Risk of cardiovascular disease following gonadotropin‐releasing hormone agonists vs antagonists in prostate cancer: Real‐world evidence from five databases

Abstract Observational studies in prostate cancer (PCa) have shown an increased risk of cardiovascular disease (CVD) following gonadotropin‐releasing hormone (GnRH) agonists, whereas randomised‐controlled trials have shown no associations. Compared to GnRH agonists, GnRH antagonists have shown less atherosclerotic effects in preclinical models. We used real‐world data from five countries to investigate CVD risk following GnRH agonists and antagonists in PCa men. Data sources included cancer registries, primary and secondary healthcare databases. CVD event was defined as an incident or fatal CVD. Multivariable Cox proportional hazard models estimated hazard ratios (HRs) and 95% confidence intervals (CIs), which were pooled using random‐effects meta‐analysis. Stratified analyses were conducted by history of CVD and age (75 years). A total of 48 757 men were on GnRH agonists and 2144 on GnRH antagonists. There was no difference in risk of any CVD for men on GnRH antagonists and agonists (HR: 1.25; 95% CI: 0.96‐1.61; I 2: 64%). Men on GnRH antagonists showed increased risk of acute myocardial infarction (HR: 1.62; 95% CI: 1.11‐2.35; I 2: 0%) and arrhythmia (HR: 1.55; 95% CI: 1.11‐2.15, I 2: 17%) compared to GnRH agonists. Having a history of CVD was found to be an effect modifier for the associations with some CVD subtypes. Overall, we did not observe a difference in risk of overall CVD when comparing GnRH antagonists with agonists—though for some subtypes of CVD we noted an increased risk with antagonists. Further studies are required to address potential confounding caused by unadjusted variables such as severity of CVD history and PCa stage.

when comparing GnRH antagonists with agonists-though for some subtypes of CVD we noted an increased risk with antagonists. Further studies are required to address potential confounding caused by unadjusted variables such as severity of CVD history and PCa stage. Degarelix, a GnRH antagonist introduced in 2010, has been associated with a lower risk of CVD in men with PCa. 9,10 Preclinical models have shown less atherosclerotic effects in mice treated with the GnRH antagonist as compared to those treated with GnRH agonists. 11 Although GnRH agonists are a GnRH inhibitor, GnRH antagonists are a GnRH blocker that completely blocks GnRH receptors. The difference in mechanism of action leads to an immediate mode of action in GnRH antagonists associated with its reduced side effects. 12 Phase II and phase III studies showed no difference in terms of efficacy and baseline testosterone levels in men receiving GnRH antagonists for 1 year compared to men receiving various GnRH agonists for their PCa. 13 Comparison of the CVD safety profile in men on GnRH agonists and antagonists has yielded inconclusive results. 8,14 Although meta-analysis of observational studies 8 have shown a lower risk of CVD in men on GnRH antagonists compared to GnRH agonists, meta-analyses of randomised-controlled trials (RCTs) have shown no such associations. 14 Moreover, these studies were not designed with CVD as a primary outcome.
A phase III RCT (PRONOUNCE; ClinicalTrials.gov identifier: NCT02663908) is recruiting to compare the risk of fatal or nonfatal CVD in 900 men with PCa receiving GnRH agonist or antagonist as primary treatment. 15 Results of the PRONOUCE trial is not expected until 2021, which justifies the need for observational evidence in the interim.
Since RCTs often exclude elderly participants and those with comorbidities, two most common characteristics of men with PCa receiving ADT, 16 real-world data used in observational studies may provide evidence applicable to the general PCa population. 17,18 Only one observational study has been conducted to date directly comparing risk of CVD between GnRH agonists and GnRH antagonists. Scailteux et al showed no difference in risk of developing stroke or myocardial infarction in men with PCa receiving either treatments; however, overall CVD was not investigated as a specific outcome in the study. 19 By combining real-world data from five countries, we designed a study with sufficient power to compare risk of CVD between GnRH agonists and antagonists in a real-world setting. Our study is the first observational study to directly compare the risk of six CVD outcomes between GnRH agonists and antagonists by using country-specific analyses from patient level data from five countries. We explored six CVD outcomes: any CVD, ischaemic heart disease (IHD), acute myocardial infarction (AMI), arrhythmia, heart failure (HF) and stroke.

