Real-world outcomes for first line next-generation hormonal agents in metastatic prostate cancer: a systematic review

Prostate cancer (PC) is the second most common global cancer in men and the fourth most common global cancer overall [1]. Approximately 10–20% of men with advanced prostate cancer will develop castrate resistant PC (CRPC) within 5 years of diagnosis and an estimated 84% will have metastases at the time of CRPC diagnosis [2]. Metastatic CRPC (mCRPC) is an advanced stage of PC characterized by disease progression and distant spread despite androgen depletion therapy (ADT) [3]. Despite access to novel therapies, 5-year survival in mCRPC is still less than 15% [4].

The treatment landscape for mCRPC has evolved over the last 2 decades with the introduction of docetaxel in 2004, and the marketing authorizations of the next-generational hormonal (NHA) agents abiraterone and enzalutamide between 2011 and 2012 [5,6]. Approvals for NHAs in the first-line (1L) treatment of mCRPC were supported by the results of two phase III placebo-controlled randomized controlled trials (RCT): COU-AA-302 (abiraterone) [7] and PREVAIL (enzalutamide) [8]. In these studies, NHAs significantly improved radiological progression-free survival (PFS) and overall survival (OS) versus placebo, achieving a median PFS of approximately 16 to 20 months [7,8] and a median OS of approximately 34 to 36 months [7,8] In many countries, the most common 1L treatment strategy is now NHA followed by chemotherapy [9–11].

Since the approval of abiraterone and enzalutamide, there has been expanding interest in the real-world outcomes of NHA treatment. Real-world evidence (RWE) provides insight into how treatments are used in everyday practice. This is important as outcomes in clinical practice can differ to those observed in RCTs due to differences in trial participants, practice and treatment availability [12]. RWE can also inform the application of evidence from RCTs to clinical decision making, providing insights beyond those addressed in RCTs, and help identify areas of current unmet medical need [13]. In this study, we report the results of a systematic literature review to identify and summarize RWE of NHA treatment in 1L mCRPC. This includes studies of treatment effectiveness, safety, health-related quality-of-life (HRQoL) and cost/healthcare resource use (HRU). Where reported, we summarize RWE of outcomes by treatment and include comparative RWE studies.

Materials & methods

The systematic literature review (SLR) methodology followed guidance from the Cochrane Collaboration [14] and the Centre for Reviews and Dissemination (York, UK) [15], and was documented in accordance with the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [16] (Supplementary Materials). An a priori (unpublished) protocol was developed that described the SLR methodology and eligibility criteria (Supplementary Table 1).

Search strategies were developed by an information specialist, and independently peer reviewed using the Canadian Agency for Drugs and Technologies in Health checklist [17]. Searches were conducted in Embase^®, MEDLINE^®, EBM Reviews and EconLit via Ovid. The searches were limited by date (2011 onwards) to cover the period after which NHAs were first approved for mCRPC and therefore available in routine ‘real world’ clinical practice. Searches were not limited by language or publication status. Database searches were conducted on 18 November 2021. Full search strategies are provided in Supplementary Tables 2–6. Supplementary searches were conducted in six conference proceedings for the last 3 years of availability, including: American Society of Clinical Oncology (ASCO), ASCO Genitourinary Cancers Symposium, American Urological Association (AUA), European Association of Urology (EAU), European Society for Medical Oncology (ESMO) and International Society for Pharmacoeconomics and Outcomes Research (ISPOR). The bibliographies of all included studies and relevant systematic reviews were also checked to identify additional studies.

Studies were eligible for inclusion based on pre-specified criteria (Supplementary Table 1). Briefly, studies had to report on adult patients with mCRPC treated in the 1L setting with at least one of abiraterone or enzalutamide and at least one clinical, safety, HRQoL or cost/HRU outcome of interest. A conservative approach to the inclusion of studies was taken when the reporting of details around treatment line was unclear. No restrictions were applied based on geography or language. Non-English language studies were translated using machine translation methodologies.

Two reviewers independently screened records at the title/abstract and full paper stages. Data were extracted by one reviewer and checked by a second reviewer. Risk of bias assessments was performed using the Cochrane ROBINS-I tool for non-randomized studies of interventions [18] (for studies reporting on clinical/safety/HRQoL outcomes) or the checklist adapted to cost of illness by Molinier et al. [19] (for cost and healthcare resource use studies). For ROBINS-I, risk of bias was assessed at the outcome level (radiographic PFS or PFS; if these outcomes were not reported, the reported primary outcome was assessed). For Molinier et al. [19], risk of bias was assessed at the study level. Any disputes during screening, extraction or risk of bias assessment were resolved by consensus or discussion with a third reviewer.

