SLCO1B1 genetic variant associated with statin-induced myopathy: a proof-of-concept study using the clinical practice research datalink.

This study aimed to determine whether patients with statin-induced myopathy could be identified using the United Kingdom Clinical Practice Research Datalink, whether DNA could be obtained, and whether previously reported associations of statin myopathy with the SLCO1B1 c.521T>C and COQ2 rs4693075 polymorphisms could be replicated. Seventy-seven statin-induced myopathy patients (serum creatine phosphokinase (CPK) > 4× upper limit of normal (ULN)) and 372 statin-tolerant controls were identified and recruited. Multiple logistic regression analysis showed the SLCO1B1 c.521T>C single-nucleotide polymorphism to be a significant risk factor (P = 0.009), with an odds ratio (OR) per variant allele of 2.06 (1.32-3.15) for all myopathy and 4.09 (2.06-8.16) for severe myopathy (CPK > 10× ULN, and/or rhabdomyolysis; n = 23). COQ2 rs4693075 was not associated with myopathy. Meta-analysis showed an association between c.521C>T and simvastatin-induced myopathy, although power for other statins was limited. Our data replicate the association of SLCO1B1 variants with statin-induced myopathy. Furthermore, we demonstrate how electronic medical records provide a time- and cost-efficient means of recruiting patients with severe adverse drug reactions for pharmacogenetic studies.

The UK Clinical Practice Research Datalink (CPRD), formerly the General Practice Research Database, is a computerized database of anonymized longitudinal medical records from primary care. In March 2011 there were more than 12 million patient records contributing more than 64 million years of prospectively collected data; the number of records is to be increased to 52 million with the transition to CPRD.[1] The information collected includes patient demographics, medical diagnoses, prescription information, referrals, and health outcomes. Although the database has been widely used in observational studies, including reports on clinical epidemiology, disease patterns, drug utilization, and outcomes research, resulting in more than 800 publications, it has never been used to obtain patient samples for biomarker analysis.

To determine whether the CPRD could be used for biomarker analysis, we focused on the pharmacogenetics of statin-induced myopathy. This was chosen as the paradigm for several reasons: first, statins (inhibitors of 5-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase) are widely used, being the cornerstone of therapy for hyperlipidemia, with proven efficacy for both primary and secondary prevention of atherosclerotic arterial disease.[2] Although generally well tolerated, a few patients develop muscle-related adverse effects, ranging from muscle pains without any elevation of plasma creatine kinase (CK)—a biomarker for muscle injury—to rhabdomyolysis, in which CK is elevated to > 10 times the upper limit of normal (ULN), which may be associated with renal impairment.[3] A systematic review of 21 clinical trials[4] suggested that mild muscle pain, myopathy, and rhabdomyolysis attributable to statin therapy occurred at an incidence of 190, 5, and 1.6 per 100,000 patient years, respectively.

Second, functional variation of the hepatic uptake transporter SLCO1B1 has been implicated in statin-induced myopathy. A genome-wide association study of 85 patients with incipient (CK level >3× ULN and >5× baseline) or definite myopathy (muscle symptoms with CK > 10× ULN) and 90 controls who were receiving 80 mg/day simvastatin showed a strong association with a noncoding single-nucleotide polymorphism (SNP; rs4363657).[5] This was subsequently found to be in nearly complete linkage disequilibrium with a nonsynonymous c. 521T>C SNP (rs4149056) that encodes a valine to arginine amino acid substitution at residue 147 (p.V147L) and defines the SLCO1B1*5 allele. This variant has subsequently been associated with statin-induced myopathy in a number of other studies.[6,7,8] The incidence of statin-induced myopathy has been reported to be 19% in individuals without any *5 alleles, 27% in heterozygous individuals, and 50% in *5/*5 homozygous individuals.[8]

Recent studies have also suggested that variation in the coenzyme Q2 (COQ2) homologue gene may also predispose individuals to statin-induced myopathy. Puccetti et al. demonstrated an association between both rosuvastatin- and atorvastatin-induced myopathy and the rs4693075 polymorphism in the COQ2 gene.[9] An association of another COQ2 variant (rs4693570) and statin-induced myalgia has also been described.[10] Variants of the COQ2 loci are directly involved in CoQ deficiency,[10] a postulated mechanism of statin-induced myopathy.[11,12]

