Journal Club With Dr. Peter Attia | Metformin for Longevity & the Power of Belief Effects

概览

This episode is the first Huberman Attia Journal Club, framed as a real-time walkthrough of how scientists and clinicians read papers: identify the question, examine methods and figures, evaluate statistics, and ask whether the conclusions are supported by the data.

The first major discussion centers on metformin and longevity. Peter Attia reviews why an earlier Bannister study helped fuel enthusiasm for metformin as a possible geroprotective drug, then contrasts it with a newer Keyes retrospective cohort study using Danish registry data and discordant twins. The main conclusion is cautious: metformin remains useful for insulin resistance and type 2 diabetes, but the transcript does not support treating it as a proven longevity drug for insulin-sensitive, highly active people.

The second major discussion focuses on belief effects using a nicotine fMRI study. Huberman argues that the study is interesting because people given the same low nicotine dose showed different subjective and brain responses depending on whether they were told the dose was low, medium, or high. The discussion extends this into broader questions about placebo, nocebo, side-effect framing, medication tapering, and how beliefs can shape physiology.

分段落总结

[00:00] Opening and Journal Club Format

[事实] Huberman introduces this as the first Journal Club episode with Dr. Peter Attia, focused on discussing and critiquing scientific papers.

[事实] The episode is set up around two papers: one on metformin and potential longevity effects, and one on placebo or belief effects involving nicotine.

[事实] Huberman says a journal club involves comparing one’s own interpretation of papers with the authors’ conclusions and highlighting key takeaways.

[推测] The format is intended not only to teach the findings, but also to model scientific skepticism for listeners.

[05:01] Why Reading Papers Is a Skill

[事实] Attia says he previously ran journal clubs inside his practice and describes paper interpretation as a skill.

[事实] He says reading only the abstract is usually not enough to understand a scientific paper.

[事实] Attia originally considered a more technical paper on cholesterol absorption but chose a simpler epidemiologic paper because it had broader relevance.

[07:27] Banter on Yerba Mate and Dreams

[事实] Attia describes a dream in which Huberman carried a special supplement drink containing dew collected from leaves.

[事实] Huberman connects the dream loosely to his real habit of drinking yerba mate, which he says he started drinking very young because his father is Argentine.

[事实] Huberman briefly describes recent intensive psychiatric analysis work and a transformative dream that affected decisions in his waking life.

[推测] This opening exchange functions as rapport-building before the technical journal club discussion begins.

[10:14] Metformin Paper and Background

[事实] Attia introduces the paper “Reassessing the evidence of a survival advantage in type 2 diabetics treated with metformin compared with controls without diabetes.”

[事实] He says the paper matters because a 2014 Bannister paper helped create excitement around metformin as a potential geroprotective drug.

[事实] Attia defines metformin as a long-used, generic, first-line drug for type 2 diabetes.

[事实] He says metformin’s mechanism is debated, but one accepted immediate effect is inhibition of complex I in mitochondria.

[11:59] Mitochondria, ATP, and Liver Glucose Output

[事实] Attia explains mitochondria as central to ATP production through oxidative phosphorylation.

[事实] He says metformin weakly inhibits part of the electron transport chain and changes the AMP-to-ADP ratio.

[事实] He says metformin clearly reduces hepatic glucose output, which is important because the liver helps regulate blood glucose.

[事实] Huberman asks whether metformin reduces conversion of food into blood glucose; Attia corrects this by saying it is better understood as changing fuel partitioning rather than blocking calories.

[17:23] Type 2 Diabetes and Insulin Resistance

[事实] Attia describes metformin as a first-line agent for type 2 diabetes.

[事实] He explains insulin resistance through muscle glucose uptake: insulin helps place a transporter at the cell membrane so glucose can enter muscle cells.

[事实] He says early insulin resistance can appear as normal blood glucose with high insulin, especially after meals.

[事实] Attia names inactivity, sleep deprivation, high cortisol, and energy imbalance as contributors to insulin resistance.

[24:20] Attia’s Personal Metformin Use and Geroprotection

[事实] Attia says he began taking metformin in May 2011 and initially took two grams per day without titrating, which made him very nauseous.

[事实] He says he was initially interested in metformin for glucose buffering, before “geroprotection” became part of his vocabulary.

[事实] Attia defines geroprotective agents as interventions aimed at underlying hallmarks of aging rather than one specific age-related disease.

[事实] He lists examples such as metformin, rapamycin, NAD, and NR as substances discussed in that category.