| Study design and population
This observational study combined data from five countries to investigate the association between GnRH agonists or antagonist (degarelix) and risk of CVD in men with PCa. All men with PCa prescribed or dispensed GnRH agonists or GnRH antagonists were included in the study. A detailed study protocol aimed to minimise heterogeneity in terms of data collection across countries by outlining exact codes for extraction of all study variables is published elsewhere. 20 Data from the United Kingdom's The Health Improvement Network (THIN) database (excluding Scotland) 21  primary healthcare database (ie, general practices and community healthcare settings) and NHSS is a secondary healthcare database (ie, hospitals and outpatient clinics), BCR and PHARMO include both primary and secondary healthcare data. [21][22][23][24][25][26][27] The SNIIRAM database is a claims database combining claims from insurance plans with the National Hospital-discharge Summaries database system (PMSI). 26 As Scotland is in the United Kingdom and there may have been some overlap of men with PCa in the THIN database and Scottish NHSS database, men with PCa with a postcode in Scotland were excluded from THIN.
The study period extended from 2010 to 2017 in the five countries.

| Exposures
Exposure was defined as a prescription or dispensation of GnRH agonists or GnRH antagonists. We included all men with locally advanced or metastatic PCa (in countries where PCa stage was available, Supplement Table 1) who started on GnRH agonists or GnRH antagonists.
Men were followed from the date of GnRH agonists or antagonist's initiation until outcome of interest, switch between agonists and antagonists and vice versa, orchiectomy, end-of-study period or death from other causes, whichever came first. No further exclusion criteria were used.

| Statistical analysis
We conducted a two-stage analysis: firstly, we obtained countryspecific hazard ratios (HRs) from Cox proportional hazards models using age as a timescale to assess heterogeneity in each country, T A B L E 2 Hazard ratios from random-effects meta-analytical models including different stratification for any CVD, ischaemic heart disease, acute myocardial infarction, arrhythmia, heart failure and stroke for the five included countries A further sensitivity analysis was also conducted to exclude men who may have been on GnRH agonists or GnRH antagonists for less than 3 months to eliminate short-term neoadjuvant or adjuvant use.
All country-specific analyses were conducted using different SAS versions in Belgium (9.4) and France (9.4) and statistics and data (STATA) versions in the United Kingdom (14C), Scotland (14) and the Netherlands (14C). The meta-analysis was conducted using STATA version 14C.
3 | RESULTS Table 1 Figure 1A Table 2) compared to the main analysis (Table 2). Sensitivity analysis with GnRH initiation date 3 months after the GnRH start date showed no statistically significant findings for arrhythmia (Supplement Table 3).

