Michiel H.F. Poorthuis*, Robin W.M. Vernooij*, R. Jeroen A. van Moorselaar and Theo M. de Reijke

Transcription

1 First-line non-cytotoxic therapy in chemotherapynaive patients with metastatic castration-resistant prostate cancer: a systematic review of 10 randomised clinical trials Michiel H.F. Poorthuis*, Robin W.M. Vernooij*, R. Jeroen A. van Moorselaar and Theo M. de Reijke *Department of Research, Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht, Department of Urology, VU University Medical Center, and Department of Urology, Academic Medical Center, Amsterdam, The Netherlands BJUI Systematic Review Quality Score (based on AMSTAR) The aim of this study is to systematically evaluate all available treatment options in chemotherapy-naive patients with metastatic castration-resistant prostate cancer (mcrpc). We systematically searched PubMed, EMBASE, and the Cochrane libraries up to 1 March 2016 for peer-reviewed publications on randomised clinical trials (RCTs). RCTs were included if progression-free survival (PFS), overall survival (OS), quality of life (QoL), or adverse events (AEs) were quantitatively evaluated. We assessed the risk of bias with the Cochrane Collaboration s tool and graded the evidence with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group s approach. We included 25 articles, reporting on 10 unique RCTs describing seven different comparisons. In one RCT, a prolonged OS and PFS ( quality) were found with abiraterone and prednisone compared to placebo plus prednisone. In one RCT, a prolonged OS and PFS ( quality) were found with enzalutamide compared to placebo. In two RCTs, a prolonged OS ( and moderate quality) was found with 223 radium compared to placebo, but its effect on PFS is unknown. In three RCTs, a prolonged OS (moderate quality) was found with sipuleucel-t compared to placebo, but no prolonged PFS (low quality). In one RCT a prolonged PFS ( quality) was found with orteronel compared to placebo, but no prolonged OS (moderate quality). In one RCT, a prolonged OS (moderate quality) was found with bicalutamide compared to placebo, but its effect on PFS is unknown. In one RCT, a prolonged PFS ( quality) was found with enzalutamide compared to bicalutamide, but its effect on OS is unknown. The best evidence was found for abiraterone and enzalutamide for effective prolongation of OS and PFS to treat chemotherapy-naive patients with mcrpc. However, taking both QoL and AEs into consideration, other treatment modalities could be considered for individual patients. Keywords metastatic castration-resistant prostate cancer, docetaxelna ıve, overall survival, progression-free survival, prostatespecific antigen Introduction Androgen responsive prostate cancer will progress into metastatic castration-resistant prostate cancer (mcrpc) despite usual care with androgen-deprivation therapy (ADT) or bilateral orchidectomy [1,2]. Most patients show initial response rates, but in the long term nearly all patients with prostate cancer develop CRPC characterised by an increase of serum PSA despite castrate levels of testosterone and the development of metastases [3]. Different therapeutic approaches have been suggested to improve the survival in patients with mcrpc. Chemotherapy with the taxane docetaxel or the semisynthetic taxane derivative cabazitaxel have been shown to prolong survival [4 6]. The use of cabazitaxel has not been reported yet as first-line treatment, whereas the use of mitoxantrone has been [7]. Nevertheless, the toxicity profiles of these chemotherapy regimens hamper BJU International 2017 BJU International doi: /bju BJU Int 2017; 119: Published by John Wiley & Sons Ltd. wileyonlinelibrary.com

2 the use of these treatments, especially as patients with mcrpc are generally of older age [4,5]. Moreover, new treatment options with less toxicity are now available. The established role of docetaxel in the treatment of mcrpc created three possible moments for new treatment options: pre-docetaxel, in combination with docetaxel, and postdocetaxel [1,4,5]. All treatment options in combination with docetaxel did not show to be of any benefit. Particularly asymptomatic or minimally symptomatic patients with mcrpc are considered eligible for pre-docetaxel treatment to postpone the initiation of cytotoxic chemotherapy [1]. First-line treatment decisions in chemotherapy-naive patients with mcrpc are challenging, because mcrpc is a heterogeneous disease and different treatment modalities have scarcely been directly compared [8,9]. For this reason, patient-specific and disease-specific factors, such as site and extent of disease involvement, that influence outcome are used to indirectly inform treatment decisions [10]. With the scarcity of direct comparative data, randomised clinical trials (RCTs) comparing available first-line treatments can help in selecting the adequate treatment modality for individual patients based on critical appraisal of study quality, differences in selection of patients in the RCTs, and varying subgroup effects. Therefore, we systematically reviewed the literature for RCTs comparing first-line non-cytotoxic therapy, including abiraterone, enzalutamide, 223 radium, sipuleucel-t, orteronel, or classic androgen receptor-blocker therapy, with placebo, prednisone, or each other in chemotherapy-naive patients with mcrpc. Materials and Methods This systematic review (SR) was performed following the Preferred Reporting Items for Systematic Reviews and Meta- Analyses (PRISMA) recommendations (Appendix S1) [11]. This SR was used as an evidence profile for the Dutch clinical practice