Plasma phospholipid trans fatty acids and risk of heart failure

Abstract

Background: Although trans fatty acids (TFAs) may increase the risk of dyslipidemia and coronary artery disease (CAD), limited data are available on their association with heart failure (HF).

Objective: Our goal was to assess associations of plasma and dietary TFAs with HF and CAD.

Design: We used a prospective, nested case-control design to select 788 incident HF cases and 788 matched controls from the Physicians’ Health Study for biomarker analyses and a prospective cohort for the dietary analyses. Plasma fatty acids were assessed by using gas chromatography, and dietary intake was estimated by using a food-frequency questionnaire. Self-reported HF was ascertained by using annual follow-up questionnaires with validation in a subsample. We used conditional logistic (or Cox) regression to estimate multivariable-adjusted ORs (or HRs) for HF and CAD.

Results: Multivariable-adjusted ORs (95% CIs) for HF across consecutive quintiles of plasma trans 18:2 (linoleic acid) fatty acids were 1.0 (reference), 1.10 (0.79, 1.54), 0.88 (0.62, 1.25), 0.71 (0.49, 1.02), and 0.67 (0.45, 0.98) (P-trend = 0.01). Each SD of plasma trans 18:2 was associated with a 22% lower risk of HF (95% CI: 6%, 36%). Plasma trans 16:1 and 18:1 were not associated with risk of HF (P > 0.05). Dietary trans fats were not associated with incident HF or CAD.

Conclusions: Our data are consistent with a lower risk of HF with higher concentrations of plasma trans 18:2 but not with trans 16:1 or trans 18:1 fatty acids in male physicians. Dietary TFAs were not related to incident HF or CAD.

INTRODUCTION

Cardiovascular disease (CVD)⁴ remains a major public health concern in both developed and developing countries (1–3). Data from the Framingham Heart Study suggest that 1 in 5 adults will develop heart failure (HF) over the course of their lifetime (4). Approximately half of those diagnosed with HF will die within 5 y (5). In the United States, health care costs due to HF are projected to reach $97.0 billion by 2030 (6). In the United Kingdom, the “direct” cost of health care for HF patients was estimated to be £716 million (1.83% of total National Health Service expenditure), additional costs associated with long-term nursing home care and secondary HF admissions accounted for another £751 million (2.0% of total National Health Service expenditure) (7).

Previous studies have shown the importance of dietary factors on CVD and its risk factors. For example, trans fatty acid (TFA) intake decreases HDL cholesterol and increases LDL cholesterol (8) and could lead to a higher risk of coronary artery disease (CAD) (9), a major risk factor for HF. Consumption of TFAs has also been shown to increase the risk of other major predictors of HF including type 2 diabetes (10), obesity (11), dyslipidemia (12), and hypertension (13). However, not all studies have reported deleterious effects with TFAs. The consumption of TFAs was not associated with risk of diabetes among male health professionals (14) and among women in Iowa (15). Another study also reported no association between TFAs and dyslipidemia (16). The role of TFAs in the development of HF has similarly not been fully elucidated.

Hence, the primary purpose of this study was to examine the relation between various plasma phospholipid TFAs (biomarkers for habitual TFA consumption) and the risk of HF in US male physicians. In a secondary aim, we evaluated the association of dietary TFA consumption [estimated by a food-frequency questionnaire (FFQ)] with incident HF and CAD.

SUBJECTS AND METHODS

Subjects

Participants in the primary analyses are members of the Physicians’ Health Study (PHS) I, a completed, randomized, double-blind, placebo-controlled trial designed to study low-dose aspirin and β-carotene for the primary prevention of CVD and cancer. A detailed description of the PHS I has been published (17). Briefly, this was a 2 × 2 factorial trial of low-dose aspirin and β-carotene. A total of 22,071 US male physicians (mostly white), aged 40–84 y, were enrolled in 1982. Each participant provided written informed consent, and the institutional review board at Brigham and Women's Hospital approved the study protocol.

In our secondary analyses, we analyzed data on participants from PHS I and II who provided an FFQ collected between 1997 and 2001. Of note, the PHS II was a randomized controlled trial in 14,641 subjects (7000 newly recruited physicians and 7641 members of the original PHS I). A detailed description of the PHS II has been previously published (18).

