Abstract
Background and objective:
This study aimed to develop and validate a prognostic model for overall survival (OS) in men with de novo metastatic hormone-sensitive prostate cancer (mHSPC), using clinical factors from phase III trials to improve survival prediction, guide clinical decision-making, and optimize trial design.
Methods:
We analyzed individual patient data (IPD) from three randomized phase III trials (four comparisons) of androgen deprivation therapy (ADT) ± docetaxel: CHAARTED (n=575), GETUG-15 (n=272), and STAMPEDE (arm A, n=689; arm C, n=347; arm E, n=351). A proportional hazards model was developed using CHAARTED and GETUG-15 as the training set and validated in STAMPEDE. Model discrimination performance was assessed by time-dependent area under the receiver operating characteristic curve (tAUC).
Key findings and limitations:
Lower level of hemoglobin, High disease volume, elevated alkaline phosphatase, and poor ECOG performance status were associated with increased risk of death. The model achieved tAUC values of 0.72 (95% CI: 0.69–0.76) and 0.70 (95% CI: 0.67–0.72) in STAMPEDE A vs. C and STAMPEDE A vs. E comparisons. Risk stratification into two or three groups showed clear differences in OS, with substantially longer OS in low-risk patients. Moreover, patients in the poor-risk group derived the greatest benefit from docetaxel, while low-risk patients derived minimal benefit. A limitation is that model was built in de novo mHSPC men.
Conclusions and clinical implications:
This validated model improves OS prediction in de novo mHSPC and supports risk-informed treatment selection. Future work should focus on external validation in contemporary settings.
Patient Summary:
This study developed and validated a model to predict survival in men with newly diagnosed metastatic hormone-sensitive prostate cancer. Using data from major clinical trials, the model identified key factors—disease volume, lab values, and performance status—that influence survival and treatment benefit. The findings may help doctors personalize treatment decisions and improve outcomes, especially in selecting patients who benefit most from chemotherapy.
Keywords: mHSPC, predictive model, overall survival, high disease volume
Introduction
Substantial efforts have been dedicated to identifying prognostic factors and developing models to predict overall survival (OS) in men with metastatic castration-resistant prostate cancer (mCRPC) [1–16]. Established prognostic factors of OS in mCRPC include prostate-specific antigen (PSA), alkaline phosphatase, serum lactate dehydrogenase (LDH), hemoglobin, albumin, pain, body mass index, and functional status [17]. The spread of cancer to lymph nodes, bones, and visceral organs (liver, lungs) is also significant, as it reflects metastasis biology and disease burden [18, 19]. Additionally, circulating biomarkers such as circulating tumor cell counts, circulating tumor DNA (ctDNA) fraction, tumor genetic subtypes from ctDNA, serum androgen levels, and plasma angiokines help differentiate treatment outcomes among mCRPC patients [20].
A few studies identified bone disease localization, performance status, PSA, and Gleason score, or grouping patients into good, intermediate, and poor prognosis categories as important prognostic factors in men with mHSPC [21–25]. More recently, investigators from the CHAARTED [26] and GETUG-15 [27] trials have identified prognostic factors, including de novo versus metachronous metastases, high versus low volume disease and body mass index.
We hypothesize that a composite of clinical factors including performance status, presence of liver metastases, hemoglobin, PSA, alkaline phosphatase, Gleason score, and the number of bone metastases, will more accurately predict OS in men with mHSPC than the currently used simple model of timing and volume of metastases. Our objective is to develop and validate a model predicting OS in men with de novo mHSPC using data from phase III clinical trials. By incorporating a broad range of clinical factors, this model aims not only to improve the precision of survival predictions but also to guide optimal treatment selection based on patient risk profiles.
