nothing. It means the absolute benefit is a great deal smaller than the way these medicines are usually described at the bedside, and reasonable people can weigh a few days of average postponement against cost and side effects differently.

Where atherosclerosis forms: endothelium and flow

If LDL concentration were the direct cause of plaque, plaque should appear fairly uniformly wherever the blood carries lipoproteins. It does not. Atherosclerosis is strikingly focal, and it clusters at the outer edges of arterial branch points.

The mechanism that best explains this is haemodynamic shear stress, the frictional force of flowing blood on the cells lining the artery. In a landmark review, higher, smooth arterial shear stress promoted a quiescent, protective endothelial phenotype, whereas the low and disturbed shear stress found at branch points promoted an atherogenic phenotype (Ref 4). The disease tracks with the mechanical-stress pattern, which is not uniform, rather than with the lipid concentration, which is. This is why my own emphasis is on the health of the endothelium and the inflammatory environment around it, rather than on the lipid number alone. Much of this endothelial and shear-stress biology comes from experimental and animal models, so it should be read as mechanism rather than as human outcome data, and humans may differ.

I want to be careful not to overclaim. Shear stress explains where plaque forms; it does not by itself prove that LDL and ApoB are irrelevant to how it progresses. The mainstream "response-to-retention" model holds that ApoB-containing particles retained in the artery wall are a necessary part of lesion growth, and that model is well supported. My honest position is that endothelial damage and inflammation set the stage, and that lipoproteins are involved once that stage is set. Where I differ from the strong lipid view is on primacy, on which comes first and which is the more useful lever, not on whether lipoproteins appear in plaque at all.

LDL versus ApoB: a better number for the same question

ApoB is increasingly promoted as the upgrade to LDL-C, and on its own terms that is fair. Each atherogenic particle carries one ApoB molecule, so ApoB counts particle number, whereas LDL-C measures only the cholesterol mass those particles carry. Two people with the same LDL-C can have very different particle counts, and ApoB captures that. Within the lipid framework, it is the more informative measurement.

My reservation is not that ApoB is a bad measurement. It is that it answers the same category of question as LDL-C: it quantifies circulating particles, which are associated with disease, rather than measuring the disease process itself. I do not routinely order it as a target to drive down, though it often appears on standard panels anyway, and it is worth understanding so that you can follow the conversation when a clinician raises it. That is a matter of interpretation and emphasis, and clinicians I respect weigh it differently.

Statins: absolute benefit and the mitochondrial question

Two things can be true at once. Statins produce a real reduction in events in secondary prevention, and that benefit is smaller in absolute terms than most patients believe. The survival-postponement figures above (Ref 8) are the clearest way to see the second point.

On mechanism and harm, statins inhibit HMG-CoA reductase, the same early enzyme in the mevalonate pathway that also produces coenzyme Q10 (ubiquinone), a molecule central to how mitochondria generate ATP. A review of the statin adverse-effect literature has argued that a mitochondrial mechanism, including reduced CoQ10, may underlie the muscle symptoms that are more frequent with statins than with placebo in randomised trials, and possibly some non-muscle effects too (Ref 9). I want to hedge this properly: that is a review-level hypothesis, not a settled account of population harm, and the mainstream position is that statins are generally well tolerated and that serious adverse effects are uncommon. My own clinical preference leans cautious, but a preference is not proof.

What I will not do is tell you to stop a statin your prescriber has recommended. No disease is caused by a deficiency of statins, and I think the framing around them is often too absolute, but the right move if you have doubts is to discuss the specifics with the clinician who knows your history, not to act on an article. If you have already had a heart attack, this is exactly the situation where the evidence for treatment is strongest and where self-directed changes carry the most risk.

Special cases where the argument gets tested

Familial hypercholesterolaemia. This is where the causal-LDL case is strongest, and it deserves respect. Heterozygous FH affects somewhere between roughly 1 in 200 and 1 in 500 people depending on the population, usually through a faulty LDL receptor, and it is associated with up to a 13-fold higher risk of coronary heart disease (Ref 10). One way to think about the underlying biology is that the receptors, the cellular "letterboxes" that take LDL out of the blood and into cells, are missing or faulty, so cholesterol accumulates in the bloodstream while cells are less able to take delivery. That framing is my interpretation, and it is not the mainstream one, which holds straightforwardly that lifelong high LDL and ApoB drive the excess risk and that lowering them reduces events in FH. I flag my disagreement honestly and add the most important sentence in this section: a confirmed FH diagnosis is precisely the case that must be managed with a prescribing clinician, not on the basis of a blog.

Menopause. Many women see LDL-C rise across the menopausal transition. This is not merely ageing. In the SWAN cohort, total cholesterol, LDL-C and ApoB rose specifically within the year around the final menstrual period, consistent with a menopause-induced change rather than gradual chronological drift, and the pattern held across ethnic groups (Ref 11). My reading is that this is a physiological recalibration as oestrogen declines and lipid handling adjusts, and that treating the number reflexively as a disease misreads it. Hormone replacement is a separate and individual conversation, and I do not see it as the default response to a post-menopausal lipid shift, though it is one tool among several and I am qualified to discuss it.

Lean, low-carbohydrate athletes. Lean people who train hard on low-carbohydrate diets often show dramatically elevated LDL-C. The pattern has been described as the "lean mass hyper-responder" phenotype, defined in one analysis as LDL-C at or above 200 mg/dL with HDL-C at or above 80 and triglycerides at or below 70, occurring in lean, metabolically healthy individuals (Ref 12). The proposed mechanism is prosaic: when little fuel comes from carbohydrate, more fat is trafficked around the body by lipoproteins, so particle numbers rise because the system is doing its job. Strikingly, a 2024 imaging study compared 80 such individuals, mean LDL-C 272 mg/dL on a ketogenic diet for a mean of 4.7 years, against matched controls with an average LDL-C 149 mg/dL lower, and found no significant difference in coronary artery calcium or CT-angiography plaque, and no correlation between LDL-C and plaque in either group (Ref 13). That is a genuinely interesting result. It is also small, short-term and cross-sectional, and the companion progression work suggests plaque can still advance in some of these individuals, so it should be read as preliminary rather than reassuring in a settled way. I would not reflexively treat a lean athlete's isolated high LDL-C without looking at the rest of their metabolic picture, but nor would I claim this question is closed.

What appears to predict risk better in practice

If the disease is fundamentally about endothelial health and inflammation, the more useful question is whether the metabolic environment is driving that inflammation. A coherent panel, read together rather than as isolated numbers, is more informative to me than an LDL-C value alone.

The triglyceride-to-HDL-C ratio is calculable from any standard lipid panel and functions as a rough proxy for insulin sensitivity. In the work of McLaughlin and colleagues, a ratio at or above 3.5 in US (mg/dL) units identified insulin resistance and a preponderance of small, dense LDL particles (Ref 14). One important practical trap: this ratio is unit-dependent, because triglycerides and HDL-C convert between mg/dL and mmol/L by different factors, so a ratio expressed in mmol/L (UK units) is numerically much lower than the same physiology in mg/dL. My own preference is to see the ratio comfortably below 1 in mmol/L units, which is stricter than commonly quoted thresholds; I treat it as a directional signal rather than a hard cut-off.

Fasting insulin is rarely ordered but is arguably the earliest number to move, often rising for years to hold glucose steady before fasting glucose drifts. HOMA-IR, derived from fasting glucose and fasting insulin together, quantifies that relationship and is the most practical measure of what is conventionally called insulin resistance. I prefer the phrase physiological insulin resistance, because in many cases the cells are not broken; they are down-regulating glucose entry as a regulated, protective response, and the "resistance" label implies a malfunction that may not be there. That is my framing, offered as a way of thinking, not as established nomenclature.

HbA1c, the roughly three-month glucose marker, is widely available and useful, but it is not reliable in isolation, and this is the methodological point I would hold myself to as firmly as I hold the lipid hypothesis. HbA1c assumes a standard red-blood-cell lifespan of about 120 days. If red-cell turnover is faster, the reading is falsely low; if slower, falsely high. Measured red-cell survival varies enough between healthy people to shift HbA1c meaningfully for the same average glucose (Ref 15). Pairing HbA1c with a reticulocyte count lets you interpret it in context. Most panels do not do this, and I include myself in the criticism when I have not.

Finally, when a direct look at the artery is genuinely warranted, the coronary artery calcium scan measures calcified plaque itself rather than a marker that travels with it. In a cohort of 25,563 asymptomatic people, a calcium score of zero was associated with very low all-cause and cardiovascular mortality across long-term follow-up (Ref 16). It is not a routine test, it involves a radiation dose, it misses non-calcified "soft" plaque, and it should only follow a conversation about whether it is justified for you specifically. The honest filter is whether the result would change the plan. For most of my patients, the dietary and lifestyle advice is the same whether the score is zero or high, and a test that does not change the decision is a test worth questioning.

What this means in practice

I try to apply one filter to every test and every intervention: does the result actually change what I would do? For most people worried about an LDL number, the sober answer is that the LDL-C in isolation changes very little, and the metabolic panel above changes rather more.

If you have been recommended a statin and want to engage rather than simply comply or refuse, these are measured questions to ask your prescriber, in a friendly and professional spirit: what is the absolute risk reduction, not the relative risk reduction, for someone with my profile; what was the baseline event rate in the trials being relied on; and were those trials conducted in people matched on the metabolic markers I have just described. If the answers are confident and grounded in outcome data, listen carefully. If they are not, that is useful information too.

Your general practitioner is not the problem here. They work inside guidelines from bodies such as NICE, the American College of Cardiology and the American Heart Association, under real medicolegal constraint and limited appointment time, and inside an information environment heavily shaped by industry funding. That structure, not the individual clinician, is what I am critiquing. If you want to see what primary care can look like when it works upstream of the lipid framing, the general-practice service evaluation by Unwin and colleagues reported around 51 per cent of a low-carbohydrate cohort achieving drug-free type 2 diabetes remission over eight years, alongside falls in weight, triglycerides and blood pressure (Ref 18), with the honest caveat that this was an uncontrolled, self-selected service evaluation rather than a randomised trial.

The foundations matter more than any of this, and they are free. Food built around real, minimally processed ingredients that does not chronically spike insulin. Sleep you wake rested from. Daylight on your skin and in your eyes. Regular movement and appropriate strength work. Social connection and lower chronic stress. Supplements are situational tools at most, not the centre of the plan, and I use very few myself. None of the foundations are glamorous, and all of them do more for the endothelium than chasing a single lipid number.

If you would like to work through your own numbers in detail, you do not have to do it with me; there are many excellent clinicians who take this kind of approach. I do offer one-to-one consultations if you would specifically like my input.

Disclosures

I run a private clinic and charge for one-to-one consultations, so I benefit financially when people choose to work with me, and you should weigh this article in that light. I hold a metabolic, low-carbohydrate view of human physiology, which colours my reading of the evidence. I have tried to present the mainstream lipid-hypothesis case at its strongest and to cite the studies that cut against my own position as fairly as those that support it. Several trials cited here were funded by pharmaceutical manufacturers, which I have noted, and one supportive reference on elderly LDL is authored by a group with a declared sceptical position, which I have also noted. I may be wrong on parts of this, and I will update publicly if better evidence emerges.

References

1. Yi SW, Yi JJ, Ohrr H. Total cholesterol and all-cause mortality by sex and age: a prospective cohort study among 12.8 million adults. Scientific Reports, 2019. https://doi.org/10.1038/s41598-018-38461-y

2. Ravnskov U, Diamond DM, Hama R, et al. Lack of an association or an inverse association between low-density-lipoprotein cholesterol and mortality in the elderly: a systematic review. BMJ Open, 2016. https://doi.org/10.1136/bmjopen-2015-010401

3. Sachdeva A, Cannon CP, Deedwania PC, et al. Lipid levels in patients hospitalized with coronary artery disease: an analysis of 136,905 hospitalizations in Get With The Guidelines. American Heart Journal, 2009. https://doi.org/10.1016/j.ahj.2008.08.010

4. Malek AM, Alper SL, Izumo S. Hemodynamic shear stress and its role in atherosclerosis. JAMA, 1999. https://doi.org/10.1001/jama.282.21.2035

5. Cohen JC, Boerwinkle E, Mosley TH, Hobbs HH. Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. New England Journal of Medicine, 2006. https://doi.org/10.1056/NEJMoa054013

6. Cannon CP, Blazing MA, Giugliano RP, et al. Ezetimibe added to statin therapy after acute coronary syndromes (IMPROVE-IT). New England Journal of Medicine, 2015. https://doi.org/10.1056/NEJMoa1410489

7. Sabatine MS, Giugliano RP, Keech AC, et al. Evolocumab and clinical outcomes in patients with cardiovascular disease (FOURIER). New England Journal of Medicine, 2017. https://doi.org/10.1056/NEJMoa1615664

8. Kristensen ML, Christensen PM, Hallas J. The effect of statins on average survival in randomised trials, an analysis of end point postponement. BMJ Open, 2015. https://doi.org/10.1136/bmjopen-2014-007118

9. Golomb BA, Evans MA. Statin adverse effects: a review of the literature and evidence for a mitochondrial mechanism. American Journal of Cardiovascular Drugs, 2008. https://doi.org/10.2165/0129784-200808060-00004

10. Nordestgaard BG, Chapman MJ, Humphries SE, et al. Familial hypercholesterolaemia is underdiagnosed and undertreated in the general population: consensus statement of the European Atherosclerosis Society. European Heart Journal, 2013. https://doi.org/10.1093/eurheartj/eht273

11. Matthews KA, Crawford SL, Chae CU, et al. Are changes in cardiovascular disease risk factors in midlife women due to chronological aging or to the menopausal transition? Journal of the American College of Cardiology, 2009. https://doi.org/10.1016/j.jacc.2009.10.009

12. Norwitz NG, Feldman D, Soto-Mota A, Kalayjian T, Ludwig DS. Elevated LDL cholesterol with a carbohydrate-restricted diet: evidence for a "lean mass hyper-responder" phenotype. Current Developments in Nutrition, 2022. https://doi.org/10.1093/cdn/nzab144

13. Budoff M, Manubolu VS, Kinninger A, et al. Carbohydrate restriction-induced elevations in LDL-cholesterol and atherosclerosis: the KETO trial. JACC: Advances, 2024. https://doi.org/10.1016/j.jacadv.2024.101109

14. McLaughlin T, Reaven G, Abbasi F, et al. Is there a simple way to identify insulin-resistant individuals at increased risk of cardiovascular disease? American Journal of Cardiology, 2005. https://doi.org/10.1016/j.amjcard.2005.03.085

15. Cohen RM, Franco RS, Khera PK, et al. Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood, 2008. https://doi.org/10.1182/blood-2008-04-154112

16. Shaikh K, Li D, Nakanishi R, et al. Low short-term and long-term cardiovascular and all-cause mortality in absence of coronary artery calcium: a 22-year follow-up observational study from large cohort. Journal of Diabetes and its Complications, 2019. https://doi.org/10.1016/j.jdiacomp.2019.05.015

17. McLaughlin T, Abbasi F, Cheal K, et al. Use of metabolic markers to identify overweight individuals who are insulin resistant. Annals of Internal Medicine, 2003. https://doi.org/10.7326/0003-4819-139-10-200311180-00007

18. Unwin D, Delon C, Unwin J, Tobin S, Taylor R. What predicts drug-free type 2 diabetes remission? Insights from an 8-year general practice service evaluation of a lower carbohydrate diet with weight loss. BMJ Nutrition, Prevention & Health, 2023. https://doi.org/10.1136/bmjnph-2022-000544

The Metabolic Roots of High Blood Pressure: Why Cutting Salt Is Rarely the Whole Story

Executive summary

• Reducing dietary salt does lower blood pressure on average, but the effect is modest and varies widely between individuals. It is larger in people who are older, salt-sensitive, or already hypertensive, and smaller in everyone else (Ref 1, Ref 2, Ref 3).

• Insulin is a sodium-retaining hormone. In a classic human study, an insulin infusion given while blood glucose was held steady reduced the kidney's sodium excretion by about half, which can expand blood volume (Ref 12).

• Chronically raised insulin, or hyperinsulinaemia, travels closely with hypertension. Whether it causes sustained high blood pressure on its own is genuinely debated, and it is probably one contributing mechanism among several rather than a single cause (Ref 14, Ref 15).

• Because dietary carbohydrate is the main stimulus for insulin, lower-carbohydrate diets tend to prompt early sodium and water loss and small reductions in blood pressure, alongside weight loss and lower triglycerides. The average blood-pressure effect is modest, and the cholesterol carried on LDL particles can rise (Ref 16).

• Magnesium supplementation lowers blood pressure by a small amount on average, roughly 2 mmHg systolic, possibly more in people who are insulin-resistant or genuinely deficient. The trials are heterogeneous and often of modest quality (Ref 4, Ref 5, Ref 6, Ref 7).

• Mineral intake is not the same as mineral status. Plant compounds such as phytate reduce absorption of some minerals, strongly for iron and zinc and more modestly for magnesium (Ref 9, Ref 10).

• Uric acid, which fructose raises, may impair nitric-oxide-dependent relaxation of blood vessels. This is mechanistically plausible but rests largely on animal and cell studies, so human causal evidence is still limited (Ref 18, Ref 19, Ref 20, Ref 26).

• Potassium relative to sodium matters more than sodium alone. Potassium lowers blood pressure most when sodium intake is high, and partially replacing sodium with potassium reduced strokes and deaths in a large trial, with an important caution in kidney disease (Ref 21, Ref 22, Ref 23).

High blood pressure, technically hypertension, is usually explained to patients in a single line: eat less salt. For some people that helps. For many it does little, and the reason, I will argue, is that salt is often a symptom rather than the root cause.

I should declare my starting point before making that case. I run a private clinic and offer paid consultations, so I have a commercial interest in people valuing this kind of analysis. I also work through a metabolic, lower-carbohydrate lens, which shapes how I read the literature. I have tried to counter that below by citing the mainstream salt evidence at its strongest and by flagging where my interpretation runs ahead of what the data can carry. Every study I mention is referenced, and I have noted the limitations of each rather than only its headline.

The salt story: real, but modest and incomplete

Two large, careful reviews anchor the conventional view. A Cochrane analysis of 34 trials found that cutting salt by roughly 4.4 g per day lowered systolic blood pressure by about 4 mmHg overall, and by around 5.4 mmHg in people with hypertension (Ref 1). A 2020 analysis of 133 trials and more than 12,000 people found a dose-response relationship, with each 50 mmol reduction in daily sodium associated with roughly a 1 mmHg fall in systolic pressure, and larger effects in older and hypertensive people (Ref 2). So salt reduction genuinely lowers blood pressure. That is not in serious dispute, and I will not pretend otherwise.

The honest caveats are that the average effect is small, the trials are statistically heterogeneous, and the response is very uneven between individuals. That unevenness has a name: salt sensitivity. Only a subset of people show a large pressure rise with sodium, and the mechanisms behind it, genetic, hormonal, renal, and immune, are still only partly understood (Ref 3). The practical problem with "everyone eat less salt" is that it treats a population average as if it were an individual prescription, and it leaves the upstream question unasked: why is this person's system handling sodium poorly in the first place? For many, the answer runs through insulin.

Insulin, the kidney, and sodium retention

Insulin does far more than move glucose. One of its jobs is to tell the kidney to hold on to sodium. The cleanest demonstration is nearly fifty years old. When researchers infused insulin into healthy volunteers while clamping blood glucose at its normal level, urinary sodium excretion fell by about half, with no change in filtration rate or in aldosterone (Ref 12). The kidney simply reabsorbed more sodium. Where sodium goes, water follows, and a larger blood volume raises pressure in the circulation, much as adding water to a closed network of pipes raises the pressure inside it. The same direction of effect appears in animal work, where insulin treatment reliably causes sodium retention in rodents (Ref 13, an animal study, so its human relevance is indirect).

Here I have to be careful, because this point is often overstated. Showing that insulin acutely retains sodium is not the same as proving that chronically high insulin causes lasting hypertension. When researchers infused insulin into dogs for weeks, blood pressure did not rise, whereas in rats it did (Ref 15). Reviews of the human data conclude that insulin resistance and hyperinsulinaemia are consistently associated with high blood pressure and plausibly act as a slow, amplifying pressure mechanism, but that obesity-related hypertension is too complex to pin on insulin alone (Ref 14, Ref 15). My reading is that hyperinsulinaemia is one important lever, particularly in people who are also salt-sensitive, rather than the only one.

Carbohydrate load and hyperinsulinaemia

If insulin is a lever on blood pressure, the main dietary control on insulin is carbohydrate. Eating carbohydrate raises blood glucose, and insulin rises to return it to its set point. Do this repeatedly against a background of excess body fat and inactivity, and cells respond less to insulin's signal, so the body secretes still more insulin to achieve the same result. That is pathological insulin resistance, and it keeps circulating insulin chronically high.

It is worth separating this from a different, benign state. A person adapted to a very-low-carbohydrate diet can also show reduced glucose tolerance, but this is an adaptive, physiological form of insulin resistance that spares glucose for tissues such as the brain. It is not the same process as the hyperinsulinaemia of metabolic disease. The distinction matters when someone worries that their fasting glucose drifted upward on a low-carbohydrate diet.

What happens to blood pressure when carbohydrate is reduced? A meta-analysis of 13 randomised trials found that very-low-carbohydrate diets produced modestly greater weight loss and a small fall in diastolic blood pressure compared with low-fat diets, along with lower triglycerides and higher HDL, although the cholesterol carried on LDL particles rose modestly (Ref 16). The blood-pressure effect was real but small on average. Part of it is likely the early natriuresis described above: as insulin falls, the kidney stops over-retaining sodium and water, and losing that extra fluid is why people often shed several pounds of water in the first week. That is a plausible mechanism rather than a proven chain in every individual. There is also a mechanistic thread linking sugar specifically to salt handling. In mice, a fructose-rich diet sensitised blood pressure to salt through an insulin-activated sodium channel in the kidney, and removing that pathway blunted the effect (Ref 17, animal work). It is a clue, not human proof, but it fits the pattern.

Magnesium: a small but plausible contribution

Magnesium relaxes blood vessels, at least in principle. It behaves as a natural calcium antagonist and supports the production of nitric oxide (NO) and prostacyclin, the local signals that tell arteries to widen (Ref 7). It helps to picture nitric oxide as the brake-release on vascular tone, and magnesium as one of the things that keeps that release working.

