Modern methods of infrared (IR) spectroscopy yield full IR absorbance spectra in arrays, forming hyperspectral images. End-to-end processing of these images via deep learning seems ideal for exploiting their high dimensionality and wealth of spatial and spectral information, but recent research suggests that convolution-based architectures may have a spatial bias. Toward the goal of improved prostate cancer tissue classification, we compare a variety of deep learning classifiers for IR spectroscopy and probe the impact of a bottleneck which compresses the spectral dimension. We find a strong correlation between model spatial receptive field and classification performance, with the highest performance achieved by a modified Vision Transformer model. Conversely, we find only limited correlation between spectral information and deep learning model performance: we find that a spectral bottleneck of just 16 features has only a negligible effect on all neural network models, including convolution-eschewing transformer architectures and a multilayer perceptron model utilizing no spatial information. Rather than any particular network component inducing a spatial bias, the breadth of architectures exhibiting little dependence on spectral information implies that tissue classification itself is characterized by only a small set of spectral features. This, in turn, suggests that success at tissue classification may be a poor benchmark in the development of deep learning models designed to effectively utilize the spectral dimension.

173 Background: Sarcopenia is common in advanced prostate cancer and worsened by androgen deprivation therapy (ADT). Routine staging CT scans provide an opportunity to screen for sarcopenia. The STAMPEDE trials have demonstrated treatment intensification beyond ADT improve cancer outcomes; however, benefits are heterogeneous. This study investigates CT-derived skeletal muscle index (SMI), a validated marker of total muscle mass, as a prognostic and predictive biomarker in the STAMPEDE docetaxel and ARPI trials. Methods: Men with newly diagnosed non-metastatic high-risk (M0) and metastatic (M1) hormone-sensitive prostate cancer (HSPC) with available staging CT imaging in the STAMPEDE docetaxel or ARPI trials were included. These trials compared standard of care (SOC) with addition of docetaxel ± zoledronic acid (ZA) or abiraterone acetate with prednisolone (AAP) ± enzalutamide (Enz). SMI (cm²/m²) was calculated as mean muscle area divided by height squared. Outcomes were overall survival (OS) in M1 patients and metastasis-free survival (MFS) in M0. Prognostic utility of SMI was evaluated in Kaplan–Meier analyses and Cox regression models. Predictive value was assessed by comparing hazard ratios of the treatment effect in Cox models for high vs low SMI cohorts. Likelihood ratio tests were used to identify treatment-SMI interactions. Continuous predictive effects were examined using multivariable fractional polynomial interaction (MFPI) models. Results: 2,267 patients (1,578 M1, 689 M0) met inclusion criteria. The median SMI was 47.2cm²/m² (IQR 42-52) in M1 patients and 48.2cm²/m² (IQR 44-54) in M0. SMI and CHAARTED burden were identified as independent prognostic biomarkers in M1 patients, with a 10 cm²/m² increase in SMI associated with a 15% reduction in risk of death (HR 0.85, 95% CI 0.79–0.92, p<0.001). SMI was not an independent prognostic biomarker in M0 patients. SMI was identified as an independent predictive biomarker of MFS benefit upon addition of AAP±Enz in M0 patients; high SMI cohorts had greater MFS benefit from addition of AAP±Enz compared with the low SMI cohort (HR 0.44 [0.3-0.66] vs 0.59 [0.38-0.91]). Likelihood ratio tests confirmed that adding a treatment-SMI interaction improved prediction of treatment benefit (χ²: 4.67, p=0.03). MFPI modelling demonstrated a significant MFS benefit with addition of AAP±Enz in M0 patients (χ²=9.95, p=0.006) with increasing SMI, however this benefit was observed in the range of 41-63 cm 2 /m 2 only. 20% of our cohort lay outside this range (13% lower, 7% higher) and did not observe MFS benefit from addition of AAP±Enz to SOC. Conclusions: SMI is an independent prognostic biomarker in trial patients with de novo metastatic HSPC. SMI is predictive of improved MFS with addition of AAP±Enz to SOC in M0 disease. Further research is required to validate our findings in real-world cohorts.

