Several trials and cohort studies have shown an increased incidence of type 2 diabetes mellitus (T2DM) in patients using statins. Whether this only applies to patients at already high risk for the development of T2DM or for all patients is still a matter of debate. In the present prospective cohort study of 4,645 patients with established vascular disease without DM at baseline, 3,057 patients used statins at baseline, of whom 1,608 used intensive statin therapy, defined as statin therapy theoretically lowering low-density lipoprotein cholesterol with ≥40%. Cox proportional hazards models were used to estimate the risk of incident T2DM with (intensive) statin therapy. Statin therapy was associated with increased risk of incident T2DM (hazard ratio 1.63; 95% confidence interval 1.15 to 2.32) when adjusted for age, gender, body mass index, plasma high-density lipoprotein cholesterol, and plasma triglyceride levels. Intensive statin therapy tended to be related to a higher risk of T2DM compared with moderate statin therapy (hazard ratio 1.22; 95% confidence interval 0.92 to 1.61, adjusted for age, gender, body mass index, plasma high-density lipoprotein cholesterol, and plasma triglyceride levels). The increase in risk was regardless of the number of metabolic syndrome characteristics or insulin resistance but was particularly present in patients with low baseline glucose levels (<5.6 mmol/L; p for interaction 2.9 × 10 −7 ). In conclusion, statin use increases the risk of incident T2DM in patients with clinically manifest vascular disease. The increase in risk was independent of the number of metabolic syndrome criteria and was even more pronounced in patients with low baseline glucose levels.
Statins are a safe and effective therapy to reduce the risk of first or subsequent vascular events, and their use is recommended for all patients at high risk of vascular events. However, recent studies in various populations have indicated an increased risk of development of type 2 diabetes mellitus (T2DM) with statin treatment. Whether only patients who are already at high risk to develop T2DM are prone to acquire T2DM because of statin therapy or whether statin therapy also conveys T2DM risk in patients who are considered to be at low risk to develop T2DM remains a matter of debate. Results of the Treating to New Targets (TNT) and Incremental Decrease in Endpoints through Agressive Lipid Lowering (IDEAL) studies indicate that only patients at already elevated risk for T2DM according to previously determined risk factors in their study (resembling metabolic syndrome criteria) are at increased risk to develop T2DM by statin therapy, but in the Justification for Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER) study, the risk of T2DM with statin therapy was independent of baseline glucose levels. For clinical practice, identification of risk factors to develop T2DM with statin therapy is of importance to guide individual therapy. This is particularly relevant for patients with established arterial disease, who are not only in need of low-density lipoprotein cholesterol (LDL-c)–lowering therapy but who are also at elevated risk of developing statin-induced T2DM. In the present study, we investigated the relation between statin therapy and the intensity of statin therapy and incident T2DM in a cohort of patients with clinically manifest vascular disease. Furthermore, we investigated whether the effect of statin therapy on incident T2DM was dependent on the presence of risk factors for T2DM.
Methods
Data were used from patients enrolled in the Second Manifestation of Arterial Disease (SMART) cohort. This is a prospective, ongoing cohort study at the University Medical Center Utrecht, The Netherlands, designed to study the presence of concomitant arterial diseases and risk factors for atherosclerosis in a high-risk population. Patients newly referred to our institution with clinically evident vascular disease or a vascular risk factor (hyperlipidemia, hypertension, or diabetes) were asked to participate. Written informed consent was obtained from all patients. The Medical Ethics Committee of the University Medical Center Utrecht approved the study. After inclusion, all patients underwent a vascular screening protocol including a health questionnaire, laboratory measurements, and physical examination. A detailed description of the study design has been published previously. For the present study, data were used from 4,645 patients enrolled in the SMART study from September 1996 to March 2011 who did not have DM at baseline, with either a history or a recent diagnosis of clinically manifest arterial disease: coronary artery disease, cerebrovascular disease, peripheral artery disease, or aneurysm of the abdominal aorta. Patients could be classified into >1 disease category. Patients who had died (n = 383) or were lost to follow-up (n = 67) before the assessment of T2DM in 2006 were excluded. Single imputation methods were used to reduce missing covariate data for body mass index (n = 8, 0.2%), plasma glucose levels (n = 30, 0.6%), systolic blood pressure (n = 21, 0.4%), total cholesterol (n = 24, 0.5%), high-density lipoprotein cholesterol (HDL-c; n = 35, 0.7%), triglycerides (n = 32, 0.7%), estimated glomerular filtration rate (n = 22, 0.5%), and smoking (n = 21, 0.5%) because complete case analysis leads to loss of statistical power and bias.
