Simvastatin is associated with superior lipid and glycaemic control to atorvastatin and reduced levels of incident Type 2 diabetes, in men and women, in the UK Biobank

Abstract Introduction Cardiovascular disease (CVD) is the leading cause of mortality in people with Type 2 diabetes mellitus (T2DM). Statins reduce low‐density lipoproteins and positively affect CVD outcomes. Statin type and dose have differential effects on glycaemia and risk of incident T2DM; however, the impact of gender, and of individual drugs within the statin class, remains unclear. Aim To compare effects of simvastatin and atorvastatin on lipid and glycaemic control in men and women with and without T2DM, and their association with incident T2DM. Methods The effect of simvastatin and atorvastatin on lipid and glycaemic control was assessed in the T2DM DiaStrat cohort. Prescribed medications, gender, age, BMI, diabetes duration, blood lipid profile and HbA1c were extracted from Electronic Care Record, and compared in men and women prescribed simvastatin and atorvastatin. Analyses were replicated in the UKBiobank in those with and without T2DM. The association of simvastatin and atorvastatin with incident T2DM was also investigated in the UKBiobank. Cohorts where matched for age, BMI and diabetes duration in men and women, in the UKBioBank analysis, where possible. Results Simvastatin was associated with better LDL (1.6 ± 0.6 vs 2.1 ± 0.9 mmol/L, p < .01) and total cholesterol (3.6 ± 0.7 vs 4.2 ± 1.0 mmol/L, p < .05), and glycaemic control (62 ± 17 vs 67 ± 19 mmol/mol, p < .059) than atorvastatin specifically in women in the DiaStrat cohort. In the UKBiobank, both men and women prescribed simvastatin had better LDL (Women: 2.6 ± 0.6 vs 2.6 ± 0.7 mmol/L, p < .05; Men: 2.4 ± 0.6 vs 2.4 ± 0.6, p < .01) and glycaemic control (Women:54 ± 14 vs 56 ± 15mmol/mol, p < .05; Men, 54 ± 14 vs 55 ± 15 mmol/mol, p < .01) than those prescribed atorvastatin. Simvastatin was also associated with reduced risk of incident T2DM in both men and women (p < .0001) in the UKBiobank. Conclusions Simvastatin is associated with superior lipid and glycaemic control to atorvastatin in those with and without T2DM, and with fewer incident T2DM cases. Given the importance of lipid and glycaemic control in preventing secondary complications of T2DM, these findings may help inform prescribing practices.


| INTRODUC TI ON
However, recent evidence indicates that statin therapy is associated with increased drug interactions, poor diabetes outcomes and potential worsening of glycaemic control in those on high-dose statin therapy. 4 This is also supported in a retrospective cohort study of 12,725 participants from the health improvement network (THIN) where concurrent statin use was associated with higher HbA1c levels, after initiation of insulin, throughout a 3-year follow-up, relative to those not in receipt of statin therapy. 5 As a drug class, statins increase the risk of developing T2DM by 10%-12%. 6 However, this effect appears to be heavily influenced by statin type and dose. A network meta-analysis of 163,039 participants revealed high-dose atorvastatin increased the odds of developing diabetes compared with low-dose atorvastatin. 7 In addition to increasing the risk of new-onset T2DM, statins have also been shown to alter the function of insulin-secreting beta cells and to increase insulin resistance, suggesting a potential diabetogenic effect for the drug class. 6 Disturbances in insulin and glucose homeostasis significantly increase the potential for major adverse cardiac events. 8 More work is required to differentiate statins with respect to their effect on blood lipids and glycaemic control, and little research has been done on how gender can influence response. The aims of this study were as follows: (1) To characterize the DiaStrat T2DM cohort in Northern Ireland (NI), in terms of comorbidities influencing glycaemic control and gender differences in lipid and glycaemic control in response to the most widely prescribed statins, simvastatin and atorvastatin. (2) To expand the research question to the UK Biobank T2DM population to assess the generalizability of the findings (as the UK Biobank includes participants from England, Scotland and Wales (but not NI)). (3) To utilize the UK Biobank dataset to investigate lipid and glycaemic control associated with statin prescription in those without diabetes and (4) To further investigate the association with incident diabetes.

| The DiaStrat cohort
The stratified medicine optimizing treatment for diabetes (DiaStrat) study is a pilot observational study. A total of 500 adults aged between 18 and 80 years, with clinically diagnosed T2DM were enrolled in the study from diabetes clinics in the Western Given the importance of lipid and glycaemic control in preventing secondary complications of T2DM, these findings may help inform prescribing practices.

