2.1 Data source

This was a retrospective non-interventional chart review of patients from four AC and six CP institutions located throughout the United States with a high volume of prostate cancer patients that integrate radium-223 into the treatment of mCRPC. Specific institutions involved in this study included Tulane Cancer Center, Tisch Cancer Institute, 21st Century Oncology, University of Pittsburgh Medical Center Hillman Cancer Center, Wichita Urology Group, Chesapeake Urology Research Associates, Advanced Radiation Centers, Associated Medical Professionals, Karmanos Cancer Institute, and Prostate Oncology Specialists. Participating providers reported their primary specialty as oncology (50%), urology (30%), or nuclear medicine (20%). “High volume” was defined as treating at least 2000 prostate cancer patients per year. Patient-level deidentified data were extracted using an electronic case report form that was completed by the participating investigator and managed by a designated contract research organization. To reduce selection bias, investigators were provided a list with instructions on how to select patients alphabetically based on randomly assigned letters. Appropriate institutional review board approval was obtained from each participating center.

2.2 Study design and patient population

Eligible patients were adults with mCRPC and a confirmed diagnosis of bone metastases who received one or more doses of radium-223 outside of a clinical trial between January 2014 and June 2017 and who had at least 4 months of follow-up data or a date of death or hospice referral recorded within the 4-month minimum follow-up time. Because diagnosis of mCRPC was by investigator determination within the medical chart, the following consistent definitions were employed: castration resistance was defined as progression in patients with castrate-level testosterone after continuous hormonal therapy, confirmed by two rises in prostate-specific antigen (PSA) levels at least 1 week apart (clinical and radiographic progression were excluded from this definition), and metastatic disease was defined as the presence of metastases confirmed by imaging tests including bone, computed tomography, or magnetic resonance imaging scans. Alternatively, use of agents approved in the metastatic setting in patients who are castrate resistant could be used to confirm metastatic disease.

2.3 Baseline variables

Baseline variables were collected at the date of radium-223 initiation. Data collected included patient demographics, such as age, site of treatment, and type of insurance; disease characteristics, such as time from diagnosis of mCRPC to initiation of radium-223 and number of bone metastases; skeletal symptoms, including incidence of SSEs; medical history (a Klabunde comorbidity score was calculated for each patient); treatment history; laboratory parameters; and radium-223-related variables, including date and number of injections and reasons for initiation and discontinuation of radium-223.

2.4 Outcomes

The primary outcomes included OS and evaluation of differences in radium-223 outcomes between sites of care. Secondary outcomes included a comparison of baseline clinical and demographic characteristics, reasons (based on a predefined list of nonmutually exclusive answers) for initiation and discontinuation of radium-223, and treatment sequencing after radium-223 between sites of care. OS (in months) was defined from the date of initiation of radium-223 to death from any cause. Patients alive at last follow-up/end of study period (November 3, 2017) were censored.

2.5 Statistical analysis

Summary statistics, including means and proportions, were used to describe baseline demographic and clinical characteristics such as treatment sequences, reasons for initiation and discontinuation of radium-223, laboratory values, and physician and patient characteristics. Differences between groups were assessed using Wilcoxon-rank sum test for continuous variables and Fisher's exact test for categorical variables and compared between site of care (AC vs. CP). OS was estimated from the date of radium-223 initiation using the Kaplan–Meier method and compared between site of care (AC vs. CP) using a log-rank test. The log-rank method was used to compare survival distributions between patients who received 1–4 versus 5–6 doses of radium-223. All analyses were conducted using SAS, version 9.4 (Cary).

3.1 Baseline demographic and clinical characteristics

A total of 200 patients were included in this analysis. Of the 200 patients, 43% received treatment at an AC and 57% at a CP. The mean age of patients at radium-223 initiation was 73.6 years overall; however, patients treated in CPs were significantly older than those treated in an AC (74.9 vs. 71.9 years; p = .031). Key median baseline laboratory values in patients treated at AC and CP settings included PSA (58.7 and 33.0 ng/dl, respectively), hemoglobin (12.0 and 12.5 g/dl, respectively), alkaline phosphatase (123 and 89.0 U/L, respectively), lactate dehydrogenase (225.0 and 208.0 U/L, respectively), and platelet count (218.5 10⁹/L and 210.0 10⁹/L, respectively). In addition, patients treated in CPs were treated with radium-223 within a shorter period of time from diagnosis of mCRPC (1.3 vs. 1.9 years; p < .001), had more comorbid conditions (Klabunde score at radium-223 initiation 1.1 vs. 0.7; p = .020), and received a greater mean number of radium-223 doses (5.4 vs. 4.8; p = .001). Tables 1 and 2 show the baseline demographic and clinical characteristics of the patients.

Most patients (61.5%) received at least 1 therapy before radium-223; the most common therapy for those treated at CPs was sipuleucel-T (n = 46, 40.4%), while the most common therapy in ACs was abiraterone acetate (n = 39, 45.3%).

3.2 Treatment patterns

The mean number of doses of radium-223 received was 5.2 (SD: 1.5), and this differed between site of care as noted above. Patients who completed 5–6 doses of radium-223 were more likely to have radium-223 initiated early in their treatment sequence for mCRPC than those who only completed 1–4 doses (p = .005). In the overall population, bisphosphonates or denosumab were commonly started before radium-223 (12.5% and 61%, respectively; 15.1% and 65.1%, respectively, at ACs; 10.5% and 57.9%, respectively, at CPs). Bisphosphonates and denosumab were also frequently continued in combination with radium-223 (10.5% and 59%, respectively; 11.6% and 61.6%, respectively, at ACs; 9.6% and 57.0%, respectively, at CPs). Also, 43% of patients received radium-223 monotherapy, and 57% received it concurrently with other agents (mainly abiraterone acetate and enzalutamide). After discontinuation of radium-223, 42.5% of patients went on to receive an additional line of mCRPC therapy. The most common therapy after radium-223 in all patients was docetaxel or cabazitaxel (n = 38, 19%), followed by abiraterone acetate (n = 23, 12%), enzalutamide (n = 17, 9%), and sipuleucel-T (n = 7, 4%) (Table 2).