BACKGROUND AND OBJECTIVE:To present a summary of the 2026 version of the European Association of Urology (EAU)-European Association of Nuclear Medicine (EANM)-European Society for Radiotherapy & Oncology (ESTRO)-European Society of Urogenital Radiology (ESUR)-International Society of Urological Pathology (ISUP)-International Society of Geriatric Oncology (SIOG) Guidelines on the treatment of relapsing, metastatic hormone-sensitive and castration-resistant prostate cancer (PCa). METHODS:The Panel performed a literature review of new data, covering the time frame between 2023 and 2025. The Guidelines were updated and a strength rating for each recommendation was added based on a systematic review of the evidence. KEY FINDINGS AND LIMITATIONS:Risk stratification for relapsing PCa after primary therapy may guide salvage therapy decisions. The range of treatment options for metastatic PCa has broadened, including androgen receptor pathway inhibitors (ARPI), metastasis-directed therapy, PARP inhibitors and their combinations, as well as PSMA-based therapy. The recommendations in the EAU Guidelines are based on clinical evidence and do not account for variations in cost, reimbursement structures, or resource availability across healthcare systems. CONCLUSIONS AND CLINICAL IMPLICATIONS:The evidence in the field of relapsing, metastatic, and castration-resistant PCa is evolving rapidly. These PCa Guidelines reflect the multidisciplinary nature of PCa management. A full version is available from the EAU Guidelines Office or online (http://uroweb.org/guideline/ prostate-cancer/).

Prostate cancers with high proliferation rates have shorter survival times but when spread has occurred (metastatic), there is increased sensitivity to hormone therapy with abiraterone.

OBJECTIVES:To evaluate the incidence of venous thromboembolism (VTE) after penile cancer surgery using national hospital data and to assess current thromboprophylaxis practices across UK specialist centres. SUBJECTS/PATIENTS AND METHODS:A retrospective cohort study was conducted using Secondary Uses Service (SUS) data on penile cancer surgeries performed in NHS hospitals in England between 2015 and 2024. A national survey of UK Penile Cancer Network (UK PeCaN) surgeons was undertaken to assess current thromboprophylaxis practice. The primary outcome was symptomatic VTE within 180 days of surgery, identified using International Classification of Diseases, 10th Revision (ICD-10) codes. Cumulative incidence of first postoperative VTE was analysed using patient-level time-to-event methods, with censoring at second surgery, 180 days or administrative end of follow-up. Survey responses were summarised using descriptive statistics. RESULTS:In this observational population-level study, 4310 patients underwent 5903 penile cancer-related procedures. A total of 143 VTE episodes were recorded over the 9-year period, corresponding to an overall crude incidence of 2.5%. In patient-level time-to-event analysis, the cumulative incidence of first postoperative symptomatic VTE was 0.21% at 30 days, 0.69% at 90 days and 1.08% at 180 days. Descriptive procedure-level analyses suggested higher unadjusted VTE rates following more extensive procedures, including lymph node dissection and total penectomy, although these estimates should be interpreted cautiously because of staged procedures and unmeasured patient-level confounding. Most VTE events occurred after hospital discharge. The survey, comprising 24 responses from 10 specialist centres, revealed substantial variation in thromboprophylaxis practice, with 71% of surgeons not using formal VTE risk assessment tools. CONCLUSIONS:Venous thromboembolism is an important postoperative complication after penile cancer surgery, particularly after more extensive procedures and staged treatment pathways. Current prophylaxis practices are inconsistent. These findings support the further collection and analysis of disease-specific and the development of procedure-specific guidelines recommending extended thromboprophylaxis in high-risk patients.

