Fasting Compared with Nonfasting Triglycerides and Risk of Cardiovascular Events in Women

ORIGINAL CONTRIBUTION

Fasting Compared With Nonfasting Triglycerides and Risk of Cardiovascular Events in Women

Sandeep Bansal, MD Context The association of triglycerides with incident cardiovascular disease re- Julie E. Buring, ScD mains controversial. Although triglyceride levels are typically obtained in the fasting Nader Rifai, PhD state, postprandial hypertriglyceridemia may play an important role in atherosclerosis. Objective To determine the association of triglyceride levels (fasting vs nonfasting) Samia Mora, MD, MHS and risk of future cardiovascular events. Frank M. Sacks, MD Design, Setting, and Participants Prospective study of 26 509 initially healthy Paul M Ridker, MD, MPH US women (20 118 fasting and 6391 nonfasting) participating in the Women’s Health Study, enrolled between November 1992 and July 1995 and undergoing follow-up N CONTRAST TO TOTAL CHOLES- for a median of 11.4 years. Triglyceride levels were measured in blood samples ob- terol, low-density lipoprotein cho- tained at time of enrollment. lesterol (LDL-C), and high- Main Outcome Measure Hazard ratios for incident cardiovascular events (non- density lipoprotein cholesterol fatal myocardial infarction, nonfatal ischemic stroke, coronary revascularization, or car- (HDL-C),I which are well-established in- diovascular death). dependent risk factors for cardiovas- 1 Results At baseline, triglyceride levels in fasting as well as nonfasting women corre- cular disease, the importance of tri- lated with traditional cardiac risk factors and markers of insulin resistance. During a median glycerides remains controversial. In part follow-up of 11.4 years, 1001 participants experienced an incident cardiovascular event this controversy reflects the fact that, (including 276 nonfatal myocardial infarctions, 265 ischemic strokes, 628 coronary revas- due to the inverse correlation of tri- cularizations, and 163 cardiovascular deaths), for an overall rate of 3.46 cardiovascular glyceride levels with those of HDL-C, events per 1000 person-years of follow-up. After adjusting for age, blood pressure, smok- adjustment for HDL-C attenuates the ing, and use of hormone therapy, both fasting and nonfasting triglyceride levels pre- relationship between triglycerides and dicted cardiovascular events. Among fasting participants, further adjustment for levels of cardiovascular disease. A recent meta- total and high-density lipoprotein cholesterol and measures of insulin resistance weakened this association (fully adjusted hazard ratio [95% confidence interval] for increasing analysis suggested that the adjusted risk tertiles of triglyceride levels: 1 [reference], 1.21 [0.96-1.52], and 1.09 [0.85-1.41] [P=.90 ratio for coronary heart disease among for trend]). In contrast, nonfasting triglyceride levels maintained a strong independent individuals in the highest third of tri- relationship with cardiovascular events in fully adjusted models (hazard ratio [95% con- glyceride levels compared with those in fidence interval] for increasing tertiles of levels: 1 [reference], 1.44 [0.90-2.29], and 1.98 the lowest third decreases from ap- [1.21-3.25] [P=.006 for trend]). In secondary analyses stratified by time since partici- proximately 2.0 to 1.5 after account- pants’ last meal, triglyceride levels measured 2 to 4 hours postprandially had the stron- ing for HDL-C levels.2 gest association with cardiovascular events (fully adjusted hazard ratio [95% confidence Ͻ A second aspect of the controversy interval] for highest vs lowest tertiles of levels, 4.48 [1.98-10.15] [P .001 for trend]), and this association progressively decreased with longer periods of fasting. stems from the manner in which triglyceride levels are typically mea- Conclusions In this cohort of initially healthy women, nonfasting triglyceride levels sured. Current national guidelines rec- were associated with incident cardiovascular events, independent of traditional car- ommend that blood for lipid profiles be diac risk factors, levels of other lipids, and markers of insulin resistance; by contrast, fasting triglyceride levels showed little independent relationship. drawn after an 8- to 12-hour fast.1 Be- JAMA. 2007;298(3):309-316 www.jama.com cause plasma triglyceride levels can increase substantially postprandially, fast- Author Affiliations: Donald W. Reynolds Center for Department of Laboratory Medicine, Children’s ing levels ostensibly avoid the variability Cardiovascular Research (Drs Bansal, Mora, and Hospital (Dr Rifai); and Department of Nutrition, Ridker), Leducq Center for Molecular and Genetic Harvard School of Public Health (Dr Sacks), Boston, associated with meals and provide a Epidemiology (Drs Rifai, Mora, and Ridker), Center Massachusetts. for Cardiovascular Disease Prevention and Division Corresponding Author: Paul M Ridker, MD, MPH, of Preventive Medicine (Drs Bansal, Buring, Mora, Center for Cardiovascular Disease Prevention, Brigham See also pp 299 and 336. and Ridker), and Division of Cardiology (Drs Mora and Women’s Hospital, 900 Commonwealth Ave E, and Ridker), Brigham and Women’s Hospital; Boston, MA 02215 ([email protected]).

