OBJECTIVES : Treatment switching, from control to treatment arm, following disease progression is common in oncology trials of novel therapies. Using standard intention-to-treat (ITT) analysis would likely bias estimates of treatment benefit. Analysis of survival data typically adjust for this bias, but such adjustments are rarely performed in analyses of health-related quality of life data (HRQoL). The aim of the study is to examine the impact of adjusting for treatment switching using EQ-5D-5L data. The AURA 3 trial was a randomised controlled trial comparing osimertinib with chemotherapy (standard care), in patients with locally advanced or metastatic EGFR and T790 mutation-positive non-small-cell lung cancer where two-thirds of the chemotherapy arm received osimertinib post-progression.

METHODS : Descriptive analyses including pareto classification of health change (PCHC) were used to compare the treatment arms. The primary analysis used a two-stage least squares instrumental variable regression to estimate the treatment effect by adjusting for treatment crossovers. Multiple imputation was applied prior to primary analysis to account for missing data. Time to deterioration (TTD) analysis, defined by minimally important difference (MID) in EQ-5D-5L utilities between treatments was also assessed after adjusting for crossover using a rank preserving structural failure time model (RPSFTM).

RESULTS : The PCHC results showed a greater proportion of osimertinib patients reporting improvement compared to chemotherapy. Estimated incremental quality adjusted life years (QALYs) for osimertinib based on ITT analysis of imputed data was 0.23 at 60 weeks. Primary analysis accounting for treatment switching increased this to 0.52. A sensitivity analysis increased this to 0.63 QALYs at 150 weeks. TTD analysis found a longer HRQoL maintenance using osimertinib, of 12.76 weeks, although this was not statistically significant.

CONCLUSIONS : This study demonstrated methods to adjust for treatment switching in the analysis of EQ-5D from clinical trials. Failure to account for crossover significantly underestimated the QALY gain for osimertinib.