Metformin

Metformin is the most prescribed type 2 diabetes drug in the world and the cheapest, oldest, best-characterised insulin sensitiser we have. It lowers blood glucose, brings down fasting insulin, modestly improves lipids, and produces a small but durable weight loss in people with insulin resistance. Off-label, it's also the centrepiece of the geroscience hypothesis that drugs targeting metabolism can slow aging itself.

Most people reading this are here for one of three reasons. Either fasting glucose, HbA1c, or fasting insulin is creeping up and you want to bend the curve before it becomes diabetes. Or you have PCOS, insulin resistance, or perimenopausal central weight gain, and you want a tool that actually addresses the underlying driver. Or you're healthy and curious whether the longevity case for metformin holds up in someone who already trains and eats well. The honest answer is that the metabolic case is rock-solid, the PCOS case is well-supported, and the healthy-person longevity case is much weaker than it was a few years ago, mostly because metformin blunts the very exercise adaptations longevity-minded people rely on. More on that below.

Deep-dive

Metformin is a biguanide. At physiological pH it carries a positive charge, which lets it accumulate inside mitochondria where it inhibits respiratory chain Complex I. Slowing Complex I drops cellular ATP and raises AMP and ADP, which activates AMP-activated protein kinase (AMPK), the master cellular energy sensor. AMPK then does the things you want a metabolic drug to do: it shuts down hepatic gluconeogenesis (your liver stops making new glucose), increases glucose uptake in muscle, ramps up fat oxidation, and inhibits lipid and protein synthesis. The first-line glucose-lowering effect comes mostly from the liver arm of this, with insulin sensitisation in muscle and adipose adding a second layer over weeks to months.

The AMPK story is the clean textbook version. The honest picture is messier. Foretz and colleagues showed metformin can suppress hepatic glucose output even in mice with the AMPK pathway knocked out, which means at least some of the glucose-lowering happens through AMPK-independent routes. Candidates include direct inhibition of mitochondrial glycerophosphate dehydrogenase, a lysosomal pathway, and a redox-based mechanism. The current consensus is that metformin works through multiple overlapping mechanisms rather than one clean lever, with the relative importance shifting depending on dose, tissue, and metabolic state. This matters because every time someone tells you metformin definitely does X via mechanism Y, the literature is usually more divided than they're letting on.

The gut may also be doing more work than the canonical liver story credits. Metformin concentrates in the intestine at far higher levels than in plasma, reshapes the microbiome (increased Akkermansia muciniphila is the most replicated finding), increases GLP-1 and PYY secretion, and reduces bile acid reabsorption. The delayed-release formulation, which keeps metformin in the gut and barely enters circulation, still lowers glucose meaningfully, which is hard to explain unless the gut effects are doing real work.

Glycaemic control and diabetes prevention. This is what metformin is licensed for and where the data is strongest. In type 2 diabetics it drops HbA1c by 1-2 percentage points and fasting glucose by 2-4 mmol/L at standard doses. In prediabetes, the Diabetes Prevention Program randomised 3,234 people with impaired glucose tolerance to lifestyle intervention, metformin 850 mg twice daily, or placebo, and over 2.8 years metformin reduced the risk of progressing to type 2 diabetes by 31% versus placebo, with intensive lifestyle hitting 58%. The effect persisted: at 15 years of follow-up, metformin still reduced diabetes risk by 18%, and the people who actually stuck with it lost and kept off around 6% of body weight, which is durable real-world weight loss from a drug that nobody markets as a weight-loss drug.

Cardiovascular outcomes. UKPDS 34 randomised 753 overweight newly-diagnosed type 2 diabetics to metformin or conventional diet, and after a median 10.7 years the metformin arm had a 36% lower all-cause mortality and a 39% lower rate of myocardial infarction. The 10-year post-trial follow-up showed the cardiovascular benefit persisted even after glycaemic differences disappeared. UKPDS is the strongest single dataset for metformin's hard-outcome benefit, but it's also old, single-trial, and limited to overweight diabetics. The DPPOS 21-year follow-up of people with prediabetes found no significant reduction in major cardiovascular events with either metformin or lifestyle versus placebo, which has tempered enthusiasm about cardioprotection in non-diabetic populations.