Data were summarized qualitatively and organised by outcome. No quantitative syntheses were conducted given the expected heterogeneity across RWE studies. To reflect actual observations, the summary of median data was based on empirical estimates only, excluding data derived through matching or adjustment. The summary of comparative effectiveness data contained only the results of studies that controlled for the effects of confounding from baseline imbalances (e.g., by matching or regression adjustment). For each outcome, the results of comparative studies were displayed on a forest plot showing the hazard ratios of abiraterone versus enzalutamide. Where required, reported hazard ratios for enzalutamide versus abiraterone were statistically inverted to yield results for the desired comparison. Waterfall plots were prepared using GraphPad Prism version 9.5.1 for Mac, GraphPad Software, San Diego, California, USA, www.graphpad.com.

Results

Overview of included studies

Electronic database searches identified a total of 3168 citations. Following the removal of 642 duplicates, 2526 citations were screened based on title and abstract. After title and abstract screening, 2201 records were excluded and 325 records were included for full-text screening. After full text review, 229 records were excluded (Supplementary Table 7). One record was identified through hand searching. Overall, 88 studies (with 97 publications) were included in the systematic review. The PRISMA flow of studies is summarized in Figure 1 .

An overview of included study characteristics is provided in Supplementary Table 8. Most studies were retrospective (n = 71), 15 were prospective and two were unclear or reported a mixed design. There were 11 studies each reporting real-world data (RWD) from the USA [20–30] and Italy [31–41]; ten studies from Canada [42–51]; and eight multinational studies [52–59]. Approximately half the studies were comparative (n = 47) and half were single arm (n = 41). The source of funding was not reported in approximately half the studies (n = 44, of which 23 were comparative and 21 were single arm).

In total, 37 studies reported on both abiraterone and enzalutamide [20,21,23,24,26–29,35,38,40,42,43,45,49–51,53,54,58,60–76]; 40 studies reported on abiraterone alone [30–32,34,36,37,39,41,44,46–48,52,55,56,59,77–100] and 11 studies reported on enzalutamide alone [22,25,33,57,101–107].

Study sample sizes ranged from ten [93] to 15,340 patients [73]. The frequency of outcome reporting by study is provided in Figure 2. Relatively few studies reported on cost or healthcare resource use (n = 9) and HRQoL (n = 9).

Substantial heterogeneity was apparent on several levels, including outcome definitions, sample sizes, follow-up periods and the disease and demographic characteristics of patients in included studies. Median age was reported by 74 studies and varied from 65 to 82 years. The presence of visceral metastases varied from 0 to 100% (n = 39), bone metastases from 29 to 100% (n = 36) and lymph node metastases from 10 to 62.3% (n = 30). A Gleason score of ≥8 was reported in 27.6 to 90.6% of patients (n = 47). The presence of pain at baseline ranged from 0 to 38.6% (n=14), with opioid use ranging from 19.1 to 54.5% (n = 5). Overall, the reporting of baseline patient characteristics varied by study (Supplementary Table 8), thereby limiting the comparison of baseline data across studies.

The ROBINS-I quality assessments rated the 85 clinical outcome studies at serious (n = 39), moderate (n = 34) or unclear (n = 12) risk of bias. No studies were rated at low or critical risk of bias (Supplementary Figure 1). Further details on the risk of bias appraisals are provided in the Supplementary Materials.

Effectiveness of abiraterone or enzalutamide in RWE studies

The median PFS and OS for abiraterone and enzalutamide in real-world clinical practice are reported in Figures 3A & 4A. Adjusted comparative data are reported in Figures 3B & 4B. Treatment duration is reported in Figure 5A, and use of subsequent therapies is reported in Figure 5B. Data for other clinical outcomes were limited and are reported in the Supplementary Materials.

