PRM79 - Systematic review of the Q-TWiST in oncology

Abstract

The quality-adjusted time without symptoms and toxicity (Q-TWiST) methodology has been used to assess the clinical benefits (prolonged [progression-free] survival) and costs (toxicities) of oncology therapies. This study was conducted to systematically review and quantitatively summarize published Q-TWiST assessments of cancer treatments. A systematic search and review was conducted in MEDLINE to identify original studies reporting the Q-TWiST information—including time with toxicity (TOX), time before disease progression without toxicity (TWiST), and time in relapse after disease progression (REL)—for all oncology treatment groups, as available. Utilities for Q-TWiST were also captured; when a base case for utilities was not selected in a study, the following was assumed: u(TWiST)=1, u(REL)=0.5, and u(TOX)=0.5. The relative gain in Q-TWiST for active treatment arms was calculated as the difference in Q-TWiST divided by mean overall survival of control arm. Relative gains ≥10% and ≥15% were considered to be a clinically important and clearly clinically important difference, respectively. Upon review of 84 initially identified articles, 39 were excluded for the following reasons: no Q-TWiST was reported (n=22), not oncology-related (n=13), other reasons (n=4). Forty-five studies were included and reported a total of 69 Q-TWiST comparisons across 10 cancers. The most commonly used utilities for Q-TWiST calculation were u(TWiST)=1, u(REL)=0.5, u(TOX)=0.5 (n=28, 62.2% of articles). Using base-case utility values, the mean (range) Q-TWiST gain was 5.2 (-6.8 to 61) months and the mean relative gain was 9.0% (-13.3% to 60.0%); 41.8% and 17.9% of studies reported relative gains ≥10% and ≥15%, respectively. Applying u(REL)= u(TOX)=0.5 to comparisons with sufficient data (n=65), the mean standardized Q-TWiST gain was 5.4 (-4.1 to 61) months and mean standardized relative gain was 8.2% (range -15.0% to 54.5%). This review of Q-TWiST for cancer therapies can serve as benchmark against which future analyses can be compared.