Predictive Models for Assessing Risks of Postoperative Cardiac and Respiratory Complications in Patients With Lung Cancer

Abstract

The aim of the study is to identify risk factors for the development of postoperative cardiac and respiratory complications in patients after lobectomy and to develop nomograms to predict such complications.
 Materials and Methods. From January 2020 to August 2023, 411 lobectomies were performed in patients with non-small cell lung cancer in the surgical department of thoracic oncology, Regional Clinical Oncology Dispensary (Ulyanovsk). The present study included 162 patients according to the designed enrolment criteria. The authors identified statistically significant risk factors for the development of severe cardiac and respiratory complications after lobectomy with systematic lymph node dissection.
 Results. Due to insufficient data cardinality, artificial data with a similar distribution were synthesized using the MICE method (multivariate imputation by chained equations). The sample is divided into training (n=1258) and validation (n=342) datasets at a ratio 80 to 20. Final testing of the initial data was carried out (n=162).
 Logistic regression for respiratory complications: respiratory complications = 23.4232 – 0.7427 × stair climbing test – 0.0259 × shuttle walking test. Based on these data, a nomogram was constructed to predict the development of respiratory complications. Error matrix characteristics: accuracy – 0.95; sensitivity – 0.86; specificity – 0.97; positive predictive value – 0.86; negative predictive value – 0.97.
 Logistic regression for cardiac complications: cardiac complications = 11.1147 – 0.1247 × stair climbing test – 0.1651 × ejection fraction + 0.1568 × BMI + 1.133 × hypertension. Based on these data, a nomogram was constructed to predict cardiac complications. Error matrix characteristics: accuracy – 0.91; sensitive-
 ty – 0.63; specificity – 0.98; positive predictive value – 0.9; negative predictive value – 0.92. 
 Conclusions. The developed prognostic models allow us to determine the risk group among patients who are planning a lobectomy for lung cancer at the preoperative stage. It also allows us to ensure high-quality prevention for these complications in patients at high risk.