Anticoagulant plasma concentrations and patient characteristics might affect the benefit–risk balance of therapy. The study objective was to assess the impact of model-predicted rivaroxaban exposure and patient characteristics on outcomes in patients receiving rivaroxaban for venous thromboembolism (VTE) prophylaxis (VTE-P) after hip/knee replacement surgery. Post hoc exposure–response analyses were conducted using data from the phase 3 RECORD1–4 studies, in which 12,729 patients were randomized to rivaroxaban 10 mg once daily or enoxaparin for ≤ 39 days. Multivariate regression approaches were used to correlate model-predicted individual rivaroxaban exposures and patient characteristics with outcomes. In the absence of measured rivaroxaban exposure, exposure estimates were predicted based on individual increases in prothrombin time (PT) and by making use of the known correlation between rivaroxaban plasma concentration and dynamics of PT. No significant associations between rivaroxaban exposure and total VTE or major bleeding were identified. A significant association between exposure and a composite of major or non-major clinically relevant (NMCR) bleeding from day 4 after surgery was observed. The relationship was shallow, with an approximate predicted absolute increase in a composite of major or NMCR bleeding from 1.08 [95% confidence interval (CI) 0.76–1.54] to 2.18% (95% CI 1.51–3.17) at the 5th and 95th percentiles of trough plasma concentration, respectively. In conclusion, based on the underlying data and analysis, no reliable target window for exposure with improved benefit–risk could be identified within the investigated exposure range. Hence, monitoring rivaroxaban levels is unlikely to be beneficial in VTE-P.