Background  Drug-induced liver injury may be immunologically mediated or metabolically induced. Peripheral eosinophilia and liver eosinophilia in suspected drug-induced liver injury generally supports the role of drug aetiology.

Aim  To assess the importance of eosinophilia and hepatic necrosis on outcome in patients with suspected drug-induced liver injury.

Methods  We performed search of MEDLINE for case reports on drug-induced liver injury associated with: amoxicillin/clavulanic acid, carbamazepine, diclofenac, disulfiram, erythromycin, flucloxacillin, halothane, isoniazid, phenytoin, sulindac and trimethoprim/sulfametoxazol.

Results  A total of 570 case reports were retrieved. Mortality/transplantation occurred in 112 (20%). Eosinophilia in peripheral blood was reported in 34% of cases, eosinophilia in liver biopsies in 40%, and hepatic necrosis in 41%. Bilirubin levels were lower in patients with peripheral eosinophilia [5.5 × upper limit of normal (interquartile range 2.9–10) vs. 7.7 (4–17); P = 0.02] and patients with liver eosinophilia [5 × upper limit of normal (2.7–10) vs. 10 (5.4–20); P = 0.003] as compared with those without eosinophilia. Eosinophilia in peripheral blood and eosinophilia in liver biopsies were more common in patients who recovered (37% vs. 15.6%; P = 0.0001 and 48% vs. 18.8%; P < 0.0001, respectively). Hepatic necrosis was present in 24% in the survivors vs. 84% in non-survivors (P < 0.0001).

Conclusions  In drug-induced liver injury, a favourable outcome was related to the occurrence of eosinophilia, whereas hepatic necrosis was associated with a poor prognosis.