Huberman LabJournal Club with Dr. Peter Attia | Metformin for Longevity & The Power of Belief Effects
CHAPTERS
- 0:00 – 14:00
Intro, Journal Club Concept, and Sponsors
Huberman introduces the first joint ‘journal club’ with Peter Attia, explains what a journal club is, and previews the two focal papers: one on metformin and longevity, and one on placebo/belief effects and nicotine. He also clarifies that the podcast is separate from his Stanford role and reads sponsor messages.
- •Journal clubs involve structured group critique of one or two scientific papers.
- •Attia will present a metformin survival paper; Huberman will present a belief‑dependent nicotine fMRI paper.
- •Audience will learn both new science and how to dissect papers like scientists and clinicians.
- •Huberman reiterates his goal to provide zero‑cost science and thanks sponsors (Helix, Levels, AG1, InsideTracker, Momentous).
- 14:00 – 24:00
Why Journal Clubs Matter & A Dream About Dew
Huberman and Attia reflect on running journal clubs in their own environments and how people often misunderstand how to interpret abstracts and headlines. Attia shares a humorous dream about Huberman obsessively carrying an elixir made with collected morning dew, which segues into a brief detour on yerba mate and dreams.
- •Most people ask, ‘Is it enough to read the abstract?’—the answer is no.
- •Journal clubs train skills in identifying limitations, methods, and over‑interpretations in papers.
- •Attia’s dew dream illustrates how rituals, supplements, and beliefs can take on quasi‑magical importance.
- •Huberman mentions recent personally transformative dreams arising from prolonged psychodynamic work.
- 24:00 – 41:00
Metformin 101: Mechanism, Mitochondria, and Type 2 Diabetes
Attia defines metformin, its history as a first‑line type 2 diabetes drug, and its debated mechanisms, focusing on partial inhibition of mitochondrial complex I and reduced hepatic glucose output. He then gives a mechanistic primer on insulin signaling, insulin resistance, and the path from early hyperinsulinemia to full‑blown type 2 diabetes.
- •Metformin is a decades‑old, generic first‑line therapy for type 2 diabetes (brand name Glucophage).
- •Its clearest acute effect is mild inhibition of mitochondrial complex I, changing AMP/ADP ratios and reducing liver glucose production.
- •Insulin binds its receptor, triggers intracellular phosphorylation, and inserts GLUT transporters (‘straws’) into muscle cells, allowing passive glucose influx.
- •Ectopic fat in muscle disrupts this signaling, leading to insulin resistance: more insulin needed to get the same glucose uptake.
- •Early warning sign: high insulin with normal glucose (hyperinsulinemia with normoglycemia).
- •Over time, both liver and pancreas become insulin resistant, hepatic glucose output rises, and beta cells fail, sometimes necessitating exogenous insulin.
- 41:00 – 52:00
Causes of Insulin Resistance & Lifestyle Levers
Attia outlines major drivers of insulin resistance—low activity, excess energy intake, ectopic fat, sleep loss, and high cortisol—and highlights Gerald Shulman’s work on intramyocellular lipid. Huberman asks for practical factors; they stress exercise and sleep as powerful modulators of glucose disposal.
- •Primary drivers: inactivity, energy surplus leading to fat in muscle/liver/pancreas, sleep deprivation, and hypercortisolemia.
- •Sleep restriction to 4 hours/night for one week can cut glucose disposal capacity by ~50%.
- •Insulin resistance starts in muscle, then extends to liver and pancreas; diabetes is a late stage of a long process.
- •Early detection demands looking at insulin levels and post‑prandial response, not just fasting glucose.
- •Lifestyle interventions—more movement, better sleep, stress management—are foundational for glucose control.
- 52:00 – 1:07:00
Metformin as a Geroprotective Drug? The Bannister Study
Attia revisits the influential 2014 Bannister paper suggesting that diabetics on metformin outlive non‑diabetic controls, which fueled enthusiasm for metformin as an anti‑aging drug. He explains the study design and the key limitation of informative censoring, comparing it to a biased smoker vs non‑smoker survival analysis.
- •Bannister et al. used UK registry data to compare type 2 diabetics on metformin only vs matched non‑diabetics.
- •They reported lower all‑cause mortality in metformin users (hazard ratio 0.85, ~15% reduction).
