Abstract
In 1385 adults with primary untreated hyperlipidemia and in a population study of 339 adults (Princeton Follow-up Study [PFS]), we hypothesized that homeostasis model assessment (HOMA) insulin resistance (IR) was a significant explanatory variable for triglycerides (TG) and that IR rose in a stepwise fashion, independent of age, race, sex, and body mass index (BMI), whereas TG categories rose from less than 150 to 150 to 200, to 200 to 500, and to more than 500 mg/dL. A third hypothesis was that TG, BMI, and the ratio of TG to high-density lipoprotein cholesterol (TG/HDL-C) were significant explanatory variables for IR and that IR was inversely associated with HDL-C quintiles and positively associated with non-HDL-C quintiles. By stepwise multiple regression with age, race, sex, BMI, and IR as explanatory variables, in the 1385 patients, positive explanatory variables for TG included BMI (partial R2 = 1.3%, P < 0.0001), sex (men higher, partial R 2 = 1.1%, P = 0.0001), and IR (partial R 2 = 0.4%, P = 0.012). In the 339 PFS subjects, positive explanatory variables for TG were IR (partial R 2 = 11.4%, P < 0.0001), race (whites higher, partial R 2 = 2.1%, P = 0.005), and sex (men higher, partial R 2 = 1.4%, P = 0.019). After adjusting for age, race, sex, and BMI, in 1385 patients, HOMA IR rose while TG categories rose, with least square means of 2.64 for the TG category less than 150 mg/dL, 3.27 for 150 to 200 mg/dL, 3.85 for 200 to 500 mg/dL, and 4.12 for more than 500 mg/dL. Similarly, in the PFS, while TG categories rose, the least square means of HOMA IR rose, with 1.68 for the TG category less than 150 mg/dL, 2.34 for 150 to 200 mg/dL, and 3.03 for 200 to 500 mg/dL. Body mass index, TG, and TG/HDL-C were significant explanatory variables for IR. Homeostasis model assessment IR is a significant, potentially reversible explanatory variable for TG in patients referred because of hyperlipidemia and in population subjects.
Insulin resistance (IR) has an effect on the modulation of plasma insulin, triglyceride (TG), and high-density lipoprotein cholesterol (HDL-C) concentrations, independent of generalized abdominal or physical endurance capacity.1In whites, a TG level of 130 mg/dL or higher and a TG to HDL-C ratio of 3 or higher predict IR in individuals with a body mass index (BMI) of 25 kg/m2 or higher, but in African Americans,2TG levels and TG/HDL-C ratio are not reliable markers of IR. Nonfasting3-5and fasting hypertriglyceridemia (HTG)6-9is an independent risk factor for coronary heart disease (CHD), particularly for women. In the prospective Bruneck population study, HOMA IR independently predicted incident cardiac events.10
The National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III)11recently identified TG levels lower than 150 mg/dL as normal, 150 to 200 mg/dL as borderline high, 200 to 500 mg/dL as high, and 500 mg/dL or higher as very high. In the current report, we studied 1385 patients referred for diagnosis and treatment of primary untreated hyperlipidemia, 816 patients with a TG level of 150 mg/dL or higher, and a free-living population study of 364 adults (the Princeton Follow-up Study [PFS]). We hypothesized that HOMA IR was a significant explanatory variable for TG9,12-24and that IR rose in a stepwise fashion, independent of age, race, sex, and BMI, whereas TG categories10rose from less than 150 mg/dL to 150 to 200 mg/dL, to 200 to 500 mg/dL, and to more than 500 mg/dL. A third hypothesis was that TG, BMI, and TG/HDL-C ratio were significant explanatory variables for IR and that IR was inversely associated with HDL-C quintiles and positively associated with non-HDL-C quintiles.
MATERIALS AND METHODS
Study Design
The study followed protocols approved by the institutional review board with signed informed consent.
Cases
After an overnight fast, blood was drawn for cholesterol, TG, and HDL-C tests, along with glucose, insulin, renal, thyroid, and liver function tests, in 2392 patients consecutively referred to the Cholesterol Center from September 27, 1987, to June 12, 2007, for diagnosis and treatment of hyperlipidemia (Table 1). A detailed written history was taken regarding prescription drug use, weekly alcohol intake, diabetes, and fasting hyperglycemia (Table 1). Referral to a regional lipid diagnosis and treatment center reflected a selection bias in the patient group to hyperlipidemia. To assess an untreated patient cohort also free of major causes of secondary hyperlipidemia, a subset of 1385 patients was focused on, with exclusions including using statin and/or TG-lowering drugs, a fasting glucose level of 126 mg/dL or higher, 6 or more drinks of alcoholic beverages per week, 10-mg/d or more prednisone, a hemoglobin A1C score of 7 or higher, hypothyroidism, exogenous estrogen use, and uremia (Table 1).