Third, in randomized controlled trials, the incidence of statin-induced myopathy is very low. For example, of 6,031 patients receiving 80 mg simvastatin, the SEARCH (Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine) study[5] identified just 49 (0.8%) patients who had developed myopathy (defined muscle symptoms with CPK > 10× ULN). Thus, it is important to explore other methods for recruiting patients from particular electronic records, in which the use of individual drugs is usually much higher than that in trials, and it represents real-world clinical practice, in which the incidence of severe adverse reactions is higher. This article thus describes the process by which the CPRD was used to identify and recruit a cohort of statin-receiving patients with and without an increase in CPK levels in the presence or absence of muscle symptoms. We have then undertaken genotyping for SLCO1B1 and COQ2 variants.

Results

Statin-induced myopathy case recruitment

A total of 76 cases were recruited between June 2010 and November 2011, and a total of 372 controls were recruited from the General Practice Research Database between June 2010 and April 2012. Clinical data are summarized in . Within the first phase of recruitment (June 2010 onward), a total of 520 potential cases of statin-induced myopathy were identified on the patient list of recruited general practice clinics. Of these, 223 (42%) were deemed suitable by the physician for inclusion. As of November 2011, 76 (34%) patients had provided adequate biological samples (blood or saliva) to the receiving laboratory. Full recruitment statistics for the 36-month study period will be subsequently reported in a future publication.

Demography

At the time of the reported event, 59 of 76 (78%) myopathy patients were receiving simvastatin; 11 (14%) were on atorvastatin, and 6 (8%) were using other statins (cerivastatin, pravastatin, rosuvastatin, or fluvastatin). In the control cohort, 222 of 372 (60%) were receiving simvastatin at the time of recruitment, 30% were on atorvastatin, and 10% received other statins (). Univariate binary logistic regression analyses () showed borderline statistically significant differences between cases and controls in terms of the statin type (P = 0.075) and previous history of type 2 diabetes (P = 0.046), asthma (P = 0.080), and hypertension (P = 0.087). These four variables were all adjusted for in the SNP-association analyses. There was no difference in the use of CYP3A4 inhibitors between cases and controls.

SNP analysis

Both SNPs conformed to Hardy–Weinberg equilibrium (P > 0.0001). The two SNPs were successfully genotyped in 99.7% (rs4693075) and 100% (rs4149056) of individuals. For logistic regression analysis of COQ2 rs4693075, 371 controls were included. On comparing the SNP model including the SLCO1B1 c.521T>C SNP (rs4149056) with the baseline model, the likelihood ratio test gave a significant P value () both when incorporating all statin-induced myopathy cases (76 cases, 372 controls; P = 0.005) and when limiting the analysis to just patients with severe myopathy (23 cases, 372 controls; P = 0.0003). Limiting analysis to only those individuals receiving atorvastatin (n = 121) demonstrated no significant association between SLCO1B1 c.521T>C (rs4149056) and risk of either myopathy (P = 0.613) or severe myopathy (P = 0.507). However, in patients receiving simvastatin (n = 281), statistically significant associations between c.521T>C (rs4149056) and risk of both myopathy (P = 0.014) and severe myopathy (P = 0.0004) were observed. Addition of the COQ2 rs4693075 to the baseline model did not give a statistically significant P value for either all myopathy (P = 0.358) or severe myopathy (P = 0.937).

Binary logistic regression () demonstrated a significant risk per SLCO1B1 c.521 C allele for all myopathy cases regardless of prescribed statin (n = 76; odds ratio (OR) = 2.08 (1.35–3.23), P = 0.005). This translates to an OR of 4.32 (1.82–10.43) for risk of all myopathy for CC carriers as compared with TT carriers. For cases with severe myopathy (n = 18), an even higher risk per C allele was observed (OR = 4.47 (1.84–10.84)), translating to an OR of 19.98 (3.38–117.50) in CC individuals vs. that in TT individuals.