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

more stable estimate for risk assess- blood sample for analysis. Partici- at least 24 hours; computed tomogra- ment. However, postprandial lipids may pants whose last meal was 8 or more phy scans or magnetic resonance im- play an important role in the patho- hours prior to their blood draw com- ages were available for most events and genesis of cardiovascular disease be- prised the fasting cohort (n=20 118), were used to distinguish ischemic from cause postprandial triglyceride-rich and those who had eaten within 8 hours hemorrhagic strokes. Coronary revas- remnant lipoproteins can penetrate the of their blood draw comprised the non- cularization included percutaneous coro- endothelial cell layer and reside in the fasting cohort (n=6391); those with un- nary interventions and coronary artery subendothelial space, where they can known time since last meal (n=1430) bypass graft surgery. All events were ad- contribute to the formation of foam were excluded from the analysis. judicated by an end points committee. cells, a hallmark of early atherosclero- In participants with more than 1 cardio- sis.3-5 Elevated postprandial levels of tri- Laboratory Methods vascular event, only the first was used in glycerides, via higher peak concentra- Blood samples were collected at enroll- these analyses. Follow-up morbidity and tions or delayed clearance, also might ment in tubes containing EDTA and mortality data were available for 97.2% represent an abnormal response to an were shipped cold overnight to a core and 99.4%, respectively, of the Wom- oral fat load that reflects insulin resis- laboratory certified by the US Centers en’s Health Study participants. tance,6-8 a condition associated with a for Disease Control and Prevention and host of metabolic abnormalities that the National Heart, Lung, and Blood In- Statistical Analyses predispose an individual to cardiovas- stitute Lipid Standardization Pro- All analyses were performed sepa- cular disease.9,10 To clarify the impor- gram. There, samples were centri- rately for fasting and nonfasting par- tance of the prandial state when mea- fuged and the plasma was stored in ticipants. For the main analyses, tri- suring triglyceride levels, we evaluated liquid nitrogen until analysis. glyceride levels were categorized into the association of fasting and nonfast- Triglyceride levels were measured en- tertiles for consistency with a recent ing levels with incident cardiovascu- zymatically, with correction for endog- meta-analysis2 and for stability of effect lar events in a large prospective co- enous glycerol,14 using a Hitachi 917 estimates by maintaining adequate hort of initially healthy women, analyzer and reagents and calibrators numbers of events in reference groups. independent of traditional cardiac risk from Roche Diagnostics (Indianapo- Because distributions differed be- factors, levels of other lipids, and mark- lis, Indiana). Triglycerides at concen- tween fasting and nonfasting partici- ers of insulin resistance. trations of 84.0 and 201.8 mg/dL (to pants, separate tertile cut points were convert to millimoles per liter, multi- defined for each cohort. To address METHODS ply by 0.0113) were determined in the whether our results might differ on the Study Participants laboratory with a day-to-day reproduc- basis of cut points used, secondary The study cohort was derived from par- ibility of 1.8% (SD, 1.6 mg/dL) and 1.7% analyses were performed dividing the ticipants in the Women’s Health Study, (SD, 3.5 mg/dL), respectively. Levels of study population into quintiles and a previously completed randomized total cholesterol and HDL-C were mea- treating triglyceride levels as a natural controlled trial of aspirin and vitamin sured enzymatically on a Hitachi 911 logarithm–transformed continuous E in the primary prevention of cardio- autoanalyzer (Roche Diagnostics, Basel, variable. vascular disease and cancer in women Switzerland), and levels of LDL-C were Differences between baseline char- aged 45 years and older. Details of the determined directly (Genzyme, Cam- acteristics of participants within each study design and primary outcomes bridge, Massachusetts). These assays are triglyceride category were analyzed have been reported.11-13 The study pro- approved for clinical use by the US Food using the Cochrane-Armitage test for tocol was approved by the institu- and Drug Administration. trend for proportions and analysis of tional review board of Brigham and variance for continuous measures. Non- Women’s Hospital, Boston, Massachu- Outcomes parametric Spearman rank correla- setts. All participants provided writ- Incident cardiovascular events were a tions were computed for triglycerides ten informed consent. composite of confirmed nonfatal myo- with other biomarkers due to the right- All participants provided baseline cardial infarction, nonfatal ischemic skewed distribution. demographic data and health histo- stroke, coronary revascularization, or Cox proportional hazard models ries and were asked to provide a blood death due to cardiovascular causes. Myo- were used to compute hazard ratios sample at enrollment. Participants were cardial infarction was defined by World (HRs) and 95% confidence intervals requested, but not required, to have the Health Organization criteria of charac- (CIs) for each tertile or quintile of tri- sample drawn in the morning before teristic symptoms accompanied by el- glyceride level, using the lowest cat- eating, and reported the number of evated levels of cardiac enzymes or by di- egory as the reference group. Linear hours since their last meal before the agnostic electrocardiographic changes. tests for trend were performed using the blood draw. Of the full cohort of 39 876 Stroke was defined as a new neurologic median triglyceride value within each women, 27 939 provided a baseline deficit of sudden onset that persisted for tertile or quintile as an ordinal vari-

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

able. All analyses were adjusted for ran- further explore the importance of the cut- tailed in TABLE 1. Among both fasting dom treatment assignment to aspirin, off time for fasting participants being de- and nonfasting participants, women vitamin E, and beta carotene. Models fined as at least 8 hours since their last with higher levels of triglycerides were were initially adjusted for potential con- meal, models were run with partici- significantly more likely to have other founders by traditional nonlipid car- pants categorized by postprandial time cardiac risk factors and markers for the diac risk factors (age, blood pressure, (Ͻ2, 2-4, 4-8, 8-12, or Ն12 hours). metabolic syndrome. However, when smoking status, and use of hormone Analyses were repeated using each in- comparing fasting with nonfasting par- therapy). To determine the role of tri- dividual component of the composite ticipants, no substantive differences glycerides independent of other lipid cardiovascular end point as the out- were seen for alcohol use, exercise fre- markers, models were then adjusted for come. To ascertain the independence of quency, total cholesterol, HDL-C, body levels of total cholesterol and HDL-C. triglycerides from HDL-C and to test for mass index, high-sensitivity C- Final multivariable models were addi- multiplicative interactions in predict- reactive protein, or glycated hemoglo- tionally adjusted for history of diabe- ing cardiovascular risk, models were run bin. Compared with the fasting group, tes mellitus, body mass index, and natu- using prespecified clinical categories of those in the nonfasting group were ral logarithm–transformed levels of HDL-C (Ͻ or Ն50 mg/dL [to convert to slightly younger, less likely to have hy- high-sensitivity C-reactive protein as millimoles per liter, multiply by 0.0259]) pertension, and had lower levels of metabolic risk factors related to insu- and triglycerides (Ͻ or Ն150 mg/dL). LDL-C. Fasting and nonfasting levels lin resistance with the potential to affect All P values were 2-tailed, and PϽ.05 of triglycerides were significantly cor- lipid levels.8,15 Proportionality of haz- was considered statistically signifi- related with other measured biomark- ards was confirmed by Wald ␹2 testing cant. All statistical analyses were per- ers, and correlation coefficients were of the interaction between natural loga- formed using SAS version 9.1 (SAS In- similar between the 2 groups (TABLE 2). rithm of triglyceride levels and natu- stitute Inc, Cary, North Carolina). During a median 11.4 years of fol- ral logarithm of follow-up time. low up, 1001 participants experienced All secondary analyses were per- RESULTS a first cardiovascular event (including formed using the final, fully adjusted The median (interquartile range) tri- 276 nonfatal myocardial infarctions, multivariable model. The possibility that glyceride levels among fasting and non- 265 ischemic strokes, 628 coronary cardiovascular risk among participants fasting participants were 115 (81- revascularizations, and 163 cardiovas- choosing not to fast differed from those 169) mg/dL and 133 (93-196) mg/dL, cular deaths), for an overall event rate who fasted was tested by including a fast- respectively. Baseline characteristics of of 3.46 per 1000 person-years of ing indicator variable in the model. To participants at study entry are de- follow-up. Hazard ratios for cardiovas-

