What's known on the subject? and What does the study add?
Over the last few years, several observational studies examined the association of statin use with the risk of biochemical recurrence of prostate cancer after definitive local therapy. The objective of our present study was to summarise available evidence on this subject using the method of meta-analysis.
Combined evidence from eight cohort studies did not definitively support the hypothesis that statins influence the risk of biochemical recurrence. However, there was considerable disagreement between individual studies in reported findings and conclusions.
• To perform a systematic review and meta-analysis of clinical studies with statin use as the exposure variable and biochemical recurrence after definitive local therapy for prostate cancer as the outcome.
• Relevant publications were identified through PubMed/Medline/Embase databases.
• Pooled estimates of the hazard ratios (HRs) were computed using the inverse-variance weighting approach.
• Heterogeneity was assessed using the Cochran's Q test.
• We identified a total of eight eligible studies, all based on the retrospective cohort design. Five of these were based on radical prostatectomy (RP) series and three on radiotherapy (RT) series.
• There was evidence of heterogeneity in the entire set of eight studies (P= 0.002) as well as in the RP series (P= 0.05) and in the RT series (P= 0.01), when these were considered separately.
• Based on the random effects inverse-variance weighting approach, pooled estimates of the HRs for the risk of biochemical recurrence in statin users v non-users were 0.91 (95% confidence interval [CI] 0.72–1.13) for the entire set of eight studies, 1.02 (95% CI 0.80–1.29) for the RP series and 0.71 (95% CI 0.44–1.16) for the RT series.
• The pooled estimates of the HRs were not significantly different from the null value in this meta-analysis; however, evidence of heterogeneity between the studies was present.
In the USA, prostate cancer is the second most commonly diagnosed malignancy in men (after skin cancer) and is the second most common cause of cancer death in men (after lung cancer), with 240 000 new cases and 33 700 deaths estimated for year 2011 . Over the last 15 years, numerous in vitro and animal studies suggested that development of prostate cancer and possibly the rate of prostate cancer progression may be affected by 3-hydroxy-3-methylglutaryl coenzyme A (HMG-coA) reductase inhibitors, commonly known as statins [2,3]. Statins are currently widely used for management of hyperlipidaemia, and their value for prevention of various cardiovascular events has been consistently shown in numerous randomised trials . The association of statins with prostate cancer incidence as a secondary end-point was also examined in six randomised trials. A meta-analysis of these trials found no significant association between statin use and prostate cancer incidence (relative risk 1.06, 95% CI 0.93–1.20) . When trials of lipophilic statins were considered separately, the relative risk of prostate cancer among statin users vs non-users was 0.99 (95% CI 0.85–1.16), while for the lipophobic statins it was 1.21 (95% CI 0.83–1.78) . Observational studies of statins and prostate cancer incidence were also included in the meta-analysis. The pooled relative risk based on six cohort studies was 0.98 (95% CI 0.89–1.09), while the pooled odds ratio based on seven case-control studies was 0.74 (95% CI 0.46–1.20) .
After publication of this meta-analysis, several additional observational studies of statins and prostate cancer incidence were published. Some of these reported lower risk of prostate cancer in statin users than in non-users [6–10], while others reported an elevated risk [11,12]. However, it can be argued that the overall incidence of prostate cancer is not the most meaningful end-point from a clinical perspective because most prostate cancers are not life-threatening and many would remain undiagnosed during the patients' lifetime if not detected by screening. Hence, instead of the overall incidence of prostate cancer, it may be of interest to examine the incidence of advanced prostate cancer in statin users vs non-users.
Unfortunately, advanced prostate cancer was not a documented end-point in the randomised trials of statins; hence Level 1 evidence for this end-point is lacking. In a meta-analysis of three cohort studies, statin users had a lower incidence of regionally advanced or metastatic prostate cancer than non-users (relative risk 0.77, 95% CI 0.64–0.93) . It should be noted that in these three studies, the point estimates were of similar magnitude; hence the findings were consistent. However, in another cohort study (published after publication of the meta-analysis), advanced prostate cancer was diagnosed more frequently in statin users compared with non-users, based on the point estimate, although this association was not statistically significant (relative risk 1.22, 95% CI 0.85–1.75) . The association of statin use with advanced prostate cancer was also examined in case-control studies. In the study published by Coogan et al., statin use was less common in men with advanced prostate cancer than in the controls, although this difference was not statistically significant. Murtola et al.  found that among men with advanced prostate cancer, the proportion of statin-users was significantly less than in the control group (odds ratio 0.75, 95% CI 0.62–0.91), suggesting that statins may have a protective effect against advanced prostate cancer. Similarly, in a recent case-control study, Marcella et al.  found that the proportion of statin users among men who died from prostate cancer was significantly less than in the control group (odds ratio 0.37, 95% CI 0.23–0.60).
