The risk of acute liver injury among users of antibiotic medications in the protect project: The results of a nested case-control study using european outpatient healthcare data

Background: The estimated incidence of antibiotic induced acute liver injury (ALI) varies widely, depending on the case definition and source population used. Objectives: We aimed to compare the risk of ALI associated with exposure to any type of antibiotic in a Spanish and United Kingdom (UK) database with access to outpatient data using the same case definitions. Methods: The case-control studies in the Clinical Practice Research Datalink (CPRD) and “Base de datos para la Investigacion Farmacoepidemiologica en Atencion Primaria” (BIFAP) were nested in two cohorts: A cohort for whom the date of the first antibiotic prescription defined the start of follow-up and a nonusing control cohort for whom a random start-date was generated. After exclusion criteria were applied, cases with ALI were identified usingmedical codes, laboratory test results and referrals to specialists. Up to 5 controls were matched to cases by age, sex, calendar date and, in CPRD only, practice. We used conditional logistic regression to compute odds ratios (OR) and 95% confidence intervals of ALI associated with current use of antibiotics compared to non-use. Results were adjusted for potential confounding variables, including smoking status, underlying diseases such as diabetes, and use of concurrent medications. A secondary analysis was performed using a broader case definition. Results: In CPRD, 263 ALI cases could be matched to 1284 controls in CPRD. In BIFAP, 124 cases were matched to 620 controls. The results of the adjusted analyses showed qualitatively similar evidence of an increased risk of ALI up to 14 days after the receipt of an antibiotic in CPRD (OR 5.7, 95% CI 3.46-9.36) and BIFAP (OR 2.6, 95%CI 1.26-5.37). Using a broader case definition, the OR was 3.59 (95% CI 2.79-4.61) in CPRD and 3.08 (95% CI 2.05-4.62) in BIFAP. Conclusions: Whilst we acknowledge the potential inaccuracy in capturing ALI using observational data, we found that the use of a robust case definition led to comparable findings regardless of the source population used.