Case and control selection for the primary aim

All incident HF cases among participants who provided baseline blood samples were eligible to serve as cases in this study. We used a risk set technique to randomly select one control subject, free of HF at the time of the index case's diagnosis of HF. Each control was matched to the respective case on race (white compared with nonwhite), age (±2 y), year of birth (±2 y), and time of blood collection (±288 d) as described previously (19). A total of 788 matched pairs were selected for the current analyses.

Blood collection

Between August 1982 and December 1984, kits for blood sampling were sent to participants with instructions to have their blood drawn and returned to Brigham and Women's Hospital using a prepaid overnight courier. The kit included a cold pack to keep the specimens cool (but not frozen) until receipt the following morning, when they were processed, separated into aliquots, and stored at −80°C. A total of 15,700 baseline blood samples were collected. There was virtually no loss to follow-up of the cohort; 100% were followed up for vital status, and >99% were followed up for nonfatal outcomes.

Ascertainment of HF and CHD in the PHS

Ascertainment of endpoints including HF in the PHS was achieved by using yearly follow-up questionnaires. A detailed description of HF validation in the PHS using self-reported information and the Framingham criteria (20) as well as against review of medical records has been published elsewhere (21). Incident cases of coronary artery bypass grafting or percutaneous transluminal coronary angioplasty were validated by the PHS Endpoint Committee, Brigham and Women's Hospital. Diagnoses of nonfatal myocardial infarction were confirmed by using WHO criteria (22). Fatal myocardial infarction was confirmed by death certificates, hospital records, and (for death outside the hospital) observers’ accounts.

Measurement of plasma phospholipid TFAs

Fatty acid profile was measured in plasma by using the method previously described by Cao et al (23). Lipids were extracted from plasma with a mixture of chloroform and methanol. Cholesteryl ester, triglycerides, and phospholipid subclasses were separated on a silica thin-layer chromatography plate in a solvent mixture of petroleum ether, diethyl ether, and glacial acetic acid. This phospholipid band was harvested and used for the formation of methyl esters by using 14% boron trifluoride in methanol and then extracted with petroleum ether. The final product was dissolved in heptane and injected onto a 100-m Varian capillary column (FAME CP7420, high-polarity 100% bonded phase; Agilent Technologies) in a Hewlett-Packard gas chromatograph with a flame ionization detector and interfaced with HP Chemstation software (Hewlett-Packard). Adequate separation of fatty acid methyl esters was obtained over an 80-min period with an initial temperature of 190°C for 25 min. The temperature was increased to 240°C at a rate of 2°C/min and held for 5 min. Fatty acid methyl esters from 14:0 through 24:1n–9 were separated, identified, and expressed as percentage of total fatty acids. The following CVs were obtained on 20 blind duplicates: 16:1n–7 trans (t) = 20.0%; 18:1n–7-9t = 2.4%; 18:1n–6t = 2.8%; 18:2cis (c)/t, 18:2t/c, and 18:2t/t = 8.5%.

Assessment of other variables

At baseline, each study participant provided information on smoking (never, former, and current smoker), exercise (How often do you exercise vigorously enough to work up sweat? Possible answers included rarely/never, 1–3 times/mo, 1 time/wk, 2–4 times/wk, 5–6 times/wk, and daily), and alcohol intake (rarely/never, 1–3 drinks/mo, 1 drink/wk, 2–4 drinks/wk, 5–6 drinks/wk, daily, and ≥2 drinks/d). Self-reported baseline weight and height were used to compute BMI (calculated as weight in kilograms divided by height in meters squared). Comorbidity in this cohort has been ascertained through annual questionnaires.

Assessment of dietary TFA intake

Information on dietary fatty acid consumption was obtained by using an FFQ administered between 1997 and 2001. Nutrients were computed by using the food composition database from the Harvard School of Public Health and manufacturers’ information. The validity and reproducibility of FFQs have been published elsewhere (24). The residual method was used to adjust nutrients for energy intake (25).

Statistical analysis

We initially examined the distribution of plasma or dietary TFAs for normality. For the primary analyses, we created quintiles of each exposure (16:1, 18:1, and 18:2 TFAs) by using the distribution of TFAs among control series. We used conditional logistic regression to compute multivariable-adjusted ORs as estimates of RR with corresponding 95% CIs. We also examined each exposure as a continuous variable and estimated the RR associated with each SD of higher exposure. On the basis of prevailing knowledge, we assessed confounding by BMI (continuous), smoking (never, past, current), alcohol intake (rarely, monthly, weekly, daily), exercise to sweat weekly or more (yes or no), previous type 2 diabetes (yes or no), and atrial fibrillation (yes or no). Because TFAs have been shown to increase LDL cholesterol and lower HDL cholesterol, we considered hypertension and CHD as potential intermediate factors between TFAs and HF. Possible interactions between each TFA and confounding variables in each model were assessed by including product terms in the model.

