If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Averaging triglycerides (TGs) improves risk prediction over single measurements.
•
TGs are associated with increased cardiovascular risk, even in the “normal” range.
•
The association between TGs and cardiovascular events differs by sex.
Background
Elevated triglycerides (TGs) are associated with increased risk of cardiovascular disease (CVD), but the best way to both measure TGs and assess TG-related risk remains unknown.
Objective
The objective of the study was to evaluate the association between TGs and CVD and determine whether the average of a series of TG measurements is more predictive of CVD risk than a single TG measurement.
Methods
We examined 15,792 study participants, aged 40–65 years, free of CVD from the Atherosclerosis Risk in Communities and Framingham Offspring studies, using fasting TG measurements across multiple examinations over time. With up to 10 years of follow-up, we assessed time-to-first CVD event, as well as a composite of myocardial infarction, stroke, or cardiovascular death.
Results
Compared with a single TG measurement, average TGs over time had greater discrimination for CVD risk (C-statistic, 0.60 vs 0.57). Risk for CVD increased as average TGs rose until an inflection point of ~100 mg/dL in men and ~200 mg/dL in women, above which this risk association plateaued. The relationship between average TGs and CVD remained statistically significant in multivariable modeling adjusting for low-density lipoprotein cholesterol, and interactions were found by sex and high-density lipoprotein cholesterol level.
Conclusions
The average of several TG readings provides incremental improvements for the prediction of CVD relative to a single TG measurement. Regardless of the method of measurement, higher TGs were associated with increased CVD risk, even at levels previously considered “optimal” (<150 mg/dL).
Despite longstanding epidemiologic data demonstrating the association between elevations in serum triglycerides (TGs) and cardiovascular disease (CVD),
many details about the relationship between TGs and CVD risk remain unanswered. First, the exact level at which risk begins to increase is unclear. When lipid guidelines for CVD prevention first emerged, elevated TGs were defined as >250 mg/dL, based on observational data showing the association between elevated TGs and CVD.
The Prospective Cardiovascular Munster (PROCAM) study: prevalence of hyperlipidemia in persons with hypertension and/or diabetes mellitus and the relationship to coronary heart disease.
Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III).
Since then, a multitude of observational studies have assessed the CVD risk associated with TGs using various measures, including quartiles of TGs and log-transformed TG level.
Plasma triglyceride level is a risk factor for cardiovascular disease independent of high-density lipoprotein cholesterol level: a meta-analysis of population-based prospective studies.
Coronary heart disease prediction from lipoprotein cholesterol levels, triglycerides, lipoprotein(a), apolipoproteins A-I and B, and HDL density subfractions: The Atherosclerosis Risk in Communities (ARIC) Study.
The exact threshold above which CVD risk increases is less clear; other work has shown increasing risk with elevated TGs above 88 mg/dL and above 133 mg/dL.
Real-world risk of cardiovascular outcomes associated with hypertriglyceridaemia among individuals with atherosclerotic cardiovascular disease and potential eligibility for emerging therapies.
Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study.
; however, specific inflection points or cutoffs, around which the rise in risk associated with TGs changes, have not been systematically evaluated with regard to incident CVD risk.
A second unclear detail about the relationship between TGs and CVD risk is that TG measurements can be highly variable, depending on when and what a patient's most recent meal was, as well as other factors. Variation among serial samples for TGs is much higher than what is seen with cholesterol (25% vs 8%).
Consequently, it is expected that serial measures may perform better than single assessments. Finally, whether the association of TGs and CVD risk is relatively stable in all patient types is unclear. Specifically, whether the association of TGs and outcomes is constant by age, sex, and factors such body mass index (BMI) or other lipid parameters is not well characterized.
Using data from 2 well-characterized national cohort studies, we performed a detailed examination into the association of TGs and CVD risk by (1) assessing whether an average of several TG measures was more associated with CVD risk than a single assessment either at baseline or using a participant's highest prior TG level; (2) evaluating the continuous association between TG levels and cardiovascular events across time to determine key inflection points; and (3) assessing whether this association was similar by age, sex, BMI, and low-density lipoprotein cholesterol (LDL-C).
