Comparison of Nutritional Risk Scores for Predicting Mortality in Japanese Chronic Hemodialysis Patients

Objective Protein energy wasting (PEW) is consistently associated with poor prognosis in hemodialysis (HD) patients. We compared the predictability of PEW as diagnosed by The International Society of Renal Nutrition and Metabolism criteria (PEW_ISRNM) and geriatric nutritional risk index (GNRI) for all-cause mortality in Japanese HD patients. As cut-off values for body mass index (BMI) for PEW have not been established in PEW_ISRNM for Asian populations, these were also investigated. Design and Subjects The nutritional status from 409 HD patients was evaluated according to ISRNM and GNRI criteria. To compare the predictability of mortality, C-index, net reclassification improvement (NRI) and integrated discrimination improvement were evaluated. Results During follow-up (median, 52 months; range, 7 months), 70 patients (17.1%) presented PEW according to ISRNM and 131 patients (32.1%) according to GNRI; in addition, 101 patients (24.7%) died. PEW_ISRNM and GNRI were identified as independent predictors of death. Addition of PEW_ISRNM and GNRI to a predictive model based on established risk factors improved NRI and integrated discrimination improvement. However, no differences were found between models including PEW_ISRNM and GNRI. When lowering the criterion level of BMI per 1 kg/m² sequentially, PEW_ISRNM at BMI <20 kg/m² maximized the hazard ratio for mortality. The model including PEW_ISRNM at BMI <20 kg/m² improved NRI compared with the model including GNRI. Conclusion PEW_ISRNM and GNRI represent independent predictors of mortality, with comparable predictability. The diagnostic criterion of BMI in the ISRNM for Japanese population might be better at <20 kg/m² than at <23 kg/m².