BACKGROUND: Nonfasting triglycerides are similar or superior to fasting triglycerides at predicting cardiovascular events. However, diagnostic cutpoints are based on fasting triglycerides. We examined the optimal cutpoint for increased nonfasting triglycerides.
METHODS: We obtained baseline nonfasting (<8 h since last meal) samples from 6391 participants in the Women's Health Study who were followed prospectively for ≤17 years. The optimal diagnostic threshold for nonfasting triglycerides, determined by logistic regression models by use of c-statistics and the Youden index (sum of sensitivity and specificity minus 1), was used to calculate hazard ratios (HRs) for incident cardiovascular events. Performance was compared to thresholds recommended by the American Heart Association (AHA) and European guidelines.
RESULTS: The optimal threshold was 175 mg/dL (1.98 mmol/L), with a c-statistic of 0.656, statistically better than the AHA cutpoint of 200 mg/dL (c-statistic 0.628). For nonfasting triglycerides above and below 175 mg/dL, after adjusting for age, hypertension, smoking, hormone use, and menopausal status, the HR for cardiovascular events was 1.88 (95% CI 1.52–2.33, P < 0.001), and for triglycerides measured at 0–4 and 4–8 h since the last meal, 2.05 (1.54– 2.74) and 1.68 (1.21–2.32), respectively. We validated performance of this optimal cutpoint by use of 10-fold cross-validation and bootstrapping of multivariable models that included standard risk factors plus total and HDL cholesterol, diabetes, body mass index, and C-reactive protein.
CONCLUSIONS: In this study of middle-aged and older apparently healthy women, we identified a diagnostic threshold for nonfasting hypertriglyceridemia of 175 mg/dL (1.98 mmol/L), with the potential to more accurately identify cases than the currently recommended AHA cutpoint.
High triglycerides may promote atherosclerosis via the accumulation of triglyceride-rich remnant particles within the endothelium (1–9). The nonfasting state provides a better indication of average lipid concentrations in the blood than the fasting state, since most people fast for only a few hours in the early morning (1, 10, 11). From a purely practical standpoint, nonfasting samples are much easier to obtain; yet fasting samples remain the standard for measurement of triglycerides and cholesterol because measuring lipids in the fasting state reduces variability and allows for a more accurate derivation of the Friedewald equation (10). However, studies suggest that the variability missed in fasting samples may capture important information about an individual's metabolic capacity (e.g., through delayed clearance of triglyceride remnants) (12, 13). An accumulation of recent evidence from prospective cohort data has shown nonfasting triglycerides to be equivalent, if not superior, to fasting lipids at predicting cardiovascular disease (12, 18). Furthermore, triglycerides and triglyceride-rich remnant lipoproteins have been associated with—and may be causally linked to—the development of atherosclerosis and even all-cause mortality (19, 23).
Thus, it is important to determine appropriate clinical cutpoints at which to diagnose and treat nonfasting hypertriglyceridemia. Current national guidelines set the borderline-high value for defining fasting triglycerides at 150 mg/dL (1.70 mmol/L) and the high value at 200 mg/dL (2.28 mmol/L) on the basis of population percentiles (8). Together, these 2 groups roughly correspond to the top one-third of the population in developed nations (8, 24, 27). Different panels have proposed various cutpoints for nonfasting hypertriglyceridemia: 175 mg/dL (1.98 mmol/L) (European Atherosclerosis Society) (28), 180 mg/dL (2.03 mmol/L) (Athens Expert Panel) (11), and 200 mg/dL (2.26 mmol/L) (American Heart Association [AHA])5 (8), although the underlying rationales for these cutpoints are unclear. It is possible that these values were extrapolated from fasting cutoffs, given that triglycerides increase by approximately 20%–30% from their baseline fasting levels and remain stable for 2–4 hours after a meal (29). To our knowledge, no evidence exists to suggest that any of the above diagnostic cutpoints for increases in nonfasting triglycerides is superior to the others at predicting cardiovascular outcomes. Therefore, in a prospective cohort of 28 345 apparently healthy women followed for ≤17 years for incident cardiovascular disease (CVD) events, we attempted to determine the optimal diagnostic cutpoint for hypertriglyceridemia in the nonfasting state.
