Oliver's mTOR Atlas Evidence Platform
Reading level
Mode

Rapamycin + AMPK-axis drugs (acarbose / metformin): a male-mouse survival signal from cross-cohort comparison, untested in humans

Evidence desert · editorial confidence 70% · evidence stands at: Lifespan in mice (cross-cohort signal, not a factorial test); no human data · Atlas ID H9

The gap

In mice, combining rapamycin with a diabetes drug called acarbose extended life more than either drug alone - but only in males; in females it was no better than rapamycin by itself. A different diabetes drug, metformin, reaches mTORC1 partly through AMPK, the cell's fuel sensor. It therefore presses on the same checkpoint that rapamycin blocks, only from another direction, so combining them might not add much. But be careful with that reasoning: two drugs hitting different targets do not automatically add up - they can still overlap further downstream, or interact in unexpected ways. Nobody in this Atlas's collection has run a human trial combining any of these.

Technical framing: Animal combination data are suggestive but the human evidence for combining mTOR inhibition with AMPK-axis geroprotectors is a desert. QUANTIFIED (STR2022, NIA Interventions Testing Program C2017 cohort): rapamycin + acarbose started at 9 months extended MALE lifespan beyond what either of the two prior ITP rapamycin-only cohorts achieved, while in FEMALES the combination was neither better nor worse than rapamycin alone - the authors relate this to the limited survival benefit acarbose alone had shown in earlier female cohorts. CAVEAT: STR2022's rapamycin-only comparison is HISTORICAL, against prior cohorts rather than a concurrent rapamycin-only arm, so 'more potent than either component' is the authors' suggestion, not a within-experiment randomised result. That is precisely why this gap remains open. HAR2014 established acarbose's male-skewed lifespan effect. The AMPK arm converges back onto mTOR mechanistically: ZHO2001 metformin acts via AMPK, GWI2008 AMPK phosphorylates raptor to gate mTORC1, HOW2017 metformin inhibits hepatic mTORC1 dose-dependently via AMPK - which raises the opposite possibility that metformin + rapamycin are partly REDUNDANT on the mTORC1 node. Human signal exists only for observational data (BAN2014: type-2 diabetics on metformin outlived non-diabetic controls - a result subject to the usual immortal-time and prescription-bias caveats of the metformin-survivorship literature, and not treated here as causal). No study IN THIS CORPUS - animal or human - cleanly separates ADDITIVE (distinct pathways) from REDUNDANT (same node) for these pairs, and this corpus holds no human combination trial. Gaps here are computed against this Atlas, not against PubMed. LOGIC NOTE: distinct proximal mechanisms do not entail additivity. Two drugs can hit different targets and still converge downstream, interact non-linearly, or show sex- and strain-dependent interaction - which is exactly what the ITP combination data themselves show.

What changed since this question was written

The mouse half has not been open since 2016, outside this corpus. The NIA Interventions Testing Program tested metformin with rapamycin (Strong, Aging Cell 2016, PMID 27312235) and rapamycin with acarbose (Strong, Aging Cell 2022, PMID 36179270); the acarbose combination gave male mice longer lifespan than rapamycin alone, with no comparable benefit in females. Both conclusions rest on historical comparison against earlier rapamycin-only cohorts rather than a contemporaneous factorial, so at least additive in males is defensible and synergistic is not. A three-drug combination also beat every single agent on healthspan measures in 20-month-old mice (Jiang, Sci Rep 2022, PMID 35508491). The category itself is also questionable: acarbose is an intestinal alpha-glucosidase inhibitor with a strong male bias, and low-dose metformin reaches AMPK through the lysosomal PEN2 route rather than complex I (Ma, Nature 2022, PMID 35197629), so grouping them as the AMPK axis imports a shared mechanism neither clearly uses at geroprotective doses.

What is still open

The human half, entirely: no trial has tested rapamycin with an AMPK-axis or glucose-lowering drug against either agent alone on any ageing endpoint, and TAME remains unfunded with no rapamycin arm. A more consequential question has appeared alongside it. Metformin abolished the VO2max gain from 16 weeks of training and blunted training-induced vascular insulin sensitivity in 91 adults (Malin, JCEM 2026, PMID 41160096). Since exercise is the best-evidenced human geroprotective intervention, an AMPK-axis drug may subtract from a stack rather than add to it. For contrast, trametinib and rapamycin combine additively for mouse lifespan in both sexes (Gkioni, Nature Aging 2025, PMID 40437307) — combination with dose-lowering may be a more tractable route to benefit without toxicity than engineering complex selectivity.

The hypothesis

In mice there is a signal: in the cohorts that were compared, pairing rapamycin with acarbose did better than rapamycin alone in males, though not in females. But those were separate cohorts run at different times rather than one experiment with both arms side by side, so "they add up" is not something the comparison can show yet. In people nothing has been tried. And there is a warning sign — in a recent human trial metformin cancelled out the fitness gains from exercise training, so adding it to a stack may take away rather than add.

Technical framing: In mice there is a signal, not a settled answer: historical ITP comparisons suggest that rapamycin with acarbose may improve male survival beyond rapamycin alone, with no comparable female benefit in the cohorts compared. That does not establish additivity. Statistical additivity requires an estimand, an interaction scale and appropriate controls, and the evidence here is cross-cohort rather than concurrent factorial, so 'at least additive' is a stronger claim than the design can carry. The female result likewise does not establish a universal absence of combination effects. What is hypothesised and untested is the human case — and the live alternative, raised by the exercise data, is that AMPK-axis drugs antagonise other geroprotective interventions in people rather than complementing them.

