Background

Because prostate cancer (PCa) is androgen-sensitive, castration by manipulating the hypothalamic-pituitary-gonadal axis to achieve the lowest testosterone levels is an important therapeutic intervention in patients with advanced cancer.¹ Yet this comes at a cost, because adverse metabolic changes, notably weight gain, insulin resistance, and dyslipidemia, possibly contribute to a higher risk of cardiovascular disease (CVD).^2–5 Current evidence linking androgen deprivation therapy (ADT) and CVD is mixed. As reported by Hu et al,⁶ 3 meta-analyses of observational studies comparing risk of cardiovascular events among patients treated with versus without ADT suggested an increased risk associated with ADT, but none of the meta-analyses of randomized controlled trials (RCTs) found this effect. The discrepancy may lie in the differences in the patient population studied in observational studies and those eligible for randomized trials, with the former including patients with clinically significant cardiovascular comorbidities at baseline who are typically excluded from randomized trials. In addition, study methods for cardiovascular endpoints in these oncology studies were possibly not as refined and specific as in cardiovascular studies. A decade ago, the FDA mandated safety labels of gonadotropin-releasing hormone (GnRH) agonists to include additional warnings about potential increased risks for diabetes and cardiovascular events, and advocated for physicians to closely monitor symptoms suggesting new CVD in patients prescribed GnRH agonists.⁷

Current literature on the safety profile of different forms of ADT is still not clear, making it difficult to draw any conclusion.¹ A meta-analysis pooling data from randomized trials comparing the GnRH antagonist degarelix with GnRH agonists favored the former in terms of cardiovascular safety.⁸ Notably, the incidence of cardiovascular events was strongly reduced in the subgroup comprising patients with preexisting cardiovascular history at baseline only. These findings further highlighted the need to consider stratifying analyses by baseline preexisting history of CVD when evaluating the cardiovascular safety of these treatments.

In this retrospective study using a Canadian provincial database, the objective was to evaluate the association between ADT type and the risk of CVD in patients newly treated with GnRH agonists or antagonist in both the subgroup with preexisting CVD and the subgroup without preexisting CVD.

Discussion

Since the initial reports linking ADT to cardiac events emerged at the beginning of the 21st century, we witnessed a surge of literature evaluating the cardiovascular risk in patients treated with ADT.¹⁸ The association between GnRH agonists and GnRH antagonist and CVD has not yet been clearly defined. Our study is the first of its kind to assess the association between these 2 compounds and CVD using the propensity score technique in a contemporary North American population. Our data suggest that the use of GnRH antagonist compared with GnRH agonists was associated with an increased risk of developing arrhythmia in patients with no prior CVD but was associated with a decreased risk of developing HF in patients with prior CVD and IHD in patients with no prior CVD, respectively. Although not statistically significant, a trend toward a decreased risk in the antagonist group was observed for other CVD, particularly MI independently of prior CVD history.

A French population-based cohort study including 35,118 patients in which 71% received GnRH agonists and 3.6% received the GnRH antagonist also found no significant difference in ischemic events (MI and ischemic stroke) between the 2 groups.¹⁹ A recent Italian observational study adjusting for confounders showed a lower tendency of developing MI and stroke among patients receiving GnRH antagonist (degarelix), although not significant, in the overall cohort and the subgroup with no prior CVD.⁵

Using real-world data from the United Kingdom, an observational study demonstrated decreased risks of any cardiac event and arrhythmia in degarelix users; however, the analyses were not adjusted for potential confounders.²⁰ A hypothesis-generating post hoc analysis of 6 pooled phase III RCTs from 2005 to 2012 with 2,328 patients comparing degarelix (64%) versus GnRH agonists (36%) showed a 40% reduction in cardiac events or death in the degarelix group during the first year of treatment, and an absolute risk reduction of 8.2% in patients with prior CVD. These findings were shown in studies with longer follow-up periods (up to 14 months) only, rendering time an important element in disease manifestation.⁸ However, the post hoc nature of this analysis and the fact that cardiac events were not the primary endpoints in the individual studies renders it inadequate for drawing definitive conclusions. Similarly, a recently published phase II RCT in men with PCa and prior CVD who were randomized to receive either GnRH agonist or antagonist for 1 year found an absolute risk reduction of 18% of developing new CVD using GnRH antagonist, although it did not show any difference in endovascular function (primary outcome).²¹ However, the small patient population (n=80) of this trial and reporting of a smaller variety of cardiovascular events (MI and stroke only) as secondary endpoints contribute to its weakness in delivering a clearer picture of this association.

On the contrary, an observational study combining 5 databases demonstrated an increased risk of developing arrhythmia in degarelix users with or without prior CVD, and an increased risk of new MI in those with prior CVD.²² Although this study stratified their analysis by prior history of CVD, it is unclear whether they adjusted for additional potential confounding variables, which is important in this particular context to minimize confounding by indication.

