This study used the UK Biobank resource (http://www.ukbiobank.ac.uk/) to analyse data that has previously been collected. The UK Biobank study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. All participants in the UK Biobank study have given informed consent for data collection, data storage and subsequent data analysis. Ethical permission for our study was given by the responsible local authorities, Regionala Etikprövningsnämden (now Etikprövningsmyndigheten, Swedish Ethical Review Authority, EPN), Uppsala, under registration number (Diarienummer, Dnr) 2017/198.

UK Biobank is an open access resource with a collection of longitudinal health-related data and biological samples [29]. It constitutes a prospective study of more than 500,000 participants aged between 40 and 69 years when attending the assessment centre, and living in the United Kingdom. Present study is based on baseline data, collected between 2006 and 2010. Generally, detailed information regarding all variables included in the study can be found on the UK Biobank website (http://www.ukbiobank.ac.uk/). Specifically, we extracted data on the following parameters at baseline: insomnia (UK Biobank Field ID (FID): 1200), chronotype (FID: 1180), age (FID: 21003), sex (FID: 31), total cholesterol level and LDL-cholesterol level (FID: 30690), body mass index (BMI) (FID 21001), education (FID: 6138), alcohol intake frequency (FID: 1558), sleep duration (FID: 1160), smoking status (FID: 20116), physical activity (FID: 884), diagnosis of diabetes (FID: 2443), diagnosis of hypertension, angina and heart attack (ID: 6150), use of hypertensive medications, statins and insulin (FID: 20003) and Townsend deprivation index (TDI) reflecting socio-economic status of the participant based on area of residence at the time of recruitment (FID: 189).

In the current analyses, participants with non-white ethnical background (n = 26,916), genetic relatedness and quality-control failure of the genotyped data (n = 138,617) were excluded (Fig. 1). Moreover, all the participants with neurological (n = 19,916) or psychiatric disorders (n = 3882), or persons who have had a stroke (n = 5197) were excluded to minimize confounding due to these factors. Following these exclusions, 206,801 participants with genetic data as well as information on medication use were included in the present study (Fig. 1). The study participants were then subjected to two independent analyses, i.e., non-genetic and genetic. In non-genetic analyses, statin medication was used as exposure while in genetic analyses, genetic variants were used as proxies for statin use.

Fig. 1.

Study population.

We analysed three SNPs (two in the HMGCR gene (rs17238484 and rs12916) and one in the PCSK9 gene (rs11591147)) that are known to be linked to cholesterol lowering and can be used as proxies for statin use and PCSK9 inhibition respectively [28, 30]. We confirmed that all SNPs were in Hardy-Weinberg equilibrium.

Use of statin medication was self-reported, and only concerned medication that is regularly taken. The prescribed statin drugs present in the UK Biobank cohort were simvastatin, atorvastatin, fluvastatin, rosuvastatin and pravastatin. However, majority of the statin users (93%) were taking lipophilic statins (i.e., atorvastatin, simvastatin and fluvastatin) and very few (7%) were taking hydrophilic statins (i.e., rosuvastatin, pravastatin). Moreover, of the lipophilic statin users, 79.0% used simvastatin, 20.7% used atorvastatin and only 0.3% used fluvastatin at the first assessment in the UK Biobank cohort.

Information regarding insomnia symptoms had been collected as part of a touch-screen questionnaire. The participants had to answer the question “Do you have trouble falling asleep at night, or do you wake up in the middle of the night?” by choosing one of the three semi-quantitative options: (1) “never/rarely”, (2) “sometimes” or (3) “usually”. For our study, we recoded this information into a binary variable by grouping together “never/rarely” and “sometimes” and treated them as controls as done previously [31], while participants reporting “usually” were treated as those having insomnia.

Information regarding chronotype was collected through a question at baseline by asking “Do you consider yourself to be?” and the participants can choose from the answers “definitely a ‘morning’ person”, “more a ‘morning’ than an ‘evening’ person”, “more an ‘evening’ than a ‘morning’ person” or “definitely an ‘evening’ person”.

The analysis included a set of covariates, which were chosen based on their link to sleep behaviour and cardiovascular health [31, 32] (Supplementary Table 1). In brief, age (years), sex (male/female), BMI (weight/(height)${}^2$, calculated from the measurement of body weight (kg) and height (cm)), physical activity (number of days/week of moderate physical activity 10+ minutes; self-reported), sleep duration (hours/day; self-reported), total cholesterol and LDL-cholesterol levels (mmol/L), Townsend deprivation index (TDI), comorbidities (diabetes, hypertension, angina, heat attack), additional medication use (insulin, hypertension), alcohol consumption (six answer options: never, special occasions only, one to three times a month, once or twice a week, three or four times a week, daily or almost daily; self-reported), smoking status (three answer options: never, previous, current; self-reported), first ten genetic principal components and UKB test centre. Education was recoded as binary depending on whether or not the participants were holding a university/college degree.