How it could be tested

Factorial mouse lifespan/healthspan study (vehicle / rapamycin / acarbose / metformin / rapamycin+acarbose / rapamycin+metformin), sex-stratified, with mTORC1 (S6K/4E-BP1) and AMPK (raptor-Ser792) readouts plus glucose tolerance. State the estimand before the experiment runs: name the survival effect scale on which additivity is defined (for example difference in median survival, or a hazard ratio on the log scale), prespecify the interaction model, and say in advance what would count as super-additive on that scale. Any combination advantage is then reported on that scale, and is not equated with synergy. Follow with a human factorial healthspan-biomarker trial on the additive pair.

Related studies

YearEvidenceStudy
2014 A Acarbose, 17-α-estradiol, and nordihydroguaiaretic acid extend mouse lifespan preferentially in males. HAR2014 Beyond rapamycin, the same ITP program found acarbose extended male median lifespan by 22% - evidence that the registry surfaces real positive hits too, not only negative results.
2001 M Role of AMP-activated protein kinase in mechanism of metformin action ZHO2001 Metformin activates AMPK, suppressing hepatic gluconeogenesis and lipogenesis.
2008 M AMPK phosphorylation of raptor mediates a metabolic checkpoint GWI2008 Found a SECOND way the energy sensor AMPK shuts mTORC1 down. Besides acting through TSC2, AMPK directly phosphorylates Raptor - the core mTORC1 subunit - to halt growth when energy runs low. This 'metabolic checkpoint' is exactly the switch that drugs like metformin and exercise tap into.
2017 M Metformin Inhibits Hepatic mTORC1 Signaling via Dose-Dependent Mechanisms Involving AMPK and the TSC Complex HOW2017 Pinned down HOW the diabetes drug metformin - a major longevity candidate - actually reaches mTOR. In the liver, metformin lowers cellular energy, and at low doses this shuts down mTORC1 specifically through AMPK and the TSC complex. Direct mechanistic bridge between a widely-used drug, energy sensing, and the mTOR pathway.
2014 H Can people with type 2 diabetes live longer than those without? A comparison of mortality in people initiated with metformin or sulphonylurea monotherapy and matched, non-diabetic controls BAN2014 Diabetic patients started on metformin had longer median survival than matched non-diabetic controls without the drug. Retrospective and observational: consistent with a survival benefit, but confounding by indication and healthy-adherer effects cannot be excluded.
2015 A Alternative rapamycin treatment regimens mitigate the impact of rapamycin on glucose homeostasis and the immune system ARR2015 Intermittent rapamycin regimens (weekly, or every 5 days) largely spared glucose tolerance, pyruvate tolerance, fasting glucose and insulin, beta-cell function and the immune system, while still inhibiting mTORC1 -- unlike daily dosing, which impaired all of them. IMPORTANT SCOPE: this study measured side effects only. It did NOT measure lifespan or any other benefit endpoint, so it shows the harm can be reduced, not that the benefit is retained. For evidence that a non-continuous schedule preserves a survival benefit, see ARR2016 (the same every-5-days regimen, started at 20 months, extended female lifespan) and BIT2016.
2025 H Influence of rapamycin on safety and healthspan metrics after one year: PEARL trial results MOE2025 First completed long-term RCT of rapamycin for healthy human aging (NCT04488601, 48 weeks, n=114). Primary endpoint (visceral fat by DXA) showed NO significant change (p=0.942) - a null result on the pre-registered primary endpoint. Secondary endpoints were more promising: women on the 10mg/week dose had significant improvements in lean tissue mass and self-reported pain. A textbook example of why the pre-registered primary endpoint, not the most exciting secondary finding, is what should drive the headline conclusion.
2022 A Lifespan benefits for the combination of rapamycin plus acarbose and for captopril in genetically heterogeneous mice STR2022 NIA Interventions Testing Program C2017 cohort. In MALE mice, rapamycin plus acarbose started at 9 months produced a longer lifespan than either of the two prior ITP cohorts treated with rapamycin alone, suggesting the combination is more potent than its components used separately. In FEMALES the combination was neither better nor worse than rapamycin alone, which the authors relate to the limited survival benefit acarbose alone had shown in earlier female cohorts. Captopril gave a small but significant lifespan increase in females (4-5%). CAVEAT (important for how this is cited): the rapamycin-only comparison is HISTORICAL, against prior cohorts, not a concurrent rapamycin-only arm in the same experiment, so 'more potent than either component' is the authors' suggestion rather than a within-experiment randomised comparison.

Revision history

ShowHide
  • 2026-08-30: The basis previously read 'No study - animal or human -', an absence claim about the whole literature; now scoped to this corpus.
  • 2026-09-05: The ITP combination publication, previously flagged as missing, was entered as STR2022 (PMID 36179270).
  • 2026-09-23: Title changed after the literature review (URL kept).
  • 2026-09-29: Title changed from 'answered in mice' to 'at least additive in male mice'; the beginner text no longer says metformin and rapamycin share AMPK.
  • 2026-10-10: Audit N09 (related A07, A21). The gap text correctly said the ITP comparison was historical rather than concurrent factorial, while the hypothesis, the title and the beginner text called the mouse question answered and the effect 'at least additive'. That is not a weak adjective but a specific statistical claim: additivity needs an estimand, an interaction scale and appropriate controls, none of which a cross-cohort comparison supplies. Hypothesis, title, beginner text and Evidence_Stands_At now describe a cross-cohort survival signal, and the proposed experiment states the effect scale and the prespecified interaction model instead of leaving additivity undefined.

← All open questions · Frontier questions

Was this page useful?