Introduction
Cardiovascular diseases (CVD) are the leading cause for overall mortality worldwide (1). One major contributing factor, besides smoking, age, sex, and blood pressure, in this context is overweight or obesity, respectively (2). To date, numerous studies have reported positive associations between overweight or central fat distribution and risk of fatal or non-fatal cardiovascular events (CVE) (3) as well as atherogenic progression (4). It has been suggested that hyperinsulinemia, which promotes the development of obesity by inhibiting lipolysis (5) and inducing lipogenesis (6), is a further major independent contributing factor for the development of CVD's and CVE's (7–9). In this context, recently published genetic studies indicated for the first time a causal link between elevated insulin levels and CVD risk (10, 11). Furthermore, several mechanistic (12) and observational studies (13) have shown positive associations between elevated insulin levels and an increase in cardiovascular and all-cause mortality in different populations. Particularly, the results of mechanistic studies indicate an atherogenic effect of persistently high insulin levels, probably via endothelial dysfunction (14), a pro-inflammatory activity of macrophages (15), suppression of autophagy and an increase in protein synthesis (16), as well as a compromised cytoprotective response to oxidative and other chemical stress (17).
So far, primarily cross-sectional studies (18–21) investigated the interrelation of insulin levels with the occurrence of CVD and CVE (22, 23) or atherogenic progression (9, 19, 24), respectively. Furthermore, prospective studies are rather rare (25), and the generalizability of findings of the examined populations is somewhat reduced [e.g., Pima Indians (26), older-aged persons with higher fasting insulin levels (27)]. Moreover, existing publications in this context are characterized by only small cohorts (28–32), and the major part of these studies focusses on insulin resistance (e.g., homeostatic model assessment of insulin resistance (HOMA-IR) index) and/or glucose levels (33–37) as potential glucometabolic predictors, and insulin levels in blood are only of inferior priority. However, the relevance of HOMA-IR as measure of total body, peripheral or hepatic insulin resistance has been questioned (38).
Therefore, more longitudinal studies are needed to investigate the impact of insulin levels compared with HOMA-IR on atherogenic progression. In the present longitudinal study, we have investigated the incidence of increased intima media thickness (IMT) and arteriosclerosis as precursors of atherogenic progression in a prospective occupational health care setting in middle-aged employees. This allowed us to evaluate metabolic and other parameters as risk factors of arterial lesions.
Materials and Methods
Study Design and Population
Details of the open-label Boehringer Ingelheim (BI) employee cohort have been described previously (39). In short, BI employees aged ≥40 and employed for at least 2 years were offered to participate in a comprehensive health check program of the BI corporate medical department every 3–5 years free of charge. As part of an occupational health care initiative this program is still ongoing. For the present analysis data from those participants who joined the program between 01.01.2006 and 31.12.2017 were analyzed. The study was approved by the ethics committee of the Ärztekammer Nordrhein (no. 2011340). All clinical investigations have been conducted according to the principles of the Declaration of Helsinki and participants gave written informed consent.
Outcomes
Anthropometrical [body-mass-index (BMI), weight], clinical (systolic and diastolic blood pressure), laboratory [triglycerides, total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), hemoglobin A1c (HbA1c), fasting blood glucose, fasting insulin, homeostasis model assessment of insulin resistance (HOMA-IR)], and behavioral (smoking) parameters were determined as previously reported (39).
Cardiometabolic diseases were defined as previously described [hypercholesterolemia (≥200 mg/dl), hypertension (systolic blood pressure: ≥140 mmHg and/or diastolic blood pressure: ≥90 mmHg or self-reported), hypertriglyceridemia (triglycerides: ≥150 mg/dl), insulin resistance (HOMA-IR: ≥2.6), hyperinsulinemia (fasting insulin: >15 μU/ml) and type 2 diabetes mellitus (fasting blood glucose: ≥126 mg/dl and/or HbA1c ≥ 6.5%)] (39).
Arteriosclerosis [occurrence of plaques in abdominal arteries (aorta abdominalis) and/or neck arteries (aorta carotis) and/or stenoses in neck arteries (aorta carotis)] and increased IMT (>1 mm) were determined as early markers of cardiovascular disease.
Statistical Analysis
Shown are means with standard deviation (SD) [mean (SD)] for normally distributed variables or median and interquartile ranges for variables with skewed distribution. Dichotomous variables were analyzed by Chi-square test or McNemar test, as appropriate.
Multivariable logistic regression analyses were performed to investigate the influence of anthropometric, clinical, and laboratory parameters on the incidence of increased IMT or arteriosclerosis expressed as odds ratio (OR) with 95% confidence interval (CI). Logistic regression analyses were adjusted for the baseline values age, sex, smoking, BMI, fasting insulin, fasting blood glucose, systolic blood pressure, diastolic blood pressure, HDL-C, LDL-C, total cholesterol, triglycerides, HbA1c, HOMA-IR (Model 1) or “BMI change” during follow-up (Model 2= Model 1 + “BMI change”) or “insulin change” during follow-up (Model 3= Model 1 + “insulin change”), or “insulin change” and “BMI change” during follow-up (Model 4= Model 1 + “BMI change” + “insulin change”). Model quality was analyzed by evaluating the area under the curve of the receiver operating characteristic curve.
