The Association between the Risk of Cardiovascular Disease and Androgen Deprivation Therapy in Patients with Prostate Cancer. A Meta-Analysis and systematic review.

Background : Androgen deprivation therapy (ADT) is widely being applied in men who suffered from prostate cancer, our aim is to evaluate whether ADT is associated with an excess risk of cardiovascular disease (CVD). Method : Studies comparing the the incidence of CVD between ADT group and control group were identified through literature search in electronic databases (Pubmed, Embase, Web of Science) until July 2019 and only observational studies and randomized controlled trials (RCT) were included. The estimating relative risk ratio (RR) and 95% confidence intervals (CI) were calculated through random effects meta-analyses. Result : A statistically significant association was detected for acute myocardial infarction (AMI) with RR = 1.22; 95% confidence interval CI, 1.05–1.43; P< 0.05. Significant relationship between coronary heart disease (CHD) and ADT was also observed, with summary RR=1.19; 95%CI, 1.03-1.38; P<0.05. ADT was associated with a risk increasement for heart failure (HF) with RR=1.15; 95% CI 1.01–1.33; P< 0.05. On the contrary, ADT was not associated with an increased risk of sudden cardiac death (SCD). Conclusions : From this study, ADT is associated with increased risk of AMI, CHD, and heart failure (HF); in contrast, this association is not detected in SCD; various modalities of ADT could significantly increase the risk of CHD, AMI, except for oral anti-androgen (AA). Our meta-analysis also suggests that the long-term application of ADT in prostate cancer patients would not result in a significant increase in AMI incidence compared with short-term. Moreover, the combined application of AA and GnRH agonists would lead to a similar risk of AMI compared with orchiectomy or GnRH agonists monotherapy whereas higher risk of CHD was detected when compared GnRH agonists plus AA with orchiectomy.

The PICO research question was: Whether patients with prostate caner (Population), treated by ADT (Intervention) compared to placebo (Comparison intervention) suffer from higher risk of CVD (Outcome), in observational and RCT studies (Study design) Studies were regarded eligible if they (1) designed as observation study or randomized controlled trial (RCT) study; (2) the research object is cardiovascular disease with AMI, CHD, HF or SCD as endpoint; (3) with sufficient data for analysis; (4) ADT type was prespecified and ADT duration was more than 6 months; (5) patients with baseline cardiac comorbidities were separated. All screening process was accomplished with the bibliographic software EndNote X9 (Clarivate Analytics)

Quality Assessment and Data extraction
All included observational studies were evaluated for risk of bias by two reviewers (ZL and RH) through Newcastle Ottawa Scale (NOS) [9] with≥7 score representing high-quality while the assessment for RCT studies was based on Cochrane Collaboration's tool for risk of bias [10]. RCTs were evaluated in terms of : Random sequence generation, Allocation concealment (Selection bias); Blinding of participants and personnel (Performance bias); Blinding of outcome assessment (Detection bias); Incomplete outcome data (Attrition bias); Selective reporting (Reporting bias) ; Anything else, ideally prespecified (Other bias). There are three levels: low, unclear, or high risk for the quality of RCT evidence. The quality evaluation was carried out by two independent reviewers (ZL and SS), and a third reviewer was consulted to adjudicate any disagreements (YY). Data were extracted by two researchers (ZL and NA) independently to minimize the bias. The following data were extracted: 1.
Basic information about study (total number of participants, number of ADT users, number of control group, follow-up duration, treatment method, definition of CVD )
3. experimental results ( outcomes measure and quantity of AMI, CHD, HF and SCD)

