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

The episode is staged as the first collaborative journal club between Andrew Huberman and Peter Attia, and Huberman spends the opening minutes explaining the format rather than the findings. A journal club, he says, is a routine of graduate and medical training in which students gather around one or two papers, critique them, and compare their own conclusions with those of the authors. The operative question is adversarial but not hostile: "Does this paper really show what it claims to show or not?" The episode is meant to make that question visible in real time, including the parts where the two hosts disagree.

Attia supplies the motivation from the other side of the table. He used to run a monthly journal club inside his practice, one person choosing a paper and presenting it formally, and the practice let it lapse as people became stretched thin. He calls the activity "a skill" and treats the common question, whether reading the abstract is enough, as a sign that the skill is rarely taught. His point is that the how-to is harder than the injunction to read critically, which is why the two papers were chosen to sit at opposite ends of a spectrum: one simple and epidemiological, one dependent on the technical details of assays and their limits.

The introduction promises two papers. The first is Attia's metformin paper; the second is Huberman's paper on the placebo effect, which he says shows that placebo responses can follow a dose-response curve, with the nicotine literature as the example. The transcribed discussion that follows, however, stays inside the metformin half. The placebo paper is set up and never delivered here, so the actual journal club on offer is a single extended exercise in reading an observational study: its baseline table, its censoring rule, its person-year arithmetic, its survival curves, and its statistical power.

Attia defines the drug before defending or attacking it. Metformin has been used for roughly fifty years in Europe and forty-plus in the United States as a first-line agent for type 2 diabetes; it was once sold as Glucophage and is now generic. Its mechanism is debated, but Attia separates the debated part from what he considers settled: metformin inhibits complex one of the mitochondrial electron transport chain. The inhibition is partial, "a weak blocker of that", not the total blockade produced by cyanide, which he invokes to show how primal the pathway is. The immediate consequence is a shift in the ratio of adenosine monophosphate to adenosine diphosphate.

Huberman offers a simplification and Attia accepts it with one correction. The simplification is that food is broken down, the breaking of bonds releases energy that cells capture as ATP, mitochondria sit at the center of that conversion, and metformin partially short-circuits the process, so less of what is eaten becomes blood glucose. Attia agrees that the drug changes fuel handling but rejects the implication that calories are being discarded. The better description, he says, is "changing the way the body partitions fuel." The distinction matters because it rules out the fantasy of a drug that lets you eat without absorbing energy, the same fantasy that briefly attached to the fat substitute Olestra before its gastrointestinal consequences ended the experiment.

The clinically relevant action is on the liver. Attia gives a number that makes the scale concrete: a person with normal blood sugar has "five grams of glucose in our total circulation", and the brain would consume that within minutes. Survival therefore depends on the liver titrating glucose output continuously. When the liver releases too much, the difference between metabolic health and frank type 2 diabetes can be as small as one teaspoon of glucose in the bloodstream. Metformin's unambiguous effect is to reduce hepatic glucose output, which is why it belongs to the first-line toolkit even though its deeper mechanism remains contested.

Attia places metformin alongside other glucose-lowering strategies rather than treating diabetes drugs as a single category. Sulfonylureas tell the body to make more insulin; exogenous insulin supplies the hormone itself; GLP-1 agonists, which Huberman identifies with Ozempic, increase insulin sensitivity while initially also increasing insulin secretion; metformin works downstream on hepatic output. Huberman asks whether berberine is "the poor man's metformin" and Attia answers yes, adding that metformin itself derives from a substance found in a lilac plant in France. The exchange is a reminder that mechanism and provenance do not establish geroprotection, which is the question the rest of the conversation is built to test.

The pathophysiology is told through muscle because muscle disposes of most glucose, roughly "75% to 80%" of it. Insulin binds its receptor and triggers a tyrosine kinase cascade that ends with a transporter inserted into the cell membrane. Attia's image is a straw pushed up through the membrane: glucose is more concentrated outside the cell than inside, so it flows down its gradient without a pump, but the cell must spend energy to place the transporter. Huberman, who studies ion channels, says he had never heard the process described that way and calls the on-demand insertion of a portal in minutes a striking piece of cell biology.

