Objective: The purpose of this study was to use CT texture analysis to predict overall survival (OS) in patients with metastatic melanoma and stable disease (SD) according to the Response Evaluation Criteria in Solid Tumors (RECIST) on initial posttherapy CT images.
Materials and methods: This retrospective study included 42 patients with metastatic melanoma who received bevacizumab therapy in the context of a randomized prospective phase II clinical trial. Target lesions on the baseline and initial posttherapy contrast-enhanced CT examinations were evaluated by CT texture analysis using TexRAD software before and after image filtering in patients with RECIST SD on initial posttherapy images. Cox proportional hazards models were used to assess the associations of CT texture analysis measurements and of other patient factors with OS. The AUC was used to evaluate predictive accuracy.
Results: In multivariate analysis (in 23 patients with RECIST SD; median OS, 1.51 years), absolute change in mean positive pixels at spatial scaling filter of 4 mm, change in tumor size, and baseline serum lactate dehydrogenase (LDH) level were predictors of OS (hazard ratio [HR] = 5.05 for decrease in mean positive pixels at spatial scaling filter of 4 mm vs increase, p = 0.007; HR = 4.14 for > 5% increase in tumor size vs otherwise, p = 0.025; and HR = 1.29 for every 100 IU/L increase in baseline LDH level, p = 0.068). A prognostic index containing these three factors was highly accurate for predicting OS at 18 months (AUC = 0.917).
Conclusion: In patients with metastatic melanoma and RECIST SD on initial post-therapy CT images, a model incorporating CT texture analysis of target lesions, tumor size changes, and baseline LDH levels was highly accurate in predicting OS.
Keywords: CT texture analysis; antiangiogenic therapy; biomarker; metastatic melanoma.