Princeton Family Study Subjects
To assess relationships between IR, TG, and TG/HDL-C in a healthy, free-living population sample, we studied 364 adults from the PFS,25a 25- to 30-year follow-up study of lipids, lipoproteins, and metabolic syndrome in former student participants in the National Heart, Lung, and Blood Institute Lipid Research Clinics Prevalence Study (1973-1976) in a suburban Cincinnati school district. Twenty-two PFS subjects with a fasting blood glucose level of 126 mg/dL or higher and 3 with missing BMI entries were excluded from the data analyses, leaving 339 subjects (Table 2).
Laboratory and Clinical Measures
Throughout the duration of the study, insulin, lipid, and glucose levels were measured after an overnight fast (≥8 hours) by the same methods for the patients and for the PFS subjects. Insulin was measured by a competitive protein-binding radioimmunoassay. Glucose was measured by a glucose oxidase method with the Hitachi 704 Chemistry Analyzer (Roche Diagnostics, Indianapolis, IN). Between-batch coefficients of variation were 9% for insulin and 4% for glucose. Homeostasis model assessment IR,26which correlates with estimates of IR measured by the euglycemic clamp technique, was used as an index of IR.26Although the HOMA IR measure is less accurate than the euglycemic clamp method, in large epidemiological studies, it is a reasonable alternative to the complicated clamp method that requires continuous intravenous administration of insulin and glucose for 3 hours for estimation of insulin sensitivity.27Blood lipids were measured in Centers for Disease Control and Prevention-standardized laboratories.25
Height, weight, and blood pressure were systematically measured. A fasting serum insulin level was identified as high for levels equal to or more than the laboratory 97.5th percentile, blood glucose was identified as high by the American Diabetes Association criteria if it is 126 mg/dL or higher,28and obesity was identified by the criteria of Flegal et al.,29with a BMI of 25 to 30 kg/m2 or higher as overweight, 30 to 40 kg/m2 or higher as obese, and 40 kg/m2 or higher as severe obesity.
Statistical Analysis
All statistical analyses were performed using SAS (version 9.1) (SAS Institute, Cary, NC).
Sample size calculations were based on the primary hypothesis of the study that IR and TG were correlated, using the data on correlations of IR to TG (r = 0.267) of Moro et al.30With α = 0.05 and β = 0.8, 107 subjects would be required.
Scatter plots were constructed to assess relationships between HOMA IR and TG in the 1385 patients (Fig. 1) and in the 339 PFS subjects (Fig. 2). Regression lines were displayed (Figs. 1 and 2).
The patients and the PFS population subjects25were categorized into the 4 NCEP ATP III11fasting TG categories (normal, <150 mg/dL; borderline high, 150-200 mg/dL; high, 200-500 mg/dL; and very high, ≥500 mg/dL; Table 6). The patients and the PFS subjects were also characterized by race- and sex-specific HDL-C and non-HDL-C quintiles (Tables 7 and 8). The Jonckheere-Terpstra nonparametric test for ordered differences among classes was used to test the association of HOMA IR with the 4 TG categories (Table 6), the HDL-C quintiles (Table 7), and the non-HDL-C quintiles (Table 8).
Stepwise regression was used with TG as the dependent variable, and age, race, sex, BMI, and HOMA IR were used as explanatory variables in the 1385 patients with untreated primary hyperlipidemia and in the 339 PFS subjects (Table 3).
To assess whether TG or TG/HDL-C were significant explanatory variables for IR and whether TG or TG/HDL-C were surrogate markers for HOMA IR, stepwise regression was used with HOMA IR as the dependent variable, and age, race, sex, TG (or TG/HDL-C), and BMI were used as explanatory variables in the 1385 patients and in the 339 PFS subjects (Tables 4 and 5).
Analysis of variance was used to compare HOMA IR with the TG categories after adjusting for age, race, sex, and BMI (Fig. 3).
