Fasting Triglycerides, High-Density Lipoprotein, and Risk of Myocardial Infarction

Over the past four decades since serum cholesterol levels were first linked to atherosclerotic disease, a number of additional markers have been identified in an attempt to better characterize the atherogenic potential of the lipid profile. Relationships of cholesterol ester–rich lipoproteins (LDL and HDL) with atherosclerosis have been clearly established. Recent data suggest that TG-rich lipoproteins (chylomicrons, chylomicron remnants, and VLDL) may also play a role in atherogenesis.¹²³⁴⁵

Despite extensive research over the past few decades, it remains unclear whether plasma TG, as a marker for the TG-rich lipoproteins, have independent value in predicting risk of cardiovascular disease. A recent National Institutes of Health consensus conference concluded that the data to support a judgment of a causal relationship of elevated TG with cardiovascular disease are “mixed.”⁶ Most case-control and prospective cohort studies that have examined the relationship of fasting TG on risk of cardiovascular disease have reported strong crude associations.⁷⁸ There appear to be complex metabolic relationships between cholesterol ester–and triglyceride-rich lipoproteins¹⁹ and control for other lipid parameters can substantially attenuate the TG association.⁷⁸⁹ In particular, control for HDL cholesterol, which is inversely correlated with TG, tends to substantially attenuate the association of TG with CHD.¹⁰

As previously outlined by Austin et al⁸ and Crique et al,¹¹ the assessment of any relationship of TG with risks of cardiovascular disease is complicated by several methodological issues. First, there is considerable within-individual variability in measured TG levels.¹² Second, the distribution of TG levels in the population is not normal.¹³ Third, TG are strongly correlated with other lipid parameters.¹⁴¹⁵ Fourth, there are complex metabolic relationships between the TG- and cholesterol ester–rich lipoproteins that may interact to increase risks of cardiovascular disease.¹⁹¹⁶ In an attempt to better understand the complex interactions of TG- and cholesterol ester–rich lipoproteins, we examined the interrelationships of the fasting TG level, other lipid parameters, and nonlipid risk factors with risk of MI in a study of 340 cases and an equal number of control subjects matched on age, sex, and neighborhood of residence.

Methods

The methods of the Boston Area Health Study have been previously described.¹⁷¹⁸ Briefly, we reviewed all admissions between January 1, 1982, and December 31, 1983, to the coronary care and other intensive care units of six suburban Boston hospitals (Emerson, Framingham Union, Leonard Morse, Mount Auburn, Newton-Wellesley, and Waltham-Weston) to identify eligible cases. Those eligible for inclusion were white men or women <76 years of age who lived in the Boston area and had no history of previous MI or angina pectoris, in whom symptoms of MI had begun within 24 hours of admission. Patients with the diagnosis of confirmed MI, based on clinical history, who had an increase in creatine kinase and were discharged alive were enrolled in the study if they were willing and able to participate and if informed consent could be obtained from the patient and the admitting physician. The research protocol was approved by institutional human subjects committees of all participating hospitals.

For each case, a control subject was chosen at random from the list of residents of the town in which the patient lived. Specifically, the name of the patient was located in the appropriate residence list of the town in which the case patient lived, and the next resident listed of the same sex and age (within 5 years) with a listed telephone number was selected as a control. Potentially eligible subjects were sent letters of invitation and then contacted by telephone. Of the eligible subjects contacted, 84% of cases and 60% of control subjects were enrolled, yielding a total of 340 case-control pairs.

All cases and control subjects were interviewed in their homes. Case patients were interviewed ≈8 weeks after their MI. Information was collected on a wide variety of potential coronary risk factors related to the time period before the MI for the case patients and before the interview for the control subjects. This information included age, sex, hypertension (defined as reported treatment for hypertension), diabetes mellitus, cigarette smoking, body mass index, family history of premature (<60 years of age) MI, dietary intake, psychological indicators, socioeconomic status, level of physical activity, and alcohol consumption. Information on diet and alcohol consumption was gathered using a semiquantitative food-frequency questionnaire.¹⁹²⁰ Psychological indicators were measured using 18 questions from the Framingham Heart Study (10 assessing type A behavior, 7 assessing anger, and 1 regarding the number of promotions gained over the past 10 years).²¹ Information on socioeconomic status included usual occupation and highest educational level attained.

