Clinicopathologic factors that influence prognosis and survival outcomes in men with metastatic castration‐resistant prostate cancer treated with Radium‐223

Abstract Background In men with metastatic castration‐resistant prostate cancer (mCRPC) with primarily bone metastases, radium‐223 (223Ra) improves overall survival (OS). However, the selection of 223Ra is not guided by specific validated clinicopathologic factors, and thus outcomes are heterogeneous. Patients and methods This retrospective survival analysis was performed in men with mCRPC treated with 223Ra at our cancer center. Demographics and disease characteristics were collected. OS was calculated using the Kaplan–Meier method (log‐rank). The potential prognostic factors were determined using both univariable (UVA) and multivariable analysis (MVA) (Cox‐regression) methods. Results In total, 150 patients with a median age of 74 years (52–93) received 223Ra between May 2015 and July 2018, and 58% had 6–20 bone metastases. Ninety‐four (63%) patients received >4 223Ra doses, and 56 (37%) received ≤4. The following pre‐treatment factors were analyzed (median [range]): eastern cooperative oncology group performance status (ECOG PS), (1 [0–3]); Albumin (ALB), (39 g/L [24–47]); alkaline phosphatase (ALP), (110 U/L [35–1633]); and prostate‐specific antigen (PSA), (49 µg/L [0.83–7238]). The median OS for all patients was 14.5 months (95% CI: 11.2–18). These factors were associated with poor survival outcomes in UVA and MVA: ALB <35 g/L, ALP >150 U/L, ECOG PS 2–3, and PSA >80 µg/L. By assigning one point for each of these factors, a prognostic model was developed, wherein three distinct risk groups were identified: good, 0–1 (n = 103); intermediate, 2 (n = 30); and poor risk, 3–4 points (n = 17). The median OS was 19.4, 10.0, and 3.1 months, respectively (p < 0.001). Conclusions Pre‐treatment ALB, ALP, ECOG, and PSA, were significantly correlated with OS and could guide treatment selection for men with mCRPC by identifying those who are most or least likely to benefit from 223Ra. Validation in an independent dataset is required prior to widespread clinical utilization.


| INTRODUCTION
Prostate cancer is the second leading cause of cancer in men and the fifth most common cause of all cancer deaths globally. 1 Bone metastases are significant in prostate cancer with approximately 30% of men developing bone metastases in the first 2 years of castration resistance and ultimately 90% of men developing bone metastases over the course of their entire disease. 2 Skeletal-related events (SRE) such as pain, fracture, spinal cord compression, hypercalcemia, and bone marrow suppression will occur in over 50% of men with advanced prostate cancer and bone metastases. 3 Taxane chemotherapy (docetaxel and cabazitaxel), androgen receptor axis-targeted (ARAT) therapies (abiraterone and enzalutamide), sipuleucel-T, and radium-223 ( 223 Ra) have demonstrated improved overall survival (OS) metastatic castration-resistant prostate cancer (mCRPC). 4-9 223 Ra is a calcium mimetic, alpha-emitting nuclide, which is taken up in bone metastases due to high bone turnover. The alpha radiation induces double-strand DNA breaks, causing localized cytotoxicity in bone metastases. 10 The ALSYMPCA trial was an international, double-blind, placebo-controlled phase III trial that enrolled patients with mCRPC and symptomatic bone metastases and randomized them to either 223 Ra or placebo. This trial reported a statistically significant improvement in median OS in patients treated with 223 Ra, demonstrating an improvement in OS by 3.6 months (HR 0.70; p < 0.0001) and delaying the time to first symptomatic SRE by 5.8 months (HR 0.66, p < 0.0001). 9 Of the men enrolled, 71% had improvement in their pain scores and activity at week 8. 11 ALSYMPCA data on quality of life (QOL) showed that 223 Ra increased EQ-5D utility scores and resulted in a slower decrease in QOL scores as compared to placebo. 12 Based on the results of this pivotal trial, 223 Ra was approved for patients with mCRPC with symptomatic bone metastases. 9,13 Prostate-specific antigen (PSA) and conventional imaging are routinely used to evaluate treatment response to other treatments, including chemotherapy and ARATs. However, during 233 Ra treatment, there are no reliable clinical methods to assess response. Hence accurate patient selection at the start of treatment will permit those who are unlikely to benefit from 233 Ra to avoid this agent while treating those who have the highest chance of benefit. Identification of prognostic factors would guide treatment decision making, improve patient outcomes, and could better inform future clinical trial design. Our aim was to analyze the survival outcomes of men with mCRPC to the bone who were treated with 223 Ra and to determine the prognostic factors that may affect their OS.