To date, only one RCT (the PRONOUNCE trial) was designed to compare head-to-head the effects of these compounds on cardiovascular health in patients with PCa with prior CVD (ClinicalTrials.gov identifier: NCT02663908). Unfortunately, the trial was stopped prematurely due to low enrollment.²³ Data available in the trial are essentially inconclusive given the limited number of patients enrolled generating large imprecision in effect estimates. It is noteworthy that the recently FDA-approved oral GnRH antagonist relugolix had shown a >50% relative risk reduction in the incidence of major adverse cardiovascular events compared with the GnRH agonist leuprolide.²⁴ However, the incidence of cardiovascular events was not the primary endpoint of the trial.

Interestingly, our data demonstrate that an antagonist compared with GnRH agonists was not associated with the risk for MI and stroke, but it was associated with a significant risk reduction of HF, particularly in those with prior CVD, and of IHD in those with no prior CVD. The effects of ADT on metabolic syndrome are unlikely to fully explain the differential effects observed between GnRH antagonist and agonists. Other factors unique to GnRH agonists, namely the testosterone flare, the effects on follicle-stimulating hormone (FSH) release, and the immune system, may provide another explanation. First, testosterone flare associated with GnRH agonists may result in a proinflammatory systemic state.²⁵ Second, users of GnRH antagonist experience a rapid and large (>90%) decrease in FSH levels, whereas users of GnRH agonists only experience a moderate (50%) decline.²⁶ Experimentally, FSH appeared to promote visceral lipogenesis and fat storage both in vitro and in vivo,²⁷ potentially contributing to atherosclerotic plaque development. A preclinical study suggested that mice treated with GnRH agonists accumulated more adipose tissue and atherosclerotic plaque compared with mice treated with a GnRH antagonist, and these differences may be explained by the lower FSH levels attained in the latter group.²⁸ Third, activation of GnRH receptors on T-cell lymphocytes potentially generates an inflammatory immune response. Preclinical studies have shown that GnRH agonists led to atherosclerotic plaque destabilization, whereas the GnRH antagonist did not.²⁹ However, clinical data showed no difference in endovascular function.²¹

Moreover, our data showed an arrhythmogenic activity associated with the use of GnRH antagonist, especially in patients with no prior CVD. An observational study similarly found an increased risk of arrhythmia in GnRH antagonist users with or without prior CVD.²² GnRH agonists were also found to elevate the risk of arrhythmia in patients with PCa, although only in those with prior CVD.³⁰ More recently, a longitudinal observational study showed subclinical derangements in cardiac parameters in patients with PCa receiving ADT with prolongation of QTc segment.³¹ Both GnRH agonists and antagonists were shown to potentially induce torsades de pointes through testosterone reduction leading to QT prolongation.^32–34 The proposed mechanism may reside in the interaction between testosterone and cardiac ion channels delaying ventricular repolarization. This finding led the FDA to issue warnings of QT prolongation for both GnRH agonists and antagonists.³² However, this cannot explain the difference we observed in our study between the 2 groups. In a brief report using the European pharmacovigilance data, the GnRH antagonist was more likely to be reported for drug-induced long QT syndrome compared with GnRH agonists.³⁵ Overall, our observation is in contrast to the safety findings from the RCTs comparing GnRH antagonist and agonists in which rates of QT prolongation were similar between groups, though the number of events was very low.^26,36 Nonetheless, given the complexity of the matter, our finding warrants further studies to elucidate the underlying mechanisms.

Our study is not without limitations. There are inherent biases associated with the use of a database, possibly introducing information biases (misclassification due to miscoding). Due to the use of administrative healthcare claims data, confounding factors related to ADT and CVD, such as like lifestyle characteristics (eg, alcohol use, smoking), pathologic data (eg, Gleason score, disease stage), and laboratory results (prostate-specific antigen level), were not measured. Despite accounting for other confounding variables through appropriate statistical methods using propensity scores, other unknown or unmeasured factors linking CVD to ADT might have been missed. These might lead to potential residual confounding even after using the IPTW method, because this does not account for these factors. Some of the changes observed between crude and IPTW adjusted hazard ratios might be explained by this. Our study has its strengths, however. First, large provincial administrative databases portray the effects of these treatments in actual clinical practice, unlike RCTs including highly selected individuals under strict protocols. Second, we evaluated a variety of cardiovascular events combined as a composite outcome and separately as individual components to allow for a more distinctive examination of the association between the type of ADT and specific types of CVD. Last, use of the IPTW method reduced the confounding bias by using balanced groups in terms of baseline variables.

Conclusions

Compared with GnRH agonists, the GnRH antagonist was found to be associated with a decreased risk of developing HF and an increased risk of developing arrhythmia in patients with PCa. Risk of IHD was also lower in patients with no prior CVD receiving GnRH antagonist. Consistent with the joint statement from the American Heart Association, American Cancer Society, and American Urological Association, the cardiovascular profiles of patients should be optimized before initiating ADT, and close follow-ups with primary care physicians or cardiologists should be arranged to monitor for CVD signs and symptoms.