Type and dose of lipid-lowering therapy were registered for all participants at baseline. To compare the intensity of different types of statins, the percentage of LDL-c reduction per individual type and dose of statin therapy were determined, based on (systematic) reviews and meta-analyses to the efficacy of statins and other lipid-lowering drugs. For the present study, intensive statin therapy was defined as statin therapy–lowering LDL-c with ≥40%. This implies that, for example, pravastatin and fluvastatin are not considered intensive statin therapy at any dose, whereas all doses of rosuvastatin are considered intensive statin therapy. For atorvastatin and simvastatin, this will depend on the dose used, with ≥20 mg atorvastatin and ≥40 mg simvastatin being intensive statin therapy. Statin use during follow-up was known for 1,093 patients. We used a logistic regression model to estimate statin use during follow-up for all participants for whom no follow-up information about statin use was present. Patients who started or stopped treatment during follow-up were assumed to have used statin treatment for half of the follow-up time. For subjects without follow-up information, we used the estimates from the logistic regression model. This resulted in a continuous scale for statin therapy, 0 meaning that no statin therapy was used at any time during the study and 1 meaning that statin therapy was used at all time during the study.
All study participants received a questionnaire every 6 months during the follow-up period to obtain information about hospitalizations and outpatient clinic visits. All available relevant data from any reported possible event were collected. To assess the incidence of diabetes, all patients who had been included until June 2006 without diabetes at baseline received a questionnaire in the period from June to December 2006 to assess the incidence of T2DM after study inclusion. After 2006, information about diabetes was biannually collected in the questionnaire. Patients reporting diabetes received a supplementary questionnaire regarding date of diagnosis, initial and current treatment (oral medication or insulin), and family history of diabetes. Patients and/or their general practitioners were contacted by telephone for further information if the answers were incomplete or unclear and also nonresponders were contacted. All diabetes cases were audited and classified by 2 independent physicians. Cross-validation with the hospital diagnosis registry revealed that none of the patients who reported not to have diabetes had a physician’s diagnosis of diabetes. Follow-up duration (years) was defined as the period between study inclusion and the date of incident T2DM, death from any cause, date of loss to follow-up, or the preselected date of March 1, 2012. From 1996 until March 1, 2012, 120 of the 4,645 patients (2.6%) were lost to follow-up.
The effect of statin use on incident T2DM was evaluated using Cox proportional hazards models. In model 1, adjustment was made for age and gender. In model 2, additional adjustment for body mass index, plasma triglyceride level, and plasma HDL-c level was made. Furthermore, a propensity score was constructed because patients using statins differ from patients not using statins, and these differences themselves may influence the risk to develop DM. To make this propensity score, the probability to use statin therapy was estimated using a logistic regression model, including baseline characteristics (except current LDL-c levels). In model 3, we adjusted for this propensity score. To compare the effect of different types and different intensities of statin therapy, we stratified these analyses by type and intensity. Pravastatin or fluvastatin and atorvastatin or rosuvastatin were grouped together because the group sizes were too small, and these types of statin therapy have shown similar effects on T2DM risk. Type and dose of statin treatment were not available during follow-up; therefore, these analyses only compare baseline differences. Furthermore, we stratified by the number of metabolic syndrome criteria present as an estimate of T2DM risk to evaluate whether statin therapy increases T2DM risk independent of baseline T2DM risk. To further investigate this, we also stratified for quartiles of each of the metabolic syndrome criteria. We used the triglyceride and glucose index (TyG index = Ln[fasting triglycerides {in mg/dl} × fasting glucose {in mg/dl}/2]) as an estimate of insulin resistance because the Homeostatis Model Assesment – Insulin Resistance (HOMA-IR) could only be calculated for patients included from July 2003 onward because of the absence of insulin measurements before July 2003. To test for interaction, that is, whether the relation between statin therapy and incident T2DM was modified by the metabolic syndrome, insulin resistance, or by any of the individual metabolic syndrome criteria, we included an interaction term in the Cox model. Furthermore, we also tested for interaction with age and gender. The proportional hazard assumption for the Cox model was tested using scaled Schoenfeld residuals, confirming proportional hazards. The statistical package R 2.13 (R Development Core Team, Vienna, Austria) was used for all analysis. A p value <0.05 was considered statistically significant.