| Patients and data collection
Men with mCRPC treated with 223 Ra at our cancer center from May 2015 to July 2018 were included in this study to permit sufficient follow-up. The study received ethics approval from our institutional review board (REB #18-5545). Demographics and disease characteristics were obtained from the electronic health records. Baseline variables such as eastern cooperative oncology group performance status (ECOG PS), number of bone metastasis on conventional imaging, lymph node (LN) status, serial PSA, lactate dehydrogenase (LDH), hemoglobin (Hb), albumin (ALB), and alkaline phosphatase (ALP) were collected. Each patient received 223 Ra 50-55 kBq/kg intravenously q 4 weeks for at least one dose to a maximum of six doses. Information on the number of pre-and post-223 Ra systemic treatments were collected. The use of bone protecting agents (BPA) such as zoledronate and denosumab was noted. PSA doubling time (PSADT) before and during 223 Ra treatment were calculated using Memorial Sloan Kettering Hospital calculator. 14 On treatment serial PSA and ALP levels were obtained. Hematological toxicities such as anemia (Hb <100 g/L), neutropenia (ANC <1 × 10 9 /L), and thrombocytopenia (platelet <100 × 10 9 /L) as well as fatigue were collected. The number of bone metastases post-223 Ra therapy was also recorded.

| Statistical analysis
Summary statistics using the median and ranges were calculated for demographic and prognostic factors. For survival calculations, time from starting 223 Ra to the event of interest was used for progression-free survival (PFS) (disease progression) and OS (death from any cause). Survival analysis was performed using the Kaplan-Meier method (log-rank). Pre-treatment clinicopathological variables were analyzed for OS associations using a cox proportional hazards model those who are most or least likely to benefit from 223 Ra. Validation in an independent dataset is required prior to widespread clinical utilization.
for univariable (UVA) and selected multivariable analysis (MVA). Hazard ratios (HR) with 95% confidence intervals (95% CI) are reported. A four-variable prognostic score model was proposed based on the MVA results. A timedependent receiver operating characteristic curve (ROC) with an area under curve (AUC) was used to assess the prognostic score's significance. All tests used a p value of ≤0.05 for significance. IBM SPSS Statistics v26 was used to conduct statistical analyses (IBM; Armonk, NY, USA).

| Patients' characteristics
In total, 150 mCRPC patients with a median age of 74 years (52-93) received 223 Ra between May 2015 and July 2018. Patient and disease characteristics are detailed in (

| Clinical outcomes
A total of 130 men had died, and 110 men had progressed at the time of analysis with a median follow-up of 39 months (range, 35-45 months). The median OS for all men was 14.5 months (95% CI: 11.2-18.0) ( Figure 1A). The median PFS was 7.3 months (95% CI: 6.6-7.9) ( Figure 1B).

| Prognostic survival model
From a prognostic model based on the four independent, pre-treatment clinicopathological factors, albumin <35 g/L, ECOG PS 2-3, ALP >150 U/L, and PSA>80 µg/L; three distinct prognostic groups were identified by assigning one point for each aforementioned variable (  Figure 2A). Interestingly, patients in the good (n = 22) or intermediate risk (n = 5) groups who achieved any PSA reduction during 223 Ra therapy had improved median OS to 23.1 months (95% CI 17.1-29), and 20.0 months (95% CI 2.5-38), respectively; p < 0.001 ( Figure  2b). The receiver operating curve for the prognostic model score and survival status produced an AUC 0.762, p < 0.001. Two patient vignettes were described in Appendix S1.