Methods
Study design and sample
We selected patients from 2 large observational cohort studies: the Atherosclerosis Risk in Communities (ARIC) study and the Framingham Offspring Study. The first visit in the ARIC study was conducted between 1987 and 1989, with follow-up visits every 2–3 years until 1998, and a final visit conducted between 2011 and 2013. In the Framingham Offspring study, participant examination 1 visits were conducted between 1971 and 1975, with follow-up visits occurring approximately every 3–4 years, and a final visit conducted between 2005 and 2008.
To evaluate the association between multiple TG measurements over time and future CVD, we selected a baseline visit from each study that would allow us to capture multiple prior TG measurements, as well as sufficient follow-up to assess events. The baseline visit for ARIC was examination 4 (1996–1998, n = 10,912), and the baseline visit for Framingham Offspring was examination 6 (1983–1987, n = 5013). These visits were chosen as baseline to allow for sufficient time with exposure (to TGs) as well as allow for sufficient follow-up time to observe cardiovascular events. From this cohort, we excluded patients younger than 40 years and older than 65 years (n = 3818 in the ARIC study, n = 2570 in the Framingham Offspring Study), patients with prevalent CVD (defined as prior myocardial infarction, angina, coronary artery disease, transient ischemic attack, stroke, percutaneous coronary intervention, peripheral artery disease, and heart failure [n = 894 in the ARIC study, n = 194 in the Framingham Offspring Study]), and patients with fewer than 2 available fasting TG measurements at baseline (n = 88 in ARIC, n = 193 in Framingham Offspring). After exclusions, the final cohort was comprised of 8068 patients (n = 6012 from ARIC, n = 2056 from Framingham Offspring).
Outcomes and exposures
Our primary end point of interest was incident hard cardiovascular events, defined as a composite of myocardial infarction, stroke, and cardiovascular death, with follow-up of up to 10 years. Baseline TGs were defined as TG level at examination 4 for the ARIC study and examination 6 for the Framingham Offspring Study. Average TGs were calculated using the average of all prior TG measurements. Max TGs were defined as the highest prior value, regardless of timing.
When TG measurements were missing, linear approximations were used to impute the values. If a subject attended the examination where TGs were missing, then the value was approximated using the slope between nonmissing visits and the date of the examination. If the subject missed the examination entirely, then the TG value was approximated as the average of the values at the examinations before and after the missing examination.
Statistical analysis
Descriptive statistics were run on both ARIC and Framingham Offspring cohorts. Continuous variables were reported as median (25th, 75th percentiles). Categorical variables were presented as the number (percentage of the total cohort) falling within the specified subcategory. The distribution of TG levels was assessed in both cohorts.
To compare single vs average TGs, univariable Cox proportional hazards models were used to assess the relationship between each measure of TG level and CVD events. Each measure was tested for linearity, and any nonlinear relationships were modeled using restricted cubic splines. The relative strength of the measures of TGs was then determined by looking at the c-index, Akaike information criterion, hazard ratio (HR) for a 1 standard deviation change, and P-value from each univariable model. The measure most closely related to CVD risk (average TGs) was then selected for further investigation.
To describe the rate of incident atherosclerotic cardiovascular disease by TG level, Kaplan-Meier event curves were generated by quartile of average TGs. To account for the nonlinear relationship between average TG and CVD risk, average TGs were modeled using 2 ways: first, using restricted cubic splines, and second, using the log2 transformation. The restricted cubic spline analysis was carried out to assess for potential inflection points in the shape of the association between TGs and CVD risk. The log2 transformation was used for multivariable modeling. Multivariable Cox proportional hazards modeling was performed to evaluate the association between the log2-transformed average TGs and CVD risk, adjusting for the following variables from the baseline examination: age, sex, BMI, history of diabetes, high-density lipoprotein cholesterol (HDL-C), LDL-C, statin use, blood pressure medication use, ounces of ethanol per week, cholesterol medication (3 levels: on statin, not on statin but on nonstatin lipid-lowering agent, no lipid-lowering agent), and cohort. The following interactions were considered in the unadjusted Cox proportional hazards model between TGs and CVD risk: age, sex, BMI, diabetes, HDL-C, LDL-C, and statin use. Where interactions were detected, unadjusted spline plots were created by subgroup. HRs with 95% confidence intervals (CIs) were presented per 1-unit change in log2 (average TGs; ie, per doubling of TGs). For interactions with continuous covariates, HRs per doubling of TGs were presented at specific values of the covariate. In the sensitivity analysis, we substituted non–HDL-C for LDL-C in multivariable modeling.