The trial evidence is real but modest. The best pooled analysis, 34 randomised double-blind trials in more than 2,000 people, found magnesium supplementation lowered systolic pressure by about 2 mmHg and diastolic by about 1.8 mmHg (Ref 4). A Cochrane review was more sceptical still, concluding that the apparent benefit was small and probably inflated by lower-quality studies (Ref 5). Some popular summaries quote considerably larger reductions, but the higher-quality pooled estimates are smaller, and it is those I would trust. Where magnesium may do more is in metabolically impaired people: in trials restricted to those with insulin resistance, prediabetes, or chronic disease, the reductions were larger, around 4 mmHg systolic (Ref 6).

Is deficiency common enough to matter? Population magnesium status is genuinely hard to measure, because blood levels are held stable at the expense of tissue stores, so a normal serum value can mask depletion. Some surveys of apparently healthy adults find a substantial fraction below reference thresholds (Ref 8, a small single study, so treat it as suggestive rather than settled). My own clinical use of magnesium is as time-limited support while the diet is corrected, not as a treatment in its own right.

Antinutrients, bioavailability, and the soil question

This undermines the simple instruction to "just eat more vegetables for your minerals": how much of a mineral you swallow is not how much you absorb. Plant foods contain compounds that bind minerals in the gut, and phytate, or phytic acid, is the best studied. In a controlled human experiment, adding phytate to bread at amounts found naturally in wholemeal cut magnesium absorption from about 33% to 13% (Ref 9).

The fair counterpoint is that phytate's grip is not equal across minerals. Reviews conclude that it strongly inhibits iron and zinc absorption, while its effect on magnesium and calcium is more modest (Ref 10). So the antinutrient argument is genuine but should not be inflated. It is a reason to value animal foods, which are largely free of these binders and deliver minerals in more absorbable forms, and a reason not to assume a leafy salad equals its label.

A related claim is that produce is simply less nutritious than it used to be. The most cited analysis compared 43 garden crops between 1950 and 1999 and found median declines of 6% to 38% for six nutrients (Ref 11). The same authors were candid about the limits, though: many individual changes were not statistically reliable, some nutrients actually rose, and the likeliest explanation is not exhausted soil but the shift to higher-yielding varieties that trade nutrient density for size. It is a modest, real signal that supports choosing nutrient-dense foods, not a dramatic story of empty vegetables.

Uric acid and fructose

Uric acid is usually discussed only in the context of gout, but it may also affect blood-vessel function. Fructose, from added sugar and from large quantities of fruit, is a notable driver of uric acid production. In cell and animal studies, raised uric acid reduces nitric oxide availability and impairs the ability of arteries to relax, and it appears to interfere specifically with insulin's own vessel-widening action (Ref 20, Ref 26, both largely animal and cell work, so human relevance is inferred). Reviews of the human epidemiology find that elevated uric acid predicts the later onset of hypertension, especially in the young, and set out a plausible mechanism through nitric oxide and the renin-angiotensin system (Ref 18, Ref 19).

I want to stay measured. Most of the direct mechanistic evidence is not from humans, and trials of uric-acid-lowering drugs to treat blood pressure have given mixed results. So the honest statement is that it is plausible that high uric acid, driven by fructose against a high-carbohydrate background, contributes to blood-pressure problems in some people, and that it can be worth measuring in the right context. It is not that lowering uric acid is a proven blood-pressure treatment.

Potassium and the sodium-potassium balance

Sodium does not act alone. It works against potassium, and the balance between the two is more informative than either in isolation. Potassium supplementation lowers blood pressure by around 3 mmHg systolic on average, and, tellingly, the effect is largest when sodium intake is high (Ref 21). A dose-response analysis confirmed the relationship but also showed it is not linear: the benefit plateaus, and very high potassium intakes are not clearly better and may carry risk in some groups (Ref 22).

The strongest evidence comes from hard outcomes rather than pressure readings alone. In a trial of nearly 21,000 people at high cardiovascular risk, replacing a quarter of ordinary salt with potassium chloride reduced strokes, major cardiovascular events, and deaths (Ref 23). The essential safety caveat is that potassium is dangerous in people with impaired kidney function or on certain blood-pressure medications, in whom the same substitution can cause harmful potassium retention. This is precisely why a blanket rule is a poor substitute for individual assessment. And, again, potassium-rich plants carry the antinutrients discussed above, so the gap between intake and absorption applies here too.

When the problem is structural, not metabolic

Everything so far concerns what I would call metabolic hypertension: pressure driven by volume, hormones, and vascular signalling, which can shift relatively quickly. There is a second category. With age and prolonged metabolic stress, arteries stiffen and lose elasticity, and stiff pipes carry higher pressure regardless of volume. Magnesium deficiency itself has been proposed as one contributor to that stiffening (Ref 7).

Structural change is slower to move, and I would not promise anyone that it fully reverses. The reasonable goal is to stop it worsening and to give the vessel wall the conditions to remodel over months: a sustained low insulin load, adequate minerals, movement, sleep, and sunlight. One popular mechanistic idea is that vitamins D3 and K2 help keep calcium in bone and out of arterial walls. It is biologically plausible, but the human evidence is thin. A vitamin K2 (menaquinone) study reporting effects on arterial stiffness, for example, was a small, uncontrolled, single-arm trial of 26 people, which cannot establish benefit (Ref 24). I mention it as a hypothesis to watch, not a recommendation.

What this means in practice

The single question I keep returning to for any test or intervention is whether it changes what I would actually do. For blood pressure, the answer reorders the usual advice.

First, look upstream. If someone has hypertension alongside excess weight, a high-carbohydrate diet, or markers of insulin resistance, the highest-leverage change is usually reducing the insulin load through diet, not shaving grams off salt. The blood-pressure fall on lower-carbohydrate eating is modest on average (Ref 16), but it arrives bundled with weight, triglyceride, and glucose improvements, and it addresses a plausible root rather than a symptom.

Second, transition at a pace that suits the person. Someone eating carbohydrate throughout the day may need months, whereas someone already eating mostly protein and fat may need only weeks.

Third, treat supplements as temporary scaffolding, not the building. Magnesium and potassium can help while the diet is corrected, within the safety limits above, but the aim is to restore the body's own ability to handle minerals from food.

Fourth, move. A meta-analysis of 93 trials and more than 5,000 people found that regular endurance and resistance training lower resting blood pressure, with the largest reductions in people who are already hypertensive (Ref 25).

Finally, a caution that matters more than any of the above. None of this is a reason to stop prescribed blood-pressure medication. If you want to change your treatment, do it with the clinician who prescribed it, using these ideas to inform that conversation rather than to replace it. Salt is not the enemy for most people. A system that cannot handle salt is the problem worth fixing.

Disclosures

I run a private clinic and offer paid consultations, so I benefit commercially when readers find this kind of analysis useful. I work primarily through a metabolic and lower-carbohydrate lens, and my interests in nutrition and metabolic medicine predispose me to emphasise insulin over sodium. I have tried to counter that bias by citing the mainstream salt-reduction evidence at its strongest and by noting the limitations of every study, including those that support my view. Nothing here is individual medical advice. If you would like help applying this to your own circumstances I offer consultations, though there are many excellent clinicians who work this way and you do not have to see me.

References

1. He FJ, Li J, MacGregor GA. Effect of longer term modest salt reduction on blood pressure: Cochrane systematic review and meta-analysis of randomised trials. BMJ, 2013. https://doi.org/10.1136/bmj.f1325

2. Huang L, Trieu K, Yoshimura S, et al. Effect of dose and duration of reduction in dietary sodium on blood pressure levels: systematic review and meta-analysis of randomised trials. BMJ, 2020. https://doi.org/10.1136/bmj.m315

3. Luzardo L, Noboa O, Boggia J. Mechanisms of Salt-Sensitive Hypertension. Current Hypertension Reviews, 2015. https://doi.org/10.2174/1573402111666150530204136

4. Zhang X, Li Y, Del Gobbo LC, et al. Effects of Magnesium Supplementation on Blood Pressure: A Meta-Analysis of Randomized Double-Blind Placebo-Controlled Trials. Hypertension, 2016. https://doi.org/10.1161/HYPERTENSIONAHA.116.07664

5. Dickinson HO, Nicolson DJ, Campbell F, et al. Magnesium supplementation for the management of essential hypertension in adults. Cochrane Database of Systematic Reviews, 2006. https://doi.org/10.1002/14651858.CD004640.pub2

6. Dibaba DT, Xun P, Song Y, et al. The effect of magnesium supplementation on blood pressure in individuals with insulin resistance, prediabetes, or noncommunicable chronic diseases: a meta-analysis of randomized controlled trials. American Journal of Clinical Nutrition, 2017. https://doi.org/10.3945/ajcn.117.155291

7. Kostov K, Halacheva L. Role of Magnesium Deficiency in Promoting Atherosclerosis, Endothelial Dysfunction, and Arterial Stiffening as Risk Factors for Hypertension. International Journal of Molecular Sciences, 2018. https://doi.org/10.3390/ijms19061724

8. Sales CH, Nascimento DA, Medeiros ACQ, et al. There is chronic latent magnesium deficiency in apparently healthy university students. Nutricion Hospitalaria, 2014. https://doi.org/10.3305/nh.2014.30.1.7510

9. Bohn T, Davidsson L, Walczyk T, Hurrell RF. Phytic acid added to white-wheat bread inhibits fractional apparent magnesium absorption in humans. American Journal of Clinical Nutrition, 2004. https://doi.org/10.1093/ajcn/79.3.418

10. Hurrell RF. Influence of vegetable protein sources on trace element and mineral bioavailability. Journal of Nutrition, 2003. https://doi.org/10.1093/jn/133.9.2973S

11. Davis DR, Epp MD, Riordan HD. Changes in USDA food composition data for 43 garden crops, 1950 to 1999. Journal of the American College of Nutrition, 2004. https://doi.org/10.1080/07315724.2004.10719409

12. DeFronzo RA, Cooke CR, Andres R, Faloona GR, Davis PJ. The effect of insulin on renal handling of sodium, potassium, calcium, and phosphate in man. Journal of Clinical Investigation, 1975. https://doi.org/10.1172/JCI107996

13. Blumenthal SA. Observations on sodium retention related to insulin treatment of experimental diabetes. Diabetes, 1975. https://doi.org/10.2337/diab.24.7.645

14. Weidmann P, Bohlen L, de Courten M. Insulin resistance and hyperinsulinemia in hypertension. Journal of Hypertension (Supplement), 1995. https://doi.org/10.1097/00004872-199508001-00010

15. Brands MW, Hall JE. Insulin resistance, hyperinsulinemia, and obesity-associated hypertension. Journal of the American Society of Nephrology, 1992. https://doi.org/10.1681/ASN.V351064

16. Bueno NB, de Melo IS, de Oliveira SL, da Rocha Ataide T. Very-low-carbohydrate ketogenic diet v. low-fat diet for long-term weight loss: a meta-analysis of randomised controlled trials. British Journal of Nutrition, 2013. https://doi.org/10.1017/S0007114513000548

17. Huang DY, Boini KM, Friedrich B, et al. Blunted hypertensive effect of combined fructose and high-salt diet in gene-targeted mice lacking functional serum- and glucocorticoid-inducible kinase SGK1. American Journal of Physiology - Regulatory, Integrative and Comparative Physiology, 2005. https://doi.org/10.1152/ajpregu.00382.2005

18. Mene P, Punzo G. Uric acid: bystander or culprit in hypertension and progressive renal disease? Journal of Hypertension, 2008. https://doi.org/10.1097/HJH.0b013e32830e4945

19. Feig DI, Kang DH, Nakagawa T, Mazzali M, Johnson RJ. Uric acid and hypertension. Current Hypertension Reports, 2006. https://doi.org/10.1007/s11906-006-0005-z

20. Choi YJ, Yoon Y, Lee KY, et al. Uric acid induces endothelial dysfunction by vascular insulin resistance associated with the impairment of nitric oxide synthesis. FASEB Journal, 2014. https://doi.org/10.1096/fj.13-247148

21. Whelton PK, He J, Cutler JA, et al. Effects of oral potassium on blood pressure: meta-analysis of randomized controlled clinical trials. JAMA, 1997. https://doi.org/10.1001/jama.1997.03540440058033

22. Filippini T, Naska A, Kasdagli MI, et al. Potassium Intake and Blood Pressure: A Dose-Response Meta-Analysis of Randomized Controlled Trials. Journal of the American Heart Association, 2020. https://doi.org/10.1161/JAHA.119.015719

23. Neal B, Wu Y, Feng X, et al. Effect of Salt Substitution on Cardiovascular Events and Death. New England Journal of Medicine, 2021. https://doi.org/10.1056/NEJMoa2105675

24. Ikari Y, Torii S, Shioi A, Okano T. Impact of menaquinone-4 supplementation on coronary artery calcification and arterial stiffness: an open label single arm study. Nutrition Journal, 2016. https://doi.org/10.1186/s12937-016-0175-8

25. Cornelissen VA, Smart NA. Exercise training for blood pressure: a systematic review and meta-analysis. Journal of the American Heart Association, 2013. https://doi.org/10.1161/JAHA.112.004473

26. Khosla UM, Zharikov S, Finch JL, et al. Hyperuricemia induces endothelial dysfunction. Kidney International, 2005. https://doi.org/10.1111/j.1523-1755.2005.00273.x

Vitamin D: A Statistical Error in the Recommended Intake, and Why a Blood Level Is Not the Same as a Benefit

Executive summary

• The public recommended intake for vitamin D — 600 IU (international units) per day for most adults, 800 IU from the age of 71 — was set by the Institute of Medicine, whose figures inform health authorities across North America and, indirectly, much of the world. It was intended to raise blood 25-hydroxyvitamin D, written 25(OH)D and read aloud as "25 hydroxy D", to at least 50 nmol/L (equivalently 20 ng/mL) in 97.5% of the population (Ref 1).

• A short 2014 analysis showed that the calculation behind this figure contained a genuine statistical error: it treated "97.5% of the study averages" as if it meant "97.5% of individuals". Read correctly, 600 IU per day would put only the average person, not the great majority, near the target (Ref 2).

• The same paper estimated that reaching 50 nmol/L in 97.5% of individuals might require something like 8,895 IU per day. The authors themselves cautioned that this number lies far beyond the doses that were actually studied and should be read as an illustration of the direction of the error, not as a dosing instruction (Ref 2). An independent group confirmed the underlying mistake (Ref 3); a later analysis put the figure much lower and showed it depends heavily on body weight (Ref 4).

• Reaching a blood number is not the same as gaining a health outcome. Large randomised trials of supplementation — VITAL and D-Health — found no reduction in cancer, cardiovascular events, fractures, or death in generally healthy adults who were not selected for deficiency (Ref 6, Ref 7, Ref 8).

• Genetic evidence suggests the risk attached to low vitamin D is real but concentrated in genuine deficiency: the benefit is in correcting a low level, not in pushing an already-adequate level higher (Ref 9). The expert guidelines themselves have moved over the past decade and still disagree with one another (Ref 5, Ref 10, Ref 11).

• I did take very high-dose vitamin D myself for several years without apparent harm. That is a single anecdote, it sat above the official upper limit, and it is not a recommendation for anyone else.

• My own preference is to prioritise the routes the body evolved to use — sensible sunlight and animal foods such as oily fish, egg yolks, and liver — and to correct genuine deficiency thoughtfully, rather than to treat a laboratory target as a goal in its own right. I declare my interests, and the limits of my confidence, below.

Introduction

Vitamin D occupies an unusual position in public health. It is the subject of confident official numbers — an intake to aim for, a blood level to reach — and, at the same time, of a large and often contradictory research literature. My aim in this piece is narrow and, I hope, defensible: to look at where the most quoted number, the recommended intake, actually comes from; to show that its derivation contained a real and independently confirmed error; and then to be equally honest about what that error does and does not license us to conclude. It is easy to move from "the recommendation was miscalculated" to "everyone is dangerously deficient and should take large doses". The first statement is well supported. The second does not follow from it, and the trial evidence does not support it.

I should declare my interests plainly. I run a private clinic and I offer paid consultations, so I have a commercial interest in people valuing clinical reasoning of the kind set out here. I sell no vitamin D product and recommend no brand. I also hold a stated dietary bias: I lean towards animal-source nutrition, and readers should weigh what follows in that light. On qualifications, so the reader knows the lens: I originally trained in dentistry, then in dermatology, and separately in anti-aging medicine and in metabolic medicine. None of that makes me right. It shapes which questions I ask.

One general point of method before the detail. A great deal of nutritional advice is built on reference ranges and target numbers whose origins are rarely examined. When a guideline states that a level "should" be a particular figure, it is worth asking where that figure came from, which people were studied to produce it, and whether the reasoning survives scrutiny. Vitamin D turns out to be an unusually clean example of why that habit matters.

What vitamin D actually is

Calling it a vitamin is slightly misleading. A vitamin, strictly, is something you must obtain from the diet because the body cannot make it. Vitamin D is largely made by the body itself: ultraviolet B light striking the skin converts a cholesterol-derived molecule into vitamin D3, which the liver then converts to 25(OH)D — the circulating form measured in blood tests — and the kidney activates further into the hormone that acts on tissues (Ref 12). In other words it behaves more like a hormone the body manufactures on demand than like a nutrient imported through food. A useful analogy is that the body is less a shopper buying a finished product and more a workshop building its own tool when the raw material, sunlight, is available.

This matters for everything that follows, because it means there are two quite different ways to raise vitamin D — through the skin, which is regulated, and through the mouth as a supplement, which is not — and because the blood test everyone quotes, 25(OH)D, is a storage marker rather than a direct measure of what the active hormone is doing in your tissues.

Where the "600 IU" number comes from — and the error inside it

The recommended intake most people have met is the Institute of Medicine's Recommended Dietary Allowance: the intake judged sufficient to meet the requirements of 97.5% of healthy people. For vitamin D that was set at 600 IU per day for those aged 1 to 70, and 800 IU thereafter, chosen to achieve a blood 25(OH)D of at least 50 nmol/L (20 ng/mL) in that same 97.5% of the population, on the basis of bone health (Ref 1).

To derive it, the committee deliberately used supplementation studies conducted in winter and at high latitudes, so that sunlight would contribute as little as possible and the effect of the dose alone could be seen (Ref 1, Ref 2). They plotted blood level against dose across those studies and read off the intake that corresponded to the lower edge of the range.

Here is the error, and it is worth stating precisely because it is specific rather than rhetorical. The committee calculated the lower confidence limit of the study averages — and then interpreted it as if it described individuals. A 2014 letter in Nutrients pointed out that "97.5% of study averages lie above this line" is a completely different statement from "97.5% of individuals lie above this line", because individual people vary far more than the averages of whole studies do (Ref 2). The everyday version: if you take the average exam mark of each of thirty classes, those class averages cluster tightly together; the individual pupils inside those classes are spread much more widely. Reading a number that describes the clustered class averages and assuming it describes the spread-out pupils will mislead you, and it will mislead you in a predictable direction — towards a figure that is too low.

Corrected, the same data implied that 600 IU per day would only get 97.5% of individuals above roughly 27 nmol/L, not 50 (Ref 2). And to actually reach 50 nmol/L in 97.5% of individuals, the authors' regression pointed to something on the order of 8,895 IU per day.

That figure has since been quoted widely, so it deserves careful handling. The authors were explicit that it "is far beyond the range of studied doses", that "caution is warranted when interpreting this estimate", and that it exceeds the official tolerable upper intake of 4,000 IU per day (Ref 2). In other words, the headline number is an extrapolation the authors themselves flagged as unreliable in magnitude. What is robust is the direction and the existence of the mistake, not the precise size of the correction. An independent group — including researchers who had spent careers on vitamin D — wrote in to confirm that the statistical reasoning was indeed wrong and reached the same conclusion using a separate cohort (Ref 3). A later analysis by the original authors, drawing on more than a hundred studies and nearly fourteen thousand people, put the intake needed to reach 50 nmol/L in most people closer to around 3,000 IU per day, rising with body weight to markedly more in people with obesity, and argued that a single population-wide allowance is neither desirable nor feasible given how much individuals differ (Ref 4).

So the faithful summary is this. The specific arithmetic behind a very widely used recommendation was demonstrably wrong, in a way two independent groups agree on, and it understated the intake needed to hit its own blood-level target. That is a genuine and somewhat remarkable finding, and it is a good reason to treat reference numbers as arguments to be examined rather than facts to be memorised. It is not, by itself, a reason to swallow nine thousand units a day.

A marker is not an outcome

This is the point where honesty requires me to slow my own argument down, including where I have made it myself. Everything above concerns a blood marker, 25(OH)D, and the intake needed to move it. Whether moving that marker in an already-healthy person changes anything they care about — living longer, breaking fewer bones, avoiding cancer or heart disease — is a separate question, and it has been tested directly.

The largest such test, the VITAL trial, randomised nearly 26,000 adults in the United States to 2,000 IU of vitamin D per day or placebo. Over about five years it found no reduction in the risk of invasive cancer of any type and no reduction in major cardiovascular events, and it recorded no excess of hypercalcaemia in the vitamin D group (Ref 6). A dedicated analysis of the same trial found no reduction in fractures — total, non-vertebral, or hip — among these generally healthy older adults, and no benefit even in the subgroup who started with lower blood levels (Ref 7). The separate Australian D-Health trial, giving 60,000 IU monthly to more than 21,000 older people for five years, found no reduction in overall mortality; if anything, its exploratory analyses raised a cautionary signal for cancer death, leading the authors to invoke the precautionary principle against routine high-dose supplementation in people who are already replete (Ref 8).

The honest reading of these trials needs its own caveat, and it cuts both ways. Each of them enrolled largely vitamin-D-replete populations who were not selected for deficiency (Ref 6, Ref 7). So they are excellent tests of one specific policy — "give extra vitamin D to broadly healthy people regardless of their level" — and that policy failed. They are weaker tests of a different question: "does correcting genuine, symptomatic deficiency help?" A trial in which most participants are already adequate cannot easily show benefit from topping them up, in the same way that filling an already-full tank tells you nothing about how far an empty car could have gone. The blood level is a fuel gauge; a full gauge confirms the tank is not empty, but adding fuel past full does not extend the journey.

Genetic evidence helps square the circle. A Mendelian randomisation study in over 300,000 UK Biobank participants — an approach that uses inherited differences in vitamin D to approximate a lifelong natural experiment, and so is less prone to the confounding that plagues observational data — found that the relationship between vitamin D and death is L-shaped: risk falls steeply as low levels rise towards about 50 nmol/L, and then flattens (Ref 9). That is a coherent story. Genuine deficiency appears genuinely harmful; correcting it plausibly matters; pushing an already-adequate level higher does little. It reconciles the observational alarm about "widespread deficiency" with the null supplementation trials, and it is, to my mind, the most defensible current position.

The guidelines have been arguing with themselves

If the science were settled, the official bodies would agree. They do not, and the disagreement is instructive rather than embarrassing.