guidelines for the treatment of patients with mcrpc and is publically available at: Search Strategy and Selection Criteria We used the electronic databases PubMed, EMBASE, and the Cochrane library (including CDSR, DARE, and CENTRAL) from 2007 to 1 March 2016 (see Appendix S2 for the complete search strategy) to identify articles reporting original data written in English or Dutch, published in peer-reviewed journals, of RCTs comparing first-line treatment, including abiraterone, enzalutamide, 223 radium, sipuleucel-t, orteronel, or classic androgen receptor-blocker therapy, compared to placebo, prednisone, or each other in chemotherapy-naive patients with mcrpc. RCTs were included if one of the following outcomes was quantified: progression-free survival (PFS), overall survival (OS), quality of life (QoL), or adverse events (AEs). Abstracts or conference correspondence were excluded since a risk of bias (RoB) assessment could not be completely performed and these references have not been necessarily subject to a peer-reviewed evaluation. Two authors (M.H.F.P. and R.W.M.V.) screened titles and abstracts for eligible studies, and subsequently reviewed full-text versions of the potentially eligible studies (Appendix S3). We crosschecked the reference lists of identified SRs and included articles until no further eligible studies were found. In case of doubt, studies were discussed in consensus meetings. If multiple articles reported the data of one RCT, we collated the data so that each unique RCT was the unit of analysis. The outcomes of the final follow-up were used. Data Extraction Two authors (M.H.F.P. and R.W.M.V.) extracted the following study characteristics with the data collection form constructed by the CoCanCPG-group ( (i) Methods: study design, conflicts of interest, setting, sample size, duration of study, and existence of a protocol; (ii) Patient characteristics: eligibility criteria and baseline patient characteristics; (iii) Intervention: used intervention and comparator; (iv) Results outcomes: for continuous outcomes: mean or median, standard deviation (SD), and number of patients for both groups. For dichotomous outcomes: event rate and total number of patients for both groups. The data collection forms of all included studies can be found in Appendix S4. Outcomes The primary outcomes comprised: OS and (clinically or radiologically assessed) PFS. The secondary outcomes comprised: QoL, assessed with either the Functional Assessment of Cancer Therapy for patients with Prostate cancer (FACT-P) questionnaire or the EuroQol-5D (EQ-5D) questionnaire, and AEs, graded according to the Common Terminology Criteria for Adverse Events. RoB Assessment and Quality of Evidence Two authors (M.H.F.P. and R.W.M.V.) assessed the RoB of the individual RCTs with the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions [12]. The RoB was assessed according to the following domains: (i) Random sequence generation; (ii) Allocation concealment; (iii) Blinding of participants and personnel; (iv) Blinding of outcome assessment; (v) Incomplete outcome data; (vi) Selective outcome reporting; (vii) Other bias. Each potential source of bias was graded as, low, or unclear RoB. A justification for this judgement has been collected and is included in Appendix S4. We planned to assess whether the outcomes were subject to publication bias by creating a funnel plot; however, this was not possible due to the few studies included [13]. 832 BJU International 2017 BJU International

3 Non-cytotoxic treatment options in chemotherapy-naive patients with mcrpc Two authors (M.H.F.P. and R.W.M.V.) graded the quality of evidence on outcome level with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system to grade the quality of evidence. This includes a judgment regarding the RoB, directness of the evidence, heterogeneity in the data, precision of effect estimates, and risk of publication bias [14]. After the assessment of the evidence, the quality of evidence was categorised as, moderate, low,or very low.to construct evidence profiles we used the GRADE profiler software [15]. Statistical Analysis We analysed the effect of treatment with risk ratios (RRs) or odds ratios (ORs) for dichotomous outcomes, with hazard ratios (HRs) for time-dependent outcomes, and with mean differences for continuous data, presented with their corresponding 95% CIs. Statistical heterogeneity between the included studies was assessed by evaluating forest plots and by using the quantitative measurement of the I 2 test. We used the classification of heterogeneity of the Cochrane Collaboration [12]. A fixed effect meta-analysis was normally used whenever the included studies in the forest plot were functionally identical and whenever solely the identified population of interest was being investigated in the included studies. Whenever the I 2 test resulted in a score of <50% a fixed effect model was applied for the forest plot, when the score was between 50% and 75% a random effect model was used, and when the I 2 was >75% we did not pool the results of the included studies. All statistical analyses were conducted with Review Manager software [13]. Results After screening of publications, 25 articles reporting the data of 10 unique RCTs were included in this SR (Fig. 1) [16 40]. An overview of baseline demographic and clinical characteristics is provided in Table 1. An overview of the outcomes is provided in Table 2. The following results are reported per comparison. Abiraterone Plus Prednisone vs Placebo Plus Prednisone One study was identified that compared abiraterone