The initial model accounted for matching factors only (age, race, date of blood collection, and year of birth). A parsimonious model also adjusted for BMI, smoking, and alcohol consumption; and a final model contained further adjustment for exercise, diabetes, and atrial fibrillation.

A P value for linear trend was obtained by creating a new variable that was assigned the median TFA value from the control series in each quintile, and then by using that new variable in the conditional logistic regression model.

For the secondary analyses, we used a Cox proportional hazards model to assess the relation of dietary TFA with incident CAD and HF. We created quintiles of TFAs by using the lowest quintile as the reference category. Proportional hazards assumption was tested by including an interaction with logarithmic person-time in the model. The models were adjusted for age, smoking, alcohol, and exercise. Potential mediation was examined by including BMI, diabetes, atrial fibrillation, and hypertension in the model. To obtain P values for linear trend, we created a new variable that was assigned the median value of the TFA for each quintile and fitted the variable in each regression model.

All analyses were completed by using SAS, version 9.2 (SAS Institute). All P values were 2-tailed, and significance was set at an α of 0.05.

RESULTS

Relation of plasma phospholipid TFAs and HF risk

Characteristics of HF cases and matched controls are shown in Table 1. The distribution of participants and covariates across quintiles of individual TFAs are shown in Table 2.

TABLE 1.

Characteristics of 788 heart failure cases and 788 matched controls in US male physicians¹

Characteristics	Cases (n = 788)	Controls (n = 788)	P value
Total trans fat (% of total fatty acids)	2.0 ± 0.62	2.1 ± 0.6	0.03
BMI (kg/m2)	25.8 ± 3.2	24.6 ± 2.5	<0.01
Age (y)	58.7 ± 8.0	58.7 ± 8.1	0.48
History of atrial fibrillation (%)	5.5	2.4	<0.01
History of CABG (%)	2.4	0.9	0.02
History of CAD (%)	4.3	2.3	0.02
History of type 2 diabetes (%)	7.9	3.3	<0.01
Hypertension (%)	37.3	24.6	<0.01
High cholesterol (%)	14.3	12.4	0.21
Current exercise (%)	71.1	75.3	0.048
Smoking status (%)			<0.01
Never	41.0	49.2
Past	45.4	41.6
Current	13.6	8.9
Alcohol (%)			0.053
Rarely	17.0	14.9
Monthly	11.4	8.1
Weekly	44.7	46.8
Daily	26.4	30.0

¹Except for BMI and age, data for other characteristics were not provided by all individuals. CABG, coronary artery bypass grafting; CAD, coronary artery disease.

²Mean ± SD (all such values).

TABLE 2.

Baseline characteristics of 1576 US male physicians across quintiles of plasma phospholipid TFAs¹

