A drug designed to lower blood sugar added 14 percent to the median lifespan of male mice. The females got more of it in their bloodstream and lived exactly as long as before.
Most drugs that get tested for lifespan extension fail. The National Institute on Aging runs a program built specifically to find that out, and the great majority of compounds it feeds to mice do nothing measurable. So when something works, and works at three independent sites, it is worth understanding what happened.
Canagliflozin worked. It is an ordinary type 2 diabetes drug, sold as Invokana, and it belongs to a class called SGLT2 inhibitors. Its job is to make you urinate sugar. That is close to the whole mechanism, and it should not obviously have anything to do with how long you live.
Why This Matters
The search for drugs that slow aging has produced a short list of survivors. Rapamycin is on it. Acarbose is on it. Canagliflozin joined in 2020.
What makes this one interesting is not the size of the effect. It is that the drug is already sitting in pharmacies, already prescribed to millions of people, and already backed by some of the largest cardiovascular outcome trials ever run. Compared with rapamycin and mTOR inhibition, where human safety data at longevity-relevant doses is thin, the human safety picture here is unusually well mapped.
It also comes with a puzzle that the field has not solved, and that puzzle is arguably more informative than the lifespan number itself.
What These Drugs Actually Do
Your kidneys filter roughly 180 grams of glucose out of your blood every day, then reclaim almost all of it before it reaches the bladder. A protein called SGLT2, sitting in the proximal tubule, does about 90 percent of that reclaiming.
Block SGLT2 and the sugar leaves. A person on one of these drugs excretes somewhere in the range of 50 to 80 grams of glucose per day, which is a few hundred calories walking out in the urine. Blood glucose falls. Insulin falls with it. The body shifts toward burning fat, and the liver produces more ketones.
That combination, fewer usable calories, lower insulin, more fat oxidation, more ketones, looks a great deal like what happens during caloric restriction and fasting. Which is why two separate review groups have argued the class should be understood as a caloric restriction mimetic rather than simply a glucose-lowering agent [1,2].
The line runs back further than the modern drugs. Phlorizin, isolated from apple tree bark in 1835, does the same thing crudely. In 1886 Josef von Mering injected it into dogs and watched them spill sugar into their urine while remaining otherwise well. In 1960 Robert Crane proposed that glucose is carried across membranes coupled to sodium, the first cotransport mechanism ever described. SGLT1 and SGLT2 were cloned in the 1990s, and the drugs followed.
The Lifespan Result
The Interventions Testing Program feeds candidate compounds to genetically heterogeneous UM-HET3 mice at three sites at once, and requires the result to hold across all of them. It is the most demanding lifespan-testing setup in the field, and it exists because so many earlier single-lab longevity claims failed to replicate.
Canagliflozin went in at 180 parts per million in chow, starting at 7 months of age. In males, median survival rose 14 percent (P less than 0.001) and the age at 90th percentile survival rose 9 percent (P less than 0.001) [3].
In females: median survival up 1 percent, 90th percentile up 1 percent. Neither result was statistically significant.
Two details are worth holding onto. The male effect was not identical everywhere. Median gains at the three sites were 15, 9 and 3 percent. And when the researchers looked at what the mice died of, the picture was not that a particular disease had been prevented. Around three quarters of treated mice and four fifths of controls died of some form of cancer, a difference that did not reach significance. The authors describe the spectrum of lesions as similar between groups "but delayed in timing."
That is the signature of something acting on aging rather than on one disease. The same things went wrong. They went wrong later.
The Part Nobody Can Explain
The obvious first guess for a male-only drug effect is dosing. Perhaps the females metabolized it faster, or absorbed less, and simply did not get enough.
The data say the reverse. At 22 months of age, plasma canagliflozin averaged 13.2 micrograms per millilitre in females against 3.6 in males, a gap of roughly three and a half times, and the same pattern held in brain and kidney tissue [3]. The authors put it flatly: the absence of a survival benefit in treated females "cannot be attributed to lower circulating levels of the drug."
So the females got more drug and got nothing from it.
It was not that the drug failed to work in them metabolically, either. Fasting glucose fell in both sexes. Glucose tolerance improved in both sexes. Fat mass dropped 41 percent in females and did not budge in males. By most of the markers a clinician would check, the drug worked better in females.
It just did not extend their lives.