BACKGROUND:Serum prostate-specific antigen (PSA) concentrations decrease after hormone therapy for prostate cancer, with the nadir serving as a potentially useful prognostic biomarker. To support clinical use, we evaluated the association between PSA nadir values and survival outcomes, stratified by pre-treatment metastatic volume or, in patients with non-metastatic cancer, stratified by lymph node status. METHODS:As part of the STAMPEDE platform trial, patients with metastatic or very high-risk non-metastatic prostate adenocarcinoma were recruited to five randomised, controlled, phase 3 trials conducted at 126 hospitals or oncology centres in Switzerland and the UK. Patients were randomly assigned to either standard of care (androgren deprivation therapy [ADT] alone or ADT plus docetaxel) or to one of five experimental treatment groups: ADT plus docetaxel with or without zoledronic acid, ADT plus abiraterone acetate with or without enzalutamide, or ADT plus prostate radiotherapy (only patients with metastatic disease). We used trial data from these participants to perform landmark analyses to test associations of PSA at 6, 12, and 24 weeks after randomisation with overall survival. Only patients with a PSA value were included in each landmark analysis. The Kaplan-Meier method was used to estimate 96-month overall survival rates and the corresponding 95% CIs for patients categorised by either metastatic volume or lymph node status. The STAMPEDE protocol platform is registered with ClinicalTrials.gov (NCT00268476), EUDRACT (2004-000193-31), and ISRCTN (ISRCTN78818544). FINDINGS:This study included 7129 patients from the STAMPEDE platform, who were recruited between Oct 5, 2005, and Sept 2, 2016; 4438 had metastases and 2691 had very high-risk non-metastatic disease. Among patients with metastasis and volumetric assessment, 2211 (55·9%) of 3956 had high-volume metastases, and among those with non-metastatic disease, 1033 (38·4%) were lymph node positive. A PSA concentration of 0·2 ng/mL or less was less frequent at 6 weeks or 12 weeks, but was associated with equivalent survival rates, compared with a PSA of 0·2 ng/mL or less at 24 weeks. Survival rates of PSA subcategories (≤0·2 ng/mL, >0·2 to 1·0 ng/mL, >1·0 to 3·0 ng/mL, and >3·0 ng/mL) differed by metastatic volume or, in patients with non-metastatic disease, by nodal status. Survival was longest for patients allocated to abiraterone with or without enzalutamide. Among patients with metastatic disease in the abiraterone with or without enzalutamide group who had a PSA of 0·2 ng/mL or less at 24 weeks, 96-month overall survival in patients with low-volume metastatic disease (64·1% [95% CI 57·8-69·8]) was higher than in patients with high-volume metastatic disease (44·6% [37·1-51·9]), but lower than in patients with non-metastatic, node-positive disease (79·4% [73·8-83·9]). 96-month overall survival was highest for patients with non-metastatic, node-negative disease (82·8% [95% CI 78·7-86·1]). INTERPRETATION:Metastatic volume or nodal status influence survival rates associated with on-treatment serum PSA categories, including for undetectable PSA. Radiological features and serum PSA could be combined to better predict survival. PSA at 24 weeks showed strongest associations with overall survival, although a PSA concentration of 0·2 ng/mL or less at any timepoint predicted favourable outcome. These findings could inform prognosis and warrant evaluation for treatment selection in clinical trials. FUNDING:Cancer Research UK, Prostate Cancer UK, UK Medical Research Council, and John Black Charitable Foundation.

BACKGROUND AND OBJECTIVE:To present a summary of the 2026 version of the European Association of Urology (EAU)-European Association of Nuclear Medicine (EANM)-European Society for Radiotherapy and Oncology (ESTRO)-European Society of Urogenital Radiology (ESUR)-International Society of Urological Pathology (ISUP)-International Society of Geriatric Oncology (SIOG) guidelines on screening, diagnosis, and treatment of clinically localised prostate cancer (PCa). METHODS:The Panel performed a literature review of all new data published in English, covering the time frame between May 2023 and 2025. The Guidelines were updated, and a strength rating for each recommendation was added based on a systematic review of the evidence. KEY FINDINGS:A risk-adapted strategy for identifying men who may develop PCa is advised, generally commencing at 50 yr of age and based on individualised life expectancy. The use of multiparametric magnetic resonance imaging to avoid unnecessary biopsies is recommended. When a biopsy is considered, a combination of targeted and regional biopsies should be performed. A new five-tier EAU classification has been introduced. Prostate-specific membrane antigen positron emission tomography imaging is the most sensitive technique for identifying metastatic spread. Active surveillance is the appropriate management for men with low-risk PCa, as well as for patients with selected favourable intermediate-risk ISUP grade group 2 lesions. Local therapies are addressed, as well as the management of persistent prostate-specific antigen after surgery. A recommendation to consider hypofractionated radiotherapy in intermediate-risk patients is provided. Patients with cN1 PCa should be offered radiotherapy to the primary tumour combined with long-term intensified hormonal treatment. CONCLUSIONS AND CLINICAL IMPLICATIONS:The evidence in the field of diagnosis, staging, and treatment of localised PCa is evolving rapidly. These PCa Guidelines reflect the multidisciplinary nature of PCa management.

