The expression (Estimated-Predicted)/Predicted creatinine generation ¿(E-P)/P¿ has been proposed as an index of compliance in continuous peritoneal dialysis (CPD). We attempted to define an (E-P)/P value that can be used as a cut-off for non-compliance and to characterize the relation of (E-P)/P to serum albumin. In 324 clearance studies, (E-P)/P had a normal distribution with a mean of +0.094 and an SD of 0.357. In these studies, there was a weak correlation between (E-P)/P and serum albumin (r = 0.12, P < 0.05), but (E-P)/P was not a predictor of serum albumin by logistic regression. In 34 CPD patients, who had an increase in the dose of CPD resulting in an increase in measured creatinine clearance from 43.8 +/- 14.0 to 66.1 +/- 17.6 L/1.73 m2 weekly (P < 0.001), (E-P)/P increased correspondingly from +0.018 +/- 0.284 to +0.153 +/- 0.369 (P = 0.018), although all subjects should be in a steady state of creatinine excretion in the second study. The rise in (E-P)/P was statistically significant in the subgroup of 17 subjects who had a decrease in serum albumin, but not in the subgroup of 14 subjects who had an increase in serum albumin from the first to the second clearance study. The (E-P)/P cut-off for a definitive diagnosis of CPD non-compliance may be around +0.400 or +0.500. The findings of this study suggest that non-compliance, as indicated by (E-P)/P, has an adverse effect on serum albumin and that non-compliance increases after a prescribed increase in the dose of CPD.