Statin treatment effectiveness and the SLCO1B1*5 reduced function genotype: Long‐term outcomes in women and men

Objective To estimate the effect of rs4149056 (SLCO1B1*5) genotype (decreases statin transport) on cholesterol control and treatment duration in male and female primary care patients prescribed common statin medications. Methods and Analysis This study comprised 69 185 European‐ancestry UK Biobank cohort participants prescribed simvastatin or atorvastatin (aged 40‐79 years at first prescription, treatment duration 1 month to 29 years, mean 5.7 years). Principal outcomes were clinically high total cholesterol (>5 mmol/L) at baseline, plus treatment discontinuation. Results A total of 48.4% of 591 females homozygous for SLCO1B1*5 decreased function genotype had raised cholesterol vs 41.7% of those with functioning SLCO1B1 (odds ratio 1.31, 95% confidence interval [CI] 1.1‐1.55, P = .001). Fewer males had high cholesterol and the genotype effect was attenuated. In primary care prescribing, females homozygous for SLCO1B1*5 were more likely to stop receiving these statins (29.5%) than women with normal SLCO1B1 (25.7%) (hazard ratio [HR] 1.19, 95% CI 1.03‐1.37, P = .01), amounting to five discontinuations per 100 statin‐years in the SLCO1B1*5 group vs four in the normal SLCO1B1 function group. This remained significant after the first year of treatment (HR for discontinuing >1 year after first prescription 1.3, 95% CI 1.08‐1.56, P = .006). In men SLCO1B1*5 was only associated with treatment discontinuation in the first year. Conclusions In this large community sample of patients on commonly prescribed statins, the SLCO1B1*5 decreased function variant had much larger effects on cholesterol control and treatment duration in women than in men. Efforts to improve the effectiveness of statin therapy in women may need to include SLCO1B1*5 genotype‐guided statin selection.


| INTRODUCTION
Elevated low-density lipoprotein cholesterol (LDL-C) level is a major risk factor for myocardial infarction and stroke. 1 Statins are the most commonly prescribed cholesterol-lowering drugs and reduce cardiovascular morbidity and mortality in higher risk patients. 2,3 However, a major barrier to effectiveness is medication nonadherence, often due to reported side effects, including muscle pain. 4 The STRENGTH study included 509 hypercholesterolaemic patients who were randomised to simvastatin, atorvastatin or pravastatin; discontinuation of treatment due to adverse effects was significant for both simvastatin (odds ratio [OR] 2.8, 95% confidence interval [CI] 1.3-6.0) and atorvastatin (OR 1.6, 95% CI 0.7-3.7) at 16 weeks follow-up. 5 However, a systematic review of randomised controlled trials (n = 74 102) found no differences between placebo and statin groups for developing muscle symptoms and discontinuation of treatment, 6 perhaps because trial participants are generally healthier than many older people pre- for affecting simvastatin or atorvastatin effectiveness or toxicity. 14 A meta-analysis of 13 atorvastatin studies found SLCO1B1*5 was associated with atorvastatin-related adverse drug reactions (OR 1.57, P = .01), 15 yet literature linking SLCO1B1*5 5,9,10 to statin's evidence is mixed and mostly focussed on shorter term outcomes (statinrelated myotoxicity is predominantly reported in the first year of treatment, with median onset of 1 month after treatment initiation). 11 A 2013 study of UK primary care recruited 77 patients with statininduced myopathy and found SLCO1B1*5 significantly increased risk of myopathy compared to controls (OR per *5 allele 2.1, 95% CI 1.3-3.2), 16 but more research is needed linking genotype and general practice (GP) data.
Though sex differences in cholesterol levels are known, with LDL generally lower in men 17 and higher total cholesterol levels in women whilst treated with statins, 18 previous studies have mainly focused on men and effects in women are understudied. [19][20][21] In addition, women have greater risk of adverse drug reactions, yet many cardiovascular risk models do not take into account female-specific factors. 22 Historically, UK guidelines for prescription of statins for cardiovascular disease (CVD) prevention used the same clinical cut-off for high cholesterol (>5 mmol/L) and LDL (>3 mmol/L) for men and women. 23 Current UK clinical recommendations are to begin atorvastatin treatment when 10-year CVD risk is >10% and to assess statin effectiveness if 40% reduction in non-HDL cholesterol is achieved 3 months after treatment initiation. 24 Women have lower body weight and a higher percentage of body fat compared to men, which might lead to higher concentrations of lipophilic drugs such as simvastatin and atorvastatin, 25 and increased risk of adverse events, which may be exacerbated by the SLCO1B1*5 decreased function genotype, which increases concentrations of drugs due to less uptake of the drug. 12 We therefore aimed to determine whether women prescribed simvastatin or atorvastatin were as likely as men to achieve cholesterol levels below clinically high cut-off points using data from the UK Biobank, a large cohort of community volunteers followed in primary care and hospital electronic medical records for over 10 years. We also tested whether the SLCO1B1*5 (rs4149056) decreased function genotype was associated with discontinuing statin treatment (in the first year and the longer term) in males and females separately. Statins are known to impact inflammation 26 and diabetes risk, 3 so we also assessed C-reactive protein (CRP), alanine aminotransferase (ALT) and HbA1c at baseline. We investigated effects on self-reported side What is already known about this subject  27 Comprehensive questionnaires on demographic, lifestyle and health information data were collected at the baseline assessment. Blood samples for genetic and biochemical analyses, and anthropometric measures were taken. This study of atorvastatin and simvastatin comprises two distinct analysis sections.
First, using the data from baseline assessment only, and second using the linked GP (primary care) data available in 230 096 (45.7%) participants ( Figure 1).

