BackgroundSecond line anti-tuberculosis drugs are substantially complex, long term, more costly, and more toxic than first line anti-tuberculosis drugs. In Ethiopia, evidence on the incidence and predictors of adverse drug events has been limited. Thus, this study aimed at assessing incidence and predictors of major adverse drug events among drug resistant tuberculosis patients on second line tuberculosis treatment in Amhara Regional State public hospitals, Ethiopia.MethodsA multi-center retrospective cohort study was conducted on 570 drug resistant tuberculosis Patients. Data were entered in to EPI-Data version 4.2.0.0 and exported to Stata version 14 for analysis. Proportional hazard assumption was checked. The univariate Weibull regression gamma frailty model was fitted. Cox-Snell residual was used to test goodness of fit and Akaike Information Criteria (AIC) for model selection. Hazard ratio with 95% CI was computed and variables with P-value < 0.05 in the multivariable analysis were taken as significant predictors for adverse drug event.ResultsA total of 570 patients were followed for 5045.09 person-month (PM) observation with a median follow-uptime of 8.23 months (Inter Quartile Range (IQR) =2.66–23.33). The overall incidence rate of major adverse drug events was 5.79 per 100 PM (95% CI: 5.16, 6.49). Incidence rate at the end of 2nd, 4th, and 6th months was 13.73, 9.25, 5.97 events per 100 PM observations, respectively. Age at 25–49 (Adjusted Hazard Ratio (AHR) = 3.36, 95% CI: 1.36, 8.28), and above 50 years (AHR = 5.60, 95% CI: 1.65, 19.05), co-morbid conditions (AHR = 2.74 CI: 1.12, 6.68), and anemia (AHR = 3.25 CI: 1.40, 7.53) were significant predictors of major adverse drug events.ConclusionThe incidence rate of major adverse drug events in the early 6 months of treatment was higher than that of the subsequent months. Age above 25 years, base line anemia, and co-morbid conditions were independent predictors of adverse drug events. Thus, addressing significant predictors and strengthening continuous follow-ups are highly recommended in the study setting.