The Health Economics of Metastatic Hormone-Sensitive and Non-Metastatic Castration-Resistant Prostate Cancer—A Systematic Literature Review with Application to the Canadian Context

Background: Health economic evaluations are needed to assess the impact on the healthcare system of emerging treatment patterns for advanced prostate cancer. The objective of this study is to review the scientific literature identifying cost-effectiveness and cost analyses that are assessing treatments for metastatic hormone-sensitive prostate cancer (mHSPC) and nonmetastatic castration-resistant prostate cancer (nmCRPC). Methods: On 29 June 2021, we searched the scientific (MEDLINE, Embase, and EBSCO) and grey literature for health economic studies targeting mHSPC and nmCRPC. We used the CHEC-extended checklist and the Welte checklist for risk-of-bias assessment and transferability analysis, respectively. Results: We retained 20 cost-effectiveness and 4 cost analyses in the mHSPC setting, and 14 cost-effectiveness and 6 cost analyses in the nmCRPC setting. Docetaxel in combination with androgen deprivation therapy (ADT) was the most cost-effective treatment in the mHSPC setting. Apalutamide, darolutamide, and enzalutamide presented similar results vs. ADT alone and were identified as cost-effective treatments for nmCRPC. An increase in costs as patients transitioned from nmCRPC to mCRPC was noted. Conclusions: We concluded that there is an important unmet need for health economic evaluations in the mHSPC and nmCRPC setting incorporating real-world data to support healthcare decision making.


Introduction
Advanced prostate cancer (PCa) is associated with poor quality of life and high mortality [1]. The health states preceding the terminal stage of PCa are metastatic hormone-sensitive prostate cancer (mHSPC) and nonmetastatic castration-resistant prostate cancer (nmCRPC). Metastatic hormone-sensitive prostate cancer is characterized by de novo metastasis while the patient is still responsive to medical or surgical castration [1]. In 2018, approximately 1200 men were diagnosed in Canada with mHSPC [2]. Nonmetastatic castration-resistant prostate cancer is characterized by rising prostate-specific antigen (PSA) levels despite castrate levels of testosterone, without metastasis being detected by conventional imaging [3]. Virtually all mHSPC and nmCRPC patients will eventually progress, develop metastasis, and present significant morbidities and paraneoplastic effects [3,4].
Targeting these health states and aiming to delay progression, the 2019 Canadian Urological Association guidelines recommend [4] the use of androgen-deprivation therapy (ADT) for newly diagnosed mHSPC patients. Docetaxel in combination with ADT (DOCE) is recommended for patients with high-volume disease and good performance status. Enzalutamide + ADT (ENZA) and apalutamide + ADT (APA) are also recommended as systemic therapy alternatives for mHSPC treatment. Abiraterone acetate + prednisone + ADT (ABI) can be considered as an option for low-volume mHSPC, but is recommended for patients presenting at least two of the following criteria: Gleason score of ≥8, presence of three or more lesions on bone scan, or presence of measurable visceral metastasis [4]. The 2020 Canadian Urological Association guidelines recommend the use of either APA, ENZA, or darolutamide + ADT (DARO) for high-risk nmCRPC patients, defined by a PSA-doubling time shorter than 10 months [3]. While these treatment options are successfully proven to delay progression and improve survival, they increase the financial burden on the healthcare system. The costs associated with novel PCa treatments are being added to the already growing burden of the disease as the incidence of PCa is increasing due to the aging population [5]. There is a need for health economic evaluations to appropriately assess the impact of these novel therapies in order to better understand the evolution of the burden associated with PCa and optimize resource allocation to improve disease management. Therefore, this systematic review is necessary to synthesize the current state of the health economic literature regarding advanced PCa.
The objective of this project is to systematically review the scientific literature identifying economic evaluation studies that are assessing the latest treatments for mHSPC and nmCRPC. Consequently, this study aims to identify potential knowledge gaps in health economic evidence for the integration of novel treatments for advanced PCa.