In 2011 the Endocrine Society issued a guideline that leant towards higher intakes and higher target levels than the Institute of Medicine, and recommended measuring 25(OH)D in people at risk of deficiency. Even then, it stated that there was not sufficient evidence to screen people who are not at risk, nor to prescribe vitamin D for cardiovascular protection (Ref 5). The Institute of Medicine, for its part, had already warned in the opposite direction: that the prevalence of vitamin D "inadequacy" had been overestimated, that higher blood levels were not consistently linked to greater benefit, that some outcomes showed a U-shaped curve with risk at both low and high levels, and that laboratory reference ranges urgently needed reassessing to avoid both under- and over-treatment (Ref 1).

By 2024 the pendulum had swung again. The Endocrine Society's newer evidence review, built on randomised trials rather than associations, found little effect of supplementation on most outcomes in healthy adults aged 19 to 74, a small mortality benefit only in those over 75, a reduction in progression from prediabetes to diabetes, and a signal that high-dose intermittent dosing may increase falls compared with lower daily doses. It found no trials that established the benefit of screening the general population with a 25(OH)D test at all (Ref 10). That shift did not go unchallenged: leading vitamin D researchers argued that confining the evidence to randomised trials discards a large body of association data on non-skeletal benefits (Ref 11). I do not think this argument is close to resolved, and where the optimal target level sits — if a single one exists — is something on which I am still collecting data rather than offering you a firm number.

On taking very high doses — my own experiment, and why it is not advice

For transparency, and because I have discussed it publicly before, I will state my own history. For a period of years I took vitamin D at doses well above the official upper limit, alongside vitamin K2, and experienced no ill effects that I could detect. I set this out because it is true and because concealing it would be its own kind of dishonesty. But I want to be careful about what it is worth, because a personal story is the weakest form of evidence there is, and I would rather you learned the reasoning than copied the behaviour.

That was many times the official tolerable upper intake of 4,000 IU per day and far above anything tested in the trials above. Vitamin D toxicity is real: because it is fat-soluble it accumulates, and the principal danger is hypercalcaemia — a rise in blood calcium that can damage the kidneys and other organs — generally described once blood levels climb above roughly 250 nmol/L (Ref 14). The reason we lean on case reports for this, rather than clean dose-response data, is that deliberately poisoning volunteers to find the threshold would be unethical (Ref 15). And those case reports show how idiosyncratic the response can be: one published patient reached a 25(OH)D of 746 ng/mL through a dosing error yet had normal blood calcium and no symptoms (Ref 15). That someone can tolerate an extreme level tells you about the width of human variation; it tells you nothing about whether you would be the person who is fine or the person in the next report who is not. I cannot see you through a screen, I do not know your kidney function or your calcium handling, and the sensible reading of the toxicity literature is caution, not reassurance.

My rationale for the co-administered K2 was mechanistic, and here the biochemistry is well established rather than speculative. Vitamin K is the cofactor that gamma-carboxylates a family of calcium-handling proteins: it activates osteocalcin, which helps deposit calcium into the bone matrix, and matrix Gla protein, which inhibits the deposition of calcium in arterial walls (Ref 16, Ref 18, Ref 19). Adequate vitamin K therefore helps steer calcium toward bone and away from blood vessels, which is exactly the concern when taking large amounts of vitamin D, and low vitamin K activity tracks with arterial stiffness and vascular calcification. It is a reasonable hypothesis that pairing the two might steer calcium sensibly. But I should apply my own standard here: when this was tested directly, a randomised trial of vitamin K2 plus vitamin D in men with existing aortic-valve calcification improved the relevant biochemical marker yet did not slow the calcification itself over two years (Ref 17). So the mechanism is plausible and the co-factor logic is defensible, but "plausible mechanism" is not "proven protection", and I hold the K2 rationale as a reasonable hedge rather than an established safeguard. If you are already taking prescribed high-dose vitamin D and are minded to change anything, speak to your prescriber before you do — that is a conversation to have with the clinician who knows your bloods, not a decision to make from an article.

Sunlight and food first

If vitamin D is something the body was designed to make from sunlight, the obvious first question is not "which supplement" but "why has the ordinary supply fallen short". Skin synthesis depends on ultraviolet B reaching the skin, which varies with latitude, season, time of day, and skin pigmentation; the further from the equator, the darker the winter, and the more melanin in the skin, the less is made from a given amount of sun (Ref 12). This is also why the Institute of Medicine built its recommendation on people with minimal sun in the first place (Ref 1). There is a neat piece of physiology worth knowing: cutaneous production has a built-in ceiling, because the same sunlight that makes vitamin D in the skin also breaks down any excess, so ordinary sun exposure does not cause toxicity — rather like a kettle that switches itself off at the boil. Oral megadoses bypass that safety catch entirely.

Where sunlight is genuinely insufficient, my preference is to look to the foods the nutrient naturally occupies before reaching for isolated high-dose tablets — and those foods are overwhelmingly animal foods. Oily fish are among the richest natural sources, with the interesting wrinkle that wild salmon carried several times the vitamin D of farmed salmon in one analysis (Ref 12). Egg yolks and liver contribute meaningfully; the vitamin travels with the fat, which is one more reason I am wary of advice that strips animal fat out of the diet. I am not persuaded that fortified plant foods are a like-for-like substitute for these sources, and I hold to that as a stated preference, not a proven hierarchy.

One physiological detail ties the food question back to the dosing question. Vitamin D is fat-soluble, and body fat acts as a reservoir that soaks it up: people with obesity achieve lower blood levels from the same intake because more of it is sequestered in fat tissue (Ref 13). That is the mundane explanation for a good deal of measured "deficiency", and it is also why any sensible intake is a function of the person, not a single national number — exactly the point the corrected statistics made from the other direction (Ref 4).

Should you get tested, and would it change anything?

A blood test is only worth doing if the result would change what you do. That is the filter I apply to every investigation, and 25(OH)D is a good case for it. For a broadly healthy person with reasonable sun exposure and a diet containing oily fish, eggs, and liver, a single number on a page is unlikely to change sensible behaviour, and the 2024 evidence review found no trial support for screening the general population (Ref 10). The situation is different if there is a real reason to suspect deficiency — very little sun exposure, living at high latitude, darker skin, obesity, malabsorption, being largely housebound, or symptoms that fit — because there the result can genuinely change management, and correcting a low level is the part of this whole story with the best evidence behind it (Ref 9).

I want to be careful not to talk over your own clinician here. General practitioners work within guidelines, finite time, and finite budgets, and interpreting a vitamin D level in the context of a whole person is exactly the kind of judgement they are trained for. If your level is genuinely low and symptomatic, that is worth acting on with them. If it is a borderline number in someone who feels well, the more useful question is usually how much sun and how much real food that person is getting, not which pill to add.

Conclusion

The story of the vitamin D recommendation is a good advertisement for reading the workings rather than the headline. A very widely used figure was built on a specific statistical error that understated its own target, and two independent groups agree the error was real (Ref 2, Ref 3). That genuinely should lower our confidence in tidy reference numbers. But the same discipline that exposes the error also restrains the conclusion: the corrected megadose figure was an extrapolation its own authors distrusted (Ref 2, Ref 4); moving a storage marker is not the same as improving health, and the trials that tested supplementation in healthy people came back null (Ref 6, Ref 7, Ref 8); and the benefit that does appear is concentrated in correcting genuine deficiency, not in chasing ever-higher levels (Ref 9). My own high-dose experiment is a single data point above the official limit and should be read as such. The position I actually hold is unglamorous: get sunlight where you can, eat the animal foods the nutrient lives in, correct real deficiency with a clinician rather than a target, and treat both the alarmist and the reassuring headlines about vitamin D with the same scepticism you would bring to the arithmetic behind them.

Disclosures

I run a private clinic and offer paid consultations, which is a commercial interest readers should weigh. I sell no vitamin D product and endorse no brand. I hold a stated dietary bias towards animal-source nutrition, disclosed above. I have described my own past use of high-dose vitamin D for transparency, not as advice. If you would like to discuss your own situation I am available, but for most people a good, regular relationship with a local clinician is worth more than any single consultation with me.

References

Identified and verified via PubMed; DOI links included.

1. Ross AC, Manson JE, Abrams SA, et al. The 2011 report on dietary reference intakes for calcium and vitamin D from the Institute of Medicine: what clinicians need to know. Journal of Clinical Endocrinology & Metabolism, 2011. https://doi.org/10.1210/jc.2010-2704

2. Veugelers PJ, Ekwaru JP. A statistical error in the estimation of the recommended dietary allowance for vitamin D. Nutrients, 2014. https://doi.org/10.3390/nu6104472

3. Heaney R, Garland C, Baggerly C, French C, Gorham E. Letter to Veugelers, P.J. and Ekwaru, J.P., A statistical error in the estimation of the recommended dietary allowance for vitamin D. Nutrients, 2015. https://doi.org/10.3390/nu7031688

4. Veugelers PJ, Pham TM, Ekwaru JP. Optimal Vitamin D Supplementation Doses that Minimize the Risk for Both Low and High Serum 25-Hydroxyvitamin D Concentrations in the General Population. Nutrients, 2015. https://doi.org/10.3390/nu7125527

5. Holick MF, Binkley NC, Bischoff-Ferrari HA, et al. Evaluation, treatment, and prevention of vitamin D deficiency: an Endocrine Society clinical practice guideline. Journal of Clinical Endocrinology & Metabolism, 2011. https://doi.org/10.1210/jc.2011-0385

6. Manson JE, Cook NR, Lee IM, et al. Vitamin D Supplements and Prevention of Cancer and Cardiovascular Disease (the VITAL trial). New England Journal of Medicine, 2019. https://doi.org/10.1056/NEJMoa1809944

7. LeBoff MS, Chou SH, Ratliff KA, et al. Supplemental Vitamin D and Incident Fractures in Midlife and Older Adults (VITAL). New England Journal of Medicine, 2022. https://doi.org/10.1056/NEJMoa2202106

8. Neale RE, Baxter C, Romero BD, et al. The D-Health Trial: a randomised controlled trial of the effect of vitamin D on mortality. Lancet Diabetes & Endocrinology, 2022. https://doi.org/10.1016/S2213-8587(21)00345-4

9. Sutherland JP, Zhou A, Hyppönen E. Vitamin D Deficiency Increases Mortality Risk in the UK Biobank: A Nonlinear Mendelian Randomization Study. Annals of Internal Medicine, 2022. https://doi.org/10.7326/M21-3324

10. Shah VP, Nayfeh T, Alsawaf Y, et al. A Systematic Review Supporting the Endocrine Society Clinical Practice Guidelines on Vitamin D. Journal of Clinical Endocrinology & Metabolism, 2024. https://doi.org/10.1210/clinem/dgae312

11. Holick MF. Revisiting Vitamin D Guidelines: A Critical Appraisal of the Literature. Endocrine Practice, 2024. https://doi.org/10.1016/j.eprac.2024.10.011

12. Chen TC, Chimeh F, Lu Z, et al. (incl. Holick MF). Factors that influence the cutaneous synthesis and dietary sources of vitamin D. Archives of Biochemistry and Biophysics, 2007. https://doi.org/10.1016/j.abb.2006.12.017

13. Wortsman J, Matsuoka LY, Chen TC, Lu Z, Holick MF. Decreased bioavailability of vitamin D in obesity. American Journal of Clinical Nutrition, 2000. https://doi.org/10.1093/ajcn/72.3.690

14. Levita J, Wilar G, Wahyuni I, et al. Clinical Toxicology of Vitamin D in Pediatrics: A Review and Case Reports. Toxics, 2023. https://doi.org/10.3390/toxics11070642

15. Chakraborty S, Sarkar AK, Bhattacharya C, Krishnan P, Chakraborty S. A nontoxic case of vitamin D toxicity. Laboratory Medicine, 2015. https://doi.org/10.1309/LM5URN1QIR7QBLXK

16. Hariri E, Kassis N, Iskandar JP, et al. Vitamin K – a neglected player in cardiovascular health: a narrative review. Open Heart, 2021. https://doi.org/10.1136/openhrt-2021-001715

17. Diederichsen ACP, Lindholt JS, Möller S, et al. Vitamin K2 and D in Patients With Aortic Valve Calcification: A Randomized Double-Blinded Clinical Trial. Circulation, 2022. https://doi.org/10.1161/CIRCULATIONAHA.121.057008

18. Villa JKD, Diaz MAN, Pizziolo VR, Martino HSD. Effect of vitamin K in bone metabolism and vascular calcification: a review of mechanisms of action and evidences. Critical Reviews in Food Science and Nutrition, 2017. https://doi.org/10.1080/10408398.2016.1211616

19. Gröber U, Reichrath J, Holick MF, Kisters K. Vitamin K: an old vitamin in a new perspective. Dermato-endocrinology, 2015. https://doi.org/10.4161/19381972.2014.968490

Muscle and Longevity: Sarcopenia, Strength and the Difference Between Correlation and Cause

Executive summary

• Muscle mass, and more importantly muscle strength, are repeatedly associated with lower death rates and better health in later life. That is an association. No study has shown, or realistically could show, that building muscle causes a longer life (Ref 2, Ref 3).

• Sarcopenia, the gradual loss of muscle with age, tracks with disability and higher mortality (Ref 1, Ref 2). The direction of cause is genuinely uncertain: illness and inactivity waste muscle, just as lost muscle may worsen the course of illness.

• Grip strength, and the ability to lower yourself to the floor and stand back up without using your hands, are useful markers of overall robustness (Ref 3, Ref 4, Ref 5). I treat them as a dashboard warning light, not the engine, and I do not train them for their own sake.

• Skeletal muscle is the body's largest site for clearing dietary glucose and a major place where fuel is oxidised (Ref 7, Ref 8). More working muscle may put you in a better position to handle carbohydrate, which is not the same as preventing diabetes.

• Muscle behaves as a secretory organ, releasing signalling molecules called myokines (Ref 9). That is real biology. The onward leap to "muscle prevents dementia" is one I am not willing to make.

• Resistance training is associated with lower mortality and, in my view, is worth doing (Ref 13). I favour training for strength and quality of movement rather than size for its own sake.

• Very large physiques, especially those built with anabolic steroids, carry a measurable cardiovascular cost (Ref 12). That is much of the answer to why some of the most muscular people still die young.

• I declare my interests plainly. I am a former competitive powerlifter, so I am naturally inclined to value this kind of training, and I sell no product mentioned anywhere in this article.

Introduction

"Muscle is longevity" has become one of those phrases that gets repeated until it sounds like a settled fact. It is not, and the honest version is more interesting than the slogan. There is a real and reasonably consistent signal in the data linking muscle, and particularly strength, to a longer and healthier life. There is also a hard limit on what that signal can prove. I want to walk through both.

I should be clear at the outset about what I am and am not claiming, because this is a subject that invites overstatement in both directions. I am not claiming that adding muscle is proven to extend your life; the studies cannot carry that weight, and I will explain why. I am also not dismissing muscle as a vanity pursuit; the association with health is genuine and, mechanistically, quite plausible. The narrower and more defensible position is this: muscle mass and strength are correlated with health and survival, building and keeping muscle is a sensible thing to do for reasons I find persuasive, but "correlated with" is not "the cause of," and the distinction matters enormously once you start deciding what to actually do with your training.

By way of background, and so my biases are visible: I trained first in dentistry, then in dermatology, and separately in anti-aging medicine and in metabolic medicine, and I hold a qualification in sports nutrition. Before medicine I competed for years as a powerlifter, at times training alongside some of the best-known lifters in the sport, including Andy Bolton, the first man to deadlift over 1,000 pounds, and I still work with athletes across a range of levels, including retired ones. That background is useful experience and it is also a bias, and I would rather you saw it than have to guess at it.

There is a question sitting underneath the whole topic, and it is worth stating in its blunt form. Does muscle make you live longer, or does muscle simply travel in the company of people who were going to live longer anyway? Hold that question in mind, because most of what follows is an attempt to answer it honestly rather than conveniently.

What sarcopenia is, and what the mortality data can say

Sarcopenia is the medical term for the age-related loss of muscle mass and strength. The current European consensus definition now places low muscle strength, rather than low muscle mass alone, at the front of the diagnosis, which is itself telling: strength, not size, is what best flags the problem (Ref 1). I am deliberately not going to quote you a headline figure for how much muscle we supposedly lose each year, because those numbers are usually extrapolated from one studied group onto the whole of humanity, and I do not think that kind of variable travels well between populations. That we tend to carry less muscle in old age than in youth is not in dispute; the precise arithmetic is less trustworthy than it looks.

What the literature does show, fairly consistently, is that people with sarcopenia have higher death rates than those without. A systematic review and meta-analysis of older nursing-home residents, for example, found sarcopenia to be a predictor of all-cause mortality (Ref 2). Set alongside the strength data below, the association is not subtle.

Here is the necessary caution, and it is the whole argument of this article in miniature. This is observational, and the arrow of cause can point either way. Losing muscle may worsen your health; equally, being ill, frail, inactive or undernourished wastes muscle, so low muscle can be a consequence of poor health rather than its cause. In all likelihood both run at once, feeding each other. A study of this kind can establish that muscle and survival move together. It cannot, on its own, tell you that putting muscle on a sick person turns them into a well one.

I will give one clinical memory, offered as illustration and not as evidence. Standing in for a senior house officer in the emergency department at North Manchester General Hospital, I saw two elderly patients arrive after similar falls. One walked away with a bruise. The other sustained multiple fractures and spent weeks in hospital. The one who walked away had, in effect, bought themselves what I think of as biological insurance and a biological pension: a body robust enough to absorb an accident that flattened someone else. That is what "longevity muscle" means to me. It is not a physique. It is a reserve.

Grip strength and getting off the floor: markers, not levers

Two measures come up again and again in this field. The first is grip strength. In a prospective study of roughly half a million people in the UK Biobank, lower grip strength was associated with higher all-cause mortality and with cardiovascular disease, respiratory disease and cancer (Ref 3). A dose-response meta-analysis has since mapped how the risk changes across the range of handgrip values (Ref 4). The second is the ability to sit down onto the floor and rise again without pushing off with a hand or knee, sometimes called the sitting-rising test; scoring poorly on it has been associated with higher all-cause mortality (Ref 5). There is also evidence that a strength measure predicts poor physical performance better than a measure of muscle mass does (Ref 6), which fits the shift in the sarcopenia definition I mentioned above.

I find these genuinely useful as markers. What I do not do is treat them as levers. I do not train my grip, or practise getting off the floor, under the specific belief that doing so will extend my life. My reasoning is the one running through this whole piece: grip strength is a proxy for something larger, namely overall strength, muscle quality, coordination and body composition. Think of it as the warning light on a car dashboard. The light tracks the state of the engine reliably, but unscrewing the bulb does not fix the engine, and polishing the light does not make the car run better. If my grip improves as a by-product of getting generally stronger, good. Chasing the marker itself, in the hope the outcome follows, is to mistake the light for the engine. I hold that view provisionally, and I am open to where the exceptions might lie.

Muscle and blood sugar: a reservoir for glucose

Across my clinical career I noticed, as a personal observation and emphatically not as a formal statistic, that patients carrying more muscle seemed less commonly diabetic. An impression from one clinician's practice describes nothing about the world at large, so treat it as a prompt for the mechanism rather than a finding.

The mechanism is reasonably well understood. Skeletal muscle is the principal site at which the body disposes of glucose from the bloodstream under the influence of insulin (Ref 8). Muscle also stores glucose as glycogen, so a larger, well-trained muscle mass acts as a bigger reservoir into which dietary glucose can be parked. Picture a sponge: a bigger, healthier sponge soaks up a spill faster and holds more of it. And because muscle is dense with mitochondria, the small structures that oxidise fuel (that is, chemically strip electrons from fuel molecules to release usable chemical fuel in the form of ATP), it also provides more of the engines needed to actually use what it takes up. Consistent with this, higher relative muscle mass has been associated, in a large cross-sectional analysis, with better insulin sensitivity and a lower prevalence of pre-diabetes (Ref 7).

Two honest caveats. First, the human muscle-mass data here are cross-sectional, a single snapshot in time, so once again they show association rather than proof of cause. Second, none of this means muscle "prevents" diabetes. What it means is that a more muscular, more active body may be in a better position to clear glucose when you eat carbohydrate. It puts you in a stronger position; it does not issue a guarantee, and how much it helps any particular person depends on their circumstances, which I cannot assess through a screen.

It is worth adding a note that often causes confusion. In someone eating very few carbohydrates, muscle can deliberately down-regulate how much glucose it lets in when insulin signals, in order to spare glucose for the tissues that most rely on it. This is a normal adaptation, best called physiological insulin resistance, and it is not the same thing as the pathological insulin resistance of metabolic disease, even though a single fasting blood test can look superficially similar. Reading the one without the other in mind is a common way to frighten a healthy low-carbohydrate eater unnecessarily.

Muscle as a gland: myokines and the brain

Contracting muscle releases signalling molecules, collectively called myokines, into the circulation; in this sense muscle behaves not just as a motor but as a secretory organ, effectively a gland (Ref 9). Some of these molecules have anti-inflammatory actions, and the biology is real and interesting. The everyday version is that a working muscle posts chemical messages to the rest of the body, and some of those messages reach the brain.

From here the popular story runs straight to "so building muscle wards off dementia," and this is where I get off. There are associations in that direction: lower grip strength has been linked with cognitive decline (Ref 10), and low grip strength appears among the risk factors identified for young-onset dementia in large datasets (Ref 11). But grip strength in these studies is doing its usual job as a marker of general health, and the people with more of it differ in many other ways, from activity levels to overall disease burden, that also affect the brain. Reading these associations as "muscle protects the brain" runs well ahead of what the study design can support. My own position is that dementia is multifactorial, that muscle is at most one strand among many, and that I would not offer muscle-building as a cognitive insurance policy. That muscle secretes bioactive molecules is not in question; what those molecules do for the aging human brain, in the doses a normal training life produces, is something I am genuinely still collecting data on.

Why the muscular can still die young: size, drugs and the heart

This is the part that punctures the slogan, and it needs saying plainly. If muscle straightforwardly caused longevity, the most muscular people would be the longest-lived, and they are not. Some visibly powerful physiques belong to people who die early. Two things reconcile that with everything above.

The first is that muscle mass is correlated with health, not a direct cause of it, so a large physique is not automatically a healthy one. The second is more specific: a good deal of extreme size is built with anabolic-androgenic steroids, and those drugs carry a real cardiovascular cost. In a cross-sectional study of long-term male weightlifters, those with years of steroid use showed reduced heart pump function and higher volumes of coronary artery plaque than non-users, with the plaque burden tracking the cumulative lifetime dose (Ref 12). Anabolic steroids also tend to lower HDL and raise LDL, the lipoprotein particle that carries cholesterol through the blood, shifting the whole picture in an unfavourable direction.