Table 1. Baseline Characteristics of Participants by Tertiles of Triglycerides, According to Fasting Statusa Fasting Nonfasting

Triglyceride Tertile, mg/dL Triglyceride Tertile, mg/dL

All Յ90 91-147 Ն148 All Յ104 105-170 Ն171 Characteristic (n = 20 118) (n = 6590) (n = 6802) (n = 6726) (n = 6391) (n = 2084) (n = 2174) (n = 2133) Age, mean (SD), y 54.5 (7.2) 53.1 (6.9) 54.8 (7.3) 55.6 (7.2) 53.3 (6.6) 52.2 (6.2) 53.5 (6.7) 54.2 (6.8) Hypertension, No. (%) 5235 (26.0) 1070 (16.2) 1718 (25.3) 2447 (36.4) 1437 (22.5) 296 (14.2) 435 (20.0) 706 (33.1) Current smoking, No. (%) 2362 (11.8) 663 (10.1) 815 (12.0) 884 (13.2) 713 (11.2) 179 (8.6) 238 (11.0) 296 (13.9) Diabetes mellitus, No. (%) 518 (2.6) 65 (1.0) 115 (1.7) 338 (5.0) 198 (3.1) 35 (1.7) 38 (1.8) 125 (5.9) Postmenopausal, No. (%) 11179 (55.6) 2996 (45.5) 3907 (57.5) 4276 (63.7) 3241 (50.9) 879 (42.3) 1166 (53.8) 1196 (56.2) Hormone use, No. (%) 8747 (43.6) 2235 (34.0) 3023 (44.5) 3489 (52.0) 2817 (44.2) 763 (36.7) 1041 (47.9) 1013 (47.7) Ն1 Alcoholic drink/d, No. (%) 2116 (10.5) 817 (12.4) 740 (10.9) 559 (8.3) 644 (10.1) 261 (12.5) 207 (9.5) 176 (8.3) Exercise Ͻ once/wk, No. (%) 11456 (57.0) 3387 (51.4) 3918 (57.6) 4151 (61.7) 3615 (56.6) 1069 (51.3) 1221 (56.2) 1325 (62.1) Total cholesterol, mean (SD), mg/dL 213 (42) 193 (35) 212 (37) 232 (43) 210 (42) 191 (34) 208 (37) 229 (45) LDL-C, mean (SD), mg/dL 126 (34) 113 (29) 128 (32) 136 (37) 120 (33) 108 (28) 122 (31) 130 (36) HDL-C, mean (SD), mg/dL 54 (15) 60 (15) 54 (14) 47 (13) 54 (15) 60 (15) 54 (15) 46 (13) Body mass index, mean (SD)b 25.9 (5.0) 24.2 (4.1) 25.9 (5.0) 27.6 (5.1) 25.9 (5.0) 24.0 (3.9) 25.7 (4.8) 27.9 (5.3) Glycated hemoglobin, mean (SD), % 5.10 (0.60) 5.00 (0.41) 5.06 (0.48) 5.22 (0.81) 5.10 (0.63) 5.00 (0.45) 5.05 (0.47) 5.26 (0.85) High-sensitivity CRP, median (IQR), mg/L 2.03 1.01 2.02 3.51 1.96 0.99 2.03 3.40 (0.82-4.39) (0.45-2.32) (0.89-4.21) (1.78-6.24) (0.78-4.33) (0.43-2.24) (0.84-4.02) (1.61-6.48) Abbreviations: CRP, C-reactive protein; HDL-C, high-density lipoprotein cholesterol; IQR, interquartile range; LDL-C, low-density lipoprotein cholesterol. SI conversion factors: To convert triglycerides to mmol/L, multiply by 0.0113; total cholesterol, LDL-C, and HDL-C to mmol/L, multiply by 0.0259; and high-sensitivity CRP to nmol/L, multiply by 9.524. aP Ͻ .001 for trend for all comparisons across tertiles for fasting and for nonfasting participants. bCalculated as weight in kilograms divided by height in meters squared.