To further understand the interaction between TGs and HDL-C, we evaluated the association between the TG-to-HDL-C ratio and CVD events in univariable and multivariable modeling, adjusting for age, sex, BMI, diabetes, non–HDL-C, statin use, and cohort. Unadjusted spline plots were created to visually assess the relationship between TG-to-HDL-C ratio and CVD events. Kaplan-Meier event rates by subgroup were also calculated stratifying by tertiles of TGs and HDL-C.
Institutional review board approval was obtained by all participating centers for the ARIC and Framingham Offspring studies. All participants provided consent before enrollment. Data were obtained from the National Institutes of Health Biologic Specimen and Data Repository Information Coordinating Center repository. All analyses were conducted by the Duke Clinical Research Institute (Duke University, Durham, NC) using SAS, version 9.4 (SAS Institute, Inc., Cary, NC).
Results
Characteristics of the study population
Our study population consisted of 8068 participants, including 6012 from the ARIC study and 2056 from the Framingham Offspring Study, with a median follow-up of 10 years (Table 1). Very few participants (7.8% overall) were on statin therapy at baseline. Compared with the ARIC population, those in the Framingham study had less diabetes, less hypertension, and more alcohol intake and included more whites. In both cohorts, use of lipid-lowering therapy was low (7.3% in the Framingham Offspring Study and 10.7% in the ARIC study). The average number of TG measurements available was 4.0 in the ARIC study and 5.8 in the Framingham Offspring Study. The median age was 58 years; 56.5% were female and 84.2% were white. The distribution of TG values was skewed, with a median TG level of 116 (25th–75th percentile 86–160) in the ARIC study and 94 (69–132) in the Framingham Offspring Study (Fig. 1). The 10th, 25th, 50th, 75th, and 90th percentiles of TGs in males were 65, 86, 118, 167, and 228 mg/dL and in females were 62, 79, 105, 144, and 196, respectively (Supplement Fig. 1).
Table 1Characteristics of the analytic cohort
Characteristic
Total (n = 8068)
Framingham Offspring (n = 2056)
ARIC (n = 6012)
P-value
Age (median 25th 75th, y)
58 (55–62)
54 (50–60)
59 (56–62)
<.001
Female (n, %)
4556 (56.5%)
1111 (54.0%)
3445 (57.3%)
0.010
White (n, %)
6790 (84.2%)
2056 (100.0%)
4734 (78.7%)
<.001
Smoking (n, %)
1308 (16.3%)
336 (16.3%)
972 (16.2%)
0.912
Diabetes (n, %)
970 (12.1%)
123 (6.0%)
847 (14.1%)
<.001
Systolic BP (median 25th and 75th, mmHg)
122, 112–135
123, 112–135
122, 112–135
0.486
Diastolic BP (median 25th and 75th, mmHg)
73, 66–79
76, 70–82
72, 65–78
<.001
BMI (median 25th and 75th)
28, 25–31
28, 25–31
28, 25–32
<.001
Ounces of ethanol per week
0, 0–2
1, 0–3
0, 0–1
<.001
Hypertension treatment (n, %)
2170 (27.0%)
393 (19.1%)
1777 (29.7%)
<.001
Lipid-lowering medication (n, %)
On statin
629 (7.8%)
120 (5.8%)
509 (8.5%)
<.001
Not on statin, on other lipid-lowering medication
161 (2.0%)
31 (1.5%)
130 (2.2%)
No lipid-lowering medication
7623 (90.2%)
1904 (92.7%)
5359 (89.3%)
Fasting glucose
99, 92–108
96, 90–104
99, 93–109
<.001
Total cholesterol (median 25th and 75th, mg/dL)
200, 178–225
204, 181–230
199, 177–224
<.001
HDL-C (median 25th and 75th, mg/dL)
48, 39–61
50, 40–61
47, 39–60
.003
Non–HDL-C (median 25th and 75th, mg/dL)
149, 126–174
152, 127–178
148, 125–173
<.001
Triglyceride-to-HDL-C ratio (median 25th and 75th)
2.29, 1.43–3.70
1.89, 1.18–3.08
2.44, 1.54–3.89
<.001
LDL-C (median 25th and 75th, mg/dL)
122, 101–145
125, 104–148
121, 100–143
<.001
Average TG (median 25th and 75th, mg/dL)
110, 82–154
94, 69–132
116, 86–160
<.001
Number of TG measurements before imputation
Mean: 4.4
Mean: 5.8
Mean: 4.0
Median: 4
Median: 6
Median: 4
Follow-up time
10 y
10 y
10 y
ARIC, Atherosclerosis Risk in Communities; BMI, body mass index; BP, blood pressure; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglyceride.
