What is the best estimate of risk per course, not per year?

The per-course risk of liver injury from amoxicillin-clavulanate is about 1 in 2,500 to 1 in 5,000 courses, not per year.

The Iceland study gives the best population rate: 19.1 cases of drug-induced liver injury per 100,000 people per year, with amoxicillin-clavulanate causing 22% of those [1]. That is about 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 Italian spontaneous reporting study found that 4% of all adverse reactions to the combination were liver damage, and those were flagged as serious [2]. Plain amoxicillin had only 1% liver injury reports, none serious [2]. The clavulanate component is the problem.

The original case series from 1989 showed 18 cases of cholestatic liver injury, all reversible after stopping the drug [4]. Patients ranged from 41 to 80 years old. Onset could be weeks after finishing the course. The 2018 mortality study of 1,135 cases found early death in 0.8% of all DILI cases, but that jumped to 37% in the severe hepatocellular subgroup [5]. Older age drove the risk: 8.9% mortality in those over 76 versus 0.8% in those under 45 [5].

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.

My call: The per-course risk of clinically apparent liver injury is about 1 in 2,500 to 1 in 5,000 courses. That is higher than most patients are told. The risk concentrates in older adults and those with specific HLA types. Do not take this drug for a viral illness or uncomplicated sinusitis. Confidence: moderate. The population incidence from Iceland is solid, but per-course rates require assumptions about prescription volume and spontaneous reporting sensitivity that the evidence cannot fully resolve.