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

cular events among fasting and non- C-reactive protein) (model 3) substan- crease in log(triglyceride level) (eg, cor- fasting participants are presented in tially reduced the association between responding to an increase in triglycer- TABLE 3 by tertiles of triglyceride lev- fasting triglyceride levels and cardio- ide level from 55 mg/dL to 150 mg/ els. In univariable analyses of event vascular events (fully adjusted HRs dL) was associated with an HR (95% CI) rates and after adjusting for age, blood [95% CIs] for increasing tertiles of of 1.11 (0.92-1.34) for fasting partici- pressure, smoking, and use of hor- fasting triglyceride levels: 1 [refer- pants (P=.30) and of 1.67 (1.18-2.35) mone therapy (model 1), both fasting ence], 1.21 [0.96-1.52], and 1.09 for nonfasting participants (P=.004). and nonfasting triglyceride levels were [0.85-1.41] [P=.90 for trend]). In con- The inclusion of other potential con- strongly associated with cardiovascu- trast, nonfasting levels maintained a founders, including alcohol consump- lar events. Among fasting participants, strong association with cardiovascular tion, exercise, and level of glycated he- adjusting for total cholesterol and events in fully adjusted models, with moglobin, had no substantive impact HDL-C (model 2) and for indicators of HRs (95% CIs) for increasing tertiles on these results. The use of LDL-C lev- insulin resistance (diabetes mellitus, of nonfasting levels of 1 (reference), els instead of total cholesterol levels in body mass index, and high-sensitivity 1.44 (0.90-2.29), and 1.98 (1.21-3.25) fully adjusted models likewise pro- (P=.006 for trend). The proportional- vided similar results. Because use of ity assumption was valid for these hormone therapy has been shown to in- Table 2. Spearman Rank Correlations of models as there was no significant crease triglyceride levels by approxi- Triglyceride Levels With Other Covariates, by 16 Fasting Statusa interaction between triglyceride levels mately 20%, analyses were also run in Covariate Fasting Nonfasting and time. the subgroup of hormone nonusers Non–HDL-C 0.56 0.57 To address the possibility that re- with similar results. Fasting status of High-sensitivity CRP 0.43 0.41 sults were driven by the choice of cut participants was not associated with risk Total cholesterol 0.41 0.41 HDL-C −0.40 −0.43 points used in the 2 groups, analyses of cardiovascular events in univariate Body mass index 0.33 0.37 were repeated categorizing the popu- analysis or in fully adjusted models. LDL-C 0.30 0.28 lation into quintiles rather than ter- In analyses stratified by postpran- Blood pressure category 0.24 0.25 Glycated hemoglobin 0.18 0.22 tiles without substantive differences dial time, women who had eaten 2 to Age 0.18 0.15 (TABLE 4). We additionally examined 4 hours prior to phlebotomy had the Alcohol use −0.12 −0.13 Exercise level −0.09 −0.11 triglyceride levels as a natural loga- strongest association between triglyc- Abbreviations: CRP, C-reactive protein; HDL-C, high- rithm–transformed continuous vari- eride levels and cardiovascular events density lipoprotein cholesterol; LDL-C, low-density li- able with similar results (Table 3); in (P Ͻ .001 for trend across tertiles) poprotein cholesterol. a P Ͻ .001 for all coefficients shown. fully-adjusted models, a 1-unit in- (FIGURE 1). This relationship became

Table 3. Association of Triglyceride Levels With Incident Cardiovascular Disease According to Fasting Status Triglyceride Tertile P Value Per Unit Increase P 12 3for Trend in Log(Triglyceride Level)a Value Fasting (n = 20 118) Triglyceride level, mg/dL Յ90 91-147 Ն148 No. of participants 6590 6802 6726 No. of events 126 262 398 Event rate per 1000 person-y 1.74 3.52 5.48 Model 1b 1 [Reference] 1.63 (1.31-2.02) 2.23 (1.82-2.74) Ͻ.001 1.94 (1.71-2.22) Ͻ.001 Model 2c 1 [Reference] 1.27 (1.02-1.59) 1.32 (1.03-1.68) .09 1.34 (1.12-1.60) .001 Model 3d 1 [Reference] 1.21 (0.96-1.52) 1.09 (0.85-1.41) .90 1.11 (0.92-1.34) .30 Nonfasting (n = 6391) Triglyceride level, mg/dL Յ104 105-170 Ն171 No. of participants 2084 2174 2133 No. of events 31 61 123 Event rate per 1000 person-y 1.35 2.55 5.34 Model 1b 1 [Reference] 1.48 (0.95-2.29) 2.53 (1.69-3.79) Ͻ.001 2.12 (1.66-2.70) Ͻ.001 Model 2c 1 [Reference] 1.31 (0.83-2.05) 1.94 (1.21-3.10) .003 1.91 (1.37-2.67) Ͻ.001 Model 3d 1 [Reference] 1.44 (0.90-2.29) 1.98 (1.21-3.25) .006 1.67 (1.18-2.35) .004 SI conversion factor: To convert triglyceride values to mmol/L, multiply by 0.0113. aFor example, corresponding to an increase in triglyceride level from 55 mg/dL to 150 mg/dL. bAdjusted for age, blood pressure, smoking, and use of hormone therapy. cAdjusted for covariates in model 1 plus total and high-density lipoprotein cholesterol. dAdjusted for covariates in model 2 plus diabetes mellitus, body mass index, and high-sensitivity C-reactive protein.

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

more attenuated as more time elapsed high levels of triglycerides in both Following food consumption, triglyc- after the participants’ last meal. Tri- groups increased risk in women with erides are transported from the small glyceride levels measured within 2 HDL-C levels below the clinical cutoff intestines via chylomicrons through the hours of a meal showed no associa- of 50 mg/dL, the magnitude of this effect bloodstream. Lipolysis of the triglyc- tion with cardiovascular risk (HR, 0.59 was greater in the nonfasting group. No erides within chylomicrons, catalyzed [95% CI, 0.15-2.28], P = .70 for trend). significant multiplicative interactions by lipoprotein lipase in tissues, trans- Similar results were obtained when lev- were found for triglycerides with forms these particles into atherogenic, els were reanalyzed as a natural loga- HDL-C among either fasting or non- triglyceride-rich remnant lipopro- rithm–transformed continuous vari- fasting participants. teins. An elevated postprandial triglyc- able. eride level, reflecting either a higher Fully adjusted HRs for extreme ter- COMMENT peak level or a delay in clearance of tri- tiles of triglyceride levels are pre- In this large-scale, prospective cohort glyceride-rich particles, can lead to an sented in FIGURE 2 for the individual of initially healthy US women, we ob- accumulation of these atherogenic par- end points of nonfatal myocardial in- served that higher nonfasting triglyc- ticles.20 The data presented in Figure 1 farction, nonfatal ischemic stroke, coro- eride levels were strongly associated correspond to the expected time course nary revascularization, and cardiovas- with an increased risk of future cardio- of postprandial triglyceride metabo- cular death. Results from the composite vascular events, independent of base- lism; in response to a meal, triglycer- cardiovascular end point were consis- line cardiac risk factors, levels of other ides and remnant lipoprotein concen- tent across the individual end points lipids, and markers of insulin resis- trations both typically increase to their studied among both fasting and non- tance. In contrast, fasting triglyceride peaks by approximately 4 hours and de- fasting women. levels showed little independent asso- cline thereafter.21 Taken together, our Because prior work has suggested that ciation with cardiovascular events. As- results support the broad hypothesis adjustment for HDL-C eliminates or sociations were particularly strong that atherosclerosis is, at least in part, substantially modifies the relationship among individuals who had their blood a “postprandial phenomenon.”20,22,23 of triglycerides with cardiovascular dis- drawn 2 to 4 hours after a meal, and this Along with the possibly direct ath- ease,17-19 we repeated our analyses strati- relationship weakened as more time erogenic effects of postprandial lipids, fied by HDL-C level (FIGURE 3). In par- elapsed postprandially. several established cardiac risk factors ticipants with normal HDL-C levels Several biological mechanisms pro- are associated with triglyceride levels, (Ն50 mg/dL), only elevated nonfast- vide a plausible explanation for the as- as seen in Tables 1 and 2. In particu- ing levels of triglycerides were inde- sociation between postprandial triglyc- lar, high levels are one manifestation pendently associated with events. While eride levels and cardiovascular disease. of the constellation of metabolic dis-

