Using health-related quality of life and treatment measures to predict 10-year mortality in older survivors of early stage breast cancer.

Abstract

e21548 Background: Older women make up the growing majority of breast cancer (BC) survivors. Guideline BC treatments are known contributors to long-term survival, and health-related quality of life (HRQOL) is emerging as not only an important outcome in survivorship care but also as a factor thought to influence mortality. However, prognostic models that include HRQOL alongside BC treatment measures are lacking. We aimed to develop a 10-year mortality risk score based on a priori chosen treatment and HRQOL variables. Methods: We studied 660 women ≥65-years old diagnosed with stage I-IIIA primary breast cancer in years 1997-1999. Data from medical and psychosocial domains were collected over 10 years from interviews, medical records, and death indexes. BC treatment variables included receipt of definitive locoregional surgery +/- radiation, chemotherapy, and tamoxifen. HRQOL variables included physical function [10-item Physical Function Index (PFI-10) from the Medical Outcomes Study Short Form-36 (MOS SF-36)]; mental health [5-item Mental Health Index (MHI-5) from the MOS SF-36]; and social support [8-item modified MOS Social Support Scale (mMOS-SSS)]. We used penalized logistic regression models to develop a 10-year mortality risk score, and investigated its discrimination (c-statistic) and calibration (observed versus predicted mortality using the Hosmer-Lemeshow (HL) test). Results: Mortality though 10-years of follow-up was 34.8% (230 of 660 women). The c-statistic of a risk score using only age, number of comorbidities, stage of BC, and BC treatment was 0.71. The c-statistic increased to 0.74 with the addition of HRQOL measures and showed good calibration (p = 0.72 from HL test). Physical function and mental health had strong independent associations with mortality (women with high PFI-10: OR 0.63, 95% CI 0.43, 0.92; women with high MHI-5: OR 0.57, 95% CI 0.39, 0.85). Conclusions: In older early stage breast cancer survivors, our risk score combining HRQOL with treatment measures showed good discrimination and calibration. HRQOL is independently associated with 10-year mortality and adds predictive ability to age, comorbidity, stage of BC, and BC treatment.