Materials (Patients) and Methods
Trials and Patients
Individual patient data (IPD) from three randomized phase III trials (four comparisons) were utilized to develop and validate the predictive model of OS in men with mHSPC. Two of these trials, GETUG-15 (NCT00104715) [28] and CHAARTED (NCT00309985) [29], compared the standard of care (SOC) with SOC plus docetaxel, whereas STAMPEDE (NCT00268476) [30] included comparisons of SOC versus SOC plus docetaxel (arms A vs. C, n=1086 patients) and SOC versus SOC plus zoledronic acid (ZA) and docetaxel (arms A vs. E, n=1090 patients). Eligible individuals for this analysis were patients who were diagnosed with de novo mHSPC. Since the GETUG-15 trial was older but similar in design to CHAARTED, we used them as the training set for our model building. This approach helps minimize potential bias related to differences in study enrollment periods. The more recent STAMPEDE trial was then used as an independent test set to assess the predictive accuracy of our model. The Duke University Medical Center Institutional Review Board approved this analysis.
Statistical Analysis
The primary endpoint was OS, defined as the time from the date of randomization to the date of death or last follow-up. Ten predictors of OS with a non-missing rate exceeding 50% in at least one of the four comparisons were considered to have sufficient information for the data imputation as higher missing rate will require more auxiliary variables for imputation. These variables were age, WHO/ECOG performance status, Gleason score, clinical T stage, nodal status, opioid analgesic use, disease volume, hemoglobin, PSA, and alkaline phosphatase. High volume was defined according to the CHAARTED criteria as the presence of visceral metastases or ≥4 bone lesions, with at least one located outside the vertebral bodies and pelvis.
We applied the least absolute shrinkage and selection operator (LASSO) [31] to 10 candidate variables, identifying eight predictors: age, WHO performance status, disease volume, Gleason score, opioid analgesic use, hemoglobin, alkaline phosphatase, and PSA. A proportional hazards (PH) model was fit in each imputed training dataset using these variables. PSA and alkaline phosphatase were log-transformed, with PSA modeled using restricted cubic splines to allow for nonlinearity; all other covariates were modeled linearly. Hazard ratios were obtained by averaging coefficients across imputations, with confidence intervals estimated via bootstrap. Model parameters were applied to the external STAMPEDE dataset to generate predicted risk scores, averaged across imputations. Model discrimination was assessed using the time-dependent area under the ROC curve (tAUC) in the test set and across clinically relevant subgroups [32]. Calibration was evaluated by comparing predicted and observed mortality at 42 and 54 months, because the observed median OS across the four randomized trials fell within this range. Patients were classified into two or three prognostic risk groups using cut points derived from cross-validation or by the number of adverse factors. Lastly, risk group–by–treatment interactions were included in PH models to estimate treatment-specific hazard ratios within each risk group. Treatment effects were pooled across trials using a two-stage random-effects meta-analysis in the training set, and estimated using a multi-arm PH model with pooled individual-level data in the test set (Supplementary Methods).
Results
Baseline characteristics
Of 2,627 patients, 2,234 with de novo mHSPC treated with ADT alone or ADT plus docetaxel ± ZA were included (Supplementary Fig. 1), comprising 847 patients in the training set and 1,387 in the external validation set. Baseline characteristics are summarized in Table 1; patients from the two STAMPEDE comparisons were similar, while differences were observed across trials in disease volume, PSA, and alkaline phosphatase levels, with substantial missing data requiring imputation (Supplementary Table 1). Overall, 79.2% of patients had Gleason scores 8–10, 63.5% had high-volume disease, and 71.2% had ECOG performance status 0. A total of 1,419 deaths occurred: 186 in GETUG-15, 320 in CHAARTED, and 479, 213, and 221 in STAMPEDE arms A, C, and E, respectively with a median follow-up among surviving patients was 72.6 months (IQR: 59.5–90.4) (Fig. 1A).
Table 1.