PFS & time on treatment

A total of 40 studies (64 datasets) reported median progression values. Of these, 25 studies (31 datasets) reported on clinical PFS, 20 studies (24 datasets) on prostate-specific antigen PFS (PSA PFS) and nine studies (nine datasets) on radiographic PFS (rPFS; definitions given in Supplementary Tables 5–7). Across all definitions, median PFS for abiraterone ranged from 3.7 to 20.9 months, and from 5.5 to 18.5 months for enzalutamide (Figure 3A). Similarly, median PFS was <12 months in 42 of 64 (66%) datasets, and between 12 and 18 months in 19 of 64 (30%) datasets. Only five datasets (8%) [22,34,57,60,84] (sample sizes: n = 145; n = 103; n = 77; n = 1171; n = 931) reported median PFS that was in line with the 16 to 20 months of median PFS seen in COU-AA-302 and PREVAIL RCTs for NHAs [7,8]. These studies included patients with generally similar baseline characteristics to those enrolled in randomized trials, including age, Gleason, PSA, ALP, LDH and time from initial diagnosis (data not shown). One single arm study from the USA reported that median PFS was not reached for enzalutamide treatment with a follow-up of 9.3 months [25].

Twelve studies reported comparative data for PFS between NHAs (Supplementary Tables 10–12), of which four studies reported adjusted comparisons, summarized in Figure 3B. The largest (n = 751) and only prospective study reported no evidence of a difference between NHAs (HR 1.04, 95% CI 0.851 to 1.271; p = 0.7) [53]. Three retrospective studies (with four datasets) reported results that favoured enzalutamide [49,62,67]; two of these were statistically significant differences [49,62]. These included one study (with two datasets) that reported a composite end point of PSA, clinical or radiological progression and the time to PSA progression in men aged 80 years or older identified from a pharmacy database in British Columbia, Canada; one study that reported PSA progression in men recruited from sites in Japan; and a third study from Brazil that was reported in abstract form only. Three studies were rated at moderate or severe risk of bias and one was rated as unclear.

Real world PFS outcomes were supplemented by median treatment duration data reported by 24 studies (33 datasets). For abiraterone, the median duration of treatment ranged from 5.3 to 14 months (18 studies) and from 5 to 14.1 months for enzalutamide (12 studies; Figure 5A & Supplementary Table 13).

Overall survival

A total of 38 studies (46 datasets) reported median OS values, ranging from 9.8 to 39.2 months for abiraterone, and from 14 to 45 months for enzalutamide (Figure 4A). The majority of datasets (26 of 46; 57%) reported survival of less than 24 months. Only three datasets (two studies) [43,85] reported survival consistent with the 34 to 36 months of median OS seen in RCTs [7,8]. Of these two studies, one was in abstract form only and one was conducted in an expert university hospital in Paris, France. Four studies for abiraterone from China, Italy or Japan [41,67,76,97] and four studies for enzalutamide from Brazil, Korea or Japan [62,67,104,105] reported that median OS was not reached. Where reported, median follow-up for the studies where OS was not reached ranged from 8.6 to 21.2 months.

Fifteen studies reported comparative survival data between NHAs (Supplementary Table 19), of which seven studies reported adjusted comparisons, summarized in Figure 4B. A single prospective study (n = 751) [53] reported no evidence of a survival difference between NHAs (HR 1.000, 95% CI 0.788 to 1.270; p = 0.9986). Retrospective studies showed improvements in survival that favoured one NHA over another, with two in favour of abiraterone [61,76] and four in favour of enzalutamide [29,60,62,72]. Two studies reported statistically significant improvements in favour of enzalutamide, including the largest study identified in the review (Scailteux 2021 [72], n = 10,308). In this retrospective study, the benefit for enzalutamide was found to reduce when restricting to recent NHA use from 2015 to 2017. The risk of bias ranged from moderate to severe in five studies, and two studies were rated at unclear risk of bias.

Subsequent treatments

In total, 15 studies (23 datasets) reported on second-line (2L) treatment use after receiving abiraterone or enzalutamide at 1L [29,41,43,49,52,53,60,72,74,83,89,94,95,100,104], summarized in Figure 5B. Overall, the proportion of patients who received at least one 2L treatment for mCRPC ranged from 8.8 to 63.3%, with most datasets (19 of 23) reporting that subsequent treatments were used in less than 50% of patients.

Safety & tolerability

Evidence for the real-world safety and tolerability of abiraterone and enzalutamide is summarized in Supplementary Tables 19 & 20. An overview of the trends in safety data is presented in Figure 6.

A total of 37 (of 88) studies (42%) reported on the safety of NHAs in 1L mCRPC. Both comparative and single arm study data were identified, and risk of bias ratings ranged from moderate to serious or unclear.

