Spatial transcriptomics-guided pathology biomarker as a predictor of benefit of adjuvant docetaxel in high-risk localized prostate cancer: NRG/RTOG 0521 (NCT00288080).

Abstract

5111 Background: The NRG/RTOG 0521 trial evaluated adding adjuvant docetaxel (DTX) to standard radiotherapy plus androgen deprivation therapy (ADT) in high-risk localized prostate cancer. Adjuvant DTX modestly improved overall survival (OS) in the trial, but the benefit was limited and not all patients benefited. Thus, biomarkers are needed to identify patients most likely to benefit from chemotherapy intensification. ST-DoxPCa (Spatial Transcriptomics–Guided Docetaxel Therapy Stratification in Prostate Cancer) is a novel artificial intelligence (AI) driven histology biomarker that predicts gene expression from H&E slides (virtual spatial transcriptomics). We assessed whether ST-DoxPCa can stratify patients in RTOG 0521 for differential benefit from adjuvant DTX. Methods: A Vision Transformer trained on paired histology–spatial transcriptomics (HEST1K) predicts a 208-gene prostate panel at spot level; predictions are distilled into a biologically informed 26-gene signature and aggregated into patient-level features. A prognostic Cox model and fixed median threshold were developed independently in a Cleveland Clinic radical prostatectomy cohort (CCF, n=352; endpoint: biochemical recurrence-free survival). This locked model and threshold were applied unchanged (no refitting or recalibration) to digitized pretreatment diagnostic biopsies from NRG/RTOG 0521 (n=350; RT+ADT n=169, RT+ADT+DTX n=181) to stratify patients into ST-DoxPCa-positive (high-risk) and ST-DoxPCa-negative (low-risk) groups. Overall survival (OS) was compared between treatment arms within each stratum. Results: ST-DoxPCa stratified patients into biomarker-defined risk groups using aggregated spatial-expression features from a biologically informed gene panel; key genes included PTEN, NKX3-1, ACPP, FASN and TMPRSS2. ST-DoxPCa-positive (high-risk) patients experienced a significant OS benefit from adding DTX to RT+ADT versus RT+ADT alone (HR=0.38, 95% CI 0.18–0.83; p=0.012), whereas ST-DoxPCa-negative (low-risk) patients derived no OS benefit (HR=1.05, 95% CI 0.67–1.63; p=0.84). Conclusions: The ST-DoxPCa model identified a subgroup with substantial OS benefit from adjuvant DTX and a subgroup with no benefit within RTOG 0521, supporting risk-aligned chemotherapy intensification using routine histology. Clinical trial information: NRG/RTOG 0521 ( NCT00288080 ) .