Materials and Methods
The study cohort was derived from participants in the Women's Health Study, a previously completed randomized controlled trial of aspirin and vitamin E in the primary prevention of CVD and cancer among 39876 apparently healthy women (30). The study protocol was approved by the institutional review board of Brigham and Women's Hospital (Boston, MA), and all participants provided written informed consent. This study was registered at Clinicaltrials.gov (identifier NCT00000479).
Baseline demographic data and health histories were obtained from women at enrollment. At that time, participants were asked to provide a blood sample if they were willing; 28 345 (71.1%) women did so, and those samples are the source of the lipid measurements in this study. The time since their last meal before the blood draw was self-reported. Fasting participants were defined as those whose last meal was ≥8 h before blood draw (n = 20 118). Nonfasting participants consisted of those who had eaten within 8 h of blood draw (n = 6391). Those with unknown time since last meal or missing baseline lipid measurements (n = 1836) were excluded from this analysis.
Blood samples were collected in tubes containing EDTA. The samples were centrifuged upon collection, and the plasma was stored in liquid nitrogen (−170 °C) until analysis. Subsequently, in a core laboratory certified by the National Heart, Lung, and Blood Institute/Centers for Disease Control and Prevention Lipid Standardization program, samples were thawed and analyzed for standard lipids as previously described (12). Direct measurement of concentrations of total cholesterol, HDL cholesterol (HDL-C), LDL cholesterol (LDL-C), and triglycerides was simultaneously performed on the Hitachi 917 analyzer (Roche Diagnostics) with reagents and calibrators from Roche Diagnostics. Triglycerides were measured enzymatically with correction for endogenous glycerol by use of a Hitachi 917 analyzer and reagents and calibrators from Roche Diagnostics. Triglycerides at concentrations of 84.0 and 201.8 mg/dL (0.95 and 2.28 mmol/L) were measured in the laboratory with a day-to-day reproducibility of 1.8% [SD 1.6 mg/dL (0.2 mmol/L)] and 1.7% [SD 2.5 mg/dL (0.3 mmol/L)], respectively. Total cholesterol and HDL-C were measured enzymatically on a Hitachi 911 autoanalyzer, and LDL-C was measured by a homogeneous direct method from Roche Diagnostics.
The primary outcome of interest was total CVD events (nonfatal myocardial infarction, nonfatal ischemic stroke, coronary revascularization, and death due to cardiovascular causes) (12, 14). Myocardial infarction was defined by WHO criteria of characteristic symptoms of chest pain accompanied by increased concentrations of cardiac enzymes or diagnostic electrocardiographic changes. Stroke was defined as a new neurologic deficit of sudden onset that persisted for ≥24 h. Computed tomography scans or magnetic resonance images of the head were available for most events and were used to distinguish ischemic from hemorrhagic strokes. Coronary revascularization included percutaneous coronary interventions and coronary artery bypass graft surgery. All events were adjudicated by an endpoints committee. In participants with >1 cardiovascular event, only the first was used in these analyses. Follow-up morbidity and mortality data were available for 97.2% and 99.4%, respectively, of the Women's Health Study participants in this study.
All analyses were performed on nonfasting participants with the exception of baseline characteristics, which included fasting and nonfasting participants. Baseline characteristics included age, hypertension status, smoking status, diagnosis of diabetes mellitus, postmenopausal status, postmenopausal hormone use, and high-sensitivity C-reactive protein (hsCRP). We analyzed differences between baseline characteristics of participants in the nonfasting and fasting populations with the t-test in 2-group comparisons or Pearson χ2 test for proportions.
Follow-up of this cohort was virtually complete through 8 years, and event rate was low. In this scenario, both logistic regression and Cox proportional hazards regression produce asymptotically consistent estimates (31), so we used the more user-friendly logistic regression to estimate the optimal threshold. We determined the optimal diagnostic threshold for nonfasting triglycerides by evaluating the area under the ROC curve (c-statistic) (32) in univariable logistic regression models, with a composite 8-year CVD event as a dependent variable and the dichotomized concentration of nonfasting triglycerides as an independent predictor. Subsequent analyses evaluated this optimal threshold in the full 17-year follow-up by use of Cox proportional hazard models for incident CVD.
