Association of Statin Use with the Risk of Incident Prostate Cancer: A Meta-Analysis and Systematic Review

Background With the growth and aging of population, the incidence of prostate cancer will increase year by year, which is bound to bring greater economic burden to the society. There has been greater interest in the anticancer effects of statin in recent years. It is controversial whether statin use is associated with the risk of prostate cancer (PCa). Thus, we conducted a meta-analysis and systematic review to explore the effects of statin use and their duration and cumulative dose on the overall incidence of PCa. Method The study was conducted according to the latest guidelines for PRISMA 2020. We searched PubMed and other databases for studies about the association of statin use with the risk of incident prostate cancer between January 1, 1990, and April 11, 2022. Two independent researchers extracted data and evaluated the quality of the studies. R x64 4.1.2 and random-effects model were used for data statistics. Relative risk (RR) and odds ratio (OR) effective values with a 95% confidence interval (95% CI) were used to assess the main results. Results The results of 6 RCT and 26 cohort studies showed that statins did not significantly associate with the incidence of PCa (RR = 0.94, 95% CI: 0.82–1.08). The similar results were obtained from 9 case-control studies (OR = 1.03, 95% CI: 0.99–1.07). However, statins were associated with a lower risk of Pca (RR = 0.44, 95% CI: 0.28–0.70) when the cumulative defined daily dose (cDDD) was high. Using statins for more than five years could be associated with a reduced incidence of Pca (RR = 0.47, 95% CI: 0.23–0.97). There was a significant heterogeneity in these studies (RCT and cohort study: I2 = 98%, P < 0.01; case-control study: I2 = 72%, P < 0.01). Conclusion We concluded that statins had a neutral association with the overall risk of PCa. High cDDD and long duration were associated with a lower risk of PCa.


Introduction
Prostate cancer (PCa) is the second most commonly diagnosed cancer and the sixth leading cause of cancer death in men, with an estimated 1.4 million diagnoses worldwide in 2020. With the growth and aging of the population, it is expected that the burden of PCa will grow to 2.3 million new cases and 740,000 deaths by 2040 [1]. Te incidence is rising in many countries, while the mortality rate is declining in developed countries and rising in developing countries [2]. As the most populous country in the world, the incidence and mortality of PCa in China have increased with age from 40 to 94 years old in the past three decades [3]. Without intervention, PCa will bring a heavy burden to society. On the one hand, preventive measures can be taken to decrease the incidence of PCa. On the other hand, we can improve the diagnosis and treatment methods to decrease the mortality of PCa. In order to alleviate the pain of patients after diagnosis, we urgently need to fnd the protective factors of PCA.
Statins are 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) reductase inhibitors, which are comprehensive lipidregulating agents. It can not only efectively decrease total cholesterol (TC) and low-density lipoprotein (LDL) but also decrease triglyceride (TG) and increase high-density lipoprotein (HDL). Te mechanism of statins is that the intracellular cholesterol synthesis is decreased by competitively inhibiting the endogenous cholesterol synthesis rate-limiting enzyme (HMG-CoA reductase) and blocking the intracellular hydroxyvalerate metabolism pathway. Statins have been shown to work through mevalonate-dependent and mevalonateindependent mechanisms [4]. Tere has been greater interest in the anticancer efects of statins in recent years. Several studies have demonstrated an increased incidence of high-risk PCa in patients with elevated TC. Statins exert their anticancer efects by regulating patients' blood lipids [5]. Lee et al. [6] reported that one mechanism by which PCa cells maintain elevated intracellular cholesterol levels is hypermethylation of the ABCA1 promoter and subsequent transcriptional silencing. In addition, a study [7] has shown that the accumulation of cholesterol esters in PCa cells results from the deletion of the tumor suppressor gene PTEN and the activation of the PI3K/AKT pathway. Since cholesterol is the precursor of androgen, cholesterol levels may also afect the development of PCa through androgen signaling pathways [8]. Terefore, statins decrease cholesterol levels through diferent mechanisms and have an anti-PCa efect.
More than 100 articles have reported that statin usage is associated with PCa, including the overall incidence of PCa and the risk of local or advanced PCa and so on. A previous meta-analysis [9] that included 6 randomized clinical trials, 6 cohort studies, and 7 case-control studies published in 2008 showed that statin usage did not decrease the overall risk of PCa. Te other meta-analysis [10] included 15 cohort studies and 12 case-control studies published in 2012 showed that statin usage associated with a lower risk of overall PCa and clinically advanced PCa3. Another metaanalysis [11]that included 6 randomized clinical trials, 7 cohort studies, and 6 case-control studies published in 2017 revealed that neither hydrophilic nor hydrophobic statins were associated with the incidence of PCa. Most recent meta-analysis published in 2021 [12] that included 10 cohort studies and 4 case-control studies indicated that statin usage was associated with a lower risk of advanced PCa.
Tus, it is controversial whether statin usage is associated with the risk of PCa in diferent populations. Furthermore, the previous meta-analysis cannot well claim the efects of statins on prostate cancer that they did not include enough studies or their result analysis was incomplete. Tis metaanalysis and systematic review are a revision and update of all the previous clinical studies, focusing more on and aiming to unravel the efects of diferent types of statins and their duration and cumulative dose on the overall incidence rate of PCa.

