|
Back to 2016 Annual Symposium ePosters
The Utility of Morphomics in Predicting Mortality after Open and Endovascular Abdominal Aortic Aneurysm Repair
Alyssa A. Mazurek, BS, Jeffrey F. Friedman, MSE, Ali Hammoud, Cameron Inglis, BA, Jacob Haugen, BS, Alex Hallway, Jason Lawton, June Ruan, Brian Derstine, MS, Jay S. Lee, MD, Stewart C. Wang, MD, PhD, Michael J. Englesbe, MD, Nicholas H. Osborne, MD.
University of Michigan Medical School, Morphomic Analysis Group, Ann Arbor, MI, USA.
OBJECTIVES:
Assessing operative risk in patients undergoing abdominal aortic aneurysm (AAA) repair has traditionally focused on pre-operative characteristics, anatomy, and comorbidities. However, these variables alone do not capture important aspects of health that influence mortality. Analytic morphomics is a novel method of risk-stratification that uses cross-sectional images to quantitatively measure domains of patient health. The benefit of morphomic measurements in addition to patient comorbidities for pre-operative risk stratification has been established in other populations. This study investigated whether morphomic measurements of core muscle size and adiposity improved the prediction of mortality in patients undergoing open and endovascular AAA repair.
METHODS:
A total of 795 patients underwent open or endovascular AAA repair between 2000 and 2012 at the University of Michigan. 722 patients (91%) had preoperative abdominal CT scans available for analysis. Manual chart review was used to identify patient demographics and comorbidities. Patients with incomplete medical records, an indication of rupture or previous aortic dissection or a suprarenal or juxtrarenal aneurysm were excluded. Analytic morphomics were used to measure cross-sectional areas and densities of psoas muscle, dorsal muscle group and subcutaneous and visceral fat at vertebral levels T12 through L5. Multivariate logistic regression models using both clinical variables and morphomics were then used to develop predictive models of mortality at 30, 90, 365 and 1095 days. Discrimination was assessed using c-statistics and model fit was assessed with Goodness of Fit (Hosmer -Lemeshow).
RESULTS:
559 patients underwent infrarenal AAA repair, including 270 patients undergoing endovascular repair and 289 patients undergoing open repair. No patients died within 30 days, 21 patients died within 90 days (4.0%), 29 patients within 180 days (5.7%), 46 patients within 365 days (9.29%) and 81 patients within 1095 days (18.7%). The predictiveness of models using clinical variables and patient comorbidities was moderate (c-statistic = 75% at all time periods). The addition of morphomic variables to the models improved the predictiveness (c-statistic = 81-89% for open and endovascular models 90-1095 days). All Hosmer-Lemeshow tests were non-significant.
CONCLUSIONS:
Morphomics are highly predictive of mid-to-late mortality following AAA repair. The addition of morphomic variables improves the discrimination of models to over 85%. Automated analysis of readily available CT scans may aid physicians in pre-operative decision-making for patients with infrarenal AAA.
Back to 2016 Annual Symposium ePosters
|