K E Y W O R D S
gender differences, glycaemic control, HbA1c, lipid control, statin, type 2 diabetes, UK Biobank

Novelty Statement
• Statins have secondary effects on glycaemia and diabetes onset risk; it is unknown if this differs between men and women.
• In DiaStrat, we found women, not men, have superior lipid and glycaemic control in response to simvastatin compared with atorvastatin.
• Validation analysis using the UK Biobank cohort indicated that men and women with T2DM achieve superior lipid and glycaemic control with simvastatin than atorvastatin.
• Simvastatin prescription reduced the risk of incident T2DM in the UK Biobank compared with atorvastatin in men and women.
• These findings may inform prescribing practices, with respect to statins, in those at-risk of and with existing T2DM.
setting. Relevant clinical information for all participants was obtained from the Northern Ireland Electronic Care Record (NIECR; Orion health) at the date of recruitment, for this cross-sectional study. Data collected included gender, date of birth (DOB), age at diagnosis, biochemical lab values, all prescription data and recorded comorbidities. Data were not available for all variables, and the revized 'n' is indicated in Tables/Figures. Three hundred and seventy four participants provided a blood sample, from which plasma was extracted, permitting analysis of c-peptide.
The present study focused on comorbid endocrine disorders in the DiaStrat cohort, which were primarily associated with lipid abnormalities, as this comorbidity was associated with inferior glycaemic control (

| UK biobank analyses
The UK Biobank (ukbiobank.ac.uk) has approximately 500,000 participants, aged 40-69 years, recruited between 2006 and 2010, from the general population of the United Kingdom. 9 We replicated part of our DiaStrat analyses using the UK Biobank by extracting data for participants with a confirmed diagnosis of T2DM (using date ICD-10 code E11 first reported, Field-ID 130708), prescribed either simvastatin or atorvastatin (n = 16,257).
For incidence T2DM, we report baseline characteristics of those who developed T2DM after baseline, as follow-up biochemical analyses were not available for such participants. The effect of simvastatin (n = 42,816; 25,593 men and 17,223 women) and atorvastatin (n = 10,241; 6310 men and 3931 women) prescription on blood lipids and HbA1c was also assessed in the absence of diabetes (ICD-10 codes E10 and E11, Field-ID 41270, Table 6).

| C-peptide enzyme-linked immunosorbent assays (ELISA)
Plasma c-peptide was measured for n = 374 participants of the DiaStrat cohort using human Alpco c-peptide ELISA kit (Alpco; Cat no. 80-CPTHU-E01.1, E10), according to the manufacturer's instructions. No c-peptide data were available for UK Biobank participants.

| Blood lipid measurement
High-density lipoprotein (HDL) cholesterol, total cholesterol and triglycerides where measured via direct laboratory assay (Cobas C-701 analyser). Low-density lipoprotein (LDL) cholesterol was calculated using the Friedewald formula.

| Statistical analysis
Biochemical changes in the DiaStrat analysis were determined in SPSS version 25 using two-way ANOVA with Bonferroni post hoc analysis. Analysis of the UK Biobank dataset was carried out in the open source software, R (https://www.R-proje ct.org/). The UK Biobank fileset was loaded in the R environment using 'ukbtools' package (https://kenha nscom be.github.io/ukbto ols/). Means, standard deviations, two-sample T-tests and one-way ANOVAs were calculated within the base R package. Significance threshold was set at p < .05 for all analysis.
In order to control for confounding variables, multivariate analysis was utilized in addition to simple bivariate analysis ( Figure 1).
We created logistic regression models to determine adjusted odds ratios associated with age, T2DM duration, BMI, blood lipids, HbA1c and c-peptide and statin prescription (with simvastatin as 1-class and atorvastatin as 0-class), as the method is ideal for our dataset.. 10 Adjusted odds ratios (ORs), 95% confidence intervals and significance are reported in Tables 3-7.
F I G U R E 1 Odds ratio of incident T2DM in men and women from the UK Biobank, without T2DM at baseline, prescribed simvastatin or atorvastatin. Individuals prescribed simvastatin and atorvastatin, without a diagnosis of T2DM, were identified at baseline within the UK Biobank cohort. Incident T2DM was noted when an individual received a T2DM diagnosis after the UK Biobank recruitment date (UK Biobank Field id 130708). ****p < .0001 vs men prescribed simvastatin. △△△△ p < .0001 vs women prescribed simvastatin  Table 2. Biguanides were the most common diabetes drug class (n = 375, 73%), and statins were the most common non-diabetes prescription. Atorvastatin was the most frequently prescribed non-diabetes medication (n = 273, 54%).