Materials and Methods
Tis systematic review and meta-analysis were reported according to the latest guidelines for PRISMA 2020 (Supplementary Material 1).

Inclusion and Exclusion Criteria.
All studies or literature included in the meta-analysis and systematic review strictly met the following PICOS (patients, interventions, comparators, outcomes, and study design) criteria: (1) Patients. Patients who were diagnosed with PCa were included. (2) Interventions. According to reliable records, there was a history of using statins. (3) Comparators. For RCT and cohort studies, the control group was participants who were not treated with statins. For case-control study, the control group was participants who was not diagnosed with PCa. (4) Outcome. In RCTand cohort studies, the participants number of exposed and nonexposed groups and their corresponding number of PCa patients need to be recorded to subsequently obtain relative risk (RR) with 95% confdence intervals (CIs). As for casecontrol studies, the participants number of case group and control group and the number of statin users and non-users in the two groups also need to be recorded, respectively, to get odds ratio (OR) with 95% confdence intervals (CIs). (5) Study Design. It included randomized controlled trials (RCTs) and cohort study and casecontrol study.

Literature Search.
Two independent researchers (M-YX and J-XS) searched PubMed, Embase-Ovid, Cochrane Library, and ClinicalTrias.gov for publications between January 1, 1990, and April 11, 2022. We also searched some grey literature in Google Academic, American Society of Clinical Oncology (ASCO) meeting summary, European Urology Society (EAU) meeting summary, and American Urology Society (AUA) meeting summary. We used keywords to search for "Statin(s)" or "Atorvastatin" or "Cerivastatin" or "Compactin" or "Fluvastatin" or "HMG-CoA" or "Lovastatin" or "Mevastatin" or "Pravastatin" or "Rosuvastatin" or "Rosvastatin" or "Simvastatin" and "PCa" or "Prostate Neoplasms" and so on. Supplementary Material 2 shows the detailed search strategies for each database. After deleting duplicated data by EndNoteX9, the search results were sorted out based on the inclusion and exclusion criteria. We use the PRISMA fowchart to describe the literature search process.

Data Extraction.
Two independent researchers (M-YX and J-XS) extracted data using a data extraction sheet that included basic information and some research factors of interest [13]. Basic information consisted of the author's name, year of publication, country, study design, and patient characteristics. We also recorded the follow-up period, age, body mass index (BMI), cholesterol, race (the percentage of white), the level of PSA, Gleason Score (GS), tumor stage, the defnition of statin use, the number of patients, the number of statin users, the number of patients diagnosed with PCa, type of statins, cumulative duration, and cumulative defned daily dose (cDDD) (the cumulative total amount of DDD over time). In addition, the data about outcomes needed to be extracted, such as the adjusted multivariate relative risk (RR) or odds ratio (OR) with 95% confdence intervals (CIs).