Table 4. Association of Triglyceride Levels With Incident Cardiovascular Disease According to Fasting Status Triglyceride Quintile P for 12345Trend Fasting (n = 20 118) Triglyceride level, mg/dL Յ73 74-98 99-132 133-184 Ն185 No. of participants 3915 3981 4133 4052 4037 No. of events 61 87 115 155 241 Event rate per 1000 person-y 1.41 1.98 2.55 3.52 5.55 Model 1a 1 [Reference] 1.49 (1.08-2.05) 2.01 (1.49-2.71) 1.93 (1.43-2.60) 3.02 (2.27-4.02) Ͻ.001 Model 2b 1 [Reference] 1.26 (0.91-1.73) 1.50 (1.10-2.04) 1.26 (0.91-1.73) 1.61 (1.16-2.25) .02 Model 3c 1 [Reference] 1.18 (0.85-1.64) 1.41 (1.03-1.93) 1.08 (0.78-1.51) 1.27 (0.90-1.78) .60 Nonfasting (n = 6391) Triglyceride level, mg/dL Յ85 86-113 114-154 155-214 Ն215 No. of participants 1273 1233 1320 1273 1292 No. of events 18 20 43 47 87 Event rate per 1000 person-y 1.28 1.47 2.97 3.38 6.27 Model 1a 1 [Reference] 0.90 (0.47-1.72) 1.78 (1.02-3.10) 1.72 (0.99-2.98) 2.81 (1.68-4.73) Ͻ.001 Model 2b 1 [Reference] 0.83 (0.43-1.61) 1.57 (0.89-2.78) 1.41 (0.79-2.55) 2.09 (1.13-3.86) .003 Model 3c 1 [Reference] 0.89 (0.45-1.75) 1.63 (0.90-2.96) 1.60 (0.87-2.95) 1.99 (1.05-3.78) .02 SI conversion factor: To convert triglycerides to mmol/L, multiply by 0.0113. a Adjusted for age, blood pressure, smoking, and use of hormone therapy. b Adjusted for covariates in model 1 plus total and high-density lipoprotein cholesterol. c Adjusted for covariates in model 2 plus diabetes mellitus, body mass index, and high-sensitivity C-reactive protein.

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

turbances associated with insulin While the relationship between non- part of the observed effect for nonfast- resistance and the metabolic syn- fasting triglyceride levels and cardio- ing levels of triglycerides may be due drome. Elevated postprandial levels of vascular events in our data persisted af- to residual confounding from unmea- triglycerides may represent an abnor- ter adjustment for many markers of sured components of the metabolic syn- mal response to an oral fat load due to insulin resistance including diabetes drome. Conversely, our final model may insulin resistance, a hypothesis sup- mellitus, HDL-C level, blood pres- provide a conservative estimate of the ported by previous case-control stud- sure, body mass index, and high- effect of triglycerides by adjusting for ies in individuals with diabetes melli- sensitivity C-reactive protein level, we closely related metabolic parameters. tus or the metabolic syndrome.24,25 cannot fully exclude the possibility that We believe that the current data demonstrating strong differences between Figure 1. Association of Triglyceride Levels With Future Cardiovascular Events, Stratified by nonfasting and fasting triglyceride lev- Time Since Last Meal els in terms of vascular risk prediction may help to explain inconsistencies in Time From No. of No. of Hazard Ratio Last Meal, h Participants Events (95% Confidence Interval) previous triglyceride studies. The 2 to <4 2797 98 4.48 (1.98-10.15) Friedewald equation for the calcula- 4 to <8 2594 92 1.50 (0.72-3.13) tion of LDL-C level using fasting lev- 8 to <12 4846 177 1.31 (0.73-2.36) els of lipids,26 coupled with the notion ≥ 12 15 272 609 1.04 (0.79-1.38) of greater reliability for triglyceride levels measured in the fasting state, have 0.5 1.0 10 Fully Adjusted HR (95% CI) led to national guidelines recommend- ing fasting lipid profiles to assess car- Hazard ratio (HR) and 95% confidence interval (CI) for highest vs lowest tertiles of triglyceride level (see Table 3 diovascular risk.1 However, by empha- for values), adjusted for age, blood pressure, smoking, hormone use, levels of total and high-density lipoprotein cholesterol, diabetes mellitus, body mass index, and high-sensitivity C-reactive protein level. sizing fasting measures, the overall association between plasma triglycer- Figure 2. Association of Triglyceride Levels With Individual Cardiovascular End Points, ides and vascular risk may be system- According to Fasting Status atically underestimated in the pub- lished literature. Our data supporting Fasting Nonfasting the importance of nonfasting levels of Myocardial infarction triglycerides are consistent with prior Ischemic stroke cross-sectional analyses correlating Revascularization Cardiovascular death postprandial levels to extent of carotid atherosclerosis21,27 and with case- All cardiovascular events control studies demonstrating higher postprandial levels among individuals 0.5 1.0 10 0.5 1.0 10 Fully Adjusted HR (95% CI) Fully Adjusted HR (95% CI) with coronary heart disease compared with healthy controls.28-32 Moreover, Hazard ratio (HR) and 95% confidence interval (CI) for highest vs lowest tertiles of triglyceride level (see Table 3 some prospective studies also have pre- for values), adjusted for age, blood pressure, smoking, hormone use, levels of total and high-density lipopro- viously demonstrated an independent tein cholesterol, diabetes mellitus, body mass index, and high-sensitivity C-reactive protein level. association between nonfasting triglyc- Figure 3. Associations of High vs Low Triglyceride Levels With Future Cardiovascular Events, eride levels and cardiovascular dis- Stratified by HDL-C Level ease.33-36 However, the use of “casual” blood draws—ie, those for which fast- Fasting Nonfasting ing was not required—in these stud- HDL-C ≥ 50 mg/dL Triglycerides < 150 mg/dL ies may have underestimated the asso- Triglycerides ≥ 150 mg/dL ciation between nonfasting triglyceride levels and cardiovascular risk by HDL-C < 50 mg/dL combining participants in different Triglycerides < 150 mg/dL Triglycerides ≥ 150 mg/dL postprandial states. As suggested by Figure 1, even stronger effects may have 0.51.0 10 0.51.0 10 been observed had these studies sys- Fully Adjusted HR (95% CI) Fully Adjusted HR (95% CI) tematically measured levels of triglycerides 2 to 4 hours after food consump- Hazard ratio (HR) and 95% confidence interval (CI) adjusted for age, blood pressure, smoking, hormone use, total cholesterol level, diabetes mellitus, body mass index, and high-sensitivity C-reactive protein level, using tion. triglyceride levels Ͻ150 mg/dL (to convert to mmol/L, multiply by 0.0113) and high-density lipoprotein cho- The results of this study suggest that lesterol (HDL-C) levels Ն50 mg/dL (to convert to mmol/L, multiply by 0.0259) as the reference group. postprandial triglyceride levels may be

