Prediction Model for Severe Complications in Patients with Failed Conservative Treatment of Acute Intestinal Obstruction Based on Clinical and Imaging Features: Development and Validation

Technical Summary: Prediction Model for Severe Complications in Patients with Failed Conservative Treatment of Acute Intestinal Obstruction

Problem Statement
Acute intestinal obstruction is a prevalent surgical emergency, with postoperative adhesions being the leading cause. While nonoperative (conservative) management is the standard initial strategy for adhesive obstruction without signs of strangulation, approximately 20–30% of patients fail this approach, requiring surgery due to unresolved obstruction or the development of severe complications such as bowel necrosis, perforation, sepsis, and reoperation. A critical clinical challenge is identifying which patients will develop these severe complications (defined as Clavien–Dindo grade ≥IIIa) during the window of conservative treatment. Delayed surgical intervention in these high-risk patients is strongly associated with increased morbidity and mortality. Current clinical judgment relies on subjective interpretation of isolated signs (e.g., peritonitis, fever, leukocytosis) or single imaging findings, which often lack sufficient sensitivity and specificity. Furthermore, existing prediction models primarily focus on the binary need for surgery rather than stratifying the risk of severe, life-threatening complications.

Methodology
This study employed a clinical prediction study design involving a retrospective cohort of 366 patients admitted to The People's Hospital of Leshan between January 2018 and January 2026.

• Population: Patients aged ≥18 with acute intestinal obstruction (adhesive or tumor-related) who underwent at least 24 hours of conservative treatment and subsequently failed, necessitating surgery. Patients presenting with immediate strangulation, necrosis, or perforation were excluded.
• Outcome Variable: The primary endpoint was the development of severe complications (Clavien–Dindo grade ≥IIIa), encompassing intestinal necrosis, perforation, sepsis, and reoperation.
• Predictors: A comprehensive set of variables was collected, including demographics, history (abdominal surgery, chronic disease), vital signs, laboratory parameters (NLR, CRP), and specific CT findings (bowel dilation, wall thickening, fecal sign, mesenteric vascular sign, ascites).
• Statistical Analysis:
  • Feature Selection: Least Absolute Shrinkage and Selection Operator (LASSO) regression was utilized to select the most relevant predictors from the candidate pool.
  • Model Construction: Multivariate logistic regression was performed on the selected variables to construct a prediction model.
  • Visualization: A nomogram was developed to visualize the logistic regression results for clinical application.
  • Validation: Model performance was assessed using Receiver Operating Characteristic (ROC) curves and calibration plots. Internal validation was conducted using 10-fold cross-validation to calculate the C-index. Decision Curve Analysis (DCA) was performed to evaluate the net clinical benefit.

Key Contributions

1. Identification of Independent Predictors: The study identified six independent predictors of severe complications in patients failing conservative treatment:
   • History of abdominal surgery.
   • Development of peritonitis signs during conservative treatment.
   • Increased heart rate.
   • Elevated 24-hour increase in Neutrophil-to-Lymphocyte Ratio (NLR) after admission.
   • Presence of the "fecal sign" on CT.
   • Presence of the "mesenteric vascular sign" on CT.
2. Development of a Quantitative Tool: The authors constructed a nomogram integrating clinical, laboratory, and imaging features to provide an objective, quantitative risk assessment tool.
3. Shift in Endpoint Focus: Unlike previous models focusing solely on the indication for surgery, this model specifically targets the prediction of severe complications (Clavien–Dindo ≥IIIa), aiming to prevent irreversible outcomes like necrosis and sepsis.

Results

• Predictor Performance: Multivariate logistic regression confirmed that all six selected variables were statistically significant independent predictors. Notably, the development of peritonitis signs during treatment showed the strongest association (Odds Ratio [OR] = 9.51), followed by the mesenteric vascular sign (OR = 4.80) and the fecal sign (OR = 3.94).
• Model Discrimination: The final prediction model demonstrated excellent discrimination, with an Area Under the Curve (AUC) of 0.894 (95% CI: 0.851–0.936).
• Internal Validation: The 10-fold cross-validation yielded a C-index of 0.872 (95% CI: 0.825–0.920), indicating robust performance and good generalizability within the dataset.
• Calibration and Utility: Calibration curves showed excellent agreement between predicted probabilities and observed outcomes. Decision Curve Analysis confirmed that the nomogram provides a significant net benefit across a range of threshold probabilities compared to treating all or no patients.

Significance and Claims
The authors claim that this study successfully developed and internally validated a prediction model that effectively stratifies the risk of severe complications in patients with acute intestinal obstruction who fail conservative treatment. By integrating dynamic clinical changes (such as the 24-hour NLR increase and development of peritonitis) with specific CT findings (fecal and mesenteric vascular signs), the model offers a more accurate and objective decision-support tool than traditional subjective judgment.

The significance of this work lies in its potential to facilitate the early identification of high-risk patients. The authors posit that using this nomogram could help clinicians adjust treatment strategies timely—such as shortening observation intervals or proceeding to earlier surgical intervention—thereby reducing the incidence of severe complications like bowel necrosis and sepsis. The study emphasizes that while the model is robust internally, its clinical utility is maximized when used to support, rather than replace, clinical judgment, particularly for patients in the "gray zone" where the decision to operate is difficult. The authors acknowledge the study's limitations, including its retrospective, single-center nature, and call for future multicenter prospective validation to confirm external validity.