Statistical methods for cost-effectiveness analysis of left-truncated and right-censored survival data with treatment delays.

The incremental cost-effectiveness ratio (ICER) and incremental net benefit (INB) are widely used for cost-effectiveness analysis. We develop methods for estimation and inference for the ICER and INB which use the semiparametric stratified Cox proportional hazard model, allowing for adjustment for risk factors. Since in public health settings, patients often begin treatment after they become eligible, we account for delay times in treatment initiation. Excellent finite sample properties of the proposed estimator are demonstrated in an extensive simulation study under different delay scenarios. We apply the proposed method to evaluate the cost-effectiveness of switching treatments among AIDS patients in Tanzania.