Fasting venous blood samples were obtained and analyzed for lipids. Venous blood was drawn into 0.1% EDTA, and plasma was obtained by centrifugation at 3000 rpm for 30 minutes at 4°C. Fresh plasma was used to determine TG level, total cholesterol, LDL cholesterol, VLDL cholesterol, and HDL cholesterol using Lipid Research Clinics methods.²²²³ Cholesterol determinations were standardized by the Lipid Standardization Program of the Centers for Disease Control and Prevention, Atlanta, Ga. Lipid determinations were made on a total of 605 subjects (306 cases and 299 control subjects) who provided an adequate venous blood sample. HDL subfractions were determined on fresh unfrozen plasma by the dextran sulfate method of Gidez on a subsample of 558 subjects (283 cases and 275 control subjects).²⁴

Matched pair and crude unmatched relative risks were calculated.²⁵ Because these were virtually identical, we judged that the matching could safely be disregarded and thereafter performed unmatched analyses. Multiple logistic regression analyses were used to estimate relative risks while simultaneously controlling for a number of coronary risk factors.²⁶ Tests for trend were conducted using logistic regression. Logistic regression models were compared using the likelihood ratio test. Because of the skewed nature of the raw TG distribution, we normalized the variable by taking the natural log of TG (logTG). Parallel models were run using both raw TG and logTG when TG were added to multivariate models as a continuous variable. To examine the interrelationships between fasting TG and other lipoprotein levels, total cholesterol, LDL, VLDL, HDL, HDL₂, and HDL₃ cholesterol were added to the risk factor model with TG one at a time. In addition, we used stepwise logistic regression to determine which lipid parameters, including ratios of total cholesterol, LDL, and TG to HDL, had the greatest predictive value. Relative risks were calculated for those in the second, third, and fourth quartiles of TG level and of the ratio of triglyceride to HDL compared with those in the first. Because of prior reports of a stronger association of TG with CHD²⁷²⁸ among women compared with men, separate analyses stratified by sex were conducted.

Results

Baseline characteristics of cases and control subjects are presented in Table 1. As expected, major coronary risk factors were more prevalent among cases than control subjects. Coronary risk factors by each quartile of TG level among control subjects are presented in Table 2. Those in the higher TG categories were more likely to be male, have hypertension, have diabetes mellitus, and have higher body mass index. In addition, they tended to be more active and consume more alcohol.

Age- and sex-adjusted levels of other plasma lipid parameters by quartile of TG level among control subjects are presented in Table 3. There were strong positive associations of TG with total cholesterol and VLDL cholesterol level as well as strong inverse associations of TG with HDL level and both of its subfractions, although the association appeared to be stronger for HDL₂ compared with that for HDL₃. There also was a weak association with LDL cholesterol level. Mean LDL particle diameter was inversely related to TG levels.

The relative risk of MI among those in the higher compared to those in the lowest quartile of TG are presented in Table 4. Age- and sex-adjusted relative risks in the second, third, and fourth quartiles were 3.2, 4.5, and 6.8, respectively (P for trend <.001). Adjustment for available coronary risk factors did not materially alter the results. As expected, adjustment for HDL attenuated the relative risks, although they remained significantly elevated in each of the higher categories (P for trend=.016). The attenuation was more pronounced in the highest quartile. Further adjustment for LDL did not materially alter the results.