- •However, the analysis censored (excluded) diabetics who stopped metformin, were lost to follow‑up, or needed stronger drugs—often sicker individuals.
- •Attia likens this to stopping counting smokers as soon as they die, underestimating smoking’s harm.
- •Result implied metformin was providing benefits beyond glucose lowering (e.g., mTOR inhibition, reduced inflammation), boosting geroprotection hype.
- 1:07:00 – 1:21:00
The Keys Danish Study: Twin Design and Metformin Reassessment
Attia presents the newer study by Keys et al. using ~500,000 Danes and both matched singletons and discordant twins to reassess metformin’s survival effects. He explains Table 1 (baseline characteristics), the massive differences in co‑medications, and the fundamental challenge of correcting for all confounders in observational data.
- •Two cohorts: (1) matched unrelated individuals with vs without diabetes on metformin; (2) same‑sex discordant twins (one diabetic on metformin, one non‑diabetic).
- •Baseline factors like age, sex, and BMI were well matched; medications were not—diabetics used far more lipid‑lowering, antihypertensive, antiplatelet, and other drugs.
- •Medication patterns reflect higher underlying disease burden and are difficult to fully adjust for statistically.
- •Epidemiology’s strength is huge sample size and long follow‑up; its weakness is no randomization, so residual confounding is inevitable.
- •They explicitly examine informative censoring, repeating analyses with and without excluding progressing metformin users.
- 1:21:00 – 1:39:00
Survival Curves, Hazard Ratios, and What the Data Really Show
They walk through crude death rates per 1,000 person‑years, Kaplan–Meier survival curves, and hazard ratios from the Keys paper. Both matched and twin analyses show clearly higher mortality in diabetics on metformin than in non‑diabetic controls, even after adjustment and when mimicking Bannister’s censoring strategy.
- •Crude mortality rates are substantially higher in diabetics on metformin vs non‑diabetics (e.g., ~25 vs ~17 deaths per 1,000 person‑years in singletons).
- •Kaplan–Meier curves show survival in metformin‑treated diabetics consistently below that of matched controls; confidence‑interval bands do not overlap meaningfully.
- •Unadjusted hazard ratios: ~1.48 (48% higher mortality) in singletons, ~2.15 (115% higher) in twins.
- •After adjusting for medications and education, HRs remain elevated (~1.3–1.8).
- •Applying informative censoring (like Bannister) attenuates but does not reverse the excess risk, suggesting Bannister’s survival advantage was likely an artifact.
- •Conclusion: metformin does not erase the survival disadvantage of diabetes relative to non‑diabetics; whether it helps *within* diabetics vs no treatment remains plausible but unproven here.
- 1:39:00 – 1:51:00
Attia’s Personal Metformin Use, Berberine, and Performance Tradeoffs
Attia shares his own history of taking metformin from 2011 to ~2018 for presumed geroprotection and why he stopped: elevated baseline lactate and impaired exercise signaling. Huberman describes experimenting with berberine to buffer ‘cheat day’ glucose spikes and experiencing probable hypoglycemia. They touch on acarbose and ITP mouse longevity data.
- •Attia self‑started metformin in 2011, initially at a full 2 g/day without titration, experiencing significant GI side effects (nausea).
- •Once he began lactate‑based zone 2 training, he noticed fasted resting lactate ~1.6 mmol/L instead of <1.0 mmol/L, consistent with mild mitochondrial blockade.
- •He stopped metformin largely due to these metabolic effects and emerging evidence it may blunt hypertrophy/strength gains.
- •Huberman used berberine (a plant‑derived metformin‑like compound) to tolerate massive ‘cheat day’ carb loads, feeling fine if he ate a lot but hypoglycemic and headachy if he didn’t.
- •Acarbose, another diabetes drug that blocks carb absorption, extends lifespan in ITP mice without reducing their body weight, implying benefits via lower glucose and insulin, not pure calorie restriction.
- •In ITP, metformin *did not* robustly extend lifespan, whereas rapamycin, 17‑alpha‑estradiol (in male mice), and SGLT2 inhibitors (e.g., canagliflozin) did.