The authors' best explanation for the male benefit is that it comes from blunting glucose peaks rather than lowering average glucose, pointing out that acarbose, which also flattens post-meal glucose spikes, produces a similarly male-skewed lifespan effect. On the sex difference itself they are honest: "the basis for the sex specificity of Cana on rodent life span remains unclear."
This matters beyond one drug. It is a clean demonstration that improving metabolic markers and extending lifespan are not the same achievement, a theme we have covered in uncoupling insulin sensitivity from longevity. It is also one more entry in a pattern the ITP keeps producing, where interventions help one sex and not the other, which is part of why women's health deserves its own lens in longevity research rather than being treated as a footnote to male data.
Starting Late Still Worked
A fair objection to any lifespan study is that treatment started early. Few people would begin taking a drug in young adulthood on the chance it helps at 80.
The ITP tested a late start. Canagliflozin begun at 16 months of age, well into mouse midlife, still produced a 14 percent median lifespan increase in males (p = 0.004) [4]. That result came from two of the three sites rather than all three.
One caveat that we could not resolve. That paper carries two published correction notices, and the contents of neither were retrievable through PubMed, PubMed Central, or the correction records themselves, which carry metadata but no visible correction text. The number above is quoted from the original full text. Treat it as provisional until the corrections can be read.
The Brain Data
Follow-up work has gone after cognition, which is where a geroprotector would need to earn its keep for most people. An earlier study in the same aged mouse strain had already reported neuroprotective effects, setting up the question of whether they tracked the lifespan split [12].
A 2025 multi-omics study in the same mouse strain layered transcriptomics, proteomics and metabolomics to see what chronic canagliflozin does to the aging hippocampus [5]. In males, it induced mitochondrial function and insulin and cGMP-PKG signalling, suppressed neuroinflammatory networks, and improved hippocampal-dependent learning and memory.
In females, the transcriptional response to neuroprotective pathways switched on, and then stopped there. It did not carry through to proteins or metabolites, and it did not rescue cognition. The same split as the lifespan data, now visible at molecular resolution.
Human observational data point the same direction. In a UK primary care cohort of adults over 60, using target trial emulation across 25,533 matched pairs, people starting an SGLT2 inhibitor had a lower dementia rate than those starting a DPP4 inhibitor: hazard ratio 0.86 (95% CI 0.79 to 0.94) in the intention-to-treat analysis, and 0.70 (0.60 to 0.82) among continuous users [6].
Read that second number carefully. People who keep taking a medication for years differ from people who stop, in health, in wealth, and in every habit that correlates with both. Healthy-adherer bias inflates as-treated estimates in a predictable direction. The intention-to-treat figure is the more trustworthy of the two, and it is considerably more modest.
What the Human Trials Actually Tested
Here is where the enthusiasm needs a hard boundary.
SGLT2 inhibitors have some of the strongest outcome data in modern medicine. In EMPA-REG OUTCOME, 7,020 people with type 2 diabetes and established cardiovascular disease were randomized to empagliflozin or placebo. Cardiovascular death fell by 38 percent, hazard ratio 0.62 (95% CI 0.49 to 0.77), with all-cause mortality at 0.68 [7]. In CREDENCE, 4,401 people with diabetic kidney disease on canagliflozin saw the composite of kidney failure, doubled creatinine and renal or cardiovascular death fall from 61.2 to 43.2 events per 1,000 patient-years, hazard ratio 0.70 (0.59 to 0.82) [8].
The benefit is not confined to diabetes. In DAPA-HF, dapagliflozin helped people with heart failure whether or not they had diabetes [9], and a prespecified analysis of DAPA-CKD found kidney and cardiovascular benefit in chronic kidney disease with and without type 2 diabetes [10].
That last point is the strongest bridge to an aging argument. These drugs are doing something that is not merely about glucose.
But every one of those trials enrolled people who were already ill. Not one randomized healthy adults and measured aging. There is no human lifespan trial, no biological age endpoint, no equivalent of what the ITP did in mice. The honest summary is that we have excellent evidence these drugs prevent specific outcomes in specific sick populations, and no direct evidence they slow aging in anyone.
Risks
The safety profile is well characterized, which is a genuine advantage over most geroprotector candidates. It is not benign [11].