Baseline Characteristics of de novo Patients of the three randomized clinical trials (four comparisons) summarized over multiple imputations. The counts in the table have been averaged over 50 imputed datasets.
| CHAARTED | GETUG-15 | STAMPEDE (A vs C) |
STAMPEDE (A vs E) |
Total | |
|---|---|---|---|---|---|
| (N=575) | (N=272) | (N=1036) | (N=1040) | (N=2234) | |
| Median Age (IQR) | 62.0 (55.0, 69.0) | 62.4 (56.8, 68.4) | 65.7 (60.6, 70.9) | 65.9 (60.5, 70.9) | 64.6 (58.9, 70.0) |
| Gleason score 8–10 (%) | 81.2 | 66.7 | 80.3 | 80 | 79.2 |
| High Disease Volume (%) | 73.2 | 64.1 | 59.3 | 59.8 | 63.5 |
| WHO Performance Status of 0 (%) | 67.0 | 73.3 | 72.7 | 71.9 | 71.2 |
| Median Hemoglobin | 13.2 (12.1, 14.4) | 12.4 (11.2, 13.3) | 13.9 (12.9, 14.7) | 13.9 (12.9, 14.7) | 13.5 (12.4, 14.5) |
| Median PSA | 16.9 (2.8, 101.1) | 47.5 (7.0, 170.0) | 110.7 (39.0, 361.1) | 110.6 (37.1, 397.5) | 73.5 (17.5, 365.7) |
| Median Alkaline phosphatase | 147 (82, 347) | 211 (108, 489) | 112 (75, 249) | 111 (74, 253) | 131 (77, 310) |
| Opioid analgesic use (%) | 61.8 | 6.4 | 60.1 | 60.6 | 54.1 |
| ADT only treatment ** | 49.7 | 52.9 | 66.5 | 66.3 | 50.0 |
| Median OS, months (95 CI) | 45.5 (42.4, 51.2) | 46.5 (40.5, 54.6) | 45.5 (43, 50.6) | 46.4 (43.7, 50.5) | 48.1 (45.3, 50.7) |
median (25th and 75th percentile)
Numbers are proportional to numbers in the SOC arm
Fig 1A.
Kaplan-Meier overall survival curves by the three randomized clinical trials (4 comparisons)
Multivariable Model Predicting OS
Hazard ratios (HRs) and 95% CI for the clinical factors in the multivariable PH model predicting OS based on data from the training set are presented in Supplementary Table 2. Higher hemoglobin was related with longer OS, while high disease volume, rising alkaline phosphatase and ECOG performance status greater than 0 were associated with increased hazard of death. Specifically, the HR for patients with high vs. low disease was 1.78 (95% CI: 1.43–2.23). The HR for patients with ECOG/WHO performance status of 1 or 2 vs. 0 was 1.32 (95% CI: 1.09–1.61). The HRs for hemoglobin and alkaline phosphatase were 0.98 (95% CI: 0.96–0.99) and 1.06 (95% CI: 1.03–1.10), respectively. The model is displayed in Fig 1B and is accessible online: De novo Metastatic Hormone Sensitive Prostate Cancer
Fig 1B.
Nomogram for the Model of Overall Survival
The tAUCs were 0.72 (95% CI: 0.69–0.76), and 0.70 (95%CI: 0.67–0.72) for STAMPEDE A vs. C, and STAMPEDE A vs. E trial comparisons, respectively. We compared the discrimination of our multivariable prognostic model, measured using the (tAUC), with that obtained using high-volume disease (CHAARTED criteria) and high-risk disease (LATITUDE criteria) in the validation datasets. Across the validation sets, the multivariable model demonstrated slightly higher discrimination compared with these simpler clinical criteria, highlighting the potential value of integrating multiple prognostic factors for individualized risk prediction (Fig 1C). The tAUC in patients who received ADT alone was 0.72 (95% CI: 0.69–0.76). In addition, the model was also evaluated for its calibration by comparing the observed and predicted OS probabilities at 42 and 54 months for training (Fig 2A–2B) and test set (Supplementary Fig 2A–2D).
Fig 1C.
Forest Plot of Integrated Time Dependent AUCs by Different Subgroups
Fig 2A-2B.