- 1:51:00 – 2:04:00
Caloric Restriction, Fasting, and the Biomarker Problem in Aging
They discuss whether short or periodic caloric restriction and prolonged fasts confer human longevity benefits. Attia notes his past regimen of repeated 3–7 day water fasts and the grim reality of muscle loss. Both emphasize that without validated biomarkers of aging, it’s nearly impossible to know whether such interventions are truly geroprotective.
- •Severe chronic caloric restriction is unlikely to be practical for most people and may not clearly extend human life even if tolerable.
- •Short fasts (e.g., 1–2 days) may improve metabolic flexibility, insulin sensitivity, and discipline but have unproven lifespan effects.
- •Attia formerly did 7‑day fasts quarterly plus 3‑day fasts monthly, now doubts net benefit due to accumulated muscle loss.
- •Major gap in geroscience: lack of reliable, actionable biomarkers of aging (beyond crude epigenetic clocks), unlike clear markers for sleep, muscle mass, or glucose.
- •They argue that solving the biomarker problem is a top research priority; without it, most anti‑aging drug and fasting protocols are not meaningfully optimizable.
- 2:04:00 – 2:18:00
How to Read a Scientific Paper Like a Scientist
Responding to the metformin discussion, Huberman outlines his four‑question framework for reading papers, and Attia adds his own workflow. They emphasize iterative reading, figure‑first strategies, checking supplemental data, and understanding power and statistics rather than relying on abstracts or media summaries.
- •Huberman’s framework: (1) What question are they asking (general and specific)? (2) What is the approach/method? (3) What did they actually find (figures, tables)? (4) Do the conclusions logically follow from the data?
- •He often reads the title and abstract, then scans figures and captions, then returns to methods and results for detail, sometimes needing to look up techniques (PCR, fMRI, Cox models).
- •Attia often goes figures‑first when familiar with a field, skipping intro/discussion; when not, he reads more linearly.
- •Supplemental figures and tables often contain key analyses (e.g., sensitivity analyses, subgroup analyses) not in the main text.
- •Bar graph ‘error bar overlap’ rules of thumb can be misleading; real inference depends on study power, effect size assumptions, and appropriate statistics.
- •Both highlight sample size and power calculations as critical to interpreting null results (no significance may mean underpowered, not ‘no effect’).
- 2:18:00 – 2:30:00
Belief Effects vs Placebo: Stress, Milkshakes, and Hotel Maids
Huberman transitions to the second paper, introducing Alia Crum’s concept of ‘belief effects’—where rich, contextual information shapes physiology and behavior beyond binary placebo. He summarizes experiments on stress mindsets, milkshake labeling, and hotel workers’ exercise beliefs as groundwork for the nicotine fMRI study.
- •Placebo is typically binary (drug vs sugar pill); belief effects involve richer, graded information that reshapes expectations and responses.
- •Stress is bad vs stress is enhancing videos change performance on cognitive and motor tasks by roughly 10–30% in opposite directions.
- •Labeling identical milkshakes as ‘indulgent’ vs ‘diet’ altered ghrelin suppression and satiety, despite identical nutrients.
- •Hotel workers told their daily physical tasks ‘count as exercise’ lost ~12% more weight than controls with similar measured activity.
- •These studies show that beliefs about stress, food, and activity can measurably alter hormones, energy balance, and performance.
- 2:30:00 – 2:41:00
Nicotine Mechanisms: Thalamus, Attention, and Reward
Before diving into the fMRI paper, Huberman recaps nicotine’s neurobiology: how acetylcholine/nicotinic receptors in the thalamus, basal forebrain, and brainstem enhance signal‑to‑noise and focus, and why nicotine can simultaneously increase alertness and bodily relaxation.
- •Nicotinic acetylcholine receptors in the thalamus improve sensory signal‑to‑noise: they increase fidelity of incoming signals (vision, audition) relative to background noise.
- •Basal forebrain cholinergic neurons and the pedunculopontine nucleus project to thalamus and cortex, modulating attention and arousal.
- •Nicotine also engages the mesolimbic dopamine pathway (reward) and ventromedial prefrontal cortex circuits for sustained attention and learning.
- •Endogenous ligand is acetylcholine; receptors are named ‘nicotinic’ because nicotine was used experimentally to characterize them, not because we’re ‘meant’ to smoke.