| Baseline assessment
The UK Biobank baseline assessment included self-reported medications; we analysed simvastatin and atorvastatin.
We also analysed self-reported symptoms associated with statin use which may cause discontinuation of the treatment. 32 We used "Headaches for 3+ months" (data field 3799), "Frequency of tiredness/lethargy in last 2 weeks" (data field 2080) and any reported pain for 3+ months (combining data fields 3404, 3571, 3741, 3414, 3773, 2956). GP list, where available) or 28 February 2016 where no deduction date was present (ie, still registered at an available practice). Data after 28 February 2016 were incomplete, depending on GP provider (see UK Biobank documentation). 33 We also evaluated prescriptions for other statins (cerivastatin, fluvastatin, pravastatin and rosuvastatin) to identify patients switching treatments from simvastatin or atorvastatin.

| GP data
Muscle symptoms were ascertained from ICD-10 codes 34 and converted to Read codes used in UK primary care records (using UK Biobank-provided diagnostic code maps), available for up to 11 years follow-up after baseline assessment. We included ICD-10 codes for myopathy, myositis or myalgia (G72.0, G72.8, G72.9, M60.8, M60.9, M79.1). We analysed the SLCO1B1*5 genetic variant rs4149056 (C allele, directly genotyped) with well-documented effects on simvastatin-and atorvastatin-related side effects in the literature, particularly on muscle symptoms. 37 Genotype data was not returned to participants as part of the study.

| Statistical analyses
Associations between genotype and biochemical variables at the baseline were tested by logistic regression, adjusting for age and the first 10 principal components of genetic ancestry to control for population substructure.
The association between genotype and discontinuation was tested using Cox's proportional hazards regression models. We also created Kaplan-Meier plots. Participants entered the model at the date of first prescription of statins and exited on the date of first incident outcome or end of records, thus providing an "intention to treat" analysis reducing any effect of genetically associated discontinuation of treatment. We tested the associations between GP-diagnosed muscle symptoms that occurred in the first 3 months and 3 months or longer after the first prescription date using time-to-event models.