plus prednisone with placebo plus prednisone [16 21]. In the COU-AA-302 study, chemotherapy-na ıve asymptomatic or mildly symptomatic patients without visceral metastases were randomly assigned to receive abiraterone acetate mg daily plus prednisone 5 mg twice daily or placebo plus prednisone 5 mg twice daily. The median duration of follow-up was 49.2 months. A significant difference in OS was found in favour of the abiraterone plus prednisone group with a HR for death of 0.81 (95% CI ) [19]. Similarly, a significant difference in PFS was found in favour of abiraterone plus prednisone group with a HR of 0.52 (95% CI ) [18]. QoL was measured with the FACT-P scale. After 1 year, a lower proportion of patients had functional status Fig. 1 Flowchart of the study selection. Identification 1749 records identified through MEDLINE via PubMed 932 records identified through EMBASE 262 records identified through Cochrane 370 duplicates removed Included Eligibility Screening 3 articles identified through additional resources 2573 records screened 260 full-text articles assessed for eligibility 25 studies included (describing 10 unique RCTs) 2313 records excluded based on title-abstract screening 238 full-text articles excluded BJU International 2017 BJU International 833

4 Table 1 Overview of baseline demographic and clinical characteristics of the patients in the included randomised clinical trials. Study No. of included patients Experimental agent Comparison Median/mean (range) age, years Exp. Comp. Exp. Comp. Patients previously treated with chemotherapy or docetaxel, n (%) or % Median PSA level, ng/ml a Sites of Duration of metastases, n (%) a follow-up, months, median Trial identification number COU-AA-302 [16 21] PREVAIL [22,23] ALSYMPCA [24 28] 223 Radium vs placebo trial [29,30] IMPACT [31,32] D9901and D9902A [33,34] ELM-PC4 [35] Bicalutamide vs placebo trial [36 39] Abiraterone acetate 1000 mg daily plus prednisone 5 mg twice a day Enzalutamide 160 mg daily 614 b 307 b 223 Radium 50 kbq/kg every 4 weeks; total of 6 injections Radium 50 kbq/kg every 4 weeks; total of 4 injections Sipuleucel-T every 2 weeks, total of 3 infusions Sipuleucel-T every 2 weeks, total of 3 infusions Orteronel 400 mg plus prednisone 5 mg twice daily Bicalutamide 80 mg daily Prednisone 5 mg twice daily plus placebo 71.0 (44 95) 70.0 (44 90) 0 (0) 42.0/37.7 Bone only: 274 (51)/267 (49) Soft tissue or node: 267 (49)/271 (50) Visceral 0 (0)/0 (0) Placebo 72 (43 93) 71 (42 93) 0 (0) 54.1/44.2 Bone: 741 (85.0)/690 (81.7) Placebo 71 (49 90) c 71 (44 94) c Previous docetaxel use: 352 (57)/174 (57) d Lymph node: 437 (50.1)/434 (51.4) Visceral disease (lung or liver): 98 (11.2)/106 (12.5) 146/173 e 2 bone: 614 (100)/307 (100) Visceral: 0 (0) 49.2 NCT ~22 NCT (planned) Placebo 73 (57 88) 72 (60 84) N/A f 167/233 g Bone: 64 (100) 18 (range 18 24) Placebo 72 (49 91) 70 (40 89) Previous chemotherapy use: 19.6% vs 15.2%. Previous docetaxel use: 15.5% vs 12.3%. 51.7/47.2 Bone only: (50.7)/(43.3) Soft tissue only: (7.0)/ (8.2) Bone and soft tissue: (41.9)/(48.5) Visceral: (0)/(0) Placebo 72 (47 85) 71 (50 87) 6.9% vs 9.0% h 50.7/45.8 Bone only: (44.5)/(26.7) Soft tissue only: (8.2)/ (13.3) Bone and soft tissue: (47.3)/(60.0) Visceral: (0)/(0) Placebo plus predisone 5 mg twice daily Placebo <75 years: 53 (52.0%) 75 years: 49 (48.0%) 71.0 ( ) 72.0 ( ) 0 (0) 55.8/55.3 Bone: 730 (93)/705 (91) Liver: 27 (3)/45 (6) Lung: 71 (9)/70 (9) Lymph node: 346 (44)/ 329 (42) <75 years: 50 (49.5%) 75 years: 51 (50.5%) 0 (0) <60: 40 (39.2%)/37 (36.6%) 60: 62 (60.8%)/64 (63.4%) Bone: 40 (39.2)/40 (39.6) Lymph node: 28 (27.5)/ 38 (37.6) Other: 2 (2.0)/3 (3.0) NCT NCT NCT N/A NCT and NCT NCT N/A 834 BJU International 2017 BJU International

5 Non-cytotoxic treatment options in chemotherapy-naive patients with mcrpc Table 1 (continued) Trial identification number Sites of Duration of metastases, n (%) a follow-up, months, median Median PSA level, ng/ml a Patients previously treated with chemotherapy or docetaxel, n (%) or % Comparison Median/mean (range) age, years Experimental agent Study No. of included patients Exp. Comp. Exp. Comp. 16.7/20.0 NCT (48 91) 71 (50 96) 0 (0) 22/21 Bone only: 83 (45)/92 (48) Soft tissue only: 36 (20)/29 (15) Bone and soft tissue: 64 (35)/69 (36) Missing: 1 (1)/1 (1) Enzalutamide 160 mg daily Bicalutamide 50 mg daily TERRAIN [40] a If possible, data are provided for the experimental group and the control group respectively. b The number of included patients without previous docetaxel use was 262 and 113 in the experimental group and the comparison group, respectively. c The median age of included patients without previous docetaxel use was 74 (49 90) and 74 (52 94) years in the experimental group and the comparison group, respectively. d These patients that had not received docetaxel, declined or were not healthy enough to receive docetaxel. e The median PSA level of included patients without previous docetaxel use was 88 (4 5837) and 98 (2 2210) lg/l for the experimental group and the comparison groups, respectively. f Patients with chemotherapy during the past 6 weeks were excluded from this study. g Two consecutive rising amounts of serum PSA was one of the inclusion criteria. h These patients received chemotherapy 6 months prior to inclusion. Comp, comparison group; Exp, experimental group; IQR, interquartile range; kbq, kilobecquerel. deterioration in the abiraterone plus prednisone group (64.8%) than in the placebo plus prednisone group (79.5%) [17]. This resulted in a RR of 0.82 (95% CI ). There were more AEs in the abiraterone plus prednisone