Genital fungal infections are the standout, running around 2 to 5 percent in men and 10 to 15 percent in women, a relative risk of 3.56. Euglycemic diabetic ketoacidosis, where ketoacidosis develops without the high blood sugar that normally signals it, occurs at 0.7 to 1.6 per 1,000 patient-years, odds ratio 2.13. It is uncommon and it is dangerous, partly because the usual warning sign is missing. Volume depletion is not raised overall but is in people with heart disease.
The amputation story is worth telling properly, because it is a case of the system working. The CANVAS trials showed roughly twice the rate of leg and foot amputations on canagliflozin. The FDA warned in 2016 and added a boxed warning in 2017. Further trials followed, the signal did not hold, and the FDA removed the boxed warning in August 2020. Pooled data now put the hazard ratio at 0.98 (95% CI 0.68 to 1.41). The caution stayed in the prescribing information; the boxed warning did not.
There is also an unresolved question in the opposite direction. Preliminary work presented as a conference abstract reported that canagliflozin increased intestinal adenoma burden in female mice bred to develop them. That is a preprint and abstract, not a peer-reviewed full paper, and it should not be treated as established. It is flagged here because it sits in the same female-specific gap as everything else in this story.
Where This Leaves You
No one should read this as a suggestion to take a prescription diabetes drug for longevity. Practically, in mid-2026, canagliflozin has no generic in the US and lists around 583 dollars a month; dapagliflozin got its first true generic approval in April 2026, with retail generic pricing in the region of 400 to 500 dollars for 30 tablets and discount pricing meaningfully lower. Those figures move constantly.
What the evidence supports is narrower and more useful. If you have type 2 diabetes, heart failure or chronic kidney disease, these drugs have outcome data most medicines can only envy, and that is a conversation to have with your doctor on those grounds alone. If you are healthy and interested in aging, the relevant finding is not "take this." It is the mechanism: blunting glucose peaks, lowering insulin, shifting toward fat oxidation. Those are levers that diet, Zone 2 training and meal timing also pull, without a prescription or a fungal infection.
And the deepest lesson is the female result. A drug improved fasting glucose, glucose tolerance and body fat in female mice, reached higher concentrations in their tissue than in males, and did not add a day to their lives. Any framework that treats better metabolic health markers as automatically equivalent to a longer life has to explain that.
Frequently Asked Questions
Do SGLT2 inhibitors extend human lifespan?
Nobody knows, because it has not been tested. The lifespan evidence is entirely from male mice in one research program. Human trials show large reductions in cardiovascular death and kidney failure, but only in people who already have diabetes, heart failure or kidney disease, and over a few years rather than a lifetime.
Why did the drug work in male mice but not female mice?
This is genuinely unresolved. The intuitive answer, that females received less drug, is contradicted by the data: their plasma levels were about three and a half times higher than males. The drug also improved their glucose control and cut their fat mass. It simply did not translate into survival. The study authors state that the basis for the sex specificity remains unclear.
Is canagliflozin safer than rapamycin as a longevity candidate?
Its side effects are better mapped, since millions of people have taken it at approved doses for years, whereas rapamycin's longevity dosing remains largely uncharted in healthy humans. Better mapped is not the same as safer. Genital infections are common, and euglycemic ketoacidosis is rare but serious and easy to miss.
Does the 14 percent lifespan gain in mice translate to 14 percent in humans?
There is no reason to assume so. Mouse lifespan gains have repeatedly shrunk or vanished in larger, longer-lived species, and this effect appeared in one sex of one strain. The gain also varied from 15 percent down to 3 percent across the three test sites, which is a useful reminder of how much a single headline number can hide.
Sources
- Hoong CWS, Chua MWJ. SGLT2 inhibitors as calorie restriction mimetics: insights on longevity pathways and age-related diseases. Endocrinology. 2021;162(8):bqab079. PubMed: 33857309.
- La Grotta R, de Candia P, Olivieri F, et al. Repurposing SGLT-2 inhibitors to target aging: available evidence and molecular mechanisms. Int J Mol Sci. 2022;23(20):12325. PubMed: 36293181.
- Miller RA, Harrison DE, Allison DB, et al. Canagliflozin extends life span in genetically heterogeneous male but not female mice. JCI Insight. 2020;5(21):e140019. PubMed: 32990681.
