The Metformin Debate

#1

Recently a paper came out: (stricly dated Sept 2025)

https://www.sciencedirect.com/science/article/pii/S1568163725001631

and an article was published interviewing probably Metformin’s greatest supporter

There was a row at ARDD about metformin and I think it was involving an author of the review and Nir Barzalai

Personally I don’t think Metformin is warranted for people without T2. However, I thought I would ask GPT5 to review both the review and the article simultaneously and see what happens.

Here’s a tight, side-by-side review of the two pieces you shared—one a peer-reviewed umbrella review in Ageing Research Reviews and the other a journalistic interview with Nir Barzilai at Lifespan.io.

1) Keys et al., 2025 — Emerging uncertainty on the anti-aging potential of metformin (Ageing Research Reviews)

What it is (1-paragraph summary).
A critical review arguing that the once-strong narrative around metformin as a gerotherapeutic now looks less certain when you examine replication, confounding, and newer trial evidence. It revisits the most influential epidemiology (e.g., the UK CPRD analysis that suggested metformin users outlived matched nondiabetics) and contrasts it with newer Danish register analyses and other follow-ups that fail to reproduce a survival advantage over people without diabetes. It also surveys RCTs and trials in non-diabetics and pre-diabetics and concludes that, as of 2025, the totality of evidence supports “emerging uncertainty” rather than confidence in anti-aging benefits outside diabetes care. (PubMed)

What’s novel.

  • A high-level, methodologically focused synthesis by a cross-disciplinary author team (clin. pharmacoepidemiology + geroscience) that centers on replication attempts and design flaws in the canonical “metformin > general population” survival claim. The paper explicitly frames how newer evidence reverses or neutralizes earlier observational findings and lays out why popular pro-longevity inferences were premature. (PubMed)
  • It also spotlights trials in non-diabetics (prevention/ancillary indications) as a separate evidentiary bucket from glycemic management, and uses those results to temper the gerotherapeutic narrative. (PubMed)

Strengths (critique).

  • Persuasive emphasis on study design: immortal-time bias, comparator choice, and changing exposure after trial closeout (e.g., DPP follow-up) are explained clearly. (PubMed)
  • Corrective context on the famous CPRD study (Bannister 2014) by pointing to register-based work that didn’t reproduce a survival advantage vs. nondiabetics, including a 2022 Danish re-assessment by the paper’s first author. (PubMed)

Limitations (critique).

  • Inevitably selective: as a corrective, it foregrounds negative/neutral data; some readers will feel it under-weights positive mechanistic and translational signals.
  • COVID-19–related randomized data (e.g., COVID-OUT’s long-COVID result) doesn’t directly test “slowing aging,” so dismissing pro-metformin enthusiasm may also risk over-correction if one equates lack of lifespan signal with lack of geroscience relevance. Still, those trials primarily inform infection/long-COVID, not aging per se. (ncats.nih.gov)

Bottom line from the paper.
When you remove biases and demand replication/RCT evidence in people without diabetes, the case for metformin as a gerotherapeutic is weak to “uncertain.” (PubMed)

2) Lifespan.io interview — “Nir Barzilai: Positive Evidence for Metformin is Mounting” (Aug 12, 2025)

What it is (1-paragraph summary).
An interview-style article presenting Barzilai’s rebuttal to the Ageing Research Reviews piece. He argues the review cherry-picks “bad” studies, that large bodies of work remain positive, and that observational contradictions (e.g., UK vs Denmark) reflect population and design differences. He cites RCT evidence around COVID-19/long-COVID and portrays TAME as a template for multi-outcome gerotherapeutic trials; he also suggests ARPA-H and big-pharma interest in TAME-like designs (e.g., for GLP-1s). (lifespan.io)

What’s novel.

  • Timely counter-narrative that collates several pro-metformin talking points in one place, including: (i) long-COVID risk reduction with early metformin in COVID-OUT; (ii) a claim of better late-life benefits (antagonistic pleiotropy) and an ITP re-analysis showing stronger effects in older animals; (iii) a stated regulatory path where TAME-like endpoints could generalize to other drug classes. These are presented as newsy updates rather than new peer-reviewed analyses. (lifespan.io)

Strengths (critique).

  • Fairly comprehensive articulation of the pro-metformin case, especially regarding breadth of effects (AMPK–mTOR, inflammation, mitochondrial complex I) and pragmatic endpoints (clustered morbidities vs single mortality). (lifespan.io)
  • Correctly notes that some highly cited studies are observational and design choices matter (e.g., post-trial treatment contamination in DPP follow-up). (lifespan.io)

Limitations (critique).

