Insulin Resistance and the Relation of a Dyslipidemia to Coronary Heart Disease. The Framingham Heart Study

Abstract

OBJECTIVE

To examine the effect of insulin resistance (IR) in subjects without diabetes on the relationship of a dyslipidemia with high triglycerides and low high-density lipoprotein-cholesterol (HDL-C) to the development of coronary heart disease (CHD).

METHODS and RESULTS

Lower and higher fasting plasma HDL-C and triglyceride concentrations (defined at the study population median) and presence or absence of IR (defined by upper quartile Homeostatic Model Assessment values) were related to the development of myocardial infarction or CHD death in Framingham Heart Study participants without diabetes or a history of CHD (n=2910) attending the 1991–95 examination. During follow-up (mean, 14 years), 128 participants experienced an incident CHD event. With Kaplan-Meier plots the incidence of CHD was significantly greater with than without IR at either lowest HDL-C or highest triglycerides (P<0.001). In multivariable Cox models, adjusted for major CHD risk factors including waist circumference, only subgroups with IR had a significantly higher incidence of CHD. Compared to a referent group without IR and higher-than-median HDL-C or lower-than-median triglycerides, the hazard ratio (HR) for incident events was significant with only IR and a lower HDL-C (HR 2.83, P<0.001) or higher triglycerides (HR 2.50, P<0.001). These findings were similar in men and women.

CONCLUSIONS

In this community-based sample exclusive of diabetes, incident CHD risk associated with plasma HDL-C or triglycerides was significantly increased only in the presence of IR.

INTRODUCTION

A dyslipidemia consisting of high triglycerides and low high-density lipoprotein cholesterol (HDL-C) is a widely-recognized lipid pattern that is frequently associated with the development of coronary heart disease (CHD). However, in contrast to plasma low-density lipoprotein-cholesterol (LDL-C), the levels of triglycerides and HDL-C that are associated with increased CHD are less sharply defined and may to a great extent depend on a number of other closely related risk factors that are frequently associated with this dyslipidemia or underlie its origins. Both high triglycerides and low HDL-C are widely known to be associated with obesity and other features that define the metabolic syndrome (MetS)¹. However, it is possible that much of the cardiovascular disease (CVD) that is associated with the MetS may be explained by the presence of insulin resistance (IR)^1,2 and there is compelling, long-standing evidence that both high triglycerides and a low plasma HDL-C are a frequent consequence of IR^2–4.

With IR, studies in man have shown there is increased synthesis and secretion of triglyceride-rich, very low-density lipoprotein (VLDL) particles by the liver⁵, an increase in plasma triglycerides^3–5, enrichment of plasma HDL particles by triglycerides and more active catabolism of triglyceride-rich HDL-C⁶. In population studies IR has been shown to independently predict both the development of a high triglyceride, low HDL-C dyslipidemia⁷ as well as new cardiovascular disease in large general populations^8–11. Reaven and colleagues have long argued that the basis for a high CHD-risk state that is frequently but not invariably associated with obesity is based on the presence of IR and have demonstrated in well-defined, relatively small subgroups that an increase in plasma triglycerides or the ratio of plasma triglycerides to HDL-C is strongly predictive of IR^12,13. In this present analysis we have sought to broaden the scope of these observations by assessing the effect of IR in the population-wide Framingham Heart Study (FHS) on the prevalence of dyslipidemia and the extent to which both low and high levels of triglycerides and HDL-C might predict the development of CHD events in the presence of IR but in the absence of diabetes. In this study we especially sought to determine the effect of IR-related dyslipidemia on the development of CHD independent of other conventional CHD risk factors and, most particularly, independent of abdominal obesity that characterizes the MetS.

METHODS

Study design and population characteristics

The design of the Framingham Offspring Study has been previously described in detail. Briefly, 5124 children of the original FHS and spouses of those children are screened for cardiovascular disease and related risk factors by periodic questionnaires, review of relevant health records, and regular examinations every 4 to 5 years. Data obtained from the Offspring Exam 5 (undertaken from 01/1991 to 06/1995), were used for this analysis. Of the 3,799 participants attending Exam 5, 377 were excluded for prevalent cardiovascular disease, 211 for the presence of diabetes, 137 were excluded for missing clinical or laboratory variables, and an additional 164 were excluded for treatment with a lipid-modifying drug. Exclusion of diabetes was made to avoid a disproportionate contribution of glucose to the HOMA measurement of IR (see below). The final sample for this study included 2,910 participants (1,289 men, 1,621 women). This study was approved by the Boston University Medical Center Institutional Review Board and written consent was obtained from each participant.