- Miller RA, Harrison DE, Cortopassi GA, et al. Lifespan effects in male UM-HET3 mice treated with sodium thiosulfate, 16-hydroxyestriol, and late-start canagliflozin. GeroScience. 2024;46(5):4657-4670. PubMed: 38753230.
- Jayarathne H, Sullivan R, Stilgenbauer L, et al. Canagliflozin reprograms the aging hippocampus in genetically diverse UM-HET3 mice and attenuates Alzheimer's-like pathology. Aging Cell. 2025;24(12):e70255. PubMed: 41047765.
- Wu CY, Iulita MF, Ganesh A, et al. Comparative dementia risk with GLP1 receptor agonists, SGLT2 inhibitors, or DPP4 inhibitors: a population-based cohort study. Alzheimers Res Ther. 2025;17(1):269. PubMed: 41420258.
- Zinman B, Wanner C, Lachin JM, et al. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med. 2015;373(22):2117-2128. PubMed: 26378978.
- Perkovic V, Jardine MJ, Neal B, et al. Canagliflozin and renal outcomes in type 2 diabetes and nephropathy. N Engl J Med. 2019;380(24):2295-2306. PubMed: 30990260.
- McMurray JJV, Solomon SD, Inzucchi SE, et al. Dapagliflozin in patients with heart failure and reduced ejection fraction. N Engl J Med. 2019;381(21):1995-2008. PubMed: 31535829.
- Wheeler DC, Stefansson BV, Jongs N, et al. Effects of dapagliflozin on major adverse kidney and cardiovascular events in patients with diabetic and non-diabetic chronic kidney disease: a prespecified analysis from the DAPA-CKD trial. Lancet Diabetes Endocrinol. 2021;9(1):22-31. PubMed: 33338413.
- Falsetti L, Zaccone V, Marra AM, et al. Beyond safety: adverse events and unanticipated advantages of SGLT2 inhibitors. Eur J Clin Pharmacol. 2026;82(4):93. PubMed: 41811498.
- Jayarathne HSM, Sullivan R, Stilgenbauer L, et al. Neuroprotective effects of canagliflozin: lessons from aged genetically diverse UM-HET3 mice. Aging Cell. 2022;21(7):e13653. PubMed: 35707855.
Funding Transparency
The split in funding here is unusually clean, and it maps onto the split in what the studies were designed to find.
The mouse lifespan work comes from the Interventions Testing Program, funded by the National Institute on Aging [3,4]. No company sells a product whose value depends on those results, and the ITP's three-site replication requirement exists specifically to make results harder to manufacture. This is about as close to disinterested evidence as the field gets.
The human trials are the opposite. EMPA-REG OUTCOME was funded by Boehringer Ingelheim and Eli Lilly, who market empagliflozin [7]. CREDENCE was funded by Janssen Research and Development, a Johnson and Johnson company, which markets canagliflozin [8]. DAPA-HF and DAPA-CKD were funded by AstraZeneca, which markets dapagliflozin [9,10]. These are well-conducted trials with independent adjudication committees, and industry funding does not make their results wrong. It does shape which questions got a hundred million dollars and which did not. Every one of these trials was designed to win a cardiovascular or renal indication. None was designed to test aging, and none will be, because there is no approval pathway to reward it.
That gap is worth naming. The most commercially valuable question about these drugs has been answered repeatedly and expensively. The most interesting one has not been funded at all.
The mechanism reviews cited here [1,2] are academic and declare no product interest, though authors publishing in this space frequently hold consultancies with the same manufacturers.
Related Reading
- Rapamycin and mTOR Inhibition for Longevity
- GLP-1 Drugs and Longevity: What the Science Actually Shows
- Uncoupling Insulin Sensitivity From Longevity
- Metabolic Health Fundamentals
- Why Women's Health Needs Its Own Longevity Lens
A drug that makes you urinate sugar bought male mice an extra 14 percent of life and bought females nothing, while working better in them by every measure a doctor would check. Until someone explains that, we should be careful about what we think metabolic health buys.
This article is for education, not medical advice. SGLT2 inhibitors are prescription medications with real risks, including genital infections and a rare form of ketoacidosis that can occur without high blood sugar. Do not start, stop or change a prescription based on animal longevity data. Talk to your doctor.
Written with the help of AI tools, shaped and verified by humans who care about getting this right.