STATA (v16) was used for analysis.
To estimate the genetically moderated treatment effect (GMTE) we used "TWIST" (Triangulation with a Study), 38 a novel pharmacogenetic causal inference approach to estimate population average effect on total cholesterol if all SLCO1B1*5 homozygotes could experience the same treatment effect as noncarriers. In brief, several assumptions common to pharmacogenetic analysis are tested (primarily that genetic variant SLCO1B1*5 does not predict whether an individual receives statin treatment, is not associated with any confounders predicting statin use or cholesterol and only affects cholesterol through the interaction with statins). From this analysis the most efficient and robust estimate of the GMTE is derived. R (v4.0.1) was used for TWIST analysis with package "twistR" (https://github. com/lukepilling/twistR).

| Sensitivity analysis
a. We conducted competing risk regression models for discontinuation or death to check whether the estimate is drastically changed when accounting for the competing risk of mortality. b. We also tested for interactions between SLCO1B1 genotype and sex with discontinuation of treatment within year 1 and discontinuation after 1 year of treatment.

| Patient and public involvement
Patients and participants are extensively involved in the UK Biobank study itself. No patients were involved in developing the research question or the outcomes tested in this analysis.

| Characteristics and associations at UK baseline assessment
There were 26 185 female and 41 445 male European-ancestry UK Biobank participants who reported atorvastatin or simvastatin treatment at baseline assessment. The mean age was 61.6 years (SD 5.7) for females and 61.4 (SD 6.1) for males (    Figure 2A for detailed estimates and Figure 2B for cumulative incidence plots of SLCO1B1*5 homozygotes association with discontinuing treatment (plots are truncated to 15 years for clarity, full plot Supporting Information Figure S1; see Supporting Information

| SLCO1B1*5 associations with GP-diagnosed muscle symptoms
A total of 110 female and 96 male participants had GP-recorded muscle symptoms in the 3 months after their first prescription of atorvastatin or simvastatin. In addition, 848 female and 1026 male F I G U R E 2 SLCO1B1*5 genotype association with discontinuing GP-prescribed simvastatin and atorvastatin treatment. Associations between SLCO1B1*5 genotype (reduced function compared to normal genotype, ie, rs4149056 CC homozygotes vs TT homozygotes) and discontinuing GP-prescribed simvastatin or atorvastatin treatment in males and females separately. (A) The number of "cases" (discontinuing treatment) and "controls" (remained on treatment) for the normal and reduced-function homozygous groups, the number of discontinuations per 100 personyears on treatment in the two groups, and the hazard ratio from Cox's proportional hazards regression models. Also shown are the associations from stratified analyses of short term (stopped less than 1 year after beginning treatment) and longer term (stopped more than 1 year after beginning treatment). See Supporting Information Table S3 for details, including for the normal/reduced (*1/*5) heterozygous group. (B) The cumulative incidence over time of discontinuing treatment in males and females, stratified by SLCO1B1*5 genotype. The x axis is censored at 15 years for figure clarity. See Supporting Information Figure S1 for complete plots participants had GP-recorded muscle symptoms more than 3 months after the first prescription of atorvastatin or simvastatin (ie, the stable treatment period). This was lower than expected based on previous literature, 2,5,9 with similar rates in the different genotype groups. In the first 3 months, there was no significant association between muscle symptoms and genotype. After 3 months, female SLCO1B1*5 heterozygotes were more likely to experience muscle symptoms compared to female normal function homozygotes (OR 1.19, 95% CI 1.03-1.4, P = .02). However, there was no significant association with muscle symptoms in males in any group of SLCO1B1*5 genotype (see Supporting Information Table S6).

