Can Bioimpedance Measurements of Lean and Fat Tissue Mass Replace Subjective Global Assessments in Peritoneal Dialysis Patients?

Abstract

Malnutrition and protein energy wasting (PEW) determined by Subjective Global Assessment (SGA) is associated with increased mortality. There is an inverse relationship between body mass and overhydration in dialysis patients. Is the predictive accuracy of SGA (for death) independent of hydration status? Can bioimpedance spectroscopy analysis of lean tissue index (LTI) and fat tissue index (FTI) accurately identify dialysis patients with protein energy wasting and increased mortality? We report an observational study of 455 peritoneal dialysis (PD) patients. We found that 96 patients (21%) were malnourished (SGA score between 1 and 5), and 192 (42%) had LTI values below 10th centile (age, gender adjusted). FTI was significantly lower in the SGA-defined malnourished cohort. By contrast, there was an inverse relationship between LTI and FTI. Malnourished (by SGA) patients were significantly more overhydrated (P<.0001), but SGA remained highly predictive of survival in multivariate analysis that included hydration status (hazard ratio: 3.12, 95% confidence interval 1.86-5.23, P<.0001). Obesity (patients with the highest 20% FTI) predicted survival (hazard ratio of death was 0.47, 95% confidence interval 0.16-0.85, P<.02) on univariate but not multivariate analysis. We have confirmed that SGA is an accurate predictor of mortality in PD patients, and its predictive value is independent of the hydration status. Predictive power of SGA was not affected when LTI and FTI were included in multivariate analysis. Patients with low LTI were different from patients with low SGA (associated with high FTI). Sensitivity and specificity of Body Composition Monitor to diagnose patients with low SGA readings were poor (area under the curve for receiver operator characteristics analysis 0.66). The phenomenon of reverse epidemiology (high FTI predicting a survival advantage) was found in our PD cohort.