PCN372 METHODOLOGIES FOR ADJUSTMENT OF THE TRIAL OVERALL SURVIVAL FOR TREATMENT SWITCHING: A REVIEW OF NICE SUBMISSIONS

Abstract

Frequently, technology appraisals (TAs) submitted to the National Institute for Health and Care excellence (NICE) include methodologies that adjust trial overall survival (OS) for treatment switching. The aim is to evaluate how NICE responds to these methods, their critique and recommendations for future submissions. This NICE HTA review included all oncology submissions that were completed as of December 2001. All reports published on the NICE website on TAs including naïve and more complex methodologies that adjust for treatment switch such as inverse probability of censoring weights (IPCW), rank-preserving structural failure time (RPSFT) models, iterative parametric estimation (IPE) and the 2-stage method were included. The included extraction items were: methodology used, acceptance by NICE of the adjusted outcomes, and overall criticism/recommendations. Out of 218 submissions, 38 were included in this review. Most commonly, more than one method was used. The RPSFT and IPCW methods were combined in 12 submissions. Among all submissions, the most common method was RPSFT (32 submissions) followed by the IPCW (15 submissions) and the 2-stage method (11 submissions). IPE and naïve methods were presented in six and five submissions, respectively. In twelve appraisals, only a RPSFT model was presented. NICE considered RPSFT a preferred methodology if the common treatment effect assumption was met. Otherwise, the 2-stage method was recognized as a better fit (5 submissions). IPCW was considered unsuitable when the percentage of patients switching treatments was high. Using more than one adjustment method was recommended. However, it appears that the choice of adjustment method(s) used or their number was not crucial for the reimbursement decisions. Treatment switching adjustment methods appear to be well accepted by NICE. Currently, NICE recommends for the future submissions to report treatment switching methodologies in a transparent way and to explore multiple adjustment methods.