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

superior to fasting levels for assess- risk. In addition, despite the known Drafting of the manuscript: Bansal, Ridker. Critical revision of the manuscript for important in- ment of cardiovascular risk, a hypoth- strong correlation between fasting and tellectual content: Bansal, Buring, Rifai, Mora, Sacks, esis consistent with the concept of ath- postprandial levels of triglycer- Ridker. 29,32,38 Statistical analysis: Bansal, Mora. erosclerosis as a postprandial disorder. ides, our data suggest that post- Obtained funding: Buring, Ridker. Strengths of the study include its large prandial levels are a more robust indi- Administrative, technical, or material support: Rifai, sample size, prospective design, ex- cator of cardiovascular risk, perhaps Ridker. Study supervision: Ridker. tended follow-up time with validated because the greater variability of post- Financial Disclosures: Dr Buring reported receiving in- outcomes, and measurement of both prandial levels captures important investigator-initiated research funding and support as principal investigator from the National Institutes of fasting and nonfasting levels of triglyc- formation about an individual’s me- Health (the National Heart, Lung, and Blood Insti- erides within the same cohort. tabolism. At a practical level, the use tute, the National Cancer Institute, and the National Institute of Aging) and Dow Corning Corporation; re- Nonetheless, limitations of this study of nonfasting triglyceride levels and the search support for pills, packaging, or both from Bayer also merit consideration. First, partici- availability of assays to directly mea- Heath Care and the Natural Source Vitamin E Asso- ciation; and honoraria from Bayer for speaking en- pants were not randomly assigned to sure LDL-C levels could allow pa- gagements. Dr Rifai reported receiving grant support fasting or nonfasting status. However, tients to have blood for a lipid profile from Merck Research Laboratories; serving as a con- as shown in Table 1, baseline charac- drawn without the need to return to the sultant to Merck Research Laboratories and Sanofi- Aventis; and receiving honoraria for speaking from Ab- teristics of fasting and nonfasting par- laboratory after an overnight fast. Ad- bott Laboratories, Dade Behring, Denka Seiken, Merck ticipants were similar for many major ditionally, the apparent value seen in Research Laboratories, Ortho Diagnostics, and Roche Diagnostics. Dr Sacks reported that he currently or in risk factors, and despite the slightly this study of triglyceride levels mea- the past 5 years has received fees for consultation or higher age, LDL-C level, and preva- sured 2 to 4 hours after food consump- lecturing on diagnosis, drug development, or treatment of hyperlipidemia from Abbott, Aegerion, Bristol- lence of hypertension among fasting tion suggests that a “triglyceride toler- Myers Squibb, Celera Diagnostics, Fournier, Johnson women, fasting status itself was not as- ance test” using a standardized meal, & Johnson, Kos, Lilly, Lipid Sciences, Merck, Pfizer, sociated with cardiovascular events. analogous to a glucose tolerance test, Quest Diagnostics, Sanofi-Aventis, and Solvay; he also reported receiving expert testimony fees related to hy- Second, despite the consistency of data warrants further evaluation as a poten- perlipidemia treatment from Kos and Pfizer, and being in Figure 1 with respect to postpran- tial indicator of a metabolic state pre- a stockholder in Lipid Sciences. Dr Ridker reported that he currently or in the past 5 years has received re- dial time and peak triglyceride levels, disposing individuals to higher risk for search funding support from multiple not-for-profit the last meal prior to phlebotomy was cardiovascular events. entities including the American Heart Association, the Doris Duke Charitable Foundation, the Leducq Foun- not given in a standardized manner. Our observations may have implica- dation, the National Cancer Institute, the National However, this potential limitation tions for the design and conduct of Heart, Lung, and Blood Institute, the Donald W. Rey- nolds Foundation, and the James and Polly Annen- would likely lead to misclassification clinical trials evaluating triglyceride- berg La Vea Charitable Trusts; that currently or in the bias and therefore result, if anything, lowering medications. To date, almost past 5 years he has received investigator-initiated in an underestimation of the true effect. all clinical trials of pharmaceutical research support from for-profit entities including Ab- bott, AstraZeneca, Bayer, Bristol-Myers Squibb, Dade- Third, given the variability of triglyc- agents targeting triglyceride levels have Behring, Novartis, Pharmacia, Roche, and Sanofi- eride levels, the single measurement of relied on fasting levels as inclusion cri- Aventis; and that he is listed as a coinventor on patents held by the Brigham and Women’s Hospital that levels at study enrollment without re- teria. However, if levels measured in the relate to the use of inflammatory biomarkers in car- peated sampling could lead to regres- fasting state are not the best marker for diovascular disease and has served as a consultant to 2,36 AstraZeneca, Dade-Behring, Isis Pharmaceutical, Sa- sion dilution bias, particularly in the the atherogenicity associated with hy- nofi-Aventis, and Schering-Plough. No other disclo- nonfasting state. Such an effect, how- pertriglyceridemia, then it is possible sures were reported. ever, also would bias the results to- that these trials might have targeted the Funding/Support: This study was supported by grants from the Donald W. Reynolds Foundation (Las Ve- ward a null finding rather than a wrong patient populations. By con- gas, Nevada) and the Leducq Foundation (Paris, France) strongly positive one. And fourth, our trast, previous studies have demon- and by grant T32 HL007575-21 from the National Heart, Lung, and Blood Institute, with additional sup- data are limited to women; thus, fur- strated the benefits of several classes of port from Abbott Laboratories (Abbott Park, Illinois). ther studies are needed before gener- drugs on postprandial elevations in tri- The Women’s Health Study was funded by grant HL- 39-43 43851 from the National Heart, Lung, and Blood In- alizing the conclusions to men. glyceride levels. Thus, based on the stitute and grant CA-47988 from the National Can- The use of nonfasting levels of tri- data presented here, future end point cer Institute. Role of the Sponsor: The funding agencies played no glycerides in risk assessment provides reduction trials of triglyceride- role in the design and conduct of the study; the col- several potential advantages to clini- lowering agents might consider par- lection, management, analysis, and interpretation of cal practice. Much of the 24-hour day ticipant inclusion on the basis of non- the data; or the preparation, review, or approval of the manuscript. is spent in the nonfasting state, espe- fasting rather than fasting triglyceride Additional Contributions: We thank the partici- cially considering the fact that triglyc- levels. pants, staff, and investigators of the Women’s Health Study for their valuable contributions. erides may take up to 12 hours to re- Author Contributions: Dr Bansal and Dr Ridker had turn to fasting levels after a meal.37 If full access to all of the data in the study and take re- sponsibility for the integrity of the data and the ac- REFERENCES postprandial triglycerides are biologi- curacy of the data analysis. cally active in atherogenesis, measure- Study concept and design: Bansal, Mora, Ridker. 1. National Cholesterol Education Program (NCEP) Ex- Acquisition of data: Buring, Rifai, Ridker. pert Panel on Detection, Evaluation, and Treatment ment of fasting levels may provide an Analysis and interpretation of data: Bansal, Mora, of High Blood Cholesterol in Adults (Adult Treatment inadequate representation of vascular Sacks, Ridker. Panel III). Third Report of the National Cholesterol Edu-