Several observational studies also examined the association of statin use with biochemical failure after definitive local therapy for prostate cancer, although a systematic review and meta-analysis of these studies has not yet been performed. Biochemical recurrence and progression of prostate cancer after definitive local therapy as an endpoint is conceptually different from the endpoint of prostate cancer diagnosis, which was the focus of the previous meta-analyses. The objective of the present study was to perform a systematic review and meta-analysis of observational studies with statin use as the exposure variable and biochemical failure after definitive local therapy for prostate cancer as the outcome.
Studies eligible for this meta-analysis were identified through the PubMed/Medline/Embase databases, using various combinations of the search keywords ‘statins’, ‘HMG-coA reductase inhibitors’, ‘prostate cancer’, ‘biochemical’, ‘recurrence’, ‘progression’, ‘failure’, ‘outcomes’, and generic names of individual statin drugs. We also examined the lists of references in recent review articles and original articles on this subject. The literature search was performed to 1 May 2012, independently by two investigators. To be included in this meta-analysis, a study had to report the association of statin use with biochemical recurrence of prostate cancer after definitive local therapy, e.g. radical prostatectomy (RP) or radiotherapy (RT). The following information was extracted from each eligible publication: type of definitive local therapy, number of subjects in each comparison group (statins vs no statins), specific types of statins used, baseline variables (e.g. age, pre-treatment PSA level, stage, and Gleason sum), definition of biochemical failure, average follow-up, and the hazard ratio (HR) for biochemical failure in statin users vs non-users (with 95% CIs and P-values).
The meta-analysis of the HRs was performed using the standard approach, described for example in Marubini and Valsecci 1995. The pooled estimate of the HR was computed using the method of inverse-variance weighting. Heterogeneity was assessed using the Cochran's Q test. A significant Q test indicates that variability in the estimates reported from different studies is greater than expected under the assumption that the studies estimate the same treatment effect. In such cases, it is often recommended that the random effects meta-analysis, estimating the average of the study-specific treatment effects, be preferred to the fixed-effects inference. For sensitivity analysis, we computed the point estimates, CIs, and P-values for the pooled HRs using both fixed and random effects weights. Publication bias was assessed with the Funnel Plot and the Egger's test . The Funnel Plot was constructed by plotting the study-specific estimates of the log-HR against their estimated standard error. Selective publication of the studies with larger estimates of the HR would be expected to produce asymmetry in the plot, particularly at the top, where the studies with larger estimated standard error are located. Statistical significance of asymmetry can be assessed with the Egger's test, with small P-values (e.g. <0.1) interpreted as evidence of asymmetry. All reported P-values are two-sided.
We identified 12 relevant publications based on 11 studies [18–29]; all with retrospective cohort design. However, one of these studies reported neither crude nor adjusted HRs for biochemical recurrence in statin users vs non-users; only an unadjusted P-value was reported from the Cox model for this comparison [18,19]. Hence, that study had to be excluded from the current meta-analysis. In addition, two studies were published only in abstract form, without sufficient details necessary for inclusion in the current meta-analysis [20,21]. Hence, these studies were also excluded, leaving eight eligible studies, all published recently (Table 1) [22–29]. We did not attempt to contact the authors of the three excluded studies for additional information because our aim was to analyse only peer-reviewed findings. Among the eight included studies, five were based on RP series and three on RT series (external beam and/or brachytherapy). In the RP studies, biochemical failure was defined according to the 0.2 ng/mL threshold, while the RT studies used the Phoenix definition of PSA nadir + 2 ng/mL.