Cancer. Diabetics have higher rates of several cancers and metformin users have consistently lower cancer risk than diabetics on other drugs. A meta-analysis of 21 studies found a 33% reduction in cancer mortality and similar reductions in colorectal, hepatocellular, and lung cancer incidence among metformin users versus non-users. The signal is real but almost all observational, with healthy-user bias and time-related bias confounding the picture. Randomised trials in non-diabetic populations have not yet shown that metformin prevents cancer. Treat the cancer story as a plausible bonus, not as a reason to start.

Longevity and the TAME trial. Animal models show lifespan extension in worms, flies, and some mouse strains. The human longevity case rests mostly on observational data: diabetics on metformin in one UK cohort actually had slightly lower mortality than matched non-diabetic controls, which prompted the geroscience hypothesis that the drug is doing something beyond glucose control. The TAME (Targeting Aging with Metformin) trial is the proposed RCT designed to test this in 3,000 older adults, but funding has been slow and the trial isn't completed, so the longevity case is still pre-data in healthy humans.

The exercise interference problem. This is the most important update for anyone using metformin off-label for longevity. Metformin inhibits Complex I, and aerobic training adaptations rely heavily on mitochondrial biogenesis and Complex I function. Konopka and colleagues randomised 53 older adults to metformin or placebo during 12 weeks of aerobic training and found metformin cut the VO2max improvement by about 50% and completely abolished the gain in muscle mitochondrial respiration. The MASTERS trial ran the resistance training version: 94 older adults on 14 weeks of progressive resistance training, with or without 1,700 mg metformin. The placebo group gained significantly more lean body mass and thigh muscle mass than the metformin group, with a trend toward blunted strength gains. The mechanism appears to be metformin-induced AMPK activation blunting the mTOR signal that drives hypertrophy, and mitochondrial inhibition blunting aerobic adaptation. A 2022 trial in people with metabolic syndrome confirmed metformin cut the VO2max improvement from 16 weeks of HIIT by 50% without affecting the insulin sensitivity gains. The takeaway: if you're training hard and your training is the primary lever you're pulling for healthspan, metformin can work against you. If you're sedentary and metabolically unwell, metformin is still pulling you in the right direction. The intersection of healthy-trainer-on-metformin is where the longevity case looks weakest.

Women. Most metformin data is mixed-sex but a few patterns matter. Women have lower body weight on average and the fixed adult dose may produce higher exposure, which a 2025 review suggests may partly explain the higher rate of side effects and discontinuation in women. In PCOS, metformin lowers testosterone by roughly 20-25%, restores ovulatory cycles in around two-thirds of women, and improves clinical pregnancy rates when combined with clomiphene. A meta-analysis found metformin alone gave lower live-birth rates than clomiphene citrate as first-line ovulation induction, but the combination outperforms either alone. The mechanism is downstream of insulin: lower insulin reduces ovarian androgen production and restores normal pulsatile GnRH. Berberine, inositol, and metformin all work on this same axis from different angles. For perimenopause and the central weight gain that comes with falling oestrogen, metformin has reasonable but not overwhelming evidence. It improves insulin sensitivity and modestly reduces visceral fat in this population. Pregnancy: metformin is widely used in PCOS to support fertility and is generally considered safe in the first trimester, with recent meta-analyses finding no increase in congenital abnormalities, but ongoing use through pregnancy is a decision to make with your obstetrician, not from a supplement blog.

Limitations of the evidence. Despite 60+ years of clinical use, metformin's exact mechanism is still debated. Most of the longevity data is observational with significant healthy-user bias, and the TAME trial that would test the longevity case in non-diabetics is incomplete. Cancer data is observational. Exercise interference data is consistent but mostly in older adults and at therapeutic doses; whether low-dose metformin in younger trained people produces the same interference is less well-studied but the mechanism gives no reason to expect a clean threshold.