Downloaded From: https://jamanetwork.com/ on 10/01/2021 FASTING VS NONFASTING TRIGLYCERIDES AND RISK OF CARDIOVASCULAR EVENTS IN WOMEN

cation Program (NCEP) Expert Panel on Detection, developing type 2 diabetes mellitus. JAMA. lenge in normolipidaemic men with and without coro- Evaluation, and Treatment of High Blood Cholesterol 2001;286(3):327-334. nary artery disease. J Cardiovasc Risk. 1997;4(2):143- in Adults (Adult Treatment Panel III) final report. 16. Hsia J, Langer RD, Manson JE, et al. Conjugated 149. Circulation. 2002;106(25):3143-3421. equine estrogens and coronary heart disease: the 31. Bjo¨ rkegren J, Boquist S, Samnegard A, et al. Ac- 2. Sarwar N, Danesh J, Eiriksdottir G, et al. Triglyc- Women’s Health Initiative. Arch Intern Med. cumulation of apolipoprotein C-I-rich and cholesterol- erides and the risk of coronary heart disease: 10,158 2006;166(3):357-365. rich VLDL remnants during exaggerated postpran- incident cases among 262,525 participants in 29 West- 17. Manninen V, Tenkanen L, Koskinen P, et al. Joint dial triglyceridemia in normolipidemic patients with ern prospective studies. Circulation. 2007;115(4):450- effects of serum triglyceride and LDL cholesterol and coronary artery disease. Circulation. 2000;101(3):227- 458. HDL cholesterol concentrations on coronary heart dis- 230. 3. Daugherty A, Lange LG, Sobel BE, Schonfeld G. Aor- ease risk in the Helsinki Heart Study: implications for 32. Schaefer EJ, Audelin MC, McNamara JR, et al. tic accumulation and plasma clearance of beta-VLDL treatment. Circulation. 1992;85(1):37-45. Comparison of fasting and postprandial plasma lipo- and HDL: effects of diet-induced hypercholesterol- 18. Gaziano JM, Hennekens CH, O’Donnell CJ, Bres- proteins in subjects with and without coronary heart emia in rabbits. J Lipid Res. 1985;26(8):955-963. low JL, Buring JE. Fasting triglycerides, high-density disease. Am J Cardiol. 2001;88(10):1129-1133. 4. Rapp JH, Lespine A, Hamilton RL, et al. Triglyceride- lipoprotein, and risk of myocardial infarction. 33. Stensvold I, Tverdal A, Urdal P, Graff-Iversen S. rich lipoproteins isolated by selected-affinity anti- Circulation. 1997;96(8):2520-2525. Non-fasting serum triglyceride concentration and apolipoprotein B immunosorption from human ath- 19. Avins AL, Neuhaus JM. Do triglycerides provide mortality from coronary heart disease and any erosclerotic plaque. Arterioscler Thromb. 1994; meaningful information about heart disease risk? Arch cause in middle aged Norwegian women. BMJ. 1993; 14(11):1767-1774. Intern Med. 2000;160(13):1937-1944. 307(6915):1318-1322. 5. Proctor SD, Mamo JC. Retention of fluorescent- 20. Ginsberg HN. New perspectives on atherogen- 34. Stampfer MJ, Krauss RM, Ma J, et al. A prospec- labelled chylomicron remnants within the intima of the esis: role of abnormal triglyceride-rich lipoprotein tive study of triglyceride level, low-density lipopro- arterial wall—evidence that plaque cholesterol may be metabolism. Circulation. 2002;106(16):2137-2142. tein particle diameter, and risk of myocardial infarction. derived from post-prandial lipoproteins. Eur J Clin 21. Boquist S, Ruotolo G, Tang R, et al. Alimentary JAMA. 1996;276(11):882-888. Invest. 1998;28(6):497-503. lipemia, postprandial triglyceride-rich lipoproteins, 35. Iso H, Naito Y, Sato S, et al. Serum triglycerides 6. Chen YD, Swami S, Skowronski R, Coulston A, and common carotid intima-media thickness in healthy, and risk of coronary heart disease among Japanese men Reaven GM. Differences in postprandial lipemia be- middle-aged men. Circulation. 1999;100(7):723- and women. Am J Epidemiol. 2001;153(5):490-499. tween patients with normal glucose tolerance and non- 728. 36. Eberly LE, Stamler J, Neaton JD. Relation of tri- insulin-dependent diabetes mellitus. J Clin Endocri- 22. Zilversmit DB. Atherogenesis: a postprandial glyceride levels, fasting and nonfasting, to fatal and nol Metab. 1993;76(1):172-177. phenomenon. Circulation. 1979;60(3):473-485. nonfatal coronary heart disease. Arch Intern Med. 7. Malmstro¨ m R, Packard CJ, Caslake M, et al. De- 23. Karpe F. Postprandial lipoprotein metabolism and 2003;163(9):1077-1083. fective regulation of triglyceride metabolism by insu- atherosclerosis. J Intern Med. 1999;246(4):341-355. 