Back to 2008 Program and Abstracts
Signal Detection: a New Statistical Method to Predict Nash in Gastric Bypass Patients
John M. Morton*, Gavitt a. Woodard, Tina Hernandez-Boussard
Surgery, Stanford School of Medicine, Stanford, CA
Background: Non-alcoholic steatohepatitis (NASH) is associated with morbid obesity, cardiac risk factor abnormalities, and metabolic syndrome. Currently, there are no adequate pre-operative predictive models for NASH identification. We present a new statistical method to address this need.
Methods: Signal detection affords the ability to identify patients at risk with both homogeneous outcomes and risk predictors and has been used in cardiac risk assessment. Potential risk factors for NASH were entered into the Signal Detection model by Kraemer et al. All patients underwent laparoscopic gastric bypass surgery with intra-operative liver biopsy at a single academic institution by a single surgeon. Preoperatively, liver function tests, cardiac risk factors, and comorbidities were assessed. We dichotomized liver pathology into NASH vs Steatosis (SS) per Brunt criteria for analysis by T-test or Chi- Square analysis as appropriate.
Results: 141 patients successfully underwent laparoscopic Roux-en-Y gastric bypass surgery. Patient demographics were mean age, 45; female, 88%; mean BMI, 49; diabetic, 38%; hypertensive, 62%; and hepatitis or alcohol abuse, 0%. Liver biopsy results were Normal (2%), Steatosis (60%) and NASH (38%). The model identified the following variables in order as most discriminating in identifying patients with NASH: ALT>31, excess weight >109 kg, Lipoprotein A>21 and Hemoglobin A1C >6.9. This model identified this subgroup as having an 81% chance of having a pathologic NASH diagnosis.
Conclusion: In this study, ALT, excess weight and hemoglobin A1C were correlated with NASH. This innovative new statistical technique affords a new ability to successfully identify gastric bypass patients with biopsy proven NASH.