The caveats apply here too, and I will not pretend otherwise: this is cross-sectional, and men who use steroids differ from those who do not in ways beyond the drugs. But the dose-response relationship with coronary plaque is exactly the kind of internal consistency that makes me take the signal seriously rather than wave it away. This is the sharp end of the vanity-versus-longevity distinction. Size pursued for appearance, and chemically forced past what the body would build on its own, is close to the opposite of what I mean by longevity muscle. My own view, held firmly, is to avoid performance-enhancing drugs wherever you possibly can.

What training for longevity muscle looks like, in practice

Before any specifics, the honest framing. Resistance training is associated with lower all-cause mortality in systematic review and meta-analysis (Ref 13). That association is, predictably, observational and subject to the healthy-user problem: people who lift weights tend to differ from people who do not in diet, income, sleep and much else. I still think the case for doing it is strong, because the mechanistic story and the associations point the same way, but I want you to see that I am reasoning from convergent plausibility, not from proof.

High-intensity interval work earns a specific mention. Brief, hard intervals drive the same molecular machinery that builds mitochondria and raises the muscle's capacity to oxidise both glucose and fat (Ref 14). I favour that over long, steady, monotonous cardiovascular sessions, both for what it does inside the muscle and for the heart.

What follows is my own approach, and a starting point for discussion, not a prescription for you specifically. I cannot see you through a screen, and the right programme depends on your history, your joints and your circumstances.

• Take the working weight to genuine muscular failure at roughly eight repetitions, for up to two sets. Quality over quantity.

• Never train two days in a row, and only train again once you have recovered from the previous session. Adaptation happens during recovery, the way plaster sets between coats, not during the session itself.

• Alternate the emphasis through the year, spending a couple of months oriented toward size and then a couple oriented toward strength, and back again.

• Prioritise form over the number on the bar. The load is a means, not the goal.

• If you develop any niggle in a joint or soft tissue, have it looked at early by a good local clinician or physiotherapist rather than training through it. Preserving mobility over decades is worth more than any single session, and this is exactly where an ongoing relationship with a clinician who knows you earns its keep.

• Avoid performance-enhancing drugs. I lifted for years mostly for strength rather than size (my own best deadlift was over 250 kg at a bodyweight around 72 kg), which I think, though I hold it as an opinion, is the more health-favourable emphasis of the two. Training to the level I once did is not itself healthy, given what it asks of the body; done at a sensible level it is, in my view, entirely reasonable. I still train alongside sports such as cricket and football, and I make no claim that my particular mix is optimal for anyone else. On the finer question of exactly how much volume is ideal, the field genuinely disagrees, and I am still collecting data rather than pretending to a certainty I do not have.

Feeding muscle: protein, and the low-carbohydrate question

You cannot build or defend muscle without the raw materials, and the raw material is protein, specifically the essential amino acids and above all leucine, which acts as a trigger for muscle protein synthesis. Higher-quality, more complete proteins do this job better, and the protein quality that matters most for muscle comes from complete animal sources: meat, fish, eggs and dairy supply the full complement of essential amino acids in the right proportions and in a highly usable form (Ref 15). Interventions that supply leucine-rich protein have been shown to improve measures of muscle in older adults at risk of sarcopenia (Ref 16). If I had to name a single dietary lever for muscle, it would be sufficient complete animal-source protein, not a shelf of supplements.

The more contested question is the metabolic state you train in. I lean toward a low-carbohydrate approach, and toward training in a fat-adapted, sometimes ketogenic state. Recent work from Professor Tim Noakes and colleagues describes how a low-carbohydrate, high-fat pattern shifts the body toward oxidising fat at higher exercise intensities (Ref 17). A 2026 review by the same group, drawing on more than a hundred years of evidence, goes further: it reports that fat-adapted athletes achieve exceptional rates of fat oxidation and equivalent performance despite low carbohydrate intake, and argues that carbohydrate is not an obligatory fuel for exercise (Ref 20). That is mechanistically the direction I find compelling, though I hold it as a lean rather than a settled verdict.

Honesty requires me to place the counter-evidence next to my own lean, not to hide it. Controlled work by Louise Burke and colleagues found that adapting to a low-carbohydrate, high-fat diet, while it did raise fat oxidation, impaired exercise economy and performance in elite endurance athletes despite adequate glycogen (Ref 18). A randomised crossover trial reported reduced anaerobic, high-intensity performance on a ketogenic diet (Ref 19). These are real results and I will not explain them away. What I will say is that some of the apparent impairment may reflect too short an adaptation period and large differences between individuals, that the field is genuinely unsettled, and that I am not claiming carbohydrate is required either to perform or to build muscle. My position is a lean, grounded in the biochemistry, not a settled verdict, and it is one of the places I am still actively collecting data. Where the evidence pulls against me, I would rather show you than bury it.

Conclusion

Muscle is worth building and worth keeping. I believe that, I train on it, and I would say the same to almost anyone who asked. But the reasons are quieter and more honest than the slogan suggests. Muscle, and especially strength, is consistently associated with a longer and more capable life; the mechanisms, from glucose handling to the myokines a working muscle secretes, are plausible; and resistance training sits alongside those associations rather than against them. What none of it establishes is direct cause. No study can, because you cannot follow a person from birth to death while holding every other variable still, and so the honest ceiling on this entire field is correlation, not proof. That some of the most muscular people die young, and that some of the longest-lived people were never notably muscular, should keep us humble about how much of the story muscle really tells.

What I take from that is not defeatist but practical. Build a body that is strong, that moves well, that can absorb an accident and clear a meal, and treat the pursuit of size for its own sake, particularly when it is drug-assisted, as a different project altogether from the pursuit of a long and healthy life. Beyond that, the best any of us can do is understand the underlying biochemistry, physiology and anatomy and make sensible decisions on top of them, adjusted for the person we actually are. I cannot make those adjustments for you through a screen, and I would be wary of anyone who claimed they could.

Disclosures

I have no commercial interest in, and I am not selling, any supplement, device or product mentioned in this article; the recommendations here cost nothing. I should also declare a relevant bias: I am a former competitive powerlifter and a long-term advocate of resistance training, which naturally disposes me to value it, and I have tried to flag throughout the places where my own lean runs ahead of the evidence. I offer private consultations for people who want to talk their own situation through, but I will put it plainly: for most people, a good, regular relationship with a local clinician who can actually examine you is worth more than a single consultation with me.

References

Identified and verified via PubMed; DOI links included.

1. Cruz-Jentoft et al. Sarcopenia: revised European consensus on definition and diagnosis. Age and Ageing, 2019. https://doi.org/10.1093/ageing/afy169

2. Zhang et al. Sarcopenia as a predictor of all-cause mortality among older nursing home residents: a systematic review and meta-analysis. BMJ Open, 2018. https://doi.org/10.1136/bmjopen-2017-021252

3. Celis-Morales et al. Associations of grip strength with cardiovascular, respiratory, and cancer outcomes and all-cause mortality: prospective cohort study of half a million UK Biobank participants. BMJ, 2018. https://doi.org/10.1136/bmj.k1651

4. López-Bueno et al. Thresholds of handgrip strength for all-cause, cancer, and cardiovascular mortality: a systematic review with dose-response meta-analysis. Ageing Research Reviews, 2022. https://doi.org/10.1016/j.arr.2022.101778

5. Brito et al. (incl. Araújo CGS). Ability to sit and rise from the floor as a predictor of all-cause mortality. European Journal of Preventive Cardiology, 2014. https://doi.org/10.1177/2047487312471759

6. Kim et al. Muscle strength: a better index of low physical performance than muscle mass in older adults. Geriatrics & Gerontology International, 2016. https://doi.org/10.1111/ggi.12514

7. Srikanthan P, Karlamangla AS. Relative muscle mass is inversely associated with insulin resistance and prediabetes: findings from NHANES III. Journal of Clinical Endocrinology & Metabolism, 2011. https://doi.org/10.1210/jc.2011-0435

8. Mastrototaro L, Roden M. Insulin resistance and insulin sensitizing agents. Metabolism, 2021. https://doi.org/10.1016/j.metabol.2021.154892

9. Pedersen BK, Febbraio MA. Muscles, exercise and obesity: skeletal muscle as a secretory organ. Nature Reviews Endocrinology, 2012. https://doi.org/10.1038/nrendo.2012.49

10. Alfaro-Acha et al. Handgrip strength and cognitive decline in older Mexican Americans. Journals of Gerontology Series A, 2006. https://doi.org/10.1093/gerona/61.8.859

11. Hendriks et al. Risk factors for young-onset dementia in the UK Biobank. JAMA Neurology, 2024. https://doi.org/10.1001/jamaneurol.2023.4929

12. Baggish et al. Cardiovascular toxicity of illicit anabolic-androgenic steroid use. Circulation, 2017. https://doi.org/10.1161/CIRCULATIONAHA.116.026945

13. Shailendra et al. Resistance training and mortality risk: a systematic review and meta-analysis. American Journal of Preventive Medicine, 2022. https://doi.org/10.1016/j.amepre.2022.03.020

14. Gibala MJ. Molecular responses to high-intensity interval exercise. Applied Physiology, Nutrition, and Metabolism, 2009. https://doi.org/10.1139/H09-046

15. Paddon-Jones D, Rasmussen BB. Dietary protein recommendations and the prevention of sarcopenia. Current Opinion in Clinical Nutrition and Metabolic Care, 2009. https://doi.org/10.1097/MCO.0b013e32831cef8b

16. Bauer et al. Effects of a vitamin D and leucine-enriched whey protein nutritional supplement on measures of sarcopenia in older adults (the PROVIDE study): a randomized, double-blind, placebo-controlled trial. Journal of the American Medical Directors Association, 2015. https://doi.org/10.1016/j.jamda.2015.05.021

17. Noakes TD, Prins PJ, Volek JS, et al. Low carbohydrate high fat ketogenic diets on the exercise crossover point and glucose homeostasis. Frontiers in Physiology, 2023. https://doi.org/10.3389/fphys.2023.1150265

18. Burke et al. Adaptation to a low carbohydrate high fat diet is rapid but impairs endurance exercise metabolism and performance despite enhanced glycogen availability. The Journal of Physiology, 2021. https://doi.org/10.1113/JP280221

19. Wroble et al. Low-carbohydrate, ketogenic diet impairs anaerobic exercise performance in exercise-trained women and men: a randomized-sequence crossover trial. Journal of Sports Medicine and Physical Fitness, 2019. https://doi.org/10.23736/S0022-4707.18.08318-4

20. Noakes TD, Prins PJ, Buga A, D'Agostino DP, Volek JS, Koutnik AP. Carbohydrate ingestion on exercise metabolism and physical performance. Endocrine Reviews, 2026 (review; fat-adapted athletes show exceptional fat oxidation and equivalent performance, and carbohydrate is not an obligatory fuel). https://doi.org/10.1210/endrev/bnaf038

Diet and Skin Wrinkles: Glycation, the Randle Cycle, and the Seed-Oil Question

Executive summary

• No food, and no way of eating, stops ageing or removes wrinkles. The honest claim is narrower: some dietary patterns may lower specific biochemical processes that nudge skin toward wrinkling, and only ever alongside sleep, daylight, stress, movement and everything else you do.

• Wrinkling is, in large part, the wearing out of the skin's structural proteins, collagen and elastin. Three mechanisms plausibly contribute: glycation (sugars cross-linking those proteins), chronic low-grade inflammation (to which fuel-selection biology may contribute), and, more speculatively, the oxidation of fragile polyunsaturated fats deposited in skin.

• Glycation is the best-supported of the three. Reducing sugars react with collagen and elastin to form advanced glycation end-products (AGEs) that stiffen the tissue and resist normal repair (Ref 1, Ref 2). Much of the receptor-level detail (RAGE signalling) comes from cell and animal work and should be held loosely (Ref 3, Ref 4).

• In observational data, higher blood glucose is associated with looking slightly older, on the order of a fraction of a year per unit of glucose, in part by way of wrinkling (Ref 7, Ref 8). That is an association, not proof of cause.

• Chronically high sugar and insulin, and the forced fuel-selection described by the Randle cycle (Ref 9, Ref 10), plausibly feed the chronic, low-grade inflammation of ageing sometimes called "inflammaging", which is a recognised risk marker whose exact causal weight is still being worked out (Ref 11, Ref 12).

• The seed-oil-and-sun idea is a hypothesis, not an established fact. Dietary linoleic acid is reflected in body fat (Ref 14), and ultraviolet light can oxidise polyunsaturated fats in skin (Ref 13), but no trial shows that eating fewer seed oils protects the face, and at least one skin-cell study found no measurable oxidation after short ultraviolet exposure (Ref 15). I am still collecting data here.

• The practical message is a set of principles, not a magic shopping list: lower dietary carbohydrate as far as is sensible for you (Ref 19), reduce the excess industrial polyunsaturated load (my opinion, not settled fact), and avoid keeping sugar and insulin permanently elevated. None of it substitutes for sleep, daylight, movement and human connection.

• Interest declared up front: I personally follow a low-carbohydrate, largely animal-based pattern of eating, so I hold a prior view that could bias me. I offer private consultations, but I sell none of the foods discussed here.

Introduction

The question this article asks is narrow and, I think, the only honest version of a popular one: does what you eat affect how your skin wrinkles, and if so, how much of that can we actually claim to know? The popular version, the one that sells, is "the foods that erase wrinkles". I have never been able to write that piece, because in the biology of ageing nothing works that way, and I would rather keep your trust than meet the demand.

Let me be clear at the outset about what I am and am not arguing, because this topic is easy to overstate in either direction. I am not claiming that any food, or the absence of any food, prevents wrinkles or reverses ageing. I am making a smaller and, I think, more defensible point: there are identifiable biochemical processes that contribute to the ageing of skin's structural proteins, diet plausibly modulates some of them, and the size of that contribution in any individual is something none of us can honestly quantify. Where the evidence is strong I will say so; where it is a mechanism I find persuasive but cannot prove, I will label it plainly as such.

A word on where I am coming from, in both senses. I trained first in dentistry, then in dermatology, and separately in anti-aging medicine and in metabolic medicine, and I have spent many years looking at the relationship between how people eat and how their skin and metabolism behave. I also owe you a declaration of interest that has nothing to do with money: I personally follow a low-carbohydrate, largely animal-based pattern of eating, and I already lean toward the view that stable, saturated animal fats are the safer default. That is a prior, and priors bias people, so please weigh what follows against it. I lean on biochemistry and mechanism more than on the messy human trials in this field, because the trials are so often confounded by uncontrolled diets and a dozen simultaneous lifestyle changes, but I hold that stance openly and I am genuinely willing to be moved by good contrary evidence.

The one mechanism worth understanding: what wears out the skin's springs

Most writing in this area hands you a list of foods. I would rather give you the single idea underneath the list, because once you have the mechanism you can reason for yourself.

Think of the dermis, the living layer beneath the surface of your skin, as a good mattress. What gives a mattress its bounce is the network of springs inside it. In skin, those springs are structural proteins, principally collagen and elastin, laid down and maintained by cells called fibroblasts. Young skin is springy because those proteins are intact and well organised. Aged, wrinkled skin is, in large part, what you get when the springs stiffen, tangle and lose the ability to repair themselves. So the useful question is not "which food is magic for my face"; it is "what wears the springs out". From a biochemical point of view there are three plausible culprits, and I will take them in descending order of how confident I am about each: sugar gumming the springs together, a low, smouldering fire of inflammation around them, and, most speculatively, fragile fats lodged in the springs and oxidising in the sun.

Glycation: sugar cross-linking the scaffolding

This is the most established of the three, and the least controversial, so it is the right place to start.

Glycation, in plain terms, is what happens when sugar in your blood reacts with your proteins without an enzyme telling it to. According to a review of the area retrieved via PubMed, glucose and fructose link the amino acids in the collagen and elastin that support the dermis and produce advanced glycation end-products, which carry the rather appropriate acronym AGEs; the process is accelerated when blood sugar is elevated and is further stimulated in skin by ultraviolet light (Ref 1). A helpful way to picture an AGE is as a tiny weld. Instead of a mattress of independent, bouncy springs, you get springs spot-welded to one another: stiffer, more brittle, and harder to service.

It is worth being precise about one thing here, because it is a point people often get wrong: glycation is not selective. It is a slow, non-enzymatic chemistry that will cross-link more or less any long-lived protein it can reach, which is exactly why collagen, one of the longest-lived proteins in the body, is such a frequent target. Nothing about the mechanism singles out skin, or singles out wrinkles; skin is simply where we can see the result.

Once formed, these welds are not easily undone. In work using mouse skin and modified collagen in the test tube, AGE-cross-linked collagen resisted the enzymes that would normally break down and recycle damaged collagen, so it accumulated (Ref 2). That is a plausible molecular reason why glycated tissue behaves like something that cannot be properly repaired, though I would flag clearly that this particular demonstration is in animal tissue and in vitro, and human skin may not behave identically.

There is a second arm to the story. Glycated proteins can bind a cell-surface receptor called RAGE (the receptor for AGEs), and that binding is described, in reviews retrieved via PubMed, as switching on oxidative and inflammatory signalling inside cells (Ref 3). The cleanest skin-specific version of this comes from experiments in cultured skin cells and animal skin, where ultraviolet exposure increased AGE-RAGE binding and, downstream, inflammatory signalling (Ref 4). I want to be upfront that this last part is largely cell-and-animal work, with added ultraviolet, rather than a demonstration in a living human face, so hold it loosely. The direction of travel is consistent, but the strength of the evidence is not the same as for the basic cross-linking.

A detail most accounts skip: not all sugars glycate at the same rate. In laboratory studies, fructose appears to be several times more reactive as a glycating agent than glucose, molecule for molecule, both in its early Maillard chemistry (Ref 5) and in its ability to cross-link collagen in the test tube (Ref 6). I would not oversell this. These are in-vitro findings, and, as ever, the honest caveat cuts the other way too: you carry far less fructose than glucose in your blood at any moment, so fructose being nastier per molecule does not straightforwardly mean it matters more in real life, because glucose is simply present in much larger amounts. The practical upshot for me is modest: I keep an eye on total sugar load rather than fixating on one sugar.

Does any of this actually show on a face? In the Leiden Longevity Study, higher non-fasted blood glucose was associated with a higher "perceived age" from facial photographs, by about 0.4 years for each 1 mmol/L increase in glucose among non-diabetic people, after adjustment for the obvious confounders (Ref 7). A later analysis from the same group found that a good part of that association ran specifically through skin wrinkling (Ref 8). I like these studies because they are a real-world echo of the mechanism, but I have to be honest about what they are: cross-sectional and observational, so they show association, not cause, and half a year of "perceived age" is a small effect.

This is also where I will gently point out an inconsistency in the usual "anti-wrinkle foods" lists, which tend to feature high-sugar items such as mango or dark chocolate. Judged as isolated ingredients those foods have real merits. But it is a little odd to hand someone a high-sugar fruit as an anti-wrinkle food in one breath while, in the next, agreeing that sugar drives the glycation that ages skin. I am not telling you mango is poison. I am saying the logic does not hold together, and now you can see why.

The honest bottom line on glycation is this. I cannot tell you what fraction of any given wrinkle was caused by sugar; there are far too many overlapping inputs for anyone to isolate that cleanly, and anyone who gives you a precise percentage is guessing. What is well supported is that glycation is real, that it stiffens and cross-links the skin's structural proteins, and that it is one genuine contributor among several.

Fuel selection, the Randle cycle, and inflammation

The second thing that plausibly wears out the springs is chronic inflammation, and this is where I get to describe a mechanism I find genuinely interesting, the glucose-fatty-acid cycle, usually called the Randle cycle.

The analogy makes it easy. Imagine a single engine that can run on either petrol or diesel, but not both flat out at once, because the two fuels share the same final set of parts. Your cells are a little like that. On the path from either fat or glucose down to ATP, the actual usable fuel currency a cell spends, there is a shared set of steps that cannot process both fuels at maximum at the same time. (When I say a cell "oxidises" a fuel, I mean the controlled, enzyme-run way it strips electrons out of that fuel to make ATP; nothing is literally on fire.) So a cell tends to be mostly running on one fuel or the other and to down-regulate its handling of the other. This competition was first described in 1963, in the paper that gave the cycle its name (Ref 9).

I should immediately add the cite-then-critique that intellectual honesty requires. The Randle cycle is a real and durable piece of physiology, but it is not the whole story of how fuels and insulin sensitivity interact. Later work makes clear that the accumulation of fat-derived signalling molecules inside muscle and liver, along with inflammation and oxidative stress, also drives changes in how cells respond to insulin (Ref 10). So the cycle is one lever, not the entire machine.

Here is why it may matter for your face. On a pattern of eating that keeps sugar and insulin chronically high, you keep pushing this fuel-selection system, and chronically elevated glucose and insulin are themselves a recognised driver of the low-grade inflammatory state described in metabolic disease, with raised interleukin-6, tumour necrosis factor-alpha and interleukin-1-beta (Ref 20). I want to be careful to label that as a mechanistic argument rather than a directly measured causal chain, because the leap from "the Randle cycle is chronically engaged" to "therefore the body is inflamed" is reasoning, not a single experiment I can point you to. What is better supported is the next link: chronic, low-grade, sterile inflammation is one of the genuinely accepted features of ageing, to the point that the field has a name for it, "inflammaging", and reviews describe how nutrient excess and overnutrition feed a closely related "metaflammation" (Ref 11).

And here the register of the evidence deserves precision, because the researchers themselves are careful. They describe inflammaging as a highly significant risk factor that sits alongside ageing, while stating plainly that its precise cause and its potential causal role in producing the damage remain largely unknown (Ref 12). So I would put my own position like this. That chronically high sugar and insulin are pro-inflammatory, I am reasonably confident about. That inflammation is bad for ageing tissue, including skin, is well supported. Exactly how much it ages your particular face, I cannot quantify, and it is probably a different amount in each of us.

I can hear the fair objections, and they deserve honest answers rather than to be waved away. One is that ectopic fat, fat stored in liver and muscle, is itself a cause of insulin resistance, so the arrow may point the other way. Another is the person who cuts carbohydrates and sees their fasting glucose rise, and concludes the whole picture is wrong. On that second point, a distinction matters: cutting carbohydrates can produce an adaptive, physiological insulin resistance, in which muscle preferentially spares glucose for the brain and its cells stop letting in glucose in response to insulin's signal, and this is not the same thing as the pathological insulin resistance of metabolic disease, even though a single fasting-glucose number cannot tell them apart. These objections are reasonable, and a proper treatment of each is longer than this article can carry; I raise them here so you know I am not pretending they do not exist.