Insulin resistance begins when intramuscular fat, meaning triacylglycerides and diacylglycerides accumulating inside the muscle cell, interrupts that signal. Attia deliberately avoids precise stoichiometry here, saying "I'm making these numbers up" before offering the illustration that a process once needing two units of insulin now needs three, then four, then five. That escalating requirement is, in his words, "That is the definition of insulin resistance." The earliest marker is therefore not high glucose but high insulin: normal glycemia with hyperinsulinemia, especially after a meal, is "the canary in the coal mine" and can precede a real problem by five to ten years.

Once the system progresses, the liver becomes insulin resistant too and overproduces glucose while muscle cannot dispose of it. Fat then accumulates in the pancreas, inflaming the beta cells that must produce more insulin, so the organ asked to compensate becomes less capable of compensating. Attia's causal chain ends where clinical medicine meets it: a person with long-standing type 2 diabetes may eventually need exogenous insulin. The chain also explains why metformin's suppression of hepatic output addresses only one link, and why the drug can look effective on glucose while remaining silent on the underlying insulin resistance.

The risk factors Attia lists are not exotic. Energy imbalance is enormous, but inactivity may be the single most important. Gerald Shulman's mechanistic trials at Yale recruited lean but inactive undergraduates precisely because the muscle biopsies had to isolate inactivity from obesity. Sleep deprivation is another: restricting people to four hours a night for a week reduces glucose disposal by about half, which Attia calls a staggering induction of insulin resistance. Hypercortisolemia and the spillover of fat from subcutaneous stores into muscle, liver, and pancreas complete the list. The practical implication is that metformin is being asked to compensate for physiology that exercise, sleep, and energy balance also shape.

Attia frames the newer paper against a 2014 study that, in his account, created the modern enthusiasm for metformin as a geroprotective agent. Bannister and colleagues drew on a UK registry, selected patients with type 2 diabetes who were on metformin alone, and matched them to people without diabetes on age, sex, socioeconomic status, blood pressure, BMI, and other available variables. The design is retrospective and observational, the standard substitute when randomization is unavailable. The outcome was all-cause mortality, which Attia calls the gold standard for a study of this kind because the goal is to delay death from anything, not to move a single biomarker.

The result was striking in the direction that made metformin famous. The crude death rate was 14.4 per thousand person-years in the metformin-treated diabetes group and 15.2 in the matched controls, and the Cox model reported a 15% relative reduction in all-cause death over 2.8 years. Attia remembers reading it in 2014 and thinking, "Holy crap. This is really amazing." The surprise is not that metformin beat a placebo. It is that people with diabetes, a condition that on average shortens life by six years, appeared to be dying at a rate indistinguishable from, or slightly better than, people without diabetes.

That inversion is the interpretive crux. If metformin only lowered glucose, the metformin group should still look worse than the non-diabetic controls, because their glucose remained higher. Attia states the logic plainly: "there's no clear explanation for it unless you believe that metformin is doing something beyond helping you lower blood glucose." The candidate off-target effects he lists are a weak inhibition of mTOR, reduced inflammation, and possible suppression of senescent cells and their secretory products. None of these is proven to mediate the survival difference; they are plausible mechanisms that make the epidemiological signal feel biologically coherent.

Attia's own relationship to the drug is part of the evidence he offers, and it cuts against his earlier confidence. He began taking metformin in May 2011 after reading about it, skipped dose titration, and went straight to "two grams a day, which is the full dose." He spent a trip to the Berkshire Hathaway shareholder meeting nauseated and unable to eat ice cream at Dairy Queen. The episode is funny, but the detail matters: the drug's gastrointestinal side effects are real, initial appetite suppression is real, and his early enthusiasm was self-experimentation rather than a verdict. By the time of this conversation he describes himself as undecided on geroprotection.