Quantitative MR imaging biomarkers of tumor heterogeneity predict prognosis in metastatic colorectal lesions.

e15121 Background: Intra-tumor heterogeneity is an independent determinant of patient survival outcomes in several tumor subtypes. Spatial variations in tumor enhancement on MRI is a macroscopic imaging marker of tumor heterogeneity. The goal of this study is to evaluate the potential role of MRI-based enhancement heterogeneity measures as predictors of survival in patients with metastatic colorectal cancer. Methods: We retrospectively analyzed T1-weighted contrast-enhanced MRI images of metastatic hepatic lesions in 41 patients (mean age 57.2 ±14.2 years) who were diagnosed with stage IV colorectal cancer between 2007 and 2013. Tumor Kras mutation status and patient survival data for up to 95 months was available for all patients. The largest metastatic hepatic lesion was identified by a radiologist and manually segmented. 14 Haralick texture features were extracted from each lesion. Cox proportional hazards regression analysis was used to assess the association between the enhancement heterogeneity measures and patient survival, with adjustment for Kras mutation status as a potential confounder. Backward stepwise feature selection was performed using p > 0.1 (Wald test) to select for statistically significant features. Results:Mean survival time was 39±3.9 months for the study population (51±4.1 months for Kras-wildtype and 37±5.1 for Kras-mutants). 68% of the patients had left-sided colon cancer; 21% had concurrent pulmonary metastases and 0.5% had concurrent brain metastasis. 61%, 27% and 12% of patients were on Flofox, Folfiri or other chemotherapeutic regimen, respectively. Texture matrix homogeneity (HR > 10; p = 0.016), inverse difference moment (HR < 0.1; p = 0.020), entropy (HR < 0.1; p = 0.033) and standard deviation (HR > 10; p = 0.006) exhibited significant independent association with patient survival. With the exception of entropy, these features maintained significant contribution to survival prediction independent of Kras mutation status. Conclusions: MRI-based quantitativeintra-metastasis heterogeneity measures are associated with patient survival and may add information beyond genetic mutation status to optimize prognosis prediction in metastatic colon cancer.