Skip to content
This GLP
← Research
Evidence

Liver cancer: half a percentage point over five years

GLP-1 starters had less liver cancer than users of four other diabetes drugs. Every difference lands under 0.6 percentage points, because the cancer is rare.

Dana Sullivan6 min read
Five-year liver cancer risk, percentage points lowervs insulin0.52%vs sulfonylureas0.39%vs metformin0.37%vs DPP-4 inhibitors0.37%Four comparators, one answer, all under 0.6 points.

This study does something most do not: it publishes its results as absolute risk differences rather than as ratios. That decision is why the numbers look small, and it is why they are useful — a hazard ratio on a rare cancer tells you almost nothing about how many people are involved, which is the gap a relative figure always leaves.

What was compared

Adults with type 2 diabetes starting a GLP-1 drug in one academic center’s clinical research network between 2010 and 2025, set against new users of metformin, insulin, SGLT2 inhibitors, DPP-4 inhibitors and sulfonylureas. [1] The primary outcome was a first diagnosis of hepatocellular carcinoma, the commonest form of liver cancer, over five years.

The numbers

Against insulin, five-year liver cancer risk was 0.52 percentage points lower, 95% CI 0.35 to 0.74. Against sulfonylureas, 0.39 points, 95% CI 0.13 to 0.64. Against metformin, 0.37 points, 95% CI 0.08 to 0.59. Against DPP-4 inhibitors, 0.37 points, 95% CI 0.06 to 0.70.

Per-protocol analyses, which account for whether people kept taking the drug, were stronger — which is expected, and is also why who keeps taking it is never a neutral variable. Results held across subgroups, sensitivity analyses and individual agents.

Why four comparators matter

The estimates against metformin, insulin, DPP-4 inhibitors and sulfonylureas sit inside a narrow band, between 0.37 and 0.52 points.

That consistency is the strongest thing here. Confounding by who gets prescribed what tends to produce a large gap against a weak comparator and a small one against a strong one — which is exactly the pattern that made another study distrust its own mortality result. Four comparators agreeing is a different shape.

What is not established

Survival. A secondary analysis looked at death after a liver cancer diagnosis and the authors describe it as limited by small sample sizes, inconclusive and exploratory, suggesting a potential benefit without demonstrating one.

Nothing in this is a reason to take a GLP-1 drug for cancer prevention, and nobody sells them for that. It sits alongside what these drugs do to a liver more generally — where the fatty liver findings are stronger and the outcomes further away.

Frequently asked

Do GLP-1 drugs prevent liver cancer?
This study found five-year risk lower by 0.37 to 0.52 percentage points against four comparator drugs. It is an observational comparison, and nobody prescribes these drugs for cancer prevention.
Why are the numbers so small?
Because liver cancer is rare. A condition affecting a small fraction of people cannot be prevented in a large fraction of them, whatever the relative reduction looks like.
Why does using four comparators help?
Confounding by who receives which prescription usually produces a big gap against a weak comparator and a small one against a strong one. Four estimates landing in a narrow band is a different pattern.
Did it improve survival?
Unknown. The mortality analysis was limited by small numbers, and the authors describe it as inconclusive and exploratory.

Sources

  1. [1] Hernández-Pérez JG, et al. (2026). Comparative risk of hepatocellular carcinoma and mortality among initiators of GLP-1 receptor agonists versus other glucose-lowering therapies: a target trial emulation Hepatology International. PMID 42584814

More in Evidence