300-Special episode: Peter on exercise, fasting, nutrition, stem cells, geroprotective drugs, & more

KEYPOINTS

• 01

  Peter sorts claims into five evidence tiers: proven, promising, fuzzy, noise and nonsense.

  For the 300th episode, Nick Stenson put listener-submitted topics, drawn from a few thousand Instagram comments, to Peter for ranking. In biology nothing is truly proven, so 'proven' means the chance it's untrue is so small it would be foolish not to act. 'Promising' has strong data with a missing piece, such as human or randomized trial evidence. 'Fuzzy' has inconsistent or contradictory data. 'Noise' has data too thin to judge, perhaps with an appealing mechanism. 'Nonsense' has been studied and found to be bunk. Peter stresses that rankings move as evidence changes.

• 02

  Rapamycin: promising, not proven.

  Rapamycin was discovered in Easter Island soil bacteria in the 1960s and championed by Suren Sehgal. The FDA approved it for organ transplantation in 1999. In 2009 the Interventions Testing Program (ITP) showed it extended life in male and female genetically diverse mice, even when started late in life, which no other drug had done. It also works in yeast, worms and flies. What's missing is human evidence that intermittent dosing, which short trials suggest can enhance rather than suppress immunity, actually extends life. Peter has taken it for about six years, yet fewer than 10% of his patients do, and he calls handing it out 'like Tic Tacs' irresponsible.

• 03

  Metformin: downgraded from promising to fuzzy.

  Bannister et al. (2014, UK data) found diabetics on metformin had about 15% lower all-cause mortality than non-diabetics, despite worse glucose control. Peter argues this was likely driven by informative censoring, since patients were dropped when their medication changed, which selected for the healthiest. Keyes et al. (2022), using Danish data that included twins discordant for diabetes, found metformin users still had 33–80% higher mortality. Metformin also failed in the ITP except when combined with rapamycin. The TAME trial may eventually test it in humans.

• 04

  NAD and its precursors: noise. Resveratrol: nonsense.

  NAD declines with age and is needed by sirtuins, but whether the decline is causal is unknown. The best human data are tiny: a four-month ALS study of nicotinamide riboside (NR) plus pterostilbene showing slowed functional decline, a 20-person Parkinson's study confounded by levodopa timing, and a small MCI study with no cognitive benefit. Peter takes no NR, NMN or infusions. Resveratrol's 2006 mouse benefit came in mice fed 60% coconut oil whose fatty livers collapsed their lungs. The rigorous ITP test at 300 ppm in food, versus under 2 ppm in wine, showed nothing, a result proponents signed off on at the time.

• 05

  VO2 max, muscle mass and strength are as close to proven as it gets.

  Peter ranks them alongside quitting smoking and controlling blood pressure. Fitness predicts lifespan better than cholesterol, blood pressure, glucose or even smoking, partly because 'you can't cram for the test': a top-2–3% VO2 max or strong grip reflects years of work. He tells patients chasing supplements that they're 'picking up pennies in front of a steam train.' Even if exercise were lifespan-neutral, he says, its quality-of-life benefits would make it worth it.

• 06

  Long-term fasting: fuzzy, and Peter stopped his own protocol.

  Caloric restriction, like rapamycin, extends life across model organisms. But the duration, depth and frequency that might work in humans are untested, and he admits his old routine of 7–10 days of water-only fasting each quarter plus 3-day fasts in other months was 'made up.' Fasting crashed his thyroid and androgen markers, and over about three years he lost roughly 16–20 lb of muscle he couldn't regain. In about 2021 he stopped, partly also to avoid fasting in front of his kids. He still uses time-restricted eating for calorie control.

• 07

  Energy balance: between promising and proven.

  A study funded by an organization Peter once ran found that equal-calorie diets with very different macronutrients don't meaningfully change energy expenditure. Thermic effects and appetite differ by food, but someone switching from 3,000 balanced calories to 3,000 keto calories who rapidly loses weight is, he believes, moving more or eating less than they realize. Changing your mind with new data is what separates a scientist from an advocate.

• 08

  'Red meat causes cancer' is nonsense in Peter's ranking, though diet patterns matter.

  In a European cohort of nearly half a million people, the heaviest red and processed meat eaters (about 160 g/day) versus near-zero eaters (about 20 g/day) had colorectal cancer risks of about 1.7% versus 1.33%. That's about 0.4 percentage points, or treating roughly 250 people to prevent one case, even assuming causation. Peter blames food-frequency questionnaires and healthy-user bias, and suspects low fiber and vegetable intake drive much of the risk. He considers the weak hazard ratios incompatible with Bradford Hill causality, unlike smoking.