Laboratory measurements

Venous blood was drawn after a 12 hour fast. Plasma cholesterol and triglyceride concentrations were measured by enzymatic methods; HDL-C after precipitation of apolipoprotein B-containing lipoproteins with dextran-sulfate magnesium; glucose using hexokinase reagent; and insulin by radio-immunoassay as total immunoreactive insulin (Coat-A-Count Insulin; Diagnostic Products, Los Angeles, CA). A measure of IR, the Homeostasis Model Assessment of IR (HOMA-IR), was calculated as described by Mathews et al¹⁴ and insulin resistance was defined by the upper quartile of HOMA-IR in sex-pooled participants without diabetes.

Outcome Events, Covariate Definitions, and Follow-up

Incident CHD events consisted of the first occurrence of nonfatal or fatal myocardial infarction (MI) or CHD death. Incident events were those that occurred from the time of a Participant’s 5^th FHS Offspring exam to 12/2009 or for a period that averaged 13.5 years for men, 14.3 years for women, and 14.0 years for the entire group.

To obtain comparable-sized groups with lower and higher values of HDL-C and triglycerides, low HDL-C and high triglycerides were defined at the baseline exam to be below and above the median value of each lipid class. Median values for HDL-C in the entire population, men, and women were 49, 42, and 56 mg/dL, respectively, and for triglycerides were 112 mg/dL, 121 mg/dL, and 108 mg/dL, respectively. The covariate of “non-HDL-C” was calculated as the difference between total cholesterol and HDL-C. The covariate of “non-triglyceride-associated cholesterol” was calculated by subtracting from the total cholesterol an average amount of cholesterol approximated by the Friedewald formula¹⁵ to be associated with fasting triglyceride-rich lipoproteins (i.e, very low-density lipoproteins), where cholesterol is approximated to be 20% of the fasting concentration of triglycerides. Other covariates included binary variables for the presence of cigarette smoking as defined by an average of one or more cigarettes per day; alcohol consumption as defined by the number of drinks/week; drug therapy for hypertension; hormone replacement therapy; waist circumference; as well as the continuous covariates of age in years and systolic blood pressure in mmHg.

Statistical Analysis

Demographic and clinical characteristics were compared by gender using a two-sample t-test for continuous variables and Chi-square test for categorical variables. Composite four-level variable models combining the levels of HOMA-IR and the level of HDL-C were created, where categories included (0)=Normal HOMA-IR, High HDL-C, (1)=High HOMA-IR, High HDL-C, (2)=Normal HOMA-IR, Low HDL-C, (3)=High HOMA-IR, Low HDL-C. Similarly, another composite variable was created combining the levels of HOMA with the levels of triglycerides as follows: (0)=Normal HOMA-IR, Low triglycerides, 1=High HOMA-IR, Low triglycerides, 2=Normal HOMA-IR, High triglycerides, 3=High HOMA-IR, High triglycerides. In all analyses, “0” was the referent category. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated for Cox proportional hazard models that were used to predict the risk of CHD with the composite HOMA-IR + HDL-C variable, adjusting first for age only and then for the following covariates: age, systolic blood pressure, hypertension treatment, cigarette smoking, alcohol use, hormone use, waist circumference, and non-HDL-C. Sex-specific Cox proportional hazard models were similarly used to predict the risk of CHD with the composite HOMA-IR + triglyceride variable, adjusting first for age only and then for the covariates shown above for the HOMA-IR + HDL-C variable but substituting non-triglyceride-associated cholesterol for non-HDL-C.

For a secondary analysis with again CHD as an outcome, we used Cox proportional hazard models containing (a) continuous HDL-C and dichotomous HOMA as risk factors; and (b) log-transformed continuous triglycerides and dichotomous HOMA as risk factors (adjustment in each model was made for those same covariates listed above). We further assessed the significance of HDL-C-HOMA and log triglycerides-HOMA interactions, respectively, in each model. Given that in both models the interaction terms approached significance (P ≤0.064), the analysis was then stratified by the level of HOMA, assessing the relationships of continuous HDL-C and continuous triglycerides to predict incident CHD.

