Aim: Leflunomide, via its active metabolite teriflunomide, is used in rheumatoid arthritis (RA) treatment, yet approximately 20 to 40% of patients cease due to toxicity. The aim was to develop a time-to-event model describing leflunomide cessation due to toxicity within a clinical cohort and to investigate potential predictors of cessation such as total and free teriflunomide exposure and pharmacogenetic influences. Methods: This study included individuals enrolled in the Early Arthritis inception cohort at the Royal Adelaide Hospital between 2000 and 2013 who received leflunomide. A time-to-event model in nonmem was used to describe the time until leflunomide cessation and the influence of teriflunomide exposure and pharmacogenetic variants. Random censoring of individuals was simultaneously described. The clinical relevance of significant covariates was visualized via simulation. Results: Data from 105 patients were analyzed, with 34 ceasing due to toxicity. The baseline dropout hazard and baseline random censoring hazard were best described by step functions changing over discrete time intervals. No statistically significant associations with teriflunomide exposure metrics were identified. Of the screened covariates, carriers of the C allele of CYP1A2 rs762551 had a 2.29 fold increase in cessation hazard compared with non-carriers (95% CI 2.24, 2.34, P = 0.016). Conclusions: A time-to-event model described the time between leflunomide initiation and cessation due to side effects. The C allele of CYP1A2 rs762551 was linked to increased leflunomide toxicity, while no association with teriflunomide exposure was identified. Future research should continue to investigate exposure-toxicity relationships, as well as potentially toxic metabolites.