	Plasma trans 16:1				Plasma trans 18:1				Plasma trans 18:2
Characteristics	Q1 (low)	Q3	Q5 (high)	P-trend	Q1 (low)	Q3	Q5 (high)	P-trend	Q1 (low)	Q3	Q5 (high)	P-trend
n	295	326	292		332	312	298		323	314	282
Plasma TFAs (% of total fatty acids)2	0.05 (0.02, 0.06)	0.08 (0.08, 0.09)	0.12 (0.11, 0.26)		0.98 (0.60, 1.16)	1.57 (1.44, 1.70)	2.37 (2.08, 4.10)		0.24 (0.15, 0.28)	0.36 (0.33, 0.39)	0.51 (0.45, 2.71)
BMI (kg/m2)	24.9 ± 2.53	25.1 ± 2.9	25.0 ± 2.6	0.98	25.5 ± 3.1	25.2 ± 2.9	24.8 ± 2.5	<0.01	25.4 ± 2.9	25.0 ± 2.8	25.0 ± 2.8	0.06
Age (y)	58.5 ± 7.6	59.5 ± 8.3	58.0 ± 7.7	0.46	58.5 ± 7.7	58.8 ± 8.1	59.1 ± 8.1	0.56	58.9 ± 7.8	58.2 ± 8.0	58.3 ± 7.9	0.05
Prevalent atrial fibrillation (%)	3.7	3.4	3.8	0.78	5.1	2.9	3.7	0.58	2.5	3.8	2.1	0.73
History of CABG (%)	1.7	1.5	1.4	0.78	1.2	0.3	1.0	0.70	2.8	2.9	0.7	0.06
History of CAD (%)	3.1	3.1	3.1	1.00	3.3	0.6	4.0	0.73	5.0	4.5	2.5	0.07
Prevalent type 2 diabetes (%)	4.4	6.4	5.5	0.34	8.1	4.2	4.0	0.05	6.2	6.1	5.0	0.49
Hypertension (%)	35.9	31.3	26.7	<0.01	36.8	28.2	25.5	<0.01	36.5	26.8	32.3	0.13
High cholesterol (%)	16.6	10.4	13.7	0.02	16.9	12.8	9.7	<0.01	18.9	13.4	11.7	0.01
Exercise (%)	70.5	74.2	73.3	0.35	64.5	76.3	78.2	<0.01	67.5	75.5	75.5	0.01
Smoking status (%)
Never	36.6	46.6	52.1	<0.01	34.3	46.2	56.4	<0.01	41.2	46.5	48.9	0.02
Past	49.2	43.3	38.4	<0.01	50.3	41.7	34.6	<0.01	50.2	42.4	39.7	0.01
Current	13.9	10.1	9.3	0.08	15.1	12.2	9.1	0.03	8.4	11.2	11.0	0.55
Alcohol (%)
Rarely	6.8	15.3	26.0	<0.01	6.0	15.7	27.5	<0.01	9.9	17.5	22.0	<0.01
Monthly	6.8	9.5	12.3	0.01	7.5	8.3	11.7	0.03	10.2	8.9	9.9	0.95
Weekly	40.3	47.9	43.5	0.26	41.3	48.1	46.0	0.27	48.3	43.3	43.3	0.21
Daily	46.1	26.7	17.8	<0.01	44.9	26.9	14.1	<0.01	31.0	29.6	24.5	0.07

¹Except for BMI and age, data for other characteristics were not provided by all individuals; quintiles were created by using TFA distribution in the control series. CABG, coronary artery bypass grafting; CAD, coronary artery disease; Q, quintile; TFA, trans fatty acid.

²Values are medians; ranges in parentheses.

³Mean ± SD (all such values).

Higher plasma concentrations of trans 16:1 fatty acids were associated with a lower prevalence of hypertension. Plasma concentrations of trans 18:1 fatty acids were inversely associated with hypertension, high cholesterol, and current smoking, whereas higher concentrations of trans 18:2 fatty acids were associated with a lower prevalence of hypercholesterolemia (Table 2).

Multivariable-adjusted ORs (95% CIs) for HF were 1.0 (reference), 1.10 (0.79, 1.54), 0.88 (0.62, 1.25), 0.71 (0.49, 1.02), and 0.67 (0.45, 0.98) across consecutive quintiles of trans 18:2 fatty acids (P-trend = 0.01) (Table 3). Each SD of higher plasma trans 18.2 fatty acids was associated with a 22% lower risk of HF (95% CI: 6%, 36%). Additional adjustment for CAD and hypertension had minimal effect on the OR (OR per SD: 0.79; 95% CI: 0.65, 0.96). Analyses for plasma trans 16:1 and trans 18:1 isomers did not show any meaningful association (P-trend 0.23 and 0.37, respectively; Table 3).

TABLE 3.

ORs (95% CIs) of heart failure according to quintiles of individual plasma phospholipid TFAs in 1576 male physicians from the PHS I¹