Table 1. Local therapy, number of study participants, and most commonly used statins
The total numbers of subjects included in each study are shown in Table 1. When all eight studies were considered together, there were a total of 2812 statin users and 10 031 statin non-users. It should be noted that in Table 1 and in all subsequent tables in this paper, subjects who started statins after completion of definitive local therapy for prostate cancer were classified as statin non-users, with the exception of the Gutt et al.  RT study where these subjects (39) were included in the ‘statin users’ group along with 150 other subjects who used statins before completion of their RT. Hence, the overwhelming majority (99%) of subjects classified as ‘statin users’ in this meta-analysis were using statins at the time of definitive local treatment for prostate cancer, while subjects classified as ‘statin non-users’ included those who either never used statins or started statins after completion of definitive local treatment for prostate cancer. Unfortunately, it was not possible to determine what proportion of the 10 031 subjects classified as ‘statin non-users’ actually used statins after RP or RT. It should also be noted, that specific types of statins used were documented in only three of the eight studies (Table 1).
Baseline characteristics of the study subjects, by statin use (as defined in the previous paragraph) are shown in Table 2. We attempted to summarise the baseline characteristics in a way that would facilitate their direct comparison between the studies. However, it should be noted that for age and PSA variables, some authors reported only the means, while others reported only the medians or percentages. We used clinical stages in Table 2 for staging summary in both RP and RT series because distribution of pathological stages in the RP series was not always clearly described. In particular, percentages of patients with ‘extracapsular extension’ and ‘seminal vesicle invasion’ were usually reported separately, without clarifying whether the ‘extracapsular extension’ category included seminal vesicle invasion as a special case or referred only to stage T3a (in other words, it was not always clear whether the reported percentages were additive or over-lapping). On the other hand, in the RP series, Gleason sum was reported more consistently based on the RP specimens than preoperative biopsies. Hence, in Table 2, pathological Gleason sum was reported for the RP series, while the clinical Gleason sum was reported for the RT series. The threshold of 6 for the Gleason sum was chosen for consistency, as it could be applied to data reported from each study.
Table 2. Baseline characteristics of the study participants
Note from Table 2 that in each study, statin users were either older than the non-users or were of about the same age. However, the pre-treatment PSA concentration tended to be lower in statin users compared with the non-users. The distribution of clinical stages was either similar in the two comparison groups or more favourable in the statin group. In contrast, the distribution of the Gleason sum was less favourable in statin users compared with the non-users in most studies.
HRs for biochemical recurrence reported from the eight studies are summarised in Table 3 and Fig. 1. In the five studies based on the RP series, the risk of biochemical recurrence among statin users compared with non-users was significantly decreased in one study (Hamilton et al., 2010 ), significantly increased in one study (Ritch et al. 2011 ) and not significantly different from the null value in the other three studies (Table 3). The Q test for heterogeneity in these five studies was marginally significant at the conventional level (P= 0.05, Table 3). Both fixed and random effects estimates of the pooled HRs were close to the null value and not close to being statistically significant (Table 4).
Table 3. HRs for biochemical recurrence, with 95% CIs and P-values
HR (95% CI)
HR, HR for biochemical recurrence rate in statin users vs non-users. Adjusted for: 1, age; 2, pre-treatment PSA level; 3, pathological stage; 4, pathological Gleason score; 5, margins; 6, race; 7, chronological year of definitive local treatment; 8, biopsy Gleason score; 9, clinical stage; 10, body mass index; 11, percentage of biopsy cores containing cancer; 12, study centre; 13, American Society of Anesthesiologists (ASA) score; 14, comorbidities; 15, prostate volume; 16, percentage of tumour by volume; 17, radiation dose; 18, perioperative hormonal therapy; 19, National Comprehensive Cancer Network (NCCN) risk group; 20, smoking; 21, prostate cancer family history; 22, aspirin use; 23, angiotensin-converting enzyme (ACE) inhibitor use. *Follow-up time for statin users and non-users, respectively.
In the three studies based on the RT series, the risk of biochemical recurrence among statin users compared with non-users was significantly decreased in two studies and not significantly different from the null value in one study (Table 3). The Q test for heterogeneity in these three studies was statistically significant at the conventional level (P= 0.01, Table 4). The pooled HR based on these studies was significantly different from the null value in the fixed but not in the random effects meta-analysis (Table 4).
When all eight studies were considered together, the pooled HR was not significantly different from the null value in the fixed or in the random effects analyses (Table 4). Note that there was substantial evidence of heterogeneity based on the point estimates of the HRs reported from individual studies (Table 3) and the highly significant Q test (P= 0.002, Table 4). The Funnel Plot, which is used for assessment of publication bias, is shown in Fig. 2. There was no significant evidence of asymmetry in the plot (Egger's P= 0.62).
