Background
Prostate cancer (PCa) is the most prevalent malignancy and remains a major healthcare problem in men in the United States [
1]. Because the development and growth of PCa cells depends on androgens [
2,
3], Androgen deprivation therapy (ADT) undoubtedly plays an important role to treat PCa, and recently, approximately 40 % of men diagnosed with PCa within 6 months have been treated with ADT in the US [
4].
ADT is a palliative therapy, including different types of treatments such as gonadotropin-releasing hormone (GnRH), oral antiandrogen (AA), orchiectomy, and two or more types above combined. Although ADT is increasingly used as a treatment for PCa, this effect on prolonging life expectancy is unclear or even negative in several clinical studies [
5,
6]. In our previous study [
7], we found that ADT was positively associated with cardiovascular disease. Because both cardiovascular and cerebrovascular diseases share many common risk factors including atherosclerosis, dyslipidemia, visceral obesity, arterial endothelial dysfunction, and hypertension [
8‐
12], ADT may also be associated with stroke. Additionally, one population-based cohort study [
13] demonstrated that, GnRH agonists could significantly increase the risk of stroke (adjusted rate ratio [RR], 1.18; 95 % confidence interval [CI], 1.00–1.39). However, conflicting results were also reported. In a nation-wide population-based cohort study [
14], authors found that ADT was associated with decreased stroke risk (adjusted hazard ratio [HR], 0.88;
P = 0.001). Therefore, there is still no consensus regarding that ADT is associated with stroke.
Based on the controversy of this clinical issue, we performed a meta-analysis and systematic review to investigate whether ADT is associated with stroke in patients with PCa.
Methods
Search strategy and study selection
We systematically searched Medline, Embase and Cochrane Library databases up to September 30th 2014, with all possible combinations of the keywords as follows:
prostate cancer or
prostate tumor or
prostate carcinoma,
androgen deprivation or
androgen suppression or
endocrine treatment or
ADT or
AST; and
stroke or
cerebrovascular or
transient ischemic attack or
hemiplegia or
TIA or
cardiovascular (Additional file
1: Methods S1). No language, date, or other restrictions was used. Publications from potentially relevant journals were complementally searched.
Studies were included if they fulfilled the following inclusion criteria: 1) Patients diagnosed with PCa only; 2) Intervention groups must include ADT (either monotherapy or combination therapy); 3) Treatments in control groups were non-ADT (e.g. radical prostatectomy, radiotherapy, active surveillance.); 4) Studies must have the data of risk estimates with 95 % CIs; 5) Studies must report comparative data. If more than one study were identified from the same population, we extracted data from all available informations, rather than just a single publication.
Data extraction and quality assessment
Two reviewers (Meng & Zhu) independently extracted the data from eligible and potentially relevant publications, with differences resolved by the third reviewer (Niu) as necessary. General characteristics of each included publication were recorded: first author’s name, year of publication, medical center, study design, sample size, population characteristics, follow-up period, interventions, definition of stroke morbidity, HRs and corresponding 95 % CIs of estimates in each comparisons. Definition of stroke was according to what descripted in each included publication. Our meta-analysis involved different types of ADT including AA, GnRH agonists, orchiectomy, and two or more types above combined.
Study qualities of the selected trials were assessed by the Jadad score [
15]. Trails were considered to be of high quality if they achieved more than 4 scores. Newcastle-Ottawa quality assessment scale (NOS) [
16] was used to assess the observational studies. Studies with more than 6 scores were considered high-quality. Two authors (Zhu & Meng) respectively addressed the assessments and discussed the discrepancies until agreement reached. Level of evidence (LOE) of all eligible publications were evaluated using the classifications of Phillips et al’s, [
17].
Subgroups analyses
In order to minimize the influence of concomitant treatments (e.g. radiotherapy and prostatectomy), subgroup analysis of ADT monotherapy vs watchful waiting or active surveillance (WW/AS) for stroke morbidity was carried out. ADT monotherapy was defined as a single therapeutic that in addition to ADT, no other previous therapy was used in intervention group. Considering the significance of existing heterogeneity in overall-analysis, additional subgroup-analyses for various types of ADT (e.g. GnRH, AA, GnRH + AA and Orchiectomy) vs non-ADT were also performed.
Statistical analysis
Using the same methods as in our previous study [
18], weighted HRs and 95 % CIs were estimate to compare all of these dichotomous variables. Different methods were employed to calculate the HRs on the basis of the data provided in the studies. When studies compared more than one types of ADT with the same control group severally (for example, GnRH vs Control, Orchiectomy vs Control), random effects meta-analyses were used to combine these results together as necessary.
Statistical heterogeneity among studies was evaluated with the Cochrane’s Q statistic [
19]. In addition, inconsistency was quantified by
I2 statistic (100 % × [(Q-df)/Q]), different
I2 values (25, 50, and 75 %) denote different levels (low, medium, and high levels) of heterogeneity [
20]. Using the Der-Simonian and Laird method, we chose random-effects models throughout this analysis no matter whether heterogeneity existed or not.
We used Begg adjusted rank correlation test and Egger linear regression test to evaluate publication bias. All meta-analyses were conducted with Review Manage (version 5.3; The Cochrane Collaboration, Oxford) and STATA software (version 11.0; College Station, Texas). Two-tailed P < 0.05 indicated significant difference statistically.