Calibration Plots for the Training Set at 42- and 54-months
Risk Groups Based on Quantiles
Patients were classified into two (low vs poor) or three (low, intermediate, poor) risk groups based on the median and tertiles of predicted risk scores. In the two-group classification, 43% of patients were poor-risk (median OS 33.8 months) and 57% were low-risk (median OS 71.0 months), with a HR of 0.39 (95% CI: 0.34–0.45) for low versus poor risk (Fig. 3A). In the three-group classification, 26.0%, 31.1%, and 42.9% of patients were classified as poor-, intermediate-, and low-risk, with median OS of 28.1, 46.1, and 76.6 months, respectively (Fig. 3B). Compared with the poor-risk group, HRs were 0.55 (95% CI: 0.47–0.64) and 0.30 (95% CI: 0.26–0.36) for the low- and intermediate risk groups, respectively. Kaplan–Meier curves and corresponding HRs stratified by STAMPEDE comparisons are shown in Supplementary Fig. 3A–3D and Supplementary Table 3.
Fig 3A-3B.
Kaplan-Meier Overall Survival Curves by the Two and Three Risk Groups in the Test Set
Mapping Risk Scores to Number of Adverse Factors
The distribution of the number of adverse factors is presented in Supplementary Tables 3A–3B and Supplementary Fig. 3E–3F. Only a small proportion of patients had no adverse factors (2.95% in the training set and 2.45% in the test set). Median OS decreased with increasing numbers of adverse factors: in the training set, median OS was NR (95% CI 67.3–NR) for patients with 0 adverse factors and 39.3 months (95% CI 33.4–45.5) for those with 4 factors; corresponding estimates in the test set were NR (95% CI 80.7–NA) and 49.8 months (95% CI 42.7–55.0), respectively (Supplementary Tables 3A–3B; Supplementary Fig. 3G–3I). Treatment benefit was greatest in the poor-risk group (≥4 adverse factors), with significant reductions in the hazard of death compared with ADT alone, whereas effects in the low- and intermediate-risk groups were smaller (Supplementary Table 3C).
Interactions between Risk Group and Treatment on OS
We also evaluated treatment interactions within the two- and three-risk group classifications (Supplementary Tables 4A–4B). The pattern of differential treatment effect by risk group was more pronounced in the STAMPEDE comparisons. Across studies, patients classified as poor-risk consistently demonstrated the greatest benefit from the addition of docetaxel to ADT with the exception of GETUG-15. Specifically, when classifying into three risk groups, HRs for poor-risk patients treated on the experimental arm versus SOC were 0.61 (95% CI:0.43–0.86) in CHAARTED, 0.72 (95% CI: 0.54–0.95) in STAMPEDE arms A vs. C, and 0.51 (95% CI:0.39–0.67) in STAMPEDE arms A vs. E (Fig 4A). In contrast, low-risk patients had unimportant HRs in all individual trials and both pooled training and test cohorts, indicating minimal or no added benefit from treatment intensification. Similar results are obtained by analyzing interaction between treatment and two-risk groups (Supplementary Fig 4A). These findings suggest that the magnitude of treatment effect varies by baseline risk group, with poor-risk patients deriving the greatest survival benefit. The corresponding Kaplan-Meier curves are presented in Fig 4B–4D and Supplementary Fig 4B–4K.
Fig 4A.
Forest Plot of Hazards Ratio Estimated Based on Contrasts from Proportional Hazards Models with Three Risk-Groups-Treatment Interaction Terms in the Training and Test Sets
Fig 4B-4D.
Kaplan-Meier Overall Survival Curves by Three Risk Groups and Treatment Arm in the STAMPEDE Comparisons
Discussion
In this article, we have developed and validated a model predicting OS in men with de novo mHSPC treated with ADT or ADT plus docetaxel with the goal of informing clinical decision-making and guiding future trial design. This study successfully incorporated a broad range of clinical factors from two randomized phase III trials to build a robust model of OS, which was subsequently validated using two comparisons from the STAMPEDE trial. Our findings underscore the importance of several baseline factors in predicting OS in mHSPC patients. High disease volume, elevated alkaline phosphatase, and poorer ECOG performance status were strongly associated with a higher hazard of death. In contrast, higher hemoglobin levels were linked to improved survival. The hazard ratio for high disease volume was particularly elevated, reflecting the substantial impact of metastatic burden on patient outcomes.