The current analyses were conducted using the IBM Statistical Package for Social Science software (SPSS) version 26 (IBM Corp., Chicago, IL, USA). The analyses were restricted to the baseline data for self-reported insomnia and chronotype, the above-mentioned SNPs, statin treatment and covariates from the UK Biobank. For all the variables wherever applicable, the answers “I don’t know” and “Prefer not to answer” were recoded as missing. The genotypes were coded as 0, 1, and 2 (based on the number of risk alleles) and included in regression models as predictor (independent) variable. Insomnia was treated as a dichotomous trait (0 representing “controls” and 1 representing “insomnia cases”) and was tested via binary logistic regression. Chronotype was treated as categorical (1–4 for the respective four answer options) or dichotomous (“definite morning person” set to control and “definite evening person” set to case). The threshold for significance was further adjusted to account for multiple testing, i.e., p 0.025 (0.05/2 parameters).

We first examined if the genetic variants of HMGCR and PCSK9 were associated with lower total cholesterol and LDL-cholesterol levels as well as lower incidence of heart attack in our study population, by using linear regression and logistic regression, respectively. Then, we compared the prevalence of insomnia and the chronotype in the whole study population by using covariate-adjusted binary and multinomial logistic regression models. Both unadjusted and adjusted analyses were conducted and the results are reported. Analyses were adjusted for age, sex, sleep duration, total cholesterol level, LDL cholesterol level, BMI, smoking, alcohol consumption, heart attack, angina, hypertension, diabetes, education, physical activity, hypertensive medication, insulin use, TDI, first ten genetic principal components and UKB test centre. The covariates were selected for their link with the sleep related outcomes and cardiovascular disease. The association of insomnia with covariates was tested by employing chi-square test for categorical variables (sex, education, alcohol intake, smoking, comorbid conditions, medications and chronotype) and linear regression for continuous variables (age, TDI, physical activity, sleep duration, BMI, total cholesterol and LDL cholesterol). Moreover, a possible interaction effect of statin treatment and the presence of one of the SNPs was investigated using a multiplicative term in the logistic regression models adjusted for the aforementioned covariates.

Lastly, as majority of the statin users (93%) used lipophilic statins, we additionally conducted the analysis for association of statin use with insomnia and chronotype after excluding the hydrophilic statin users (n = 2739). Since we observed no significant difference in the findings, the results based on entire sample (i.e., including both hydrophilic and lipophilic statin users) are presented throughout the manuscript.

Insomnia, defined as having trouble falling asleep at night or waking up in the middle of the night on a regular basis, was reported by 30.5% of individuals in our study population (mean age $\pm$ SD 57.5 $\pm$ 7.9 years). The prevalence was higher in women compared to men (33.8% versus 26.2%, respectively).

As we described in a previous study [22], statin users differed from controls in a range of health-related parameters (Table 1). In the current study, statin users had 0.95 mmol/L lower LDL cholesterol levels compared to controls [95% CI, 0.94 to 0.96] in the unadjusted analyses and after the adjustment with age and sex, the circulating LDL cholesterol was further decreased to 1.0 mmol/L [95% CI, 0.99 to 1.01]. With respect to sleep, statin users more often reported insomnia symptoms compared to controls, the unadjusted odds ratio (OR) was 1.15 [95% CI, 1.12 to 1.17; p = 2.89 $\times$ 10${}^{-41}$] and the adjusted OR was 1.07 [95% CI, 1.03 to 1.11; p = 1.42 $\times$ 10${}^{-4}$] (Fig. 2A).

We first validated that the HMGCR and PCSK9 gene variants were associated with lower circulating cholesterol as well as a lower incidence of coronary artery diseases in our study population, as shown in other studies [27, 28, 33]. The demographic characteristics of individuals carrying different HMGCR and PCSK9 gene variants are shown in Supplementary Table 2 and Supplementary Table 3.