Based on prior reports that the interrelationship of triglyceride-rich apoprotein C-III–containing lipoprotein particles (largely VLDL) and cholesterol ester–rich apoprotein C-III–containing particles (largely HDL) predicts angiographic progression of atherosclerosis,²²⁹ we defined a ratio that likely captures similar information: TG level (which roughly approximates VLDL) divided by HDL (TG/HDL). TG levels were log transformed to normalize distribution. Data on the risk of MI by quartile of TG/HDL are presented in Table 5. Compared with those in the lowest, those in the highest quartile had a 16.0-fold increased risk of MI (95% CI=7.7 to 33.1; P for trend across quartiles <.001) after multivariate adjustment. Stepwise logistic regression was also used to assess the predictive value of the TG/HDL ratio compared with LDL/HDL (previously the strongest lipid predictor of risk of MI in this data set) and total cholesterol/HDL. Although all three remained highly significant independent predictors, triglyceride/HDL entered first (data not shown).

LDL subclass was previously determined in a subset of 197 of the 680 cases and control subjects (101 cases and 96 control subjects) in the current study.³⁰ Compared with those with pattern A subjects (predominance of large buoyant LDL), pattern B subjects (predominance of small dense LDL) had significantly higher TG (232.0 versus 124.7 mg/dL; P<.0001) and lower HDL (32.3 versus 43.4 mg/dL; P<.001) levels. There was a 3.3-fold (95% CI=1.6 to 6.8) increased risk of MI among those with pattern B; however, this relationship was substantially attenuated after control for TG (RR=1.9; 95% CI=0.8 to 4.5) and HDL (RR=2.1; 95% CI=1.0 to 4.8).

There were apparent quantitative although not qualitative differences in the relationship of TG and HDL by sex. The association of TG with risk of MI appeared to be stronger for women than men, although the association of HDL appeared to be stronger for men than women. For men, there was a 2.6-fold (95% CI=1.7 to 3.9)–increased risk of MI for each unit change in logTG after multivariate adjustment compared with 4.5-fold (95% CI=1.7 to 12.1) for women. However, there were no apparent sex differences in the association of TG/HDL with risk of MI (data not shown). There were no qualitative differences in the predictive value of TG or TG/HDL among those with high compared with low LDL cholesterol.

Discussion

Our data are consistent with earlier reports that elevated fasting TG levels are strongly associated with risk of MI.¹⁰ Although elevated TG were associated with hypertension, diabetes mellitus, physical activity, increased alcohol intake and male sex, adjustment for these and other nonlipid risk factors did not materially alter the crude estimates.

Although fasting TG generally represent a strong predictor of CHD after control for nonlipid factors, adjustment for other lipid parameters substantially attenuates the association. Control for total cholesterol attenuates the relationship to nonsignificance in some but not all prior studies; however, adjustment for total cholesterol may not be appropriate.⁹¹¹ A proportion of the total cholesterol will be carried by TG-rich VLDL particles. Therefore, adjustment for total cholesterol may represent overadjustment. Similarly, adjustment for the TG-rich VLDL levels would not be appropriate. Adjustment for HDL substantially attenuates the association of TG with risk of CHD,^7891011 consistent with our findings. Although there is significant attenuation after control for HDL cholesterol, our data suggest that fasting TG remain an independent predictor of MI even after control for HDL.

The attenuation of the TG effect after control for HDL cholesterol may be due to true confounding or more likely is the result of metabolic interactions. Recent data suggest that there are complex metabolic interrelationships between the TG- and cholesterol ester–rich lipoproteins.¹²³¹ The relationship of TG levels with other lipid parameters is likely to be metabolic in nature. TG levels are elevated in the setting of decreased lipoprotein lipase activity. This leads to higher chylomicron remnant and VLDL levels (both of which may be atherogenic) and lower HDL levels (which clearly promote atherogenesis). Thus, the ratio of TG/HDL may be a valuable marker for abnormal TG metabolism. In addition, lower lipoprotein lipase activity could prolong circulation time of VLDL and may result in increased density of VLDL particles. The subclass patterns of LDL may be dependent in part on VLDL density. A predominance of small dense LDL particles (LDL subclass pattern B), which appear to be more atherogenic, is strongly associated with elevated TG levels and lower HDL levels. Smaller LDL diameter may be the result of smaller more dense VLDL precursors resulting from abnormal TG metabolism. This possible mechanistic explanation of the current findings remains somewhat speculative and warrants further basic and epidemiological investigation.

