Advertisement

MS2 TREATMENT SWITCHING ANALYSES ON PATIENT-LEVEL DATA TO INFORM TRANSFERABILITY OF A TRIAL-BASED HEALTH ECONOMIC ANALYSIS IN METASTATIC COLORECTAL CANCER: A CASE STUDY USING PATIENT-LEVEL DATA FROM THE FIRE-3 TRIAL

      Objectives

      Geographic transferability of health economic analyses (HEA) may be facilitated by addressing differences in treatment and patient characteristics of an original trial population vs the population of interest. For a health technology assessment–driven HEA, applicability of the FIRE-3 trial comparing cetuximab to bevacizumab added to FOLFIRI (CET+F; BEV+F) as first-line treatment of RAS wild-type metastatic colorectal cancer was explored. In FIRE-3, a proportion of patients received panitumumab (PAN) as a subsequent treatment. Treatment-switching analyses were used to adjust overall survival (OS) for later-line treatment effects in jurisdictions where PAN is unavailable.

      Methods

      The two-stage method (TSM) was applied to adjust trial outcomes. In scenario analyses, excluding (EXC) or censoring (CEN) PAN-receiving patients from FIRE-3 OS was explored. TSM adjusts patients’ survival times (ST) by multiplying ST beyond start of second-line treatment with an acceleration factor (AF), which represents relative effectiveness of PAN vs other subsequent treatments (first step). AFs were derived from the FIRE-3 trial (internal [INT]) and external (EXT) published data. Resulting counterfactual ST estimates were calculated for each patient according to the first-line treatment received (second step). Recensoring was applied to correct for imbalances resulting from AF <1. Counterfactual ST were compared, and hazard ratios (HRs) were calculated for each method. Bootstrap analyses were performed to estimate uncertainty over the outcomes.

      Results

      The FIRE-3 trial HR (95% CI) for CET+F vs BEV+F was 0.697 (0.539-0.903). HRs adjusted for subsequent PAN use were 0.682 (0.526-0.885; TSM INT), 0.696 (0.530-0.914; TSM EXT), 0.702 (0.531-0.929; EXC), and 0.714 (0.540-0.944; CEN). P-values for all HRs were ≤0.05.

      Conclusions

      Adjusted HRs showed OS improvement favouring CET+F. Bootstrap analyses confirmed deterministic results. Consequently, data from the FIRE-3 trial can be applied in HEA in a range of jurisdictions to calculate robust ICERs for CET+F vs BEV+F.