trans 16:1			trans 18:1			trans 18:2
Quintile range2	Model 1	Model 2	Quintile range2	Model 1	Model 2	Quintile range2	Model 1	Model 2
Q1 (0.018–0.062)	1 (ref)	1	Q1 (0.578–1.158)	1	1	Q1 (0.159–0.274)	1	1
Q2 (0.063–0.076)	1.03 (0.75, 1.41)	0.88 (0.63, 1.24)	Q2 (1.159–1.433)	0.93 (0.68, 1.26)	0.88 (0.63, 1.23)	Q2 (0.275–0.333)	1.15 (0.84, 1.58)	1.10 (0.79, 1.54)
Q3 (0.077–0.089)	1.14 (0.83, 1.58)	1.04 (0.73, 1.47)	Q3 (1.434–1.701)	0.88 (0.64, 1.19)	0.87 (0.62, 1.22)	Q3 (0.334–0.386)	0.87 (0.63, 1.20)	0.88 (0.62, 1.25)
Q4 (0.090–0.108)	1.08 (0.79, 1.49)	0.96 (0.67, 1.37)	Q4 (1.702–2.071)	0.87 (0.64, 1.20)	0.87 (0.61, 1.24)	Q4 (0.387–0.452)	0.76 (0.54, 1.07)	0.71 (0.49, 1.02)
Q5 (0.109–0.219)	0.81 (0.57, 1.16)	0.74 (0.49, 1.10)	Q5 (2.072–3.817)	0.81 (0.59, 1.10)	0.83 (0.58, 1.19)	Q5 (0.453–2.705)	0.65 (0.46, 0.94)	0.67 (0.45, 0.98)
P-trend	0.34	0.23		0.17	0.37		<0.01	0.01
TFA modeled as continuous variable3	0.95 (0.84, 1.07)	0.91 (0.80, 1.04)		0.91 (0.82, 1.01)	0.92 (0.82, 1.03)		0.76 (0.63, 0.91)	0.78 (0.64, 0.94)

¹Model 1 adjusted for matching factors; model 2 adjusted as for model 1 with additional adjustment for BMI, smoking, alcohol, exercise, diabetes, and atrial fibrillation analyzed by using conditional logistic regression. PHS I, Physicians’ Health Study I; Q, quintile; ref, reference; TFA, trans fatty acid.

²Values are % of total fatty acids.

³ORs per SD increase in TFAs (16:1 = 0.03% of total fatty acids, 18:1 = 0.55% of total fatty acids, 18:2 = 0.18% of total fatty acids).

Relation of dietary TFAs (from the FFQ) and incident HF and CAD

The average age at the time of dietary assessment was 66 y, and mean follow-up was ∼10 y. Baseline characteristics by quintiles of total dietary TFAs are provided in Table 4. In a multivariable-adjusted model, we found no association between dietary TFAs with incident HF: multivariable-adjusted HRs (95% CIs) across consecutive quintiles of dietary TFAs were 1.0 (reference), 1.13 (0.90, 1.42), 1.03 (0.82, 1.30), 1.18 (0.94, 1.48), and 1.11 (0.88, 1.39) (Table 5).

TABLE 4.

Characteristics of 18,750 individuals by quintiles of total dietary TFAs (from FFQs)¹

	Dietary TFAs
Characteristics	Q1 (low)	Q2	Q3	Q4	Q5 (high)	P-trend
n	3750	3750	3750	3750	3750
Total dietary TFAs (g/d)2	0.88 (0.02, 1.14)	1.35 (1.14, 1.53)	1.70 (1.53, 1.87)	2.05 (1.87, 2.28)	2.61 (2.28, 6.10)
BMI (kg/m2)	25.0 ± 3.23	25.6 ± 3.3	26.0 ± 3.5	26.2 ± 3.6	26.6 ± 3.9	<0.01
Age (y)	65.9 ± 8.5	66.0 ± 8.9	66.1 ± 9.2	66.4 ± 9.5	66.6 ± 9.9	<0.01
Total energy intake (kcal)	1674.6 ± 542.1	1686.3 ± 524.4	1705.9 ± 528.2	1714.2 ± 509.8	1656.1 ± 503.7	0.74
History of atrial fibrillation (%)	6.3	6.8	7.0	7.2	7.8	0.01
History of type 2 diabetes (%)	5.0	5.5	5.1	5.9	7.9	<0.01
Hypertension (%)	39.2	42.0	43.0	43.6	44.0	<0.01
Exercise (%)	70.4	64.7	62.2	60.1	54.1	<0.01
Smoking status (%)
Never	56.8	53.7	54.7	55.3	53.5	0.05
Past	41.2	43.5	42.3	40.5	41.6	0.36
Current	2.0	2.8	3.0	4.2	4.9	<0.01
Alcohol (%)
Rarely	20.1	14.6	14.6	16.2	20.5	0.23
Monthly	6.4	7.1	7.2	7.4	10.1	<0.01
Weekly	36.1	37.1	37.4	40.2	39.4	<0.01
Daily	37.4	41.2	40.8	36.3	30.0	<0.01

¹Some participants had missing data on smoking (n = 13), exercise (n = 319), diabetes (n = 21), hypertension (n = 90), and alcohol use (n = 99). FFQ, food-frequency questionnaire; Q, quintile; TFA, trans fatty acid.

