OBJECTIVES:To validate the SAPS 3 admission prognostic model in patients with cancer admitted to the intensive care unit (ICU). DESIGN:Cohort study. SETTING:Ten-bed medical-surgical oncologic ICU. PATIENTS AND PARTICIPANTS:Nine hundred and fifty-two consecutive patients admitted over a 3-year period. INTERVENTIONS:None. MEASUREMENTS AND RESULTS:Data were prospectively collected at admission of ICU. SAPS II and SAPS 3 scores with respective estimated mortality rates were calculated. Discrimination was assessed by area under receiver operating characteristic (AUROC) curves and calibration by Hosmer-Lemeshow goodness-of-fit test. The mean age was 58.3+/-23.1 years; there were 471 (49%) scheduled surgical, 348 (37%) medical and 133 (14%) emergency surgical patients. ICU and hospital mortality rates were 24.6% and 33.5%, respectively. The mean SAPS 3 and SAPS II scores were 52.3+/-18.5 points and 35.3+/-20.7 points, respectively. All prognostic models showed excellent discrimination (AUROC>or=0.8). The calibration of SAPS II was poor (p<0.001). However, the calibration of standard SAPS 3 and its customized equation for Central and South American (CSA) countries were appropriate (p>0.05). SAPS II and standard SAPS 3 prognostic models tended somewhat to underestimate the observed mortality (SMR>1). However, when the customized equation was used, the estimated mortality was closer to the observed mortality [SMR=0.95 (95% CI=0.84-1.07)]. Similar results were observed when scheduled surgical patients were excluded. CONCLUSIONS:The SAPS 3 admission prognostic model at ICU admission, in particular its customized equation for CSA, was accurate in our cohort of critically ill patients with cancer.