Although the CHAARTED and LATITUDE risk classifications are commonly used to stratify patients with mHSPC, they were developed using arbitrary variables rather than using modeling approaches. In contrast, our model was developed using rigorous statistical methods and demonstrated good discriminative ability, with tAUC values of 0.72, and 0.70 for the STAMPEDE arms A vs.C, and STAMPEDE arms A vs. E comparisons. Furthermore, our model provided better discrimination than models based on the CHAARTED and LATITUDE criteria, which achieved lower tAUC values of 0.62 and 0.61, respectively. These values indicate the capacity of the model to differentiate between patients with varying survival outcomes within the de novo subgroup. The calibration plots further supported the model’s predictive accuracy, showing close alignment between the predicted and observed survival probabilities at 42 and 54 months.
We constructed and validated two- and three-risk prognostic groups based on either the quantiles of the risk score or the number of adverse prognostic factors. The predictive accuracy was slightly lower for risk groups based on adverse factors than for those based on the quantile scores, although both tAUCs were lower than that for the continuous risk score. Stratifying patients into these risk groups demonstrated the model’s ability to identify distinct survival probabilities: patients in the poor-risk category had substantially shorter median OS than those in the low-risk group, with hazard ratios indicating a three-fold higher risk of death. Future clinical trials in men with mHSPC could use either model-based quantile risk or the number of adverse factors to guide patient selection, with randomization stratified according to two- or three-level prognostic risk categories.
Prior analyses have shown that patients with high-volume disease benefit most from docetaxel plus ADT [27, 33]. Our analysis suggest an interaction between treatment and risk group: patients in the poor-risk group showed substantially greater benefit from docetaxel plus ADT, while those in the low-risk group experienced minimal to no added benefit. These findings highlight the value of risk stratification in optimizing treatment decisions and support the use of enrichment in clinical trial design.
This validated model provides a valuable complementary tool to the established prognostic models for clinicians in the design of future clinical trials in mHSPC and to personalize treatment plans for mHSPC patients. By accurately predicting OS, it can help guide decisions regarding the intensity and type of treatment, such as who might benefit from the addition of docetaxel to a combination of ADT alongside an androgen receptor pathway inhibitor (ARPI). We propose patients with de novo mHSPC in the poor risk group defined as having 4 or higher adverse factors will be those who benefit from adding docetaxel to ADT plus ARPI and this is planned to be tested as part of the STOPCaP collaboration once the data is available for analysis. Additionally, the model can aid in patient counseling by providing more precise survival estimates, thereby enhancing shared decision-making processes.
Despite its strengths, this model has several limitations. The predominance of patients with ECOG performance status 0 reflects a fitter trial population, potentially limiting generalizability. Missing data for some variables required multiple imputation, which introduces uncertainty and relies on assumptions about the missingness mechanism, although no evidence of nonrandom missingness was observed. Additionally, several potentially important predictors could not be included due to high missingness or lack of collection across trials, which may have improved model performance. Finally, the model was developed and validated in de novo mHSPC clinical trial populations treated with ADT with or without docetaxel, which may not fully represent the broader mHSPC patient population encountered in routine clinical practice.
Conclusions
Future research should focus on external validation of the model in diverse, real-world patient cohorts being treated with ADT plus ARPI as the back-bone to confirm its generalizability in current treatment paradigms and assess the ability to identify patients who benefit from additional agents such as docetaxel. Additionally, incorporating emerging biomarkers, such as circulating tumor DNA and genetic subtypes, may further refine the model and enhance its predictive accuracy. Prospective studies evaluating the impact of model-guided treatment decisions on patient outcomes are also needed to establish the clinical utility of this model.
Supplementary Material
Fig 3C-3D.