| DISCUSSION
Our study is the first to combine real-world data from the United Our study showed that there was an increased risk of developing any CVD, AMI and arrhythmia in men on GnRH antagonists with a prior CVD history compared to GnRH agonists. Our final results are in contrast to our preliminary findings 30 using four of the five countries, which looked at the proportion of men developing a CVD event in the exposure groups, without accounting for age or follow-up period. The difference in methods may have reflected the results as we pooled country-specific HRs in our final meta-analysis of five countries.
This is the first study to show an increased risk of developing CVD subtypes in men on GnRH antagonists compared to GnRH agonists. Although this is significant, it is important to note that degarelix, the GnRH antagonist, was a new drug during the study period with strict prescription guidelines that tailored the drug to specific PCa population (ie, to those with preexisting CVD). Even though we have accounted for history of CVD by history of CVD indicator stratified analysis, we may have selected a population with underlying CVD risk leading to an increased risk of CVD in men on GnRH antagonists observed in our study. Moreover, we only accounted for history of CVD indicator 12 months prior to GnRH initiation due to data unavailability, which may also have limited the number of CVD events in both arms of the study. The generic history of CVD indicator variable may have led to a loss of granularity in terms of assessment of history of CVD. This may have contributed to confounding by indication in our study due to the inability to adjust for precise history and severity of prevalent CVD. Moreover, men with more advanced stage PCa who are given more aggressive forms of PCa treatments may already have had elevated CVD risks due to their disease stage and treatments. By not accounting for PCa stage in our study, we may have missed some of the information indicative of treatment history and potential increased CVD risks.
Our findings showing an increased risk of CVD subtypes in men on GnRH antagonists compared to GnRH agonists contradicts prior literature showing an increased CVD risk among men using the GnRH agonists compared to GnRH antagonists. [9][10][11]31 Although GnRH agonists work by inhibiting the release of GnRH, antagonists work by inhibiting both GnRH and follicle-stimulating hormone (FSH) receptors. [32][33][34] The additional inhibition of FSH receptors by GnRH antagonists may reduce the risk of a recurrent CVD due to FSH receptors' role in lipid metabolism and in fat accumulation. 35 However, results from our real-world data showed contradictory effects, which may be partly explained due to confounding by indication.
In our methodological protocol, we used the Risk of Bias in Non-Randomised Studies of Interventions (ROBINS-I) tool to assess our study design, which emphasised three main forms of biases: misclassification of study variables, channelling or indication and unmeasured confounding. 20 We avoided misclassification bias, where possible, by following a standard protocol 20 to extract study variables from the five countries involved. Channelling or indication bias 36 occurs when a physician prescribes GnRH antagonists to men with a history of CVD based on previous evidence. 37 We were unable to avoid channelling bias in our study because data on physician preferences were not available. In addition to no information on physician preference, heterogeneity in guidelines for prescription of GnRH antagonists in the five countries may have also affected the study results (discussed in George et al 20 ).
In this large prospective cohort study, we found it difficult to fully homogenise study variables (Supplement Table 4). A large proportion of this complexity was attributed to the varied data sources used in our study. THIN from the United Kingdom was the most distinct database used in the study, due to its data derivation exclusively from primary healthcare settings. 21 As degarelix was a fairly new drug on the market 38 and a much smaller proportion of men were degarelix users in individual countries, there was a need to combine data from all countries in our analysis. We attempted to account for this heterogeneity in the data sources by not only setting up a standard protocol 20 but also by focusing analyses on data that was fully available in all five countries (ie, history of CVD indicator). Moreover, further sensitivity analyses excluding the United Kingdom (Supplement Table 2) and a delayed start date of 3 months after GnRH initiation date (Supplement Results for some subtypes of CVD were limited due to the data sources that they were obtained from. For example, the acute nature of AMI means that it is usually recorded in an acute hospital setting. 40 The United Kingdom had no AMI events for analysis because the THIN database originates from a primary healthcare setting. Therefore, further assessment of hospital registries is needed to understand the risk of developing AMI in men with PCa on GnRH analogues.
A key strength of our study was the use of data from five countries that made study results applicable to the general PCa population.
Moreover, the use of different types of databases (primary healthcare, secondary healthcare, cancer registries and claims databases) also ensured the inclusion of rare, adverse events that may not have been identified in an RCT. However, our study also highlighted the challenges involved of using real-world data. Although the potential for real-world data is large in the healthcare setting, differences in data sources need supervised reconfiguration of data for real-world data to achieve its full potential. The way forward for researchers using realworld data is to combine and analyse "big data" from databases from various institutions into a single platform through projects such as the GetReal Initiative 41 and Prostate cancer dIagnOsis and treatmeNt Enhancement through the power of big data in EuRope (PIONEER), 42 which are part of the Big Data for Better Outcomes (BD4BO). 43

| CONCLUSION
Our study across five countries provided little support for a difference in risk of any CVD between GnRH antagonists compared to GnRH agonists, but there was some evidence of increased risk of certain CVD subtypes. Since our results are based on real-world data, they may be more applicable to the general PCa population who are on hormonal treatment. However, the potential for indication bias in our study needs to be addressed through RCTs, such as the PRO-NOUNCE trial.

CONFLICT OF INTEREST
JGK is an employee of the PHARMO Institute for Drug Outcomes Research, which is an independent research institute that performs financially supported studies for government and related healthcare authorities and pharmaceutical companies. The other authors declare that there is no conflict of interest.

DISCLAIMER
The views expressed in this article are the author(s) own and not that of any particular institution or funder.

DATA AVAILABILITY STATEMENT
The data are not publicly available due to privacy or ethical restrictions. The results presented in this article are from the meta-analysis (stage two) which combine country-specific hazard ratios as combining data from all countries at an individual level was not possible due to ethical and legal restrictions on data sharing. Datasets that are minimally required to replicate the outcomes of the study will be made available upon reasonable request.

ETHICS STATEMENT
Ethical approval was obtained for all data used in the study from relevant ethical committees and healthcare authorities in the five countries.