Abstract
Background Accurate prediction of the response of patients with metastatic castration-resistant prostate cancer (mCRPC) to lutetium 177 prostate-specific membrane antigen-617 (Lu-PSMA) therapy would be helpful for treatment selection. Purpose To evaluate the incremental value of total tumor volume (TTV) in prediction models developed for patients receiving Lu-PSMA. Materials and Methods Previously developed prediction models using baseline variables (time since diagnosis, chemotherapy status, hemoglobin, dichotomized number of metastases, tumor mean standardized uptake value, and bone and liver involvement) were externally validated in patients treated with Lu-PSMA between June 2022 and April 2024. The concordance index (ie, C index) for overall survival (OS), prostate-specific antigen (PSA) progression-free survival (PSA-PFS), and 50% or greater decline in PSA from baseline (PSA50) was assessed. The predictive value of prostate-specific membrane antigen PET/CT-derived TTV was evaluated by re-estimating previous models (Cox regression for OS and PSA-PFS; logistic regression for PSA50) and replacing the dichotomized number of metastases with the continuous number of metastases or TTV. Results Data from 168 patients (median age, 73 years [IQR, 68-78 years]) were analyzed. External validation of the original models achieved C indexes of 0.70 (95% CI: 0.64, 0.77) for OS, 0.68 (95% CI: 0.63, 0.73) for PSA-PFS, and 0.72 (95% CI: 0.64, 0.80) for PSA50. Replacing the dichotomized number of metastases with TTV increased the C index for OS by 0.04 (95% CI: 0.01, 0.07; P = .002). A higher TTV was associated with shorter OS (hazard ratio, 1.35 [95% CI: 1.08, 1.68]), whereas there was no evidence of an independent association between dichotomized number of metastases and OS (hazard ratio, 1.23 [95% CI: 0.64, 2.36]; P = .53). A higher TTV was associated with shorter PSA-PFS (hazard ratio, 1.25 [95% CI: 1.10, 1.42]) and lower PSA50 (odds ratio, 0.70 [95% CI: 0.53, 0.91]). Conclusion Previously developed models predicted OS, PSA-PFS, and PSA50 in patients with mCRPC receiving Lu-PSMA. The inclusion of TTV produces prediction models with better discriminatory ability than the dichotomized or continuous number of metastases. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Vargas and Woo in this issue.