Association of geriatric nutritional risk index with infection-related mortality in patients undergoing hemodialysis: The Q-Cohort Study

Abstract

The geriatric nutritional risk index (GNRI) is a simple but useful nutritional marker for all-cause mortality and cardiovascular mortality in patients undergoing hemodialysis (HD). However, whether the GNRI can predict infection-related mortality in patients undergoing HD remains unclear, and there is insufficient evidence regarding whether the GNRI improves the predictive value for risk assessment beyond the existing conventional nutritional markers. Here, we investigated the association between the GNRI and infection-related mortality in patients undergoing HD and evaluated the predictive value of GNRI. A prospective cohort study was performed on a total of 3436 Japanese HD patients aged ≥18 years. Patients were divided into four groups by quartiles of GNRI: (Quartile 1 [Q1], >100.2; Q2, 95.9-100.2; Q3, 90.8-95.8; Q4, <90.8). We estimated the relationship between GNRI and all-cause mortality and infection-related mortality using a Cox proportional hazards model. To assess the additional predictive value of the GNRI in risk assessment, we compared the c-statistic, net reclassification improvement, and integrated discrimination improvement among serum albumin, serum creatinine, and the GNRI. During follow-up period (median, 4.0 years), a total of 564 patients died; 120 of these patients died of infectious disease. All-cause mortality and infection-related mortality increased linearly with lower GNRI levels. After adjusting for confounding risk factors, the GNRI was an independent predictor of infection-related mortality as well as all-cause mortality (hazard ratio [HR], 5.89; 95% confidence interval [CI], 2.85-13.8; P<0.001 for Q4 vs. Q1, HR, 2.62; 95% CI, 1.23-6.24; P=0.01 for Q3 vs. Q1). Additionally, when the GNRI was incorporated into a model with potential risk factors instead of serum albumin, the c-statistic increased significantly (0.811 vs. 0.821, P=0.03), and the net reclassification improvement and integrated discrimination improvement was 0.26 (P=0.005) and 0.005 (P=0.01). This association was more apparent in the older patients (0.739 vs. 0.760, P=0.02) than in the younger patients (0.916 vs. 0.912, P=0.35). Similar results were observed between serum creatinine and the GNRI, but the difference did not reach statistical significance. Lower GNRI levels are an independent risk factor for infection-related mortality in patients undergoing HD. Moreover, addition of the GNRI to models with standard risk factors significantly improves the predictive ability of infection-related mortality, especially in older patients.