Comparison of mobility indices for predicting early death in older patients with cancer: The physical frailty in elderly cancer cohort study

Abstract

Background: To assess and compare the ability of five mobility indices to predict 6-month mortality in older patients with cancer. Methods: All consecutive ambulatory older patients with cancer referred for a geriatric assessment before a cancer treatment decision were included in a prospective two-center cohort study (Physical Frailty in Elderly Cancer) between 2013 and 2017. The mobility indices compared were the short physical performance battery, gait speed, hand grip strength, the one-leg stance balance test, and repeated falls. The primary endpoint was 6-month overall mortality. The adjusted hazard ratio (95% confidence interval [CI]) for each mobility index was estimated using a multivariate Cox proportional hazard model adjusted for sex, the Cumulative Illness Rating Scale for Geriatrics, the body mass index, cancer site/extension, and the provision of supportive care alone. The models' predictive performances were assessed in terms of Harrell's C index, net reclassification improvement, and the standardized net benefit. Results: A total of 603 patients included (mean age: 81.2 ± 6.1 years; women: 54%; metastatic cancer: 45%). In multivariate analyses, an impairment in any of the mobility indices (with the exception of repeated falls) was independently associated with 6-month mortality following a geriatric assessment; the adjusted hazard ratio [95% CI] ranged from 2.35 [1.34-4.13] for the one-leg stance balance (C index: 0.74) to 3.03 [1.93-4.76] for the short physical performance battery (C index: 0.77). For each mobility index, inclusion in the multivariate model improved significantly the latter's prediction of 6-month mortality. Conclusions: Among mobility tests, short physical performance battery had the best discriminative value for predicting 6-month mortality in older patients with cancer.