The trends for any adverse event (AE) and any serious AE (SAE) are provided in Figure 6A & B, respectively. While the rates of AE varied between studies, no marked differences were observed between treatments: rates of AE ranged up to 72% in abiraterone and up to 77% in enzalutamide; and rates of SAE ranged up to 23% in abiraterone and up to 26% in enzalutamide. However, treatment with enzalutamide was associated with numerically higher rates of treatment discontinuation (up to 40.8%), AEs >grade 3 (up to 17%) and AE-related deaths (up to 7.0%) (Supplementary Table 19). With respect to individual AEs, comparative data suggested that rates of hypertension (15 studies) and seizures (four studies) were similar for abiraterone and enzalutamide. The reporting of falls and fractures (one study) or ischemic heart disease (two studies) was limited but suggested advantages for abiraterone (Supplementary Table 20).

Health-related quality of life

Nine studies reported on HRQoL [28,34,39,44,54,57,58,69,92]. Of these, four studies reported comparative data [28,54,58,69] and risk of bias ratings ranged from serious (n = 3) to unclear (n = 1).

Generally, HRQoL was improved in patients receiving abiraterone compared with enzalutamide across all instruments identified; however, evidence was limited (one to two studies per HRQoL outcome of interest; Supplementary Figure 3 & Supplementary Table 9).

Economic outcomes

Ten studies reported on economic outcomes, including cost and healthcare resource use [21,26,27,40,50,51,65,70,83,95]. Of these, three reported from a USA setting [21,26,27], two from Canada [50,51], and one each from Germany [65], Italy [40], Colombia [70], Spain [83] and France [95]. Five studies reported adjusted comparative data [27,50,51,65,95], four studies reported unadjusted comparative data [21,26,40,70] and one study reported single arm (non-comparative) data [83]. NHA-specific economic outcomes were reported for the period of 2012 to 2016, with one study reporting non-NHA outcomes from 2003.

Eight studies reported on direct medical costs [26,27,40,65,70,83,95,108] (Supplementary Table 21). Of these, six reported comparisons between NHAs [26,27,40,65,70,108]. Two studies reported lower direct medical costs in patients treated with abiraterone compared with enzalutamide, with one study reporting a statistically significant difference (n = 5973) and one study not reporting formal statistical testing (n = 144) [21,70]. One study (n = 837) reported higher direct medical costs for patients treated with abiraterone compared with enzalutamide [65] and three studies reported mixed results depending on the outcome assessed (n = 3351 [26], n = 9700 [40] or n = 3230 [27]). None of the identified studies reported costs explicitly using generic abiraterone, which became available in the USA from 2018 [109] and Europe in 2021 [110–112].

Healthcare resource use was reported in seven studies [21,27,50,51,65,70,83], of which five reported comparative data [21,27,51,65,70] (Supplementary Table 22). Two retrospective studies reported that HRU was significantly (n = 3525) or numerically (n = 3405) higher for abiraterone compared with enzalutamide [21,51] and three retrospective studies reported mixed results for HRU depending on the specific outcome assessed (n = 944 [65], n = 144 [70] or n = 3230 [27]). Two studies reported a numerically longer duration of hospital stay for abiraterone versus enzalutamide (n = 3230 [27] or n = 144 [70]).

Indirect costs, in the form of sick leave payments, were reported in one study from Germany (n = 837) [65] (Supplementary Table 23). Comparisons across treatments suggested numerically higher costs for abiraterone versus enzalutamide. No formal statistical testing was reported.

Two studies reported on cost drivers [40,65] (Supplementary Table 24). One study reported that the costs of medication followed by inpatient care and sick leave payments were the most significant cost drivers in mCRPC (based on n = 837 patients) [65]. A second study reported that the cost of treatment (drugs and administration) was the main cost driver, which represented more than 77% of costs (based on n = 9700 patients) [40].

Discussion

RWE has an important role in providing insight into the use of novel therapeutic products in clinical practice. This systematic review identified a relatively large number of studies reporting RWE for NHAs in 1L mCRPC, covering a broad range of geographies and outcome types, including clinical, economic and quality of life. With the breadth of studies identified, there was heterogeneity in populations and outcomes, highlighting the complexity and variability of outcomes for prostate cancer patients. In particular, a high variability was observed in the baseline characteristics of evaluated patient populations. For example, rates of visceral metastases ranged from 0 to 100% and rates of bone metastases ranged from 28 to 100%, clearly depicting diverse patient populations.

