Incorporating patient-reported outcome data into a hematopoietic cell transplant survival calculator.

Abstract

7045 Background: The Center for International Blood and Marrow Transplant Research (CIBMTR) provides a validated 1-year Overall Survival (OS) calculator to estimate outcomes for individual patients prior to allogeneic Hematopoietic Cell Transplant (HCT) to inform risk. The calculator considers pre-HCT clinical and demographic characteristics, but not patient-reported outcome (PRO)s. Pre-HCT PRO scores have been associated with OS. We hypothesized that adding the baseline PRO score to the calculator will enhance its performance in predicting OS. Methods: In addition to established covariates,we considered pre-HCT PRO scores, collected on five (published) prospective randomized clinical trials performed through the Blood and Marrow Transplant Clinical Trials Network (BMT CTN). Measures used were the SF36 physical component score (PCS) and mental component score (MCS), and the FACT-BMT Trial outcome index (TOI). We also investigated inclusion of a single question from the SF36: In general, would you say your health is: excellent/very good/good vs fair/poor. Patients were ≥18. Results: 1,033 patients were included, 341 (33%) completed only the SF36, 339 (33%) only the FACT-BMT and 353 (34%) both, per study design. HCT occurred 2004-2014, and the population was diverse with respect to disease, stage, demographics, and transplant characteristics. When adjusted for clinical characteristics, the PCS was significantly predictive of mortality (HR=0.88, 95%CI 0.81-0.96, p=0.0021), while the MCS and TOI were not. The single general health question was also significantly associated with mortality (HR=0.52, 95%CI 0.35-0.74, p=0.0004). Addition of PRO scores to the calculator did not result in a significant change in the model’s predictive ability as measured by Harrell’s concordance statistic (Table). Conclusions: In this large analysis, we confirmed the significant, independent association of pre-HCT PRO scores with OS. We showed, for the first time in this setting, that a single general health question is as accurate as the full SF36 measure for predicting survival, but we did not find support for the hypothesis that adding PROs to the calculator improved its performance. Estimates of baseline PRO measure effect on mortality and model concordance statistics at 1-year post-HCT. [Table: see text]