Discussion
In this longitudinal study using the repeated measurements of fasting glucose level in nationwide health examinations, there was a statistically significant association between statin therapy and increase in fasting glucose in non-diabetic individuals (more adherent and intensive statin users had higher fasting glucose). Non-statin lipid lowering medications of fibrate and ezetimibe were not associated with change in fasting glucose. When we analyzed individual statin types, all types of statins were positively associated with an increase in fasting glucose although the effects of pravastatin, lovastatin, and fluvastatin were not significant. Our findings are in line with the current consensus that statin can induce insulin resistance, hyperglycemia, and new-onset DM [
17,
18]. However, our study should not be interpreted as to merely avoid statins due to the hyperglycemic effect. Overall evidence show that the benefits of statins far outweigh the potential hazards [
19‐
21]. Statin therapy was associated with less good glycemic control in diabetes and pre-diabetes, but there was a much lower risk of major cardiovascular events [
22,
23]. Real-world studies consistently indicate that statins are frequently suboptimal and under prescribed in populations with high cardiovascular risk [
24‐
27].
When we plotted estimated change in fasting glucose over time (Fig.
2), more adherent and intensive use of statins had positive effects on the change in fasting glucose level, but there was no significant interaction between statins and time interval. That means that use of statins may induce an increase in fasting glucose, but the slope of increase in fasting glucose over time was not associated with the statin therapy. Therefore, we supposed that there is no reason to avoid statins for high risk cardiovascular patients due to the concern for additional hyperglycemia over time induced by long-term statin therapy. This finding was consistent with a prior meta-analysis evaluating the change in hemoglobin A1c (HbA1c) by statins in DM patients; there was no significant interaction between the duration of statin treatment and increase of HbA1c [
28]. In addition, we found that the increase in fasting glucose was influenced by an unhealthy lifestyle of smoking, higher alcohol consumption, and lower physical activity. Higher systolic blood pressure, higher BMI, and lower household income were also associated with increase in fasting glucose. These findings support the current recommendations highlighting the importance of lifestyle modifications to prevent impaired glucose tolerance and DM [
29].
With expanded indications for statin in the recent guidelines, statins have become the first-line therapy for cardiovascular protection [
30]. As more patients take statins, there is awareness of unwanted side effects as well as benefits of statins [
31]. Due to there are pathophysiologic and epidemiologic data for a linkage between hyperglycemia and increased cardiovascular or all-cause mortality in diabetic and non-diabetic patients, physicians should be cautious about the potential risk of statin-mediated hyperglycemia and new-onset DM [
32,
33]. To maintain the cardiovascular benefit and minimalize the potentially adverse hyperglycemic effect with statins, multi-strategy approach should be considered, including monitoring of fasting glucose, control of predisposing factors to glucose intolerance and unhealthy life-style, proper selection of the statin (type and dosage) and concomitant medications [
21,
34]. Screening and risk stratification for diabetes on the initiation of statins and periodical monitoring of fasting glucose would be helpful for early diagnosis and proper glucose control in patients on statin therapy [
33,
35]. Sustainable lifestyle changes can improve control of both glucose metabolism and cardiovascular risk [
36]. Encouraging healthy dietary pattern, physical activity, body weight control, and quitting smoking could be effective to prevent impaired glucose tolerance and DM [
29,
37]. There are a number of medications with diabetogenic property such as thiazide diuretics and steroids; concomitant use of the medications may further increase the risk of impaired glucose metabolism and DM [
38,
39].
To explain the diabetogenic effect of statins, numerous mechanisms have been proposed. Glucose and lipid metabolism are interconnected in many ways [
40]. Gene variants involved in lipid metabolism (Niemann-Pick C1-Like 1, Proprotein convertase subtilisin/kexin type 9, and HMGCR) are significantly associated with obesity, hyperglycemia and DM [
41‐
43]. Lower prevalence of type 2 DM in patients with familial hypercholesterolemia suggests genetic links between glucose and lipid metabolism [
44]. Statins-induced cholesterol-dependent conformational changes in glucose transporter (GLUT) proteins can lead to impaired glucose uptake in the cells and insulin resistance [
45]. Inhibition of de-novo cholesterol synthesis by statins results in deleterious inflammation and oxidation within islet β-cells, which lead to cellular apoptosis and impaired insulin secretion [
46]. Biosynthesis of the isoprenoid side chain of coenzyme Q10 is suppressed by statins, which has been implicated in the downregulation of GLUT4 synthesis, mitochondrial oxidative stress, delayed adenosine triphosphate production, and apoptosis of β-cells, causing impaired glucose-stimulated insulin secretion [
19,
47,
48]. Simvastatin and atorvastatin have been shown to reduce the level of adiponectin, which is a hormone with anti-inflammatory and anti-diabetogenic properties secreted by adipocytes [
19,
49]. One of the most common complaints with statins is muscle-related symptoms (muscle pain, myopathy, myalgia, and fatigue) [
50]. Statin-related muscle complaints are frequently exacerbated by exercise, which further limits physical activity and exercise performance [
51]. Since physical inactivity is associated with abdominal adiposity, body weight gain, and glucose intolerance, limited physical activity by statins may be a mediator of glucose intolerance [
23,
52,
53].