To assess the relationships of HOMA IR and TG to BMI categorized by 25 or less, 25 to 30, 30 to 40, and 40 kg/m2 or higher, analysis of variance was used with adjustment for age, race, and sex; least square means and SEs of IR (Fig. 4) and TG (Fig. 5) were displayed.
To control for false positives related to multiple comparisons in Figures 3 to 5, Hochberg-Benjamini31adjusted P values were checked. All of the P values shown in Figures 3 to 5 were significant below the Hochberg-Benjamini threshold.
RESULTS
Patients and PFS Subjects
After exclusions to rule out secondary causes of HTG, and excluding patients with statin and TG-lowering drug use, the Cholesterol Center patient cohort was reduced from 2392 to 1385 patients with primary untreated hyperlipidemia (Table 1). The characteristics of these 1385 patients and the 339 PFS subjects are displayed in Table 2. The patients differed from the PFS subjects for total and low-density lipoprotein (LDL) cholesterol, TG, and fasting serum insulin levels and HOMA IR (Table 2).
In the 1385 patients with primary untreated hyperlipidemia, HOMA IR was positively correlated with TG (r = 0.11, P < 0.0001; Fig. 1). The R 2 for the direct correlation between IR and TG was 0.012, so that a 1.2% variation of TG is explained by IR (Fig. 1). There was a closer IR/TG correlation (r = 0.34, P < 0.0001) in the PFS subjects (R 2 = 0.116), so that an 11.5% variation of TG is explained by IR (Fig. 2).
The least square mean (SE) HOMA IR rose in patients in step with the TG categories, with the highest adjusted IR in patients with a TG level of 500 mg/dL or higher, and in the PFS subjects with a TG level of 200 to 500 mg/dL (Fig. 3). There were only 5 PFS subjects with a TG level of 500 mg/dL or higher.
In the 1385 patients with primary untreated hyperlipidemia, significant explanatory variables for TG included BMI, sex (TG higher in men), and IR (Table 3). In the PFS cohort, IR was a significant explanatory variable, as was race (TG level higher in whites) and sex (TG level higher in men; Table 3). Although statistically significant in the patient group (P = 0.012), the partial R 2 of IR as an explanatory variable for TG was 0.4%, whereas it was 11.4% (P < 0.0001) in the PFS subjects (Table 3).
In the stepwise regression models run separately for men and women, BMI was the only significant explanatory variable for TG in female patients and was the most significant variable in male patients (Table 3). In the PFS cohort, IR was the most significant explanatory variable for TG in women and the only significant variable in men (Table 3).
In both patients and in the PFS cohort, BMI was the major predictive variable for IR, with TG being a significant but much less important explanatory variable (Table 4). When the regression model was run separately for men and women, BMI was the most significant explanatory variable for IR, but TG was a significant but less important variable in male patients and in male and female PFS subjects (Table 4).
With IR as the dependent variable and TG/HDL-C, age, race, sex, and BMI as the explanatory variables (Table 5), BMI was the predominant explanatory variable, with TG/HDL-C also significant but much less important in both the patients and the PFS subjects.
After categorizing by BMI, patients and PFS subjects had a progressive increase in IR as BMI categories rose (after covariance adjusting for age, sex, and race; Fig. 4).
The NCEP ATP III criteria11for an elevated TG level (≥150 mg/dL) captured 815 (59%) of 1385 patients with primary untreated hyperlipidemia and 109 (32%) of 339 of the free-living PFS sample (Table 6). In both the patient and the PFS cohorts, the percentage of subjects with a HOMA IR equal to or more than the 97.5th percentile for healthy subjects in our center (5.54) rose with TG category (Table 6).
After categorizing by BMI, the patients and the PFS subjects had increasing TG levels while BMI categories rose (Fig. 5).
In both the patient and PFS cohorts, the percentage of patients with a HOMA IR equal to or more than the 97.5th percentile for healthy subjects in our center fell as the HDL-C quintile increased (Table 7).
In both the patient and the PFS cohorts, the percentage of patients with a HOMA IR equal to or more than the 97.5th percentile for healthy subjects in our center rose with non-HDL quintiles (Table 8).