  • It’s an interview, not a systematic review: many claims are assertions without primary citations, e.g., “20-something” mouse lifespan studies, “monkey study in Cell delaying transcriptomic aging by eight years,” and broad statements such as metformin being “first line for macular degeneration” (not standard of care). These would need verification and may overstate the consensus. (lifespan.io)
  • The COVID-OUT data are real but specific: metformin reduced long-COVID incidence (~41%) and some acute outcomes when started early; mortality effects are inconsistent across syntheses, and the primary aging question remains untested. (ncats.nih.gov, PMC)
  • Methodological criticisms of the Danish replication are countered largely with post hoc explanations (e.g., obesity rates, time-to-event bias) rather than direct re-analyses. (lifespan.io)

Bottom line from the article.
Barzilai remains bullish: he frames metformin as a broad gerotherapeutic with mounting evidence and argues the skeptical review is unduly negative or methodologically naïve. But the piece is opinion-forward and lightly referenced. (lifespan.io)

Where the two pieces actually disagree

  • The CPRD survival claim (metformin users outliving matched nondiabetics):
    • Keys et al.: Not reproduced in Danish registers; likely confounded/biased. (PubMed)
    • Barzilai interview: UK vs Denmark ≠ like-for-like; still accepts UK result as supportive. (lifespan.io)
  • Relevance of COVID-19 trials to “aging”:
    • Keys et al.: These don’t address aging endpoints; thus limited weight for geroscience claims. (PubMed)
    • Barzilai: Treats COVID-OUT as strongly supportive of gerotherapeutic action (resilience/immunity). (lifespan.io)
  • Animal data (ITP & primates):
    • Keys et al.: Emphasizes that robust lifespan extension in standard models has been inconsistent; hence caution. (PubMed)
    • Barzilai: Asserts age-restricted benefits (older > younger) and cites a non-specific primate transcriptomic claim; verification needed. (lifespan.io)

My synthesized view (weighing the arguments)

  1. For non-diabetic, generally healthy adults, the balance of evidence remains neutral to skeptical.
    The most influential pro-longevity observational result (CPRD 2014) has not been robustly replicated against nondiabetic comparators in independent datasets; Danish register emulation did not find superior survival vs the general population. That knocks out the strongest simple narrative that “metformin users outlive controls” and aligns with the ARR review’s conclusion of emerging uncertainty. (PubMed)
  2. Metformin has clear clinical value in T2D and signals in specific contexts (e.g., long-COVID prevention when given early), but these do not yet equate to a proven gerotherapeutic effect.
    COVID-OUT is a well-run RCT showing ~41% relative reduction in long-COVID incidence with early metformin, but meta-analyses of mortality in COVID remain mixed—and none of this directly demonstrates slowed aging or delayed multi-morbidities in non-infectious settings. It’s supportive of resilience/host-response effects, not proof of “anti-aging.” (ncats.nih.gov, PMC)
  3. Barzilai’s mechanistic case (AMPK/mTOR/mitochondria, anti-inflammatory actions) is biologically plausible, and some animal and human biomarker results are encouraging. But in 2025, mechanistic plausibility + heterogeneous observational wins ≠ clinical proof of geroprotection in non-diabetics. The interview advances hypotheses and programmatic roadmaps (e.g., TAME) rather than new adjudicating data. (lifespan.io)

Net take:

  • If your bar is “should clinicians routinely use metformin to slow aging in non-diabetic adults?” — not yet, on evidence grounds. The ARR review’s methodological cautions carry more weight than an interview-based rebuttal. (PubMed)
  • If your bar is “is metformin worth testing in well-designed aging-endpoint RCTs?” — yes. There’s enough mechanistic and context-specific clinical signal to justify TAME-like trials, but policy or practice shouldn’t jump ahead of outcomes. (lifespan.io)

Citations (key items)

  • Keys MT, Hallas J, Miller RA, Suissa S, Christensen K. Emerging uncertainty on the anti-aging potential of metformin. Ageing Res Rev. 2025;111:102817. (PubMed)
  • Bannister CA et al. Can people with type 2 diabetes live longer than those without? Diabetes Obes Metab. 2014. (UK CPRD.) (PubMed)
  • Keys MT et al. Reassessing the evidence… (Danish registers replication). Int J Epidemiol. 2022. (PubMed)
  • Lifespan.io interview: Nir Barzilai: “Positive Evidence for Metformin is Mounting”, Aug 12, 2025. (lifespan.io)
  • COVID-OUT/NCATS summary of RCT findings for long-COVID risk reduction with early metformin. (ncats.nih.gov)
  • 2025 systematic review/meta-analysis on metformin in COVID-19 (mixed mortality results). (PMC)

If you want, I can pull specific trial tables (population, dose, endpoints, effect sizes) to see exactly where the contradictions arise and what a “TAME-style” primary endpoint would need to show to change the balance.