Kaplan-Meier Overall Survival Curves by Two Risk Groups and Treatment Arm in the STAMPEDE Comparisons
Acknowledgments
This manuscript was prepared using data from CHAARTED [ID - 2650] from the NCTN/NCORP Data Archive of the National Cancer Institute’s (NCI’s) National Clinical Trials Network (NCTN). Data were originally collected from clinical trial NCT00268476 (E3805) [Androgen Ablation Therapy With or Without Chemotherapy in Treating Patients With Metastatic Prostate Cancer (CHAARTED)]. Additionally, the authors gratefully acknowledge the STAMPEDE and GETUG-15 trial teams for providing access to the individual patient data used in this analysis, and extend their deep appreciation to all the patients who participated in these trials.
ClinicalTrials.gov Identifier: NCT00104715, NCT00268476, NCT00309985
Funding
This research was supported in part by the United States Army Medical Research Awards HT9425-23-1-0393 and HT9425-25-1-0623, the National Institutes of Health Grants R01 CA256157 and R01 CA249279, the Prostate Cancer Foundation, and Prostate Cancer UK (Research Innovation Award (RIA16-ST2-020). LR and PJG are supported by the UK Medical Research Council (MC_UU_00004/06), Prostate Cancer UK (RIA16-ST2-020; MA-TIA23-007) and PJG by a Prostate Cancer Foundation-John Black Charitable Foundation Young Investigator Award. All analyses and conclusions in this manuscript are the sole responsibility of the authors and do not necessarily reflect the opinions or views of the clinical trial investigators, the NCTN, the NCORP or the NCI.
Conflict of Interest
SH has served as a member of the Data Safety Monitoring for BMS, BeOne, CG Oncology, Janessen, and Sanofi, and has served on advisory boards for AZ and Pfizer. She received research funding (institution) from the American Society of Clinical Oncology and Prostate Cancer Foundation.
Footnotes
Declaration of generative AI and AI-assisted technologies in the writing process
During the preparation of this work the author(s) used Chat GPT in order to reduce the number of words to 2500. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication.
Presented in part at the 2024 ESMO Annual Meeting September 2024 (Abstract 1615P).
Data Availability
The clinical data from the CHAARTED trial are available through the NCTN NCORP Data Archive: https://nctn-data-archive.nci.nih.gov/. The data from GETUG-15 and STAMPEDE are available from the trial investigators and sponsors through the MRC office, subject to appropriate data sharing agreements.
References
- [1].Halabi S, Small EJ, Kantoff PW, Kattan MW, Kaplan EB, Dawson NA, et al. Prognostic model for predicting survival in men with hormone-refractory metastatic prostate cancer. J Clin Oncol. 2003;21:1232–7. [DOI] [PubMed] [Google Scholar]
- [2].Halabi S, Lin CY, Small EJ, Armstrong AJ, Kaplan EB, Petrylak D, et al. Prognostic model predicting metastatic castration-resistant prostate cancer survival in men treated with second-line chemotherapy. J Natl Cancer Inst. 2013;105:1729–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Halabi S, Lin CY, Kelly WK, Fizazi KS, Moul JW, Kaplan EB, et al. Updated prognostic model for predicting overall survival in first-line chemotherapy for patients with metastatic castration-resistant prostate cancer. J Clin Oncol. 2014;32:671–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Halabi S, Yang Q, Roy A, Luo B, Araujo JC, Logothetis C, et al. External Validation of a Prognostic Model of Overall Survival in Men With Chemotherapy-Naïve Metastatic Castration-Resistant Prostate Cancer. J Clin Oncol. 2023;41:2736–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Armstrong AJ, Lin P, Higano CS, Sternberg CN, Sonpavde G, Tombal B, et al. Development and validation of a prognostic model for overall survival in chemotherapy-naïve men with metastatic castration-resistant prostate cancer. Annals of Oncology. 2018;29:2200–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Armstrong AJ, Lin P, Tombal B, Saad F, Higano CS, Joshua AM, et al. Five-year survival prediction and safety outcomes with enzalutamide in men with chemotherapy-naïve metastatic castration-resistant prostate cancer from the PREVAIL trial. European urology. 2020;78:347–57. [DOI] [PubMed] [Google Scholar]