Figure 1Distribution of baseline TGs in Framingham Offspring and ARIC participants free of CVD. Histogram of distribution of prior average TGs at baseline in Framingham Offspring and ARIC. ARIC, Atherosclerosis Risk in Communities; TG, triglyceride; CVD, cardiovascular disease.
Table 2 shows results from univariable analyses of different TG measures and cardiovascular events. Compared with baseline TGs, the average serum TG level was more highly correlated with CVD events (c-statistic, 0.60 vs 0.57) and had a higher c-statistic than TGs at baseline (c-statistic, 0.57) or maximum prior TGs (c-statistic, 0.58). Therefore, the average serum TG level was chosen as the primary exposure variable of interest for subsequent analyses.
Relationship nonlinear; modeled using restricted cubic splines using 4 knots.
(mg/dL)
0.584
10,234.1
<.0001
1.01 (1.00–1.01)
136.2 mg/dL
AUC, area under the curve, defined as triglycerides × years of exposure using trapezoid rule; COV, coefficient of variation defined as standard deviation/mean; CVD, cardiovascular disease; HR, hazard ratio; SD, standard deviation; TG, triglyceride.
∗ Relationship nonlinear; modeled using restricted cubic splines using 4 knots.
Characteristics of patients by quartile of average TGs are presented in Supplement Table 1. Those in the highest quartile of TGs included more males, more patients with diabetes, higher rates of use of cholesterol medication, higher fasting glucose, higher non–HDL-C, and higher LDL-C than those in the lowest quartile.
Figure 2 shows Kaplan-Meier results for cardiovascular risk by quartile of average prior TGs (<82 mg/dL, 82–110 mg/dL, 110–153 mg/dL, and ≥153 mg/dL), with increasing event rates by increasing average TG level (P < .0001). Figure 3 shows the shape of the relationship between TGs and CVD risk in univariable analysis using restricted cubic splines. Overall, the risk of CVD increased in proportion with the elevation of the average TG level until an inflection point around 150 mg/dL, at which point, the slope of the relationship was attenuated. In univariable analyses, each doubling of serum TGs increased CVD risk by 65% (HR: 1.65, 95% CI: 1.47–1.85). In multivariable modeling including LDL-C, the association remained statistically significant with a 24% increase in CVD risk per doubling of TGs (HR: 1.24, 95% CI: 1.06–1.45). In a sensitivity analysis, we substituted non-HDL as an alternative to LDL-C. After adjusting for non-HDL, the association between TGs and CVD events was no longer statistically significant (HR: 1.14 per doubling of TGs, 95% CI: 0.97–1.34, P = .11).
Figure 2Kaplan-Meier analysis of time-to-CVD event by quartile of average TG. Kaplan-Meier curves modeling probability of event over the study period by quartile of prior average TG. CVD, cardiovascular disease; TG, triglyceride.
Figure 3Cubic splines model of association between average TG and CVD risk. Univariable association between average TG and CVD risk. ∗Dotted lines represent 95% CI. CI, confidence interval; CVD, cardiovascular disease; TG, triglyceride.
Table 3 shows results of interaction analyses for the association between TGs and CVD risk. Statistically significant interactions were found between the average TG level and sex (P = .01), as well as between the average TG level and HDL-C (P = .01). The relationship between the average TG level and CVD risk was stronger for women (HR: 1.79, 95% CI: 1.50–2.14 vs HR: 1.34, 95% CI: 1.15–1.55) and with successively higher levels of HDL-C (HR: 1.68, 95% CI: 1.37–2.06; HR: 1.55, 95% CI: 1.29–1.86; and HR: 1.32, 95% CI: 1.13–1.53 at HDL-C levels of 60 mg/dL, 50 mg/dL, and 40 mg/dL, respectively). Figure 4A displays the association between TGs and CVD risk in men and women. In men, the risk appeared to increase with increasing TGs to around 100 mg/dL, whereas in women, the risk increased to around 200 mg/dL. Figure 4B shows the association between TGs and CVD risk by the HDL-C group. In those with low HDL-C, the association between the TG level and risk was steepest until around 100 mg/dL, whereas in those with higher HDL-C, the CVD risk increased with increasing TG levels. Interactions between TGs and age, BMI, diabetes, LDL-C, and statin use were not statistically significant.