Evidence from basic research suggests that statins may influence the rate of prostate cancer recurrence by several mechanisms involving the synthesis of cholesterol and isoprenoids. Statins decrease the rate of cholesterol production by inhibiting HMG-coA reductase, an enzyme that catalyses the conversion of HMG-coA to mevalonate in the cholesterol biosynthesis pathway. Because cholesterol is a precursor of androgens, lowering the rate of its production may reduce the levels of circulating testosterone and the intra-prostatic concentration of dihydrotestosterone, which is a strong ligand for the androgen receptor [2,3]. Cholesterol is also a major component in lipid rafts, which are involved in various signalling pathways (e.g. PI3K/Act and epidermal growth factor receptor) regulating prostate cancer cell survival and proliferation [2,3]. In addition to cholesterol-mediated effects, statins may influence the rate of prostate cancer progression by inhibiting the synthesis of isoprenoids, such as geranylgeranyl pyrophosphate and farnesyl pyrophosphate, thus interfering with signalling mediated by certain proteins in the Ras super-family. Most consequences of this interference seem to be antiproliferative in nature and would be expected to reduce prostate cancer aggressiveness [2,3]. However, it has been noted that pro-neoplastic effects of statins on prostate cancer are also conceivable . Because in vitro and animal studies cannot definitively determine whether the net effect of statins on prostate cancer in humans is pro- or anti-neoplastic, evidence from epidemiological research must be examined to address this question.
The objective of our present study was to perform a systematic review and meta-analysis of observational studies with statin use as the exposure variable and biochemical recurrence after definitive local therapy for prostate cancer as the outcome. We found that published observational studies reported conflicting results about the association of statin use with biochemical recurrence. The pooled estimates of the HRs were not significantly different from the null value in this meta-analysis; however, substantial evidence of heterogeneity between the studies was present. The strengths of this meta-analysis include a large number of subjects treated at different institutions, with the same PSA thresholds for definition of biochemical failure for a given local treatment method (RP or RT), and with reasonably long follow-up for analysis of biochemical failure rates in most studies. However, the following limitations of the present analyses must be recognised.
First of all, it should be noted that all studies included in our meta-analysis were observational in design. Hence, in each study, there was a possibility of residual confounding, which could either attenuate or exaggerate the observed association of statin use with recurrence relative to the true causal effect. Substantial variability between the studies in the estimated HRs (Table 3) suggests that the studies were not estimating the same treatment effect, which may be due to differences in the magnitude of uncontrolled confounding, or other factors. Although all studies with a given local treatment method performed adjustment for similar sets of baseline covariates (Table 3), differences in reported HRs could also be influenced by variation between the studies in the types of statins used, the dose and the duration of use and the follow-up protocols. Because these studies were retrospective in design, the follow-up often was not standardised and specific types and doses of statins and duration of their use usually were not documented (Table 1). In the Hamilton et al.  study, there was a dose-response relationship between the risk of biochemical recurrence and the statin dose equivalent, but not the duration of use. In the Gutt et al.  study, the risk of biochemical recurrence did not differ significantly according to the dose of statins used. In the Mondul et al.  study, subjects who used statins for at least 1 year (but not those who used statins for <1 year) had a lower risk of recurrence than the non-users based on the point estimates; however, these estimates were extremely imprecise based on CIs and not significantly different from the null value. Because most studies included in the present meta-analysis did not incorporate information on statin types, doses, and duration of use, the impact of these factors on the risk of biochemical recurrence of prostate cancer remains unclear.
