A novel prognostic model for Japanese patients with newly diagnosed bone‐metastatic hormone‐naïve prostate cancer

Abstract Objectives To evaluate the prognosis of newly diagnosed patients with metastatic hormone‐naïve prostate cancer (mHNPC) and develop a novel prognostic model based on ChemoHormonal Therapy Versus Androgen Ablation Randomized Trial for Extensive Disease in Prostate Cancer (CHAARTED) risk classifications. Patients and methods We retrospectively analyzed the data of 578 newly diagnosed mHNPC patients initially treated with androgen deprivation therapy. We evaluated three clinical factors, namely, CHAARTED risk classifications (high‐volume disease [HVD] vs low‐volume disease [LVD]), Gleason scores (GS, 9‐10 vs ≤8), and hemoglobin (Hb, ≤13.0 g/dL vs >13.0 g/dL), for their prognostic potential in predicting time to castration‐resistant prostate cancer (TTC) and overall survival (OS) of mHNPC patients by multivariate analysis. Moreover, we developed a novel prognostic model that consisted of significant prognostic factors. Results Of the entire cohort, the median TTC and OS values were 18.3 and 67.5 months, respectively. HVD, GS 9‐10, and Hb ≤13.0 g/dL were independent poor prognostic factors for both TTC and OS. We developed a novel prognostic model which could stratify mHNPC patients into four risk groups according to the numbers of poor prognostic factors: group 1, LVD with low‐risk (LVD patients without GS 9‐10 and Hb ≤13.0 g/dL); group 2, LVD with high‐risk (LVD patients with GS 9‐10, Hb ≤13.0 g/dL, or both); group 3, HVD with low‐risk (HVD patients without GS 9‐10 with or without Hb ≤13.0 g/dL); and group 4, HVD with high‐risk (HVD patients with GS 9‐10 with or without Hb ≤13.0 g/dL). The median TTC and OS of groups 1, 2, 3, and 4 were 124.8, 36.4, 17.9, and 11.2 months, and 117.2, 94.2, 67.9, and 46.2 months, respectively. A significant difference in TCC and OS was found between all groups. Conclusion We developed a prognostic model for mHNPC patients that consisted of CHAARTED risk classifications, GS, and Hb. Our prognostic model could significantly stratify the prognosis of patients with LVD and HVD into two groups each. This model might be a good reference for shared decision making between patients and physicians on the initial treatment for mHNPC.


| INTRODUC TI ON
Prostate cancer is currently one of the most common cancers, as more than 1.3 million cases were newly diagnosed worldwide in 2018, 1 and the incidence of prostate cancer has been particularly increasing in northeast Asian countries. 2 One reason for the increase in the number of prostate cancer patients is the spread of prostate-specific antigen (PSA) screening, which has led to earlier detection and a decrease in the mortality rate of prostate cancer. 3,4 However, there are still quite a few patients with distant metastases at first diagnosis, and appropriate treatment is required especially for these newly diagnosed (de novo) metastatic hormone-naïve prostate cancer (mHNPC) patients. For mHNPC patients, androgen deprivation therapy (ADT), along with the upfront use of docetaxel or androgen receptor targeting agent (ARTA) including abiraterone acetate, apalutamide, and enzalutamide, is now the standard of care (SOC). [5][6][7][8] In addition, Phase III Systemic Therapy for Advanced or Metastatic Prostate cancer: Evaluation of Drug Efficacy (STAMPEDE) trial arm H showed that local radiotherapy to the prostate improved the survival of mHNPC patients with a low-tumor burden. 9 Thus, local radiotherapy is considered a SOC for mHNPC patients with low tumor burden.
Currently, many treatment strategies for mHNPC patients are available, but the appropriate treatment remains unclear because those patients are known to have a wide spectrum of progression risks. [10][11][12] Therefore, predicting the survival of mHNPC patients is important when making treatment strategies for them. In this study, we aimed to investigate the prognosis of Japanese patients newly diagnosed with mHNPC treated with primary ADT and to develop a novel prognostic model for these patients.