Table 3Interaction analysis of association between TGs and CVD risk
Figure 4Association between the average TG level and CVD risk based on (A) sex and (B) HDL-C level. CVD, cardiovascular disease; HDL-C, high-density lipoprotein cholesterol; TG, triglyceride.
Supplement Figure 3 shows Kaplan-Meier event rates stratified by tertiles of average TGs and HDL-C. The risk was lowest in those in the lowest tertile of TGs and highest tertile of HDL-C. In univariable analysis, TG-to-HDL-C ratio was associated with CVD risk, with a steep increase in risk up to a ratio around 4 (HR: 1.44, 95% CI: 1.33–1.56) and no additional increase in risk beyond that point (Supplement Fig. 2). In unadjusted analysis, the association between the TG-to-HDL-C ratio and 10-year CVD risk was significant (1.48, 95% CI: 1.37–1.60), and in multivariable analysis, this relationship was attenuated but remained significant after adjustment for age, sex, BMI, antihypertensives, cholesterol medications, ethanol use, diabetes, hyperlipidemia, and non-HDL (HR: 1.16, 95% CI: 1.05–1.28).
Discussion
Although elevated TGs have long been known to be associated with increased risk of CVD, how to best quantify TG-associated CVD risk remains unclear. In this analysis of a large, prospectively followed up sample of patients from 2 observational cohort studies, we found that using the average of multiple prior TG measurements improved risk prediction more so than a single measurement. In addition, the association of TGs and CVD risk continued well below TG levels of 150 mg/dL, consistent with what has been seen previously.
Real-world risk of cardiovascular outcomes associated with hypertriglyceridaemia among individuals with atherosclerotic cardiovascular disease and potential eligibility for emerging therapies.
Our study extends these findings by modeling risk by average TGs to show that risk behaves linearly before an inflection point of about 150 mg/dL in a cohort of participants free of pre-existing CVD.
Measuring and interpreting TGs
Our finding that average TG levels over time proved to be a better approximation of CVD risk than a one-time measurement, a maximum value, or variability across values is not surprising, in light of the inherent variability of TGs and given how the biology of TG impacts its own measurement. Because fasting status impacts TG levels, TGs may vary significantly between measurements. Averaging TGs minimizes the noise in these measurements and decreases the impact of outliers. With increasing capture of all laboratory values in an electronic health record, the use of average prior TGs may become more clinically viable. Future epidemiologic studies and clinical trials using TGs as an inclusion criterion may consider using average TGs rather than single point-in-time measurements.
When patients receive reports of their lipid levels, these reports are often accompanied by descriptors of how their levels compare with “normal” ranges. Typically, TG levels <150 mg/dL are described as “normal” or “optimal” in these laboratory reports; however, we found that the association between TG levels and CVD risk continued well below TG levels of 150 mg/dL. Thus, even at what is considered within a “normal” range, risk is increased with increasing TG levels, suggesting that widely accepted “normal” TG ranges may not be biologically optimal. Our data are consistent with those of other studies that have found increasing CVD with increasing TG levels starting below 150 mg/dL.
Real-world risk of cardiovascular outcomes associated with hypertriglyceridaemia among individuals with atherosclerotic cardiovascular disease and potential eligibility for emerging therapies.
This supports the American Heart Association's statement that an “optimal” fasting TG level is less than 100 mg/dL and further suggests that an “optimal” level may be even lower.
The variability across previous studies may reflect differences in the study populations used. As we found, the slope of the relationship between TGs and CVD risk can vary across groups, including non-HDL and sex. Another explanation for this variability may be differences in the method used to select the reference group for analysis. For this reason, rather than binning patients by groups, we chose to analyze TGs as a continuous measure.