One more thread ties this section to the last. When the human body does synthesise fat from surplus carbohydrate, the process (de novo lipogenesis) is stimulated by higher glucose and insulin, and its principal product is palmitate, a saturated fatty acid (Ref 17). I will come back to why that particular fact is quietly interesting.

Fragile fats and the sun: a hypothesis, clearly labelled

The third idea is the most speculative thing I will say, so I want a large, honest label on it before I begin. What follows is a hypothesis I find compelling, not a fact. I could be wrong, and I would be perfectly content to be shown so.

To follow it you need the difference between two kinds of fat. A saturated fat is a sturdy, stable molecule with no chemically weak points. An unsaturated fat, and especially a polyunsaturated one, contains one or more carbon-to-carbon double bonds, and each double bond is a fragile spot, a place that is comparatively easy to attack and oxidise. The kitchen version of this is familiar: saturated fats such as butter or tallow are solid and keep well, while polyunsaturated seed oils are liquid and eventually go rancid, and "rancid" is just a homely word for oxidised. Think of the double bond as a weak link in a chain: the more of them a fat carries, the easier the chain is to break.

The chain of the hypothesis runs as follows. First, the polyunsaturated fats you eat do not simply pass through; they are built into your tissues, and the amount of linoleic acid (the main dietary polyunsaturated fat) stored in body fat is used as a biomarker of long-term dietary intake precisely because intake is reflected in the tissue (Ref 14). Second, in skin, ultraviolet light can oxidise polyunsaturated fats, and a skin-focused review describes how this lipid oxidation generates reactive breakdown products that then form damaging adducts on collagen and elastin and contribute to the changes of photoageing (Ref 13). Third, the oxidised breakdown products of these fats are themselves biologically active and can be pro-inflammatory rather than inert (Ref 16). So the individual links are each plausible.

Now the honest counter-weight, because this is exactly the sort of tidy story that deserves suspicion. The review evidence above is largely mechanistic and drawn from cell and tissue models, not from a human face over years. More pointedly, at least one study exposing human skin cells to short bursts of ultraviolet A found the expected lipid oxidation did not appear, and reported no detrimental consequence over that window (Ref 15). And the biomarker paper that tells us dietary linoleic acid is reflected in body fat is the same kind of paper that, in a large cohort, found no association between that stored linoleic acid and heart attacks (Ref 14), which should caution anyone, myself included, against a simple "polyunsaturated fat is straightforwardly harmful" narrative. Above all, there is no trial showing that eating fewer seed oils gives you younger skin or protects your face from the sun. That experiment does not exist. So from here on this is observation and mechanism, not proof, and I am still collecting data.

What I can offer is an anecdote, clearly flagged as the weak evidence it is. In myself, in some of my patients, and in patients colleagues look after, a pattern keeps recurring: people who shift toward mostly animal-sourced saturated fats and cut the seed oils report that they stop burning in the sun the way they used to. It happened to me. I cannot lean on that, because someone who changes their diet usually changes ten other things at once, their sun habits, their weight, their sleep, so the anecdote is exactly the kind that can mislead, and I hold my own observations loosely. I mention it not as proof but because I cannot honestly pretend I have not seen it.

A necessary word on sunscreen, because this is where the wrong message is easy to take. I am not telling anyone to abandon sunscreen and go and bake. The thing that genuinely damages and ages skin is the burn itself. My own position is simply that, since changing how I eat, I personally burn less than I did and reach for sunscreen less than I once did. There is a much-quoted observation that skin cancer rates did not fall as neatly as expected after sunscreen became widespread, but I will not leave that hanging as a spooky mystery, because the honest reading is that the comparison is heavily confounded: people stay out longer when they feel protected, older formulations blocked the burning wavelengths while letting other ageing ones through, detection improved, and the heaviest sunscreen users are often those at highest risk to begin with. The honest summary is that the evidence here is not settled either way, so I treat my own experience as an experiment of one and nothing more. Please do not copy me on the strength of an article. If you are very fair-skinned, or you have ever had a skin cancer, do not take my personal experiment as your protocol; that is a conversation for you and a clinician who can actually examine you.

The Mediterranean question

All of this leads to a slightly heretical question, which I want to ask as a question rather than bang a drum about. Is the Mediterranean diet as unambiguously good for skin as we are usually told?

There are two honest problems. The first is definitional: what, precisely, is "the Mediterranean diet"? I have an office in Milan and spend a good deal of time there, and every Italian I meet eats differently; one has pasta daily, the next weekly. The label is far fuzzier than the confident headlines imply. The second problem is that some of the celebrated early data has real cracks. A commentary retrieved via PubMed points out that fieldwork behind the famous Cretan data overlapped with periods of Greek Orthodox fasting, when people were genuinely eating far less fat than usual, which would have made the everyday diet look leaner and healthier than it really was (Ref 18). I will cite it and then be fair about it: that commentary documents the fasting overlap itself; it is a strong reason for scepticism, not a full re-audit of the original study, and it is a short letter rather than a large re-analysis.

Set the population studies aside, though, because I find the biochemistry more persuasive anyway, and here is the quietly interesting fact I promised earlier. When the human body synthesises its own fat to store, the principal product is a saturated fatty acid, palmitate (Ref 17). Left to its own devices, the body builds the sturdy, stable molecule. So when I am told that a large glug of polyunsaturated oil is the single healthiest fat I can eat, I find myself asking, genuinely, whether that is as obvious as it sounds, given that saturated fat is what my own physiology chooses to make and store. I am not declaring olive oil bad, and I would not want that sentence quoted as if I had. I am inviting you to sit with the question, and I remain open to evidence in either direction.

What this looks like in practice

Now the practical part, and I am aware of a trap: I opened by gently mocking "five foods" lists, so the last thing I should do is hand you my own. Principles travel; lists do not.

The first principle is to lower dietary carbohydrate as far as is sensible for you. I will state one thing as physiology rather than opinion: there is no such thing as an essential carbohydrate. Your body can manufacture the glucose it needs from other substrates, which is why a published argument in the nutrition literature can reasonably ask whether dietary carbohydrate is required at all, in contrast to the essential amino acids and essential fatty acids that you genuinely must obtain from food (Ref 19). Fewer dietary carbohydrates means less of the sugar available to drive glycation, and steadier fuel that avoids constantly jamming the fuel-selection system described above. How far is "sensible" depends on you, and I cannot set that number through a screen.

The second principle, which I will flag clearly as my opinion rather than settled fact, is to reduce the excess industrial polyunsaturated load, chiefly refined seed oils. My reasoning is the evolutionary and biochemical one above, not a completed trial, and you should treat it as such.

The third principle is to avoid keeping sugar and insulin permanently elevated across the day, which is the same Randle idea applied to how you actually eat, and which lowers the smouldering inflammation that goes with chronic over-fuelling.

What does that look like as food? For me it is animal-sourced and simple: well-salted ruminant meat, fish, the natural fats that come with them, and unrefined salt. That is the core of how I eat. But I would rather you did not take it as gospel because I said it. There is genuine debate here, and you will find intelligent, qualified people who will tell you close to the opposite. Do not believe me, or anyone, merely because we have collected qualifications; go and understand the biochemistry, the physiology and the human history for yourself, and then decide.

I should also say plainly that I am not proposing you eat like this, joylessly, for the rest of your life. If a way of eating makes you miserable, that has its own costs. If the larger pattern is sound, a small, deliberate, occasional deviation you genuinely enjoy is not going to undo it. I follow this fairly strictly, roughly nine parts in ten, and the remaining tenth is modest and non-alcoholic in my case, an occasional Coca-Cola or some dessert. It is not individually "healthy", and I would not pretend otherwise, but in the context of an otherwise sound life it is a reasonable trade for enjoyment taken in sensible amounts. Where you draw that line is yours to decide.

On supplements, my rule is simply food first, with supplements reserved for specific situations rather than used as a default; I will not name products here.

And one human note, which is also where I will mention consultations, briefly. Knowing these principles is the easy part; living them around a real job, a real family and real tastes is the genuinely hard part, and if you find it difficult, that is not a personal failing. You do not need to see me for this. For most people a good, regular relationship with a local clinician is worth more than any single consultation with me, and there are many excellent clinicians who can help. I do offer private consultations for those who specifically want to work through their own situation with me, but I would rather you heard the humble version of that sentence than the salesy one.

Conclusion

Let me bring this back to where it started, honestly. I cannot promise you that eating this way will stop you getting wrinkles, and nobody being truthful can. What I can say is that you may be able to lower the specific biochemistry that pushes skin toward wrinkling: less glycation, less of the smouldering inflammation that accompanies chronically high sugar and insulin, and, if the seed-oil hypothesis holds up, fewer fragile fats sitting in the skin waiting to be oxidised. That is worth doing. It is not a magic trick, and I cannot tell you, for your face in particular, how much difference it will make, because I cannot see you through a screen and because the honest answer differs from person to person.

And keep the proportions right, because your face largely keeps the score of how you live. The best diet in the world will not rescue skin running on broken sleep, no daylight, chronic stress and no real human connection. Get those foundations in first, in whatever combination fits your life. Eat real food that is not spiking your insulin all day, get sensible daylight on your skin and eyes, sleep until you are genuinely rested, move, and stay close to the people you love. No plate and no product gets to skip those. Layer this biochemistry on top of them, and you have stacked the odds sensibly in your favour. That is about as honest as I can make it.

Disclosures

I sell none of the foods or supplements discussed in this article, and I have no financial interest in any of them. I do, however, hold a relevant non-financial bias that you should weigh: I personally follow a low-carbohydrate, largely animal-based pattern of eating, and I already lean toward stable animal fats as the safer default, so I am not a neutral observer of this evidence. I lean on mechanism and biochemistry more than on the confounded human trials in this field, and I have tried to flag every point where I am reasoning rather than citing a settled result. I offer private consultations for those who want to discuss their own situation, though for most people a good, regular relationship with a local clinician is worth more than any single consultation with me.

References

Identified and verified via PubMed; DOI links included.

1. Danby FW. Nutrition and aging skin: sugar and glycation. Clin Dermatol. 2010;28(4):409-411. https://doi.org/10.1016/j.clindermatol.2010.03.018 (Narrative review of glycation in skin ageing.)

2. Nowotny K, Grune T. Degradation of oxidized and glycoxidized collagen: role of collagen cross-linking. Arch Biochem Biophys. 2013;542:56-64. https://doi.org/10.1016/j.abb.2013.12.007 (Mouse skin and in-vitro collagen; AGE cross-linking resists degradation. Animal and in-vitro finding.)

3. Yamagishi S, Fukami K, Matsui T. Evaluation of tissue accumulation levels of advanced glycation end products by skin autofluorescence: a novel marker of vascular complications in high-risk patients for cardiovascular disease. Int J Cardiol. 2015;185:263-268. https://doi.org/10.1016/j.ijcard.2015.03.167 (Review; AGE-RAGE interaction and oxidative/inflammatory signalling.)

4. Oh S, Lee SY, Jang JW, Son KH, Byun K. Fermented fish collagen diminished photoaging-related collagen decrease by attenuating AGE-RAGE binding activity. Curr Issues Mol Biol. 2024;46(12):14351-14365. https://doi.org/10.3390/cimb46120860 (UV-irradiated keratinocytes and animal skin; AGE-RAGE binding drives inflammatory signalling. Cell and animal work.)

5. Dills WL. Protein fructosylation: fructose and the Maillard reaction. Am J Clin Nutr. 1993;58(5 Suppl):779S-787S. https://doi.org/10.1093/ajcn/58.5.779S (Review; early Maillard chemistry proceeds faster with fructose than glucose. In-vitro chemistry.)

6. Fujimori E. Cross-linking and fluorescence changes of collagen by glycation and oxidation. Biochim Biophys Acta. 1989;998(2):105-110. https://doi.org/10.1016/0167-4838(89)90260-4 (In-vitro rat-tail collagen; fructosylation more effective than glucosylation at cross-linking. In-vitro finding.)

7. Noordam R, Gunn DA, Tomlin CC, et al. High serum glucose levels are associated with a higher perceived age. Age (Dordr). 2013;35(1):189-195. https://doi.org/10.1007/s11357-011-9339-9 (Observational, cross-sectional; ~0.40 years higher perceived age per 1 mmol/L glucose in non-diabetic adults. Association, not causation.)

8. van Drielen K, Gunn DA, Noordam R, et al. Disentangling the effects of circulating IGF-1, glucose, and cortisol on features of perceived age. Age (Dordr). 2015;37(3):9771. https://doi.org/10.1007/s11357-015-9771-3 (Observational; higher glucose associated with higher perceived age, in part mediated by skin wrinkling.)

9. Randle PJ, Garland PB, Hales CN, Newsholme EA. The glucose fatty-acid cycle. Its role in insulin sensitivity and the metabolic disturbances of diabetes mellitus. Lancet. 1963;1(7285):785-789. https://doi.org/10.1016/s0140-6736(63)91500-9 (Original description of the Randle cycle.)

10. Delarue J, Magnan C. Free fatty acids and insulin resistance. Curr Opin Clin Nutr Metab Care. 2007;10(2):142-148. https://doi.org/10.1097/MCO.0b013e328042ba90 (Review; the Randle cycle is one mechanism among several for insulin resistance.)

11. Franceschi C, Garagnani P, Parini P, Giuliani C, Santoro A. Inflammaging: a new immune-metabolic viewpoint for age-related diseases. Nat Rev Endocrinol. 2018;14(10):576-590. https://doi.org/10.1038/s41574-018-0059-4 (Review; inflammaging and nutrient-driven metaflammation.)

12. Franceschi C, Campisi J. Chronic inflammation (inflammaging) and its potential contribution to age-associated diseases. J Gerontol A Biol Sci Med Sci. 2014;69(Suppl 1):S4-S9. https://doi.org/10.1093/gerona/glu057 (Review; inflammaging is a significant risk factor, but its causal role remains largely unknown.)

13. Negre-Salvayre A, Salvayre R. Post-translational modifications evoked by reactive carbonyl species in ultraviolet-A-exposed skin: implication in fibroblast senescence and skin photoaging. Antioxidants (Basel). 2022;11(11):2281. https://doi.org/10.3390/antiox11112281 (Review; UV-A drives polyunsaturated-fat oxidation and reactive carbonyl adducts on collagen and elastin. Largely mechanistic and in-vitro/ex-vivo.)

14. Nielsen MH, Frydenberg M, Bork CS, et al. Linoleic acid in adipose tissue and the risk of myocardial infarction: a case-cohort study. Eur J Nutr. 2021;60(7):3639-3646. https://doi.org/10.1007/s00394-021-02526-y (Observational cohort; adipose linoleic acid is a biomarker of long-term dietary intake, but was not associated with myocardial infarction in this cohort.)

15. Leung KS, Chan HF, Leung HH, et al. Short-time UVA exposure to human keratinocytes instigated polyunsaturated fatty acid without inducing lipid peroxidation. Free Radic Res. 2017;51(3):269-280. https://doi.org/10.1080/10715762.2017.1300885 (In-vitro keratinocytes; short UVA exposure did not induce measurable lipid peroxidation. Counter-evidence.)

16. Mazzocchi A, De Cosmi V, Risé P, et al. Bioactive compounds in edible oils and their role in oxidative stress and inflammation. Front Physiol. 2021;12:659551. https://doi.org/10.3389/fphys.2021.659551 (Review; oxidised metabolites of dietary oils can be biologically active and pro-inflammatory.)

17. Smith GI, Shankaran M, Yoshino M, et al. Insulin resistance drives hepatic de novo lipogenesis in nonalcoholic fatty liver disease. J Clin Invest. 2020;130(3):1453-1460. https://doi.org/10.1172/JCI134165 (Human clinical study; de novo lipogenesis yields palmitate, a saturated fat, and is stimulated by glucose and insulin.)

18. Sarri K, Kafatos A. The Seven Countries Study in Crete: olive oil, Mediterranean diet or fasting? Public Health Nutr. 2005;8(6):666. https://doi.org/10.1079/phn2005765 (Commentary/letter; Cretan fieldwork overlapped with Orthodox fasting, a confounder of the original diet picture.)

19. Westman EC. Is dietary carbohydrate essential for human nutrition? Am J Clin Nutr. 2002;75(5):951-953. https://doi.org/10.1093/ajcn/75.5.951 (Commentary/letter arguing there is no dietary requirement for carbohydrate, unlike essential amino acids and fatty acids.)

20. Silveira Rossi JL, Barbalho SM, Reverete de Araujo R, et al. Metabolic syndrome and cardiovascular diseases: going beyond traditional risk factors. Diabetes Metab Res Rev. 2021;38(3):e3502. https://doi.org/10.1002/dmrr.3502 (Review; chronic hyperglycaemia and hyperinsulinaemia drive metaflammation with raised IL-6, TNF-alpha and IL-1-beta.)

Magnesium: Which Form, Which Test, and Who Actually Needs It

Executive summary

• Magnesium is not a fringe micronutrient. It acts as a cofactor for more than 300, and by some more recent tallies over 600, of the body's enzymatic reactions, including the generation of ATP (the molecule cells spend to do their work), the control of nerve and muscle excitability, and the synthesis of DNA and protein (Ref 1, Ref 2, Ref 3).

• Inadequate intake is common. Depending on the population and the cut-off used, somewhere between roughly a third and a half of people surveyed fall short of requirements or show subclinical deficiency. No single figure should be treated as exact (Ref 5, Ref 6, Ref 7).

• The standard serum magnesium test is a weak guide to whole-body stores, because only about one per cent of the body's magnesium sits in the blood. A "normal" serum result can sit alongside genuine tissue depletion. Red blood cell (RBC) magnesium is a better proxy, though it is not flawless either (Ref 6, Ref 8).

• The figure on a supplement label is usually the weight of the whole compound, not the elemental magnesium that does the work. Reading the elemental figure is the single most useful habit a buyer can adopt.

• Whether the different salts (citrate, glycinate, oxide, taurate, malate, L-threonate) genuinely differ in usefulness is less settled than social media suggests. My own view is that form probably matters, but I hold it loosely and would revise it if stronger head-to-head data appeared.

• For specific problems the evidence ranges from reasonable (blood pressure, migraine, constipation) through preliminary and often industry-funded (L-threonate for cognition) to frankly weak (malate for fibromyalgia, oral magnesium for pregnancy leg cramps). I have tried to grade each honestly rather than flatten them into one confident claim.

• The safety point I hold most firmly: long-term acid-suppressing drugs (proton pump inhibitors) can deplete magnesium, occasionally to dangerous levels. If you take one, this is worth testing, not worth stopping on your own.

• I have no commercial interest in any magnesium product, no brand affiliation, discount code, or affiliate link. What follows is a reading of the biology and the trials, and all of the references were retrieved and checked through PubMed.

Introduction

Magnesium is having a moment. It is on every shelf, in every wellness feed, and increasingly on the list of things patients arrive already taking. When something reaches that level of popularity, my instinct is not to join the enthusiasm or to dismiss it, but to slow down and ask a narrower question: what does the biochemistry actually require, what have the trials actually measured, and where does the confident marketing outrun the evidence?

I should declare my position at the outset. I do not sell magnesium, I hold no affiliation with any brand, and I have no financial interest in whether you supplement or not. My interest is clinical and, frankly, educational. I trained first in dentistry, then in dermatology, and separately in anti-aging medicine and in metabolic medicine, and across that work the same pattern keeps appearing: a mineral that is genuinely central to physiology, routinely under-measured, and then oversold in a way that obscures the few things about it that are actually useful to know.

This piece tries to separate those threads. It covers what magnesium does, how deficiency presents and why it is easy to miss, why the standard blood test can mislead, how to read a label, whether the fashionable distinctions between forms hold up, who genuinely stands to benefit, and how to think about testing and dosing. Throughout, I have tried to state what a study found and then, in the same breath, what it does not establish. That is the only honest way to write about a supplement.

What magnesium actually does

It helps to hold three roles in mind, because the rest of the article hangs off them.

The first is making ATP. Adenosine triphosphate is the small molecule a cell spends whenever it does anything, and magnesium is required to form and use it; biologically active ATP exists largely as a magnesium-ATP complex (Ref 1, Ref 2). A reasonable analogy is a rechargeable battery: ATP is the charged battery the cell drains to power its machinery, and magnesium is part of what lets the battery be assembled and discharged at all. When it is scarce, it is not surprising that unexplained fatigue is one of the symptoms people report.

The second is controlling excitability at the cell membrane. Magnesium regulates the movement of calcium and potassium across membranes and behaves, in effect, as a natural brake on over-excitable cells (Ref 2, Ref 4). Picture a doorman on the calcium channel: when magnesium is present, the door does not fly open at every knock. When it is low, nerve and muscle cells fire too readily, which is the mechanistic thread running through cramps, twitches, restless legs, a racing or skipping heartbeat, and that hard-to-place sense of being wired.

The third is building things. Magnesium is needed for the synthesis of DNA and protein and for cell division, and it helps stabilise the membranes of the mitochondria where ATP is made (Ref 1, Ref 3). From a longevity standpoint that reframes it: less a garnish on the plate, more part of the structural scaffolding a cell needs to copy and repair itself.

None of this is controversial. It is textbook physiology, and it is the reason a real deficiency can show up in so many different systems at once.

How common is shortfall, and why it is easy to miss

Estimates vary because the definition varies. Surveys that measure dietary intake against requirements tend to find that a large minority, sometimes around half of the sample, fall short (Ref 6). Studies that measure blood or cellular magnesium in apparently healthy people find a meaningful proportion below the reference limit (Ref 7). A fair summary is that inadequacy is common enough to take seriously and variable enough that you should distrust any single headline percentage, including the ones I have just cited. What is not in doubt is that this is a genuine, widespread shortfall rather than a marketing invention (Ref 5).

The reasons are unglamorous: modern diets, food processing, and the fact that several common medications quietly increase losses. I will come back to the medications, because one of them matters more than the others.

The blood test problem: serum versus red-cell magnesium

This is the point I most want a reader to leave with, because it changes how you interpret a result you may already have been given.

The magnesium test a doctor orders by default is serum magnesium, the amount circulating in the blood plasma. The difficulty is that only about one per cent of the body's magnesium is in the blood at all; the remaining ninety-nine per cent is held in bone, muscle, and the inside of cells (Ref 1, Ref 6). The body defends the blood level tightly, pulling magnesium out of storage to keep serum in range. So serum can read comfortably normal while the tissue that actually uses magnesium is running low. An analogy: checking the water pressure at a single tap tells you the tap works, not how much water is left in the reservoir behind the house.

A better proxy is red blood cell (RBC) magnesium, which reflects what is inside cells and tracks tissue status more faithfully. I use it in preference to serum when the question is genuine status rather than an acute emergency.