Results
Baseline characteristics according to baseline statin use are listed in Table 1 . The mostly used statins in the patients using moderate statin therapy were pravastatin (36%) and (low dose) simvastatin (43%). In patients using intensive statin therapy, mostly used statins were (high dose) simvastatin (51%) and atorvastatin (34%). Patients using no statin therapy were, in general, included earlier in the study than patients using moderate statin therapy or intensive statin therapy. Furthermore, patients with cerebrovascular disease or peripheral artery disease were less likely to use statin therapy at baseline than patients with coronary artery disease.
| Variable | Statin therapy | ||
|---|---|---|---|
| No (n=1588) | Moderate (n=1449) | Intensive (n=1608) | |
| Age (years) | 58.4 ± 11.3 | 59.7 ± 10.2 | 59.0 ± 9.7 |
| Men | 1092 (69%) | 1092 (75%) | 1231 (77%) |
| Body mass index (kg/m 2 ) | 26.1 ± 3.9 | 26.8 ± 3.7 | 27.0 ± 3.7 |
| Plasma glucose (mmol/L) | 5.7 ± 0.8 | 5.8 ± 0.9 | 5.8 ± 0.8 |
| Systolic blood pressure (mmHg) | 141 ± 22 | 139 ± 21 | 139 ± 21 |
| Total cholesterol (mmol/L) (mg/dL) | 5.6 ± 1.2 217 ± 46 | 4.8 ± 1.0 186 ± 39 | 4.3 ± 1.0 166 ± 39 |
| LDL-cholesterol (mmol/L) (mg/dL) | 3.6 ± 1.0 139 ± 39 | 2.8 ± 0.8 108 ± 31 | 2.4 ± 0.8 93 ± 31 |
| HDL-cholesterol (mmol/L) (mg/dL) | 1.26 ± 0.40 49 ± 15 | 1.24 ± 0.35 48 ± 14 | 1.25 ± 0.37 48 ± 14 |
| Triglycerides (mmol/L) (mg/dL) | 1.4 (1.0-2.1) 124 (89-186) | 1.4 (1.0-2.0) 124 (89-177) | 1.2 (0.9-1.8) 106 (80-159) |
| Estimated GFR (mL/min/1.73 m 2 ) ∗ | 77 ± 18 | 76 ± 16 | 77 ± 17 |
| HsCRP (mg/L) † | 3.3 ± 3.0 | 2.6 ± 2.6 | 2.4 ± 2.5 |
| Current smoker | 638 (40%) | 408 (28%) | 504 (31%) |
| Metabolic syndrome | 710 (45%) | 746 (52%) | 770 (48%) |
| Triglyceride and glucose index ‡ | 8.8 (8.4-9.2) | 8.8 (8.4-9.1) | 8.6 (8.3-9.0) |
| Use of antiplatelet agents | 935 (59%) | 1184 (82%) | 1436(89%) |
| Use of blood pressure-lowering agents | 842 (53%) | 1225 (85%) | 1303 (81%) |
| Type of statin: | |||
| Pravastatin | 0 (0%) | 514 (36%) | 0 (0%) |
| Fluvastatin | 0 (0%) | 68 (5%) | 0 (0%) |
| Cerivastatin | 0 (0%) | 7 (1%) | 0 (0%) |
| Simvastatin | 0 (0%) | 628 (43%) | 813 (51%) |
| Atorvastatin | 0 (0%) | 232 (16%) | 547 (34%) |
| Rosuvastatin | 0 (0%) | 0 (0%) | 248 (15%) |
| Localisation of vascular disease: | |||
| Coronary artery disease | 609 (38%) | 1076 (74%) | 1123 (70%) |
| Cerebrovascular disease | 555 (35%) | 286 (20%) | 469 (29%) |
| Peripheral arterial disease | 446 (28%) | 196 (14%) | 187 (12%) |
| Abdominal aortic aneurysm | 168 (11%) | 96 (7%) | 80 (5%) |
| Mean time since inclusion in the cohort (years) | 7.7 ± 3.9 | 6.3 ± 3.5 | 3.8 ± 2.9 |
∗ Glomerular Filtration Rate, estimated by the Modification of Diet in Renal Disease (MDRD) equation.
‡ Triglyceride and glucose index calculated as Ln[fasting triglycerides(in mg/dL) x fasting glucose (in mg/dL)/2].