group. In all, 290 patients (54%) had grade 3 or 4 AEs in the abiraterone plus prednisone group vs 236 patients (44%) in the placebo plus prednisone group [17]. This resulted in a RR for AEs of 1.22 (95% CI ). All four outcomes were graded as a quality of evidence (Table 3). Enzalutamide vs Placebo One study was identified that compared enzalutamide with placebo [22,23]. In the PREVAIL study, chemotherapynaive asymptomatic or mildly symptomatic patients were randomly assigned to receive enzalutamide 160 mg daily or placebo daily; 12% of the included patients had visceral metastases. This study was stopped after the planned interim analysis, because of the observed benefit in the two coprimary outcomes OS and PFS. The median duration of follow-up was ~22 months. A significant difference in OS was found in favour of the enzalutamide group with a HR of 0.71 (95% CI ) [22]. Similarly, a significant difference in PFS was found in favour of the enzalutamide group with a HR of 0.19 (95% CI ) [22]. Both the FACT-P and the EQ-5D scale were used to measure QoL. Patients treated with enzalutamide had significantly better scores on both scales [23]. This resulted in a RR of 1.73 (95% CI ) measured by FACT-P and a RR of 1.74 (95% CI ) measured by EQ-5D. There were more AEs in the enzalutamide group. In all, 43% of patients in the enzalutamide group had grade 3 or 4 AEs vs 37% of the patients in the placebo group [22]. This resulted in a RR of 1.16 (95% CI ) in favour of the placebo group. All four outcomes were graded as a quality of evidence (Table 4). 223 Radium vs Placebo Two studies were identified that compared 223 radium with placebo [24 30]. In the ALSYMPCA study, 921 patients with two or more bone metastases and no known visceral metastases were randomly assigned to receive six injections of 223 radium or placebo [24 28]. This RCT included 526 (57%) patients with previous docetaxel use, and docetaxelnaive patients considered unfit for docetaxel. The planned duration of follow-up was 36 months. In the study of Nilsson et al. [29,30], 64 chemotherapy-naive patients with multiple bone metastases or one painful lesion with two consecutive rising amounts of PSA were randomly assigned BJU International 2017 BJU International 835

6 Table 2 Overview of outcomes Study OS PFS QoL AEs COU-AA-302 [16 21] PREVAIL [22,23] ALSYMPCA [24 28] 223 Radium vs placebo trial [29,30] IMPACT [31,32] D9901/ D9902A [33,34] ELM-PC4 [35] Bicalutamide vs placebo trial [36 39] TERRAIN [40] Deaths at end of follow-up: Abiraterone: 354/546 (65%) Placebo: 387/542 (71%) HR 0.81 ( ) OS at 12 months: Enzalutamide 797/872 (91%) Placebo: 701/845 (83%) HR 0.71 ( ) Median: 223 Radium: 16.1 ( ) months Placebo: 11.5 ( ) months HR 0.69 ( ) c OS at 24 months: 223 Radium: 10/33 (30%) Placebo: 4/31 (13%) HR 0.48 ( ) Median: Sipuleucel-T: 25.8 months Placebo: 21.7 months Deaths at end of follow-up: Sipuleucel-T: 210/341 (61.6%) Placebo: 121/171 (70.8%) HR 0.78 ( ) d Sipuleucel-T: 23.2 ( ) months Placebo: 18.9 ( ) months HR 1.50 ( ) Median: Orteronel: 31.4 (28.6 NC) Placebo: 29.5 (27.0 NC) HR 0.92 ( ) OS at 5 years: Bicalutamide: 75.3% Placebo: 63.4% HR 0.78 ( ) Not reported Radiographic PFS: Abiraterone: 292/546 (53.5%) Placebo: 352/542 (64.9%) HR 0.52 ( ) Radiographic PFS at 12 months: Enzalutamide: 65% Placebo: 14% HR 0.19 ( ) Not reported FACT-P total score deterioration at 1 year: Abiraterone: 354/546 (64.8%) Placebo: 431/542 (79.5%) RR 0.82 ( ) FACT-P: Improvement at any time during the trial: Enzalutamide: 327/826 (40%) Placebo: 181/790 (23%) RR 1.73 ( ) a EQ-5D: Improvement at any time during the trial: Enzalutamide: 224/812 (28%) Placebo: 99/623 (16%) RR 1.74 ( ) a FACT-P: Improvement at week 24: b 223 Radium: 57/314 (18.2%) Placebo: 10/120 (8.3%) RR 2.18 ( ) a Grade 3 4 AEs: Abiraterone: 290/542 (54%) Placebo: 236/540 (44%) RR 1.22 ( ) Any Grade 3 Enzalutamide: 374/871 (43%) Placebo: 313/844 (37%) RR 1.16 ( ) a Grade 3 AEs 223 Radium: 145/253 (57.3%) Placebo: 77/130 (59.2%) RR 0.97 ( ) a EQ-5D: Meaningful improvement at 24 weeks: b Radium: 75/343 (21.9%) Placebo: 20/131 (15.3%) RR 1.43 ( ) a Not reported Not reported Haematological grade 3 4 AE: 223 Radium: 3/33 (9.1%) Placebo: 2/31 (6.5%) RR 1.29 ( ) a Objective disease progression: Sipuleucel-T: 14.6 weeks Placebo: 14.4 weeks HR 0.95 ( ) Clinical disease progression: HR 0.92 ( ) Sipuleucel-T: 11.1 ( ) weeks Placebo: 9.7 ( ) weeks HR 1.26 ( ) Radiographic PFS: Orteronel: 13.8 ( ) months Placebo: 8.7 ( ) months HR 0.71 ( ) Not reported Median PFS: Enzalutamide: 15.7 ( ) months Bicalutamide: 5.8 ( ) months HR 0.44 ( ) Not reported Not reported Not reported FACT-P total score difference between baseline and 24 weeks: Bicalutamide: 4.86 (SD 18.44) Placebo: 1.67 (SD 17.97) P = FACT-P: improvement at any time during study: Enzalutamide: 22% Bicalutamide: 33% No risk estimate calculable Serious AEs: 223 Radium: 8/33 (24.2%) Placebo: 14/31 (45.2%) RR 0.52 ( ) a Grade 3 5 AEs: Sipuleucel-T: 107/338 (31.7%) Placebo: 59/168 (35.1%) RR 0.90 ( ) a Grade 3 4 AEs: Sipuleucel-T: 49/147 (33.3%) Placebo: 21/67 (31.3%) RR 1.06 ( ) a Grade 3 4 AEs: Orteronel: 465/784 (59%) Placebo: 313/770 (41%) RR 1.46 ( ) Not reported Grade 3 5 AEs: Enzalutamide: 73/183 (40%) Bicalutamide 72/189 (38%) RR 1.05 ( ) a Self-estimated risk ratio (RR). b These data are only provided for the intention-to-treat population. c Subgroup of patients without previous docetaxel use (that were unsuitable for or declined docetaxel). d In the subgroup of patients that did not previously receive docetaxel, no significant difference in OS was found. However, this was only visualised in a figure without quantified data. NC, not calculable. 836 BJU International 2017 BJU International