Quality Assessment. Two researchers (M-YX and J-XS)
independently assessed the quality of cohort and casecontrol studies using the Newcastle-Ottawa scale (NOS). Te NOS scale which includes eight items consists of three parts (selection, comparability, and outcome). A score of 7 to 9 is defned as high quality, and a score less than 7 is defned as low quality. We adapted RoB2 Excel to evaluate the quality of RCTs. RoB2 set up fve evaluation domains: the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Tere are many diferent signal problems in each domain. Researchers need to make judgments and answer these questions objectively when evaluating the bias risk of an RCT. Tere are generally fve answers to signal questions: yes (Y), probably yes (PY), probably no (PN), no (N), and no information (NI). According to the reviewers' answers to the signal questions, the bias risk in each domain can be divided into three levels: "low risk," "certain risk," and "high risk." Any diferences are resolved by re-evaluating the original article with the third researcher.

Statistical Analysis.
In RCT and cohort studies, the estimated RR efect values with 95% CI were used to assess the potential association between statin usage and PCa risk, whereas the OR efect values were used in case-control studies. We used Cochrane's Q and I 2 statistics to test the heterogeneity. Te random-efects model should be adopted due to expected heterogeneity among diferent study populations. In order to verify the publication bias of studies, we conducted Begg's and Egger's tests [14]. We conducted subgroup analyses of countries and statins type to fnd potential sources of heterogeneity. Te diference between studies may be related to age, BMI, race (the percentage of white) (the percentage of white) and cDDD (the percentage of statin users whose cDDD >600), so we conducted metaregression analysis to test this hypothesis. In addition, we performed sensitivity analysis and cumulative meta-analysis to estimate the stability of our meta-analysis by gradually omitting or adding included studies.
We used the "meta" package in R x64 4.1.2 to perform meta-analysis [15,16]. All the statistical tests were bilateral, and the P value <0.05 indicated statistical signifcance.

Search Results and Study Characteristics.
We have retrieved 1,250 publications from electronic databases and grey literature. We remained 731 after removing repetitive literature. After a preliminary reading of the title and abstract, we retained 110 articles. After screening according to strict criteria, 57 articles were excluded. Ten, we read the full text of each record. After excluding 12 publications due to a lack of data about the incidence of PCa, 41 studies were fnally included in this meta-analysis and systematic review ( Figure 1).
Te characteristics of all studies included are shown in Supplementary Material 3. Te literature included in this study was published from 1998 to 2022. Tis study included 6 randomized controlled trials [17][18][19][20][21][22], 26 cohort studies , and 9 case-control studies [49][50][51][52][53][54][55][56][57] which included a total of 3,697,172 participants. Tere was a total of 12 countries (Israel, Scotland, Australia, Canada, China Taiwan, Denmark, Finland, Japan, the Netherlands, Sweden, UK, and USA) in this meta-analysis. Te top three countries with the most studies were the United States, Finland, and the United Kingdom, with 20, 5, and 3 studies, respectively. Participants used hydrophilic and hydrophobic statins, including fuvastatin, pravastatin, rosuvastatin, atorvastatin, cerivastatin, simvastatin, and lovastatin. It should be noted that "statin usage" was defned diferently in diferent studies. Some studies had shown that participants could be regarded as statin users as long as they conducted a selfreport or kept a record of drug purchases to prove that they had used statins. In other studies, researchers defned statin users as participants who used statins with a clear type, dose, and time.

Quality Assessment of Included Studies.
As for the quality of the included studies, more details are presented in Tables 1  and 2. NOS was used to access the quality of 36 observational studies, and the scores of the included studies ranged from 5 to 9. Te bias risk map and bias risk summary map (Supplementary Material 4) in ROB2 excel vividly demonstrated the literature quality of 6 RCT studies [58]. Except for some concerns about the selection of the reported result and some high-risk aspects of the measurement of the outcome, the other three domains were low risk. Furthermore, overall bias was low risk. However, among the included studies, the casecontrol study had the lowest evidence grade [59].