[27:09] The Bannister Study and Informative Censoring

[事实] Attia says the Bannister study used a UK registry to compare people with type 2 diabetes on metformin against matched controls without diabetes.

[事实] The matching attempted to account for variables such as age, sex, socioeconomic status, blood pressure, and BMI.

[事实] Attia highlights informative censoring as a major limitation: metformin patients who stopped the drug, were lost to follow-up, or progressed to more intensive medication were excluded.

[推测] This censoring may have made the metformin group look healthier than a real-world population of people with type 2 diabetes.

[30:32] Bannister’s Mortality Result

[事实] Attia says the Bannister study found crude death rates of 14.4 deaths per 1,000 person-years in the metformin diabetes group and 15.2 in controls.

[事实] He says this was surprising because type 2 diabetes generally shortens life expectancy.

[事实] He reports that the Bannister paper showed a 15% relative reduction in all-cause mortality over a median 2.8 years for people with diabetes taking metformin.

[事实] Attia says he once considered this suggestive that metformin might have benefits beyond glucose lowering, including possible effects on mTOR, inflammation, and senescent-cell signaling.

[36:28] The Keyes Study Design

[事实] Attia says Keyes and colleagues attempted to redo the Bannister analysis using a larger Danish health registry sample.

[事实] He says the study sampled about half a million people and included both matched unrelated individuals and same-sex discordant twins.

[事实] The twin design compared twins where one had diabetes and was on metformin and the other did not have diabetes.

[推测] The twin comparison strengthens control over genetics and early-life environment, but it still cannot make the study equivalent to random assignment.

[39:46] Medication Confounding

[事实] Attia points out that baseline characteristics were mostly well matched except for medication use.

[事实] People with type 2 diabetes on metformin were much more likely to be on lipid-lowering drugs, antiplatelet therapy, and antihypertensive medication than controls.

[事实] Huberman and Attia discuss this as a core limitation of epidemiology: without random assignment, not all confounders can be removed.

[推测] The medication imbalance suggests the metformin group had broader health burdens than the non-diabetic controls.

[41:39] Crude Mortality and Person-Years

[事实] Attia explains crude mortality per 1,000 person-years as a way to normalize deaths across different follow-up durations.

[事实] In the Keyes study, he says the non-twin control group had 16.86 deaths per 1,000 person-years, while the metformin diabetes group had 24.93.

[事实] For twins, he says the metformin diabetes group had 24.73 deaths per 1,000 person-years after age adjustment compared with 12.94 in co-twins without diabetes.

[事实] Huberman clarifies that the result does not mean metformin itself “kills you,” because the treated group had diabetes and other health issues.

[48:07] Kaplan-Meier Curves and Confidence Intervals

[事实] Attia explains Kaplan-Meier curves as survival curves with time on the x-axis and cumulative survival on the y-axis.

[事实] He says the curves for people with diabetes taking metformin remain below the curves for non-diabetic controls.

[事实] The shaded areas around the curves represent 95% confidence intervals.

[事实] Huberman adds that error bars and confidence intervals should not be interpreted simplistically without appropriate statistics.

[53:01] Statistical Power and Sample Size

[事实] Huberman and Attia discuss the importance of statistical power before interpreting whether a result is significant or not.

[事实] Attia says a study must be powered to detect an expected difference; failing to find significance does not rule out smaller differences.

[事实] Huberman notes that larger sample sizes generally reduce variability and that epidemiology uses scale to compensate for lack of experimental control.

[推测] This section is meant to teach listeners how to avoid overreading graphs or p-values without understanding study design.

[57:56] Keyes Hazard Ratios

[事实] Attia says the unadjusted hazard ratio in the Keyes singleton analysis was 1.48, meaning a 48% higher mortality risk for the metformin diabetes group versus matched non-diabetic controls.

[事实] After adjustment for medications and other factors, the singleton hazard ratio fell to about 1.32 or 1.33.

[事实] In the twin analysis, the unadjusted hazard ratio was 2.15, and adjusted models still showed substantially higher mortality risk for the twin with diabetes on metformin.

[事实] Informative censoring reduced the hazard ratios somewhat but did not reproduce Bannister’s finding of lower mortality.

[63:45] What the Keyes Study Does and Does Not Prove

[事实] Attia says the Keyes paper makes clear that metformin did not erase the mortality disadvantage associated with type 2 diabetes in that population.

[事实] He also says the study does not prove metformin failed to help, because it did not compare diabetics on metformin against diabetics taking nothing.

