New Model Improves Survival Predictions for Metastatic Prostate Cancer

A novel prognostic model that incorporates genetic alterations from circulating tumor DNA (ctDNA) may improve the way doctors predict overall survival (OS) in men with metastatic castration-resistant prostate cancer (mCRPC). This clinical-genetic (CG) model enhances an existing clinical model by integrating key genetic features, offering more precise survival predictions and better patient stratification.

Susan Halabi, PhD, a professor in the Department of Biostatistics and Bioinformatics, Duke Cancer Institute Center for Prostate and Urologic Cancers, at the Duke University School of Medicine in Durham, North Carolina, and colleagues developed this advanced model using data from the A031201 phase 3 trial, which tested enzalutamide with or without abiraterone in patients with mCRPC. Traditionally, the clinical model relied on variables such as performance status, disease site, opioid analgesic use, lactate dehydrogenase, albumin, hemoglobin, prostate-specific antigen, and alkaline phosphatase. The new CG model aims to improve prediction accuracy by adding genetic data.

In this study, presented at the 2024 American Society of Clinical Oncology Annual Meeting, 776 patients provided plasma samples that were analyzed using a 69-gene targeted DNA-sequencing assay to detect ctDNA pathogenic genetic alterations (PGAs). The researchers used a random survival forest method to identify the most significant genetic features, which were then included alongside the clinical variables in the final model. The model’s effectiveness was measured using the time-dependent area under the receiver operating characteristic curve (tAUC).

The enhanced CG model included several genetic markers, such as gains in AR and the AR enhancer, MYC, RSPO2, and losses or PGAs in ZBTB16, PTEN, MSH6, PPP2R2A, NKX3-1, TP53, FANCA, RB1, APC, CHD1, and BRCA2, as well as the ichorCNA tumor fraction. The CG model achieved a tAUC of 0.77, compared with 0.72 for the clinical model alone, indicating a significant improvement in predictive accuracy.

The study also classified patients into 3 and 4 prognostic risk groups based on their predicted risk, with notable differences in median OS and hazard ratios. In the 3-risk group model, patients in the low-risk category had a median OS of 58.9 months (95% CI), while those in the intermediate- and poor-risk categories had a median OS of 35.5 months and 19.3 months, respectively. The 4-risk group model offered even finer stratification, with the low-risk group showing a median OS of 64.2 months and the poor-risk group demonstrating a median OS of 17.0 months.

The findings reveal that the CG model not only identifies novel ctDNA PGAs that are prognostic of OS but also provides a robust method for classifying patients into risk groups. This study underscores the potential of ctDNA analysis in enhancing the precision of cancer prognostics and personalizing oncology care.