Kaplan-Meier plots, unadjusted for covariates, were used to show the cumulative incidence of CHD events by the levels of the HOMA-IR + HDL-C variable and HOMA-IR + triglycerides variable. Log rank tests were performed to assess the difference in the incidence of CHD by the level of these composite variables. Relations between HOMA-IR values and plasma insulin and the continuous variables of age, systolic blood pressure, waist circumference, BMI, and plasma values of glucose, HDL-C, triglycerides, and LDL-C were determined using Pearson correlations.

In all analyses, a two-tailed P-value of <0.05 was considered statistically significant. SAS software Version 9.1 (SAS Institute, Cary, NC) was used for all analyses.

RESULTS

Baseline characteristics of the study group are summarized for men and women in Table 1. Of note, men on average had higher values of BMI, waist circumference, blood pressure, plasma glucose, LDL-C, and triglycerides then women and lower values of HDL-C. Plasma insulin values were higher in men and the frequency of IR, defined by the upper quartile of HOMA-IR, was also greater in men. In this population, continuous values for HOMA-IR were significantly correlated with age (r=0.10), plasma insulin levels (r=0.98), glucose (r=0.52), HDL-C (r=−0.34), triglycerides (r=0.36), LDL-C (r=0.07), systolic blood pressure (r=0.30), BMI (r=0.50) and waist circumference (r=0.49), all with P<0.001.

Table 1.

Population Characteristics at Baseline.

Characteristic	Men (n=1289)	Women (n=1621)	P-value
Age, mean (SD), y	53.8 (9.8)	54.1 (9.8)	0.39
Body mass index, mean (SD), kg/m2	28.0 (4.1)	26.2 (5.2)	<0.001
Waist, mean (SD), cm	98.6 (10.8)	85.6 (13.6)	<0.001
Blood pressure, mean (SD), mm Hg
Diastolic	77.1 (9.7)	72.6 (10.0)	<0.001
Systolic	127.5 (16.5)	121.9 (19.0)	<0.001
Smoking, current, n, %	252 (19.6)	313 (19.3)	0.87
Alcohol use, drinks/week, n (SD)	7.9 (9.6)	3.6 (5.6)	<0.001
Total cholesterol, mean (SD), mg/dL	201.3 (34.4)	205.0 (36.8)	0.007
LDL-C, mean (SD), mg/dL	128.9 (30.9)	122.9 (33.6)	<0.001
HDL-C, mean (SD), mg/dL	44.0 (11.5)	57.3 (15.0)	<0.001
Triglycerides, mean (SD), mg/dL	149.9 (103.6)	123.9 (69.04)	<0.001
Non-HDL-C, mean (SD), mg/dL	157.4 (35.8)	147.6 (38.8)	<0.001
Non-triglyceride-C, mean (SD), mg/dL*	171.4 (34.0)	180.2 (34.0)	<0.001
Fasting glucose, mean (SD), mg/dL	96.9 (8.9)	92.8 (9.6)	<0.001
Fasting insulin, mean (SD), pmol/L	9.9 (8.6)	7.6 (6.6)	<0.001
HOMA-insulin resistance, n, % †	422 (32.7)	320 (19.7)	<0.001
Hypertension therapy, n, %	183 (14.2)	215 (13.3)	0.47
Hormone replacement therapy, n, %		276 (17.1)

Abbreviations: LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; non-triglyceride-C, non-triglyceride cholesterol; HOMA, homeostasis model assessment

^*Non-triglyceride-C defined as cholesterol not associated with fasting triglyceride-rich lipoproteins or as approximated by the Friedewald formula¹⁵ calculated as 20% of the concentration of fasting plasma triglycerides less the concentration of total cholesterol.

^† HOMA-insulin resistance was calculated using the equation of Matthews et al¹⁴ and defined here as the upper quartile of the sex-pooled population without diabetes.

As shown in Figure 1, panel A, across a quartile range of increasing plasma HDL-C concentrations IR was most prevalent in both men and women at the lowest values of HDL-C and least prevalent with HDL-C levels that were highest (P <0.001 for IR trend in both sexes). In contrast, as shown in panel B, with increasing quartiles of triglycerides IR increased in prevalence in both men and women (P <0.001 for IR trend in both men and women). At the lowest quartile of HDL-C or highest quartile of triglycerides approximately 50% of men and 40% of women had IR.

Fig 1.