BACKGROUND:Effective prognostication improves selection of patients with prostate cancer for treatment combinations. We aimed to evaluate whether a previously developed multimodal artificial intelligence (MMAI) algorithm was prognostic in very advanced prostate cancer using data from four phase 3 trials of the STAMPEDE platform protocol. METHODS:We included patients starting androgen-deprivation therapy in the docetaxel, docetaxel plus zoledronic acid, abiraterone, or abiraterone plus enzalutamide trials. Patients were recruited at 112 sites. We combined all standard-of-care control patients (including those allocated to standard of care [SOC-ADT] consisting of testosterone suppression with luteinising hormone-releasing hormone agonists or antagonists, and radiotherapy when indicated), and we combined the rest of the patients into docetaxel-treated or abiraterone-treated groups. Patients had either metastatic disease or were at very high-risk of metastatic disease, determined by node-positivity or, if node-negative, by T stage, serum prostate-specific antigen (PSA) level, and Gleason score. We used the locked ArteraAI Prostate MMAI algorithm that combined these clinical variables, age, and digitised prostate biopsy pathology images. We performed Fine-Gray and Cox regression adjusted for treatment allocation and cumulative incidence analyses at 5 years to evaluate associations with prostate cancer-specific mortality (PCSM) for continuous (per SD increase) and categorical (quartile-Q) scores. The STAMPEDE platform protocol is registered with ClinicalTrials.gov, NCT00268476. FINDINGS:Of 5213 eligible patients recruited from Oct 5, 2005, to March 31, 2016, 3167 were included in this analysis (1575 [49·7%] with non-metastatic disease, 1592 [50·3%] with metastatic disease; median follow-up 6·9 years [IQR 5·9-8·0]) with all datapoints available for score generation. The MMAI algorithm (per SD increase) was strongly associated with PCSM (hazard ratio [HR] 1·40, 95% CI 1·30-1·51, p<0·0001). On ad-hoc inspection, the highest scoring quartile of patients in each disease and treatment allocation group (MMAI Q4; vs the bottom three quartiles, Q1-3) had the highest PCSM risk in both patients with non-metastatic disease (HR 2·12, 1·61-2·81, p<0·0001) and those with metastatic disease (HR 1·62, 1·39-1·88, p<0·0001). MMAI quartile stratification split patients categorised by disease burden into groups with notably different risks of 5-year PCSM: patients with non-metastatic disease that were node-negative could be further stratified by MMAI score quartile Q1-3 (3%, 2-4) versus Q4 (11%, 7-15), those with non-metastatic disease that were node-positive could be stratified by Q1-3 (11%, 8-14) versus Q4 (20%, 13-26), those with metastatic disease with low-volume could be stratified by Q1-3 (27%, 23-31) versus Q4 (43%, 36-51), and those with metastatic disease with high-volume could be stratified by Q1-3 (48%, 44-52) versus Q4 (68%, 62-75). INTERPRETATION:Diagnostic prostate biopsy samples contain prognostic information in patients with, or at high-risk of, radiologically overt metastatic prostate cancer. MMAI algorithm combined with disease burden improves prognostication of advanced prostate cancer. FUNDING:Prostate Cancer UK, UK Medical Research Council, Cancer Research UK, John Black Charitable Foundation, Prostate Cancer Foundation, Sanofi Aventis, Janssen, Astellas, Novartis, Artera.

5001 Background: The STAMPEDE trials showed that adding abiraterone acetate + prednisolone (AAP) ± enzalutamide (ENZ) to standard of care androgen deprivation therapy (SOC) improves metastasis-free survival (MFS) in high-risk non-metastatic (M0) prostate cancer (PCa) patients (pts). However, variable responses & adverse events underscore the need for prognostic & predictive biomarkers. We evaluated performance of a validated MMAI algorithm (ArteraAI Prostate Test v1.2) to identify pts who benefit most from the addition of AAP ± ENZ (ARPI). Methods: High-risk M0 STAMPEDE pts treated with SOC+ARPI (N=555) or SOC (N=781) with sufficient quality H&E biopsy images & clinical data (T stage, age, PSA) were included. MMAI score association with PCa specific mortality (PCSM, primary outcome measure) & distant metastasis (DM) was analyzed using Fine-Gray regression & cumulative incidence curves, with other cause mortality treated as competing risks. MFS was assessed using Cox regression & Kaplan-Meier curves. An optimal cut-point was identified via grid search to maximize ARPI benefit separation across biomarker positive (pos, MMAI in top quartile) & negative (neg) subgroups. Hazard ratios [95% CI] & p values are reported. Results: PCSM median follow-up was 6.0 years (N=1336). Continuous MMAI scores were statistically significantly associated with poorer PCSM (1.65 [1.43-1.90], p<0.001), MFS (1.42 [1.29-1.56], p<0.001) & DM (1.54 [1.36-1.74], p<0.001). Using clinically-established prognostic cut-offs, 89% of pts were MMAI high-risk. The optimal ARPI MMAI cut-point identified 334 biomarker-pos pts who had significantly higher PCSM than biomarker-neg pts. A statistically significant biomarker-treatment interaction for PCSM (p-int=0.04) revealed that biomarker-pos pts treated with ARPI had improved PCSM (0.42 [0.24-0.74], p=0.003), while biomarker-neg pts did not derive a treatment benefit (0.85 [0.56-1.29], p=0.45). Estimated 5-year PCSM was 9% for biomarker-pos pts receiving ARPI vs. 17% with SOC, compared to 4% & 7% for biomarker-neg pts, respectively, with similar results observed in M0N0 pts (Table 1). Conclusions: For the first time, we demonstrate that a validated MMAI algorithm can identify high-risk non-metastatic PCa pts most likely to benefit from the addition of ARPI. Notably we identify a positive biomarker-treatment interaction in the highest MMAI score quartile, which in cases of clinical equipoise could inform clinical decision-making. We highlight MMAI’s potential to optimize treatment decisions & spare biomarker-neg pts from unnecessary therapy & toxicities. Clinical trial information: NCT00268476 . Estimated 5-yr absolute risk reduction from ARPI vs SOC-treated patients by biomarker groups in M0 (M0N0) pts. Biomarker-neg Biomarker-pos PCSM 3% (1%) 8% (9%) MFS 2% (-1%) 17% (16%) DM 5% (3%) 12% (15%)