In the subset of patients (n = 78) who received chemotherapy (specifically docetaxel or cabazitaxel), 56% (n = 44) received it before or during radium-223 administration and 44% (n = 34) received it after radium-223. Notably, patients who were treated at an AC were more likely to receive either docetaxel or cabazitaxel at some point in their treatment sequence (66% overall; 34.9% before radium-223 initiation, 31.4% after) compared to patients who were treated in a CP (18% overall; 8.8% before radium-223 initiation, 9.6% after).

3.3 Reasons for radium-223 initiation and discontinuation

Participating investigators could report multiple reasons for initiation or discontinuation of radium-223. In addition to the presence of bone metastases, the most common reason selected for radium-223 initiation in the total population was disease progression (43%) (Figure 1A). Other reasons selected for initiation included to prolong survival (32.5%), treatment guidelines (30.5%), patient symptoms (25.5%), impact on quality of life (23.5%), and unknown (7.5%). Patients treated at an AC compared to a CP were more likely to be initiated on radium-223 due to treatment guidelines (39.5% vs. 23.7%; p = .0199) and less likely due to patient symptoms (12.8% vs. 35.1%; p = .0003) or impact on quality of life (8.1% vs. 35.1%; p < .0001). There were no differences in other reasons for initiation.

In total, 72% of patients completed treatment, defined as receiving 5–6 doses, with radium-223 (n = 143); significantly fewer patients treated at an AC completed therapy compared to those in a CP (59.3% vs. 80.7%; p = .0014). Overall, for patients who did not complete radium-223 treatment, the most frequent reason for discontinuation was progression (28%). Other reasons included decreased quality of life (23%), adverse events (19%), death (14%), and unknown reasons (16%). By site of care, patients treated at an AC were more likely to discontinue therapy due to adverse events (10.5% vs. 2.6%; p = .0323) or experience disease progression (15.1% vs. 4.4%; p = .0117) than those treated in a CP (Figure 1B).

3.4 Overall survival

At a median follow-up for all patients from the last dose of radium-223 of 8.8 months (interquartile range [IQR]: 4.5, 15.1), 42% of patients had died. The estimated median OS from radium-223 initiation was 21.2 months (95% confidence interval: 19.6, 29.2) (Figure 2A). The median OS was not significantly different based on whether patients received radium-223 at a CP versus AC (21.6 vs. 20.7 months; p = .306) (Figure 2B). Notably, OS was significantly longer for patients who received 5–6 doses of radium-223 versus those who received 1–4 doses (23.3 vs. 6.4 months, respectively; p < .001).

3.5 Laboratory values

Laboratory values were assessed at initiation of radium-223, discontinuation, and 3 months postdiscontinuation to identify any signs of prolonged bone marrow suppression. Due to the retrospective nature of the data collection, limited values were available for inclusion at each of these time points. Despite the majority of radium-223 being administered in combination with other agents, there were no laboratory measurements indicating prolonged bone marrow suppression from initiation to discontinuation to 3 months postdiscontinuation, including hemoglobin (median: 12.2 g/dl [IQR: 11.1, 13.0; n = 141]; 11.5 g/dl [IQR: 9.9, 12.4; n = 137]; 10.4 g/dl [IQR: 8.8, 12.2; n = 77]) and platelet counts (median: 215 10⁹/L [IQR: 170, 258; n = 138]; 192 10⁹/L [IQR: 151, 230; n = 132]; 160 10⁹/L [IQR: 107, 235; n = 75]).

4.1 Limitations

As with any retrospective study, there are inherent limitations. This study was descriptive and not powered for comparisons between subgroups. Also, comparisons may be biased by unobserved differences between groups or immortal time bias.³¹ For example, it appears patients treated earlier with radium-223 before chemotherapy had improved survival, but because the subgroups were defined by post-baseline criteria, the criteria may not be independent of the outcome and additional variables may be driving the perceived difference, such as disease burden, stage at treatment initiation, or other unaccounted for variables such as access to care. An attempt to mitigate selection bias was made by incorporating a randomization method for eligible patient selection. Patients in the AC cohort had numerically higher baseline PSA and alkaline phosphatase laboratory values than those in the CP cohort, but we did not compare these values statistically. These differences may reflect a higher tumor burden in patients seeking care at AC, a selection bias in which patients chose to receive treatment at an AC as a second opinion or pursuing a clinical trial, or others. While these factors may have influenced the makeup of our cohorts, it also reflects real-world practice differences and therefore, our results showing no difference in outcomes between practice sites remain relevant and reassuring. While this analysis focused on patients treated at ACs vs CPs, there are other factors that may affect treatment patterns within each type of location, including primary specialty of provider (urology, oncology, nuclear medicine), geographic location (urban, suburban, rural), number of years of experience of treating provider, and others that were not included in the analyses in this study. Recent studies have shown an OS benefit in patients with mCRPC that have genetic aberrations associated with homologous recombination deficiency (HRD) receiving radium-223. As we did not have this data, we were unable to control for the presence or absence of HRD in our cohorts or its potential impact on our results.^{32, 33} Finally, data collection was dependent upon the completeness of the information contained in each chart and reported on the electronic case report form.