[事实] Attia says his intuition is that metformin may help people with diabetes, but not nearly as much as suggested by earlier enthusiasm.

[事实] He says a randomized controlled trial is needed to answer the question more definitively.

[68:58] Animal Data and the ITP

[事实] Attia discusses the Interventions Testing Program as a rigorous NIH-funded animal-testing framework for geroprotective compounds.

[事实] He says rapamycin showed lifespan extension in the ITP, including when started late in life in mice.

[事实] He says metformin failed in the ITP.

[事实] Attia also mentions 17-alpha estradiol and canagliflozin as interventions with positive ITP findings.

[71:35] Why Attia Stopped Taking Metformin

[事实] Attia says he stopped taking metformin about five years before the recording.

[事实] He says he noticed elevated resting fasted lactate while doing lactate testing for zone two training.

[事实] He links this to metformin acting as a weak mitochondrial toxin, shifting pyruvate toward lactate.

[事实] He says metformin may attenuate exercise performance, hypertrophy, and strength, though that topic is left for another time.

[73:21] Berberine and Acarbose

[事实] Huberman says he previously used berberine during a slow-carbohydrate diet period to buffer the effects of high-carbohydrate cheat days.

[事实] He reports that berberine made him feel better when eating large amounts of carbohydrate, but caused severe headaches if he did not eat carbohydrates soon afterward.

[事实] Attia describes acarbose as a drug that works in the gut by reducing glucose absorption.

[事实] Attia says acarbose showed a survival benefit in the ITP, and the mice lived longer without weighing less.

[78:13] Glucose Management, SGLT2 Inhibitors, and Diet

[事实] Huberman summarizes that insulin and blood glucose should be neither too high nor too low, and that regular exercise is a key tool for maintaining this system.

[事实] Attia says the only glucose-related drug he takes is an SGLT2 inhibitor, based on mechanistic studies, ITP results, and human trial data in non-diabetics with heart failure.

[事实] Attia says glucose buffering can be supported through sleep, exercise, and thoughtful eating without necessarily maintaining a calorie deficit.

[事实] He says he no longer believes macronutrient composition matters as much as he once did, assuming energy balance and food quality are reasonable.

[82:06] Current Position on Metformin

[事实] Attia says he still prescribes metformin to insulin-resistant patients.

[事实] He says he does not view metformin as a great tool for someone who is insulin sensitive and exercises a lot.

[事实] He says he would reconsider his view if the TAME trial shows a geroprotective benefit in non-diabetics randomized to metformin versus placebo.

[推测] His current stance is cautious rather than anti-metformin: useful in the right metabolic context, unproven as a general longevity drug.

[83:43] Fasting and Aging Biomarkers

[事实] Huberman asks whether short periods of caloric restriction or fasting have evidence for longevity benefit.

[事实] Attia says the short answer is no, partly because short duration may be insufficient and longer fasts also have downsides.

[事实] Attia says he previously performed repeated multi-day water fasts but now recognizes cumulative muscle loss as a downside.

[事实] He says geroscience lacks good biomarkers for true aging, making it difficult to know whether fasting, metformin, or rapamycin are moving the system in the right direction.

[87:36] How to Read Scientific Papers

[事实] Huberman recommends reading papers more than once and often one figure at a time.

[事实] He describes four guiding questions: what question the paper asks, what approach was used, what was found, and whether the conclusions are supported by the findings and methods.

[事实] Attia says he often starts with title, abstract, and figures, then moves into the text and methods when needed.

[事实] They note that supplemental information can contain important details not visible in the main article.

[93:43] Second Paper: Belief Effects and Nicotine

[事实] Huberman introduces the second paper as an unpublished bioRxiv paper, meaning it had not yet completed peer review at the time discussed.

[事实] He says the paper studies how beliefs about drugs affect people biologically, not just subjectively.

[事实] He distinguishes belief effects from simple placebo effects by saying belief effects can scale with the type and amount of information a person receives.

[事实] He names Ali Crum’s work as background for this area.

[97:18] Examples of Belief Effects

[事实] Huberman describes studies where people told stress is harmful performed worse under stress, while people told stress enhances performance performed better.

[事实] He says belief framing produced roughly opposite performance shifts, approximately 10% to 30% in either direction.

[事实] He describes a milkshake study in which ghrelin suppression differed depending on whether people believed the shake was high or low calorie, even though it was the same shake.

[事实] He describes a hotel-worker study where workers informed that their daily work counted as beneficial exercise lost more weight despite similar activity levels.