The extent of heterogeneity present in the RWE literature for mCPRC is both a limitation and an important finding of this study. As a limitation, the heterogeneity greatly limited our ability to compare results across studies and to identify strong trends or conclusions. As an important finding, it highlights the considerable challenge of comprehensively synthesising RWE for use in reimbursement and clinical decision making in mCRPC. Ultimately, the appropriate use of RWE to inform decisions requires judgement on the representativeness of data to the locale of interest, which itself can vary significantly. To this end, our review provides an important resource for clinical and reimbursement decision makers seeking to identify relevant local RWE in mCRPC while providing further insight on the range of outcomes possible across diverse prostate cancer populations.

Despite the challenges of synthesising RWE, a number of general trends and evidence gaps were identified. There was evidence that, despite the use of NHAs, median PFS and OS in real-world clinical practice remain limited and consistently fall below the median durations reported in RCTs. These findings may arise from differences in patient characteristics between trial and RW practice, as well as differences in post-progression treatment. In general, studies that reported median durations in line with RCTs were conducted in specialist settings and/or included patients with similar baseline characteristics to the RCTs. Where reported, the use of subsequent treatments after progression on NHAs was low, with typically fewer than one in two mCRPC patients receiving treatment after 1L. Overall, these data illustrate the persistently poor RW outcomes of mCRPC patients and highlight the remaining unmet need for improved 1L treatment in this setting.

The review identified a number of studies reporting on the comparative effectiveness of NHAs. The majority of included studies reported estimates of treatment effect without adjustment for differences in patient characteristics across groups, potentially leading to biased effect estimates. In accordance with best practice, we prioritised data to those studies reporting adjusted analyses or those that were of prospective study design [113]. In studies that had adjusted for baseline population differences, there was a lack of consensus on the similarity of effect for NHAs in clinical practice, which may be explained by several factors. Firstly, differences in the effectiveness of NHAs may be due to between-study differences in patient characteristics, such that studies showing benefit may include patients who are selectively more responsive to one NHA over another. Secondly, the observed benefits may not be treatment-related but due to selection bias unaccounted for in the adjustment analysis. Most studies were retrospective by design meaning that study authors had to rely on available data from registries, which may not capture all confounding factors between groups. Only one comparative study reporting on both OS and PFS was prospectively designed and reported no difference in outcome between NHAs. These results are consistent with other comparative studies, such as network meta-analyses of RCTs, which have reported no significant difference in OS between NHAs [114,115].

Alongside the effectiveness data, the review identified studies reporting RW safety and tolerability data for NHAs. As with the effectiveness outcomes, few studies reported adjusted comparative data for safety. Based on comparative studies, safety evidence suggested that the rate of AEs leading to study drug discontinuation, ischemic heart disease and falls and fractures may be numerically lower for abiraterone versus enzalutamide; and that the rate of hypertension, any serious AEs, any grade ≥3 or 3/4 AEs, AEs leading to death and AEs leading to hospitalisation may be similar for abiraterone and enzalutamide. However, the data for safety and tolerability were based on limited numbers of studies and patients.

The review identified 19 studies reporting on the economic or HRQoL impact of NHAs. Economic studies generally showed mixed results with few clear trends in cost or HRU between NHAs. Importantly, none of these studies evaluated the impact of generic abiraterone (available since 2018 in the USA [109]; 2021 in Europe [110–112]) on economic outcomes, highlighting an area of further research. The available comparative HRQoL RWE (four studies) suggests potentially improved HRQoL (BPI-SF, EQ-5D, EORTC-QLQ-C30 and FACT-PC scales) for abiraterone versus enzalutamide. None of these studies reported results with adjustment for population differences across NHAs, which may have biased results. Future HRQoL studies should include adjustment for baseline characteristics when comparing results across treatments. Unlike other end points, the majority of HRQoL studies (n = 8) had prospective data collection. The limited availability of HRQoL data in existing RWE registries may have contributed to the low number of retrospective studies identified by the review. Improved availability of HRQoL data in registries may increase the scope and breadth of future HRQoL study in mCRPC.

We identified eight other systematic reviews that reported on abiraterone or enzalutamide for the treatment of mCRPC. Two SLRs were focussed exclusively on sequential treatment outcomes (abiraterone followed by enzalutamide or vice versa) [116,117], and all SLRs included fewer studies than this SLR (four to 22 included studies). Therefore, it appears that this SLR represents the most comprehensive real-world assessment of the evidence around 1L abiraterone and enzalutamide treatment outcomes in mCRPC patients to date.

Limitations of evidence

There were relatively few comparative studies that reported incremental outcomes with adjustment for differences in populations across treatment groups, and very few studies that were prospectively designed. The studies identified were judged to be at either moderate, serious or unclear risk of bias, with confounding identified as the most common source of bias (Supplementary Figure 1). Limited reporting of the funding source was also identified for many studies, a common issue in RWE [118].