Interactions with sex and HDL-C
Another series of questions revolves around whether the association of TGs and CVD risk varies as a function of age, sex, BMI, diabetes, LDL-C, HDL-C, or statin use. Although diabetes and BMI are both associated with increasing TGs, we found no interaction between these factors and the association between TGs and CVD; this finding suggests that regardless of diabetes status or BMI, the risk associated with increasing TGs was similar. On the other hand, 2 statistically significant interactions were found based on sex and HDL-C, with the association between CVD and TGs being stronger in women and individuals with higher HDL-C levels. We show that the relationship between TGs and CVD risk appeared steeper in men than in women. In women, the TG-to-CVD risk association was flatter but plateaued at a higher value than in men. Given the smaller sample size of adults with TGs more than 150 mg/dL, the exact location of this inflection point by sex remains unclear. However, importantly, in both men and women, increasing TGs were associated with increased CVD risk even among those with TGs well below 150 mg/dL. Sex-specific differences in the TG-to-CVD association are not unprecedented. TG levels and CVD risk are known to vary significantly by sex. Previous observational data have demonstrated that TGs are more strongly associated with CVD risk in women than in men.
Plasma triglyceride level is a risk factor for cardiovascular disease independent of high-density lipoprotein cholesterol level: a meta-analysis of population-based prospective studies.
On the other hand, trials of fibrate therapy have found conflicting results for a sex-based interaction in the association between TG-lowering therapy and CVD events.
Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study.
Elevated triglyceride level is independently associated with increased all-cause mortality in patients with established coronary heart disease: twenty-two-year follow-up of the Bezafibrate Infarction Prevention Study and Registry.
One explanation for sex-based variance may be related to hormonal differentiation between men and women, leading to differences in the relative atherogenicity of TGs, as well as the type and composition of particles contributing to increasing TGs. Overall, the differences we find between TGs and CVD risk in men compared with women suggest that using different cutoffs by sex may allow for more accurate risk prediction.
We also identified a key interaction between HDL-C and TGs. The interaction between HDL-C and TGs is interesting in light of prior trials demonstrating a greater impact of TG-lowering therapies on CVD event rates in participants with lower HDL-C.
Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study.
In this analysis, overall increasing TGs were associated with greater increases in CVD risk among those with higher HDL-C; this association may be partially explained by the multivariable modeling performed adjusting for BMI and diabetes status. Low HDL-C and high TGs are both part of metabolic syndrome and often coexist. Therefore, in patients with lower HDL-C, some of the TG-related risks may be explained by the presence of metabolic syndrome and accounted for in multivariable modeling. Importantly, regardless of the HDL-C level, increasing TGs were associated with increasing CVD risk. Nevertheless, both TGs and HDL-C appear to have a role in CVD risk prediction. Isolated low HDL-C has not been shown to be a predictor of CVD independent of TGs and LDL-C,
Elevated triglyceride level is independently associated with increased all-cause mortality in patients with established coronary heart disease: twenty-two-year follow-up of the Bezafibrate Infarction Prevention Study and Registry.
We also demonstrate that greater TG-to-HDL-C ratio associates with increased CVD risk with an inflection point that occurs around a ratio of 4. This relationship remains significant after adjustment for a variety of key variables, supporting that TG elevation and reduced HDL-C in combination represent a higher risk phenotype, but only up until a certain point where conferred risk then levels off.
Our data support the use of TGs to identify patients at highest risk of CVD events. However, whether TGs themselves are a modifiable risk factor for CVD or if they serve as a marker for other risk factors remains unclear. Adjusting for factors associated with both CVD and high TGs such as diabetes, BMI, alcohol use, LDL-C, and HDL-C attenuated but did not fully account for the association between TGs and CVD events. However, using non–HDL-C, rather than LDL-C, in our models, did appear to attenuate the apparent independent association between TGs and CVD risk. That non–HDL-C, but not LDL-C, appeared to account for much of the association between TGs and CVD suggests that much of the apparent risk “captured” by TG measurements may be due to elevations in very-low-density lipoprotein and IDL particles, TG-rich lipoproteins, and apolipoprotein B (apoB)–containing lipoproteins. This is consistent with genetic studies suggesting that the risk associated with elevated TGs can be accounted for through their association with elevations in apoB.