[100:57] Nicotine and Attention Circuits

[事实] Huberman says nicotine can enhance cognitive performance, while warning that delivery methods such as smoking or vaping can be harmful.

[事实] He explains that nicotinic acetylcholine receptors are named after nicotine but normally respond to acetylcholine.

[事实] He says nicotine and acetylcholine can enhance signal-to-noise in sensory processing through brain regions including the thalamus.

[事实] He connects nicotine’s effects to reward pathways, dopamine, and attention-related prefrontal circuits.

[106:39] fMRI Study Design

[事实] Huberman says the study used experienced smokers who abstained before the experiment.

[事实] Participants used vape pens and then entered an fMRI scanner while performing a market-style decision task involving investing or shorting.

[事实] The task was designed to engage thalamus, reward circuitry, and ventromedial prefrontal cortex.

[事实] The researchers measured nicotine exposure and brain activation, using fMRI as a blood-flow-based proxy for neural activity.

[113:16] Same Nicotine Dose, Different Beliefs

[事实] Huberman says all participants actually received the same relatively low nicotine dose.

[事实] Participants were told they had received a low, medium, or high nicotine dose.

[事实] Their subjective feeling of nicotine strength matched what they were told.

[事实] Huberman says thalamic activation scaled with instructed belief about nicotine dose.

[117:42] Dose-Dependent Brain Response

[事实] Huberman highlights figure two as showing that belief about nicotine strength induced a dose-dependent response in the thalamus.

[事实] He says the study did not include a zero-nicotine condition in which participants were told they received high nicotine.

[事实] He says this is important because drug studies usually examine dose-response curves, and here the dose-response pattern appeared from belief about dose.

[推测] The missing zero-dose condition limits how far the results can be generalized to pure placebo effects.

[120:48] Prefrontal Pathway and Study Limits

[事实] Attia notes that thalamic differences looked weaker than the pathway results.

[事实] Huberman says figure four showed significant scaling in the pathway from thalamus to ventromedial prefrontal cortex.

[事实] They discuss the small sample sizes typical of fMRI studies and the possibility of being underpowered for some comparisons.

[事实] Huberman says belief about nicotine strength did not modulate reward-pathway dopamine response in the same way.

[124:22] Implications for Tapering and Medication Beliefs

[事实] Huberman suggests the findings could be relevant to nicotine tapering if a person believes they are receiving more nicotine than they are.

[事实] Attia asks how the experiment might differ in non-smokers; Huberman says nicotine-naive people might respond strongly even to a small dose.

[事实] Huberman says the study is not definitive, but he values it because it begins to explore dose-dependent belief effects.

[事实] They extend the implication to drugs, behavioral interventions, vaccines in general, psychoactive drugs, antidepressants, metformin, and ADHD medications.

[130:47] Nocebo Effects and Side-Effect Framing

[事实] Huberman says belief about side effects can shape how tolerable a drug feels and how much relief people report from the target symptoms.

[事实] He warns against lying to people, but also against framing side effects in a way that pushes people toward nocebo responses.

[事实] Attia says some people seem especially affected by nocebo responses and may experience many expected side effects.

[事实] Attia also says medical training can distort perceived prevalence of rare conditions because students spend so much time studying them.

[134:03] Closing

[事实] Huberman and Attia say they want to do another journal club in Austin.

[事实] Huberman closes by thanking Attia and asking listeners to subscribe, review, comment, check sponsors, sign up for the newsletter, and follow his social media.

[推测] The closing reinforces that this is intended as the start of a recurring collaborative format.

播客点评/总结

The episode’s strongest value is its method: the hosts do not simply report study conclusions, but show how to question design, confounding, censoring, hazard ratios, power, figures, and the gap between results and author claims.

The metformin section is especially useful for listeners interested in longevity because it separates clinical utility in insulin resistance from unproven geroprotection in healthy, insulin-sensitive people. Its limitation is that the central human evidence discussed remains observational, and the transcript repeatedly stresses that a randomized trial is needed.

The belief-effects section is memorable because it moves placebo discussion from subjective expectation into measurable brain activation. Its limitations are also clear in the transcript: the nicotine paper was not yet peer-reviewed, involved a small fMRI-style sample, and lacked a zero-nicotine control condition.

[推测] This episode is best suited for listeners who want to learn how to think through biomedical evidence, not for someone seeking a simple medication protocol. It will be most valuable to people interested in longevity science, metabolic health, placebo and nocebo effects, neuroscience, or critical reading of medical papers.