Nutritional status of older patients on hemodialysis: Which nutritional markers can best predict clinical outcomes?

Abstract

OBJECTIVE\nThe aim of this study was to investigate nutritional status in older patients undergoing maintenance hemodialysis (MHD) to determine the prevalence of nutritional markers indicating protein-energy wasting (PEW) as assessed by subjective global assessment (SGA) and other methods, and to explore which nutritional markers can best predict clinical outcomes.\n\n\nMETHODS\nThe study included 173 patients (median age 69 y; 65% men; 38% diabetes) undergoing MHD for >3 mo. Nutritional markers included SGA, malnutrition-inflammation score (MIS), geriatric nutritional risk index (GNRI), handgrip strength (HGS), midarm muscle circumference (MAMC), triceps skinfold thickness (SKF), calf circumference, and albumin. Associations between PEW (diagnosed by different measures and thresholds) and risk for hospitalization (by Poisson regression) and all-cause mortality (by Cox proportional hazards model) were analyzed.\n\n\nRESULTS\nDepending on methods and thresholds used, the prevalence of nutritional markers indicatingPEW varied from 6.9% to 59.5%. In the Poisson models adjusted for age, sex, dialysis length, and diabetes, low SGA, HGS, albumin, and high MIS score were associated with high hospitalization events, whereas in the bivariate Cox regression models adjusted for the same variables, low SGA, GNRI, BMI, calf circumference, and high MIS score were associated with high hazard ratio (HR) for mortality. In addition, in the multivariate models, SGA showed the strongest association with mortality (HR, 2.32; 95% confidence interval [CI], 1.27-4.24) and together with MIS (HR, 2.09; 95% CI, 1.20-3.64), the highest values of C-statistics.\n\n\nCONCLUSIONS\nAmong older MHD patients, the prevalence of nutritional markers indicating PEW varies substantially depending on methods applied. SGA, MIS, BMI, GNRI, calf circumference, and HGS predicted worse outcomes. SGA and MIS showed the strongest association with hospitalization and mortality risk in the adjusted models.