The prevalence of insulin resistance (IR) by the 4^th quartile of HOMA-IR is shown across quartiles of fasting plasma HDL-C and across quartiles of fasting plasma triglycerides for men and women. Darker shaded bars represent the percentage of men and lighter-shaded bars the percentage of women with IR at each quartile division. Values at the base of each quartile division show the concentration range of HDL-C or triglycerides for that quartile.

During a mean follow-up period of 14 years (maximum 18.4 years), there were 128 new cases of MI or CHD death (83 in men and 45 in women). The incidence of a CHD event was determined in the presence versus absence of IR with the entire sample divided at the median of HDL-C and the median of triglyceride values into lower and higher range groups. In Table 2 the incidence rates of CHD are shown with and without IR, at higher and lower levels of HDL-C and triglycerides, for the entire study sample and also for men and women separately. HDL-C values tended to be lower and triglyceride values higher with IR than without IR. For the entire study group, the 10-year incidence of CHD appeared generally greater at both lower and higher ranges of either plasma HDL-C or triglycerides in the presence of IR as compared to the absence of IR.

Table 2.

Cumulative Incidence Rates of MI or CHD Death by HDL-C and Triglyceride Values and Presence or Absence of Insulin Resistance.

Group	Lipid Level	IR	N	HDL-C or Triglyceride mean ± SD	CHD events/person-years of follow-up	Age-adjusted 10-year cumulative incidence (95% CI)
HDL-C
All	≥median	no	1283	62.2±13.2	33/12307	0.01 (0.007, 0.021)
	<median	no	883	42.3±7.8	32/8347	0.02 (0.010, 0.031)
	≥median	yes	227	55.1±13.3	12/2057	0.04 (0.010, 0.060)
	<median	yes	513	38.4±7.8	51/4732	0.07 (0.043, 0.093)
Men	≥median	no	543	53.02±9.7	23/5164	0.02 (0.009, 0.038)
	<median	no	323	35.4±4.5	17/2970	0.03 (0.008, 0.048)
	≥median	yes	151	48.4±6.6	12/1347	0.06 (0.016, 0.095)
	<median	yes	270	33.8±5.07	31/2442	0.07 (0.038, 0.110)
Women	≥median	no	740	69.0±11.1	10/7142	0.01 (0.001, 0.013)
	<median	no	560	46.3±6.4	15/5376	0.02 (0.004, 0.029)
	≥median	yes	76	68.4±13.2	0/710	-
	<median	yes	234	43.6±7.1	20/2290	0.06 (0.026, 0.092)
Triglycerides
All	<median	no	1246	78.5±20.0	34/11939	0.02 (0.008, 0.023)
	≥median	no	920	173.7±75.4	31/8714	0.02 (0.008, 0.027)
	<median	yes	200	89.6±17.1	11/1849	0.04 (0.013, 0.076)
	≥median	yes	540	173.7±75.4	52/4940	0.06 (0.039, 0.086)
Men	<median	no	523	82.0±21.8	24/4924	0.02 (0.009, 0.038)
	≥median	no	344	198.7±92.8	16/3210	0.03 (0.008, 0.047)
	<median	yes	120	91.5±18.7	8/1090	0.05 (0.005, 0.088)
	≥median	yes	302	235.1±125.7	35/2699	0.08 (0.041, 0.110)
Women	<median	no	724	75.9±18.2	10/7014	0.01 (0.001, 0.019)
	≥median	no	577	158.8±58.0	15/5504	0.01 (0.002, 0.020)
	<median	yes	80	86.7±14.0	3/759	0.04 (0.000, 0.084)
	≥median	yes	240	197.1±85.9	17/2240	0.05 (0.017, 0.074)

Abbreviations: CHD, coronary heart disease; CI, confidence intervals; HDL-C, high density lipoprotein-cholesterol; MI, myocardial infarction; IR, insulin resistance. IR was defined by the upper quartile of HOMA-IR levels in subjects without diabetes. Mean values ± SD of HDL-C and triglycerides are shown as mg/dL

As shown by Kaplan-Meier plots (Fig. 2) the cumulative incidence of a CHD event was greatest for groups with IR and either lowest HDL-C or highest triglyceride values and was significantly different than the CHD incidence rates for the groups with comparable levels of these lipids but without IR. All log rank tests were statistically significant (P<0.001).

Fig 2.

