Background
Statins, 3-hydroxy-3-methyl-glutaryl coenzyme A reductase (HMGCR) inhibitors, reduce low-density lipoprotein cholesterol (LDL-c) significantly, resulting in a commeasurable reduction in morbidity and mortality from cardiovascular disease (CVD) [
1,
2]. Beyond their effectiveness as a cardiovascular intervention via lipid modification, pleiotropic effects of statins have long been suggested [
3‐
5]. Cardiovascular-related pleiotropic effects of statins may include beneficial effects via a range of potentially inter-related factors, including inflammatory responses [
6,
7], endothelial function [
5], possibly prothrombin time, and sex hormones [
8]. More recently, Mendelian randomization (MR) studies have indicated that statins may reduce the risk of cancer by a lipid-independent pathway [
9] as well as specifically reducing epithelial ovarian cancer [
10]. Taken together, these studies highlight potential sex differences in the mechanisms underlying statins’ protective effects on CVD and overall mortality. However, exactly what mechanisms underlie statins’ complex sex-specific effects and how they might affect the positioning of statins in disease prevention and treatment remains unclear. Randomized trials are rarely designed or powered to investigate mechanisms or pleiotropic effects, although effects of statins on body weight and diabetes have been identified from a meta-analysis of trials [
11,
12].
To identify statins’ pleiotropic effects, previous studies have used genetically mimicked statins to assess statins’ metabolomic profile [
13], compared it with that of proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors [
14], and compared statin’s lipoprotein signature with that of cholesteryl ester transfer protein (CETP) inhibitors [
15]. However, none of these studies has been comprehensive across the phenotype or sex-specific, when differences by sex are evident for the incidence of CVD [
16] and some cancers [
17,
18], highlighting the possibility of sex-specific pathways and sex-specific impacts. To fill this gap, we conducted a phenome-wide association study (PheWAS), a summary statistics-based [
19] genotype-to-phenotype approach [
20,
21], to assess systematically the sex-specific associations of genetically mimicking statins with a wide range of conditions and related phenotypes, using the largest available genome-wide association studies (GWAS), with validation where possible. The use of genetic mimics largely avoids confounding owing to the random allocation of alleles at conception [
22]. To assess if any pleiotropic effects identified from the PheWAS were unique to statins, we also assessed whether these pleiotropic effects were evident for (a) genetically mimicking the major lipid modifiers in common use, i.e., PCSK9 inhibitors and ezetimibe, and (b) LDL-c.
Discussion
Consistent with previous findings about the effects of statins [
13], the rs12916 T allele was associated with lower LDL-c, total cholesterol, and ApoB; higher HbA1c [
33]; and greater adiposity [
12,
34,
35], particularly higher BMI, with little difference by sex. Our study adds by assessing a wider range of phenotypes sex-specifically. We found the rs12916 T allele was associated with higher BMR, and with platelet attributes in both sexes and with lower SHGB and calcium particularly in women, which did not extend to genetically mimicking the major lipid modifiers in current use and largely were not a consequence of LDL-c.
Few RCTs have assessed the effects of statins on BMR, SHBG, or serum calcium, but statins reducing SHBG in women have been previously reported [
36,
37]. Previous MR studies have suggested that calcium increases the risk of ischemic heart disease (IHD) [
38‐
40] and SHBG reduces it [
40], so these mechanisms together might have a relatively neutral effect on IHD in women. Sex-specific effects of calcium and SHBG on IHD have not been fully assessed, so how these effects would affect specifically women is unknown, although broadly statins appear to have the same effects on IHD in men and women after accounting for testosterone [
8]. SHBG inactivates sex hormones, so lowering SHBG might increase the availability of sex hormones and increase the risk of any related conditions. A recent MR study showed that lower SHBG increases the risk of estrogen-positive breast cancer [
41]. Effects of SHBG on other cancers in women have not been systematically examined. Concerns were raised about the possibility of statins increasing the risk of breast cancer in women more than a quarter of century ago [
42], but are not feasible to investigate in trials. However, a recent MR study did not suggest that genetically mimicking statins increases breast cancer risk [
9]. An MR study has suggested that higher BMR increases the risk of colorectal cancer [
43]. Recent MR studies have shown that genetically mimicking statins reduces cancer overall, but did not provide sex-specific estimates [
9]. Correspondingly, genetically mimicking effects of statins reduced epithelial ovarian cancer [
10]. The role of lowering calcium in cancer is unclear and has not been extensively examined.
Few RCTs have examined the effects of statins on platelet attributes. An RCT indicated that statins might reduce platelet count [
44], consistent with our findings. We also found that rs12916 T reduced platelet crit, and increased MPV. A cohort study found platelet count was positively associated with CVD risk and mortality [
45], as did an MR study [
46]. Few RCTs have examined these questions, but MPV is increasingly realized to be important to CVD [
47,
48], and platelet crit may be associated with stroke [
49], so these may be additional mechanisms by which protective effects of statins are actuated.