5003 Background: Docetaxel (Doce) is effective for metastatic (M1) PC but its effect is varied. Combining Doce and hormone therapy can improve overall survival (OS) but is not appropriate for all. We previously reported the mRNA Decipher test predicts benefit from Doce. We then used transcriptome-wide data to interrogate differential associations with outcome for biologically-relevant pathways. Methods: PTEN inactivation using a previously described signature (Liu et al JCI, 2021; active score <= 0.3, inactive: score > 0.3) and Decipher score (high > 0.8, lower <= 0.8) was determined from transcriptome-wide expression data generated in a clinically-accredited lab on prostate tumor from M1 patients (pts) randomized 1:1 to ADT vs ADT + Doce +/- zoledronic acid (ZA) or ADT vs ADT + Abi (abiraterone acetate + prednisone) in the STAMPEDE protocol (Oct 2005-Jan 2014). Cox survival models were fitted with an interaction between treatment allocation and PTEN activity, adjusted for age, WHO PS, pre-ADT PSA, Gleason score, T-stage, N stage (N0, N1), metastatic volume (CHAARTED definition, high [HV] or low [LV]). Hypotheses were tested using partial likelihood ratios. Primary endpoint was OS. Results: We generated transcriptome-wide profiles on 832 M1 pts with no notable differences from the full M1 trial cohort (N=2224). 657 (79%) were reported to have died. 50% of tumors were classified as PTEN inactive (N=419). PTEN mRNA score distribution was similar across HV and LV disease (p=0.310). PTEN inactivity associated with shorter OS in pts allocated ADT+Abi (N=182; HR=1.56, 95%CI: 1.06-2.31) but not in pts allocated ADT+Doce+/-ZA (N=279; HR=0.93, 95%CI: 0.70-1.24). We found strong evidence (p=0.002) of an interaction between PTEN inactivation and Doce sensitivity: PTEN inactive pts benefited from Doce (HR=0.57, 95% CI 0.42-0.76) unlike PTEN active pts (HR=1.05, 95% CI 0.77-1.43). This was consistent in LV (N=244; PTEN inactive HR=0.53, 95% CI 0.33-0.86; PTEN active HR=0.82, 95% CI 0.48-1.40) and HV (N=295; PTEN inactive, HR=0.59, 95% CI 0.39-0.88; PTEN active HR=1.23, 95% CI: 0.83-1.81). In pts randomized to Abi, treatment effect was uniform (PTEN inactive HR=0.52, 95% CI 0.36-0.73; PTEN active HR=0.55, 95% CI 0.39-0.77; p=0.784). We estimated adding Doce for tumors classified as PTEN inactive and high Decipher reduced the hazards of death by 45% (HR 0.55, 99% CI 0.34-0.89). Conclusions: Prostate tumors classified as high Decipher and PTEN inactive have a 45% reduction in hazard of death when Doce is added to ADT. This biomarker should be tested in pts considered for triplet therapy of ADT + Abi + Doce.