Another issue that needs to be considered in the context of this meta-analysis is the possibility of publication bias. It is generally thought that the ‘positive’ studies (i.e. those reporting statistically significant findings) are more likely to be published than the negative studies. The possibility of publication bias is sometimes investigated using ‘funnel plots’ and associated tests of significance [31,32]. Application of this methodology to the present data did not reveal clear evidence of publication bias; however, these methods are most reliable when the number of studies is large, and even in that case, factors other than the publication bias can significantly influence the interpretation (e.g. the likelihood of publication can be influenced by the study quality in addition to statistical significance of reported findings, etc.) [31,32].To investigate the possibility of publication bias further, we attempted to find all studies on statin use and prostate cancer recurrence that were presented in an abstract form but were not published in a peer-reviewed journal. Two such studies were identified, but the corresponding abstracts did not contain information necessary for inclusion of these studies in the meta-analysis. However, we could determine from the abstracts that the association of statin use with prostate cancer recurrence was not statistically significant in both studies [20,21], although one of the studies had a mean follow-up of only 16 months in both comparison groups . In addition, one study which resulted in two journal publications also had to be excluded from the meta-analysis because the HR for the association of statin use with biochemical recurrence was not reported (but was stated to be ‘not statistically significant’) [18,19]. We did not attempt to contact the authors of the three excluded studies for additional information because our aim was to analyse only peer-reviewed findings. In general, the quality of information obtained through personal communication is very difficult to assess; hence, our meta-analysis was based only on published data.
In addition to these considerations, it must also be noted that in general, biochemical recurrence, which was the focus of the present meta-analysis is used as an end-point in prostate cancer studies because it is a marker of treatment failure, indicating disease progression. However, in many patients, radiographically detectable or symptomatic metastases appear many years after biochemical recurrence . Hence, the influence of statins on the rate of biochemical recurrence may differ in magnitude from their influence on the rate of progression to radiographically or clinically apparent metastatic disease. This is possible, given that statins have been associated with lower PSA concentrations not only at initial prostate cancer diagnosis  (Table 2), but also in the general male population [35–37]. Most studies included in the present meta-analysis did not examine progression to radiographically or clinically apparent metastases or prostate cancer-specific mortality in statin users vs non-users as separate end-points. However, Gutt et al.  and Kollmeier et al.  reported that they found no differences in metastases-free, prostate cancer-specific, or over-all survival in statin users vs non-users, and Mondul et al. , reported a HR of 0.85 (95% CI 0.38–1.92) for the end-point of progression to metastases or death. Of course, analysis of metastases-free or disease-specific survival of patients with prostate cancer requires much longer follow-up than analysis of biochemical recurrence.
Interestingly, the association of statin use with prostate cancer mortality, symptoms, PSA response, and quality of life has been recently examined in patients who participated in the docetaxel trial . Unlike patients included in the present meta-analyses, all patients in the docetaxel study already had castration-resistant prostate cancer. Among these patients, statin use was not associated with any of the disease-specific end-points, or overall survival . Of course, participants of this study were not randomly assigned to statins vs control, which is a limitation. According to another recent report, based on a cohort of patients with prostate cancer managed with watchful waiting, the mean PSA velocity did not differ significantly between statin users and non-users (P= 0.25) . In another observational study, Katz et al.  examined a cohort of patients with prostate cancer treated with RP or RT and reported lower all-cause mortality in statin users vs non-users, after adjusting for age, certain comorbidities, prostate cancer characteristics, and smoking.
Given available evidence from observational studies, including the present meta-analysis, would it be feasible to examine the association of statins with prostate cancer recurrence and progression in randomised trials? Randomising patients to statins vs placebo at the time of definitive local therapy for prostate cancer and following them for recurrence and progression would be very challenging because this study would require many years of follow-up with long-term exposure to statins among men who do not have cardiovascular indications for statin therapy. Furthermore, during the follow-up time, many patients in the control group would inevitably develop cardiovascular conditions for which statins are indicated, resulting in substantial contamination of the control arm. Perhaps a more feasible approach would be to enrol patients who are about to start hormonal therapy for biochemical progression after definitive local treatment of prostate cancer and randomise them to statins vs control, with follow-up for appearance of radiographic evidence of progression, prostate cancer-specific survival, and all-cause survival. Because there is evidence of adverse effects of hormonal therapy on lipid metabolism with increased risk of cardiovascular disease , the potential benefits of statin therapy in this setting may not be limited to prostate cancer-specific outcomes and may include reduction in the incidence of fatal and non-fatal cardiovascular events.
In conclusion, published observational studies reported conflicting results about the association of statin use with biochemical recurrence. The pooled estimates of the HRs were not significantly different from the null value in this meta-analysis; however, substantial evidence of heterogeneity between the studies was present. Considering limitations of observational studies, it may be appropriate to investigate the effect of statins on prostate cancer recurrence and progression in randomised trials.
Funding for this study was provided by the Ashley Family Foundation.