7 Non-cytotoxic treatment options in chemotherapy-naive patients with mcrpc Table 3 Summary of findings for the comparison: Abiraterone plus prednisone vs placebo plus prednisone. Patient or population: Chemotherapy-naive asymptomatic or mildly symptomatic patients with mcrpc and without visceral metastases Intervention: Abiraterone plus prednisone Comparison: Placebo plus prednisone Outcomes Illustrative comparative risks* Relative effect Assumed risk Corresponding risk Placebo plus prednisone Abiraterone plus prednisone No of participants (studies) Quality of the evidence (GRADE) Comments PFS 649 per per 1000 ( ) OS 714 per per 1000 Deaths ( ) QoL 795 per per 1000 FACT-P ( ) deterioration after 1 year Toxicity AE grade per per 1000 ( ) HR 0.52 ( ) HR 0.81 ( ) RR 0.82 ( ) RR 1.22 ( ) 1088 (1 study) 1088 (1 study) 1088 (1 study) 1082 (1 study) *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. to receive four injections of 223 radium or placebo. The duration of follow-up was 18 months, ranging from 18 to 24 months. Both studies reported data on the OS. A significant difference in favour of the 223 radium group was found in OS in the subgroup analysis of patients without previous docetaxel use (that were unsuitable for or declined docetaxel) in the ALSYMPCA study with a HR of 0.69 (95% CI ) [25]. However, the power is unproven, as the trial was not designed to compare 223 radium with placebo within this specific subgroup. Similarly, a significant effect in the study of Nilsson et al. [29,30] was found, with a HR of 0.48 (95% CI ). We were not able to meta-analyse these data, as the study of Nilsson et al. did not provide values of variance (e.g. SD or 95% CI). Both studies did not report data on PFS. Both the FACT-P and the EQ-5D scale were used to measure QoL in the intent-to-treat population in the ALSYMPCA study [28]. For the FACT-P, an improvement at 24 weeks in total score was found in 18.2% (57/314) of the patients treated with 223 radium and in 8.3% (10/120) of the patients treated with placebo [28]. This resulted in a RR of 2.18 (95% CI ). For the EQ-5D, an improvement at 24 weeks in total score was found in 21.9% (75/343) of the patients treated with 223 radium and in 15.3% (20/131) of the patients treated with placebo [28]. This resulted in a RR of 1.43 (95% CI ). The study of Nilsson et al. [29,30] did not report on QoL. The ALSYMPCA study measured grade 3 or 4 AEs in the intent-to-treat population and Nilsson et al. [29,30] measured serious AEs. In the ALSYMPCA study, 57.3% (145/253) of the patients in the 223 radium group and 59.2% (77/130) of the placebo group had grade 3 or 4 AEs [25]. This resulted in a RR of 0.97 (95% CI ). In the study of Nilsson et al. [30], 24.2% (8/33) of the patients in the 223 radium group and 45.2% (14/31) of the patients in the placebo group had serious AEs. This resulted in a RR of 0.52 (95% CI ). Although, both studies reported data on AEs, it was not possible to pool the data, due to the difference in the definition of measuring AEs. The outcomes of OS and QoL in the ALSYMPCA study were graded as a quality of evidence. The outcomes of AEs in the ALSYMPCA study and OS in the study of Nilsson et al. [29,30] were downgraded to a moderate quality of evidence, because relatively few patients were included in the study and the CI includes two different directions, respectively. Similarly, the outcomes of AEs and serious AEs in the study of Nilsson et al. [29,30] were downgraded to a low quality of evidence, because relatively few patients were included and no blinding after 12 months (total follow-up 24 months) was performed, respectively (Table 5). Sipuleucel-T vs Placebo Three studies were identified that compared sipuleucel-t with placebo [31 34]. In all studies, patients were randomly assigned to receive sipuleucel-t every 2 weeks with a total of three infusions or placebo. In the IMPACT study, 512 asymptomatic or mildly symptomatic patients without visceral metastases were included [31]. This RCT included 19.6% patients post-chemotherapy in the sipuleucel-t group. The median duration of follow-up was 34.1 months. Small et al. [34] included 127 patients in the D9901 study. Higano et al. [33] integrated the data of the D9901 study with the identical small phase 3 trial D9902A. In all, 225 patients with BJU International 2017 BJU International 837

8 Table 4 Summary of findings for the comparison: Enzalutamide vs placebo. Patient or population: Chemotherapy-na ıve asymptomatic or mildly symptomatic patients with mcrpc Intervention: Enzalutamide Comparison: Placebo Outcomes Illustrative comparative risks* Assumed risk Placebo Corresponding risk Enzalutamide Relative effect No of participants (studies) Quality of the evidence (GRADE) Comments PFS at 12 months 191 per per 1000 (31 48) OS at 12 months 830 per per 1000 Survival ( ) QoL 229 per per 1000 FACT-P ( ) QoL 159 per per 1000 EQ-5D ( ) Toxicity 371 per per 1000 AE grade 3 ( ) HR 0.19 ( ) HR 0.71 ( ) RR 1.73 ( ) RR 1.74 ( ) RR 1.16 ( ) 1633 (1 study) 1717 (1 study) 1616 (1 study) 1435 (1 study) 1715 (1 study) *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. radiological evidence of metastases, with a 3 months chemotherapy-free interval, and no visceral metastases or pain from bone metastases were included [33]. The duration of follow-up was not reported. In the IMPACT study, a significant difference in OS in favour of the sipuleucel-t group was found with a HR of 0.78 (95% CI ) [31]. In the subgroup of