#270 ‒ Journal club with Andrew Huberman: metformin as a geroprotective drug, the power of belief, and how to read scientific papers

Episode Summary

Executive Summary: Peter Attia and Andrew Huberman discuss how to read scientific papers through two contrasting journal-club papers: a retrospective metformin study that challenges earlier claims of geroprotective benefit, and a preprint showing that beliefs about nicotine dose can measurably alter brain activity and task performance. The episode emphasizes epidemiologic limitations, statistical interpretation, and the power of expectation in biology.

Main Topics: Metformin as a potential geroprotective drug (Priority: 5/5): Attia reviews metformin’s mechanism, its role in type 2 diabetes, and why it became a longevity candidate after earlier observational data suggested lower mortality. Reassessment of the Bannister metformin findings (Priority: 5/5): The Keys et al. paper is presented as a more rigorous reanalysis using Danish registry data and discordant twins, finding no mortality advantage that reverses the diabetes-associated risk. How to interpret epidemiologic studies (Priority: 5/5): The hosts explain censoring, confounding, hazard ratios, Kaplan-Meier curves, confidence intervals, and why large observational datasets still cannot fully replace randomized trials. Belief effects versus placebo effects (Priority: 4/5): Huberman distinguishes belief effects from classic placebo effects, arguing that expectations can change physiology and not just subjective experience. Nicotine dose beliefs and brain activation (Priority: 5/5): A preprint shows that telling smokers they received low, medium, or high nicotine changes thalamic and thalamocortical activation in a dose-dependent way, despite identical nicotine exposure. Practical implications for longevity and self-experimentation (Priority: 4/5): The discussion broadens to metformin, berberine, acarbose, SGLT2 inhibitors, fasting, caloric restriction, and the need for biomarkers of aging before strong anti-aging claims can be made.

Key Arguments: Metformin is a valuable diabetes drug, but its proposed longevity benefit remains unproven without randomized trial data. The Bannister study likely overstated metformin’s survival benefit because of informative censoring and other observational biases. The Keys study is more convincing because it uses a larger registry and discordant twins, yet it still cannot isolate metformin’s causal effect. Observational studies can suggest hypotheses, but confounding variables such as comedications, health status, and behavior can distort conclusions. Beliefs about a drug’s dose can alter brain physiology, not merely subjective feelings, implying that expectation is biologically active. Nicotine’s effects on attention-related circuits appear to scale with what users believe they received, even when the actual dose is constant. Current geroscience lacks reliable biomarkers of aging, making it hard to know whether interventions like fasting or metformin are truly slowing aging.

Data Points: Bannister study crude death rate (metformin group): 14.4 deaths per 1,000 patient-years - Referenced as the original observational finding that fueled excitement about metformin Bannister study crude death rate (control group): 15.2 deaths per 1,000 patient-years - Matched controls without diabetes in the original UK registry analysis Bannister study all-cause mortality reduction: 15% relative reduction over 2.8 years - The headline result that suggested possible geroprotection Type 2 diabetes lifespan impact: ~6 years shorter life expectancy - Attia notes the actuarial burden of type 2 diabetes compared with non-diabetics Keys study singleton sample size: 7,842 metformin users and 7,842 matched controls - Non-twin cohort in the Danish registry reanalysis Keys study twin sample size: 976 metformin users and 976 co-twins - Discordant twin analysis used to reduce genetic/environmental confounding Keys study crude mortality (singletons, metformin): 24.93 deaths per 1,000 person-years - Unadjusted mortality rate in diabetics on metformin Keys study crude mortality (singletons, controls): 16.86 deaths per 1,000 person-years - Matched non-diabetic controls in the singleton analysis Keys study hazard ratio (singletons, unadjusted): 1.48 - 48% higher annual mortality risk for metformin users with diabetes versus controls Keys study hazard ratio (singletons, adjusted): 1.32 to 1.33 - After adjusting for medications, marital status, and education Keys study hazard ratio (twins, unadjusted): 2.15 - Twin analysis showing substantially higher mortality risk in the diabetic twin on metformin Keys study hazard ratio (twins, adjusted): 1.70 to 1.80 - Adjusted twin models still showed elevated mortality risk Nicotine study sample size estimate: ~20 participants per belief condition for 90% power - Authors’ power estimate for detecting the observed effect size Nicotine study correlation coefficient: r = 0.27 - Reported for one thalamic belief-response relationship, described as modest

Pivotal Quotes: "This is a very important limitation of this study." — Peter Atiyah: On informative censoring in the Bannister metformin paper "What we believe about the effects of a drug, presumably, in addition to what we believe about how much we're taking and what those effects ought to be, clearly are impacting at least the way that our brain reacts to those drugs." — Andrew Huberman: Summarizing the nicotine belief-effect paper "We don't have a biomarker that gives us any insight into whether or not we're moving in the right direction." — Peter Atiyah: On the central limitation of geroscience and anti-aging interventions

Implications: Listeners should be cautious about extrapolating longevity claims from observational studies and should expect stronger evidence from randomized trials like TAME. The nicotine paper suggests expectations can shape biology, with broad implications for drug design, adherence, and clinical communication.