| Diagnosis of a lipid abnormality in the DiaStrat cohort, particularly in women, is associated with increased HbA1c
There were 217 participants diagnosed with hyperlipidaemia (Table 3). At a cohort level bivariate analysis revealed, those with a diagnosed lipid abnormality were older (64 ± 9 vs 60 ± 11 years, p < .0001), had increased duration of T2DM (15 ± 7 vs 9 ± 7 years, (p < .001) and reduced LDL (1.7 ± 0.8 vs 2.0 ± 0.8, p < .01) compared with those without a lipid abnormality ( Note: DiaStrat represents a cohort of T2DM participants recruited from secondary care clinics in northern Ireland. Total number of values available per variable (total), characteristic mean values ± standard deviation (mean (SD)), and percentage of total (%) are illustrated for the complete cohort, men and women. *p < .05, ***p < .001 and ****p < .0001 compared with men or women.
Women without a lipid abnormality had increased HDL compared with the total cohort with (p < .001) and without (p < .001) a lipid abnormality and men with (p < .001) and without (p < .01) a lipid abnormality. Similarly, LDL was increased in women with a lipid abnormality compared with the total cohort with a lipid abnormality (p < .01) and to diagnosed men (p < .01). Total cholesterol was highest (4.2 ± 0.9 mmol/L) in women without a lipid abnormality and significantly increased compared with the total cohort with a lipid abnormality (p < .01) and with diagnosed men (p < .001). HbA1c was lower in women without a diagnosed lipid Ezetimibe Cholesterol absorption inhibitor 33 7 Note: DiaStrat represents a cohort of T2DM participants recruited from secondary care clinics in northern Ireland. Diabetes drug class, quantity prescribed and percentage of total cohort prescribed each diabetes drug class (A), and; drug name, drug class, quantity prescribed and percentage of total cohort prescribed non-diabetes drugs (B).
abnormality than the combined cohort with a lipid abnormality (p < .05) and men (p < .05) and women (p < .01) with a diagnosed lipid abnormality. In contrast, women with a diagnosed lipid abnormality had the highest recorded HbA1c (70 ± 18 mmol/ mol [8.6%]), which was significantly higher than the combined cohort without a lipid abnormality (p < .01) and women without TA B L E 3 Diagnosis of a lipid abnormality, particularly in women, is associated with increased HbA1c in the DiaStrat cohort

| Simvastatin is associated with reduced HbA1c in UK Biobank participants without diabetes
In men without a diabetes diagnosis, multivariate analyses revealed that there was no difference in HDL, LDL or total cholesterol between those prescribed simvastatin and atorvastatin; however, triglycerides (p < .05) and HbA1c (p < .0001) were reduced in men prescribed simvastatin compared with atorvastatin (Table 6). In women, there was also a significant difference in LDL (p < .01), total cholesterol (p < .05), triglycerides (p < .01) and HbA1c (p < .0001), associated with simvastatin compared with atorvastatin, as illustrated in Table 6.

