Introduction/Background Nutritional status is directly associated with the long-term prognosis of cancer patients as well as the perioperative outcome, including infectious morbidity. Prognostic nutritional index (PNI), a predictor of nutritional status, is considered to be an important prognostic indicator in cancer patients and this fact has been also observed in gynecological cancer as well.
Methodology We conducted a prospective observational study of gynecologic oncology patients undergoing surgical procedure between January 2019 and December 2021. Patient with extremely low body mass index (BMI <18 kg/cm2) were excluded. Multivariate predictive analysis for postoperative infectious diseases was performed using logistic regression, naïve Bayes, classification and regression trees, random forest and neural network analysis with the Python software. Parameters that were considered included patient age, body mass index (BMI), ECOG status, smoking, presence of systemic disease, use of enhanced recovery after surgery protocol, preoperative PNI and postoperative CRP.
Results Overall, 209 gynecological cancer patients were included in the present study. Of those, 43 women (20.6%) developed perioperative infections, including 16 patients with surgical site infection, 12 patients with urinary tract infections, 8 women with respiratory infections and 7 women with other causes. Preoperative PNI performed better than post-operative white blood cell count in detecting patients with postoperative infectious morbidity, however it was inferior to postoperative C-reactive protein (AUC: .562, .375 and .723 respectively). Classification and regression tree and random forest analysis achieved an outstanding performance in detecting the risk of perioperative infectious morbidity (AUC .979 and .990 respectively). PNI ranked first in the information gain and Gini coefficient analysis.
Conclusion Concluding, PNI may be able to predict postoperative morbidity in gynecologic oncology patients undergoing surgical procedures; however, its use as a single factor in a multivariate analysis setting has moderate predictive accuracy and should be avoided.
Disclosures The authors report no conflicts of interest. The present study was not funded.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.