During follow-up, 353 patients developed T2DM, the overall incidence of T2DM was 1.29 per 100 patient-years. Statin therapy increased the risk of incident T2DM with 63% when adjusted for age, gender, body mass index, plasma HDL-c, and plasma triglyceride levels ( Table 2 ). This risk persisted when adjusted for a propensity score.
| Events/N | Incidence per 100 person year | HR Model 1 | HR Model 2 | HR Model 3 | |
|---|---|---|---|---|---|
| Statin use vs. no statin use | 353/4645 | 1.29 | 1.71 (1.22-2.41) | 1.63 (1.15-2.32) | 1.66 (1.14-2.42) |
| Intensive statin vs. usual dose | 212/3057 | 1.37 | 1.18 (0.90-1.57) | 1.22 (0.92-1.61) | 1.21 (0.90-1.63) |
| Simvastatin vs fluvastatin/ pravastatin | 137/2023 | 1.31 | 1.20 (0.84-1.73) | 1.26 (0.88-1.82) | 1.25 (0.86-1.81) |
| Atorvastatin /rosuvastatin vs fluvastatin/ pravastatin | 117/1609 | 1.38 | 1.38 (0.94-2.03) | 1.32 (0.90-1.93) | 1.41 (0.95-2.13) |
Baseline use of intensive statin therapy compared with usual dose statin therapy tended to be related to an increased risk of T2DM, although this finding was not statistically significant. When types of baseline therapy used at baseline were compared, simvastatin and especially atorvastatin or rosuvastatin confer a higher risk of developing T2DM than pravastatin or fluvastatin. However, this trend was only weak for baseline use of simvastatin and of borderline significance for baseline use of atorvastatin or rosuvastatin.
Because incident T2DM was initially not registered in this cohort, we performed sensitivity analyses by excluding patients who were included >7 years ago or by adjusting for years since inclusion in the cohort. This led to similar results. The increased risk of T2DM with statin therapy was independent of the number of metabolic syndrome criteria present ( Table 3 ). The p for interaction by number of metabolic syndrome criteria was 0.91. When stratified according to TyG index, a measure of insulin resistance, the risk of T2DM with statin therapy was elevated in the upper TyG index quartile. However, the p for interaction by TyG index was not statistically significant (p = 0.62).
| Parameter | Events / N | Incidence per 100 person year | HR Model 1 | HR Model 2 |
|---|---|---|---|---|
| Number of metabolic syndrome criteria | ||||
| 0-2 | 72/2416 | 0.50 | 1.84 (0.88-3.83) | 1.65 (0.79-3.47) |
| 3 | 92/1188 | 1.33 | 1.51 (0.76-2.99) | 1.33 (0.67-2.65) |
| 4 | 121/762 | 2.77 | 1.58 (0.86-2.93) | 1.57 (0.85-2.92) |
| 5 | 68/276 | 4.59 | 1.28 (0.55-2.94) | 1.62 (0.67-3.95) |
| Triglyceride and glucose index ∗ | ||||
| Quartile 1 (<8.37) | 12/1147 | 0.22 | 1.28 (0.23-7.29) | 1.23 (0.21-7.42) |
| Quartile 2 (8.37-8.72) | 48/1145 | 0.70 | 1.67 (0.68-4.11) | 1.11 (0.49-2.80) |
| Quartile 3 (8.72-9.10) | 84/1171 | 1.15 | 1.45 (0.72-2.91) | 1.34 (0.67-2.67) |
| Quartile 4 (≥9.10) | 209/1181 | 2.79 | 2.16 (1.35-3.45) | 2.37 (1.46-3.83) |
∗ Triglyceride and glucose index calculated as Ln[fasting triglycerides(in mg/dL) x fasting glucose (in mg/dL)/2].
When the variables included in the metabolic syndrome were considered separately, the effect of statin therapy was similar in increasing quartiles of each of the variables ( Table 4 ). The only exception was glucose level, with a larger increase in the risk of T2DM with statin use for patients with low baseline glucose levels compared with patients with higher baseline glucose levels. The p for interaction by baseline glucose level was 2.9 × 10 −7 , adjusted for age, gender, body mass index, plasma triglycerides, and HDL-c. When adjusted for age, gender, and a propensity score interaction term, this interaction remained statistically significant (p = 0.006). In an exploratory analysis to explain this finding, the increased risk to develop T2DM with statin treatment seemed confined to patients with low baseline glucose levels but a high TyG index. No effect modification by age or gender was present (p for interaction 0.35 and 0.93, respectively, adjusted for age, gender, body mass index, plasma triglycerides, and HDL-c).