37. Campos H, Khoo C, Sacks FM. Diurnal and acute lin in the liver in NIDDM. Diabetologia. 1997;40(4): 24. Kim HS, Abbasi F, Lamendola C, McLaughlin T, patterns of postprandial apolipoprotein B-48 in VLDL, 454-462. Reaven GM. Effect of insulin resistance on postpran- IDL, and LDL from normolipidemic humans. 8. Ginsberg HN. Insulin resistance and cardiovascu- dial elevations of remnant lipoprotein concentrations Atherosclerosis. 2005;181(2):345-351. lar disease. J Clin Invest. 2000;106(4):453-458. in postmenopausal women. Am J Clin Nutr. 2001; 38. Mohanlal N, Holman RR. A standardized triglyc- 9. Trevisan M, Liu J, Bahsas FB, Menotti A; Risk Fac- 74(5):592-595. eride and carbohydrate challenge: the oral triglycer- tor and Life Expectancy Research Group. Syndrome 25. Kolovou GD, Anagnostopoulou KK, Pavlidis AN, ide tolerance test. Diabetes Care. 2004;27(1):89-94. X and mortality: a population-based study. Am J et al. Postprandial lipemia in men with metabolic syn- 39. Parhofer KG, Barrett PH, Schwandt P. Atorva- Epidemiol. 1998;148(10):958-966. drome, hypertensives and healthy subjects. Lipids statin improves postprandial lipoprotein metabolism 10. Isomaa B, Almgren P, Tuomi T, et al. Cardiovas- Health Dis. 2005;4:21. in normolipidemic subjects. J Clin Endocrinol Metab. cular morbidity and mortality associated with the meta- 26. Friedewald WT, Levy RI, Fredrickson DS. Estima- 2000;85(11):4224-4230. bolic syndrome. Diabetes Care. 2001;24(4):683-689. tion of the concentration of low-density lipoprotein 40. Schaefer EJ, McNamara JR, Tayler T, et al. Ef- 11. Buring JE, Hennekens CH. The Women’s Health cholesterol in plasma, without use of the preparative fects of atorvastatin on fasting and postprandial lipo- Study: summary of the study design. J Myocardial ultracentrifuge. Clin Chem. 1972;18(6):499-502. protein subclasses in coronary heart disease patients Ischemia. 1992;4:27-29. 27. Teno S, Uto Y, Nagashima H, et al. Association versus control subjects. Am J Cardiol. 2002;90(7):689- 12. Ridker PM, Cook NR, Lee IM, et al. A random- of postprandial hypertriglyceridemia and carotid intima- 696. ized trial of low-dose aspirin in the primary preven- media thickness in patients with type 2 diabetes. Dia- 41. Boquist S, Karpe F, Danell-Toverud K, Hamsten tion of cardiovascular disease in women. N Engl J Med. betes Care. 2000;23(9):1401-1406. A. Effects of atorvastatin on postprandial plasma li- 2005;352(13):1293-1304. 28. Patsch JR, Miesenbock G, Hopferwieser T, et al. poproteins in postinfarction patients with combined 13. Lee IM, Cook NR, Gaziano JM, et al. Vitamin E Relation of triglyceride metabolism and coronary ar- hyperlipidaemia. Atherosclerosis. 2002;162(1): in the primary prevention of cardiovascular disease and tery disease: studies in the postprandial state. Arte- 163-170. cancer: the Women’s Health Study: a randomized con- rioscler Thromb. 1992;12(11):1336-1345. 42. Ooi TC, Cousins M, Ooi DS, Nakajima K, Ed- trolled trial. JAMA. 2005;294(1):56-65. 29. Weintraub MS, Grosskopf I, Rassin T, et al. Clear- wards AL. Effect of fibrates on postprandial remnant- 14. Stinshoff K, Weisshaar D, Staehler F, Hesse D, Gru- ance of chylomicron remnants in normolipidaemic like particles in patients with combined hyperlipidemia. ber W, Steier E. Relation between concentrations of patients with coronary artery disease: case control Atherosclerosis. 2004;172(2):375-382. free glycerol and triglycerides in human sera. Clin study over three years. BMJ. 1996;312(7036):936- 43. van Wijk JP, de Koning EJ, Castro Cabezas M, Ra- Chem. 1977;23(6):1029-1032. 939. belink TJ. Rosiglitazone improves postprandial triglyc- 15. Pradhan AD, Manson JE, Rifai N, Buring JE, Rid- 30. Braun D, Gramlich A, Brehme U, Kahle PF, Schmahl eride and free fatty acid metabolism in type 2 diabetes. ker PM. C-reactive protein, interleukin 6, and risk of FW. Post-prandial lipaemia after a moderate fat chal- Diabetes Care. 2005;28(4):844-849.

Downloaded From: https://jamanetwork.com/ on 10/01/2021