By varying the dichotomization threshold from 100 to 300 mg/dL (1.13 to 3.39 mmol/L) by increments of 25 mg/dL (0.28 mmol/L), we obtained the concentration of nonfasting triglycerides that optimized the c-statistic and equivalently the Youden index (sum of sensitivity plus specificity minus 1). This value has the optimal balance of sensitivity and specificity. Next, in a multivariable analysis, we used Cox proportional hazards regression models to compare dichotomized triglycerides at this optimal cutpoint value with the following alternative cutpoints selected by expert panels: 175 mg/dL (1.98 mmol/L) (European Atherosclerosis Society) (27), 180 mg/dL (2.03 mmol/L) (Athens Expert Panel) (11), and 200 mg/dL (2.26 mmol/L) (AHA) (8). We used 3 multivariable models to control for potential confounders and/or mediators. Model 1 adjusted for age, postmenopausal status, hormone replacement therapy use, smoking status, and hypertension (defined as history of hypertension or use of antihypertensive medication). To determine the predictive value of triglycerides independent of other lipids, model 2 was additionally adjusted for total cholesterol and HDL-C. Model 3 adjusted for covariates in model 2 plus diabetes mellitus, body mass index (BMI), and hsCRP, as these variables may be in the causal pathway for the association of triglycerides with CVD.
Additionally, to validate the results and avoid overoptimism, we compared 3 different nonfasting triglyceride thresholds in the 3 models using 10-fold cross-validation and bootstrapping: differences in cross-validated c-index between the 3 thresholds were calculated, and their 95% CIs were estimated by the bootstrap method. Because there was almost no censoring, c-index is an appropriate performance measure of the survival model (32). A better threshold is expected to produce higher c-index, cross-validation mitigates overoptimism, and bootstrap produces a CI with a better coverage probability in this setting (33, 36).
To further explore the association of time since last meal with future CVD events in nonfasting individuals, we used Cox proportional hazard regression models to study effect modification of nonfasting triglycerides by time since last meal (0 to <4 h or 4 to <8 h). To assess interaction between nonfasting triglycerides and hours since last meal, we included a cross-product term. To ascertain the independence of triglycerides from HDL-C and test for multiplicative interactions in predicting CVD risk, the fully adjusted model for the optimal threshold of nonfasting triglycerides was also run with prespecified clinical categories of HDL-C [less than, greater than, or equal to 50 mg/dL (1.30 mmol/L)]. All P values were 2-tailed, and P < 0.05 was considered statistically significant. All statistical analyses were performed with SAS version 9.3 (SAS Institute Inc.).
Baseline characteristics of participants in the study (Table 1) were similar between fasting and nonfasting individuals for all of the variables examined (at a 0.05 significance level). Compared with the fasting group, the nonfasting participants tended to be slightly younger, were less likely to have hypertension, had lower concentrations of LDL-C, and were more likely to have diabetes.
Of the 6391 nonfasting participants, 136 developed incident CVD in 8 years, and 353 developed incident CVD in 17 years. The ROC curve over all concentrations of nonfasting triglycerides is shown in Fig. 1 (c-statistic = 0.656). The optimal threshold for triglyceride value for predicting the incidence of CVD during this duration was 175 mg/dL (1.98 mmol/L), corresponding to the maximum Youden index and c-statistic by use of dichotomized triglycerides (Table 2) over a composite 8-year CVD event period that minimized censoring (censoring rate 94%). It also corresponded to the value proposed by the European Atherosclerosis Society (27). The other proposed values of 180 mg/dL (2.03 mmol/L) and 200 mg/dL (2.26 mmol/L) were also examined, but each produced a lower c-statistic than 175 mg/dL (1.98 mmol/L). Because model fitting and model evaluation were performed in the same data set, we additionally implemented a bootstrap with 10-fold cross-validation to avoid overoptimism. Results are reported in Table 3.