Eligibility Criteria
We built our inclusion criteria around the population of male patients that have been clinically diagnosed with mHSPC or nmCRPC. We considered all interventions that were recommended for mHSPC and nmCRPC in the Canadian Urological Association guidelines. For outcomes, we targeted costs, the burden of disease, or cost-effectiveness results that referred to Health Canada-approved treatments for mHSPC and nmCRPC, regardless of the country of origin of these studies. As treatment guidelines may differ in different jurisdictions, we did not stratify our analyses further than by health state (mHSPC or nmCRPC). Within our eligibility criteria, we considered studies using data from clinical trials as well as studies using real-world data to capture the full extent of the literature.
The inclusion criteria that were used for study selection were cost-effectiveness analysis, cost-of-illness analysis, health technology assessment (HTA), economic evaluation, and disease-burden analysis (analysis that estimates the financial impact of PCa). We excluded studies referencing only mCRPC without analyzing mHSPC or nmCRPC, other reviews, meta-analyses, and studies that did not present costs. Additionally, we excluded budget impact analyses (BIAs) because they are highly payer-specific, and they consider costs of given products, projected market shares, incidence, prevalence, and indication restrictions [6,7]. Budget impact analyses report on the affordability of a particular health technology for a specific payer based on their purchasing power, and therefore they lack transferability between payers and healthcare systems. Furthermore, BIAs contain confidential elements that are often not publicly disclosed [8].

Literature Search
We searched MEDLINE, Embase, EBSCO and the grey literature (National Institute for Health and Care Excellence (NICE) Evidence database) on 29 June 2021. As data collection was initiated prior to study registration, this systematic review was not eligible for registration in PROSPERO and does not have a registration number. Based on our search strategy and database verification, there is no similar registered study in PROSPERO prior to the submission date of this manuscript. Our search strategy was centered around three concepts and was reviewed by an experienced librarian. The first concept was designed to capture economic evaluations, models, and cost analyses and is based on the Canadian Agency for Drug and Technologies in Health (CADTH) search filter developed for literature reviews [9]. The second concept aimed to capture the advanced stages of PCa and was constructed by combining the medical subject heading (MeSH) terms and keywords such as "Prostatic Neoplasms", "Neoplasm Metastasis" and "Castration-Resistant" referring to mHSPC and nmCRPC. Since mCRPC is the terminal stage of advanced PCa, we included it in the search criteria to ensure the capture of studies referencing the pre-mCRPC period. This wider search strategy allows for a thorough review of the literature and captures studies reporting on mHSPC or nmCRPC that might have been wrongfully tagged as mCRPC. The third concept represented the combination of search terms for medications and therapies that are currently approved in Canada for the treatment of advanced PCa. The full search strategy and results for MEDLINE are available in Appendix A, Table A1 and were adapted for the other databases of interest. We considered all original research publications and abstracts published in English from 2010 to the present day, to capture all relevant publications.

Study Selection
Search results were uploaded into Covidence [10], a web-based licensed software designed to facilitate and improve literature reviews. Duplicates were detected and removed automatically by Covidence [10]. Two reviewers (IY, JJG) independently conducted a title and abstract screening to retain pertinent articles that satisfied the inclusion criteria. Conflicts were resolved by consulting with a third independent reviewer (AD). Full-text review was then performed independently by two reviewers (IY, JJG). We rejected irrelevant studies and documented the reason for rejection. Conflicts at that stage were resolved by discussion among the two reviewers. The third reviewer (AD) was consulted when an agreement was not reached.

Data-Collection Process
Data items were collected by an extraction form (available in Appendix A, Table A2) that we adapted from Wijnen et al. [11] to fit our specific study objective as recommended. When multiple references reported data from the same study, only the final or most mature report was considered. Data extraction was validated by a second reviewer (JJG).

Data Items
When available, we extracted the following information: the reference health state, the type of analysis, the study base type (model vs. trial-based), the intervention, the comparator or the current standard of care, the perspective, the methods of cost measurement, the costs, the methods of effect measurement, the effects in life years gained (LYGs) or quality-adjusted life-years (QALYs), the incremental cost-effectiveness ratio (ICER), and the sensitivity analysis. Additionally, we sought data regarding the year of valuation, the time horizon, the discounting rate, the authors, the preferred strategy, the type of publication, the setting, and the sponsor.