I want to be careful here, because it would be easy to oversell RBC testing the way serum is oversold. It is not a perfect gold standard. In acute illness, for example, red-cell magnesium can be pushed around by changes in the red cells themselves and should be read with caution (Ref 8). No single blood test captures whole-body magnesium cleanly; the honest position is that RBC is generally the more informative of the two commonly available options, not that it is infallible.

A word about the clinicians ordering these tests. If your doctor told you your magnesium was normal on a serum test, that is not a failing on their part. Serum magnesium is what standard laboratory pathways offer, it is what guidelines and local formularies point to, and a busy GP is working within those constraints. The useful move is not to distrust the doctor but to ask, specifically, which test was run, and whether an RBC magnesium is available to you privately or on request.

Reading the label: elemental versus compound weight

Before comparing forms, there is a trap on almost every label. When a product says "400 mg Magnesium Citrate," that 400 mg is the weight of the entire compound, the magnesium plus whatever it is bound to. The elemental magnesium, the part that does the physiological work, is only a fraction of it.

The fractions follow directly from molecular weight, so they are chemistry rather than opinion. Magnesium oxide is roughly sixty per cent elemental magnesium by weight. Magnesium citrate is around sixteen per cent. Magnesium bisglycinate sits near fourteen per cent. Magnesium L-threonate is lower still, roughly eight per cent, which is why threonate products ask you to take one and a half to two grams of compound to deliver little more than a hundred milligrams of elemental magnesium.

The practical rule is simple and, unlike much of what follows, not in scientific dispute: find the elemental magnesium figure on the label. If a product hides it and quotes only the compound weight, treat that opacity as a reason to look elsewhere. A well-run manufacturer states the elemental dose plainly.

Do the forms really differ? Bioavailability, honestly

Here the internet is more confident than the data. A popular claim is that magnesium oxide is almost useless because only about four per cent of it is absorbed, while organic salts such as citrate and glycinate are far superior. There is a real signal underneath this, but the picture is messier than the slogan.

Some controlled comparisons do find organic or specially formulated salts raising blood magnesium more smoothly than oxide (Ref 10). Others find the opposite: in one short randomised comparison, magnesium oxide actually produced a larger rise in blood magnesium than citrate or carbonate (Ref 9). Absorption also depends on dose, on whether it is taken with food, and on how depleted the person is to begin with. So the specific "four per cent" figure is best treated as a widely repeated number rather than a settled constant, and the ranking of salts by usefulness is genuinely contested.

My own reading is that form probably does matter, both for absorption and for the co-molecule attached (glycine and taurine, for instance, have effects of their own). But I hold that as a working view, not a certainty. If good head-to-head trials showed the differences to be smaller than I think, I would be glad to update. Keeping a position revisable is part of keeping it honest, and this is an area where I am still collecting data rather than closing the question.

Matching the form to the problem

With that caution in place, here is how I think about the common forms. The pattern below leans partly on mechanism, partly on trials, and partly on consistent patient report, and I have tried to flag which is which.

Sleep and a low, unexplained tension: magnesium glycinate. Magnesium is bound to glycine, an amino acid with its own calming profile, and the glycinate form tends to be gentle on the gut. For general repletion and sleep it is my usual default. The trial evidence for magnesium and sleep is real but modest: a systematic review of older adults found magnesium shortened the time taken to fall asleep by around seventeen minutes versus placebo, while noting that the underlying studies were small and of low quality (Ref 22). In other words, a reasonable, low-risk thing to try, not a sedative.

There is one caveat I rarely see mentioned, and I offer it as mechanism rather than established fact. Vitamin B6 is a cofactor for the enzyme that converts glyoxylate to glycine; when B6 is lacking, more glyoxylate is diverted into oxalate instead, so in someone deficient in B6 a high glycine load could, in theory, nudge oxalate production upward (Ref 30). I would keep this in proportion, though: the same work found no association between vitamin B6 intake and kidney-stone risk in the general population, so this is a mechanistic flag rather than a demonstrated hazard. If you have a history of kidney stones or a known B-vitamin deficiency it is worth being aware of, which is part of why several glycinate products include B6.

Palpitations, ectopic beats, and blood pressure: magnesium taurate. The rationale is appealing: taurine has membrane-stabilising effects on heart muscle cells, and magnesium calms the same cells through calcium-channel control, so pairing them looks synergistic. I want to be precise about the evidence, because it is easy to overstate. Randomised trials show that magnesium supplementation in general lowers blood pressure by a modest amount, on the order of two millimetres of mercury for systolic pressure at a median dose of around 370 mg per day (Ref 12). Separately, randomised trials show that taurine on its own lowers blood pressure in people with high-normal readings (Ref 13). What we do not have is a large randomised trial of the specific magnesium taurate salt for blood pressure; the direct evidence for the combined membrane effect is mechanistic and, in the cleanest experiments, done in animal heart cells rather than in people (Ref 14). So it is reasonable to try, and colleagues report patients helped by it alongside conventional treatment, but I would frame it as plausible symptom management rather than a cure. Palpitations and hypertension have many causes, so this is a conversation to have with the doctor supervising your care, not a reason to change cardiac medication yourself.

Constipation: magnesium citrate, or oxide. This one has the most straightforward evidence. Magnesium salts draw water into the bowel by osmosis, acting as a gentle osmotic laxative. A randomised trial found magnesium oxide as effective as senna, and clearly better than placebo, for chronic constipation (Ref 15). If cost matters and the goal is gut motility, this is an effective, inexpensive use of magnesium. I would not reach for citrate first for sleep or cardiac purposes, but for the bowel it does the job.

Fatigue, fibromyalgia, and chronic pain: magnesium malate. Malate is bound to malic acid, a molecule that sits inside the Krebs cycle, the pathway cells use to extract usable fuel as ATP, which is the theoretical reason it is favoured for fatigue states. Honesty requires flagging that the human evidence is thin. The best-known trial of a malic-acid-plus-magnesium product in fibromyalgia found no clear benefit at the blinded, controlled dose; improvement appeared only later in an unblinded, higher-dose phase, which is a much weaker kind of evidence (Ref 16). A Cochrane review of combination treatments could not find enough good data to support it (Ref 17). I mention malate because patients do report benefit and the mechanism is coherent, but I would present it as an adjunct worth a careful trial, not a proven treatment. One practical note that does hold up in experience: take it in the morning, as some people find it mildly stimulating late in the day.

Cognition and brain ageing: magnesium L-threonate. Threonate was designed to raise magnesium levels in the brain, and it does appear to cross into the central nervous system more readily than other forms, rather like a molecule carrying the right passport for the border. The human trials are genuinely interesting and genuinely preliminary. Small randomised studies report improvements in memory and cognition, including in older adults, and a separate trial reports better sleep and daytime function (Ref 18, Ref 19, Ref 20). Two cautions belong next to those results. First, several of these studies are small and were conducted or funded by parties with a commercial interest in the ingredient, which does not invalidate them but does lower how much weight I place on them. Second, much of the mechanistic and disease-model work is in mice, not people, and animal findings do not always translate (Ref 21). My position is that threonate may modestly help rather than reverse age-related cognitive change, that it is best seen as an add-on for cognitive goals rather than a first choice, and that its low elemental content makes it an expensive way to correct a simple deficiency. Some users report vivid dreams, which is harmless but worth knowing.

Getting value from vitamin D. This interaction is underappreciated and well grounded. The enzymes that activate vitamin D require magnesium as a cofactor, so if magnesium is low you may not get the full benefit of the vitamin D you are taking (Ref 11). If you supplement vitamin D, I would treat adequate magnesium as part of the same protocol in most circumstances, rather than an optional extra.

The form to be wary of: magnesium oxide in cheap multivitamins. Oxide is inexpensive and high in elemental magnesium by weight, and it does have a legitimate use as a laxative. My reservation is narrower: it is the default filler in many low-cost combination products, and, absorption debate notwithstanding, if your magnesium supplement never seems to do anything it is worth checking whether oxide is the only form in it.

Who genuinely stands to benefit

Most people do not need a supplement if their diet is adequate. Some groups are different, and here specificity matters more than enthusiasm. In every case below, I am describing patterns, not diagnosing you, because I cannot assess you through a screen and your circumstances may differ.

Older adults. Absorption tends to fall with age and gut transit changes, and this group is also more likely to be taking the acid-suppressing drugs discussed below. I lean towards gentler forms such as glycinate or malate, starting low, around 100 mg elemental, and increasing slowly if needed and tolerated.

Perimenopausal and menopausal women. This is an under-studied group that, in my experience and that of colleagues, often responds well, with magnesium potentially easing several overlapping complaints (disturbed sleep, low mood or tension, and the bone-health picture) at once. Glycinate suits the sleep and mood component; a form like malate suits fatigue. I would add that diet and lifestyle do the heavy lifting here and that magnesium is one tool among several, not a hormonal substitute.

People taking proton pump inhibitors (omeprazole, lansoprazole and similar). This is the safety point I feel most strongly about, and it is rarely mentioned to patients. Long-term PPI use lowers magnesium, because reducing stomach acid impairs magnesium absorption in the gut (Ref 24). At the extreme, there are documented cases of dangerously low magnesium causing serious heart-rhythm disturbances in long-term users, which is why regulators have advised checking magnesium in people on these drugs (Ref 25). I would put two honest caveats beside that. First, across the millions of people taking PPIs this severe outcome is uncommon, and the observational data show association more than proven cause (Ref 26). Second, and importantly, none of this is a reason to stop a PPI on your own. If you take one, the sensible step is to have your magnesium checked, ideally RBC rather than serum, and to raise supplementation with the prescriber who started it. The message is monitor, not self-deprescribe.

People with kidney disease. Here the caution runs the other way. The kidneys regulate how much magnesium is excreted, so when kidney function is impaired, magnesium can accumulate to harmful levels. If you have kidney disease, magnesium supplementation needs medical supervision and is not something to start from a video or an article.

Athletes, sauna users, and heavy sweaters. Intense training and sweating do increase magnesium losses, and correcting a genuine shortfall is reasonable. I would temper expectations about performance, though: a review of supplementation trials in athletes found that, once a real deficiency is excluded, extra magnesium does not reliably improve strength or endurance (Ref 27). So this is about replacing what is lost and supporting recovery, not about ergogenic magic.

Pregnancy. I want to correct something I have said too loosely in the past, and that is often repeated online, namely that magnesium is one of the best-supported supplements in pregnancy. The picture is more mixed than that. The strong evidence in pregnancy is for intravenous magnesium sulfate to prevent seizures in pre-eclampsia, which is a hospital treatment, not an over-the-counter supplement (Ref 1). For the everyday claim, oral magnesium for pregnancy leg cramps, the trials are inconsistent and of low certainty: one randomised trial found no benefit over placebo, and a Cochrane review concluded the evidence does not clearly establish that it works (Ref 28, Ref 29). Magnesium may still be reasonable in pregnancy, but any supplementation should be discussed with your midwife or obstetrician, and I would no longer describe the oral evidence as strong.

Testing, and a reasonable therapeutic trial

Putting the testing question together: the default serum magnesium your doctor orders reflects blood plasma, with a typical reference range around 0.7 to 1.0 millimoles per litre, and because the body defends that level it can look normal despite real depletion (Ref 6). RBC magnesium is the more informative option and can be requested privately or, sometimes, on the standard pathway. You will see "optimal" RBC ranges quoted online that are tighter than the laboratory's normal range; I would treat those optimal figures as opinion rather than established thresholds, because a reference range tells you where the population sits, not necessarily where you feel best.

There is also a pragmatic route that does not require paying for tests. If you have several of the symptoms described earlier, a time-limited trial of a well-absorbed form, glycinate being my usual choice, is entirely reasonable and is often cheaper than testing. In that sense the trial itself yields information: if adequate magnesium in a sensible form relieves the symptoms, that tells you something useful. It is worth saying plainly that this is a way of gathering evidence about your own response, not a diagnosis, and not a cure for whatever underlying issue produced the symptoms.

Dosing, and what to expect

I will give rough figures, because the vague "200 to 400 mg" advice is not very actionable, while stressing that these are starting points to discuss with your own clinician, not prescriptions for an individual I have not met. They are drawn from the pattern across many trials rather than any single one.

For general repletion and sleep, magnesium glycinate starting around 200 mg elemental at night, titrated towards 300 to 400 mg over a few weeks if tolerated. For blood-pressure support, magnesium taurate in the region of 125 to 150 mg elemental twice daily with food, in the range most trials of magnesium have used. For fatigue states, magnesium malate 200 to 300 mg elemental in the morning rather than at night. For cognitive goals, magnesium L-threonate following the product's dosing, accepting that its low elemental content is the point of the format.

Two expectations matter more than the exact numbers. First, repletion is slow. Restoring magnesium inside cells and tissues takes weeks, not days, so you should judge the effect over roughly six to twelve weeks rather than after the first fortnight. You will sometimes see a very precise half-life figure quoted for this; I have not been able to trace it to a source I trust, so I would rather give the honest guidance, allow a couple of months, than repeat a number I cannot stand behind. Second, tolerability is dose-dependent: loose stools are the osmotic effect at high doses, particularly with citrate and oxide, and the fix is to reduce the dose and build up more slowly, often splitting it between morning and evening.

Food first

I should be clear about where I actually land, because it is not on a shelf of bottles. My preference, for myself, is to address the root cause, which is diet, before reaching for a supplement. In keeping with the way I eat and the way I advise, I favour getting minerals from animal foods and from a good unrefined salt rather than from the plant sources conventional lists reach for first. In practice, for me, that means using a quality unrefined salt, such as a Celtic sea salt, generously on real food, which supplies a spread of minerals together, tastes good, costs little, and treats the underlying pattern rather than a single number. Whole, salted animal foods do a great deal of this work quietly.

Supplements are tools for specific situations, several of which I have described above. They are not, in my view, a substitute for a diet that would not have produced the shortfall in the first place. That is also why there are no brand names, discount codes, or affiliate links anywhere in this piece.

Conclusion

Magnesium rewards precision and punishes slogans. The parts worth holding onto are unglamorous: it is a genuinely central mineral, shortfall is common, the routine blood test can miss it, and the number on the label is rarely the number that matters. The fashionable distinctions between forms probably carry some truth, but less cleanly than they are sold, and the evidence for specific uses runs from reasonable to weak in a way that deserves to be named rather than blurred. If you have been told your magnesium is normal but still have the symptoms, ask which test was used. If you take a long-term acid-suppressing drug, have your magnesium checked. And if you decide to supplement, buy on the elemental figure, choose the form to fit the problem, give it a couple of months, and remember that the aim is to correct a diet, not to replace one.

Disclosures

I have no commercial interest in any magnesium product or supplement brand, and no discount codes or affiliate links. I offer private consultations for people who want to think through their own situation, but I would put that in proportion: for most people a good, regular relationship with a local clinician is worth more than a single consultation with me, and there are many excellent clinicians who can help with this. Nothing here is individual medical advice, and decisions about medication, including any drug you already take, should be made with your own prescriber.

References

Identified and verified via PubMed; DOI links included.

1. de Baaij JHF, Hoenderop JGJ, Bindels RJM. Magnesium in man: implications for health and disease. Physiological Reviews, 2015. https://doi.org/10.1152/physrev.00012.2014

2. Fatima G, Dzupina A, Alhmadi HB, et al. Magnesium Matters: A Comprehensive Review of Its Vital Role in Health and Diseases. Cureus, 2024. https://doi.org/10.7759/cureus.71392

3. Volpe SL. Magnesium in disease prevention and overall health. Advances in Nutrition, 2013. https://doi.org/10.3945/an.112.003483

4. Bara M, Guiet-Bara A, Durlach J. Regulation of sodium and potassium pathways by magnesium in cell membranes. Magnesium Research, 1993. https://pubmed.ncbi.nlm.nih.gov/8274363/

5. Matek Sarić M, Sorić T, Juko Kasap Ž, et al. Magnesium: Health Effects, Deficiency Burden, and Future Public Health Directions. Nutrients, 2025. https://doi.org/10.3390/nu17223626

6. Costello RB, Elin RJ, Rosanoff A, et al. Perspective: The Case for an Evidence-Based Reference Interval for Serum Magnesium: The Time Has Come. Advances in Nutrition, 2016. https://doi.org/10.3945/an.116.012765

7. Hermes Sales C, Azevedo Nascimento D, Queiroz Medeiros AC, et al. There is chronic latent magnesium deficiency in apparently healthy university students. Nutrición Hospitalaria, 2014. https://doi.org/10.3305/nh.2014.30.1.7510

8. Martin BJ, Lyon TD, Fell GS, McKay P. Erythrocyte magnesium in elderly patients: not a reliable guide to magnesium status. Journal of Trace Elements in Medicine and Biology, 1997. https://doi.org/10.1016/S0946-672X(97)80009-0

9. Ivanovic ND, Radosavljevic B, Zekovic M, et al. Effects of short-term magnesium supplementation on ionized, total magnesium and other relevant electrolytes levels. Biometals, 2022. https://doi.org/10.1007/s10534-022-00363-y

10. Pajuelo D, Meissner JM, Negra T, Connolly A, Mullor JL. Comparative Clinical Study on Magnesium Absorption and Side Effects After Oral Intake of Microencapsulated Magnesium Versus Other Magnesium Sources. Nutrients, 2024. https://doi.org/10.3390/nu16244367

11. Uwitonze AM, Razzaque MS. Role of Magnesium in Vitamin D Activation and Function. Journal of the American Osteopathic Association, 2018. https://doi.org/10.7556/jaoa.2018.037

12. Zhang X, Li Y, Del Gobbo LC, et al. Effects of Magnesium Supplementation on Blood Pressure: A Meta-Analysis of Randomized Double-Blind Placebo-Controlled Trials. Hypertension, 2016. https://doi.org/10.1161/HYPERTENSIONAHA.116.07664

13. Sun Q, Wang B, Li Y, et al. Taurine Supplementation Lowers Blood Pressure and Improves Vascular Function in Prehypertension: Randomized, Double-Blind, Placebo-Controlled Study. Hypertension, 2016. https://doi.org/10.1161/HYPERTENSIONAHA.115.06624

14. Zhao L, Lou J, Wu H, Yin Y, Kang Y. Effects of taurine-magnesium coordination compound on ionic channels in rat ventricular myocytes of arrhythmia induced by ouabain. Biological Trace Element Research, 2012 (animal / in-vitro study). https://doi.org/10.1007/s12011-011-9317-1

15. Morishita D, Tomita T, Mori S, et al. Senna Versus Magnesium Oxide for the Treatment of Chronic Constipation: A Randomized, Placebo-Controlled Trial. American Journal of Gastroenterology, 2021. https://doi.org/10.14309/ajg.0000000000000942

16. Russell IJ, Michalek JE, Flechas JD, Abraham GE. Treatment of fibromyalgia syndrome with Super Malic: a randomized, double blind, placebo controlled, crossover pilot study. Journal of Rheumatology, 1995 (benefit only in unblinded high-dose phase). https://pubmed.ncbi.nlm.nih.gov/8587088/

17. Thorpe J, Shum B, Moore RA, Wiffen PJ, Gilron I. Combination pharmacotherapy for the treatment of fibromyalgia in adults. Cochrane Database of Systematic Reviews, 2018. https://doi.org/10.1002/14651858.CD010585.pub2

18. Zhang C, Hu Q, Li S, et al. A Magtein, Magnesium L-Threonate, -Based Formula Improves Brain Cognitive Functions in Healthy Chinese Adults. Nutrients, 2022 (combination formula; industry-linked). https://doi.org/10.3390/nu14245235

19. Liu G, Weinger JG, Lu ZL, Xue F, Sadeghpour S. Efficacy and Safety of MMFS-01, a Synapse Density Enhancer, for Treating Cognitive Impairment in Older Adults: A Randomized, Double-Blind, Placebo-Controlled Trial. Journal of Alzheimer's Disease, 2016 (industry-sponsored). https://doi.org/10.3233/JAD-150538

20. Hausenblas HA, Lynch T, Hooper S, et al. Magnesium-L-threonate improves sleep quality and daytime functioning in adults with self-reported sleep problems: A randomized controlled trial. Sleep Medicine: X, 2024 (industry-affiliated). https://doi.org/10.1016/j.sleepx.2024.100121

21. Liao W, Wei J, Liu C, et al. Magnesium-L-threonate treats Alzheimer's disease by modulating the microbiota-gut-brain axis. Neural Regeneration Research, 2023 (animal model, APP/PS1 mice). https://doi.org/10.4103/1673-5374.391310

22. Mah J, Pitre T. Oral magnesium supplementation for insomnia in older adults: a Systematic Review and Meta-Analysis. BMC Complementary Medicine and Therapies, 2021 (low to very-low quality evidence). https://doi.org/10.1186/s12906-021-03297-z

23. von Luckner A, Riederer F. Magnesium in Migraine Prophylaxis: Is There an Evidence-Based Rationale? A Systematic Review. Headache, 2017 (Grade C, possibly effective). https://doi.org/10.1111/head.13217

24. Gommers LMM, Hoenderop JGJ, de Baaij JHF. Mechanisms of proton pump inhibitor-induced hypomagnesemia. Acta Physiologica, 2022. https://doi.org/10.1111/apha.13846

25. Chrysant SG. Proton pump inhibitor-induced hypomagnesemia complicated with serious cardiac arrhythmias. Expert Review of Cardiovascular Therapy, 2019. https://doi.org/10.1080/14779072.2019.1615446

26. William JH, Danziger J. Proton-pump inhibitor-induced hypomagnesemia: Current research and proposed mechanisms. World Journal of Nephrology, 2016 (association, causation not proven). https://doi.org/10.5527/wjn.v5.i2.152

27. Newhouse IJ, Finstad EW. The effects of magnesium supplementation on exercise performance. Clinical Journal of Sport Medicine, 2000. https://doi.org/10.1097/00042752-200007000-00008

28. Araújo CAL, Lorena SB, Cavalcanti GCS, Leão GLS, Tenório GP, Alves JGB. Oral magnesium supplementation for leg cramps in pregnancy: An observational controlled trial. PLoS One, 2020 (no benefit over placebo). https://doi.org/10.1371/journal.pone.0227497

29. Luo L, Zhou K, Zhang J, Xu L, Yin W. Interventions for leg cramps in pregnancy. Cochrane Database of Systematic Reviews, 2020 (low / very-low certainty evidence). https://doi.org/10.1002/14651858.CD010655.pub3

30. Ferraro PM, Taylor EN, Gambaro G, Curhan GC. Vitamin B6 intake and the risk of incident kidney stones. Urolithiasis, 2017 (supports the B6 / glyoxylate / oxalate mechanism; found no association between B6 intake and stone risk in the general population). https://doi.org/10.1007/s00240-017-0999-5

Ozempic Face: Fat Compartments, the Extracellular Matrix, and Why Rapid Weight Loss Shows in the Face First

Executive summary

• "Ozempic face" is not a marketing invention. It is the visible result of rapid, whole-body fat loss on GLP-1 receptor agonists (semaglutide, tirzepatide and related drugs) appearing in a place where small volume changes are impossible to hide: the face.

