Pneumatosis Intestinalis Predictive Evaluation Study (PIPES)

Abstract

Pneumatosis intestinalis (PI) is associated with numerous adult conditions, ranging from benign to life threatening. To date, series of PI outcomes consist of case reports and small retrospective series. We conducted a retrospective multicenter study, involving eight centers, of PI from January 2001 to December 2010. Demographics, medical history, clinical presentation, and outcomes were collected. Primary outcome was the presence of pathologic PI defined as confirmed transmural ischemia at surgery or the withdrawal of clinical care and subsequent mortality. Forward logistic regression and a regression tree analysis was used to generate a clinical prediction rule for pathologic PI. During the 10-year study period, 500 patients with PI were identified. Of this number, 299 (60%) had benign disease, and 201 (40%) had pathologic PI. A wide variety of variables were statistically significant predictors of pathologic PI on univariate comparison. In the regression model, a lactate of 2.0 or greater was the strongest independent predictor of pathologic PI, with hypotension or vasopressor need, peritonitis, acute renal failure, active mechanical ventilation, and absent bowel sounds also demonstrating significance. Classification and regression tree analysis was used to create a clinical prediction rule. In this tree, the presence of a lactate value of 2.0 or greater and hypotension/vasopressor use had a predictive probability of 93.2%. Discerning the clinical significance of PI remains a challenge. We identified the independent predictors of pathologic PI in the largest population to date and developed of a basic predictive model for clinical use. Prospective validation is warranted. Epidemiologic study, level III.