3 Likes
#2

Screenshot 2025-08-13 at 11.44.38 AM
Screenshot 2025-08-13 at 11.44.38 AM1188×892 173 KB

3 Likes
#3

Hasn’t been a debate for me for awhile. Ther are much better drugs out there that do what some people hope metformin might do. I haven’t seen any data so far that would make me change my mind.

1 Like
#4

I felt that way for a while but some compelling information is coming out lately that I’m liking:

#5

I used to take 500mg XR morning and night. Now I only take it at night, at least six hours after any training and 14 hours before any training.

Morning dose replaced with Canagliflozin.

I’ll see how his program goes and maybe make adjustments.

1 Like
#6

I feel the effects on inflammation alone are worth the trade-offs, especially since I notice no effects on my exercise. I definitely notice the effects on inflammation within a few days.

3 Likes
#7

I started taking it because I found evidence it can help with asthma. It also pairs better with rapamycin than berberine.

I will like berberine but I’ve replaced one of the doses I take with metformin sustained release.

1 Like
#8

I feel like I read a version of this post every 6 months somewhere. This is just the August 2025 edition of metformin as possibly, maybe being good for longevity.

In contrast, I don’t have to squint too hard to see that SGLT2 inhibitors and GLP1 agonists are good for longevity.

2 Likes
#9

I’m certainly more bullish on GLP1 agonists and SGLT2 inhibitors.

#10

See also this thread: What's the Deal with Renewed Interest in Metformin?

1 Like
#11

SGLT inhibitors are a no-brainer:

  • Life extension in mice
  • Very safe
  • Approval for heart failure, T2D, CKD and other age-related conditions (WIP)
  • Mechanistic understanding
  • Lower all-cause mortality in longitudinal studies in sick populations
  • Mendelian randomization

I don’t think any other drugs tick all these boxes.

3 Likes
#12

Any point stacking SGLT2i with GLP1RA?

#13

It seems that the GLP-1RA + SGLT2i combination offers synergies and benefits in T2D and CKD but there’s not enough data to conclude yet. We know it’s safe though.

1 Like
#14

I’m not convinced lifelong GLP-1 administration is a good idea for healthy people at ideal body weight.

I see them more as a temporary tool to make cuts easy.

2 Likes
#15

Plenty of evidence emerging they are beneficial for more than just fat loss imo

#16

Despite their wife benefits, they might still be detrimental in some people such as those with optimal weights. Some GLP-1RAs also cause depression (semaglutide especially). And as far as I know there is no evidence that they extend lifespan in animals. So they’re wonderful drugs for obese and overweight people and people with diabetes or CKD but for others, so far, their benefits are unproven. In the phase 3 exenatide trial in Parkinson’s for instance, people on exenatide were slightly worse off than those on placebo. If Parkinson’s is a “disease of aging” then that’s not the type of result you want to see. That being said I think there is a wide intraclass variability among GLP-1RAs and it’s possible that the newest one (tirzepatide?) are better. Time will tell…

1 Like
#17

I wouldn’t bother with anything before semaglutide, and see no reason to use semaglutide over tirzepatide. I’m looking forward to more studies being done. We likely won’t see a lifespan study done until orally bioavailable small molecule orforglipron is out since people don’t want to inject mice. That, or some kind of autoinjector for mouse studies is invented.

2 Likes
#18

For mechanistic reasons alone I would expect modern GLP1 to have a similar life-extending effect in mice as acarbose.

#19

That’s very interesting, because I was thinking the opposite. I’m probably wrong. I see SGLT2i and acarbose as directly limiting glucose levels, mechanistically, so more on the same plane. And GLP1 hormonal pathway, different plane. But it’s likely I misunderstood.

2 Likes
#20

Both GLP1 and acarbose prevent glucose spikes due to delayed digestion while SGLT2i increase the rate of glucose excretion.