| Study population
We retrospectively identified 593 newly diagnosed Japanese de novo mHNPC patients with bone metastases. These patients were initially treated with ADT and were registered in the Metropolitan Prostate Cancer Group (MPCG) database between January 2004 and December 2015. Fifteen patients were excluded due to a lack of data and 578 patients were ultimately evaluated.
All patients had a histologically confirmed diagnosis of prostate adenocarcinoma. Metastatic sites were evaluated by computed tomography (CT) and bone scan using 99-technetium methylene diphosphonate/hydroxymethylene diphosphonate before any treatment. ADT was carried out by medical castration with luteinizing hormone-releasing hormone (LHRH) agonist or antagonist combined with bicalutamide (80 mg daily, approved dose in Japan).
Patients did not receive upfront abiraterone, apalutamide, enzalutamide, or docetaxel for mHNPC. Clinical data including patients' age, PSA, Gleason scores (GS), hemoglobin (Hb), CT findings, bone scan findings, time to castration-resistant prostate cancer (TTC), and overall survival (OS) for mHNPC were obtained from electronic medical records. GS was determined by a pathologist at each facility according to Gleason grading. 13 Hb levels were examined prior to the start of ADT. The patients were classified into two survival risk groups defined by the ChemoHormonal Therapy Versus Androgen Ablation Randomized Trial for Extensive Disease in Prostate Cancer (CHAARTED) trial 14 : High-volume disease (HVD) was defined as the presence of visceral metastases or ≥4 bone lesions with ≥1 beyond the vertebral bodies and pelvis and low-volume disease (LVD) was defined as a non-HVD. 6 Metastatic castration-resistant prostate cancer (mCRPC) was defined according to PCWG-2. 15 After progression to the mCRPC state, all patients were administered with an LHRH agonist or antagonist continuously and subsequently treated according to each attending physician's treatment strategy.
Flutamide, oral steroids, estramustine phosphate, and docetaxel were mainly used for those mCRPC patients until 2014. Since 2014, ARTA including abiraterone acetate and enzalutamide, docetaxel, and cabazitaxel was mainly used for those mCRPC patients; in addition, radium-223 was used since 2016 as appropriate according to each attending physician's choice. Bone-modifying agents such as denosumab and zoledronic acid were also used for mCRPC patients according to the attending physicians' choice. In the terminal stage, palliative therapy, pain control with morphine, and palliative external-beam radiation were used as appropriate.

| Prognostic factors and prognostic model
We evaluated three clinical factors, namely, CHAARTED risk classifications 6 (HVD vs LVD), GS (9-10 vs ≤8), and Hb (≤ 13.0 g/dL vs >13.0 g/dL) as prognostic significance for TTC and OS of mHNPC patients by univariate and multivariate analyses. Moreover, we developed a novel prognostic model that consisted of statistically significant prognostic factors.

| Statistical analysis
A Kaplan-Meier (KM) product-limit estimator was used to assess the TTC and OS of mHNPC patients. A generalized Wilcoxon test was used to analyze the differences in TTC and OS between the groups. stratify the prognosis of patients with LVD and HVD into two groups each. This model might be a good reference for shared decision making between patients and physicians on the initial treatment for mHNPC.
For detecting the prognostic factors for TCC and OS, univariate and multivariate analyses were performed using a Cox proportional hazards regression model.
We derived relative risks and 95% confidence intervals (95% CI). All tests were two-sided; an alpha value of .05 was considered significant. The statistical software "EZR" (version 1.40; Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (version 3.5.2; The R Foundation for Statistical Computing, Vienna, Austria), 16 was used for drawing the KM curve.
Other analyses were all conducted using IBM SPSS Statistics software for Windows version 26 (IBM Corp., Armonk, NY, USA).

| Patients' characteristics
Patients' characteristics are shown in

| Clinical factors with prognostic significance for TTC and OS
The KM curves for TTC and OS of the entire cohort are shown in Hb≤13.0 g/dL, were found to be independent significant poor prognostic factors for both TTC and OS.

| D ISCUSS I ON
In this study, we developed a prognostic model for mHNPC patients that consisted of CHAARTED risk classifications, GS, and Hb. LVD and HVD were first suggested in the clinical Phase III CHAARTED trial. 6 In the CHAARTED trial, the median OS of newly diagnosed mHNPC patients treated by ADT (control arm) with LVD and HVD was 59. 8  Cytoreductive prostatectomy for newly diagnosed mHNPC patients is being tested in nine prospective randomized clinical trials. 25 As described above, many treatment strategies are available for newly diagnosed mHNPC patients currently; however, choosing the appropriate treatment for these patients according to various risks is important because they are known to have a wide spectrum of clinical progression risks. 12 In our study, we developed a novel prognostic model that consisted of HVD, GS 9-10, and Hb≤13.0, which demonstrated as statistically significant poor prognostic factors. All variables have been also reported as prognostic factors for mHNPC patients. [10][11][12]26 Although   Finally, next-generation imaging (NGI) for prostate cancer detection such as prostate-specific membrane antigen-positron emission tomography (PET), choline PET, and fluciclovine PET, has been developed recently, and these techniques are more sensitive to metastases detection than conventional imaging such as CT and bone scan. 28 Moreover, as applications of NGI continue to increase, the treatment of strategy for patients with few metastases has changed.
MDT based on NGI is expected to improve survival. Our prognostic model was developed based on clinical information obtained by conventional imaging including CT, bone scan, and magnetic resonance imaging. When NGI would be clinically applied in the near future, a new prognostic model for NGI should be developed.
In conclusion, our novel prognostic model might be the first model that could stratify the finer survival risk of mHNPC patients based on LVD and HVD, with two groups each. This prognostic model might be a good reference for shared decision making between patients and physicians of the initial treatment for mHNPC.
To confirm our results, a prospective study is warranted.