Comparison of accuracy between pre-hemodialysis and post-hemodialysis levels of nutritional factors for prediction of mortality in hemodialysis patients

Background: The nutritional status of hemodialysis (HD) patients is usually assessed using pre-HD laboratory data. However, it remains unclear whether the most diluted laboratory value is appropriate for assessment. We compared the pre-HD and post-HD laboratory data for their accuracy in predicting mortality.

Methods: Maintenance HD patients (n = 96 700; men, 61.5%) were enrolled. The outcome events were one-year and five-year mortalities. Their laboratory data included body mass index (BMI), and serum albumin, creatinine, and blood urea nitrogen (BUN) levels. Bootstrap resampling was used to compare the accuracy in predicting the mortalities between pre-HD and post-HD levels using area under receiver operating characteristic curves (AUCs) adjusted for baseline characteristics.

Results: The mean age ± standard deviation was 65.7 ± 12.2 years, and the vintage was 8.3 ± 6.7 years. The numbers of patients who died were 6442 (6.7%) in one year and 30 965 (32.0%) in five years. The adjusted AUCs for predicting the one-year and five-year mortalities showed that the pre-HD albumin and creatinine levels and the pre-HD BMI and BUN levels were more accurate than the post-HD levels (each p < 0.0001). The pre-HD albumin and creatinine levels showed the highest adjusted AUC for predicting one-year mortality [0.613 (95% CI 0.598, 0.629)] and five-year mortality [0.591 (95% CI 0.586, 0.595)], respectively.

Conclusions: Pre-HD albumin and creatinine levels are more accurate than post-HD levels and other nutritional indices in predicting one-year and five-year mortalities in HD patients.