- [7].Templeton AJ, Pezaro C, Omlin A, McNamara MG, Leibowitz-Amit R, Vera-Badillo FE, et al. Simple prognostic score for metastatic castration-resistant prostate cancer with incorporation of neutrophil-to-lymphocyte ratio. Cancer. 2014;120:3346–52. [DOI] [PubMed] [Google Scholar]
- [8].Dehm SM, Tindall DJ. Alternatively spliced androgen receptor variants. Endocrine-related cancer. 2011;18:R183–R96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Pritchard CC, Mateo J, Walsh MF, De Sarkar N, Abida W, Beltran H, et al. Inherited DNA-repair gene mutations in men with metastatic prostate cancer. New England Journal of Medicine. 2016;375:443–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Robinson D, Van Allen EM, Wu Y-M, Schultz N, Lonigro RJ, Mosquera J-M, et al. Integrative clinical genomics of advanced prostate cancer. Cell. 2015;161:1215–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Wyatt AW, Azad AA, Volik SV, Annala M, Beja K, McConeghy B, et al. Genomic alterations in cell-free DNA and enzalutamide resistance in castration-resistant prostate cancer. JAMA oncology. 2016;2:1598–606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Annala M, Vandekerkhove G, Khalaf D, Taavitsainen S, Beja K, Warner EW, et al. Circulating tumor DNA genomics correlate with resistance to abiraterone and enzalutamide in prostate cancer. Cancer discovery. 2018;8:444–57. [DOI] [PubMed] [Google Scholar]
- [13].De Laere B, van Dam P-J, Whitington T, Mayrhofer M, Diaz EH, Van den Eynden G, et al. Comprehensive profiling of the androgen receptor in liquid biopsies from castration-resistant prostate cancer reveals novel intra-AR structural variation and splice variant expression patterns. European urology. 2017;72:192–200. [DOI] [PubMed] [Google Scholar]
- [14].Nixon AB, Liu Y, Yang Q, Luo B, Starr MD, Brady JC, et al. Prognostic and predictive analyses of circulating plasma biomarkers in men with metastatic castration resistant prostate cancer treated with docetaxel/prednisone with or without bevacizumab. Prostate Cancer and Prostatic Diseases. 2024:1–8. [DOI] [PubMed] [Google Scholar]
- [15].Ryan CJ, Halabi S, Ou S-S, Vogelzang NJ, Kantoff P, Small EJ, et al. Adrenal androgen levels as predictors of outcome in prostate cancer patients treated with ketoconazole plus antiandrogen withdrawal: results from a cancer and leukemia group B study. Clinical Cancer Research. 2007;13:2030–7. [DOI] [PubMed] [Google Scholar]
- [16].Ryan CJ, Molina A, Li J, Kheoh T, Small EJ, Haqq CM, et al. Serum androgens as prognostic biomarkers in castration-resistant prostate cancer: results from an analysis of a randomized phase III trial. Journal of clinical oncology. 2013;31:2791–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Halabi S, Ou SS, Vogelzang NJ, Small EJ. Inverse correlation between body mass index and clinical outcomes in men with advanced castration-recurrent prostate cancer. Cancer. 2007;110:1478–84. [DOI] [PubMed] [Google Scholar]
- [18].Halabi S, Kelly WK, Ma H, Zhou H, Solomon NC, Fizazi K, et al. Meta-Analysis Evaluating the Impact of Site of Metastasis on Overall Survival in Men With Castration-Resistant Prostate Cancer. J Clin Oncol. 2016;34:1652–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Halabi S, Vogelzang NJ, Kornblith AB, Ou SS, Kantoff PW, Dawson NA, et al. Pain predicts overall survival in men with metastatic castration-refractory prostate cancer. J Clin Oncol. 2008;26:2544–9. [DOI] [PubMed] [Google Scholar]