Dosage:

Standard starting dose: 500mg once daily with the largest meal of the day for 1-2 weeks, then 500mg twice daily, then titrate to 1500-2000mg/day in divided doses if needed. The clinical target for diabetes and prediabetes is 1500-2000mg/day, which is what most outcome data is based on

Maximum dose: 2550mg/day for immediate release, 2000mg/day for extended release. Going above 2000mg rarely adds much glycaemic benefit and reliably worsens GI side effects

Off-label longevity/metabolic use in non-diabetics: Often dosed lower, 500-1000mg/day, on the theory that you want enough AMPK activation to shift metabolism without maximally inhibiting Complex I. There's no proper dose-response data in healthy users so this is rational-but-unproven territory

Always with food: Metformin with food significantly reduces GI side effects. Taken on an empty stomach the nausea and diarrhoea rate roughly doubles. The fed state also smooths the post-meal glucose excursion, which is part of the point

Extended release (Glucophage XR, generic) vs immediate release: Both work equivalently for glucose control. A retrospective cohort found switching from IR to XR cut GI side effects from 26% to 12% on equivalent doses. If you're getting GI issues on IR, switch to XR before giving up. XR is dosed once daily with the evening meal

Women: A 2025 review noted women on standard adult doses may be over-exposed relative to body weight and have higher discontinuation due to side effects. If you're a smaller-framed woman, starting at 500mg daily and staying there if it works is reasonable. There's no formal dose adjustment but the principle of lowest effective dose applies harder

If you train seriously: Take metformin on rest days or as far from your training window as possible. The exercise interference data is on chronic continuous use, but separating peak metformin levels from peak training stimulus is the most defensible compromise. If your primary goal is mitochondrial adaptation and strength, consider whether metformin is the right tool at all.
Berberine hits AMPK through the same mechanism without the same depth of mitochondrial inhibition at typical doses and may be a better fit

Cycling: No established cycling protocol. Most clinical use is continuous. Off-label longevity protocols sometimes cycle (5 days on, 2 days off, or 3 weeks on, 1 week off) on the unproven theory that intermittent dosing preserves training adaptations while keeping the metabolic benefit. Reasonable hypothesis, no data

Here's what you can expect:

In the first 1-2 weeks, expect some GI noise: loose stools, mild nausea, occasional cramping, a metallic taste. This is the most common reason people quit, and it almost always settles within 2-4 weeks if you start low and titrate slowly. Switching to extended release fixes it for most people who can't tolerate IR.

Fasting glucose typically drops 0.5-1.5 mmol/L within 2-4 weeks. HbA1c moves more slowly because it's a 3-month rolling average, so meaningful changes show up at 8-12 weeks. Fasting insulin and HOMA-IR usually improve in parallel with glucose. If you're prediabetic or have insulin resistance, this is where the biggest visible win sits.

Weight loss is real but modest. In the DPP, metformin produced about 2-3 kg of weight loss over the first year, mostly through mild appetite suppression mediated by GDF15, a hormone metformin upregulates that acts on hindbrain appetite circuits. Don't expect GLP-1 levels of weight loss. If you're chasing weight loss specifically,

Retatrutide or other GLP-1/GIP agonists are an order of magnitude more effective.

Lipid changes are modest. Triglycerides drop a little, LDL is roughly neutral or slightly down, HDL barely moves. If lipids are your main concern, this is the wrong tool.

Subjectively, most people on metformin notice very little day-to-day. It's not a stimulant, it's not a mood drug, it doesn't give you energy. It quietly changes your numbers. The exception is the GI side effects in the first weeks, and the occasional brain fog or fatigue, which is worth knowing about because some of that can be B12 depletion creeping in over months.