Statistical Analysis and Subgroups Analyses
The heterogeneity of included studies was evaluated using the Cochrane I 2 statistics as well as a ''remove-one'' analysis. P < 0.05 was defined as invalid assumption of homogeneity. A frequentist meta-analysis according to random effects model was performed for all outcomes to assess relative risk (RR) and their 95% confidence interval (95%CI) using the method of the Mantel-Haenszel test.
The outcomes were then presented as forest plots.  Figure 1 represents a flowchart demonstrated the selection process that met our criteria aforementioned, a total of 1541 potential records were initially identified from databases through the search terms listed above. 531 duplicate studies were excluded. 50 were considered closely correlated to the concept of this study and underwent full text review. 14 studies were removed because they did not mention AMI, CHD, HF or SCD separately as outcomes. 2 were excluded due to insufficient data, 7 studies were excluded because ADT duration was shorter than 6 months, 6 did not specifically report the type of ADT. Additionally, we remove 9 retrospective researches from our study, 2 were excluded for using the same databases as studies that were already included.  Figure 2 and no research was excluded due to low quality. According to the Beggs' funnel plot, the likelihood of publication bias was low, with a p value of 0.097 for the slope coefficient. (Figure 3) The ''remove-one'' analysis did not demonstrated influences of one specific study.

Association between ADT and AMI
8 studies were regarding the relationship between ADT and AMI.

Association between ADT and CHD
There were a total of 532,919 participants from all the 7 included studies investigating the relationship between ADT and CHD of which 157,165 received ADT and 375,754 were control groups.
A significantly increased CHD risk was found in the prostate cancer patients treated with ADT.

Association between ADT and SCD HF
Subgroup analyses of the relationship between ADT and SCD were also investigated with total of 182,403 identified events from 5 studies, among them 79,881 were ADT users and 102,522 were from control groups. We identified that SCD was not significantly associated with ADT usage. (RR=1.13; 95% CI 0.89-1.45; P< 0.05) . (Figure 10) Subgroup analysis for different types of ADT was not performed as there was only one study [14] reported exact ADT method. Figure 11 showed the forest plot of HF for patients with or without ADT treatment. Pooled data for HF were available from 6 studies with a total of 334,093 patients (97,925 with ADT exposure and 236,168 without ADT exposure). Based on our results, ADT was associated with a higher incidence of HF compared to control group. (RR=1.15; 95% CI 1.01-1.33; P< 0.05). Subgroup analysis was also not performed for ADT type because of the small number of trials that reported such outcomes.

Subgroup by duration of ADT
In order to reduce the impact of inconsistent endpoints on our conclusion, whether the duration of ADT application was associated with a significant effect on the risk of AMI events was also explored.
The included studies were separated into different groups based on the duration of ADT application,