- [20].Halabi S, Guo S, Luo B, Yu C, Knutson TP, Kobilka A, et al. Pathogenic Genomic Alterations in Circulating Tumor DNA Predict Overall Survival in Men with Metastatic Castrate-resistant Prostate Cancer. Eur Urol. 2025. [Google Scholar]
- [21].Gravis G, Boher J-M, Fizazi K, Joly F, Priou F, Marino P, et al. Prognostic factors for survival in noncastrate metastatic prostate cancer: validation of the glass model and development of a novel simplified prognostic model. European urology. 2015;68:196–204. [DOI] [PubMed] [Google Scholar]
- [22].Tangen CM, Hussain MH, Higano CS, Eisenberger MA, Small EJ, Wilding G, et al. Improved overall survival trends of men with newly diagnosed M1 prostate cancer: a SWOG phase III trial experience (S8494, S8894 and S9346). The Journal of urology. 2012;188:1164–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Iversen P Antiandrogen monotherapy: indications and results. Urology. 2002;60:64–71. [Google Scholar]
- [24].Roy S, Sun Y, Wallis CJD, Morgan SC, Grimes S, Malone J, et al. Development and validation of a multivariable prognostic model in de novo metastatic castrate sensitive prostate cancer. Prostate Cancer Prostatic Dis. 2023;26:119–25. [DOI] [PubMed] [Google Scholar]
- [25].Fujiwara H, Kubota M, Hidaka Y, Ito K, Kawahara T, Kurahashi R, et al. A novel prognostic model of de novo metastatic hormone-sensitive prostate cancer to optimize treatment intensity. Int J Clin Oncol. 2024;29:1574–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [26].Kyriakopoulos CE, Chen Y-H, Carducci MA, Liu G, Jarrard DF, Hahn NM, et al. Chemohormonal therapy in metastatic hormone-sensitive prostate cancer: long-term survival analysis of the randomized phase III E3805 CHAARTED trial. Journal of Clinical Oncology. 2018;36:1080–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Gravis G, Boher JM, Chen YH, Liu G, Fizazi K, Carducci MA, et al. Burden of Metastatic Castrate Naive Prostate Cancer Patients, to Identify Men More Likely to Benefit from Early Docetaxel: Further Analyses of CHAARTED and GETUG-AFU15 Studies. Eur Urol. 2018;73:847–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Gravis G, Fizazi K, Joly F, Oudard S, Priou F, Esterni B, et al. Androgen-deprivation therapy alone or with docetaxel in non-castrate metastatic prostate cancer (GETUG-AFU 15): a randomised, open-label, phase 3 trial. Lancet Oncol. 2013;14:149–58. [DOI] [PubMed] [Google Scholar]
- [29].Sweeney CJ, Chen YH, Carducci M, Liu G, Jarrard DF, Eisenberger M, et al. Chemohormonal Therapy in Metastatic Hormone-Sensitive Prostate Cancer. N Engl J Med. 2015;373:737–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].James ND, Sydes MR, Clarke NW, Mason MD, Dearnaley DP, Spears MR, et al. Addition of docetaxel, zoledronic acid, or both to first-line long-term hormone therapy in prostate cancer (STAMPEDE): survival results from an adaptive, multiarm, multistage, platform randomised controlled trial. The Lancet. 2016;387:1163–77. [Google Scholar]
- [31].Tibshirani R Regression Shrinkage and Selection Via the Lasso. Journal of the Royal Statistical Society: Series B (Methodological). 1996;58:267–88. [Google Scholar]
- [32].Uno H, Cai T, Tian L, Wei L-J. Evaluating prediction rules for t-year survivors with censored regression models. Journal of the American Statistical Association. 2007;102:527–37. [Google Scholar]
- [33].Vale CL, Fisher DJ, Godolphin PJ, Rydzewska LH, Boher JM, Burdett S, et al. Which patients with metastatic hormone-sensitive prostate cancer benefit from docetaxel: a systematic review and meta-analysis of individual participant data from randomised trials. Lancet Oncol. 2023;24:783–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The clinical data from the CHAARTED trial are available through the NCTN NCORP Data Archive: https://nctn-data-archive.nci.nih.gov/. The data from GETUG-15 and STAMPEDE are available from the trial investigators and sponsors through the MRC office, subject to appropriate data sharing agreements.