Subgroup Analyses and the Meta-Regression.
In order to fnd the source of heterogeneity, we conducted a subgroup analysis ( Figure 3). In the subgroup analysis of the study design, the P value for interaction between the two types was 0.18, indicating that the study design did not contribute to the heterogeneity of the literature whereas diferent countries were one of the sources of heterogeneity because the P value for interaction in the subgroup analysis of countries was <0.01. Studies conducted in Denmark (RR � 0.65, 95% CI: 0.46-0.92) and Israel (RR � 0.64, 95% CI: 0.59-0.70) had shown that statins could be associated with decreasing the incidence of PCa. Te studies of Scotland (RR � 1.51, 95% CI: 1.09-2.09) and Sweden (RR � 1.62, 95% CI: 1.52-1.73) showed that statins were associated with a higher incidence of PCa. Given that P value for interaction in the subgroup analysis of statin type was 0.13, there was no signifcant diference in the incidence of PCa among diferent statins.
In addition, we used meta-regression analysis for some continuous variables to further fnd the covariates. Te meta-regression models of publication year, followup time, age, BMI, population percentage of BMI <30 kg/ m 2 , race (the percentage of white), PSA level, GS, and population percentage of cDDD >600 were constructed. For RCT and cohort studies, we found that race (the percentage of whites) (P � 0.0430) signifcantly afected the efects of statins on the incidence rate of PCa (Figure 4). Tere was no signifcant relationship between the incidence of PCa and year (P � 0.9890), follow-up period (P � 0.8267), age (P � 0.8852), BMI (P � 0.9106), and cDDD <600 (P � 0.8718) (Supplementary Material 5). For case-control studies, there was no relationship between the incidence of Pca and the incidence of the year (P � 0.8254) and follow-up period (P � 0.8254).

Efect of cDDD and Duration of Statins on
PCa. 4 studies recorded the cDDD for statins of each group of participants in detail [32,38,40,48]. And our results showed that statins were associated with a lower risk of PCa (RR � 0.44, 95% CI: 0.28-0.70) when the cumulative defned daily dose (cDDD) was high (cDDD >600) (Figure 5(a)). Te results of 3 cohort studies [27,45,48] that recorded data from participants who had used statins for long periods showed that using them for more than fve years was associated with a decreased incidence of PCa (RR � 0.47, 95% CI: 0.23-0.97) ( Figure 5(b)).