• The face is not padded evenly. It is organised into discrete fat compartments, and when these deflate quickly the overlying skin is left with more surface area than it has structure to fill.

• The problem is compounded from two directions. The skin's elastic scaffold, elastin in particular, renews extremely slowly and cannot re-tension at the speed fat is lost; and a meaningful share of the weight lost on these drugs is lean tissue, including muscle that contributes to facial contour.

• GLP-1 receptors are distributed across many tissues, not just the gut and pancreas. I regard the long-term consequences of chronically stimulating that system as incompletely understood, and I would rather say so than pretend otherwise.

• The rate of weight loss, more than which specific drug is used, is in my view the most useful predictor of facial change. I have not been able to find direct, head-to-head data on facial outcomes comparing semaglutide and tirzepatide.

• There are sensible measures that may reduce the facial cost, such as adequate protein from complete animal sources, resistance training, and attention to the skin, but none of them abolish it, and I cannot individualise them for a reader I cannot examine.

• Declared interest: I provide some of the aesthetic treatments discussed here, I offer private consultations, and I am, to be plain, not an enthusiast for these drugs. All three facts should colour how you read my recommendations.

• My settled preference is to help people change body composition through diet and training rather than manage the facial consequences of a drug. If you are already taking one, though, the decision to continue, adjust or stop belongs with your prescriber, not with a video or an article.

Introduction

I have patients on these drugs in my clinic most weeks, and roughly a year ago something changed in what they came in worried about. It was no longer only the number on the scale. It was their face. People told me they looked a decade older, that their cheeks had hollowed and their jawline had softened, and at least one patient sat in front of me in tears saying they regretted starting. The phenomenon acquired a nickname, "Ozempic face", faster than the profession produced a sober account of it. This is my attempt at that account.

I want to be careful at the outset about what I am and am not claiming, because this is an emotive subject and easy to distort. I am not saying these drugs do not cause weight loss; they plainly do. I am not saying every person who takes one will see their face change, nor that the change is always severe or permanent. What I am saying is narrower and, I think, defensible: the facial changes people describe are mechanistically predictable, they follow from how and how fast fat is lost, and the confident reassurance that there is nothing to see here does not survive contact with the anatomy and the biology. Where the evidence is thin, I will say so rather than fill the gap with certainty.

A word on where I am coming from, in both senses. I trained in dentistry, then in dermatology and separately in anti-aging medicine and in metabolic medicine; I hold a separate qualification in sports nutrition, which is directly relevant here because so much of this comes down to muscle and protein; and I have spent over a decade in aesthetic medicine and seen more than ten thousand patients. I also have interests I need to declare plainly, because they cut in more than one direction. I provide some of the aesthetic treatments I will discuss, so I have a commercial reason to think well of them. I offer private consultations. And I am, candidly, not a fan of these drugs, which is a bias too, and you should weigh it alongside the others. I have tried to let the anatomy and the retrieved evidence lead, and to flag my opinion as opinion when that is all it is.

What "Ozempic face" actually describes

The face is not a smoothly padded balloon. According to anatomical work retrieved from PubMed, the subcutaneous fat of the face is partitioned into discrete compartments separated by connective-tissue boundaries, rather than forming one continuous even layer (Ref 1). Later anatomical reviews have mapped how these compartments sit in the cheek, around the eye and in the temple, and how their volume and position shift with age (Ref 2). The ones that matter most for the "Ozempic face" appearance are the midface (malar) fat, the buccal fat of the cheek, and the temporal fat at the side of the forehead. Together they provide much of the youthful convexity of the face: the gentle fullness over the cheekbone, the smooth transition into the lower eyelid, the filled-out temple.

When you lose fat from the body, you lose it from these compartments too, because there is no mechanism that spares the face. And here is the crux: a small absolute loss of fat is trivial on the abdomen and conspicuous on the face. Think of a well-stuffed cushion. Take a handful of filling out of a large floor cushion and nobody notices; take the same handful out of a small scatter cushion and the cover visibly sags and creases. The face is the scatter cushion. This is why two people can lose the same number of kilograms and only one of them shows it in the face.

How these drugs cause weight loss, and a sharper reading

GLP-1 receptor agonists produce genuinely large weight loss. In the STEP 1 trial retrieved from PubMed, once-weekly semaglutide at 2.4 mg produced a mean loss of around 15% of body weight over 68 weeks, far more than placebo (Ref 3). The mainstream explanation for how they do this has three strands: they reduce appetite and food intake through receptors in the brain; they slow the rate at which the stomach empties, so you feel full for longer on less; and they enhance the glucose-dependent release of insulin while suppressing glucagon. According to a review retrieved from PubMed, the slowing of gastric emptying is a well-established action of these agents, with one important wrinkle: for long-acting agents such as semaglutide the gastric-emptying effect is subject to tachyphylaxis, meaning it wanes with sustained exposure, so chronic weight loss is driven more by the appetite and intake effects than by a permanently slowed stomach (Ref 4). The everyday picture is of a slowed conveyor belt: food moves through more slowly, fullness arrives sooner and lingers, and total intake falls. With less fuel coming in, the body draws on stored fat, oxidising it, which is to say dismantling it enzymatically by stripping electrons from its molecules to release usable fuel, and the fat mass shrinks.

In my videos I have emphasised one particular strand of this: the effect of these drugs on the smooth muscle of the gut, and the biochemistry underneath it. The GLP-1 receptor is a G-protein-coupled receptor of the stimulatory (Gs) type, so when it is activated it switches on an enzyme called adenylyl cyclase, and that enzyme consumes ATP, the cell's fuel currency, to manufacture a second messenger called cyclic AMP, or cAMP (Ref 17). In smooth muscle a rise in cAMP tends to promote relaxation. Alongside this, the retrievable experimental evidence shows a direct effect on gut motility: the GLP-1 receptor agonist exendin-4 reduced the contraction of colonic smooth muscle, acting through nitrergic and purinergic nerve signalling (nitric oxide and ATP acting as messengers) and by inhibiting calcium channels in the muscle cells (Ref 5). I would flag two things honestly. First, that motility study was done in rats and in isolated tissue, so it demonstrates a mechanism rather than proving its magnitude in a living human. Second, I want to be precise about the ATP point, because it is easy to overstate: adenylyl cyclase uses ATP as the raw material to build cAMP, and that cAMP then changes how the muscle behaves. That is an ordinary second-messenger cascade, not the wholesale depletion or poisoning of the muscle's energy supply, and I would rather describe the mechanism accurately than reach for the more dramatic version. Either way, the downstream point holds: motility is dampened, the gut moves less, and eating falls.

Why the face shows it first: systemic loss, not spot loss

You cannot choose where the fat comes off. Spot reduction, the idea that exercising or targeting a region preferentially strips fat from it, does not hold up. In a study retrieved from PubMed, twelve weeks of resistance training of one arm did not produce localised fat loss in that arm when measured by MRI; the fat loss was generalised across the body rather than confined to the trained limb (Ref 6). The same logic applies in reverse to the face: there is no way to lose weight from everywhere but the face, and no drug or diet directs fat away from it. What you can influence is the overall rate of loss, which, as I will come to, is the lever that matters most.

So the face deflates alongside everything else, but shows it more, for the cushion reason above. In some people the midface and temporal compartments visibly hollow, the cheekbone loses its soft cover, and the under-eye becomes more skeletal. None of this requires anything exotic. It is ordinary fat loss in an unforgiving location.

The skin cannot keep pace: elastin, collagen and the matrix

If fat loss were the whole story, the skin would simply shrink to fit. It does not, and the reason lies in the biology of the skin's scaffolding. The relevant proteins are collagen, which provides tensile strength, and elastin, which provides recoil, the ability to spring back after being stretched. According to a review retrieved from PubMed, elastin has a very low turnover: the body lays down most of its elastic fibres early in life, production falls after maturity, and there are very few interventions capable of meaningfully rebuilding the elastic-fibre network once it degrades. The same review notes that the combination of subcutaneous fat loss and degraded elastic fibres is precisely what produces looser, sagging skin (Ref 7).

An everyday analogy: elastin is like the elastic in a waistband. When it is new it stretches and snaps back; after years of wear it has been stretched too many times and no longer recovers, so when the thing it was holding gets smaller, the fabric just hangs. Rapid facial fat loss stretches and then abandons an elastic scaffold that renews too slowly to re-tension in time. Collagen is more renewable than elastin, but here I want to be precise about a point that is often muddled: collagen is not something you straightforwardly eat. It is a protein your own fibroblasts build, a triple helix rich in the amino acids glycine and proline. Animal foods are the best source of the raw amino-acid materials to build it, but the building is done by your cells, and whether they do it well depends on their condition, not merely on supply. When fat disappears faster than the collagen-and-elastin matrix can remodel, you get the double problem my colleagues and I see in clinic: volume loss underneath and lax skin on top.

Lean mass, and the muscles of the face

A second, less obvious contributor is muscle. It is now well described that a meaningful proportion of the weight lost on these drugs is not fat but lean tissue. According to a review retrieved from PubMed of incretin-based weight-loss drugs, treatment reduces both fat mass and lean (fat-free) mass, and the review's central interest is whether resistance exercise can shift that balance toward preserving muscle (Ref 8). A real-world study retrieved from PubMed put numbers on it: over three months on semaglutide, participants lost on average about 4.1 kg, of which roughly 1.4 kg, about a third, was lean mass rather than fat (Ref 9). I should give the honest counterpoint that the same study found that, as a proportion of the now-smaller body, the lean and skeletal-muscle share actually rose, so body composition on that measure improved; it is the absolute loss of muscle, and where some of it comes from, that concerns me for the face. It was also a small, short, retrospective study from a commercial wellness setting using compounded drug, so I would not lean on its exact figures too hard.

Why does this matter above the neck? Because muscles such as the masseter and temporalis contribute to the contour of the lower face and temple, and muscle in general is subject to this loss. Preserving it is the part of the picture you have the most control over, and it rests on two levers, both grounded in evidence retrieved from PubMed. The first is resistance training, which is the signal that tells the body its muscle is load-bearing and worth keeping (Ref 8). The second is protein, and specifically enough of the amino acid leucine, which is the strongest single dietary trigger of muscle-protein synthesis; a systematic review retrieved from PubMed found that the muscle-building response after a meal tracks with its leucine content (Ref 10). A per-meal leucine threshold of roughly two to three grams is often quoted, though I would treat that as a useful rule of thumb rather than a precise cut-off, because the literature is not unanimous. This gets harder with age: older adults show what is called anabolic resistance, needing more protein to mount the same muscle response, which is well reviewed in the PubMed literature (Ref 11).

On the source of that protein, I will be plain and let the numbers speak. Protein quality can be measured objectively rather than argued about, and on digestibility-corrected scoring retrieved from PubMed, animal-source proteins (in that study, dairy proteins) generally out-score plant proteins such as pea, soya and wheat (Ref 12). My own practice is to build meals around complete animal proteins that are free of anti-nutrients and come with their fat, because for muscle preservation on a drug that suppresses appetite, protein density per mouthful matters. If you avoid animal foods for ethical, religious or other reasons, that is entirely your business; I am pointing at a measurement, not issuing an instruction. One further practical point, well reviewed on PubMed: because these drugs cut how much people eat, they carry a real risk of protein and micronutrient inadequacy, which is another reason to make the food that does get eaten count (Ref 13).

Beyond the face: wide receptors, and the honest limits of what we know

There is a larger question sitting underneath the cosmetic one. GLP-1 receptors are not confined to the gut and pancreas. According to a receptor-mapping study retrieved from PubMed, they are expressed across a wide range of peripheral tissues (Ref 14). I would flag immediately that this particular map was made in mice using a reporter system, so it tells us where the receptor can be expressed in a rodent, not the functional importance of every site in a living human; whether cosmetically relevant receptors sit in human skin, and do anything meaningful there, is not something I would state with confidence. My unease is more general, and I will put it as opinion: I do not think it is obviously wise to stimulate a single receptor system continuously, for years, across the whole body, and I am still collecting data before I would say anything stronger than that.

On the related claim that low-dose use might be good for longevity: no intervention has ever been shown in a human lifespan trial to make people live longer, for the straightforward reason that such a trial is essentially impossible to run. So "good for longevity" is not an established fact for any drug, and I would be cautious about anyone who tells you otherwise. My own view, and I flag it as a view, is that continuous pharmacological appetite suppression is not the same thing as a well-constructed diet, and I am not persuaded it is good for you. If that turns out to be wrong, I will change my mind when the data arrive.

Semaglutide versus tirzepatide, and why the rate matters more

A question I am asked constantly is which drug is worse for the face, semaglutide (Ozempic, Wegovy) or tirzepatide (Mounjaro). The honest starting point is that tirzepatide generally produces more weight loss. In a head-to-head trial retrieved from PubMed, tirzepatide produced greater weight reduction than semaglutide, with the largest tirzepatide dose giving around five and a half kilograms more weight loss (Ref 15). I have to give the caveats, because they matter: that trial was in people with type 2 diabetes, its main purpose was blood-sugar control rather than weight, and it compared tirzepatide against a 1 mg dose of semaglutide, which is lower than the 2.4 mg dose used for weight loss. Indirect comparisons at the higher weight-loss doses retrieved from PubMed point the same way, more weight loss with tirzepatide, but indirect means the two drugs were compared through placebo across separate trials rather than in the same study, which is weaker evidence (Ref 16).

Here is the inference I will and will not make. If greater weight loss means, all else being equal, more facial volume lost, then more potent drugs carry more facial risk. But that is my reasoning from mechanism, not a measured finding, because I have not been able to find a study directly comparing facial outcomes between these drugs. What I will commit to, from clinical observation, is that the rate of weight loss looks like a better predictor of facial change than the identity of the drug. Losing fat fast, on anything, gives the elastic scaffold less time to adapt. Picture emptying a water balloon: let it down slowly and the rubber contracts and keeps a reasonable shape; empty it fast and you are left with a slack, wrinkled bag. If you are choosing or titrating between these drugs, that is a conversation to have with your prescriber, weighing your starting muscle mass, your metabolic situation and your baseline facial volume.

What may reduce the facial cost, before and during treatment

I want to be useful here without pretending I can prescribe for someone I cannot see. Everything below is general; none of it is individual advice, and how it applies to you depends on your circumstances, which I cannot assess through a screen. The theme is simple: slow the loss, defend the muscle, support the skin.

• Protein and training, started early. The muscle-preservation levers from the previous section, enough complete animal protein with adequate leucine, and regular resistance training with progressive overload, are most effective if they are in place before and during the weight loss, not bolted on after the face has already changed. If you are able to train in a way appropriate to your body, compound movements that load large amounts of muscle are a sensible backbone, but whether and how you should train depends on your training history, joints and medical background.

• Rate. Where you and your prescriber have latitude, a more gradual loss gives collagen, and to the limited extent it can, elastin, more time to remodel. Faster is not better here.

• Skin support. There is a reasonable case for measures that nudge the skin to maintain itself: well-tolerated topical agents such as niacinamide, and procedures such as controlled micro-needling or light-based (photobiomodulation) treatments that aim to stimulate the skin's own collagen production. I would treat all of these as marginal helpers rather than fixes, and I would be wary of anyone promising more.

• Body-composition scans. People often ask whether they should get a DEXA scan to track muscle. My filter is always the same: will the result change what I do? If you are already eating well, training and losing weight gradually, a scan with its small radiation dose usually will not change the plan, so in most cases I do not think it is justified. But that is a judgement to make with your own clinician for your own situation.

If the changes have already happened

If the face has already changed, there is a ladder of options, and I am going to lay them out from free upward, precisely because this is an area where clinics, mine included, have a financial interest in steering you toward the expensive end.

At the free end is time. When these drugs are stopped, weight is commonly regained: a systematic review and meta-analysis of GLP-1 discontinuation found a consistent metabolic rebound, with body weight and waist circumference climbing back over the months after stopping (Ref 18). With that regained weight, in my own practice I have seen the facial compartments refill to a degree over the following six to twelve months. I would be honest that the facial part specifically is a clinical observation rather than a measured trial outcome, that it does not always return to exactly how it was, and, importantly, that the decision to stop, continue or change a drug is one to make with your prescriber first, not on your own and not on my say-so. General practitioners in particular work within guidelines and time pressures that constrain them; the point is to have the conversation, not to bypass it.

Above that sits ordinary skincare, which will slowly support skin quality over months but will not refill lost fat. Then come clinical treatments, and here I am openly biased because I provide some of them. For lost volume, the usual tools are dermal fillers, which are reversible if the right ones are chosen and near-immediate in effect, or, for larger deficits, fat transfer. For skin laxity rather than volume, there are collagen-stimulating (biostimulatory) injectable treatments and energy-based devices, up to ablative lasers that tighten by removing tissue. I provide some biostimulatory injectable treatments myself, which is exactly why you should not take my enthusiasm at face value and should seek an independent view. I have deliberately not named brands or products in this article, because the principles matter more than the labels and the labels change.

Older patients, men and women, and the limits of generalisation

A few briefer points that come up repeatedly, with the caveat that they are tendencies, not predictions about you specifically.

Is it worse with age? In my clinical experience, yes, and the biology is consistent with it: older adults start with less muscle reserve, show the anabolic resistance to leucine noted above (Ref 11), and have slower collagen turnover and a more degraded elastic scaffold to begin with (Ref 7). The preventive work matters more the older you are, and recovery, if changes occur, tends to be slower and less complete.

Is it different for men and women? Men, on average, carry more subcutaneous facial fat than women of similar age and weight, which may buffer the early changes, but the underlying mechanisms are identical, and a buffer delays a process rather than preventing it. As with everything here, "on average" is doing a lot of work, and I cannot tell you where you personally sit on that distribution without seeing you.

Can you lose weight with no facial effect at all? Not reliably, and not by targeting. What you can do is slow the rate, defend the muscle and support the skin, which together shift the odds rather than guarantee an outcome.

Conclusion

"Ozempic face" is not the drug doing something mysterious to your face. It is the drug doing something quite understandable to your whole body, removing fat, and some muscle, faster than an ageing and slowly-renewing skin scaffold can keep up, in the one place where small changes are impossible to hide. Seen that way, most of the argument dissolves: the changes are real, they are mechanistically predictable, and they are partly modifiable but not abolishable. You can slow the loss, protect muscle with protein and training, and support the skin, and you can climb a ladder of treatments afterwards if you choose. But I would be failing you if I did not say the thing I actually believe: I would rather help someone reach a good body composition through diet and training, which is free and carries none of these trade-offs, than manage the facial cost of a drug I am not enthusiastic about. I have seen patients lose very large amounts of weight that way, eating to fullness and spending nothing. If you are on one of these drugs, none of that is a reason to stop abruptly; take it to your prescriber. I cannot see you through a screen, and the right answer is the one that fits you, made with someone who can.

Disclosures

I have direct commercial interests relevant to this article. I provide some of the aesthetic treatments I have described, including biostimulatory injectable treatments, so I profit if readers pursue them. I offer private consultations and paid education. And I hold a candid bias against these drugs on health grounds, which shapes my framing as surely as any financial interest would. I have tried to counter all three by naming no brands, by laying out free and low-cost options before expensive ones, and by grounding claims in retrieved evidence. I am happy to consult privately with anyone who wants to discuss their own situation, but I would say honestly that for most people a good, regular relationship with a local clinician is worth more than a single consultation with me.

References

Identified and verified via PubMed; DOI links included.

1. Rohrich and Pessa. The fat compartments of the face: anatomy and clinical implications for cosmetic surgery. Plast Reconstr Surg, 2007. https://doi.org/10.1097/01.prs.0000265403.66886.54

2. Cotofana and Lachman. Anatomy of the Facial Fat Compartments and their Relevance in Aesthetic Surgery. J Dtsch Dermatol Ges, 2019. https://doi.org/10.1111/ddg.13737

3. Wilding et al. Once-Weekly Semaglutide in Adults with Overweight or Obesity (STEP 1). N Engl J Med, 2021. https://doi.org/10.1056/NEJMoa2032183

4. Phillips, Rayner, Jones and Horowitz. Measurement of gastric emptying in diabetes. J Diabetes Complications, 2014. https://doi.org/10.1016/j.jdiacomp.2014.06.005

5. Yan et al. Effects of exendin-4 on colonic motility in rats and its underlying mechanism. Neurogastroenterol Motil, 2018 (animal, isolated-tissue study). https://doi.org/10.1111/nmo.13482

6. Kostek et al. Subcutaneous fat alterations resulting from an upper-body resistance training program. Med Sci Sports Exerc, 2007. https://pubmed.ncbi.nlm.nih.gov/17596787/

7. Baumann et al. Clinical Relevance of Elastin in the Structure and Function of Skin. Aesthet Surg J Open Forum, 2021. https://doi.org/10.1093/asjof/ojab019

8. Locatelli et al. Incretin-Based Weight Loss Pharmacotherapy: Can Resistance Exercise Optimize Changes in Body Composition? Diabetes Care, 2024. https://doi.org/10.2337/dci23-0100

9. Chun et al. Weight loss and body composition after compounded semaglutide treatment in a real world setting. Diabetes Obes Metab, 2025 (small retrospective real-world study). https://doi.org/10.1111/dom.16162

10. Wilkinson et al. Association of postprandial postexercise muscle protein synthesis rates with dietary leucine: A systematic review. Physiol Rep, 2023. https://doi.org/10.14814/phy2.15775

11. Paddon-Jones et al. Protein and healthy aging. Am J Clin Nutr, 2015. https://doi.org/10.3945/ajcn.114.084061

12. Mathai, Liu and Stein. Values for digestible indispensable amino acid scores (DIAAS) for some dairy and plant proteins may better describe protein quality than values calculated using PDCAAS. Br J Nutr, 2017. https://doi.org/10.1017/S0007114517000125

13. Urbina et al. Micronutrient and Nutritional Deficiencies Associated With GLP-1 Receptor Agonist Therapy: A Narrative Review. Clin Obes, 2026. https://doi.org/10.1111/cob.70070

14. Andersen et al. Using a Reporter Mouse to Map Known and Novel Sites of GLP-1 Receptor Expression in Peripheral Tissues of Male Mice. Endocrinology, 2021 (animal study). https://doi.org/10.1210/endocr/bqaa246

15. Frías et al. Tirzepatide versus Semaglutide Once Weekly in Patients with Type 2 Diabetes (SURPASS-2). N Engl J Med, 2021. https://doi.org/10.1056/NEJMoa2107519

16. Hankosky et al. Tirzepatide 10 and 15 mg versus semaglutide 2.4 mg in people with obesity or overweight with type 2 diabetes: an indirect treatment comparison. Diabetes Obes Metab, 2025 (indirect comparison). https://doi.org/10.1111/dom.16401

17. Stoicovy RA, Cora N, Perez A, et al. Cyclic adenosine monophosphate critically modulates cardiac GLP-1 receptor's anti-inflammatory effects. Inflammation Research, 2024 (cited for the GLP-1 receptor / Gs / adenylyl cyclase / cAMP pathway). https://doi.org/10.1007/s00011-024-01950-0

18. Tzang et al. Metabolic rebound after GLP-1 receptor agonist discontinuation: a systematic review and meta-analysis. eClinicalMedicine, 2025 (weight regain after stopping; does not assess facial change). https://doi.org/10.1016/j.eclinm.2025.103680

Hyaluronic Acid: Molecular Weight, CD44, and Why It Is Not Simply a Moisturiser

Executive summary

• Hyaluronic acid (HA) is marketed as a straightforward hydrating ingredient. Biologically it is more than that: it is a signalling molecule whose effects depend on its molecular weight.