In subanalyses, based on the assumption that BMI and fasting insulin are possible major contributing factors for the incidence of increased IMT and arteriosclerosis, both variables were divided into quartiles and changes from baseline to follow-up were parted in tertiles. Furthermore, relative risk (RR) and adjusted relative risk (ARR) for increased IMT and arteriosclerosis were calculated. ARR was analyzed due to Mantel-Haenszel analysis stratifying for significant covariables. RR analyses were based on the strongest predictors identified due to the logistic regression analysis. For this purpose, fasting insulin, BMI and age were recoded into dichotomous variables defined as hyperinsulinemia (fasting insulin: >15 μU/ml), overweight (BMI: ≥25 kg/m2) and older age (>44.5 years).
Level of significance was set at α = 0.05 from two-sided tests. Statistical analyses were performed using GraphPad Prism 6.04 (GraphPad Software, San Diego, CA, USA) and SPSS 22.0 (SPSS Inc., Chicago, IL, USA).
Discussion
In the present longitudinal study middle-aged employees were followed up in average for 5 years as part of an occupational health care program. As expected, there was incidence or progression of atherogenic lesions, defined by increased IMT and atherosclerosis. Multivariable logistic regression models confirmed common risk factors at baseline such as age, sex, smoking, as well as BMI. We identified fasting insulin blood levels at baseline as a further strong predictor of atherogenic progression. Moreover, comprehensive modeling revealed that in addition insulin changes over time have a significant association with the incidence of increased IMT and arteriosclerosis. Fasting insulin is controversially discussed as a reliable marker for predicting cardiovascular events as not all long-term follow-up studies identified fasting insulin levels as a critical factor for developing CVD (40). Differences between the studies may account for the inconsistency. Welin et al. (follow-up for 8 years) included older aged men (n = 595) with high normal or elevated fasting insulin levels (27). Hargreaves et al. (follow-up for 12 years) considered only data of 83 persons in their final logistic regression models of whom 11 developed coronary heart disease (32). Liu et al. did not find associations of fasting insulin concentrations with incident electrocardiogram abnormalities in Pima Indians (n = 47 cases in 994 study participants) (26). In smaller cohorts, the authors did not find associations of fasting insulin levels with the progression of carotid intima media thickness (72 adolescent study participants for 2 years) or in 374 young adults after 13 years, as well as in 84 postmenopausal women after 5 years of follow-up, respectively (28–30). The pronounced obesity-related hyperinsulinemia and insulin resistance in Pima Indians may have obscured the association of fasting insulin with markers of CVD risk (41). Another aspect is that we find the increase of fasting insulin levels during follow-up as particularly strong risk factors for endothelial lesions, and this parameter has not been analyzed in these studies.
There are also confirmative studies to the present findings. Strong associations of fasting insulin with a 5-year progression of carotid atherosclerosis (i.e., IMT, plaque presence) have been shown for middle-aged European and Asian persons (31). Furthermore, high-dosages of exogenic induced insulin associated with an increasing CVD risk in older hospitalized patients with type 2 diabetes in an observational study (22). Moreover, patients with atrial fibrillation and insulin-requiring diabetes had an increased thromboembolic risk (stroke/systemic embolism) at 1 year compared to those patients with diabetes but without insulin therapy (23). There are also cross-sectional studies demonstrating significant correlations of fasting insulin levels with different entities of early arteriosclerosis such as carotid artery stiffness in 83 middle-aged patients with hypertension, or increased carotid IMT in 100 healthy middle-aged persons as well as coronary calcification in 443 middle-aged men and women (18, 20, 21).
High fasting insulin levels and hyperinsulinemia, respectively, may promote atherogenesis by inducing endothelial dysfunction (by reducing the activation of endothelial NO synthase) (14), pro-inflammatory activity of macrophages (15), suppression of autophagy and an increase in protein synthesis (e.g., by upregulating the mechanistic target of rapamycin complex 1) (16), as well as a compromised cytoprotective response to oxidative and other chemical stress (by suppressing the nuclear factor Nrf2) (17).
