A Terminal Decline Model for Fast Track Palliative Care Trials (S757)

Abstract

•Correctly analyze the quality of life and survival data.•Apply appropriate modeling approach for fast-track palliative care randomized clinical trials. In randomized fast-track trials (delayed-start), the fast-track arm receives intervention immediately after randomization and the control arm receives intervention after a certain period following randomization during which time standard care is provided. Such trials are more acceptable to patients, families, and physicians because no one is denied access to the intervention, while still allowing for a rigorous assessment of a palliative care intervention. Many issues remain with respect to implementing such designs and analyzing the longitudinal quality of life and survival. In this paper, we will propose a novel approach for jointly modeling quality of life data and survival to analyze fast-track designs in palliative care. We have developed a novel joint modeling approach where time-varying coefficients will be used in a both the survival sub model and the piecewise mixed effects sub model for the trajectory of the longitudinal quality of life at the end of life. Quality of life at fixed time point prior to death, quality of life conditional on being alive and quality-adjusted life years are estimated and compared between the intervention and control arms. ENABLE III is a recently completed fast-track trial at Dartmouth comparing palliative care intervention with the delayed start arm receiving palliative care after 3 months. Preliminary results show that the risk of death (hazard ratio [HR] (95% CI)) for immediate participants over the first year was 0.72 (95% CI, 0.57-0.89; p=0.003). The difference of the primary patient-reported outcomes at 3 months was generally increased but not statistically different: for the FACIT-pal (immediate group estimated means [95% CI] 129.9 [126.6, 133.3] vs delayed group 127.2 [124.1, 130.3]; overall p=0.34) and the QUAL-E symptom impact scale (11.4 [10.8, 12.1] vs 12.2 [11.6, 12.8]; overall p=0.09). The joint modeling approach is an efficient and interpretable method for fast track palliative care study designs.