Assessments from HTA Agencies
By reviewing the grey literature, we captured assessments of interest that contained cost-effectiveness analyses from the United Kingdom's NICE and the Scottish Medicines Consortium (SMC). To complement this information and reflect the Canadian governmental assessment of therapies for advanced PCa, health economic analyses of the target medications were extracted from the CADTH and Institut National de l'Excellence en Santé et en Services Sociaux (INESSS) databases.

Risk-of-Bias Assessment
We performed a risk-of-bias assessment on the selected studies using the Consensus on Health Economic Criteria (CHEC) extended checklist [12,13] for critical appraisal of the quality of the economic evaluations (available in Appendix A, Table A3) as recommended by the Cochrane collaboration [11]. Through this questionnaire, we evaluated potential sources of bias, structural assumptions for modeling, outcome valuation, and if study conclusions were supported by their results. The CHEC extended checklist was used because of its high scrutiny and its ability to assess model-based economic evaluations [11]. We classified the studies as "Excellent", "Good", "Fair" and "Poor" based on their score in the risk-of-bias assessment questionnaire. This grading system, which has not been validated, considered that all the items of the questionnaire carried the same weight. The questionnaire items were judged dichotomously: 1 point was awarded if a study satisfied an item from the questionnaire; no point was awarded if item fulfillment was unclear, unspecified, or insufficient. Therefore, we quantified the quality of the studies by their total score (maximum score of 20) to be able to identify the higher-quality studies. Studies that scored 17 or higher were considered of excellent quality, 15-16 of good quality, 13-14 of fair quality, and 12 or lower of poor quality.

Transferability Analysis
Furthermore, we evaluated the transferability of the economic evaluations, which is the ability to hold true for different populations or settings [14] by using the Welte checklist [15]. The Welte checklist was used due to its ability to assess trial and model-based economic evaluations as well as the fact that it uses clear cut-off points to determine if a study is transferable [11]. The Welte checklist is a decision chart for assessing and improving the transferability of economic evaluation results between countries [15]. This decision chart includes knockout criteria, a checklist of transferability factors, and a component that evaluates the uncertainty of transferred results. The knockout criteria are defined by three characteristics that a study needs to satisfy for its results to be transferable to the study country, and they are used as cut-off points to determine transferability. Studies were grouped by the country-specific setting of the conducted analysis and transferability to the Canadian setting was assessed. The International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Country-Specific Pharmacoeconomic Guidelines were used as a reference for evaluating the methodological characteristics [16]. Healthcare system characteristics were evaluated through the provided information within the retained references. Population characteristics were evaluated through an online search [17][18][19].

Effect Measures
As we extracted crude effectiveness measures in either LYGs or QALYs, we did not use any synthesis methods to report these outcomes. Additionally, we extracted costs and ICERs. In cost analyses, cost components were reported as they were reported by the original authors. When probabilistic sensitivity analyses were available, they were reported as the probability that an ICER is inferior to the prespecified willingness-to pay-threshold.

Synthesis Methods
No statistical analyses were performed in the reporting of costs or outcomes. All costs were converted to 2021 Canadian dollars and adjusted for inflation by using historical currency exchange rates [20] and the Canadian historical consumer price index, respectively [21]. On the rare occasions that the year of cost valuation was not reported, the year of publication was considered the year of valuation. When discounting rates were not reported, we assumed that the analysis was conducted using recommended local discounting rates. No extrapolation was performed for missing data; therefore, only data retrieved from publications were reported.