Limiting this analysis to individuals receiving simvastatin only demonstrated a similar risk to that observed for all statins, with a per-C-allele OR for all myopathy (n = 59) of 2.13 (1.29–3.54; P = 0.014). For simvastatin-induced severe myopathy (n = 18), the OR was 4.97 (2.16–11.43). Stratification of simvastatin patients (all myopathy) into those receiving <40 mg/day (n = 24) or ≥40 mg/day (n = 35) showed an increased risk for c.521C-allele carriers in the ≥40-mg/day group (OR = 3.23 (1.74–5.99), P = 0.0002)), whereas no significant risk was observed in the <40-mg group (OR = 1.03 (0.45–2.36), P > 0.05). For severe myopathy in patients receiving ≥40 mg/day simvastatin (n = 13), the OR per-C-allele was 6.28 (2.38–16.60; P = 0.0004). In patients receiving <40 mg/day (n = 5), no significant association was observed with severe myopathy (OR = 1.84 (0.34–9.86)).

Meta-analysis

A total of seven studies, including our own, were included in the initial meta-analysis of myopathy risk for SLCO1B1 c.521C carriage for any statin (). The overall OR for myopathy risk was 2.18 (1.39–3.43). Limiting the analysis to those studies (n = 4) reporting genotype frequency in patients receiving simvastatin, the combined OR was marginally higher at 3.25 (1.72–6.12). Three studies reported frequencies of SLCO1B1 in atorvastatin-receiving patients. The combined OR for myopathy was not significant at 1.54 (0.80–2.97).

Discussion

The recruitment of patients with severe adverse drug reactions to pharmacogenomic studies is complicated by the facts that these reactions are rare and there is no systematic process for identifying patients. The use of electronic health records therefore represents an opportunity to undertake such studies, but, to date, electronic health records have not been used to identify patients with severe and rare phenotypes. Part of the problem here is that the phenotypes in the databases may be inadequate, leading to capture of heterogeneous patient groups and thus the identification of no or weak associations. It is well known that phenotype standardization is crucial in order to disentangle the signals from noise.[13]

To evaluate whether electronic health care record databases can be used to recruit patients with severe adverse drug reactions, we first chose CPRD as the database to undertake this feasibility study because of the quality of data contained within, which has resulted in a large number of important drug safety findings (http://www.cprd.com). We then chose statin-induced myopathy as the paradigm adverse drug reaction. Although statin-induced myopathy can present with many different clinical manifestations,[3] and indeed previous pharmacogenetic studies have used different end points (), our inclusion criteria were simple, based on an increase in CPK levels. A previous study in Scotland using electronic records used a composite definition of intolerance based on increases greater than 50% from baseline in alanine transaminase and/or 1–3× ULN in CPK, with an accompanying prescription change.[7] This perhaps represents a milder intolerance phenotype as compared with our definition of CPK > 4× ULN. The utility of our approach is shown by the fact that over a period of 16 months, after administrative startup, we were able to recruit 76 patients with statin-induced myopathy, of whom 23 were of a more severe phenotype, denoted by CPK > 10× ULN or rhabdomyolysis. The CPRD (as of October 2009) recorded 127,268 individuals receiving a statin with a concurrent CK measurement recorded. Of those, 953 (0.75%) had CK > 4× ULN concurrent with statin prescription (T.v.S., unpublished data), an incidence comparable with that reported previously.[4]

Our results show that the rs4149056 SNP in SLCO1B1 is associated with statin-induced myopathy. This is in accordance with previous findings,[5,6,7,8] confirming the utility of our approach. We have shown that possession of at least one copy of the C-allele (CT/CC) is a significant risk factor for statin-induced myopathy (CK > 4× ULN), with an observed OR per C allele of 2.09 (1.27–3.45). The risk per C allele of severe myopathy (CK > 10× ULN/rhabdomyolysis; n = 23) was greater still, with an OR of 4.47 (1.84–10.84).

Our data replicate those of Link et al.,[5] who recruited cases and controls from a randomized trial setting, showing that our cases recruited through CPRD, an observational database, are comparable. However, our cases differ from those recruited by Link et al. in two important aspects: (i) the observations first made by Link et al. were in patients receiving 80 mg/day simvastatin, whereas the mean daily dose in this study was lower (33.4 ± 19.7 mg); and (ii) only 78% of our cases with myopathy were on simvastatin, with 22% receiving other statins, including atorvastatin (in 14% of cases). Limiting the analysis to those receiving simvastatin only demonstrated an association between SLCO1B1 c.521T>C and both all myopathy cases (OR = 1.92 (1.08–3.42)) and those with severe myopathy (OR = 4.99 (1.72–14.50)). However, the association was observed only in those patients receiving ≥40 mg/day simvastatin (), indicating the importance of dose–genotype interaction. Despite the differences, the per-C-allele OR of 4.5 for high-dose (80 mg/day) simvastatin-induced myopathy (defined as CK > 3× ULN) by Link et al.[5] was highly comparable with that observed in our study for the equivalent phenotype (CK > 4× ULN) with ≥40 mg/day simvastatin (4.97; 95% confidence interval: 2.16–11.43).