Unadjusted Kaplan-Meier curves, showing the cumulative incidence of CHD events in the entire study sample (N=2910) with and without insulin resistance (IR) and with lower or higher values of fasting plasma HDL-C) (panel A) and lower or higher values of plasma triglycerides (TG) (panel B). Insulin resistance was defined in the entire study group without diabetes by the upper quartile of HOMA-IR values. Lower or higher values of plasma HDL-C and triglycerides were defined at the median plasma concentration of these values for the combined study group of men and women (i.e., at 49 mg/dL for HDL-C and at 112 mg/dL for triglycerides). Mean HDL-C and triglyceride values for each of these subgroups (with and without IR) are shown for each curve within parenthesis.

The relation of CHD incidence to the presence and absence of IR across a full range of HDL-C and triglyceride levels is shown by Cox proportional hazards regression analysis in Table 3. All Cox models were constructed with a referent subgroup without IR and either highest levels of HDL-C or lowest levels of triglycerides. Models were adjusted for age alone and additionally for multiple commonly-associated risk factors. Included in these adjustments with HDL-C as the variable of interest is non-HDL-C and for triglycerides as the variable of interest, non-triglyceride-related cholesterol.

Table 3.

Incidence of Coronary Heart Disease by Presence or Absence of Insulin Resistance at Lower or Higher Values of Plasma HDL-C or Triglycerides.

Groups	IR	HDL-C	HR (95% CI) Age-adjusted	P-value	HR (95% CI) Multi-adjusted	P-value
All Participants	yes	< median	3.60 (2.32, 5.59)	<0.001	2.83 (1.70, 4.70)	<0.001
	yes	≥ median	1.73 (0.89, 3.38)	0.11	1.46 (0.73, 2.91)	0.29
	no	< median	1.56 (0.96 2.53)	0.08	1.24 (0.74, 2.07)	0.42
Men	yes	< median	2.82 (1.64, 4.84)	<0.001	2.29 (1.24, 4.26)	0.009
	yes	≥ median	2.03 (1.01, 4.09)	0.048	1.73 (0.82, 3.65)	0.15
	no	< median	1.32 (0.70, 2.46)	0.39	1.09 (0.56, 2.10)	0.81
Women	yes	< median	5.88 (2.75,12.57)	<0.001	5.32 (2.08, 13.60)	<0.001
	yes	≥ median	0 (0, 0)	0.98	0 (0, 0)	0.99
	no	< median	2.15 (0.97, 4.79)	0.06	1.88 (0.78, 4.49)	0.16
Groups	IR	Triglycerides	HR (95% CI) Age-adjusted	P-value	HR (95% CI) Multi-adjusted	P-value
All Participants	yes	≥ median	3.01 (1.94, 4.66)	<0.001	2.50 (1.52, 4.11)	<0.001
	yes	< median	1.75 (0.89, 3.47)	0.11	1.60 (0.79, 3.25)	0.19
	no	≥ median	1.18 (0.72,1.92)	0.51	0.96 (0.58, 1.57)	0.86
Men	yes	≥ median	2.62 (1.55, 4.42)	<0.001	2.30 (1.27, 4.15)	0.006
	yes	< median	1.50 (0.67, 3.33)	0.32	1.35 (0.58, 3.12)	0.48
	no	≥ median	1.02 (0.54, 1.91)	0.96	0.84 (0.44, 1.60)	0.59
Women	yes	≥ median	3.91 (1.78, 8.58)	<0.001	3.08 (1.23, 7.75)	0.02
	yes	< median	2.37 (0.65, 8.62)	0.19	2.44 (0.64, 9.37)	0.19
	no	≥ median	1.39 (0.62, 3.10)	0.43	1.05 (0.45, 2.45)	0.91

Abbreviations: CI, confidence interval; HDL-C, high-density lipoprotein cholesterol; HR, hazard ratio; IR, insulin resistance

^*Median values of HDL-C are 49 mg/dL for all subjects; 42 mg/dL for men; and 56 mg/dL for women. Median values of triglycerides are 112 mg/dL for all subjects; 121 mg/dL for men; and 108 mg/dL for women.

^† Multivariable-adjusted models were adjusted for age, sex, systolic blood pressure, waist circumference, active smoking, alcohol consumption, treatment of hypertension, hormone replacement therapy, and for non-HDL-C with HDL-C as the variable of interest and for non-triglyceride-related cholesterol with triglycerides as the variable of interest.