Summary
Through our literature search, we captured 1330 records from our database search and 305 grey-literature records, which resulted in 1505 nonduplicate citations of original research articles, abstracts, or reports that were screened for relevance ( Figure 1 based on PRISMA reporting guidelines [22]). Among those, 213 (13%) database records and 129 (7.9%) grey-literature records were retained for full-text screening. The final analysis included 23 (1.4%) database records and 19 (1.2%) grey-literature records. Of these, 24 studies referred to mHSPC and 20 to nmCRPC. currency exchange rates [20] and the Canadian historical consumer price index, respectively [21]. On the rare occasions that the year of cost valuation was not reported, the year of publication was considered the year of valuation. When discounting rates were not reported, we assumed that the analysis was conducted using recommended local discounting rates. No extrapolation was performed for missing data; therefore, only data retrieved from publications were reported.

Assessments from HTA Agencies
Through the NICE and SMC databases, five HTA reports for mHSPC and five for nm-CRPC were captured. The Canadian HTA entities (CADTH and INESSS) have published five reports for nmCRPC and four reports for mHSPC, which all contained cost-effectiveness analyses, except for one report regarding darolutamide that included a cost-minimization analysis.

Economic Evaluations
Willingness-to-pay thresholds referred to in this paragraph are those considered by the original authors and reflect local standards. In the mHSPC setting, 11 studies evaluated DOCE and 10 of them analyzed ADT alone as an alternate option ( Table 2). On the other hand, Pelloux-Prayer et al. (2020) [34] assessed treatment sequencing. They identified the sequence of DOCE, followed by ABI, as being the cost-effective option for asymptomatic and mildly symptomatic patients when compared to DOCE followed by ENZA (ICER of 708,983 CAD/QALY). In symptomatic patients, repeating DOCE compared to cabazitaxel (CABA) after the failure of DOCE was the preferred option, as the CABA sequence was associated with an excessive ICER of 1,869,295 CAD/QALY. Docetaxel was analyzed versus ABI in five studies [33][34][35][36]41], and against ENZA in two studies [34,66]. There seems to be a consensus that DOCE is the cost-effective treatment for mHSPC compared to ADT alone, with ICERs ranging from 9045 CAD/QALY to 70,459 CAD/QALY. The two studies that did not consider DOCE as cost-effective are a Chinese [40] and a Brazilian [41] study that reports ICERs exceeding the local willingness-to-pay thresholds (20,301 USD/QALY and 33,000 USD/QALY, respectively). A study by Zheng et al. (2021) [39] evaluated the cost-effectiveness of ENZA compared to ADT and rejected ENZA with ICERs of 538,940 CAD/QALY in the US perspective and 281,948 CAD/QALY in the Chinese perspective, as they exceeded local willingness-to-pay thresholds.    In the nmCRPC setting, two cost-effectiveness analyses evaluated APA in comparison to ENZA [53,54] (Table 2) Regarding the HTAs conducted by governmental authorities in the mHSPC setting, ENZA, APA, DOCE, and ABI were assessed. These evaluations fall in line with the published literature, identifying DOCE as the cost-effective treatment for mHSPC when compared to the alternatives. Reported ICERs are within the acceptable range when comparing APA, ABI, and ENZA to ADT. However, comparing these novel therapies against DOCE yields high ICERs (200,000 CAD/QALY and more). These high ICERs occasionally lead to favorable recommendations for reimbursement based on the provided clinical benefit and improved quality of life. These favorable recommendations are often made conditionally to the attenuation of the financial burden through price reductions or patient access schemes.
In the nmCRPC setting, CADTH and INESSS both identify APA, DARO, and ENZA as more effective treatments compared to ADT, and are associated with ICERs ranging from CAD184,879 to 1,237,896 per QALY [24,26,46,47]. However, these treatments received positive recommendations based on their abilities to improve quality of life and delay metastases with the condition that the financial burden is reduced. The evaluations conducted for DARO vs. ADT by the SMC and ENZA vs. ADT by NICE were associated with ICERs of 31,927-47,890 CAD/QALY [50] and 24,996 CAD/QALY [49], respectively, led to favorable recommendations. These lower ICERs in comparison to the Canadian assessments are in part due to patient access schemes. In their HTA of ENZA for nmCRPC, NICE [49] concluded that ENZA in combination with ADT is not cost-effective vs. ADT alone at the provided list price. They recommended APA and DARO for reimbursement in the nmCRPC setting [48,52]. However, these treatments were associated with excessive ICERs, and NICE's recommendations were made conditional to financial rebates provided by the manufacturers.