• This note explains that high-molecular-weight HA and low-molecular-weight HA behave very differently, and that the body generates low-molecular-weight fragments in damaged or inflamed tissue.

• It describes CD44, the principal cell-surface receptor for HA, and the signalling that HA sets off when it binds.

• It sets out, carefully, what the cancer literature does and does not say about HA and CD44, and where the boundary of the evidence lies.

• It asks what this means for a cosmetic serum whose molecular weight you cannot see on the label.

• It closes on the more basic question of whether topical HA is needed at all, given that the skin produces its own.

• I have no commercial interest in any HA product, and I do not sell skincare.

Introduction

Hyaluronic acid is one of the most heavily promoted ingredients in modern skincare, almost always on a single claim: that it holds water, and therefore hydrates. That claim is not false, but it is a small part of a larger biological picture, and the larger picture is routinely left off the bottle. HA is not an inert sponge. It is a molecule the body uses to send signals between cells, and the signal it sends depends on its size.

I should be clear at the outset about what I am and am not arguing, because this topic is easy to misrepresent. I am not claiming that a hyaluronic acid serum causes cancer. I am making a narrower and, I think, more defensible point: HA is biologically active in a way that depends on its molecular weight, one of the receptors it acts through is deeply involved in tumour biology, and the marketing premise that HA is a simple, inert hydrator does not survive contact with the science. I trained in dentistry, then in dermatology and separately in anti-aging medicine and in metabolic medicine, and I have taught the biology of the extracellular matrix to clinicians. I have no financial interest in any skincare product.

What hyaluronic acid actually is

HA is a glycosaminoglycan, a long sugar chain, and it is one of the most abundant components of the extracellular matrix, the scaffolding between cells. The body makes it continuously through a family of enzymes called hyaluronan synthases, and in the skin the cells that produce it are the fibroblasts of the dermis. Crucially, the body makes it at a controlled size, in the right place, in the right amount. Its native, freshly synthesised form is a very long, high-molecular-weight chain.

This matters because the size is not incidental. It is, in effect, part of the message.

Molecular weight determines what HA does

The single most important fact about HA, and the one skincare marketing omits, is that high-molecular-weight HA (HMW-HA) and low-molecular-weight HA (LMW-HA) do close to opposite things. According to research retrieved from PubMed, native HMW-HA tends to be anti-inflammatory and immunomodulatory, while the shorter LMW-HA fragments are potent pro-inflammatory signals; the size of the molecule effectively sets the balance between the two (Ref 1). The same reviews describe how those long chains are broken down into short fragments in the first place: in damaged, infected or inflamed tissue, enzymes called hyaluronidases and reactive oxygen species chop HMW-HA into LMW-HA (Ref 1, Ref 5).

A useful way to picture this is that the length of the chain is a status report. A long, intact chain signals "tissue is healthy and quiet." A field of short fragments signals "tissue has been damaged," and the cell responds accordingly. The molecule is not just occupying space and holding water; it is telling the cells around it what state the neighbourhood is in.

I should be honest that this picture, while well supported, is not perfectly uniform. At least one study found that HA of various molecular weights did not activate macrophages as the pro-inflammatory model would predict (Ref 6). I would not lean on that result too heavily, though, given how it was produced: it used immortalised mouse macrophage cell lines rather than primary human cells, examined a single cell type through a narrow set of readouts, and was conducted entirely in a dish. Cell lines are convenient, but they do not always behave like the living tissue they are standing in for. The size-dependent story remains the mainstream reading of the evidence; I include this dissenting result because leaving it out would be dishonest, not because I think it overturns the picture.

CD44: the receptor that reads the signal

For a signal to be received, there must be a receptor. HA has several, but the principal one is a cell-surface protein called CD44. It sits in the membrane like a sensor on the outside wall of the cell, and when HA binds to it, it triggers changes on the inside. According to work retrieved from PubMed, HA is the main ligand for CD44, and that binding activates signalling pathways that drive cell proliferation, improve cell survival, reorganise the internal skeleton of the cell and increase cell movement (Ref 2). HA also engages other receptors, including the toll-like receptors and RHAMM, which is part of why its effects are so context-dependent.

The point worth holding onto is the one Dr Abs's critics missed when this was discussed on video: HA is a signal, and CD44 is how the cell reads it. Treating HA as though it were a passive humectant ignores the entire receptor biology.

What the cancer literature does, and does not, say

This is the part that needs stating precisely, because it is the part most easily distorted. The involvement of the HA-CD44 axis in cancer is not fringe speculation; it is a substantial, mainstream body of work. According to reviews retrieved from PubMed, HA is often enriched in tumours, in some tumour types the amount of HA present tracks with how malignant the tumour is, and HA-CD44 binding activates intracellular signalling (including Rho-family GTPase pathways) that promotes tumour-cell adhesion, growth, survival, migration and invasion (Ref 4). CD44 itself is one of the best-known markers of cancer stem cells, and its signalling is linked to metastasis and to the epithelial-to-mesenchymal transition that lets tumour cells spread (Ref 3). Separately, the fragmentation of HMW-HA into LMW-HA in the tumour environment is associated with the breakdown of the barrier between blood vessels and tissue, a step in angiogenesis and metastasis (Ref 5).

Now the necessary caveats, because they are what separate an honest account from a scare story. First, much of this evidence comes from cell lines and animal models and from the study of established tumours; it describes what HA and CD44 do inside cancer biology, not what a cosmetic ingredient does to healthy skin. Second, the literature is not tidy: one of the CD44 reviews notes openly that there are contradictory findings on whether high or low CD44 expression predicts a worse outcome (Ref 3). Third, and most importantly, none of this demonstrates that applying an HA serum to intact skin causes cancer, and I want to be unambiguous that I am not claiming it does. Whether cosmetic HA even penetrates living skin in a way that engages these receptors at all is itself an open question.

I want to address directly the way this argument has been misrepresented, because it matters. I have never said that hyaluronic acid causes cancer. I have said something quite different and quite precise: that HA is a molecular-weight-dependent signalling molecule, that its low-molecular-weight fragments act through receptors such as CD44, and that this same receptor system is deeply involved in tumour biology. Those are not the same statement, and collapsing the second into the first is not a paraphrase, it is a failure to read. It is the difference between saying a molecule participates in signalling pathways relevant to cancer and saying the molecule gives you cancer, and anyone who cannot hold that distinction has no business summarising a scientific argument for other people. The precision of the words is not pedantry here, it is the entire point. If you are going to quote a clinician on cancer biology, the least you owe your readers is to represent what was actually said, rather than the cartoon of it that is easier to be outraged by.

What the evidence does establish is narrower and still meaningful: HA is a size-dependent, receptor-active signalling molecule, and one of those receptors is central to tumour progression. That is a long way from "it just holds water."

What this means for a bottle of serum

Return, then, to the cosmetic product. If HA's biological effect depends on its molecular weight and its conformation, the obvious question is: what molecular weight is in the bottle? In practice, you usually cannot tell. Labels advertise the presence of hyaluronic acid; they rarely specify the size distribution, the conformational state, or the purity, and there is no reason to assume it matches the finely controlled material the body makes for itself. My point is not that this is proven to be harmful. My point is that the premise underpinning the whole category, that HA is a simple, inert, water-binding ingredient, is not a premise the biology supports, and a molecule with this much signalling activity deserves more scrutiny than a marketing line allows.

Do you need it at all?

There is a more basic question underneath all of this. The skin already makes its own HA, at the right size and in the right place, through its fibroblasts. If those cells are working well, the requirement to apply more from outside is not obvious. If they are not working well, then adding a single molecule to the surface does not address why they have stopped, and it is unlikely that HA is the only thing they have stopped producing.

My own clinical preference, and I put it as a preference rather than a proven prescription, is to focus on the conditions that let skin cells function properly, rather than on layering individual finished molecules onto the surface. The details of that are a separate topic. The narrower conclusion of this note stands on its own: hyaluronic acid is a signalling molecule whose behaviour depends on its size, not the passive moisturiser it is sold as.

Conclusion

Hyaluronic acid is genuinely interesting biology, and that is precisely why the "just adds water" framing is inadequate. High- and low-molecular-weight HA do different, sometimes opposing things; the body generates the pro-inflammatory short fragments when tissue is damaged; CD44 and related receptors read those signals; and the HA-CD44 axis is woven through tumour biology. None of that means a serum causes cancer, and I have been careful not to say so. It does mean that a molecule this active should not be treated as inert, and that the confidence of the marketing is not matched by the caution of the science.

Disclosures

I have no commercial interest in any hyaluronic acid product or skincare brand. I offer private consultations for those who want to discuss their own situation, though for most people a good, regular relationship with a local clinician is worth more than any single consultation with me.

References

Identified and verified via PubMed; DOI links included.

1. Hoarau A, Polette M, Coraux C. Lung Hyaluronasome: Involvement of Low Molecular Weight HA (LMW-HA) in Innate Immunity. Biomolecules, 2022. https://doi.org/10.3390/biom12050658

2. Chen C, Zhao S, Karnad A, Freeman JW. The biology and role of CD44 in cancer progression: therapeutic implications. Journal of Hematology & Oncology, 2018. https://doi.org/10.1186/s13045-018-0605-5

3. Hassn Mesrati M, Syafruddin SE, Mohtar MA, Syahir A. CD44: A Multifunctional Mediator of Cancer Progression. Biomolecules, 2021. https://doi.org/10.3390/biom11121850

4. Bourguignon LYW. Hyaluronan-mediated CD44 activation of RhoGTPase signaling and cytoskeleton function promotes tumor progression. Seminars in Cancer Biology, 2008. https://doi.org/10.1016/j.semcancer.2008.03.007

5. Singleton PA. Hyaluronan regulation of endothelial barrier function in cancer. Advances in Cancer Research, 2014. https://doi.org/10.1016/B978-0-12-800092-2.00007-1

6. Krejcova D, Pekarova M, Safrankova B, Kubala L. The effect of different molecular weight hyaluronan on macrophage physiology. Neuro Endocrinology Letters, 2009 (contradictory result: LMW-HA did not activate macrophages). https://pubmed.ncbi.nlm.nih.gov/20027154/

Why Your Face Sags After 40: It Might Be Your Skull, Not Just Your Skin

Executive summary

• This note examines why the face changes shape with age, and questions the common assumption that facial ageing is primarily a problem of the skin.

• It reviews the evidence that the facial skeleton itself remodels and recedes with age, drawing on three-dimensional CT imaging.

• It explains the structural consequence: as the bony scaffold recedes, the overlying skin and fat lose their support, which contributes to what we perceive as sagging.

• It examines the reported association between facial wrinkling and low bone density, and is careful about what that association can and cannot show.

• It assesses the two practical responses: dermal filler (what it can and cannot do), and protecting the skeleton itself through nutrition, mechanical load, micronutrients and dental health.

• It considers the specific case of women around menopause, when falling oestrogen accelerates bone loss.

• It closes with a measured practical summary, and discloses my commercial interests in full.

Introduction

When a patient in her forties or fifties tells me her face is "dropping", she almost always means her skin. The working assumption, shared by patients and by a good deal of my own profession, is that facial ageing is a surface problem: collagen is lost, the skin loosens, and it sags. That is part of the story. It is not the whole of it. A substantial and under-discussed contributor is that the facial skeleton beneath the skin changes shape with age. The bones recede, and the soft tissue that was draped over them loses its support.

I should state my background and my biases at the outset. I trained originally in dentistry, a discipline that spends years on the anatomy of the skull and face. I subsequently completed postgraduate training in dermatology, a qualification in anti-aging medicine, and a separate qualification in metabolic medicine. I also offer private consultations. I set all of this out now so that you can weigh what follows accordingly.

The facial skeleton is not fixed

It is tempting to think of the skull as inert, but bone is living tissue that is continually rebuilt. Two populations of cells do this work: osteoclasts, which dissolve old bone, and osteoblasts, which lay down new bone. A useful analogy is a demolition crew and a building crew working the same site in shifts. In youth, with sex hormones and nutrition in reasonable order, the two are roughly matched. Most people reach peak bone mass somewhere around the mid-twenties, depending on diet and lifestyle, and thereafter the balance tips: a little more is removed than is replaced. The bones of the face are not exempt from this.

What the imaging shows

The clearest evidence comes from three-dimensional CT studies that compared the facial skeletons of younger and older adults. The best known, by Shaw and colleagues, reported that with age the angle of the midface flattens, the eye sockets enlarge (particularly at the outer edges), the pyriform aperture (the bony opening of the nose) widens, and the jaw loses height (Ref 1). These are not subtle differences visible only to specialists; on a scan, an older facial skeleton can often be distinguished from a younger one by shape alone.

The limitation to keep in mind is that these are cross-sectional comparisons of different people at different ages, not the same faces tracked over time. They demonstrate a consistent pattern of age-related change; they do not, on their own, quantify how fast it happens in any one individual.

Why receding bone produces sagging

The structural logic is straightforward. Consider a marquee, the kind erected for an outdoor event. The skin is the canvas, the fat and muscle beneath it are the lining, and the facial bones are the poles. If a pole is shortened, the canvas above it slackens and drapes, not because the canvas has deteriorated but because it has lost its support. The same principle applies to a face whose underlying bone has receded.

How much of any individual's sagging is due to bone rather than skin cannot be judged from a distance, and I would not attempt to. In younger people the dominant factor is more often skin laxity and the redistribution of facial fat. From the forties onward, loss of bony support becomes a more plausible contributor, and for many people it is a combination of both. The point of this article is not that skin is irrelevant, but that the skeleton is routinely left out of the conversation.

Skin and bone: a shared loss of collagen

There is an intriguing line of evidence linking the two. As part of the KEEPS hormone study, a team at Yale examined women in early menopause who were not taking hormone therapy and reported that the more numerous and deeper a woman's facial wrinkles, the lower her bone density at the hip, spine and heel, independent of age and body mass (Ref 2).

This finding is easy to over-interpret, so it is worth stating its limits plainly. It is an association, not a demonstration of cause and effect; the sample was not large; and no mechanism was established by the study itself. Facial wrinkling is therefore not a substitute for a bone-density (DEXA) scan. What makes the association biologically plausible, rather than merely coincidental, is that skin and bone are both built substantially from collagen, and both appear to lose it in parallel. It is reasonable to regard skin quality and skeletal quality as two visible expressions of the same underlying process, while stopping short of claiming that one predicts the other in an individual.

Why this is rarely discussed in the clinic

If skeletal change is so central, why is it seldom raised? The explanation is largely structural, and I make it without wishing to disparage colleagues, most of whom were taught this anatomy alongside me. A patient presents with a visible complaint, and the tools available within a short consultation act on the visible complaint. There is an injectable for lost volume; there is no injectable for a maxilla that has receded over fifteen years. There is also an incentive structure worth acknowledging honestly: the skeletal side of facial ageing is preventive and slow, and it does not generate revenue in the way a procedure does. None of this requires bad faith; it is simply where the incentives lead.

Response 1: dermal filler, and its limits

This explains why filler is so often reached for when a patient presents with midface flattening, and in many cases that is a reasonable response. Filler cannot replace bone, but it can be placed over an area of bony recession to restore volume, and in competent hands, with an appropriate product, the result can look natural for a period of time.

The important qualification is that filler is being asked to compensate for a foundation that continues to recede beneath it. As the bone loss progresses, more filler tends to be required, and this is one route to the over-filled appearance that most practitioners, myself included, dislike. The filler has not malfunctioned; it has been asked to perform a structural role it was not designed for. My own position is not against filler, which I use in practice, but in favour of treating the foundation and the surface as separate questions, and of being candid with patients that volume placed on a shrinking base is a holding measure rather than a correction.

Response 2: protecting the skeleton

The more durable response is to look after the skeleton itself. Three levers have reasonable support.

Raw materials. Bone is, by weight, largely collagen, and collagen is built from amino acids, predominantly glycine and proline, arranged in a triple helix. Adequate high-quality protein is therefore not optional. Animal foods provide these amino acids in their most bioavailable form, along with the minerals and fat-soluble vitamins that bone metabolism draws on. The relative bioavailability of animal versus plant protein is not seriously disputed in the nutrition literature, and it is the basis on which I encourage patients to prioritise animal sources.

Mechanical load. Bone adapts to the forces placed on it, a principle long known as Wolff's Law. Osteoblasts are prompted to build bone by mechanical stress; remove the stress, as happens in prolonged bed rest or in spaceflight, and bone is lost. The best randomised evidence in postmenopausal women, the LIFTMOR trial, found that supervised heavy resistance and impact training improved bone density in a group often advised to avoid such loading (Ref 3). The important caveat for the face is that loading is largely site-specific: a heavy squat loads the spine and hips, not the maxilla. For the facial skeleton the relevant load is chewing, discussed below, and the wider benefit is systemic.

Micronutrients. The long-standing advice to supplement calcium and vitamin D has produced, at best, modest results, and there is a reasonable concern that calcium taken without adequate vitamin K2 may be deposited in arterial walls rather than bone. The biochemistry of K2 is well characterised: it activates the proteins that direct calcium toward bone. The human trial evidence, however, is mixed. A three-year trial of MK-7 in postmenopausal women with osteopenia found no benefit over placebo on bone density (Ref 4), whereas a study by Knapen and colleagues reported improved bone microarchitecture (Ref 5), and pooled analyses sit between the two (Ref 6). My reading is that the trials are limited by an unavoidable problem: it is not possible to control the background diet of free-living participants, so a null result cannot distinguish between "K2 does not help" and "the rest of the diet was too poor for it to help." I therefore weight the established biochemistry more heavily than I might otherwise, while acknowledging that this is a judgement, not a settled conclusion. The practical position is that micronutrients support the diet and the loading; they do not substitute for them.

Dental health and the jaw. The jaw is part of the facial skeleton, and its principal daily load is chewing. When the diet shifts to soft foods, that load is withdrawn, and the bone thins. The most striking clinical illustration is tooth loss: the section of jawbone that supported an extracted tooth resorbs once it stops receiving load, a well-recognised observation in dentistry. Retaining natural teeth, and eating foods that require chewing, both help to preserve the lower face.

Women, oestrogen and menopause

The picture changes markedly for women around menopause. Oestrogen restrains the bone-dissolving osteoclasts; when it falls, that restraint is lifted, and the years immediately following the final period are when bone is lost most rapidly. Skeletal facial ageing may therefore accelerate over the same interval.

Hormone replacement therapy is one legitimate response, and I am qualified to prescribe it. My own clinical preference, stated as a preference rather than a directive, is to address diet, load and the wider metabolic picture first, because I have seen these measures resolve a good deal of menopausal symptom burden in practice, after which the original case for prescribing may be weaker. Where HRT remains the right choice for a particular woman, it is a reasonable one, and the nutritional and loading measures described above matter regardless of that decision.

The soft tissue, and a disclosed interest

None of the above should imply that the skin is unimportant. Above the bone, the fibroblasts of the dermis produce collagen, elastin and fibronectin, and their output declines with age. Supporting that output is a large topic in its own right.

Here I must disclose an interest. The nutraceutical I personally use to support skin quality, Nutrakos, is formulated to stimulate the body's own production of multiple collagen types, together with elastin and fibronectin. I have a personal bias to it, so please treat this as a disclosed bias and evaluate the evidence for yourself rather than on my say-so.

Conclusion

The practical message is unglamorous. The measures that best preserve the shape of the face over decades are the same ones that preserve the skeleton generally: sufficient high-quality protein, genuine mechanical loading of the body and regular chewing for the jaw, an adequate micronutrient base, attention to oestrogen status around menopause, and the retention of one's own teeth. Filler has a place, but as a surface measure layered on top of these, not as a substitute for them. None of this requires perfection; optimising the large factors most of the time leaves ample room for ordinary enjoyment. It does, however, require attention to a structure that the usual conversation about ageing skin leaves out.

Disclosures

I offer private consultations, and I run an online community in which I discuss this kind of material in more depth; details for both are available via my site. I would add that a regular, long-term relationship with a good local clinician is, for most people, more valuable than any single consultation with me.

References

1. Shaw RB Jr, et al. Aging of the facial skeleton (three-dimensional CT analysis). Plastic and Reconstructive Surgery, 2011. https://pubmed.ncbi.nlm.nih.gov/17230106/

2. Pal L, et al. (KEEPS ancillary analysis), reported by Yale Medicine: deeper facial wrinkles associated with lower bone density in early-menopausal women; presented at the Endocrine Society annual meeting, 2011. https://medicine.yale.edu/news-article/not-just-skin-and-bones-wrinkles-could-predict-womens-bone-fracture-risk/

3. Watson SL, Weeks BK, Weis LJ, Harding AT, Horan SA, Beck BR. High-Intensity Resistance and Impact Training Improves Bone Mineral Density and Physical Function in Postmenopausal Women With Osteopenia and Osteoporosis: The LIFTMOR Randomized Controlled Trial. Journal of Bone and Mineral Research, 2018;33(2):211-220. https://doi.org/10.1002/jbmr.3284

4. MK-7 (vitamin K2), three-year randomised trial in postmenopausal osteopenia (null result). https://pubmed.ncbi.nlm.nih.gov/33030563/

5. Knapen MHJ, et al. MK-7 and bone microarchitecture. https://pubmed.ncbi.nlm.nih.gov/27625301/

6. Vitamin K2 and bone health, pooled meta-analysis. Frontiers in Public Health, 2022. https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.979649/full