| DISCUSSION
We aimed to assess the success of cholesterol control in men and women taking simvastatin or atorvastatin, including examining the contribution of the SLCO1B1*5 (rs4149056) genetic variant that impairs intracellular transport of these statins. In UK Biobank participants who self-reported taking simvastatin or atorvastatin at baseline assessment, 42% of females and 25% of males had clinically high total cholesterol (>5 mmol/L), despite receiving treatment (OR 2.2, 95% CI 2.11-2.26, P = 5 Â 10 À439 ), with consistent results for LDL also observed. We found that females homozygous for SLCO1B1*5 (ie, with reduced protein function) were more likely to have high cholesterol compared to common "normal function" homozygotes. In the linked GP electronic clinical records data, female SLCO1B1*5 reduced-function homozygotes were more likely to discontinue simvastatin or atorvastatin treatment: five discontinuations per 100 patient-years on statins compared to four per 100 in the normal function genotype group.
The difference between males and females could be due in part to males being more likely to adhere to statin therapy, 39 and is consistent with previous reports of females having higher total cholesterol levels whilst treated with statins. 18  We included patients treated for up to 29 years because likelihood of discontinuation of statin treatment is impacted by adverse events, in addition to whether the patient meets the target cholesterol reduction. A systematic review of randomised controlled trials (RCTs) with 74 102 subjects found that statin therapy was not associated with discontinuation of treatment compared with placebo, 10 yet our analysis shows that for a subset of patientsespecially females carrying the SLCO1B1*5 genotypediscontinuation is more likely.
As the studied statins are mainstays of CVD prevention, and nonadherence is a major barrier to treatment effectiveness, prescribing an appropriate statin without high risk of adverse events and with higher efficacy at first intervention could reduce discontinuation and improve control of cholesterol, especially in women. We used a novel pharmacogenetic causal inference framework (TWIST 38  This significant interaction between SLCO1B1*5 genotype and sex could be due to females having lower mean muscle mass and body weight, and higher percentage of body fat compared to males, leading to higher concentrations of simvastatin and atorvastatin, 25 with SLCO1B1*5 therefore causing increased discontinuation of treatment due to side effects. 12 Although we observe raised cholesterol levels in SLCO1B1*5 homozygotes at the UK Biobank baseline assessment, and increased likelihood of discontinuing GP-prescribed simvastatin and atorvastatin therapy, we found limited evidence of SLCO1B1*5 associations with GP-diagnosed muscle symptoms. This could be due to underreporting of statin-associated pain (by the patients themselves or underrecording by GPs in the clinical record; 3.4% of participants prescribed simvastatin or atorvastatin received a relevant GP diagnosis, including myalgia and myositis) compared to previous studies with systematic ascertainment of muscle effects that reported up to 25% of patients with muscle symptoms. 5,9,42 A recent systematic review of RCTs reported that statins were not associated with clinically confirmed muscle disorders, consistent with our results (patients may report muscle symptoms, but these are not clinically confirmed). 43 Additionally, our ability to analyse dose was limited due to data availability (numbers of homozygotes stratified by dose were low and GP instructions on number of tablets to be taken was missing) and SLCO1B1*5 has been linked to muscular complaints (and atorvastatin intolerance), especially in patients receiving high doses of atorvastatin. 44 However, statins are known to impact inflammation 26 and diabetes risk 3 : we found that female SLCO1B1*5 homozygotes (but not male) treated with atorvastatin or simvastatin at baseline assessment had higher CRP, ALT and HbA1c, further emphasising the increased importance of appropriate prescribing in females.
Additionally, it is thought that a "nocebo" effect is common, where patients treated with statins report more statin related symptoms. 4 48 In a recent cost-consequence analysis patients with pre-emptive SLCO1B1 testing prior to statin initiation experienced lower rates of statin-associated muscle symptoms and were less likely to discontinue, although the strategy was US$96 more costly per patient compared to usual care. 49 It is noted elsewhere that cost-effectiveness increases in multigene pharmacogenetic testing. 50 In conclusion, in the large UK Biobank community volunteer study, women prescribed atorvastatin or simvastatin were more likely to still have clinically raised cholesterol levels than men. In women,

COMPETING INTERESTS
All authors declare no support from any organisation for the submit-

DATA AVAILABILITY STATEMENT
The genetic and phenotypic UK Biobank data are available upon application to the UK Biobank (www.ukbiobank.ac.uk/register-apply).
The derived data fields used in our analysis will be available via the UK Biobank, search for application number 14631. We are not able to share these directly.