Acarbose: an old diabetes drug with the best mouse lifespan data outside rapamycin

A drug that works entirely inside the gut, taken for an effect measured over decades in mice. The mechanism is unglamorous and the longevity data is among the strongest in the field — and it came with a wrinkle that rarely gets mentioned.

It works before absorption, not after

Established clinical use Alpha-glucosidase enzymes on the intestinal brush border break disaccharides and oligosaccharides into absorbable glucose. Acarbose competitively inhibits them, so carbohydrate digestion is slowed and shifted further down the gut. The effect is a blunted post-meal glucose rise rather than a lower fasting glucose, and the drug barely enters the circulation at all — which is why it has essentially no systemic pharmacology and why the app holds no half-life for it.

Two practical consequences follow directly. It must be taken with the first bite of a meal, because it has to be present when the carbohydrate arrives; taken afterwards it does nothing. And it does nothing at all on a meal without carbohydrate, which makes it a poor fit for anyone eating low-carbohydrate — there is no substrate to slow.

The gastrointestinal effects come from the same mechanism. Carbohydrate that reaches the colon undigested gets fermented by gut bacteria, and the flatulence and bloating that follow are the drug working rather than a side effect in the usual sense. They diminish over weeks as the microbiome adapts, which is why starting low and titrating is standard.

What the ITP found, and the part that gets left out

Established clinical use The National Institute on Aging’s Interventions Testing Program is the multi-site programme designed specifically to weed out lifespan results that do not replicate. Acarbose extended median lifespan in genetically heterogeneous mice, and the effect was substantial — on the order of twenty per cent in males. In later work, acarbose combined with rapamycin outperformed either alone.

The wrinkle is that the effect was strongly sex-dependent: much larger in males than in females, where it was modest. That pattern recurs across several ITP interventions and nobody has a settled explanation for it. Anyone quoting the twenty per cent figure without saying which sex it applies to is quoting half the result.

Animal-only or theoretical And it is a mouse result. There is no human longevity trial of acarbose, and the human evidence is for glycaemic control in type 2 diabetes plus a cardiovascular signal in impaired glucose tolerance that has been argued over. The mechanistic story — fewer glucose excursions, less glycation, less oxidative stress — is plausible and is not the same as a demonstrated outcome.

What to watch

The app’s panel is fasting glucose, HbA1c and liver enzymes. HbA1c is the marker that reflects what this drug does over time, and the HbA1c page covers why it and fasting glucose can disagree — which is exactly what a drug that flattens post-meal peaks without lowering the fasting value should produce. Continuous glucose monitoring, where available, shows the effect far more directly than either.

Liver enzyme elevation is rare and dose-related and is the reason it is on the panel. Hypoglycaemia is not a risk from acarbose alone, but it is worth knowing that if hypoglycaemia occurs alongside another glucose-lowering drug, ordinary table sugar will not correct it quickly — the drug is blocking the enzyme that breaks it down, so glucose itself is needed.

It requires a prescription and the longevity use is off-label. Whether it applies to a particular person is a prescribing decision, and anything on the drawbacks list below belongs in that conversation.

The dosing rows the app records

Established clinical use Reproduced from the app’s reference so you can see what it holds, not as a recommendation. Which row applies to a particular person, if any, is a clinical decision this page does not make — and rows describing supraphysiological or post-cycle use are filtered out before this table is built, so what you see here is a subset.

What the app monitors alongside it

The panel below is the app’s own monitoring note for Acarbose, verbatim. The analytes in it that have a page here are linked; those pages cover what each one measures, which assay produced it and how to read a trend.

• Hematocrit on testosterone therapy
• HbA1c and fasting glucose

Drawbacks and risks the app records

From the app’s own entry, and note how many items are the same fact: this drug acts on carbohydrate in the gut, so timing and diet decide whether it does anything at all.

• GI side effects — flatulence and bloating (reduced carb fermentation)
• Must be taken with first bite of each meal
• Less effective on low-carb diets (less substrate to slow)
• Not a dramatic effect in isolation
• Requires prescription
• GI side effects limit dose escalation
• GI side effects (gas, bloating) especially initially
• Must be taken with first bite of meal — timing critical
• Reduces absorption of some medications
• Not useful without carbohydrate intake
• Uncomfortable GI effects at higher doses

Questions people actually ask

Why must it be taken with the first bite?

Because it competes with carbohydrate for the same intestinal enzymes and has to be present when the carbohydrate arrives. Taken after a meal it does nothing.

Does it work on a low-carbohydrate diet?

Barely. There is nothing for it to slow. That is a genuine limitation for anyone whose diet is already low in carbohydrate.

How big was the mouse effect?

Around twenty per cent in median lifespan in males in the NIA programme, and considerably smaller in females. The sex difference is real and is usually omitted when the figure is quoted.

What if I have a hypo while taking it?

Acarbose alone does not cause hypoglycaemia, but if one occurs alongside another glucose-lowering drug, table sugar will not correct it quickly — the enzyme that breaks it down is what the drug is blocking. Glucose itself is what works.

Where the load-bearing numbers come from

Named rather than linked. Publisher URLs move, and a citation that resolves to a 404 two years from now is worse than one you can search for by name — every entry below is findable from the title and year alone.