patients that did not previously receive docetaxel, no significant difference in OS was found. However, this was only visualised in a figure without quantified data. No significant difference in PFS was found. In the IMPACT study, the median time to objective disease progression was 14.6 weeks in the sipuleucel-t group and 14.4 weeks in the placebo group, resulting in a HR of 0.95 (95% CI ) [31]. No significant difference in clinical disease progression was found, with a HR of 0.92 (95% CI ) [31]. In the D9901 and D9902A studies, the median OS was 23.2 months in the sipuleucel-t group and 18.9 months in the placebo group with a HR of 1.50 (95% CI ) [33]. In the D9901 and D9902A studies, the median time to progression was 11.1 weeks in the sipuleucel-t group and 9.7 weeks in the placebo group with in a HR of 1.26 (95% CI ) [33]. In the D9901 and D9902A studies, the HR is defined as the risk in patients treated with placebo divided by the risk in patients treated with sipuleucel-t. Therefore, a HR of 1 indicates a er risk for patients treated with placebo relative to sipuleucel-t. Consequently, a pooled analysis of the HRs of the three studies is not possible. None of the studies reported data on QoL. Both studies measured AEs. In the IMPACT study, 31.7% (107/338) of the patients in the sipuleucel-t group and 35.1% (59/168) of the placebo group had grade 3, 4 or 5 AEs [31]. This resulted in a RR of 0.90 (95% CI ). In the D9901 and D9902A studies, 33.3% (49/147) of the patients treated with sipuleucel-t and 31.3% (21/67) of the placebo group had grade 3 or 4 AEs [33]. This resulted in a RR of 1.06 (95% CI ). Similarly, the pooled result was not significant with a RR of 0.94 (95% CI ; I 2 = 0%) (Fig. 2). The outcome of PFS was downgraded to a low quality of evidence, because no description of randomisation, allocation concealment, blinding, and selective outcome reporting was provided. The outcome AEs was downgraded to a low quality of evidence, because the CI included two different directions. The outcome OS was downgraded to a moderate quality of evidence, because no description of randomisation, allocation concealment, blinding, and selective outcome reporting was provided (Table 6). Orteronel Plus Prednisone vs Placebo Plus Prednisone One study was identified that compared orteronel with placebo [35]. In the ELM-PC4 study, chemotherapynaive patients with radiographic nodal, bone or visceral metastases were randomly assigned to receive orteronel 400 mg daily plus prednisone 5 mg twice daily or placebo plus prednisone 5 mg twice daily [35]. The median duration of follow-up was 20.7 months. No significant difference in OS was found, with a HR of 0.92 (95% CI ) [35]. A significant difference in radiographic PFS was found, with a HR of 0.71 (95% CI ) [35]. 838 BJU International 2017 BJU International

9 Non-cytotoxic treatment options in chemotherapy-naive patients with mcrpc Table 5 Summary of findings for the comparison: 223 Radium vs placebo. Patient or population: Patients with mcrpc and 2 bone metastases and no known visceral metastases. This RCT included patients with previous docetaxel use and docetaxel-naive patients considered unfit for docetaxel Intervention: 223 Radium Comparison: Placebo Outcomes Illustrative comparative risks* Assumed risk Placebo Corresponding risk 223-Radium Relative effect No of participants (studies) Quality of the evidence (GRADE) Comments PFS No evidence available OS (continuous) 16.1 months 11.5 months HR 0.69 ( ) 474 (1 study) Not possible to calculate the mean difference since no details on statistical variability were provided. OS at 24 months 129 per per 1000 (35 114) HR 0.48 ( ) 64 (1 study) moderate QoL meaningful improvement at week 24 by FACT-P total score QoL meaningful improvement at week 24 by EQ-5D utility score 83 per per 1000 (96 343) 153 per per 1000 ( ) Toxicity 592 per per 1000 ( ) Toxicity 65 per per 1000 Haematological AE grade (15 467) 3 4 Serious AEs 452 per per 1000 ( ) RR 2.18 ( ) RR 1.43 ( ) RR 0.97 ( ) RR 1.29 ( ) RR 0.52 ( ) 434 (1 study) 474 (1 study) moderate 383 (1 study) moderate 64 (1 study) low, 64 (1 study) low, *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. High risk of bias because of no blinding of patients and personnel after 12 months with 24 months of follow-up. Lower interval of the CI results in different conclusion than the upper limit. No data on QoL were reported. The ELM-PC4 study measured grade 3 or 4 AEs. In all, 59% (465/784) of the patients in the orteronel plus prednisone group and 41% (313/770) of the prednisone group had grade 3 or 4 AEs [35]. This resulted in a RR of 1.46 (95% CI ). The outcomes of PFS and AEs were graded as a quality of evidence. The outcome of OS was downgraded to a moderate quality of evidence, because the effect between the control and intervention group did not differ significantly. Consequently, recommendations about the use of the intervention can differ between the lower and upper limit of the CI (Table 7). Bicalutamide vs Placebo One study was identified that compared bicalutamide with placebo [36 39]. In this RCT, 203 chemotherapy-naive patients were randomly assigned to receive bicalutamide 80 mg daily or placebo. The median duration of follow-up was 62.4 months. A significant difference in OS was found, with a HR of 0.78 (95% CI ) in favour of the bicalutamide group [39]. No data on PFS were reported. The FACT-P was used to measure QoL [38]. No significant difference between the bicalutamide and placebo group was found in the improvement at 24 weeks compared to baseline. No data on AEs were reported. The outcome of OS was graded as a moderate quality of evidence, as no details about randomisation, blinding, and allocation were provided. The outcome of QoL was downgraded to a low quality of evidence, because no details about randomisation, blinding, and allocation were provided. Next to that, no risk estimate could be calculated hampering the evaluation of size and direction of the effect (Table 8). Bicalutamide vs Enzalutamide One study was identified that compared bicalutamide with enzalutamide [40]. In the TERRAIN study, 375 asymptomatic or mildly symptomatic patients with at least two bone lesions or BJU International 2017 BJU International 839