| DISCUSSION
We identified three different prognostic groups based on a combination of four pre-treatment clinical and laboratory factors that affected OS in our cohort of 150 patients treated with 223 Ra during the last 5 years. These factors were ALB less than 35 g/L, ALP above 150 U/L, PSA above 80 µg/L, and ECOG PS score of 2-3. Our analysis identified good, intermediate, and poor risk groups, and these factors may inform the selection of 223 Ra for men with mCRPC who are most likely to receive a substantial benefit. This real-world cohort analyzed in this study had similar outcomes to those reported in the ALSYMPCA trial, where patients treated with 223 Ra had a median OS of 14.0 months. In addition, ALSYMPCA reported a median OS of 6.2 months in those who received ≤4 doses and 17.9 months with >4   15 reported a baseline prognostic survival model for patients with progressive mCRPC treated with 19 consecutive therapeutic protocols, using some prognostic variables such as age, Karnofsky performance status (KPS), and baseline laboratory values such as PSA, Hb, LDH, ALB, and ALP. On MVA, only KPS, Hb, ALP, ALB, and LDH were significantly associated with OS (p < 0.05), however, age and PSA were not.
Low pretreatment ALB is a well-recognized factor associated with poor survival in cancer patients, including prostate cancer, as reported in a systematic review by Gupta et al. 16 Interestingly, even at in early stage prostate cancer, low pretreatment ALB can predict a higher pathological T-stage (T3 or T4) after radical prostatectomy as reported by E Richter et al. 17 A MVA by Li et al 18 identified a prognostic nutritional index that included ALB. Elevated pretreatment ALB was a favorable prognostic factor for PFS, cancer-specific survival, and OS of the metastatic prostate cancer patients.
Bone-specific prognostic factors, including a history of previous skeletal-related event 19 , presence of pathologic fracture 20 , urinary N-telopeptide, and bone-specific ALP levels 21 , have also demonstrated a prognostic value.
A number of studies have reported factors associated with prognosis to 223 Ra treatment. Sartor et al 22 reported an exploratory study from the phase III ALSYMPCA trial in an attempt to correlate baseline patient variables and survival on 223 Ra. On MVA of these prognostic variables, men who presented with poor baseline ECOG PS, LDH >245 U/L, or ALP >131 U/L were associated with worse survival outcomes. Whereas men on placebo receipt with elevated baseline LDH >266 U/L and ALP >153 U/L were found to be associated with a statistically higher death rate. Notably, PSA was found to have decreased in 27% of 223 Ra patients, compared with 18% in our cohort. Wongetal et al 23 evaluated 64 patients with mCRPC who received 223 Ra based on retrospective analysis and found on MVA, 3 factors prior to 223 Ra were associated with better OS, which were baseline ALP < 115 U/L, no prior chemotherapy, and ≤ five bone metastases.  Our prognostic model showed an acceptable level of performance to define the prognostic groups, which is similar to that proposed by Halabi et al. 27 They proposed an updated mCRPC model using CALGB-90401 training set, which included opioid analgesic use, LDH > 1x the upper limit of normal, disease site, ECOG PS, albumin, Hb, PSA, and ALP.
Previous attempts to create a prognostic model for mCRPC patients who received 223 Ra produced a similar AUC to our study. Frantellizzi et al 28 studied retrospectively 92 patients with mCRPC and symptomatic bone metastases who were treated with 223 Ra. On UVA, prognostic baseline variables included body mass index, ECOG PS, Hb, and ALP values were associated with OS. However, after MVA, only ECOG PS and Hb were strongly associated with OS. In a far more complicated model, those authors tested a high number of variable combinations and discovered an optimal prognostic score. The model was achieved by combining three prognostic scores (3-PS): ECOG PS, PSA ≥ 20 µg/L, and Hb < 120 g/L and produced an AUC 0.784; p < 0.001. They reported zero, one, and two points for baseline ECOG PS (0,1, ≥ 2), respectively, one point for baseline PSA ≥20 µg/L, and one point for baseline Hb < 12 g/dL. In a subgroup analysis of survival based on 3-PS, cohort stratification showed median OS >31, 11, 9, and 4 months for the scores 0-1, 2, 3, and 4, respectively. Unlike our model, which clearly describes very distinct groups, this model has two reasonably similar groups with an OS only 2 months apart for those who scored 2 or 3 on the 3-PS model. The small number of patients included in the analysis limits the robustness of the findings. Our findings must be confirmed in a larger number of patients in an independent dataset. Furthermore, the variables identified by our model are not specific to 223 Ra, and so the biologic rationale for the impact of these variables on survival with 223 Ra is unclear.
In conclusion, pre-treatment ALB, ALP, ECOG PS, and PSA, have significant correlations with OS and provide prognostic information that could be used to select patients who are most or least likely to benefit from 223 Ra. Our model is simple, sensitive, and defines very distinct prognostic groups. The use of an independent dataset for validation should be promoted. Patients who received >4 doses of 223 Ra, or systemic treatment after 223 Ra or who have any PSA response during 223 Ra therapy had better survival outcomes. The presence of any hematological toxicities after completion of 223 Ra therapy was associated with poor survival outcomes.