Sensitivity Analysis and Publication Bias for Included
Studies. In order to test the stability of our results, we conducted a sensitivity analysis ( Figure 6)  Records excluded:       rafts and further changes in cell signal transduction by reducing the content of cholesterol. Finally, the consumption of cholesterol leads to the loss of membrane integrity [65]. In this meta-analysis, we included a total of 41 studies on statin usage and the incidence of PCa, and we used RR and OR as efect values. RR was mainly used in randomized controlled trials or prospective cohort studies to study the efect of statin exposure on the outcome of PCa by designing follow-up studies [66]. Te OR value was mainly used in retrospective case-control studies to study the correlation between the case group (PCa patients) and the control group (non-PCa population).
6 RCT studies did not show a signifcant association between statin use and PCa. Tese 6 studies explored the relationship between statin use, all-cause mortality, and specifc cancer incidence in hypercholesterolemia patients. Cardiovascular events were the main outcome, so there were basic diseases in patients. 10 of 26 cohort studies reported that statins were associated with a lower incidence of PCa, which was in line with our expectations. Only one of nine case-control studies indicated that statins associated with a lower incidence rate of PCa [56]. In addition, a previous meta-analysis showed that statins decreased overall PCa risk [10]. However, the results of two other meta-analyses [9,11] did not support the hypothesis that statins are associated with the overall risk of PCa, which was consistent with our conclusion.
Many of these studies relied on questionnaires to obtain exposure data and lacked screening adjustments [55]. Nevertheless, the defnition of statin exposure varies across studies. Some studies defne exposure as the use of statins before PCa diagnosis, which is the most appropriate defnition, and some defne it as having ever used statins, while others defne it as the use of certain statins within a specifed period of time. If the defnition of exposure includes the condition that patients still use statins after diagnosis with PCa, it may undermine the hypothesis that statins are benefcial to reduce the incidence of PCa. In addition, among these three types of studies, the level of evidence in case-control studies was the lowest [59]. Te real results should be more biased towards the combined results of RCTs and cohort studies.
We identifed documented cDDD in three cohort studies, and we found that high cDDD (cDDD >600) is signifcantly associated with a reduced risk of PCa. However, cDDD was the efect value accumulated by the defned daily dose over time, so it was difcult to identify whether long-    Journal of Oncology term duration or a large dose led to the reduction of PCa risk. Ten, to explore the impact of duration on the risk of PCa, we recorded and analyzed the specifc duration of each group. In 3 cohort studies and 4 case-control studies, some participants used statins for more than 5 years. By analyzing the total RR efect value of these cohort studies and the total OR efect value of these case-control studies, we found that when the duration was more than 5 years, the overall risk of PCa in cohort studies decreased signifcantly, and the overall risk of PCa in case-control studies decreased but not statistically signifcant. Long-term use of statins had a more stable efect on the anticancer efect in the body. It shows high heterogeneity in the overall analysis. In some subgroups, heterogeneity was decreased, indicating that research types and countries were important infuencing factors. Researchers could dictate that participants use a certain type of statins in an RCT. However, in observational studies, the type of statins used by participants could not be intervened. In addition, the defnition of statin use was diferent in diferent studies. Tus, statins type was also one of the sources of heterogeneity.
Meta-regression showed that the proportion of whites was positively correlated with the risk of PCa. On the one hand, this was due to the diferent genetic susceptibilities of diferent races. On the other hand, Caucasians were more likely to be diagnosed with PCa because of diferent levels of medical care in diferent countries.
Concerning confounding factors, we put forward some thoughts. Tere was a direct correlation between the malignant degree of PCa and PSA [67]. Te higher the malignant degree of the tumor, the higher the serum level of prostaglandin-specifc antigen. PSA screening was associated with overdiagnosis of low-risk PCa [47]. Some studies also indicated that serum PSA levels decreased signifcantly among statin users [68][69][70][71][72]. However, taking aspirin and/or antidiabetic drugs at the same time may change the anticancer efect of statins [27,41]. Moreover, we speculate that diferent education levels, income levels, and social status may also be confounding factors for the incidence of PCa.
For the prevention of PCa, we not only pay attention to statins but it is more important that residents maintain good living habits. Te goal of the World Cancer Research Fund/ American Institute for Cancer Research (WCRF/AICR) Report is to improve health and reduce the global cancer burden through diet, nutrition, and physical activity [73], which is consistent with certain of the purposes of our study. Te Cohort of Swedish Men (COSM) revealed that compliance with WCRF/AICR recommendations signifcantly reduced the overall cancer risk [74]. Te treatment of cancer is becoming more complex, and diet, nutrition, and physical activity play an important role as auxiliary means.
Compared with previous meta-analysis, ours included more qualifed literature, and the sample size was greatly increased, making the conclusion more universal. Moreover, we not only explored the efects of statin usage on the overall risk of PCa but also further explored the association between cDDD, the duration of statins, and the risk of PCa.
Some limitations should be noted. First, some data were incomplete in diferent studies. For example, the follow-up period was not recorded in 3 studies, 16 studies lacked age data, 32 studies lacked BMI data, 25 studies lacked race composition data, 33 studies did not record participants' PSA baseline value. And the type of statins in 16 studies was unknown. Second, due to the inclusion of too many studies, it was difcult to unify each study's design methods and the defnition of statin exposure, so that the heterogeneity between studies was

Conclusion
In summary, we conducted a meta-analysis to explore the efects of statin use on the overall incidence of PCa. We found that statins had a neutral efect on the overall risk of PCa, and high cDDD and long duration were associated with  a lower risk of PCa, which may contribute to preventing PCa and decreasing its incidence.

Data Availability
All data generated or analyzed during this study are included in this published article and referenced articles are listed in the References section.

Conflicts of Interest
Te authors declare that they have no conficts of interest.