Early prediction of outcomes has major potential implications regarding the management of dialysis-related peritonitis. The outcomes of 565 consecutive episodes of peritonitis complicating peritoneal dialysis between August 2001 and July 2005 were evaluated in relation to the dialysate cell counts. Discriminatory power, based on the area under the receiver-operating characteristic (ROC) curves, of the cell counts was assessed. The findings then were validated externally in a cohort of 217 peritonitis episodes from another dialysis unit. During the study period, 565 episodes of peritonitis were included for analysis, 465 of which had treatment success defined as complete resolution of peritonitis without the need for Tenckhoff catheter removal. Of the remaining 100 episodes (treatment failure), 70 required Tenckhoff catheter removal and 30 had peritonitis-related death. The peritoneal dialysate total white blood cell count on day 3 of peritonitis predicted treatment failure independent of standard risk factors, and it had a higher area under the ROC curve than the dialysate white cell count on day 1 (0.80 versus 0.58; P < 0.0001). Using a peritoneal dialysate white count cut point > or = 1090/mm3 on day 3, the sensitivity was 75% and the specificity was 74% for the prediction of treatment failure (defined as catheter loss or peritonitis-related death). In multiple logistic regression analyses, peritoneal dialysate white count > or = 1090/mm3 on day 3 was an independent prognostic marker for treatment failure after adjustment for conventional risk factors (hazard ratio 9.03; 95% confidence interval 4.40 to 18.6; P < 0.0001). Number of years on peritoneal dialysis; diabetes; gram-negative organisms; and Pseudomonas, fungal, or Mycobacterium species were other independent risk factors that were predictive of treatment failure. Findings from an independent validation set of peritonitis (217 episodes after exclusion of Mycobacterium and fungal causes) also favored the peritoneal dialysate white count on day 3, as compared with day 1 and day 2, to predict treatment failure. Area under the ROC curve for the white counts on day 3 was 0.98 (95% confidence interval 0.95 to 0.99) in the validation set. This study demonstrated and cross-validated the superiority of peritoneal dialysate white cell count on day 3 to predict outcomes of dialysis-related peritonitis. These results call attention to the value of validating prognostic factors of peritonitis complicating peritoneal dialysis.