| Simvastatin is associated with fewer incident T2DM cases than atorvastatin in UK Biobank participants
Baseline characteristics of individuals prescribed simvastatin and atorvastatin who developed T2DM after initial recruitment (blood sample collection) to the UK Biobank, are illustrated in Table 7.
Multivariate analyses revealed that HbA1c was only significantly TA B L E 5 Characteristics of men and women with T2DM from the UK Biobank prescribed simvastatin and atorvastatin   Figure 1).
Furthermore, the odds ratio associated with incident T2DM in men was significantly greater than for women for both simvastatin OR in both men and women.
In 2008, rosuvastatin was linked to increased diabetes risk in the JUPITER study, and the link between statin treatment and glycaemic control has been of concern since. 10 It is established that high HbA1c is correlated with elevated lipids. 11 This often translates into patients being prescribed high-intensity statins, such as atorvastatin. 12 Studies have linked statin therapy to diabetes onset and have highlighted high-dose therapy and LDL level as the main factors influencing diabetes incidence. 13 The CARDS study found that atorvastatin negatively affected HbA1c in participants with diabetes, but found no effect in a simvastatin-treated group. 13 Consistently, we show that both atorvastatin and simvastatin reduce LDL and total cholesterol in men and women, but simvastatin was associated with lower HbA1c than atorvastatin. The effect of statins on glycaemic control is controversial with prior reports of beneficial effects, 14 negative effects 4 or no impact at all. 15 The mechanisms behind the differential effects of statins are not well understood, particularly in relation to the opposing effects of atorvastatin and simvastatin on glucose metabolism. 16 Most studies show atorvastatin reduces LDL and total cholesterol levels without influencing blood glucose in individuals with diabetes. 17 The effect of simvastatin appears to be dependent on dose or the cohort profile. In those with T2DM and hypercholesterolemia, simvastatin doses of 80 mg/day result in a ~10% increase in plasma glucose after 2 months, 18 whilst lower dose (<20 mg/day) statins have less impact on glycaemia. 19 Other work has reported that 20 mg/day simvastatin negatively affects insulin sensitivity but has no effect on insulin or glucose levels after 4 weeks. 20  Note: Prescription information was extracted from data obtained at the initial recruitment appointment for the UK Biobank. All participants with a confirmed diagnosis of T1DM or T2DM (ICD-10 code E11, field id 41270) were excluded. * p < .05, ** p < .01, **** p < .0001 vs men prescribed atorvastatin. effect on insulin sensitivity, fasting insulin levels or HOMA-B levels.
Hydrie and colleagues 22 reported people with T2DM and insulin resistance showed improved glycaemic control after receiving 40mg/ day simvastatin for 3 months, which is consistent with our observations. Our study also supports the VYTAL 23 study's finding that simvastatin has greater lipid-lowering effects than atorvastatin and the VOYAGER 24 study's observation that simvastatin causes a superior (2.1% greater) reduction in LDL in women than men.
We utilized the UK Biobank to assess if the sex specific effect of simvastatin on lipid control and HbA1c was evident in the wider UK population with a T2DM diagnosis. In the UK Biobank analysis, simvastatin was associated with superior lipid control in both men and women. The differences observed may be attributed to dosing inconsistencies. 12 In the DiaStrat cohort, both simvastatin and atorvastatin were prescribed at an average dose of ~40 mg/day, which is considered high-intensity therapy. 12 The standard dose of atorvastatin is 10 mg/day 25 29 In another UK study, with 2 million participants followed over 15 years, statin-associated risk was significant (HR 3.6), with no difference between moderate or intensive therapy. 30 Other work has shown high-intensity statins such as atorvastatin and rosuvastatin are associated with higher risk than moderate intensity statins, such as simvastatin. 19 These findings correspond with the present study which reports reduced incidence of T2DM in the UK Biobank with simvastatin but not atorvastatin, and a higher risk in men compared with women when exposed to either drug.
This analysis has several limitations. DiaStrat recruitment was from one geographical region, and the cohort may represent a severe T2DM phenotype due to the fact that recruitment was TA B L E 7 Characteristics of men and women prescribed simvastatin and atorvastatin with incident T2DM in the UK Biobank   Future studies should not impose an upper age limit, given the relevance of older adults to the topic. This was also the case in the UK Biobank, which excludes those >70 years. This limitation may be addressed in future longitudinal studies as the UK Biobank participants age. The high rate of CVD, obesity and related CVD risk evident in this cohort commonly results in patients starting on intensive therapy. 12 The generalizability of findings, however, was substantiated by follow-up analyses within the UK Biobank, representing a significantly larger cohort. Prescription data within the UK Biobank were obtained via verbal interview, which undoubtedly impacts reliability. Consideration needs to be made that prescription guidelines for DiaStrat would have differed from UK Biobank.
Historically, simvastatin would have been prescribed at a higher dose, over atorvastatin. Prior to 2010, the Joint British Societies' guidelines (JBS 2) aimed for LDL of <2 mmol/L in high-risk individuals rather than the current 40% reduction in non-HDL cholesterol. 32 Furthermore, comorbidities and associated polypharmacy may play a role in our observed findings. There are few longitudinal studies looking at the effect of individual statins and further prospective studies are warranted. A recent longitudinal study (11 years), in non-diabetic patients, has reported that atorvastatin and simvastatin increased the fasting blood glucose. 33 Given the potential importance of reducing incident T2DM and improving glycaemic control in established T2DM, outcomes reported here should be investigated in randomized controlled trials. The present study cannot definitively establish that simvastatin and atorvastatin are responsible for differences in lipids and HbA1c observed, rather that they are associated with this observation. We aimed to address confounding but utilizing a multivariate analysis approach; well-designed prospective studies will determine the reproducibility of our observations and potentially identify other unmeasured confounders.

| CON CLUS ION
In the DiaStrat cohort, simvastatin and atorvastatin were associated with reduced LDL and total cholesterol in T2DM participants, whilst simvastatin was associated with superior glycaemic control in women. In the UK Biobank, superiority of simvastatin over atorvastatin, in terms of glycaemic and lipid control, was observed in both men and women. Furthermore, in individuals without T2DM at baseline, atorvastatin is associated with increased risk of incident T2DM when compared with simvastatin. Whilst causality cannot be established within the present study, our observations suggest that simvastatin is associated with superior lipid-lowering and HbA1c properties in those at-risk of, and diagnosed with, T2DM and may, after confirmatory clinical trials, inform prescribing practices in this population.

ACK N OWLED G EM ENTS
We would like to extend our sincere gratitude to the participants of this study for their invaluable contribution and to the Western Health Social Care Trust, Northern Ireland, for facilitating access to participants and clinics. This study was approved by the Research

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interests to declare.