Side effects & risks:

GI side effects are the dominant issue. In the DPP, gastrointestinal symptoms occurred at 77.8 events per 100 person-years on metformin versus 30.7 on placebo. Diarrhoea, nausea, bloating, metallic taste, and loose stools. Starting low, titrating slowly, taking with food, and switching to extended release if needed addresses this for most people. About 5-10% of users genuinely cannot tolerate it at any dose

Vitamin B12 deficiency is the most clinically meaningful long-term issue. Metformin interferes with calcium-dependent B12 absorption in the ileum. Prevalence of B12 deficiency in long-term metformin users runs around 6-30% depending on duration and dose, with one Pakistani case-control study finding 31% deficiency in metformin users versus 8.6% in controls and B12 levels dropping further the longer you're on it. Risk goes up with dose, duration past 4 years, and age. B12 deficiency causes the same neurological symptoms as diabetic neuropathy (numbness, tingling, balance issues) which often gets misattributed to the underlying diabetes. Check B12 yearly and supplement if low or trending down.
Vitamin B12 covers supplementation

Exercise adaptation interference as covered above. Blunts VO2max gains from aerobic training by roughly 50% in older adults, blunts hypertrophy and lean mass gains from resistance training. This is a feature of the mechanism, not a bug that can be dosed around. If maximising training response is a primary goal, this is a real trade-off

Lactic acidosis (MALA) is the historically feared but practically rare complication. Incidence is around 3 per 100,000 person-years, similar to sulfonylureas. Mortality when it does happen is significant (around 30%). Nearly all cases occur in people with acute kidney injury, sepsis, severe heart failure, alcohol intoxication, or hypoxia, where metformin accumulates and oxidative phosphorylation is already compromised. Stop metformin if you have an acute illness with vomiting, dehydration, or any AKI risk. Stop it 48 hours before contrast-enhanced imaging or major surgery

Renal function sets the safety boundary. Current FDA guidance allows metformin down to eGFR 30 mL/min/1.73 m2, with dose reduction below 45. Below 30 it's contraindicated. Check eGFR before starting and at least annually, more often if you're over 65 or have any kidney history

Hypoglycaemia is rare on metformin alone because it doesn't push insulin secretion. Risk rises if combined with sulfonylureas, insulin, or in lean people with already-good insulin sensitivity. If you're metabolically healthy and lean, you may notice some symptoms of low glucose (shakiness, mild brain fog) at standard doses

Alcohol stacks with metformin to increase lactic acidosis risk, particularly with binge drinking. Light-to-moderate alcohol is generally fine; heavy or binge drinking on metformin is a hard no

Drug interactions are less of an issue than with many drugs. The main ones are: contrast dye (hold for 48 hours peri-procedure), other glucose-lowering drugs (additive hypoglycaemia risk), and cimetidine (reduces metformin clearance, watch for accumulation)

Pregnancy: Widely used in PCOS through pregnancy and considered low-risk based on current data, but this is a decision for an obstetrician, not a self-prescribed call

Hepatic and cardiac function: Severe liver disease or decompensated heart failure are contraindications because both compromise lactate clearance and oxidative phosphorylation

🩸

Blood markers

Fasting glucose, HbA1c, fasting insulin, HOMA-IR, baseline before starting and at 12 weeks. These are your efficacy markers. HOMA-IR is the most sensitive early signal if you're using metformin in a non-diabetic insulin-resistance context.

Vitamin B12 and methylmalonic acid (MMA), baseline and yearly. Serum B12 misses a chunk of functional deficiency, especially as you age, so MMA (or homocysteine if you can't access MMA) catches early deficiency before serum B12 drops. This is the single most important long-term check on metformin.

eGFR and creatinine, baseline and annually, more often if you're over 65 or have kidney history. Metformin dosing depends on this.

ALT, AST, baseline and yearly. Not because metformin damages the liver but because severe liver dysfunction changes the lactate-handling math.

Full lipid panel, baseline and 12 weeks. Don't expect big changes but you want a reference.

Total and free testosterone, SHBG, LH/FSH if using for PCOS, baseline and at 12 weeks. The androgen drop is the marker that the underlying insulin-driven hyperandrogenism is responding.

For most people, the essential set is fasting glucose/HbA1c/fasting insulin, eGFR, and B12 (with MMA) at baseline and annually thereafter. Everything else is optional based on your context.

Metformin is a prescription medication. Where you obtain it is your business; how you use it is the part this page is trying to help with.