In the present study BMI also significantly contributed to the risk of developing increased IMT and arteriosclerosis. These results are in line with the current literature as BMI is often used in predicting models as a prognostic factor for CVD progression and the incidence of CVE's (42). Besides age, sex, and smoking (1), obesity has been discussed as a major factor for the development of atherosclerosis (43) and hyperinsulinemia and/or insulin resistance are seen as accompanied consequences of excessive weight gain. However, based on the present results, especially when considering changes of fasting insulin and BMI over time, higher fasting insulin levels exhibit a stronger association with increased IMT and arteriosclerosis than BMI. Adjusting for changes of fasting insulin blood levels during follow-up eliminated BMI as risk factor but not vice versa. These findings are similar to a meta-analysis of hyperinsulinemia vs. risk of cardiovascular mortality in men and women with type 2 diabetes which also reported an association independent of BMI (44). Based on the current findings, the broadly accepted causal model between obesity, the resulting hyperinsulinemia, and atherogenic progression is obsolete. Moreover, fasting insulin levels should be also evaluated at least in a similar fashion in comparison to BMI from a clinical perspective for preventing cardiovascular diseases. Furthermore, the current results support the potential role of high insulin levels for atherogenic progression and that maybe obesity is rather a consequence than the actual risk factor for developing hyperinsulinemia and in the further course arteriosclerosis (8).
In the present study, multivariable logistic regression modeling identified smoking as an additional independent risk factor for developing increased IMT and arteriosclerosis. Interestingly, hyperinsulinemia was stronger associated with arteriosclerosis than smoking. Further adjusted relative risk analyses found that relative risks conferred by smoking and hyperinsulinemia were not additive. This finding indicates that smoking and hyperinsulinemia target the same atherogenic mechanism. Hyperinsulinemic persons tend to have an insulinotropic effect by smoking (45) and chronic smokers with (46) or without diabetes (47) have a significantly reduced insulin sensitivity which in turn led to hyperinsulinemia.
Furthermore, HOMA-IR did not associate with the incidence of increased IMT or arteriosclerosis in the present study. Studies supporting insulin resistance as a risk factor for atherogenic progression differ in part significantly to the present study, especially in case of population size (n <100) or age (≥ 70 years) (33–35), study design (e.g., cross-sectional design) (37), follow-up period (1 year) (48), and fasting insulin levels were not treated as a confounder in the statistical analyses. Interestingly, Hanley et al., the only study which equally investigated HOMA-IR and fasting insulin with a predictive approach (follow-up after 8 years), found that, both, HOMA-IR and fasting insulin were comparable predictors (with similar OR's in comparison to the present study) for the incidence of cardiovascular events in middle-aged persons of the San Antonio Heart Study (49). Possible reasons for the difference to the present study might be ethnical (> 50% Hispanic population) or behavioral (e.g., > 25% active smoker) as well as the primary endpoint of the study (incidence of precursors of CVD vs. incident cardiovascular events). A recently published observational and cross-sectional study with a large cohort (n = 5,764) indicates that insulin resistance not consistently associates with the presence of coronary artery disease and that hyperglycemic conditions can have a significant impact in this context (36). The variable findings with regard to HOMA-IR may be due to the weak correlation with direct measures of insulin sensitivity (e.g., euglycaemic–hyperinsulinaemic clamp) (38).
In the present study, fasting blood glucose did also not correlate with the incidence of increased IMT or arteriosclerosis. Other works demonstrated similar results that postprandial glucose levels rather than fasting blood glucose associate with atherogenic progression (50). This interrelation supports the assumption that higher fasting insulin levels or a hyperinsulinemic status rather than elevated fasting blood glucose levels mediate the development of arteriosclerosis (51). Furthermore, fasting blood glucose, triglycerides, HDL cholesterol, as well as diastolic blood pressure did not deteriorate despite the overall BMI increase of 0.3 kg/m2 (≈ 1 kg). However, subanalyses revealed that participants experiencing an increase in body weight above 1 kg had expected worsening of triglycerides, fasting blood glucose, HDL cholesterol, and diastolic blood pressure levels. A possible explanation for this data course is that a small increase in BMI of ≤ 0.3 kg/m2 over 5 years in a younger working cohort does not inevitably lead to a deterioration of clinical parameters (Supplementary Table 2).
There are strengths and limitations of the present study which should be considered. Although the results are based on a large number of people, it must be taken into account that these are all middle-aged persons. Focusing on this particular cohort might lead to a selection bias as persons in this age tend to have less prevalent endothelial lesions. However, the longer observation period revealed a substantial rise in the number of persons with increased IMT and arteriosclerosis. Furthermore, the results of the present study describe associations and cannot prove causal relationships. Moreover, it needs to be considered that information about medication was not recorded (e.g., antidiabetic, antihyperlipidemic, antihypertensive drugs). Besides medication, there is also missing information regarding ethnic background as well as a detailed recording of comorbidities (e.g., polycystic ovary syndrome, chronic viral hepatitis, chronic renal failure, fatty liver disease, adrenal and pituitary diseases, infectious and oncological diseases).
Strengths of the present work comprise the observation period of 5 years and the large number of persons followed up.
Taken together, fasting insulin levels should be considered in the assessment and prognosis of atherosclerotic progression in addition to the common risk factors age, sex, smoking, and BMI. From a clinical perspective, HOMA-IR appears to be of inferior relevance for predicting precursors of cardiovascular disease. Furthermore, more attention should be paid to the prevention or treatment of hyperinsulinemia because of the associated health risks. One possibility would be the recommendation of a lifestyle favoring low insulin levels as part of general prevention guidelines.