Cost-Analysis Studies
Among the studies that conducted a cost analysis in the mHSPC setting, Hu et al. [56] identified that using DOCE instead of ABI would represent a cost-saving alternative in China (Table 3). Wong et al. [58] reported the cost of treating mHSPC with ABI to vary from CAD 540,299 to CAD 797,544 for a period of 42 to 44 months. Treating mHSPC patients with ENZA resulted in costs of CAD 225,387 to CAD 602,822 for a period of 12-36 months. This analysis identified the main cost factor as the duration of the mHSPC state.   (Table 3). Freedland et al. [60] observed that the yearly cost per patient increased from CAD 5121 to CAD 16,014 after the onset of nmCRPC in the US. Shah et al. [63] assessed the increase in cost due to adverse events in nmCRPC that reached CAD 63,619 compared to CAD 47,212 per patient without adverse events. Central nervous system adverse events were an important cost driver. Four studies analyzed the cost increase as patients transitioned from nmCRPC to mCRPC [60][61][62]64]. George et al. [61] reported an increase in PCa-related costs from CAD 556 to CAD 3675 and all-cause medical costs that increased from CAD 1883 to CAD 5460 for nmCRPC and mCRPC, respectively. Wu et al. [64] reported an increase in the medical and pharmacy costs within the Medigap and commercially insured patients. Medicare Advantage and Medigap are both supplementary private insurance plans that beneficiaries can opt for. They differ in the fact that Medigap policies are neither provided nor endorsed by the United States Government, while Medicare Advantage plans are provided by government-approved private companies [67,68].

Results from Real-World Data Studies
This review captured seven studies using real-world data to conduct health economic evaluations. There were two publications assessing the mHSPC setting and 5 assessing the nmCRPC setting. Additionally, it is important to mention that all these studies were cost analyses. Furthermore, none of the studies using real-world data conducted a direct comparison between treatments. Instead, these studies focused on reporting the financial impact caused by various elements. Shah et al. [63] reported the increase in costs due to adverse events while others evaluated cost differences due to the transition from nm-CRPC to mCRPC [60][61][62]64]. All the real-world studies, with the exception of the study by Svensson et al., were conducted in the United States and used the Veterans' Health Administration (VHA) database or private insurance databases. The study by Svensson et al., on the other hand, was conducted in Sweden [59].

Risk-of-Bias Assessment
Results from the risk-of-bias assessment are reported in Table 4. We classified 12 studies as excellent, 6 as good, 2 as fair, and 3 as poor. Issues relating to generalizability, ethics, and distribution were the predominant sources of bias.

Transferability Assessment
Studies were grouped by country of origin of the conducted analysis. Transferability of economic studies from Brazil, China, Columbia, France, Greece, Mexico, Sweden, the United Kingdom, and the United States to the Canadian setting was evaluated (Appendix A, Table A4). The correspondence between study country and Canada is summarized in Table 5. General knockout criteria were respected throughout all of the countries of interest [15]. When focussing on the methodological characteristics of the analyses, all the reference countries present unbiased or slight underestimates [16]. This is due to the use of higher discount rates, as current discounting in Canada is fixed at 1.5% by the CADTH guidelines [69]. Additionally, the studies do not consider the cost of productivity loss from a societal perspective. When analyzing the healthcare-system characteristics, technology availability was consistent across all the studied countries. Price variability and absolute and relative prices of healthcare, however, seem to be important sources of bias affecting the transferability to the Canadian setting. Population characteristics of Greece, France, Sweden, and the United Kingdom presented high correspondence to the Canadian ones [15]. However, Brazil, China, Columbia, Mexico, and the United States presented a few differences that might bias the transferability of the studies and yield lower ICERs. Transferring results from studies conducted in these countries is therefore subject to a potential bias that may lead to an underestimation. Therefore, these results should be transferred with caution considering this greater uncertainty. Disease incidence and prevalence, life expectancy, work-loss time, health-status preference, and productivity are potentially sources of transferability bias that could over-or underestimate results leading to erroneous conclusions [15,[17][18][19].

