HSR&D Citation Abstract
Search | Search by Center | Search by Source | Keywords in Title
Analytic morphomics predicts overall survival in patients with hepatocellular carcinoma.
Singal AG, Zhang P, Lakshmi A, Waljee AK, Sharma P, Barman P, Krishnamurthy V, Wang L, Wang SC, Su GL. Analytic morphomics predicts overall survival in patients with hepatocellular carcinoma. Poster session presented at: Digestive Disease Week Annual Conference; 2015 May 16; Washington, DC.
Analytic morphomics predicts overall survival in patients with hepatocellular carcinoma
View Session Detail
Presentation Number: Sa1857
Author Block: Amit G. Singal1 , Peng Zhang2 , Lakshmi Ananthakrishnan3 , Akbar K. Waljee2 , Pratima Sharma2 , Pranab Barman2 , Venkat Krishnamurthy2 , Lu Wang2 , Stewart C. Wang2 , Grace L. Su2
1 Internal Medicine, University of Texas Southwestern, Irving, Texas, United States; 2 University of Michigan, Ann Arbor, Texas, United States; 3 UT Southwestern Medical Center, Dallas, Texas, United States
Abstract: Background: There is no universally accepted system for accurately assessing prognosis among patients with hepatocellular carcinoma (HCC). Analytic morphomics, a novel process to measure body composition using computational image processing algorithms, has been used to link patient phenotypes to clinical outcomes in patients with cirrhosis and may offer a unique method to assess prognosis among HCC patients.
Aims: To derive and validate a prognostic model for patients with HCC using analytic morphomics and objective clinical information
Methods: Using computed tomography (CT) scans from a cohort of patients with newly diagnosed HCC at the Veterans Affairs Ann Arbor Healthcare System between January 2006 and December 2013, we derived a prognostic model using a combination of analytic morphomics and patient clinical data. Using elastic net regularization and cross validation for variable selection, we performed multivariate Cox regression to develop the model. We validated the model using an independent external cohort of racially diverse patients with HCC diagnosed at Parkland Health and Hospital System, the safety-net institution of Dallas County, between January 2005 and March 2012. Model performance was assessed in the derivation and validation cohorts using C-statistics and a modification of the net reclassification improvement.
Results: The derivation cohort consisted of 204 HCC patients (median age 60.8 years, 100% male, and 44% Caucasian) and validation cohort had 225 HCC patients (median age 57.0 years, 79% male, and 26% Caucasian). Median survival was lower in the validation cohort (5.4 vs. 16.8 months) related to higher proportions of Child Pugh C cirrhosis (25% vs. 12%) and lymph node and/or distant metastases (26% vs. 10%). We derived a model using analytic morphomics, routine clinical information and TNM stage with good prognostic accuracy (c-statistic 0.80, 95%CI 0.71 - 0.89) that was significantly better than either TNM (C-statistic 0.71, 95%CI 0.61 - 0.80, p = 0.007) or Barcelona Clinic Liver Cancer (C-statistic 0.66, 95%CI 0.56 - 0.76, p = 0.001) systems. Despite marked differences in tumor stages, liver dysfunction, and prognosis between the two cohorts, our model continued to perform with good accuracy in the external validation cohort (C-statistic 0.75, 95%CI 0.68-0.82), which was better than TNM (C-statistic 0.67, 95%CI 0.60 - 0.74) or BCLC (C-statistic 0.70, 95%CI 0.62 - 0.78) systems.
Conclusion: Analytic morphomics, combined with basic readily available clinical data, can serve as a highly accurate prognostic model for patients with HCC.