OP437 Use Of Real-World Evidence In Survival Analysis Adjusting For Treatment Crossover In Cutaneous T-Cell Lymphoma

Abstract

Real-World Evidence is useful for validating crossover adjustment approaches, particularly when the adjustment is required because a trial does not accurately reflect a health technology assessment (HTA)-relevant population. We use the MAVORIC trial advanced stage mycosis fungoides and Sézary syndrome cutaneous T-cell lymphoma population and data from the Hospital Episodes Statistics to explore and validate crossover adjustment methods.

Introduction

The MAVORIC trial compared mogamulizumab to vorinostat in patients with mycosis fungoides (MF) or Sézary syndrome (SS), subtypes of cutaneous T-cell lymphoma. However, the treatment comparison within MAVORIC may not represent an HTA relevant population from a UK perspective: (i) 72.6 percent of patients randomized to vorinostat switched to mogamulizumab and (ii) vorinostat is not used in current clinical practice in the UK. This study explores methods to adjust treatment effect estimates using different crossover adjustment methods and Real-World Evidence.

This medicine is subject to additional monitoring. This will allow quick identification of new safety information. See www.mhra.gov.uk/yellowcard for how to report side effects.

Methods

An advanced stage (stage ≥IIB MF and all SS) population was included. Three methods were considered for treatment crossover adjustment. A synthetic control arm was created using the Hospital Episodes Statistics (HES) dataset. Predicted survival for the MAVORIC control arm, post-crossover adjustment, was compared to the HES to inform the selection of the appropriate methods for adjustment. A direct comparison between mogamulizumab (reweighted to represent the distribution of MF/SS patients in the HES) and the synthetic control was also conducted.

Results

Following crossover adjustment of the vorinostat arm, using the inverse probability of censoring weighting method, the overall survival (OS) hazard ratio (HR) estimate for mogamulizumab vs. vorinostat was 0.45 (95% confidence interval (CI): 0.19, 1.07). This adjustment method was considered the most appropriate based on an assessment of assumptions and a comparison of OS between the adjusted vorinostat data and the HES data. The OS HR estimate for reweighted mogamulizumab vs. synthetic control from HES was 0.33 (CI: 0.21, 0.50).

Conclusions

Real World Evidence from the HES database can be used to validate crossover adjustment methods and to better reflect current clinical practice in the UK. Results using both methods support each other.