Atorvastatin was the second most common drug implicated in our case group, reflecting its usage in comparison with simvastatin. However, unlike in simvastatin-treated patients, there was no significant association between the SLCO1B1 c.521T>C variant and either myopathy or severe myopathy in atorvastatin-treated patients. This is consistent with a previous study that showed that the association was stronger for simvastatin than for atorvastatin.[6] Our meta-analysis of studies in Caucasians, including our data (), also shows that there was a higher risk with simvastatin (OR = 3.25 (1.72–6.12)) than with atorvastatin (OR = 1.54 (0.80–2.97)), regardless of daily dose, in carriers of the SLCO1B1 polymorphism. Pathophysiologically, this would be consistent with the fact that this polymorphism has the greatest effect on simvastatin (area under the curve is 221% higher in patients with the c.521CC genotype than in patients with the c.521TT genotype) but also has a smaller effect on atorvastatin (mean increase in area under the curve of 173%), and a very small, if any, effect on the other statins.[14] We did not have enough patients treated with the other statins to undertake any meaningful drug-specific analyses.

Recent studies[9,15] have shown that variation in the COQ2 gene also predisposes an individual to statin-induced myopathy. However, we could not replicate the association with the COQ2 rs4693075 polymorphism in our patient group. Previous studies included patients mainly receiving atorvastatin and rosuvastatin.[9] In our study, just 13 (17%) of the statin-intolerant patients and 4 (16%) of the severe myopathy cases were receiving either atorvastatin or rosuvastatin. As such, we did not have sufficient statistical power to test this particular hypothesis. On the basis of the minor allele frequency of 0.35 observed in our atorvastatin-tolerant patients, we would require 135 cases and controls in order to have a study with 80% power to detect an OR of 2 and a significance value of 0.05.

The percentage of suitable statin-induced myopathy patients, identified within general practices, from whom biological samples were ultimately received (34%) was actually better than we had expected (20–25%). A previous study using spontaneous reports under the UK yellow card scheme to obtain biological samples from patients with terodiline-induced cardiotoxicity demonstrated a success rate of 25%.[16] Of course, we need to strive for higher recruitment rates for future studies, but interest in taking part in research studies by medical professionals is always tempered by the lack of time available. However, it should also be noted that a huge amount of time was saved through the more rapid identification of cases using the database, which would not have been possible through manual case-note searching.

In conclusion, there are clear time and cost benefits in using electronic patient records, such as the CPRD, for recruiting patients for genetic studies, particularly for rare phenotypes, such as statin-induced myopathy. There are also clinical benefits because the recruited patients will be from a real-world setting, and hence the effects of clinical factors such as concomitant medications can be evaluated. The electronic Medical Records and Genomics (eMERGE) network has already demonstrated the applicability of electronic medical records to identifying genomic loci associated with a population trait, white blood cell counts.[17] Others have applied a similar methodology to the identification of patients for pharmacogenetic studies of drugs such as warfarin.[18] In terms of the clinical utility of the genetic association between the SLCO1B1 polymorphism and statin-induced myopathy, there is now convincing evidence for simvastatin, but not for other statins, for which more studies are needed. A recent Clinical Pharmacogenetics Implementation Consortium guideline has made some recommendations regarding dosing and choice of statin in patients with the variant SLCO1B1 genotype.[19]

Methods

Study design

From a cohort of ~600,000 patients receiving statins identified in the CPRD (http://www.cprd.com), a case–control design was used to identify suitable patients for the study. Participation was restricted to Caucasians ≥18 years of age and with the first-ever statin prescription at least 1 year after the start of CPRD data collection. Potential cases were selected from the database if they discontinued their implicated statin therapy and demonstrated an increase in CPK > 4× ULN.

