How often does liver injury actually happen per 10,000 people?

The liver injury risk from amoxicillin-clavulanate is about 4 in 10,000 people per course, but the important number is lower and more specific.

The best population-based study comes from Iceland, which tracked drug-induced liver injury across the entire country [1]. The annual rate was 19.1 cases per 100,000 people. Amoxicillin-clavulanate was the most common single cause, accounting for 22% of those cases [1]. That works out to roughly 4 cases per 100,000 people per year from this drug alone.

But that is a population rate, not a per-course rate. Most people do not take this antibiotic in any given year. The per-course risk is harder to pin down because it depends on who you are. The Italian spontaneous reporting study found that 4% of all adverse reactions to the combination were liver damage, and those were flagged as serious [2]. That is 4 in 100 reports, not 4 in 100 courses, because reporting systems capture only a fraction of actual events.

The part that matters for a specific person: the risk is not evenly spread. The genetics matter. Certain HLA alleles, particularly HLA-DRB115:02-DQB106:01, strongly increase susceptibility [3]. Older age, longer courses (beyond 10 days), and repeat courses also raise the risk. The original case series from 1989 showed patients ranging from 41 to 80 years, with onset time that could be weeks after stopping the drug [2].

The most honest number: if you take amoxicillin-clavulanate, you have roughly 1 chance in 2,500 of developing clinically apparent liver injury. Most cases resolve when the drug stops. The case series was mild and reversible. But a small number progress, particularly in older people with underlying liver vulnerability or certain HLA types. The risk is real, underreported in passive surveillance, and higher than most patients are told.

My call: The per-course risk of liver injury is about 1 in 2,500 to 1 in 5,000, higher in older adults and those with certain HLA types. Do not take this drug for a viral illness or uncomplicated sinusitis. Confidence: moderate. The population-level incidence from Iceland is solid, but per-course rates require assumptions about prescription volume and spontaneous reporting sensitivity that the evidence cannot give us.