Hazard ratios (HRs) and 95% CIs for the 17-year follow-up were estimated for the optimal cutoff of 175 mg/dL (1.98 mmol/L) in association with incident CVD events both crudely (Fig. 2) and adjusted for multiple factors (Table 4). After adjusting for age, history of hypertension, smoking, use of hormone therapy, and postmenopausal status (model 1), nonfasting triglyceride concentrations ≥175 mg/dL (≥1.98 mmol/L) were strongly associated with CVD events (HR 1.88, 95% CI 1.52–2.33, P < 0.001). In model 2, after adjusting for total cholesterol and HDL-C in addition to the variables in model 1, the association was somewhat attenuated, but nonfasting triglycerides ≥175 mg/dL (≥1.98 mmol/L) remained significantly associated with CVD (HR 1.36, 95% CI 1.06–1.75, P = 0.02). After adjusting for variables in the causal pathways (diabetes, BMI, and CRP) in addition to model 2 variables, the association for incident CVD events over 17-year follow-up for nonfasting triglycerides ≥175 mg/dL (≥1.98 mmol/L) was further attenuated (HR 1.25, 95% CI 0.96–1.62, P = 0.10). There was no statistical interaction between hours since last meal (0 to <4 and 4 to <8 h) and the association of triglycerides with CVD events (P for interaction >0.05 for all 3 models).
In this prospective cohort of 28 345 apparently healthy women followed for ≤17 years, we observed that the optimal threshold for the diagnosis of hypertriglyceridemia in the nonfasting state was 175 mg/dL (1.98 mmol/L), which more accurately predicted CVD compared with the currently recommended AHA value of 200 mg/dL (2.26 mmol/L). To our knowledge, this is the first study that has prospectively validated a diagnostic cutpoint for nonfasting triglycerides in relation to incident CVD events in a healthy population. Furthermore, the association of the identified threshold with incident CVD was not affected by postprandial duration.
Increased nonfasting triglycerides are associated with higher cardiovascular risk in several studies (12, 18). The most plausible explanation for this increased risk is that nonfasting triglycerides signify the presence of atherogenic remnant lipoproteins. These lipoproteins contain a degree of cholesterol that is not accounted for in typical fasting triglyceride samples or LDL-specific measurements. Because all human cells can degrade triglycerides but not cholesterol, it is likely that the cholesterol content of the triglyceride-rich remnant particles enters the arterial intima and contributes to atherosclerosis. There is evidence to suggest that, once trapped inside the intima, remnant particles may be trapped inside the arterial wall preferentially to LDL, simply because of their larger size and attachment to extracellular proteoglycans (6). Unlike with LDL particles, triglyceride-rich remnant molecules can be taken up directly by macrophages, leading to foam cell formation (33). Another novel mechanism by which triglycerides may predispose an individual to CVD involves the concept that lipoprotein lipase activity at the surface of triglyceride-rich remnant particles acts on the vascular endothelium or within the intima to precipitate the release of free fatty acids, resulting in local injury and inflammation (1, 33, 37).
This study has several clinical implications. Practitioners who would like to incorporate nonfasting lipid measurements into their practice are hobbled by having to rely on fasting triglyceride cutpoints, which have not been studied or validated in nonfasting populations. Patient compliance may become a hindrance to the interpretation of fasting samples. Furthermore, the use of simplified diagnostic criteria with clinical relevance will be more accessible to the increasingly overburdened physician (38). This study builds on the evidence that nonfasting samples can accurately capture prognostic data for both triglycerides and cholesterol by establishing validated cutpoints that can now be used to help guide clinical decision making (12, 39). Indeed, a nonfasting lipid profile has been the standard in Denmark since 2009 (1).
