LAUSR

OBJECTIVE A geriatric assessment (GA) assesses functional age of older patients with cancer and is a well-established tool predictive of toxicity and survival. The objective of this study was to investigate the prognostic value of individual GA items. MATERIALS AND METHODS 546 patients with cancer ≥ 65 years completed GA from 2009 to 2014 and were followed for survival status for a median of 3.7 years. The GA consisted of function, nutrition, comorbidity, cognition, psychological state, and social activity/support domains. GA items with p < 0.05 in univariable analyses for overall survival (OS) were entered into multivariable stepwise selection procedure using a Cox proportional hazards model. A prognostic scale was constructed with significant GA items retained in the final model. RESULTS Median age was 72 years, 49% had breast cancer, and 42% had stage 3-4 cancer. Three GA items were significant prognostic factors, independent of traditional factors (cancer type, stage, age, and Karnofsky Performance Status): (1) "limitation in walking several blocks", (2) "limitation in shopping", and (3) "≥ 5% unintentional weight loss in 6 months". A three-item prognostic scale was constructed with these items. In comparison with score 0 (no positive items), hazard ratios for OS were 1.85 for score 1, 2.97 for score 2, and 8.67 for score 3. This translated to 2-year estimated survivals of 85%, 67%, 51% and 17% for scores of 0, 1, 2 and 3, respectively. CONCLUSIONS This three-item scale was a strong independent predictor of survival. If externally validated, this could be a streamlined tool with broader applicability.