10 Fig. 2 Meta-analysis of sipuleucel-t vs placebo for the outcome of AEs (grade 3). Study D9901 and D9902A IMPACT Experimental Control Risk Ratio Events Total Events Total Weight M-H, Fixed, 95% CI % 1.06 [0.70, 1.62] % 0.90 [0.70, 1.17] Risk Ratio M-H, Fixed, 95% CI Total patients % 0.94 [0.76, 1.18] Total events Heterogeneity: Chi 2 = 0.43, df = 1 (P = 0.51); I 2 = 0% Test for overall effect Z = 0.50 (P = 0.61) Favours sipuleucel-t Favours placebo Table 6 Summary of findings for the comparison: Sipuleucel-T vs placebo. Patient or population: Asymptomatic or mildly symptomatic patients with mcrpc and without visceral metastases Intervention: Sipuleucel-T Comparison: Placebo Outcomes Illustrative comparative risks* Assumed risk Placebo Corresponding risk Sipuleucel-T Relative effect No of participants (studies) Quality of the evidence (GRADE) Comments PFS Unknown Unknown IMPACT: HR 0.95 ( ) D9901 and D9902A: HR 1.26 ( ) OS Unknown Unknown IMPACT: HR 0.78 ( ) D9901 and D9902A: HR 1.50 ( ) 3 studies low, 3 studies moderate QoL No evidence available Toxicity AE grade 3 or 4 AEs 340 per per 1000 (259 to 402) RR 0.94 ( ) 720 (3 studies) low, Continuous outcomes are not provided in the studies and HRs of the two studies could not be pooled, because in the IMPACT study, the risk estimate was defined as the risk in patients treated with sipuleucel-t divided by the risk in patients treated with placebo and in the D9901 and D9902A studies, the risk estimate was defined as the risk in patients treated with placebo divided by the risk in patients treated with sipuleucel-t. Continuous outcomes are not provided in the studies and HRs of the two studies could not be pooled, because in the IMPACT study, the risk estimate was defined as the risk in patients treated with sipuleucel-t divided by the risk in patients treated with placebo and in the D9901 and D9902A studies, the risk estimate was defined as the risk in patients treated with placebo divided by the risk in patients treated with sipuleucel-t. *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. IMPACT: unclear risk of bias due to no description of allocation concealment. D9901 and D9902A: unclear risk of bias due to no description of randomisation methods, allocation concealment, blinding methods, and protocol. Lower interval of the CI results in different conclusion than the upper limit. soft tissue metastases were randomly assigned to receive bicalutamide 50 mg daily or enzalutamide 160 mg daily [40]. The median duration of follow-up was 16.7 months in the bicalutamide group and 20.0 months in the enzalutamide group. No data on OS were reported. A significant difference in PFS was found, with a HR of 0.44 (95% CI ) [40]. The FACT-P was used to measure QoL. In all, 22% of the patients in the enzalutamide group and 33% of the patients in the bicalutamide group had improvement at any time during the study [40]. No risk estimate was reported and we were not able to calculate a risk estimate. The TERRAIN study measured grade 3, 4 or 5 AEs. In all, 40% (73/183) of the enzalutamide group and 38% (72/189) of 840 BJU International 2017 BJU International

11 Non-cytotoxic treatment options in chemotherapy-naive patients with mcrpc Table 7 Summary of findings for the comparison: Orteronel plus prednisone vs placebo plus prednisone. Patient or population: Chemotherapy-naive patients with mcrpc and radiographic nodal, bone or visceral metastases Intervention: Orteronel plus prednisone Comparison: Placebo plus prednisone Outcomes Illustrative comparative risks* Assumed risk Placebo Corresponding risk Orteronel Relative effect No. of participants (studies) Quality of the evidence (GRADE) Comments PFS Unknown Unknown HR 0.71 ( ) 1554 (1 study) Continuous outcomes are not provided in the studies OS Unknown Unknown HR 0.92 ( ) 1554 (1 study) moderate Continuous outcomes are not provided in the studies QoL No evidence available Toxicity AE grade per per 1000 ( ) RR 1.46 ( ) 1554 (1 study) *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). Lower interval of the CI results in different conclusion than the upper limit. GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. Table 8 Summary of Findings for the comparison: Bicalutamide vs placebo. Patient or population: Chemotherapy-naive patients with mcrpc Intervention: Bicalutamide Comparison: Placebo Outcomes Illustrative comparative risks* Assumed risk Placebo Corresponding risk Bicalutamide Relative effect No of participants (studies) Quality of the evidence (GRADE) Comments PFS No evidence available OS N/A N/A HR 0.78 ( ) 203 (1 study) moderate Continuous outcomes are not provided in the studies QoL The mean QoL in the intervention groups was 203 (1 study) low 3.19 er ( er) Toxicity 0 No evidence available *The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. Unclear risk of bias because of details lacking regarding randomisation, allocation concealment, blinding, and a protocol. the bicalutamide group had AEs [40]. This resulted in a RR of 1.05 (95% CI ). The outcome of PFS was graded as a quality of evidence. The outcome QoL was graded as a moderate quality of evidence, because the upper and lower limit of the CI indicate different directions (Table 9). Discussion In this SR, 10 unique RCTs reporting on seven different comparisons, assessing the efficacy and safety of first-line therapy for chemotherapy-naive patients with mcrpc, were included. To summarise, the best available evidence for the outcomes OS and PFS was found for abiraterone plus prednisone and for enzalutamide. 