Discussion
Although the occurrence of stroke in men undergoing ADT with PCa has been an emerging problem over recent years, the relationship between ADT and stroke morbidity is still unclear. This meta-analysis including five population-based observational studies showed that ADT has a tendency to increase the risk of stroke. Evidence was directly proved by Azoulay et al. [
13], showing that ADT could significantly increase the risk of stroke over a median follow-up of 3.9 years in men with newly diagnosed PCa (HR = 1.34,
P = 0.0001). Another cohort study [
24] involving 29,443 ADT users, and 19,527 with surveillance showed the standardized mortality ratios of stroke was 1.17.
ADT is considered to be effective when serum testosterone is declined to the recommended levels of 50 ng/dl, according to the 2012 NCCN (National Comprehensive Cancer Network) guidelines [
26]. However, However, as reported in our previous study [
7], low level of serum testosterone is likely related to many stroke risk factors including high triglyceride and low-density lipoprotein cholesterol levels, endothelial dysfunction and proinflammatory factors [
12,
27‐
29]. In addition, previous studies [
11,
30] showed that testosterone deficiency was significantly associated with hypertension, high body mass index, hypercoagulable states, and hyperfibrinogenemia [
31]. All of these adverse effects may put patients at a high risk of stroke.
Out of the six studies we analyzed, only one [
14] did not show the positive relationship between ADT and stroke (HR = 0.88;
P = 0.001). This inconsistency was likely due to the contamination bias caused by radical prostatectomy. To reduce this bias, a sensitivity analysis was performed comparing ADT monotherapy with WW/AS. When ADT users undergoing other treatments were excluded, more significantly increased risk of stroke was observed in ADT monotherapy users (Fig.
2b).
There may be bias in the results due to different types of ADT that were used in some studies [
13,
23‐
25]. Therefore, we carried out subgroup analyses stratified by different types of ADT in order to reduce this heterogeneity, and showed that stroke morbidity was significantly associated with GnRH alone, GnRH plus AA, and prostatectomy. The US Food and Drug Administration announced a safety warning that GnRH agonists could increase the risk of stroke in men receiving these drugs for treating PCa [
1]. As previously reported [
32], GnRH agonist may cause the development of metabolic syndrome, which in turn could accelerate the atherosclerotic process and then lead to increased stroke morbidity. One included cohort study [
23] investigating the relationship between GnRH and stroke over a median follow-up of 2.6 years, concluded that GnRH was significantly associated with stroke morbidity (adjusted HR = 1.18,
P = 0.03). All of these listed above was in accordance with our findings.
This meta-analysis and systematic review has several strengths. First, the included studies were all large-scale observational studies with long term of follow-up. Second, if the HRs were not available in eligible studies, all the data which could be used to calculate these were adjusted for the durations of follow-up. Finally, funnel plots showed balance in our assessment of publication bias. Begg’s and Egger’s tests also indicated that no significant publication bias existed (Table
2). Additionally, there was no obvious publication bias as shown in Additional file
1: Figure S2, since points are distributed around the verticals. Therefore, the findings in this meta-analysis can be considered credible.
Table 2
Pooled Results and Publication Bias for All Comparisons
Stroke morbidity
| | | | | | | |
ADT vs Non-ADT | 5 | 74538/85947 | <0.001 | 85 | 1.13 (0.95–1.33) | 0.806 | 0.261 |
AA vs Non-ADT | 3 | 5078/47309 | 0.010 | 78 | 1.06 (0.71–1.57) | 1.000 | 0.653 |
GnRH vs Non-ADT | 3 | 49292/47309 | 0.930 | 0 | 1.20 (1.12–1.28) | 1.000 | 0.125 |
GnRH plus AA vs Non-ADT | 3 | 13906/47309 | 0.360 | 3 | 1.23 (1.13–1.34) | 0.296 | 0.501 |
Orchiectomy vs Non-ADT | 4 | 7963/67616 | 0.060 | 59 | 1.37 (1.33–1.64) | 0.734 | 0.456 |
However, we acknowledge that several limitations should be taken into consideration with the results found in this meta-analysis. First, all eligible reports were retrospective observational studies, which may introduce recall limitation, so the integrity of records may weaken the reliability of the results to some extent. Second, selection bias may have influenced our results. To minimize this bias, we carried out a predesigned search strategy with independent selection, and data was extracted by two reviewers. Third, incomplete data in some included publications [
24,
25] may have influenced the overall result. As described in detail in our previous study [
7], we have tried to minimize this limitation as much as possible. Furthermore, the stroke definition (ischemic, hemorrhagic, or TIA) was not specified in some studies [
13,
14,
24], introducing potential bias in stroke incidence estimate. However, most of events in these eligible studies were defined as ischemic events, and this bias is possibly minimized because these overall stroke rates were similar to the study [
23] only including ischemic events as the endpoint. Finally, the certain characteristics of patients that may contribute to stroke were different in each included study, which might confound the presented results. Therefore, adjusted data were extracted when available to minimize the bias.
Competing interests
No competing interests exit in the submission of this manuscript, and manuscript is approved by all authors for publication. All authors have contributed significantly, and are in agreement with the content of the manuscript.
Authors’ contributions
NY had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. MF, ZS and ZJ made substantial contributions to concept, design, and acquisition of data, statistical analysis and interpretation of data. V helped to edit English expression. WL and ZD drafted the manuscript, ZY revised this manuscript critically for important intellectual content and offering a lot of revise opinions. All authors read and approved the final manuscript.