Ultimately, a shift to increased use of non-HDL or apoB to assess lipid-related CVD risk may decrease the need to measure TGs to evaluate CVD risk. However, current clinical practice continues to rely on the measurement of LDL-C, TG, and HDL-C, not non–HDL-C, to assess lipid-related CVD risk.
Limitations
This study has several limitations. First, despite the large sample size, relatively few patients had TG levels above 250 mg/dL; therefore, CIs are wide around the shape of the relationship between TGs and CVD risk greater than this level. Thus, this study cannot answer the question of whether increasing TGs more than 250 mg/dL continues to increase risk. However, owing to the large numbers of patients with TGs < 200 mg/dL, the study is well powered to demonstrate the association between TGs and CVD risk in what would otherwise be considered a “normal” level. Second, this analysis is limited to individuals free of CVD and cannot be extrapolated to those with prevalent disease. Third, all of the measurements used were from fasting samples. Several studies have shown that nonfasting TG measurements may also provide important data regarding CVD risk.
Fourth, we did not adjust for all potential factors associated with TG levels including certain medications (ie, hormone replacement therapy, beta blockers). These factors may both mediate and confound the relationship between TGs and CVD.
Finally, we evaluated the concentration of TGs, but owing to a lack of apoB measurements at the baseline examination used in Framingham, we were unable to determine the composition of the TG-containing particles in these subjects. There are significant differences in the atherogenicity of chylomicrons compared with very- low-density lipoprotein cholesterol particles, and using TG concentration overall fails to capture the differences in these particle types.
Conclusion
In conclusion, we demonstrate a direct linear relationship between average TGs and CVD risk even at levels well below what would be considered “normal”, that is, 150 mg/dL, potentially due to the association between TGs and elevations in non–HDL-C. Our work adds to the body of evidence supporting an independent relationship between TG level and CVD risk and expands on previous observations of lowered CVD risk in patients with TGs far below the 150 mg/dL cutoff. We find this relationship to be most pronounced in women and those with higher HDL-C, suggesting that TGs may play a stronger role in these specific populations.
Acknowledgments
The authors would like to thank Erin Campbell, MS, for her editorial contributions to this manuscript.
Authors' contributions: T.A. participated in study design, interpretation of the data, and drafting and critical revision of the manuscript; E.D.P. and N.J.P. participated in study conception, study design, interpretation of data, and critical revision of the manuscript; H.M. and D.M.W. participated in study design, data analysis, acquisition of data, interpretation of data, and critical revision of the manuscript; S.P. and C.G. participated in interpretation of data and critical revision of the manuscript; A.M.N. participated in study conception, study design, acquisition of the data, interpretation of the data, and drafting and critical revision of the manuscript; all authors have approved of the final submitted version of the manuscript.
Conflict of interest: T. Aberra reports no relevant conflicts of interest. E.D. Peterson reports receiving research grants from Amgen, Sanofi, AstraZeneca, and Merck and served as a consultant/member of the advisory board of Amgen, AstraZeneca, Merck, and Sanofi Aventis. N.J. Pagidipati reports receiving research grants from Alexion Pharmaceuticals, Inc., Amarin Pharmaceutical Company, Amgen, Inc., AstraZeneca, Baseline Study LLC, Boehringer Ingelheim, Duke Clinical Research Institute, Eli Lilly & Company, Novo Nordisk Pharmaceutical Company, Regeneron Pharmaceuticals, Inc., Sanofi S.A., and Verily Life Sciences Research Company. H. Mulder reports no relevant disclosures. D.M. Wojdyla reports no relevant disclosures. S. Philip reports being an employee and stock shareholder of Amarin Pharma, Inc. C. Granowitz reports being an employee and stock shareholder of Amarin Pharma, Inc. A.M. Navar reports receiving research grant from Amarin, Janssen, Amgen, Sanofi, and Regeneron Pharmaceuticals and served as a consultant/member of the advisory board of Amarin, Amgen, Esperion, Novartis, New Amsterdam, Pfizer, BI, Novo Nordisk, AstraZeneca, Janssen, The Medicines Company, Sanofi, and Regeneron.
Financial disclosure
This study was funded by Amarin, and Dr. Navar receives support from NIH, United States K01HL133416. The study sponsor co-authors participated in the design of the study, interpretation of the results, and contributed to the manuscript. DCRI investigators controlled all analyses and the decision to publish.