As shown in Table 3, in fully-adjusted analysis in the entire study group the risk of a CHD event was significantly increased with IR at the low levels of HDL-C (i.e. below the median) compared to the referent group (no IR and HDL-C above the median). In contrast, in the absence of IR, the risk of a CHD event for low HDL-C was not significantly increased compared to high HDL-C. Results in men and women separately closely approximated the results of the entire group, showing an increase in CHD risk compared to the referent category in the presence of IR and a low HDL-C but not at higher levels of HDL-C or in the absence of IR.

Results for triglycerides (Table 3) show in fully-adjusted analysis that compared to a referent group without IR and a low level of triglycerides, there was a significant increase in the risk of CHD events with IR and high triglycerides. In contrast, in the absence of IR even with triglycerides in the same high range, the risk of CHD events was not increased. These results were similar in men and women with, again, a significant increase in the risk of CHD events associated with the presence of IR and high triglycerides as compared to those with no IR and triglycerides below the median.

Secondary analysis using continuous HDL-C, dichotomous HOMA (presence vs. absence of IR) and their interaction, have shown the interaction approaching significance (P= 0.064). Therefore, analysis assessing the relationship of HDL-C to CHD risk was stratified by the level of HOMA. For those with IR, increasing continuous HDL-C was significantly associated with decreasing risk of CHD (HR=0.94, P<0.0001). For those without IR, continuous HDL-C was also significantly associated with CHD risk (HR=0.98, P=0.049), although both the magnitude of the effect was lower. Similarly, the interaction between HOMA and log triglycerides on CHD risk also approached significance (P=0.052), therefore the analysis assessing the relationship of triglycerides to CHD risk was also stratified by the level of HOMA. In those with IR, increased log triglycerides was significantly associated with increased CHD risk (HR=1.95, P=0.018). In those with no IR, the association was not significant (HR=0.62, P=0.15).

DISCUSSION

In our large community-based sample we have found in analyses excluding diabetes and adjusted for conventional CHD risk factors as well as abdominal obesity that characterizes the metabolic syndrome, that in the presence of IR lowest values of HDL-C as well as highest values of triglycerides were associated with an increased incidence of MI and CHD death. Conversely, we have found that in the absence of IR comparably high levels of triglycerides or low levels of HDL-C that ordinarily might be expected to be associated with an increase in CHD, were not significant risk factors for CHD. In this study, an increase in CHD with IR occurred similarly in men and women with either lower HDL-C or higher triglyceride values whereas in the absence of IR neither lower HDL-C nor higher triglycerides were associated with an increase in CHD events.

Our findings generally support multiple studies by Reaven and associates^4,12,16,17 demonstrating in selected subgroups that IR and not obesity that is associated with lower HDL-C and higher triglyceride values best defines a high-CHD risk state. Furthermore, our present results appear to conform to more general studies showing that IR (or hyperinsulinemia as a surrogate marker for IR) will independently predict the development of cardiovascular (CV) disease in general populations^8–11 as well as in populations with diabetes¹⁸. Although there is evidence that high fasting insulin concentrations, independent of body weight, will predict the development of decreased HDL-C and increased triglyceride concentrations⁷ there appears to be just one population study other than this present one which has demonstrated differences in CV disease outcomes in relation to different concentrations of HDL-C and triglycerides in the presence and absence of IR. In that population, the placebo-treated group of men in the Veterans Affairs HDL-Intervention Trial¹⁹ with uniformly low values of plasma HDL-C (≤40 mg/dL), the 5-year incidence of a major CV event was significantly greater in the presence of IR than without IR at both higher and lower levels of either HDL-C or triglycerides.

There is substantial experimental evidence that both higher levels of plasma triglycerides and lower levels of HDL-C can be a consequence of IR. Studies in man have shown that with IR there is increased synthesis and secretion of triglyceride-rich, VLDL particles by the liver⁵. These studies have further shown that there is a strong linear relationship between hepatic triglyceride production and plasma triglyceride levels that is dependent on circulating insulin levels but is independent of obesity³. With increased plasma triglycerides HDL-C levels are also frequently low and two mechanisms, both involving the activity of triglyceride lipases, have been proposed to account for this reduction in HDL-C as a consequence of IR. By one process, IR may result in increased cholesteryl ester transport protein (CETP)-mediated triglyceride exchange between VLDL and HDL, enriching HDL with triglycerides and making these particles more susceptible to catabolism by hepatic lipase^6,20. By another process, IR may decrease plasma lipoprotein lipase²¹ that ordinarily catalyzes the catabolism of plasma VLDL and, as a by-product of VLDL catabolism, generates new material for HDL formation from the surface components of plasma VLDL^22,23.