Each additional HMGCR rs17238484-G allele was associated with 0.047 mmol/L lower LDL-cholesterol levels [95% CI, 0.040 to 0.053; p, 2.50 $\times$ 10${}^{-45}$] and 0.057 mmol/L lower total cholesterol levels [95% CI, 0.049 to 0.065; p = 3.16 $\times$ 10${}^{-40}$] after adjustment with age and sex. The risk for heart attack was lowered with each additional allele by 7.0% [95% CI, 0.89 to 0.98; p = 0.004] after adjustment with age, sex, LDL and total cholesterol [27, 33]. Similarly, each additional HMGCR rs12916-T allele was associated with 0.056 mmol/L lower LDL-cholesterol levels [95% CI, 0.050 to 0.061; p = 3.67 $\times$ 10${}^{-86}$] and 0.069 mmol/L lower total cholesterol levels [95% CI 0.061 to 0.076; p = 5.4 $\times$ 10${}^{-78}$] as well as a lower risk for heart attack (OR [95% CI], 0.93 [0.90 to 0.97]; p = 0.001).

Similarly, the PCSK9 rs11591147-T allele was associated with 0.29 mmol/L lower LDL-cholesterol [95% CI, 0.26 to 0.31; p = 1.30 $\times$ 10${}^{-158}$] and 0.34 mmol/L lower total cholesterol levels [95% CI, 0.31to 0.37; p = 9.92 $\times$ 10${}^{-133}$]. Moreover, each additional rs11591147-T allele was also associated with 38.0% lower risk for heart attack [95% CI, 0.52 to 0.74; p = 4.13 $\times$ 10${}^{-8}$].

After validating our genetic instruments, we analysed the association of rs17238484, rs12916 and rs11591147 genotypes with insomnia symptoms. Interestingly, gene variants in HMGCR and PCSK9 were both associated with insomnia, but in a differential manner (Fig. 2B‒D, Table 2). We observed that each additional rs17238484-G allele was associated with 2.0% lower odds for insomnia [95% CI, 0.96 to 0.99; p = 0.014] (Fig. 2B). Similarly, the rs12916-T allele was associated with 2.0% lower odds for insomnia [95% CI, 0.96 to 0.99; p = 0.002] (Fig. 2C). In contrast, we observed that each additional PCSK9 rs11591147-T allele was associated with 8.0% higher odds for insomnia [95% CI, 1.02 to 1.14; p = 0.014] (Fig. 2D).

Fig. 2.

Association of statin use, HMGCR rs17238484 and rs12916 as well as PCSK9 rs1159147 alleles with self-reported insomnia symptoms. Odds ratios of insomnia with (A) statin use and (B‒D) different copies of respective gene variants of HMGCR (B: rs17238484, C: rs12916 and D: rs1159147) were analysed via binary logistic regression, adjusted for age, sex, sleep duration, plasma lipid concentration, BMI, smoking, alcohol consumption, heart attack, angina, hypertension, diabetes, education, physical activity, hypertension medication, insulin medication, Townsend deprivation index, first ten genetic principal components and UK Biobank test centre. The threshold for significance after correction for multiple testing was p 0.025. CI, confidence interval.

We did not observe any association of either statin treatment or HMGCR gene variants with chronotype (Fig. 3A‒C). In contrast, each additional copy of the PCSK9 rs11591147-T allele was associated with 15.9% greater odds for evening preference [95% CI, 1.06 to 1.30; p = 0.001] (Fig. 3D). Moreover, when the analyses were restricted to compare only definitely evening preference with definitely morning preference, PCSK9 rs11591147-T allele was associated with 17% greater odds for evening preference [95% CI, 1.05 to 1.29; p = 0.003] (Table 3).

Fig. 3.

Association of statin use, HMGCR rs17238484 and rs12916 as well as PCSK9 rs1159147 alleles with self-reported chronotype. Odds ratios of the strength of self-reported morning or evening preference (mild or definite) with (A) statin use and (B‒D) different copies of respective gene variants of HMGCR (B: rs17238484, C: rs12916) and PCSK9 (D: rs1159147) were analysed via multinomial logistic regression, adjusted for age, sex, sleep duration, total plasma cholesterol, LDL cholesterol, BMI, smoking, alcohol consumption, heart attack, angina, hypertension, type 2 diabetes, education, physical activity, hypertension medication, insulin medication and Townsend deprivation index, first ten genetic principal components, UK Biobank test centre. The threshold for significance after correction for multiple testing was p 0.025 (significant p values are written in bold letters). CI, confidence interval.

No significant interaction was observed between statin treatment and genetic variation in HMGCR or PCSK9 genetic variants on the risk of having insomnia. Only a nominal significant interaction effect was observed between statin treatment and HMGCR rs12916-T (the allele with the greater positive effect on insomnia symptoms) on insomnia symptoms [P${}_{i\mathit{}n\mathit{}t}$ = 0.025] (Table 2).