Abstract
Four hundred twenty-two cancer patients who underwent major surgery were studied. At admission, nutritional status was evaluated in all patients by assessing serum albumin (SA), total iron-binding capacity (TIBC), total lymphocyte count (TLC), serum cholinesterase activity (CHE), and weight loss (WL). All patients received perioperative short-term antibiotic prophylaxis and postoperative total parenteral nutrition. Prognostic ability of nutritional indicators was assessed by receiver-operating characteristic (ROC) curve analysis. The area beneath the ROC curve (Az) is an index of predictor performance when its value ranges from 0.5 (chance performance) to 1 (perfect prediction). Specificity, sensitivity, Youden index, and predictive values were determined for each nutritional parameter within a wide range of potential threshold values. Postoperative septic complications were observed in 85 (20.14%) patients. The Az values for the considered nutritional parameters ranged from 0.52 to 0.57 and that showed the low predictive ability of the parameters. When sensitivity and specificity for each nutritional parameter were examined at different thresholds, a clearly more predictive cutpoint was not observed, but ranges of values with a similar predictivity were observed. Significant ranges of predictivity were found for SA (33 to 35 g/L), for TIBC (2200 to 2300 μg/L), for TLC (2100 to 2200 million/L), for CHE (1700 to 1900 U/L), and for WL (7% to 12%). The higher values of Youden index were as follows: 1.183 for WL (cutoff 11%), 1.150 for TLC (cutoff 2100 million/L), and 1.145 for SA (cutoff 35 g/L). In conclusion, ROC curve analysis showed that the nutritional parameters had a low predictive ability. However, an examination of different thresholds showed that within closed ranges of values the parameters provided a significant prognostic performance. (Journal of Parenteral and Enteral Nutrition
Get full access to this article
View all access options for this article.