Statins increasing BMI [
12,
34] and HbA1c [
33,
50] have been reported previously, consistent with our findings. We also found that the rs12916 T allele affected BMR, fat-mass, and fat-free mass. Genetically mimicked PCSK9 inhibitors and ezetimibe were not associated with BMI or BMR, and genetically mimicked PCSK9 inhibitors were not associated with HbA1c, consistent with previous findings [
51]. LDL-c had minor effects on BMR, anthropometrics, and most body composition traits. Given statins adversely impact body composition and glycemic traits more strongly than other lipid modifiers but have similar effects on IHD per unit change in LDL-c, it suggests that statins have greater effects via lipid and/or non-lipid mechanisms than other lipid modifiers.
Differences in pleiotropic effects of genetically mimicked statins compared to PCSK9 inhibitors and ezetimibe may be related to the differences of their mechanisms of action. PCSK9 inhibitors reduce the degradation of LDL receptors by blocking PCSK9 [
52], resulting in lower levels of circulating LDL-c [
53]. Ezetimibe only inhibits the absorption of cholesterol [
54]. Statins target cholesterol synthesis [
55], including de novo synthesis in Leydig cells, so statins specifically affect hormones, although the mechanism by which statins might affect SHBG in women is less clear.
This study aimed to identify pleiotropic effects of statins, and whether they might be mediated by the target of statins, LDL-c, by testing the associations of LDL-c with any pleiotropic effects. We found the pleiotropic effects of statins did not appear to be driven by LDL-c, and so are specific effects of statins. Exactly, what drives these pleiotropic effects of statins has not been definitively established, nor has their interrelationships, which may well be complex. A possibility is that the pleiotropic effects of statins are driven by effects of statins on BMI. However, statins raise BMI in both sexes (Table
1), while genetically mimicked effects on calcium and SHBG were specific to women, suggesting a more complex explanation. A trial of gastric bypass suggested that BMI increases platelet count [
56] but genetically predicted BMI does not appear to affect platelet count in women (Additional file
1: Table S14). However, BMI is well-known to play a crucial role in BMR in both sexes [
57]. Using multivariable MR, the pleiotropic effects of genetically mimicking statins on BMR were not independent of BMI (Additional file
1: Table S15), suggesting that any effects of statins on BMR are due to statins raising BMI.
Despite a comprehensive sex-specific scan in the largest available studies, this study has some limitations. First, not all phenotypes of interest were available for the main analysis or for replication, such as very-low-density lipoprotein and some other lipid sub-fractions, and some body composition traits; however, effects of statins on lipid fractions have been examined before [
13‐
15]. Second, this study is systematic and comprehensive but is also agnostic. As such, it uses a stringent test for significance to avoid chance findings, so may not replicate all known effects of statins. Instead, this study may provide information about unknown or overlooked potential effects of statins, which is important because of the very widespread use of statins globally. The multiple comparison cut-off using a Bonferroni correction is suitable for agnostic studies, as here. An agnostic study design is most appropriate for identifying pleiotropic effects which have not been considered before rather than for replicating findings based on known physiological pathways, which can be evaluated on different criteria in the context of all the other evidence. The largest available GWAS of IHD does not show rs12916 associated with IHD at genome-wide significance [
58], which does not invalidate the GWAS or the role of statins in preventing IHD; instead, the GWAS provides information about other, possibly overlooked, factors that could be relevant to IHD. Similarly, here, this study does not invalidate known relations of statins with IHD or testosterone, which have been demonstrated in a meta-analysis of RCTs [
59,
60], but provides additional insight about potentially relevant pleiotropic effects of statins. Overall, this study design is most appropriate for identifying pleiotropic effects of statins rather than replicating known effects. We cannot exclude the possibility that some novel effects of statins have been missed, which could be addressed by repeating this study when larger sex-specific genetic studies are available. Third, we used rs12916 and associated genetic variants as a surrogate for the pharmacological effects of statins [
12], which mimics a life-long small dose of endogenous statins [
8,
10,
13], so the MR estimates represent life-long inhibition of HMGCR [
8,
10] and do not necessarily reflect the effects of statin treatment which generally starts in middle age [
10,
61]. These estimates are usually different in magnitude from the short-term effects of pharmacologic interventions in an RCT [
62] although similar effect sizes have been seen for genetically mimicked statins and use of statins [
13]. Fourth, the UK Biobank is not representative of the UK population. However, no confounding and no selection bias are the criteria for an internally valid study of associations, not population representativeness [
63]. The UK Biobank has shown similar associations to an equivalent population-representative study [
64]. However, as with any study recruited in middle to older age, the UK Biobank is missing people who died before recruitment from their genetic make-up, from a condition of interest, or from a competing risk of such a condition, which may generate selection bias particularly for conditions that share etiology with diseases that cause death before recruitment [
65], so effects of statins on late-onset diseases may have been missed. The UK Biobank study undoubtedly comprises healthy volunteers, less vulnerable to disease, which may bias associations with disease towards null. Fifth, the underlying studies mostly concern populations of European descent, due to data availability, which may limit the generalizability of these findings to other populations. It would be extremely beneficial to validate these findings in consortia more representative of the global population.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.