Session Item

Prostate cancer (PCa) patients, who receive salvage radiotherapy (sRT) due to biochemical recurrence (BCR) after surgery experience heterogeneous response rates. In search of novel non-invasive biomarkers, this study aims to develop a CT-based radiomic signature to predict BCR after sRT guided by positron-emission tomography targeting prostate-specific membrane antigen (PSMA).

Ninety-nine patients, who underwent 68Ga-PSMA11-PET/CT guided sRT from three high volume centers in Germany were included in this retrospective multicenter study. Patients had PET-positive local recurrences and were treated with intensity-modulated sRT. Patients with evidence of nodal or distant metastases in PSMA-PET or received ADT prior to PSMA-imaging were not eligible. PSMA-PET/CT were performed with iodine-based contrast-enhanced CTs using 120 kVp and exposure of 100–400 mAs (dose modulation) for attenuation correction. Images were acquired at portal venous phases approximately 80 seconds after injection of contrast agents. Median slice thickness of CT images was 3 mm (range 1.5 – 5) with a median voxel of 0.97 x 0.97 mm (range 0.68 x 0-68 – 1.17 x 1.17x).
Radiomic features were extracted from volumes of interests on CT, guided by focal PSMA PET uptakes. After pre-processing, clinical-, radiomics- and combined clinical-radiomics models were developed combining different feature reduction techniques and Cox proportional hazard models within a nested cross validation approach.

Median follow-up  (FU) was 29 (range 3-79) months, in which 26% of patients experienced BCR. The radiomic models outperformed clinical models and combined clinical-radiomics models for prediction of BCR with a C-index of 0.71 compared to 0.53 and 0.63 in the test sets, respectively. In contrast to the other models, the radiomic model achieved significantly improved patient stratification in Kaplan Meier analysis and improved receiver operator characteristics analysis (ROC) at 24 months of FU. The radiomic and clinical-radiomic model achieved a significantly better time-dependent net reclassification improvement index (0.392 and 0.762, respectively) compared to the clinical model. Decision curve analysis demonstrated a clinical net benefit for both models. Mean intensity was the most predictive radiomic feature.

Figure 1 shows results of the receiver operator characteristics analysis for the clinical signatures, radiomic signatures and clinical-radiomics signature at 24 months of follow up based on repeated nested cross validation (test results in the outer fold) results.

This is the first study to develop a CT-based radiomic model to predict BCR after PSMA-PET guided sRT. The radiomic models outperformed clinical models and might contribute to guide personalized treatment decisions.