Ratios of cholesterol ester–rich lipoprotein levels (total cholesterol/HDL and LDL/HDL) are well-established predictors of CHD,⁶³² and a high ratio may be a good indicator of abnormal cholesterol metabolism. In the Physicians’ Health Study, a 1-unit increase in the LDL/HDL ratio was associated with a 53% increase in risk of MI.³² Similarly, in our study both ratios were strong independent predictors with a 1-unit increase in the TC/HDL and LDL/HDL associated with 49% and 75% increases in risk of MI, respectively. Our data suggest that TG/HDL ratio may be an important, albeit crude, marker of abnormal TG metabolism, which may provide valuable additional information about the atherogenic potential of a lipid profile. Although there may be some quantitative sex differences in the relationship of TG and HDL with risk of MI, there are no apparent sex differences for the relationship the TG/HDL with risk of MI.

Several limitations should be considered in the interpretation of these results. First, results are based on a single measurement of fasting lipids. The considerable within-individual variability in TG measurements would result in substantial regression dilution bias, which would tend to underestimate any true association between elevated TG and MI. Second, only survivors of MI were included in this study because of the need to assess adequately risk factor information on a large number of lifestyle variables as well as to allow transient lipoprotein alterations to return to normal before plasma samples are obtained. The result of this process is likely to underestimate the impact of many coronary risk factors due to the selection of the healthiest patients of MI. Third, the timing of the blood drawing could affect lipid levels because MI is known to acutely affect lipid metabolism. For this reason, we used blood specimens collected ≈10 weeks after hospital discharge, rather than those drawn in-hospital. Several interventions after MI may have altered lipid levels, including dietary modifications, exercise, and treatment for elevated cholesterol. Thus confounding by intervention after MI cannot be ruled out. The overall impact of these interventions would likely have resulted in only modest changes in TG and HDL levels among cases. Risk factor modifications, in general, could lower or raise TG and HDL levels. We could not control for drug treatment after MI because only limited information on medication use after the MI was available. However, adjustment for prior treatment for elevated cholesterol, as well as β-blocker and thiazide use, did not materially alter the results. Furthermore, our findings are consistent with those reported in previous prospective studies. Finally, this study was underpowered to detect modest differences in effect between men and women given the fact that only 32% of the study population was female. Furthermore, the sample was composed primarily of white participants; therefore, generalizability to minorities may be limited.

In conclusion, although imprecision in TG measurements, within-individual variability, and complex interactions between TG and lipoprotein levels may obscure the impact of TG in the development of CHD, our data indicate that elevated fasting TG represent a useful marker for risk of CHD, particularly when HDL levels are considered. The strong association of the ratio of TG/HDL with risk of CHD suggests a metabolic interaction between the TG- and cholesterol ester–rich lipoproteins in increasing risk of MI, although further basic and epidemiological studies are warranted to explore mechanistic explanations for these findings. Although other laboratory measures, such as chylomicron remnants and LDL subclass, may provide additional information about the atherogenicity of the lipoprotein profile beyond the standard lipoprotein profile, a simple TG measure may provide a great deal of information in this regard. If these findings are confirmed in other studies, trials testing interventional strategies specifically designed to correct abnormalities in TG metabolism may be warranted. Ultimately, screening and treatment guidelines may require modification to allow greater attention to be paid to fasting TG levels.

Selected Abbreviations and Acronyms

Reprint requests to Dr J. Michael Gaziano, Brigham and Women’s Hospital, 900 Commonwealth Ave E, Boston MA 02215-1204.