Discussion
The correlation between ADT and cardiovascular toxicity is a debatable topic in PCa treatment.
Multiple researches explored this relationship with different outcomes.
Our results indicated that ADT was associated with a significant increase in risk of AMI, CHD, and HF, but not found to be associated with SCD; the individual administration of each type of ADT was associated with AMI and CHD, except for AA; AA alone was only significantly associated with the increased risk of AMI, but not CHD. In addition, using ADT more than 5 years would not lead to an increased risk of AMI compared with less than 5 years. Multiple earlier studies have proved that GnRH agonists were associated with a higher risk of CVD compared with orchiectomy [44], whereas our study indicated orchiectomy monotherapy could also lead to cardiac events. Orchiectomy can lead to a low level of testosterone though it remains unclear how low testosterone levels causes major cardiovascular events. Callou de Sá et al. [44] indicated that men with CHD had higher oestradiol and free oestrogen index (FEI) levels. As a matter of fact, high level estrogen can increase heart attack risk by accelerating coagulation and platelet aggregation in coronary arteries [45]. A nationwide, population-based study, provided exclusive evidence to show that closed risk of fatal CVD was observed for men treated by GnRH agonists compared with orchiectomy [25]. Further, larger population-based trial is needed to determine whether interventions like orchiectomy that raise estrogen levels might promote the progression of CHD, so that clinical doctors could be conscious of the serious potential risks of orchiectomy and ensure medical safety when deciding the type of ADT for patients.
Our subgroup analysis for the duration of ADT indicated that long-term ADT was not associated with an excess risk of AMI when regarding 5 year as threshold. In an observational study, Efstathiou et al.
[23] indicated that long-term ADT administration could not increase the risk of cardiovascular mortality compared with short-term in men with locally advanced PCa. In another study, patients from control group received short-term androgen suppression while patients who received 2-5 years of further treatment were recruited as experiment group, the result demonstrated that no significant difference was observed in overall mortality between short-term group and long-term group [22]. On the other hand, evidence from another more resent study suggested the cardiovascular risk factors such as hyperglycemia, frank diabetes, and MetS are more likely to occur in patients with long-term ADT over 12 months, while with short-term ADT, (3-6 months) users are only affected by temporary insulin resistance [31]. Therefore, the optimum duration of ADT application should be adapted to comorbidities conditions and risk factors of individual patients.
Based on our results, the administration of ADT was not associated with SCD, but led to an increased The previous literature also showed conflicting evidence evaluating the link between ADT and HF.
Multiple previous studies revealed men treated with ADT were associated with significantly increased of fatal and nonfatal HF in all patients especially for those with pre-existing CVD which met our results [32]; whereas only one propensity-score matching cohort study from Canada demonstrated that incidence of HF was not higher among ADT users with HR= 0.95; 95%CI, 0.90-1.00. Research which explored the impact of ADT on cardiomyocyte contractility at molecular level, and the results of testosterone therapy for HF at clinical level is needed to specify the possible relationship between HF and ADT.
This study has limitations. First, although we have strictly followed the PRISMA guidance, tried our best to apply the most extensive keywords and conducted the selection as impartial as possible, we know that some potential studies may still be neglected in our search, which would compromise our results. Second, the heterogeneity of our study was high, probably due to included studies used different criteria for patient recruitment, different defination of CHD, AMI, SCD, HF among studies and different treatment method for PCa, and because they included disease of variable severity. We could not explore the source of the heterogeneity through subgroup analyses because of the limited quantity of included studies. However, the ''remove-one'' sensitivity analysis suggested that our findings are convincible. The funnel plot also indicated that there is no publication bias. Future studies should aim to recruit consecutive cases under ADT treatment with more detailed subgroup analysis and assess cardiac risk using standarlized instruments that have cross-centre validity. Third, our analysis pooled together both RCT and observational studies, and subgroup analyses based on study design were not performed in the AMI and CHD subgroup, because all the included studies are designed observationally for AMI and CHD subgroup. Fourth, some included studies do not provide direct data for analysis, so we have to calculate and extract the data we needed by ourselves, this may have an impact on the overall result to some extent.

Conclusions
Our analysis demonstrates an increased risk of CHD, AMI, HF for ADT application; ADT is not connected with SCD occurrence and it requires further exploration; various ADT modalities have different impact on CVD risk in different level, the combined application of AA and GnRH agonists would not significantly increase the risk of AMI compared with individual method; AA is less likely to trigger onset of CHD compared with other methods but would significantly lead to AMI. In addition, this study also reports that that short-term use and long-term use of ADT lead to similar risks of AMI.
Therefore, ADT is connected with a significant negative impact on life quality. Cautions and periodic cardiovascular elevation are necessary for patients before the ADT starting. More experimental and epidemiological studies are needed to discriminate the SCD effects on different types of ADT.

Declarations
Ethics approval and consent to participate: Not applicable Consent to publish: Not applicable Availability of data: All data is included in the article and additional files.

Competing interests:
The authors declare that there are no conflicts of interest.  Acute myocardial infarction risk associated with ADT Figure 5 RRs of acute myocardial infarction related to different types of ADT.

Figure 6
GnRH agonists plus AA compared with other method for acute myocardial infarction Coronary heart disease risk associated with ADT Figure 8 RRs of coronary heart disease related to different types of ADT.

Figure 9
GnRH agonists plus AA compared with other method for coronary heart disease Figure 10 Sudden cardiac death risk associated with ADT Figure 11 Heart failure risk associated with ADT Figure 12 RRs of acute myocardial infarction related to different duration of ADT application.