Potential controls were selected if they had been receiving statins for at least 3 months with no previous above-normal serum CPK measurements. General practitioners were contacted with a list of potential cases and/or controls identified from their practices. After being given the opportunity to decline involvement, they were first asked to review the list and remove any patients they considered unsuitable. They were then asked to contact suitable patients by letter requesting participation. Consenting case patients were randomized and invited to provide either a saliva sample (by post) or a blood sample (by visit to the practice). Controls provided only blood samples. All samples were then forwarded to the University of Liverpool for processing. To preserve anonymity, patient and practice identifier codes were used throughout the recruitment process, and all patient contact was through the general practitioner only.

Ethical approval was obtained from the National Research Ethics Committee North West 2—Liverpool Central, and approval to use the CPRD data was obtained from the Independent Scientific Advisory Committee at the Medicines and Healthcare Products Regulatory Agency. In addition, site-specific approval to contact the GP practices was obtained for each of the 138 primary-care trusts across the United Kingdom. Local informed consent was obtained from all study subjects or their guardians in accordance with the Declaration of Helsinki.

Genomic DNA was extracted from 5 ml of whole blood or 2 ml of saliva (collected using the Oragene DNA Sampling kit, DNAGenotek, Ontario, Canada) using the Chemagic Magnetic Module 1 system per the manufacturer's protocol (Chemagen Biopolymer-Technologie, Baesweiler, Germany).

A total of 448 individuals were genotyped for the rs4149056 SNP in SLCO1B1 and rs4693075 in COQ2 using commercially available TaqMan real-time PCR SNP genotyping assays with 1× Genotyping Master Mix (both from Applied Biosystems, Carlsbad, CA). Subsequently, 20 ng of genomic DNA per reaction was genotyped according to the manufacturer's protocol using an ABI 7900HT real-time PCR system (Applied Biosystems, Carlsbad, CA); Ten percent of the samples were run in duplicate to ensure concordance of genotype.

A univariate analysis of association between all nongenetic variables considered to be of a priori interest and case–control status was first undertaken. The χ2 test was used for categorical variables and Student's t-test for continuous variables. Any variable demonstrating a statistically significant association (P < 0.10) was carried forward and adjusted for in the SNP association analyses.

To test for association with each SNP in turn, two multiple logistic regression models were fitted. The first (the baseline model) included all univariately significant (P < 0.10) nongenetic variables. The second (the SNP model) was the same but also included a covariate to represent the SNP (either rs4149056 or rs4693075). An additive effect of the variant allele was assumed. Homozygote wild type was coded as “0,” heterozygote as “1,” and homozygote variant allele as “2”.

To test for association with the SNP, the likelihood ratio test was used to compare the SNP model with the baseline model. A P value <0.025 (0.05 corrected for two tests of associations using the Bonferroni approach) was assumed to represent statistical significance of the SNP.

Sensitivity analyses were undertaken by separately limiting cases to those classified as having either plasma CK > 10× ULN or rhabdomyolysis (n = 23; termed “severe myopathy”). All statistical analyses were undertaken using SPSS version 17.0.

A search of PubMed (http://www.ncbi.nlm.nih.gov/pubmed accessed January 2012) using the search terms “SLCO1B1” and “statin” yielded 108 publications, of which 96 were original research articles. Inspection of titles and abstracts identified six research articles that defined the frequency of the SLCO1B1 rs4149056 polymorphism in an entirely, or predominantly, Caucasian population of statin-induced myopathy. Studies were included regardless of the suspect statin investigated, dose, and myopathy phenotype observed (as described in ). Due to the high degree of heterogeneity among the included studies (I2 = 84.1%), a DerSimonian-Laird random effects model was applied to the meta-analysis in StatsDirect version 2.6.8 (StatsDirect, Altrincham, UK)

Author Contributions

D.F.C. and H.O. wrote the manuscript. T.v.S., G.M., and M.P. designed the research. D.F.C., J.C., M.H., and H.O. performed the research. D.F.C, H.O., A.L.J., G.M., T.v.S., and M.P. analyzed the data.

Study Highlights

Table 1

Case–control comparison of nongenetic clinical variables

Table 2

Multiple logistic regression analysis of statin-induced myopathy risk and SLCO1B1 p.V174A and COQ2 rs4693075 genetic variants