Summary of Results
The emergence of novel treatments for advanced PCa led to an improvement in survival and quality of life. Given the high costs of these medications, health economic evaluations are needed to maximize the clinical benefit for patients while controlling the financial burden on the healthcare system. The Canadian setting was used as a reference throughout the manuscript, given the fact that Canada has a robust health-technology assessment process that is extensive, well-referenced, expert-reviewed, and used as a benchmark for HTA worldwide. Through this project, we reviewed the scientific and grey literature for health economic studies targeting the latest treatments for mHSPC and nm-CRPC approved by Health Canada, analyzed their potential benefit in the management of PCa in Canada, and identified knowledge gaps. This systematic literature review identified 24 and 20 health economic studies in the health states of mHSPC and nmCRPC, respectively, with the predominant type of analysis being cost-effectiveness analysis. The risk-of-bias assessment confirmed that the retrieved studies are of good quality in general. While only a few academic studies were conducted from a Canadian perspective, transferability analysis suggested that results from foreign studies would incorporate a small to medium level of bias if interpreted in the Canadian setting.
Our study identified 142 references, 80 of which included cost-effectiveness analyses for mCRPC, that were excluded from our analysis. Relative to the well-established health economic literature in mCRPC, the health economic literature for mHSPC and nmCRPC is still immature and there is a need for increased efforts to provide evidence-based support to healthcare decision-making. There is a significant unmet need for health economic evaluations that target mHSPC and nmCRPC and carry through disease progression until death while integrating all active treatment options and that are adapted to the Canadian setting.

mHSPC
The current literature review demonstrated that DOCE in combination with ADT was determined to be the most cost-effective treatment in the mHSPC setting. Compared to DOCE, comparators such as ENZA, APA, and ABI yield ICERs that are exceeding the predefined willingness-to-pay thresholds due to small incremental effectiveness benefits that are outweighed by considerably higher costs. This was also underlined in the cost analysis by Hu et al. 2019 [56], where the costs associated with ABI were 3 times greater than the costs of DOCE, CAD 259,909 vs. CAD 80,754 in the healthcare system perspective and CAD 64,510 vs. CAD 18,823 in the patient perspective. Hu et al. 2019 [56] ranked as a study of good quality according to our risk-of-bias assessment, but its results might be an underestimation of the costs according to our transferability analysis. It is important to mention that manufacturers often provide rebates to improve these ICERs. In Canada, the pan-Canadian Pharmaceutical Alliance (pCPA) is an organization comprised of provincial, territorial, and federal governments that aims to increase the value of publicly funded drug programs through their combined negotiating power [70]. Joint negotiations led by the pCPA for the reimbursement of ABI, APA, DARO, and ENZA for mHSPC and nmCRPC [71] have led to listing agreements and private discounted prices for these medications. Furthermore, patent expirations give birth to generic products that are available at lower prices. As of 2021, generic versions of abiraterone acetate are available on the Canadian market, some of which cost 73% less than the brand name product [72]. These lower prices will undoubtedly have an important impact, potentially making ABI the cost-effective option, as the price of abiraterone acetate was identified to have a major impact on the ICER [31,33,35,36,56]. Given the general trend, quality, and relatively good transferability of the retrieved studies, we can conclude that DOCE is the cost-effective treatment for mHSPC. These results could potentially be reversed if cost rebates on new acquisition prices are considered.

