Protein catabolic rate in patients on continuous peritoneal dialysis. A multivariate predictive model.

Protein catabolic rate (PCR) and PCR normalized to standard weight (PCRN) are important indices of nutrition in patients on continuous peritoneal dialysis. The purpose of this study was to test whether urea clearance is among the predictors of PCR and PCRN in a multivariate analysis. Stepwise logistic regression was used to develop separate models for low PCR and low PCRN on a set of 143 urea kinetic studies in 92 patients on continuous peritoneal dialysis. The regression models were tested on an independent sample of 189 urea kinetic studies in 102 patients on continuous peritoneal dialysis by deriving the area under a receiver operating characteristic curve. In the derivation set, low serum urea, high serum creatinine, low urine and dialysate drain volumes, and low body surface area were identified as predictors of PCR < or = 50 g daily. The area under the receiver operating characteristic curve in the validation set was 0.930 (95% confidence interval: 0.915-0.945). Low serum urea, male gender, high body mass index and low urea fractional clearance (KT/V) were predictors of PCRN < or = 0.80 g/kg daily. The receiver operating characteristic area for this model was 0.948 (95% confidence interval: 0.926-0.970). Logistic regression analysis was repeated twice after adding urea nitrogen excretion normalized to standard weight (UNEN) as a candidate variable. This process identified low UNEN, male gender, and obesity as the predictors of low PCRN, and low UNEN, male gender, low urine volume, low drain volume normalized by body water, and high serum albumin as predictors of low KT/V urea. The authors conclude that PCR and PCRN can be predicted by models that incorporate serum azotemic indices, body size and composition, and direct or indirect measurements of urea clearance. Small body size and lean body composition predict low PCR but high PCRN values. Both PCRN and KT/V urea are predicted by UNEN. Multivariate analysis cannot, therefore, rule out the hypothesis that PCRN and KT/V are linked mathematically.