We examined the predictive value of urea kinetics for patient outcomes in CAPD by measuring dialysis index (DI; a means of quantifying CAPD dose using urea kinetics), KT/V and normalized protein catabolic rate (PCRN) on 222 occasions in 76 new patients at the time of starting CAPD and at subsequent six month intervals. We investigated how these indices altered with time and in relation to each other, and how they correlated with a wide range of subsequent patient outcomes. DI, KT/V and PCRN all tended to decrease with time on CAPD (P less than 0.0004, less than 0.0001 and 0.0005, respectively). DI and KT/V were highly correlated with each other (r = 0.89, P less than 0.0001) and both correlated with PCRN (r = 0.57, P less than 0.0001 and r = 0.60, P less than 0.0001, respectively). DI and KT/V both correlated inversely with subsequent values for serum creatinine (P less than 0.0001), urea (P less than 0.0002), potassium (P less than 0.02) and phosphate (P less than 0.002), and directly with bicarbonate (P less than 0.0001). PCRN correlated inversely with serum creatinine (P less than 0.0002) and directly with urea (P less than 0.0001) and with the number of blood transfusions received (P less than 0.03). None of these indices correlated with levels of hemoglobin, PTH, alkaline phosphatase or albumin, or with nerve conduction velocity or any other subsequent clinical outcomes including death, technique failure, hospital days, peritonitis rate and subjective indices of fatigue, pruritus and insomnia. We conclude that the urea kinetic model is predictive of some biochemical outcomes but not of clinical outcomes in CAPD patients.