nmCRPC
In the nmCRPC setting, the results from this literature review inform that APA, DARO, and ENZA are considered cost-effective when compared to ADT alone. Furthermore, these three medications have similar ICERs compared to ADT alone, because they have demonstrated similar efficacy in clinical trials [73][74][75] and have similar drug acquisition prices in Canada [76]. It would be relevant to conduct a cost-effectiveness analysis with real-world data to compare their effectiveness in the Canadian setting. A Japanese real-world evidence analysis studied ENZA's effectiveness through a long-term medical records review [77]. In this study, Fujiwara et al. reported similar overall survival and slightly inferior progression-free survival when benchmarking against the PROSPER, PREVAIL, and AFFIRM clinical trials [77], which can be an indication that the effectiveness of ENZA would yield similar cost-effectiveness results if conducted with real-world data.
Cost analyses show an increase in healthcare costs as patients progress to metastatic disease underlining the importance of delaying progression. This increase is perceived in the inpatient and outpatient settings by Seal et al. [62] and Freedland et al. [60], where inpatient costs can be increased by up to threefold per patient per year after the appearance of metastasis. This increase can be perceived in the medical, pharmaceutical, inpatient, and outpatient costs [60,62,64].

Real-World Data Studies
As this review captured only a few health economic studies (i.e., cost analyses) using real-world data, it appears that clinical trials remain the main data source for conducting cost-effectiveness analysis in the nmCRPC and mHSPC settings. Real-world data represented the data source of choice for cost analysis, where researchers were able to determine the financial impact of transition between health states or the increased costs of treatment due to adverse events. The use of real-world data from administrative databases allows researchers to capture larger sample sizes, has greater external validity, and is more representative of clinical practice as patients outside of clinical trials tend to be older and have more comorbidities relative to trial patients.

Risk-of-Bias Assessment
The risk-of-bias assessment demonstrated that the selected studies were of good quality with a few exceptions. In general, studies did not satisfy the following criteria of the checklist: assumptions, costs measure methods, generalizability, and ethical and distributional issues. This underreporting can be explained by a lack of consideration or by the fact the authors conscientiously omitted the specification to comply with publication-specific constraints. This is an important aspect to acknowledge, since certain records are conference abstracts. In those cases, it would be impossible to report the full extent of the scientific effort. The CHEC extended checklist was selected for the risk-of-bias assessment as it is proven to be of greater scrutiny than others and it is recommended by the Cochrane collaboration [11]. Furthermore, the CHEC extended checklist is not only suitable for assessing modeling analysis but also cost analysis, which was one of its main advantages over the ISPOR questionnaire to assess relevance and credibility by Caro et al. [78] The Philips checklist [79] was another suitable option; however, because of its numerous criteria, it is not recommended for use in the assessment of a large number of studies.

Strengths
To our knowledge, this is the first systematic review that combines the health economic evaluations of mHSPC and nmCRPC. Furthermore, this study is the only one that considers governmental reports while conducting transferability analysis to the Canadian perspective. Through our literature review, we have encountered a similar review conducted by Grochtdreis et al. in 2018 [80], where the authors searched for cost-effectiveness analyses and cost-of-illness analyses targeting treatment for the CRPC and mCRPC. Quality assessment was conducted by using the CHEERS checklist and the risk of bias was assessed by the Bias in the Economic Evaluations checklist [80]. While this study was of great methodological quality, it did not consider the grey literature or analyses from HTA agencies, and nor did it conduct a transferability analysis. Furthermore, Grochtdreis et al. [80] did not extend their search to the mHSPC health state.
Through our review, we identified significant knowledge gaps. For instance, very few studies consider mHSPC and nmCRPC simultaneously in their analysis, the primary reason being that these are mutually exclusive health states that require a specific indication for a drug to be used. That being said, there is an important androgen receptor-axis-targeted therapies (ARAT) usage overlap in the mHSPC and the nmCRPC settings. Moreover, as both health states eventually lead to mCRPC, considering them jointly integrates a more complete spectrum of the disease. Additionally, as cost-effectiveness analysis is often used to justify treatment reimbursement, analyses were designed to compare active adjunct treatment plus ADT to ADT alone. Given the growing landscape of treatments for advanced PCa, future health economic models should not only consider ADT as the standard of care but also consider the other active treatments that are given in combination with ADT, as was conducted in CADTH's pharmacoeconomic report of APA for mHSPC [23]. In this study, APA was benchmarked against DOCE, ABI, and ADT alone. Furthermore, some studies are conducted from the societal perspective that may be biased as they do not provide indirect costing components such as productivity loss to the patient and the healthcare provider. However, patient productivity loss is likely to be low, given that PCa is a disease of old age with the average age of diagnosis being above 65 [81]. Nonetheless, this should be acknowledged in the design and discussed by the authors as it is an important part of the societal perspective.