Appendix
Supplement Table 1Characteristics of patients by baseline quartile of prior average TG characteristics
Characteristic
Q1
Q2
Q3
Q4
N = 2016
N = 2017
N = 2019
N = 2016
Age
57, 54–61
58, 55–62
59, 56–62
59, 56–62
Female
1267 (62.8%)
1213 (60.1%)
1137 (56.3%)
939 (46.6%)
White
1612 (80.0%)
1641 (81.4%)
1723 (85.3%)
1814 (90.0%)
History of smoking
274 (13.7%)
343 (17.0%)
334 (16.6%)
357 (17.8%)
History of diabetes
104 (5.2%)
158 (7.8%)
245 (12.2%)
463 (23.0%)
Systolic BP
119, 109–130
123, 111–134
123, 113–136
126, 115–138
Diastolic BP
72, 66–79
73, 67–80
72, 66–79
73, 67–80
BMI
26, 23–29
27, 24–31
29, 26–32
29, 27–33
Ounces of ethanol per week
0, 0–3
0, 0–2
0, 0–2
0, 0–2
Treatment for hypertension
348 (17.3%)
478 (23.7%)
613 (30.4%)
731 (36.4%)
Cholesterol medication
On statin
46 (2.3%)
105 (5.2%)
162 (8.0%)
316 (15.7%)
No statin, on other cholesterol medication
15 (0.7%)
18 (0.9%)
36 (1.8%)
92 (4.6%)
Not on cholesterol medication
1951 (97.0%)
1892 (93.9%)
1817 (90.2%)
1603 (79.7%)
Fasting glucose
95, 89–101
98, 92–106
100, 93–109
104, 96–119
Total cholesterol
189, 169–210
199, 179–223
206, 183–230
207, 185–235
HDL-C
59, 49–71
51, 42–62
46, 38–55
39, 33–47
Non–HDL-C
127, 108–150
146, 125–168
158, 135–181
167, 144–193
Triglyceride:HDL-C ratio
1.10, 0.87–1.40
1.85, 1.49–2.29
2.82, 2.29–3.45
5.15, 3.99–6.90
LDL-C calculated
112, 93–135
125, 105–146
129, 107–151
123, 102–147
BP, blood pressure; BMI, body mass index; TG, triglyceride; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol.
TG quartiles defined as Q1: min to 81.75; Q2: 81.75 to 110.19; Q3: 110.2 to 153.5; Q4: 153.5 to max.
Supplement Figure 1Distribution of TG levels in each cohort by sex bar graph and superimposed linear graph representing the distribution of TG levels in each cohort stratified by sex. ARIC, Atherosclerosis Risk in Communities; TG, triglyceride.
Supplement Figure 3CVD Kaplan-Meier event rates by tertiles of average TG and HDL-C. A 3D bar plot representing Kaplan-Meier event rates by average TG and by HDL-C. CVD, cardiovascular disease; HDL-C, high-density lipoprotein cholesterol; TG, triglyceride.
The Prospective Cardiovascular Munster (PROCAM) study: prevalence of hyperlipidemia in persons with hypertension and/or diabetes mellitus and the relationship to coronary heart disease.
Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III).
Plasma triglyceride level is a risk factor for cardiovascular disease independent of high-density lipoprotein cholesterol level: a meta-analysis of population-based prospective studies.
Coronary heart disease prediction from lipoprotein cholesterol levels, triglycerides, lipoprotein(a), apolipoproteins A-I and B, and HDL density subfractions: The Atherosclerosis Risk in Communities (ARIC) Study.
Real-world risk of cardiovascular outcomes associated with hypertriglyceridaemia among individuals with atherosclerotic cardiovascular disease and potential eligibility for emerging therapies.
Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study.
Elevated triglyceride level is independently associated with increased all-cause mortality in patients with established coronary heart disease: twenty-two-year follow-up of the Bezafibrate Infarction Prevention Study and Registry.
30079-9&id=gr1.jpg){kind=link}
30079-9&id=gr2.jpg){kind=link}
30079-9&id=gr3.jpg){kind=link}
30079-9&id=gr4.jpg){kind=link}
30079-9&id=fx1.jpg){kind=link}
30079-9&id=fx2.jpg){kind=link}
30079-9&id=fx3.jpg){kind=link}