e17073 Background: The NCCN began recommending germline genetic testing for individuals with high-risk or advanced prostate cancer (PCa) in 2017. In 2019, our survey of physicians within the PCCTC showed that the adoption of germline testing was not widespread: 62% of physicians considered testing for all men with metastatic PCa (mPCa) and 12% considered testing for men with high risk localized PCa. In 2024 we conducted a follow up survey to evaluate changes in perception of germline testing uptake, clinic workflow and barriers. Methods: A 22-item survey was distributed via email to physicians associated with the PCCTC and was open for responses between 5/24/2024 and 7/11/2024. We collected data on physicians’ personal practices around germline testing in PCa, perceived barriers to testing and uptake of the testing. Results: 379 physicians received the germline genetic testing current practice survey, and 90 (24%) completed the survey. A total of 40% (36/90) of participating physicians reported that they refer eligible patients to a separate department for genetic testing and counseling, and more than half (70%) reported taking personal responsibility for some or all genetic education and testing of their eligible patients. 21% (19/90) reported obtaining genetic testing results through patient enrollment in a research study. When asked for a best estimate of all PCa patients who meet current NCCN criteria for germline testing who have completed testing, 28% (22/80) of physicians responded with a range of 51-75%. The remaining responses of participating physicians are split with 26% estimating >76%, 25% estimating 26-50%, and 21% estimating <26% of patients who qualify receive germline genetic testing. The majority of responders reported considering some mPCa patients for germline genetic testing: 81% reported considering all mPCa patients; 34% considered testing mPCa patients with a family history, and 38% considered testing only those for whom results would influence treatment (FDA-approved targeted therapy or trial candidates). More than half of the participating physicians (54%) considered germline genetic testing for some PCa patients with high-risk localized PCa. Based on participant responses and free text comments, cited barriers in streamlining genetic testing were: clinical workflow, time and space availability, access to genetic counselors, out of pocket cost, insurance coverage, access to resources for provider/patient education, among others. Conclusions: Physician uptake of germline genetic testing has increased since our initial survey in 2019. A greater percentage of participating physicians report ordering germline testing in their clinic and providing pre- and post-test counseling. Clinic flow and access to genetic counselors remain prominent barriers.

One of the major limitations for clinical applications of infrared spectroscopic imaging modalities is the acquisition time required to obtain reasonable images of tissues with high spatial resolution and good signal-to-noise ratio (SNR). The time to acquire a reasonable signal to noise spectroscopic scan of a standard microscope slide region of tissue can take many hours. As a trade-off, systems can allow for discrete wavenumber acquisitions, sacrificing potentially vital chemical bands in order to reach specific acquisition targets. Recent instrumentation developments now allow for the full fingerprint imaging of entire microscope slides in under 30 minutes, enabling rapid, high quality spectroscopic imaging of tissues within clinical timeframes without sacrificing frequency bands. Here we compare the data from a novel QCL microscope to an FTIR microscope covering multiple aspects of spectroscopic imaging of a large, clinically relevant, prostate cancer tissue cohort (N = 1281). Comparisons of hyperspectral data acquisition quality in both achieved signal to noise and image contrast alongside the capacity for unsupervised and supervised modelling of tissue constituents are reported. We conclude that it is now possible to collect full fingerprint spectra and derive clinically relevant data in a timeframe suitable for translation into the pathology laboratory without the need to resort to discrete frequency imaging with subsequent loss of information.

Advanced prostate cancers respond to hormone therapy but outcomes vary and no predictive tests exist for informed treatment selection. To identify novel biomarker-treatment pairings, we examined associations between biological pathways and 14-year survival outcomes of patients randomized in practice-changing phase 3 trials (testing docetaxel or abiraterone). We included transcriptome-wide expression signatures and immunohistochemistry markers (Ki-67 and PTEN) on prostate tumors from 1,523 patients (832 metastatic). Tumor androgen receptor signaling is associated with longer survival, whereas increased proliferation predicted shorter survival. In a pre-specified analysis, the previously identified decipher RNA signature was both prognostic and predicted survival benefit from docetaxel for metastatic cancers (biomarker-docetaxel interaction p = 0.039). Additionally, transcriptome-based classification of PTEN inactivation identified tumors more likely to have PTEN protein loss (p = 4 × 10-37) and metabolically perturbed metastatic cancers that had shorter survival with hormone therapies (p < 0.001) but exhibited docetaxel sensitivity (biomarker-docetaxel interaction p = 0.002). Transcriptome classifiers predict docetaxel benefit and could be clinically implemented for improved patient management.