Limitations
This systematic review was based on peer-reviewed methods designed specifically for health economic articles and was conducted with great scrutiny [11]. However, as with all systemic reviews, this study has certain limitations. Because the number of captured studies was relatively low and because they did not always report results by subgroup of patients based on disease severity, we could not stratify our analyses beyond the health states of mHSPC and nmCRPC. As this review protocol was not registered in PROSPERO, it was not peer-reviewed and may incorporate a certain level of bias. To overcome this bias, the review protocol was designed to have wide inclusion criteria and cover various databases, including the grey literature. By reviewing the grey literature, conference abstracts, and reports that are not peer-reviewed, the research exposes itself to biases. Correctly assessing the quality of these publications is not possible as some of these publications are not reporting their full protocols and results, either due to publication-length limits or confidentiality agreements. To tackle this problem, other literature reviews have excluded conference abstracts and governmental HTAs [80]. We decided to include grey literature in our analysis to preserve a high level of sensitivity in our analysis. We were, however, faced with a challenge when assessing the risk of bias in abstracts and governmental reports. For abstracts, we considered that all unreported items from the CHEC extended checklist were omitted and therefore might have underestimated the quality of some publications. While we considered all the items of the CHEC extended checklist to carry the same weight, this grading scheme has not been validated. It is important to mention that the criteria list of the CHEC extended checklist is regarded as a minimum standard [13]. A good-quality health economic study should therefore satisfy all the items. Consequently, the CHEC extended checklist is not intended to be used as a grading system and these results should be interpreted with caution. Through our analysis, we did not capture a single study that satisfied all the items, and only five publications had one unsatisfactory item. This indicates that there is an unmet need for high-quality publications in the field.
We decided to exclude governmental HTA reports from the risk-of-bias assessment analysis because of the high level of underreporting due to confidentiality agreements. Furthermore, HTA reports from CADTH and INESSS were not captured by our search and were added manually to satisfy the scope of this analysis. This could potentially lead to article-selection bias or the omission of certain reports. It is important to mention that HTA entities do not provide sufficient information for model reconstruction and model validation by peer scientists because of confidentiality agreements with treatment manufacturers. However, their results remain important for consideration, serving as a robust benchmark for academic research. Ignoring them will lead to a significant study-selection bias.
Another limitation of this study is that we were not able to integrate cost-effectiveness thresholds in the analysis because they are country-or healthcare-system-specific. The United Kingdom's NICE uses an official explicit cost-effectiveness threshold of GBP 20,000 to GBP 30,000 per QALY. In the United States, this threshold is between USD 50,000 to USD 100,000, while in Canada the same threshold is being referred to, but in Canadian dollars. Although the United States and Canada have historically referred to these thresholds without officially endorsing them; certain medications exceeding these thresholds have been judged cost-effective. Furthermore, converting these thresholds from their local currency to 2021 CAD may result in significant bias and is not considered a recommended practice as they have not been updated to reflect the current country-specific purchasing power. We have decided therefore not to benchmark our results against these thresholds that are not always explicitly endorsed and that might be biased as they have not been updated to reflect the current value of money and country-specific purchasing power.

Conclusions
This literature review describes the current state of health economic studies on mHSPC and nmCRPC. We identified docetaxel plus ADT to be the cost-effective treatment for mH-SPC in most of the retained publications. Enzalutamide, apalutamide, and darolutamideall in addition to ADT-were associated with similar ICERs when compared to ADT alone. Additionally, through the risk-of-bias assessment and transferability analyses we found that while the current literature provides guidance, study results cannot be applied directly to the Canadian healthcare system without incorporating a certain degree of bias. Finally, we conclude that the scientific literature is immature. We identify an important unmet need for health economic evaluations in the mHSPC and nmCRPC settings incorporating Canadian real-world data to support healthcare decision-making to effectively manage advanced PCa.