As in other studies in predominantly white, general populations^24,25 our results in the FHS show that IR as defined by the fourth quartile of HOMA-IR was less frequent in women (19.7%) then in men (32.7%). This gender-related difference in the prevalence of IR is consistent with a generally lower prevalence of an intra-abdominal (or visceral) pattern of obesity in women then in men^26,27. Since, as we show, the frequency of both lower plasma values of HDL-C and higher values of triglycerides is increased with IR, it seems possible that women may less frequently have both lower levels of HDL-C and higher levels of triglycerides then men because of a less frequent presence of lR.

In our present analysis we did not attempt to assess the extent to which the MetS in comparison to IR might predict the development of new CHD. This comparison has previously been published in the FHS population but for a broader endpoint of CVD that included cerebrovascular and peripheral vascular events as well as coronary disease²⁸. In that study MetS predicted outcomes as well as or better than the HOMA-IR measure of IR. However, given the negative to highly variable relationship of low HDL-C to especially the incidence of ischemic stroke^29,30 the apparent superiority of MetS in predicting events in that analysis might be explained by the inclusion of non-CHD cases. Furthermore, while waist circumference (or BMI) is a key distinguishing feature of the MetS and is significantly correlated with insulin levels, the strength of this relationship is relatively weak. In our current analysis with a correlation of HOMA-IR to waist circumference and BMI of 0.49 and 0.50, respectively, only 23–25% of the variability in HOMA-IR levels was explained by waist or BMI measurements which closely conforms to comparisons in two large series of healthy individuals^17,31 where insulin-mediated, plasma glucose concentrations were related to body weight.

It has been suggested because of the strong association of a high triglyceride or low HDL-C to IR that an increased ratio of triglycerides to HDL-C might provide a clinically useful surrogate measure of IR^12,13. However, in a previous analysis from the FHS the ratio of triglycerides to HDL-C was shown to lack both substantial precision and sensitivity compared to a high HOMA-IR measurement in the prediction of new-onset CHD³². The combination of an increased waist circumference and high triglycerides³³ has also been suggested as a marker of increased visceral fat and a clinically useful index of increased CHD risk. While this index was proposed as a surrogate measure for a triad of laboratory measurements that included hyperinsulinemia³⁴, there appears to be no evidence in any general population that this more readily available index will perform as well in assessing CHD risk as directly measuring plasma insulin levels.

Implications, Strengths, and Limitations of Findings

Our results should not be interpreted to implicate IR as directly causal for a CHD event, with or without a dyslipidemia. It is clear from a number of studies that IR is associated with a heightened inflammatory and prothrombotic state associated with abnormalities in a number of laboratory markers of vascular injury, thrombosis, and fibrinolysis that may result in clinical disease³⁵. Although we have adjusted for essentially all of the elements of the MetS and found a fourth quartile value of HOMA-IR to independently predict the development of CHD in conjunction with lower values of HDL-C and higher values of triglycerides, our results should not also imply that the MetS, as a clinical construct, is not useful in identifying individuals who may be at higher risk for a CHD event.

Our findings were related to only the traditionally “hard” CV endpoints of confirmed MI or CHD death. Consequently, although our primary outcome (in sex-merged analysis) may appear to be based on relatively low numbers of events these events were selected to exclude both the possibly less definitive endpoint of angina and the endpoint of stroke, which has an uncertain relationship to dyslipidemia.

As a limitation of this study, we recognize that our analysis was constructed to obtain approximately equally-sized groups with lower and higher values of HDL-C and triglycerides similarly in both men and women. We therefore did not use values for these lipids that are more commonly used in clinical practice to characterize low and high-risk CHD states. We further recognize that although our results were obtained in a large community-based population, not selected for particular traits, the FHS population is predominantly a white population and the results obtained in this analysis may differ in other ethnic groups.

Conclusions

Our present study provides clear evidence that for the endpoints of MI and CHD death the risk associated with lower and higher levels of HDL-C and triglycerides can be more precisely defined in conjunction with a measure of IR. Although fasting plasma insulin measurements have yet to be standardized, as we show in this analysis even a locally-defined measure of IR that is used in conjunction with lower levels of HDL-C and higher levels of triglycerides is apt to more accurately identify an individual at-risk for CHD than the use of just these lipid measures alone.