MAGELLAN LONGEVITY

Peter Attia's Longevity Claims, Checked Against the Papers

Sixty-nine claims from fifteen appearances. Attia comes out of this better than anyone else in the series — and the single most important thing I found is that most of what circulates as "Attia was wrong" was said by someone else on his show.

The short version

Of the 69 claims pulled from Attia's appearances, only 49 are Attia speaking. Twelve belong to Jason Fung, eight to other guests. When you separate them, the picture inverts:

The same 69 claims, sorted by who actually said them.
SpeakerClaimsSupportedOverstatedContradictedRate wrong
Attia himself4924408%
Jason Fung (guest)1236158%
Other guests842025%

An 8% error rate, with zero flatly contradicted claims, is a better record than most textbooks. If you take one thing from this page: when Peter Attia states a number, it is usually right. That is not faint praise in this field.

It also means the interesting questions about him are not accuracy questions. They are about the four places he does overreach, about what happens when he hands his microphone to someone with a 58% error rate, and about a conflict of interest he discloses but keeps.

The four places he overreaches

Protein at 1 gram per pound. "If you weigh 150 pounds, are you getting 150 grams of protein a day?" That is roughly 2.2 g/kg. The best available meta-analysis and meta-regression found that resistance-training gains from added protein plateau at about 1.6 g/kg per day DOI. Attia's target is about 40% above the breakpoint. For a healthy person with normal kidneys this is not dangerous, just unnecessary and expensive — but it is worth noting that Bryan Johnson, who gets far more criticism, uses exactly the meta-analytic number.

"Some 90% of adults at some point are going to struggle with insomnia." Standard epidemiology puts insomnia symptoms at around 30% of adults and chronic insomnia at 6–10%. Ninety percent is far above any published prevalence. This is the kind of number that sounds harmless and quietly recalibrates what an audience thinks is normal.

"Decaf coffee can have as much as 15 to 30% of the caffeine that regular coffee has." He notes on air that he could not find the reference, which is honest. The measured answer: decaffeinated brewed coffee and espresso contained 0 to 13.9 mg of caffeine per 16 oz serving DOI — typically under 5% of a regular coffee, not 15–30%.

"Lean mass loss happens at 8 to 10% per year." Taken literally that would consume a person. It is almost certainly a transcription artifact or a slip for "per decade," and Attia gives the correct figure — about 1% per year — in a different episode. I include it because it is in the record, not because I think he believes it.

The guest problem

This is the substantive criticism, and it is not about Attia's knowledge. It is about what a long-form interview does to a listener.

Attia's episode with Jason Fung is nearly three hours of two credentialed physicians in agreement-shaped conversation. A listener has no way to distinguish the host's claims from the guest's, and the guest's claims do not hold up:

Claims made by Jason Fung on Attia's podcast, checked.
ClaimThe record
Insulin-treated diabetics "all gained like 30 pounds — the DCCT showed that" DCCT reported 5.1 kg — about 11 lb — in the first year of intensive therapy, versus 2.4 kg standard. Roughly a third of the stated figure. DOI
In the glucose-vs-fructose beverage study, the fructose group "became diabetic" on OGTT Stanhope 2009 found fructose-sweetened beverages increased visceral adiposity and decreased insulin sensitivity over 10 weeks. Subjects did not become diabetic. DOI
"We could never show any benefit to tight glucose control in type 2 diabetes" Too absolute. UKPDS showed microvascular benefit; the genuine controversy is about macrovascular and mortality endpoints. DOI
"80 to 90% of those type 2 diabetics get off all [medication]" with fasting No study supports this. The underlying evidence is a case series of three patients.
Relaying Valter Longo: seven days of fasting "resets your immune system" The source paper used 48–72 hour fasting cycles in mice, plus a small human chemotherapy pilot. DOI
Insulin causes sodium reabsorption at the proximal tubule Insulin's antinatriuretic effect is classically localised to the distal nephron, though it acts at multiple sites. A defensible correction, but a narrower one than our own notes originally claimed.

One more, from a different episode: in the training roundtable, Gabrielle Lyon says animal foods provide "nearly 100% of our calcium and bioavailable iron, zinc, selenium." NHANES analysis shows dairy supplies the majority of dietary calcium but nowhere near all of it, and plant foods contribute substantially to iron, zinc and selenium intake DOI.

None of this is Attia's fault in the sense of him being wrong. It is his responsibility in a different sense: he is the reason those statements reached a large audience with an implicit endorsement, and he did not push back on any of them in the moment.

Where he is unusually good

Attia is right about the thing that matters most, and he is right about it more forcefully than most cardiologists are in public. His position — that apoB is the number to track, that the evidence for lowering it is "as unambiguous as any data are in medicine," that statins reach most of their effect at low doses — is mainstream lipidology stated plainly. He takes a statin himself.

I flag this because "Attia is a statin skeptic" is a persistent and completely false claim about him. The criticism he actually receives from other physicians is the reverse: that he over-medicalises prevention.

Also checking out: the melatonin supplement-content data (measured content ranged from -83% to +478% of label DOI), his sub-1 mg melatonin dosing guidance, the beta-blocker/melatonin interaction, FDA zolpidem dosing, CBT-I efficacy, the LIFTMOR resistance-training bone data, and the entire cardiorespiratory-fitness mortality section.

The conflicts he does disclose

AG1. Attia has confirmed he is an investor and adviser to AG1 while recommending it daily. The disclosure is real and repeated, which is more than many do. But disclosure does not neutralise a conflict, it just makes it visible, and the underlying product question — whether a 75-ingredient proprietary blend delivers effective doses of anything — remains unanswered by the company.

Rapamycin. He has been among the drug's most prominent advocates and funded his own trial of rapamycin plus exercise. In May 2026 he published the result on his own site under the heading "Disappointing results": the drug "didn't enhance functional gains from training, and appears to have modestly blunted them."

Publishing that is genuinely to his credit and I want to say so clearly. It is also worth sitting with what it means. Cardiologist Eric Topol, quoted in The New Yorker, had warned: "This isn't like a vitamin… It's a serious drug that can potentially lower your body's ability to fight off infections," and noted his own patients were reading Outlive and requesting prescriptions. Topol's concern — that "his followers aren't going to be able to detect which recommendations are firmly grounded in evidence" — is the whole problem in one sentence, and it applies to a figure whose individual claims are, as this review found, mostly accurate.

What I think this adds up to

Attia is the most reliable narrator in this series and the one whose influence carries the most clinical risk, and those are not contradictory. Accuracy at the sentence level is not the same as calibration at the recommendation level. A person can get every citation right and still push an audience toward a monitoring-and-intervention posture that the evidence does not support — full-body MRIs with their false-positive burden, CGMs in non-diabetics, off-label rapamycin.

The practical read: take his lipid material seriously, take his exercise material seriously, treat the protein number as a ceiling rather than a target, and remember that when a guest is talking, the accuracy guarantee does not transfer.

How this was checked

Fifteen of Attia's most-viewed appearances were transcribed and each factual claim pulled verbatim with a timestamp. Claims were attributed to the person who actually said them; guest claims are reported as guest claims and counted separately. Each citation was verified against the PubMed record — author, year, journal, finding. Article metadata retrieved from PubMed. Full methodology →

Evidence appraisal, not medical advice. No physician–patient relationship is created by reading this.