Predictors of 1-Year Mortality in a Prospective Cohort of Elderly Patients With Cancer

Background. Mortality prediction is crucial to select the optimal treatment in elderly cancer patients. Our objective was to identify cancer-related factors and Comprehensive Geriatric Assessment (CGA) findings associated with 1-year mortality in elderly inpatients and outpatients with cancer. Methods. We prospectively included patients aged >= 70 years who had solid or hematologic malignancies and in whom the CGA was performed by geriatricians in two French teaching hospitals. We identified independent predictors of 1-year mortality after study inclusion, using multivariate Cox models stratified on inpatient/outpatient status. We built three multivariate Cox models, since strong correlations linked activities of daily living (ADL), Eastern Cooperative Oncology Group Performance Status (ECOG-PS), and timed get-up-and-go test (GUG) results; and since physicians' preferences for these three assessments vary. A sensitivity analysis was performed using multiple imputation. Results. Of the 993 patients (mean age, 80.2 years; 51.2% men), 58.2% were outpatients and 46% had metastatic disease. Colorectal cancer was the most common malignancy (21.4%). Mortality rates after 6 and 12 months were 30.1% and 41.2%, respectively. In all models, tumor site and metastatic status (p <.001), age > 80 years (p <.05), higher number of severe comorbidities (p <.05), and malnutrition (p <.001) were associated with death independently from impaired ECOG-PS (p <.001), ADL (p <.001), and GUG (p <.001). The adverse effect of metastatic status differed significantly across tumor sites, being greatest for breast and prostate cancer (p <.001). Multiple imputation produced similar results. Conclusion. The predictors of 1-year mortality identified in our study may help physicians select the optimal cancer-treatment strategy in elderly patients.