477 Apples Are Higher Risk Than Pears: Subcutaneous Fat Measurement Using Analytic Morphomics and the Risk of Infectious Complications in Abdominal Surgeries for Crohn's Disease

Abstract

BACKGROUND/AIMS: In decision-making for medical vs. surgical therapy in Crohn's disease (CD), it is important to consider the likelihood of serious post-operative complications. Body habitus and BMI have been implicated as important factors for incurring post-operative complications, however these factors have undergone limited studies in CD. AnalyticMorphomics is a novel method of computational image analysis providing quantifiable and reproducible measurements of organs and body tissues. The aim of this study was to determine whether morphomic assessment of body tissues was predictive of major infectious complications following CD-related surgeries. METHODS: We conducted a retrospective review of a large tertiary center's medical records to identify adults with a diagnosis of CD (ICD 555.x) who underwent Crohn's-related abdominal surgery from 2000-2010. Charts were reviewed to verify the diagnosis of CD, medications, demographics, laboratory data, and outcomes. CT scans within 30 days of surgery underwent analytic morphomic analysis for fat and muscle quantification and total body cross-sectional area at several spinal levels. Surgical outcomes of 1) placement of a new post-operative abdominal drain; 2) use of intravenous antibiotics for greater than 10 days; or 3) re-operation within 30 days were considered infectious complications. Logistic regression was used to model these outcomes, focusing on clinical predictors and image-based measures of abdominal fat and muscle. The best models were chosen using pseudo-R^2 in Stata 11. RESULTS: There were 269 subjects identified with complete data on 229 subjects; infectious complications occurred in 23%. Univariate analysis showed hemoglobin, albumin, surgery type, distribution of subcutaneous fat and body area, but not psoas area, BMI, prednisone use, or anti-TNF use were predictors of infectious complications. A multivariate model including hemoglobin, the distribution of subcutaneous fat (differential of fat area between T12 and L3 spinal levels), and admission on a low risk day produced a c-statistic of 0.76. The PPV of the model is 71.4%, and the NPV was 81.4%. CONCLUSIONS: Patients with CD have predictable risk for infectious post-operative complications. Having a higher distribution of subcutaneous fat cross sectional area at T12 compared to L3 (an shaped distribution of abdominal fat) increases risk, along with low hemoglobin, and being admitted on high-risk days of the week (likely a proxy for urgent surgery). The apple distribution of abdominal fat could be a marker of past steroid use, metabolic syndrome, or vascular disease (and poor mesenteric blood flow). This is the first report of the distribution of abdominal fat as a risk factor for infectious complications of CD surgery, and the first application of analytic morphomics to surgery risk in Crohn's disease.