There are scarce data on the effects of individual statins on DM and diabetogenic potency. Previous meta-analyses with randomized and observational studies have shown no clear difference between statin types in terms of DM incidence [
5,
54]. The Women’s Health Initiative showed that an increased risk of DM was observed for all types of statins [
55]. Therefore, all statins seem to have a diabetogenic potency considered as a medication class effect [
17]. However, individual statins have different mechanisms of action and pharmacokinetic properties, and experimental data suggest that different statins have varying effects on glucose metabolism [
56]. Based on available clinical and experimental data, it is generally accepted that atorvastatin, rosuvastatin, simvastatin, lovastatin, and fluvastatin have an unfavorable influence on glycemic parameters and risk of new-onset DM [
57].
Some data suggested that pravastatin may have relatively weak diabetogenicity compared to that of other statins [
3]. In a network meta-analysis, pravastatin had the lowest risk of inducing new-onset DM among the high-dose statin therapy (atorvastatin 80 mg, rosuvastatin 20 mg, simvastatin 40 mg, and pravastatin 40 mg), but the difference was not statistically significant [
58]. The West of Scotland Coronary Prevention Study showed that pravastatin therapy resulted in a 30% reduction in the risk of new-onset DM compared to placebo [
59]. The relatively lower risk of DM with pravastatin was also reported in retrospective studies using the real-world data [
60,
61]. Unlike other statins, pravastatin may increase plasma adiponectin levels and improve insulin sensitivity mediated by the elevation of calcitriol [
62]. The hydrophilicity of pravastatin may have little effect on membrane-embedded proteins involving glucose metabolism such as the GLUT [
63]. However, the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER) trial showed that the risk of new-onset diabetes increased by 30% in patients treated with pravastatin compared with those given a placebo [
64].
Pitavastatin is the newest statin and has been shown to be well-tolerated with fewer side effects and a relatively low drug interaction profile [
65]. A recent meta-analysis of randomized controlled trials with pitavastatin suggested that pitavastatin did not adversely affect glucose metabolism or increase risk of developing diabetes compared to the controls [
66]. In the Japan Prevention Trial of Diabetes by Pitavastatin in Patients With Impaired Glucose Tolerance (J-PREDICT) trial, treatment with pitavastatin reduced the risk of new‐onset DM by 18% in patients with impaired glucose tolerance [
67]. The Japanese long-term prospective post-marketing surveillance LIVALO Effectiveness and Safety (LIVES) study demonstrated that treatment with pitavastatin lowered HbA1c (from 8.1% at baseline to 7.4% at 6 months) in the patients with poorly controlled diabetes [
68].
In our study, all classes of statins were associated with an increase in fasting glucose although the effects of pravastatin, lovastatin, and fluvastatin were not significant. In our data, the hyperglycemic effects of pitavastatin does not seem to be lower than that of other statins in contrast to the favorable reports for pitavastatin on glucose metabolism. A recent retrospective cohort study in Korea also reported that pitavastatin had the highest risk of new-onset DM compared to other statins [
69].
In a number of prior studies, there are inconsistent findings about the diabetogenic property of individual statins. The discrepancy could be originated from difference in baseline characteristics of study population, coexisting risk factors, follow-up duration, medication adherence and so on. It is known that statin-induced DM risk is higher in pre-diabetes and patients with predisposing factors for DM such as old age and obesity [
34]. The short-term and long-term effects on glucose metabolism may be different in accordance with type of statins and underlying characteristics of patients [
70]. The diabetogenic effect of statins may be varied in genetic background [
71]. In patients who were prescribed statins, discontinuation and poor adherence were common in clinical practice, which limits assessing the diabetogenic effect of statins [
72]. The effect of statins on glucose metabolism are complex and interconnected by multiple pathways; individual statins may have a heterogenous impact on the multiple processes [
11]. To obtain further knowledge of individual statins on glucose metabolism, there is need for further extensive studies.
This study had both strengths and limitations. From a population-based nationwide health screening program, we collected data from a large cohort of 379,865 patients over a 10-year study period. Serial data of more than 2,000,000 measurements of fasting glucose were analyzed. Besides data on fasting glucose, we also included detailed data on blood pressure, BMI, and lifestyle. The significant effects of an unhealthy lifestyle on fasting glucose including smoking, higher alcohol intake, higher BMI and lower physical activity may support the relevance of this study. In Korea, statins must be prescribed by physicians and refill programs at pharmacies are not allowed. Using the health insurance claim data, prescription records for individual statin could be accessed. However, the actual intake of statins in subjects might be different from the prescription records. Fortunately, there was good correlation between prescription and real exposure to drugs in prior studies [
73,
74]. Our study was retrospectively performed and due to the observational nature, there was a possibility of hidden bias due to uncollected data. We excluded a large proportion of subjects with DM or who had received anti-diabetic medication during the study period to eliminate the potential bias caused by the effects of this medication on glucose level. The item for physical activity lacked data for the type and duration of exercise. This study was based on Koreans; the response to statins might be varied in other genetic population. Therefore, our results should be interpreted with caution and further studies are needed to confirm these findings in other populations.