Lack of improvement of prognostic performance of weight loss when combined with other parameters.

This prospective study was designed to evaluate the ability of single and combined prognostic parameters in predicting postoperative infections in cancer surgical patients. The evaluation was based on multiple logistic analysis and receiver operating characteristic (ROC) curve analysis. The Younden's index (YI) was used to select threshold values of the parameters. This analysis was applied in 398 patients undergoing curative elective surgery for gastric, colorectal, or pancreatic cancer. At admission, the percentage of body weight loss, serum albumin, total lymphocyte count, total iron-binding capacity, and serum cholinesterase activity were evaluated. In all patients, the type and rate of postoperative infection were recorded. Multiple logistic analysis showed weight loss as the most predictive variable (p = 0.02), when taken individually. Its best cutoff value was 10% (YI = 1.27, p = 0.00001, ROC area = 0.62). When serum albumin was added as the second-best variable, with a threshold of 35 g/L, the combined YI was 1.27, and the ROC area was 0.65 (p NS vs. weight loss). Total lymphocyte count dichotomized at 2200 million/L was the third variable added to weight loss and serum albumin (YI = 1.31, ROC area = 0.59). In conclusion, weight loss split at 10% appears as the only index with a moderate prognostic performance that is worth evaluating in the preoperative nutrition assessment. A nonsignificant improvement of predictive ability was obtained by the combination of serum albumin, total lymphocyte count, total iron-binding capacity, or serum cholinesterase activity.