DISCUSSION
In our current study, both in patients referred for diagnosis and treatment of primary hyperlipidemia and in healthy free-living subjects, IR was a significant, independent, explanatory variable for TG. However, IR was a much stronger independent predictor of TG (11.4%) in the PFS cohort than in the patient cohort (0.4%). For the direct correlation between IR and TG in patients, 1.2% of the TG was explained by IR compared with 11.5% for the PFS cohort. We speculate that the weaker relationship of IR to TG in the patient cohort reflects other competing variables more likely to be concentrated in the patient cohort including obesity, a significant independent predictor of TG in the patient but not PFS cohort, and genetic contributions to HTG,32-40much more frequent in the patient than in PFS cohort. Because patients and PFS subjects with a fasting serum glucose level of 126 mg/dL or higher were excluded from our analyses, the high IR portion of the present study population is enriched with individuals with high insulin relative to glucose.
In our current study, as TG categories11rose from normal (<150 mg/dL) to borderline high (150-200 mg/dL), to high (200-500 mg/dL), and to very high (>500 mg/dL), IR rose in a stepwise fashion independent of age, race, sex, and BMI. Insulin resistance and TG were significantly correlated. These findings were congruent with previous reports.10,12-24Although correlation does not impute causation, we speculate that the uniform associations of IR with TG in the current study may reflect underlying biological directionality, in agreement with many other studies.1,10,12-24,41-45Moreover, in keeping with putative biological directionality of IR to TG, decreasing IR reduces TG levels.15,16,46-51
In the current study, TG/HDL-C was a significant explanatory variable for IR, consistent with previous reports that IR leads to HTG, low HDL-C, and more small dense LDL molecules, the atherogenic triad.13,41-45Laws and Reaven1reported that IR modulates plasma insulin, TG, and HDL-C concentrations, independent of BMI, waist-to-hip ratio, or physical endurance capacity. Reaven13also reported that IR was associated with increased postprandial accumulation of remnant lipoproteins, elevated plasminogen activator inhibitor activity, and increased levels of adhesion molecules that promote binding of mononuclear cells to the endothelium. Resistance to insulin-mediated glucose uptake has been described as the basic metabolic abnormality both in patients with endogenous HTG and in nonobese subjects with a TG level lower than 175 mg/dL.52Plasma glucose, insulin, and TG increase significantly with each tertile of IR.53
Insulin resistance is associated with high levels of the TG/HDL-C ratio.54Triglyceride55and TG/HDL-C ratio56are surrogate predictors of IR. McLaughlin et al.8,9have reported that TG and TG/HDL-C ratio are closely correlated with IR. In the current study, congruent with previous reports,8,9,55,56TG and TG/HDL-C were significant explanatory variables for IR in both the patients and the PFS subjects after covariance adjusting for race, sex, age, and BMI. However, fasting TG and the TG/HDL-C ratio are not reliable markers of IR in African Americans,57,58probably because of a lack of effect of IR on postheparin lipoprotein lipase that clears TG.58
In 1998, in the Bruneck general population study, Bonora et al.58reported that IR was very common, found in 84.2% of patients with high TG and low HDL-C and in 83.9% of patients with non-insulin-dependent diabetes. After a 15-year follow-up, Bonora et al.10subsequently reported that HOMA IR was associated with development of cardiovascular disease, independent of all classic and several nontraditional CHD risk factors, and suggested that reducing IR may be an important approach to reduce CHD risk. In a cross-sectional study of 293 nondiabetic patients referred for diagnosis and treatment of hyperlipidemia, Glueck et al.59reported that fasting serum insulin independently and uniformly improved the prediction of arteriosclerotic cardiovascular disease status beyond traditional risk factors and lipid variables. Lawlor et al.60have reported independent associations between fasting insulin and stroke and CHD in older women. Lebovitz61emphasized that "insulin resistance was a common link between type 2 diabetes and cardiovascular disease." In patients with the IR syndrome, Reaven13has proposed that IR imparts cardiovascular risk.
The NCEP ATP III criteria11for elevated TG (≥150 mg/dL) captured 59% of 1385 patients with primary untreated hyperlipidemia referred to our center for diagnosis and treatment of hyperlipidemia and 32% of the free-living PFS population sample. This emphasizes the high frequency of HTG in the general free-living population and in patients referred for diagnosis and treatment of hyperlipidemia when the TG cut point11of 150 mg/dL or higher is used.
Patients referred for diagnosis and therapy for HTG, as in the current report, often have concomitant hyperinsulinemia, hyperglycemia, obesity, and low HDL-C, major components of the metabolic syndrome,15all of which need to be treated. The goals of treatment in patients with high TG and concomitant IR are focused on the reduction of cardiovascular disease risk,11,42with concomitant focus on optimal low HDL-C targets.11