127 FRAIL-AF RANDOMISED CONTROLLED TRIAL: RESULTS S2: SENSITIVITY ANALYSIS. Visual assessment of the cumulative incidence curves for the primary composite outcome of major bleeding and clinically relevant non-major bleeding revealed that the proportional hazard assumption for treatment arm was not met. In line with the sensitivity analysis prespecified in the statistical analysis plan, a causespecific Cox regression model was fitted that included a treatment by time period interaction. More specifically, a time-dependent coefficient for treatment arm was included in the model. Following the primary analysis, the renal function stratum was included as a fixed factor in the model and the thrombosis services stratum was included as stratification factor (allowing a separate baseline hazard function for each thrombosis service). Based on visual inspection of the cumulative incidence curve, a step-function was used with time periods defined as period 1 (0-100 days of follow-up) and period 2 (101-365 days of follow-up). In this sensitivity analysis, the hazard ratio (HR) for period 1 was 1.17 (95% confidence interval (CI) 0.70-1.96). For period 2 the HR was 2.10 (95% CI 1.40-3.16). 7
RkJQdWJsaXNoZXIy MTk4NDMw