Several limitations were noted at the study level. First, the treatment line at which abiraterone or enzalutamide were administered was not always specified, leading to 17% of records being excluded during full paper screening. In some studies, 1L treatment was not explicitly stated, but could be reasonably assumed from author descriptions (Supplementary Table 3). This approach was in line with the shift in nomenclature and classification of prostate cancer from a chemotherapy-naive versus post-docetaxel classification to a line of therapy classification based on PCWG3 recommendations [119,120]. For key outcomes (OS, PFS), there was no meaningful difference in median survival comparing studies where 1L was stated versus assumed (Supplementary Table 25). Nonetheless, including these studies may have introduced heterogeneity in terms of treatment line. Some studies allowed ADT prior to 1L abiraterone or enzalutamide therapy, whereas other studies considered ADT to represent a 1L therapy itself [86]. Second, the data were immature in nine studies (i.e., median ‘time to event’ outcome was not reached for one or more treatment arms) [25,41,52,62,67,76,97,104,105], which limited treatment comparisons. Third, in some studies, the prescription of abiraterone or enzalutamide was used as a proxy for capturing patients at the mCRPC stage [22]. Finally, there was considerable heterogeneity in the definitions of PFS across studies, which included clinical, radiological and PSA, alongside composite end points that combined two or more definitions. To support consistency across studies, there remains a need to establish best practice in the collection and estimation of RW PFS in mCRPC.

Strengths & limitations of review

Key challenges identified by this systematic review included: limited information on some effectiveness and economic outcomes of interest; and the expected heterogeneity of populations and study design present between RWE studies. Both these limitations prevented robust conclusions being made on most outcome data. No evidence was found for the clinical outcomes, PFS2 or time to opiate use; the safety outcome, posterior reversible encephalopathy; or the economic outcome, direct non-medical costs.

The review captured papers up until November 2021 and therefore does not capture RWE of newly approved treatments in mCRPC, including poly (ADP-ribose) polymerase (PARP) inhibitors plus NHA. For countries such as the USA, Canada and Italy, a relatively large number of studies reporting RWE for NHAs were identified. There was also comprehensive evidence identified from France, Spain and China. Several multinational studies were also identified. However, there was limited evidence identified from Australia, Asia, Central America, India and South America. No evidence was found from Africa or the Middle East. Where RWE are currently lacking, the transferability of RWE from other countries requires careful consideration of multiple factors [121].

Inherent to all RWE studies, the data retrieved by this review provides a historical overview of treatments and outcomes in 1L mCRPC. As treatment practice changes, with the transition to greater use of NHAs prior to pre-mCRPC and the introduction of new regimens for mCRPC, a change in RW outcomes can be expected. Further studies are required to continue to monitor RW outcomes and ensure unmet needs are identified and addressed.

Research implications

The evidence retrieved by this review may support the future health technology assessment of new medicines for mCRPC, as well as guide the prioritisation and design of future RWE studies. RWE reveals the challenges of adherence and administration of new drugs and can fill the research gaps from trial data (e.g., actual dose, persistence) [113]. Although the evidence in this review has limitations, these are inherent in any review of RWE and add to the growing urgency to improve the reporting of RWE and follow newly established guidance, including making clear the funding source for research [122,123]. The review highlights the research need for more adjusted comparative analyses following best practice on RWE reporting and more prospective study designs.

Conclusion

To our knowledge, this study provides the most comprehensive summary to date of published RWE for 1L NHA treatment in mCRPC, contributing to our understanding of the current state of RWE and unmet medical need in this setting. Significant heterogeneity in the design of studies, the characteristics of included patients and associated treatment outcomes greatly limited our ability to draw definite conclusions on RWE trends. A qualitative summary of reported outcomes suggests variable but limited median progression-free survival (3.7 to 20.9 months) and OS (9.8 to 45 months) from NHA treatment, with most patients receiving only one line of therapy for mCRPC. Median survival in RW practice tended to be shorter than observed in RCTs, which likely reflects differences in patient characteristics and post-progression treatment across settings. Most safety outcomes were either similar for abiraterone and enzalutamide or favoured abiraterone, while limited HRQoL data tended to favour abiraterone. These results highlight the importance of developing novel therapeutic options to improve outcomes in the first-line setting. High-quality RWE studies are urgently needed to provide greater certainty and further inform the real-world effectiveness of current 1L treatment options in mCRPC.