- •Peripherally, nicotine relaxes skeletal muscle and can feel calming, while centrally it sharpens cognition—a key reason it is so reinforcing.
- 2:41:00 – 2:53:00
The Nicotine Belief fMRI Study: Design and Task
Huberman outlines the Gu et al. experimental design: experienced smokers abstain from nicotine, then vape what they are told is low, medium, or high nicotine while actually all receiving the same low dose. They then perform a financial prediction task in the fMRI scanner designed to engage thalamic and reward circuitry.
- •Participants are regular smokers, washed out for ~2 days to clear nicotine; CO and blood nicotine levels are measured to verify abstinence and dose.
- •Everyone receives the same low‑dose nicotine vape, but instructions differ: ‘low,’ ‘medium,’ or ‘high’ nicotine strength.
- •A financial task (predicting stock/price movements and deciding to go long or short) incentivizes accurate performance with real monetary outcomes.
- •fMRI measures the BOLD (blood oxygen level‑dependent) signal as a proxy for neural activity, focusing on thalamus, ventromedial PFC, and reward regions.
- •They control for possible confounds like reaction time differences and sensory/visual processing that might explain brain activation differences.
- 2:53:00 – 3:07:00
Results: Dose-Dependent Belief Effects in Thalamic–Prefrontal Circuits
Huberman walks through the key figures: subjective ratings map onto instructed dose, thalamic activation trends by belief, and, most strikingly, thalamus–ventromedial prefrontal connectivity scales cleanly with believed nicotine strength. Reward‑area activation does not differ significantly across belief conditions.
- •Participants’ subjective feeling of ‘how much nicotine’ they got tracks exactly with what they were told, despite identical blood nicotine levels.
- •Within the thalamus, belief strength shows a modest correlation with activation; low statistical power likely blunts some effects.
- •Connectivity from thalamus to ventromedial prefrontal cortex shows a clear graded effect: low < medium < high belief groups, each significantly different.
- •Belief about dose *does not* significantly modulate activity in classic reward regions (mesolimbic dopamine pathway).
- •This implies that expectations preferentially modulate attentional and learning circuits rather than raw reward signaling, at least under these conditions.
- •Huberman notes a missing but intriguing control: giving zero nicotine but telling people they got a high dose, analogous to the classic ‘non‑alcoholic beer’ prank.
- 3:07:00 – 3:21:00
Implications: ADHD Drugs, Tapering, Blood Pressure, and Beyond
They extrapolate the nicotine findings to other domains: ADHD medications, smoking cessation, and conditions like hypertension where central processes influence peripheral physiology. Huberman suggests that belief‑shaped brain responses could make lower doses of some drugs more effective than expected, if framed correctly.
- •For ADHD, a child who has benefited from stimulants might maintain circuit‑level gains on a reduced dose if they believe it is the same or equally potent—potentially lowering side effects.
- •For nicotine cessation, gradual dose tapering while maintaining the belief of a stable dose might moderate withdrawal by stabilizing central circuit responses.
- •Attia notes that beliefs could conceivably influence conditions like essential hypertension, where central neural mechanisms play a role but are poorly understood.
- •Framing side effects as evidence a drug is working (from Crum’s work) can reduce their perceived aversiveness and increase adherence and perceived benefit.
- •Nocebo effects (negative expectations causing side effects) can make some people experience nearly every listed symptom for a new medication.
- •Ethically aligning truthful framing with beneficial beliefs is a nuanced clinical communication challenge.
- 3:21:00
Wrap-Up: Why This Journal Club Matters
Huberman and Attia close by reiterating the value of rigorous, transparent paper dissection for anyone consuming health science. Huberman thanks Attia, invites feedback and questions, mentions newsletter and supplement partners, and encourages continued engagement with science.
- •The metformin discussion illustrates how a single famous observational paper can mislead practice if its limitations aren’t appreciated.
- •The nicotine fMRI study shows that beliefs can shape dose–response relationships at the neural level, broadening the concept of placebo.
- •Critical reading habits (figures, tables, methods, supplemental data) are accessible to non‑scientists and can vastly improve health decision‑making.
- •Huberman plugs his Neural Network newsletter for free protocols and summaries, and notes that he posts distinct content on social media.
- •They signal an intention to continue these joint journal club episodes in the future.
