Azathioprine treatment and the risk of extrahepatic malignancy in patients with autoimmune hepatitis: a nationwide cohort study using Korean claims data

Azathioprine (AZA) is widely used for maintenance treatment of autoimmune hepatitis (AIH), but its association with extrahepatic malignancy remains uncertain. We evaluated the association between early AZA prescription and subsequent extrahepatic malignancy in a nationwide Korean claims cohort. We conducted a retrospective cohort study using Health Insurance Review and Assessment Service (HIRA) data from 2008 to 2020. Patients were divided into an early AZA group, defined by at least one AZA prescription within 180 days after diagnosis, and a no-early-AZA group. Group assignment was fixed at the index date (Day 180); patients who started AZA later remained in the no-early-AZA group. Follow-up began 180 days after the initial AIH diagnosis. Cox regression was used to estimate hazard ratios (HRs) adjusted for age, sex, diabetes mellitus, cirrhosis, and corticosteroid use, overall and by malignancy site. The cohort included 8,280 patients: 3,059 in the early AZA group and 5,221 in the no-early-AZA group. Extrahepatic malignancy developed in 136 patients in the early AZA group (1.36 per 100 person-years) and in 219 patients in the no-early-AZA group (1.23 per 100 person-years; incidence rate ratio, 1.11; 95% CI, 0.89–1.37; P = 0.356). In the adjusted model, the hazard of extrahepatic malignancy in the early AZA group was not statistically significantly different from that in the no-early-AZA group (adjusted HR, 1.09; 95% CI, 0.79–1.51; P = 0.600). No statistically significant difference between the two groups was found for any individual malignancy site, including lymphoma/multiple myeloma (8 vs. 12 events; adjusted HR, 1.78; 95% CI, 0.49–6.51) and non-melanoma skin cancer (10 vs. 16 events; adjusted HR, 1.18; 95% CI, 0.38–3.71). This analysis did not detect a statistically significant association between early AZA prescription and subsequent extrahepatic malignancy, overall or by site, compared with no early AZA prescription. Given the exposure definition, the possibility of later AZA use in the comparison group, the mean follow-up of about 4 years, and the wide confidence intervals, no conclusion on long-term oncologic safety can be drawn.