Clinical Implications of Systemic Corticosteroid Treatment on Mortality and Surgical Outcomes in Patients with Gastrointestinal Perforation: A Retrospective Cohort Study

Abstract

Background:

Systemic corticosteroids are commonly used in critically ill patients. Although steroids may reduce the duration of shock and mechanical ventilation (MV), their use in postoperative surgical patients raises concerns regarding complications such as anastomotic leakage, impaired wound healing, and infection. This study aimed to evaluate the impact of postoperative corticosteroid therapy on clinical outcomes in patients undergoing surgery for gastrointestinal perforation and receiving intensive care.

Methods:

We conducted a single-center retrospective cohort study of 354 patients treated between January 2013 and March 2023. Postoperative corticosteroid use was identified, and clinical outcomes were compared between treated and untreated patients. To reduce confounding, propensity score matching was performed based on initial and postoperative sequential organ failure assessment scores and the Acute Physiology and Chronic Health Evaluation II score. A generalized linear model and a random forest (RF) regression model were applied to explore associations between steroid-related variables and the duration of MV.

Results:

Of the 354 patients, 37 (10.5%) received corticosteroids. In the unmatched cohort, the steroid group had higher mortality and complication rates. These differences were not significant after matching. In the matched cohort, the steroid group showed longer MV duration. No steroid-related variable was significantly associated with ventilation duration in regression analysis. Total hydrocortisone-equivalent dose emerged as the strongest predictor of prolonged MV duration in the RF model.

Conclusion:

Postoperative systemic corticosteroid therapy was independently associated with prolonged MV without adversely affecting mortality or overall morbidity. Although parametric analyses did not confirm a clear dose–response relationship, cumulative steroid dose demonstrated the highest predictive importance for ventilator dependence in machine learning modeling.