Instructor/Biostatistician University of South Dakota Sioux Falls, South Dakota, United States
Background: Antibiotics, such as amoxicillin-clavulanate, are a known cause of idiosyncratic drug-induced liver injury (DILI). A deeper understanding of the clinical and genetic determinants of risk may help clinicians reduce the frequency of DILI by deploying automated decision support in routine practice. Methods: Retrospective study using clinical practice-based data in the All of Us Research Program at NIH. Potential cases of antibiotic-related DILI were identified using a lab-based phenotyping algorithm applied to the de-identified electronic health records (EHR) of 119,812 patients exposed to any of the five most frequently prescribed antibiotics in the U.S. Results: A total of 745 potential case patients had developed abnormal liver function tests 0 to 90 days after starting amoxicillin, amoxicillin-clavulanate, cephalexin, azithromycin or ciprofloxacin. The racial distribution of these potential cases was similar to the racial distribution of the entire cohort. The ratio of potential case patients reporting Hispanic ethnicity versus non-Hispanic ethnicity (ratio = 27:100 in potential cases) was greater than the ratio for the entire cohort (ratio = 20:100). Following the rigorous application of exclusion criteria based on phenome-scanning, final case counts for acute DILI were 22 cases with amoxicillin, 30 with amoxicillin-clavulanate, < 20 with cephalexin, 24 cases azithromycin and 24 with ciprofloxacin. Final confirmed case rate was highest for amoxicillin-clavulanate (case rate = 8.15 DILI cases per 10,000 subjects exposed to amoxicillin-clavulanate (95% Confidence Interval 5.71, 11.64). Conclusion: EHR-linked research cohorts can be efficiently studied to (1) identify DILI cases related to antibiotic use and (2) quantify the distribution of relevant covariates. Phenotyping efforts such as these will facilitate characterization of the underlying genetic architecture for DILI within EHR-linked biobanks.