Given that hypertriglyceridemia is a cardiovascular risk factor of evolving significance, it is necessary to create diagnostic tools to assess when these concentrations place a person at risk for clinically important endpoints (myocardial infarction, ischemic stroke, or death by cardiovascular causes). Nonfasting triglycerides provide a more accurate assessment of an individual's average metabolic state than fasting triglycerides (11). One explanation for this phenomenon is that triglycerides do not return to basal concentrations until ≥8 h after a meal, and clearance of triglycerides from the bloodstream can be delayed for ≥12 h in patients with insulin resistance or a predisposition to producing remnant particles (10). Thus, we spend the vast majority of our time in nonfasting conditions. However, fasting samples have been the standard for measurement of triglycerides and cholesterol because measuring these lipids in the fasting state (a) reduces variability and thus increases precision and (b) allows for a more accurate derivation of the Friedewald equation for calculating LDL-C. Recent data also suggest that nonfasting LDL-C has prognostic value similar to that of fasting LDL-C (40). Triglyceride-rich remnant molecules, composed of triglycerides, cholesterol, and proteins, are associated with increased cardiovascular risk in multiple studies (12, 18). New advances in genetics have shown that triglycerides are only 1 component of the causal pathway to CVD and that mutations involving lipoprotein metabolism and function directly affect phenotype (20, 23). This has important implications for future therapeutic targets aimed at reducing triglyceride-rich remnant molecules.
The strengths of the study include its large sample size, prospective design, and extended follow-up time. Application of rigorous methods such as bootstrap and 10-fold cross-validation further supports the optimality of the 175 mg/dL (1.98 mmol/L) threshold. Nonetheless, limitations of this study also merit consideration. First, participants were not randomly assigned to fasting or nonfasting status. The participants chose whether to fast and for how long, which may introduce sampling bias. However, the baseline characteristics of all the participants were similar, and there was no interaction of hours since the last meal within our nonfasting cohort. Also, there was no standardization of the meal given (e.g., proportion of fat); however, standardization would result, if anything, in an underestimation of the true effect. Given the variability of triglyceride concentrations, the single measurement of concentrations at study enrollment without repeated sampling could lead to regression dilution bias, but this would again bias the results toward a null finding. Our study population was all women and limited to mostly whites. Further studies should examine outcomes in men and other ethnicities. Finally, this study focuses on diagnosis, but areas of therapeutic intervention continue to be controversial. The most important lifestyle modification (and least controversial) is to lose weight through eating less and exercising more (1). Other treatments for lowering triglycerides include ω3 fatty acids (fish oils), statins, fibrates, and niacin (8).
In summary, this is the first study to identify a diagnostic threshold for nonfasting hypertriglyceridemia in a large, prospective cohort of apparently healthy individuals. As the evidence for the link between triglycerides and cardiovascular disease increases, identifying early points of intervention in the prevention of CVD are crucial for preventive public health efforts. Given the study entry criteria, additional studies should be done to assess the generalizability of our results in women younger than 45 years, men, and more ethnically diverse populations.
We thank the participants, staff, and investigators of the Women's Health Study for their valuable contributions.
↵5 Nonstandard abbreviations:
- American Heart Association;
- cardiovascular disease;
- HDL cholesterol;
- LDL cholesterol;
- high-sensitivity C-reactive protein;
- body mass index;
- hazard ratio.
(see editorial on page 1123)
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:
Employment or Leadership: None declared.
Consultant or Advisory Role: P.M Ridker, ISIS, Boston Heart, Genzyme, and Jannson; S. Mora, Genzyme, Lilly, Pfizer, Quest Diagnostics, and Cerenis Therapeutics.
Stock Ownership: None declared.
Honoraria: P.M Ridker, Lilly and Genentech.
Research Funding: The Women's Health Study is supported by grants HL117861, HL43851, HL 080467, and CA 47988 from the National Heart, Lung, and Blood Institute (NHLBI) and National Cancer Institute (NCI). A. Akinkuolie, NHLBI (T32 HL007575); P.M Ridker, Novartis, AstraZeneca, Aegerion, ISIS, Boehringer, Pfizer, NHLBI, NCI, and Amgen; S. Mora, NHLBI, AstraZeneca, and Atherotech Diagnostics.
Expert Testimony: None declared.
Patents: P.M Ridker, coinventor on patents held by Brigham and Women's Hospital.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
- Received for publication March 30, 2015.
- Accepted for publication May 26, 2015.
- © 2015 American Association for Clinical Chemistry