223 Radium and bicalutamide showed a prolonged OS, but their effect on PFS is unknown. Sipuleucel-T showed a prolonged OS, but no effect on PFS. Orteronel showed a prolonged PFS, but no effect on OS. PFS was prolonged with enzalutamide compared to bicalutamide, but its effect on OS is unknown. The critical appraisal of study quality, differences in selection of patients in the RCTs, QoL, and AEs should be considered when treatment possibilities are discussed with patients once they develop mcrpc. BJU International 2017 BJU International 841

12 Table 9 Summary of findings for the comparison: Enzalutamide vs bicalutamide. Patient or population: Asymptomatic or mildly symptomatic patients with mcrpc and at least two bone lesions or soft tissue metastases Intervention: Enzalutamide Comparison: Bicalutamide Outcomes Illustrative comparative risks* Assumed risk Bicalutamide Corresponding risk Enzalutamide Relative effect No of participants (studies) Quality of the evidence (GRADE) Comments PFS Unknown Unknown HR 0.44 ( ) 203 (1 study) Continuous outcomes are not provided in the studies OS No evidence available QoL 0 No evidence available Toxicity AE grade per per 1000 ( ) RR 1.05 ( ) 372 (1 study) moderate *The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI, confidence interval; RR, risk ratio; HR, hazard ratio. GRADE Working Group grades of evidence: High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. Lower interval of the CI results in different conclusion than the upper limit. Appropriate Selection of Patients Selection of patients is a challenge, especially, as guidance for treatment decisions by clinically validated prognostic characteristics are not available [8]. The TERRAIN study compared enzalutamide with bicalutamide, but all other RCTs compared the experimental agent to placebo and/or prednisone. Differences in selection of patients and varying subgroup effects in the RCTs can indirectly inform treatment decisions. The COU-AA-302 study, PREVAIL study and TERRAIN study restricted inclusion to asymptomatic and minimally symptomatic docetaxel-naive patients, establishing the role of separate treatment with enzalutamide or abiraterone in the pre-chemotherapy setting. The addition of abiraterone in patients treated with enzalutamide is currently under investigation in the phase III ALLIANCE study (NCT ). In addition, in the PREVAIL study, 204 of the (12%) included patients had visceral metastases. In those patients, a trend was seen towards an increase in OS in patients treated with enzalutamide compared to placebo with a HR of 0.82 (95% CI ). The role of orteronel is less established, because, although radiographic PFS was prolonged, no OS benefit was found in the ELM-PC4 study [35]. Only patients with a PSA level of >50 ng/ml at baseline showed a significantly prolonged OS [35]. The ELM-PC4 study included 277 of (17.8%) patients with visceral metastases [35]. However, the effect of treatment with orteronel was not measured separately for these patients. Based on these clinical data the company decided to stop the development of orteronel in June The ALSYMPCA study included patients that were either pretreated with docetaxel or were unsuitable for or declined docetaxel. Moreover, most of the patients had more than six symptomatic bone metastases. The role of 223 radium is therefore at this moment only proven for the defined group of patients that had already received docetaxel. For this reason, the National Institute for Health and Care Excellence (NICE) has not recommended 223 radium in docetaxel-naive patients [41]. The role of 223 radium in addition to abiraterone or enzalutamide for a broader indication is currently being investigated in a phase III trial (NCT ), as is optimisation of the dosing schedule in a phase II trial (NCT ). In the first two trials comparing sipuleucel-t to placebo, a significant prolongation of OS was found, but no benefit in terms of PFS. Similar results were found in the IMPACT trial. Based on the OS benefit, sipuleucel-t was approved by the USA Food and Drug Administration (FDA), but it is not available in Europe [1,2,42]. Sequence and Combination of Different Treatment Options The efficacy and safety of the sequential use of different treatment options is a clinical important question, which remains to be answered and cannot be solved by the present analysis. Sipuleucel-T has been proposed for early stages of mcrpc, as patients with lower PSA levels are more likely to benefit from sipuleucel-t [8,32,43]. Data from observational studies suggests cross-resistance between enzalutamide and abiraterone [44 46]. A 50% decline in PSA level was found in 18 34% of the patients treated with enzalutamide after abiraterone, but without prior docetaxel treatment. Based on expert consensus and available evidence at this moment, enzalutamide should not be considered after progression on abiraterone due to cross-resistance [43]. Although data from 842 BJU International 2017 BJU International