AUTHORS
Ajibade Ashaye, MD, MS, MBA, MPH, Ling Shi2, Christopher Dieyi3, Shien Guo4
AFFILIATIONS
1 Oncology Clinical Science, Takeda Development Center Americas, Inc., Cambridge, Massachusetts, United States
2 PPD Evidera Patient-Centered Research, Thermo Fisher Scientific, Waltham, Massachusetts, United States
3 Oncology Clinical Science, Takeda Development Center Americas, Inc., Cambridge, Massachusetts, United States
4 PPD Evidera Patient-Centered Research, Thermo Fisher Scientific, Waltham, Massachusetts, United States
ABSTRACT
Objectives
In the phase 3 PhALLCON trial, mixed models for repeated measures (MMRM) analyses demonstrated significantly improved patient-reported outcomes (PROs) with ponatinib versus imatinib. However, conventional MMRM analyses, which compare average changes in PRO domain and summary scores between treatment groups do not indicate how consistently individual patients benefit, are difficult to interpret without established thresholds for meaningful change, and do not identify which specific PRO items drive treatment differences, especially for multi-item domains and summary scores. We applied win ratio (WR) analysis, a patient-level pairwise comparison approach, to complement conventional analyses by evaluating the likelihood and drivers of favorable health-related quality of life (HRQoL) outcomes.
Methods
This post hoc analysis used PRO data from the phase 3 PhALLCON study in adults with newly diagnosed Philadelphia chromosome–positive acute lymphoblastic leukemia randomized to ponatinib or imatinib, in combination with reduced-intensity chemotherapy. PROs were assessed using the Functional Assessment of Cancer Therapy–Leukemia (FACT-Leu). WR analyses compared all possible patient pairs using change from baseline at the end of induction (C4D1) and consolidation (C10D1). For each FACT-Leu item (5-point Likert scale), pairwise comparisons were classified as win, loss, or tie based on the more favorable change from baseline in item score; results were aggregated to domain and summary levels based on the net balance of wins and losses. WR results were interpreted as the likelihood that a randomly selected patient treated with ponatinib would experience a more favorable outcome than one treated with imatinib.
Results
A total of 202 patients were included in the WR analysis population (ponatinib n=135; imatinib n=67). At C4D1, ponatinib increased the likelihood of favorable patient-level outcomes in physical well-being (PWB: WR 1.66; ~62% probability of more favorable outcome; p=0.014) and leukemia-specific symptoms (LeuS: WR 2.18; p<0.001), with corresponding improvements in Trial Outcome Index (TOI: WR 1.85; p=0.002) and total score (WR 1.54; p=0.028). These findings indicate consistent patient-level benefit primarily driven by improvements in fatigue, gastrointestinal symptoms, and leukemia-related concerns. No statistically significant differences were observed for social/family well-being (SWB), emotional well-being (EWB), functional well-being (FWB), or FACT-G.
At C10D1, WR provided additional insight by identifying significant patient-level benefit in EWB (WR 1.71; p=0.024), a finding not detected in prior MMRM analyses, highlighting the additional patient-level insight provided by the WR approach. In addition, ponatinib-treated patients were more likely to experience favorable outcomes in PWB (WR 2.61; ~72% probability of more favorable outcome; p<0.001), and LeuS (WR 1.69; p=0.020), with significant separation across all summary scores (FACT-G WR 1.88; p=0.011; TOI WR 2.21; p=0.001; total WR 1.95; p=0.004). No statistically significant differences were observed for SWB or FWB.
Conclusions
WR analysis demonstrates consistent patient-level HRQoL benefit with ponatinib and complements conventional PRO analyses by showing that statistically significant group-level HRQoL advantages were also reflected in consistent patient-level benefit. This approach provides an interpretable probability of benefit and identifies key symptom domains driving treatment differences. Applied to PhALLCON, WR offers additional clinical insight beyond mean-based analyses and complements conventional PRO analyses by providing patient-level interpretation without reliance on predefined meaningful change thresholds.
DISCLOSURES
Christopher Dieyi: Takeda Development Center Americas, Inc. - Current Employment