Association of 18F-FDG PET characteristics and survival outcomes using whole body tumor analysis in patients (pts) with metastatic genitourinary (GU) malignancies.

Abstract

4600 Background: 18F-FDG PET scans are widely used for staging and monitoring pts with metastatic GU malignancies. In addition to identifying lesions, 18F-FDG PET scans provide semi-quantitative parameters (SUVmax, SUVmean, total lesion glycolysis (TLG)) that can be used to characterize disease burden and treatment response. These parameters can demonstrate how treatment response varies across different patients as well as across different lesions within the same patient. This study was done to determine which parameter provides the best prognostic information for pts with metastatic GU malignancies. Methods: 101 patients with metastatic GU malignancies from two separate, prospective trials with at least two 18F-FDG PET scans were analyzed. All regions of interest were contoured for both the baseline and follow-up 18F-FDG PET scans. Regions of interest (ROI) were quantified and matched between timepoints using TRAQinform IQ technology (AIQ Solutions) to identify which were new, increasing, stable, decreasing, or disappeared. Imaging features were extracted from each patient, including basic features (SUVmax, SUVmean, TLG, and changes in these features) and heterogeneity (intrapatient heterogeneity of disease and response) features. Cox regression was used to determine univariate predictive power of each measure. The TRAQinform Profile was calculated to predict either progression-free (PFS) or overall survival (OS) using 5-fold cross-validation of a random survival forest. The performance of individual features and the TRAQinform Profile was evaluated using the c-index. Results: 54pts were treated with single agent cabozantinib (cabo) on a phase II trial and 47 pts were treated with cabozantinib and nivolumab, with or without ipilimumab (cabonivo +/- ipi), on a phase I trial. Most pts had urothelial carcinoma (N = 65). Other histologies included prostate (4), penile (1), renal cell carcinoma (RCC) (6), renal medullary carcinoma (4), testicular (4), adenocarcinoma of the bladder (8), squamous cell carcinoma of the bladder (2), sarcomatoid bladder (1), sarcomatoid RCC (1), and small cell carcinoma (4 bladder, 1 prostate). 2389 ROIs were analyzed, with a range of 1 to 151 per patient. The TLG at the follow-up image was the strongest predictor of both PFS (c-index = 0.62) and OS (0.69), followed by number of increasing ROI (0.61, 0.64), and percentage of all ROIs classified as new or increasing (0.57, 0.63). SUVmax (0.49, 0.47) and SUVmean (0.47, 0.56) at follow-up were found to be weaker predictors of PFS and OS. TRAQinform Profile was able to predict the responder’s vs suboptimal responders to study treatment (c-index = 0.73, 0.75). Conclusions: The TRAQinform Profile analysis of 18F-FDG PET scans provided both prognostic and predictive information for patients with metastatic GU malignancies treated with either cabo or cabonivo +/- ipi.