Why Peter Attia Changed His Mind on Keto

A hedge fund manager once handed Peter Attia a phrase that has outlasted most of his trades: strong convictions, loosely held. The idea is simple. Put real money, or a real clinical opinion, behind a thesis, but hold it loosely enough that the moment new evidence shows up, you can let it go without grief (02:50). On this episode of his podcast The Drive, marking roughly eight years and 400 episodes, Attia, a physician known for his long-form interviews on longevity and for his book Outlive, uses that phrase to revisit two of his most public positions: the ketogenic diet and extended fasting. What he finds is less a story of being wrong than a story of having had an incomplete explanation.

Attia says his book should not read the same way a decade from now. If it did, he argues, that would mean either he had stopped learning or the science had stopped moving, and neither option sits well with him (03:48). Medical opinion, in his view, should not update smoothly. It lurches.

Scientific progress is, you know, by its very nature, somewhat glacial and episodic. It doesn't just progress at the same rate. [04:43]

Three forces push him to revise a position, he says: the published literature, conversations with experts he interviews on the podcast, and his own patients, who he calls teachers (05:11). Sometimes those three sources converge and reinforce what he already believed. Sometimes they add nuance and carve out exceptions. And sometimes, as with nutrition, they change his mind completely (05:37).

From insulin to energy balance

Attia's original thinking on diet started with the insulin model of obesity, the idea that carbohydrate intake drives insulin, insulin drives fat storage, and fat storage drives metabolic disease, the condition he calls the foundational problem behind nearly every other chronic illness (06:43). Cutting carbohydrates step by step eventually forces a choice: go all the way into ketosis, a metabolic state where the body burns fat for fuel and produces ketone bodies, or stay in a middle zone that feeds neither the brain's glucose needs nor its ketone needs well (07:13). Attia chose to go all the way. He stayed in ketosis for three years, logging extensive personal data, with an average morning fasting beta-hydroxybutyrate (a ketone body used to measure depth of ketosis) of 1.8 to 2 millimoles per liter, rising higher through the day as he ate mostly fat (09:14). He calls it the best he ever felt.

I abandoned the ketogenic diet basically because I was just tired of being that restrictive in my eating. [08:50]

That line matters because it separates two questions people tend to merge: did the diet fail, and did he quit it. He quit it. It did not fail. To this day he calls ketogenic eating effective. What changed is his explanation for why.

I used to believe that the diet worked through insulin suppression, but today, I don't think that's the case. [09:42]

The insulin model predicted that lowering insulin should, on its own, fix metabolic dysfunction: less visceral fat, better insulin sensitivity, resolved liver fat. Attia now reads the data differently. He thinks metabolic dysfunction, visceral fat accumulation, and declining insulin sensitivity are downstream results of energy imbalance, meaning eating more energy than the body uses, and that chronically elevated insulin, or hyperinsulinemia, simply rides along with that imbalance rather than causing it (10:30). The clearest evidence, he says, is watching someone on a ketogenic diet who is still eating in surplus. Their insulin drops, as expected, but they do not lose weight. Some even gain it (11:18). If suppressed insulin were the actual lever for fat loss, that should not be possible.

So why does keto work when it does work? Attia now credits satiety. Many people find the foods allowed on a ketogenic diet, heavy in fat and protein, more filling than carbohydrate-dense foods, which makes it easier to eat in a deficit without consciously counting anything (09:42). There may also be something to burning fatty acids for fuel directly supplying energy during a weight-loss phase (10:30). The mechanism changed. The usefulness of the diet, for the right person, did not.

One consequence of this shift is that Attia has become far less dogmatic about which diet a patient should choose. He no longer has a strong preference between ketogenic, low-fat, or mostly pescatarian eating patterns, as long as one line holds: a minimum of 1.4 grams of protein per kilogram of body weight, regardless of what else is on the plate (11:44).

I really am pretty adamant about people maintaining the right amount of protein, and at a minimum, I want them hitting one point four grams per kilogram of body weight. [11:44]

Fasting and the analog nature of autophagy

A parallel story runs through his views on extended fasting. For about three years, Attia ran a personal protocol: a fast of seven to ten days once a quarter, and a three-day fast in each of the other two months of that quarter (12:36). He was not trying to lose weight, he says; his weight was already where he wanted it. He was chasing autophagy, the process by which cells break down and recycle old intracellular proteins and structures, literally self-eating in order to reuse worn-out parts (13:10). The theory, well supported in preclinical, animal-based research, is that under nutrient scarcity the body recycles its worst material first: cells heading toward senescence, meaning they have stopped dividing and lost normal function, or cells carrying mutations (13:37).

The trouble is dosing. Ketosis has a rough, agreed-upon threshold: somewhere between 0.5 and 1 millimole of beta-hydroxybutyrate at rest marks the entry point (14:43). Autophagy has no comparable blood marker and no clean switch.

Autophagy is not digital, it's analog. It's not like you have no autophagy and then you cross a threshold. [14:19]

Based on conversations with researchers and preclinical data, Attia now estimates that roughly three days of fasting is where autophagy rises meaningfully beyond what exercise alone produces (14:43). That is an approximation drawn from animal studies and expert opinion, not a measured human cutoff, and he says plainly that he has not reviewed this literature in a couple of years, so the estimate could be out of date.

Given that uncertainty, why stop a protocol he once ran for three years straight? Attia points to two things together: the ambiguous evidence for exactly how much fasting is needed, and the practical cost of repeatedly going without food for a week or more at a time (15:09). The fuller answer, along with his updated thinking on metformin, rapamycin, whole-body MRI, and strength training, is reserved for paying subscribers; this preview ends just as he starts to unpack it.

What carries across both stories, keto and fasting, is not regret. Attia still eats low-carbohydrate at times. He still thinks fasting promotes autophagy. The thing that moved is the explanation underneath the behavior, and the confidence with which he holds any single mechanism as the whole answer. That is the actual content of strong convictions, loosely held: keep the practice that works, and be willing to rewrite the theory behind it the moment the data ask you to.