²Values are medians; ranges in parentheses.

³Mean ± SD (all such values).

TABLE 5.

HRs (95% CIs) of heart failure according to quintiles of dietary TFAs (from FFQs)¹

Quintile of TFA ( g/d )	No. of cases/person-years	Model 1	Model 2	Model 3
Q1 (ref): 0.87 (0.025, 1.12)	144/40,378	1.00	1.00	1.00
Q2: 1.33 (1.13, 1.50)	165/39,830	1.14 (0.91, 1.42)	1.10 (0.88, 1.38)	1.13 (0.90, 1.42)
Q3: 1.69 (1.51, 1.84)	156/39,469	1.04 (0.83, 1.30)	0.99 (0.79, 1.25)	1.03 (0.82, 1.30)
Q4: 2.04 (1.85, 2.26)	184/38,610	1.26 (1.01, 1.57)	1.12 (0.90, 1.40)	1.18 (0.94, 1.48)
Q5: 2.60 (2.27, 5.32)	182/37,857	1.21 (0.97, 1.51)	1.03 (0.82, 1.29)	1.11 (0.88, 1.39)
P-trend	—	0.05	0.80	0.39
Modeled as continuous variable2	—	1.07 (1.00, 1.15)	1.01 (0.94, 1.08)	1.03 (0.96, 1.10)

¹Quintile values are medians; ranges in parentheses. Number of subjects per model: model 1, 19,768; model 2, 19,301; and model 3, 19,214. Model 1 adjusted for age; model 2 adjusted as for model 1 plus additionally adjusted for diabetes, smoking, alcohol, exercise, BMI, and atrial fibrillation; and model 3 adjusted as for model 2 plus additionally adjusted for coronary artery disease and hypertension. Analyses were performed by using proportional hazards regression. FFQ, food-frequency questionnaire; Q, quintile; ref, reference; TFA, trans fatty acid.

²ORs per SD increase in dietary TFAs (0.69 g/d).

In a subanalysis, we also evaluated the relation of dietary TFAs with CAD and found no meaningful relation: multivariable-adjusted HRs (95% CIs) across consecutive quintiles were 1.0 (reference), 0.91 (0.78, 1.07), 0.86 (0.74, 1.01), 0.94 (0.81, 1.10), and 0.86 (0.73, 1.01) (Table 6).

TABLE 6.

HRs (95% CIs) of CAD according to quintiles of dietary TFAs (from FFQs)¹

Quintile of dietary TFA (g/d)	No. of cases/person-years	Model 1	Model 2	Model 3
Q1 (ref): 0.88 (0.02, 1.13)	331/36,911	1.00	1.00	1.00
Q2: 1.35 (1.14, 1.52)	308/36,567	0.94 (0.81, 1.10)	0.92 (0.78, 1.07)	0.91 (0.78, 1.07)
Q3: 1.70 (1.53, 1.86)	299/36,391	0.90 (0.77, 1.05)	0.87 (0.74, 1.01)	0.86 (0.74, 1.01)
Q4: 2.05 (1.87, 2.27)	327/35,569	1.00 (0.86, 1.17)	0.94 (0.81, 1.10)	0.94 (0.81, 1.10)
Q5: 2.61 (2.28, 6.10)	306/34,894	0.95 (0.81, 1.10)	0.87 (0.74, 1.02)	0.86 (0.73, 1.01)
P-trend	—	0.73	0.17	0.13
Modeled as continuous variable2	—	0.99 (0.94, 1.04)	0.96 (0.92, 1.02)	0.96 (0.91, 1.01)

¹Quintile values are medians; ranges in parentheses. Number of subjects per model: model 1, 18,750; model 2, 18,321; and model 3, 18,302. Model 1 adjusted for age; model 2 adjusted as for model 1 plus additionally adjusted for smoking, alcohol, exercise, and BMI; model 3 adjusted as for model 2 plus additionally adjusted for diabetes and atrial fibrillation. Analyses were performed by using proportional hazards regression. CAD, coronary artery disease; FFQ, food-frequency questionnaire; Q, quintile; ref, reference; TFA, trans fatty acid.

²ORs per SD increase in dietary TFAs (0.69 g/d).

DISCUSSION

In this study, higher concentrations of plasma phospholipid trans linoleic acid (trans 18:2) were significantly associated with a lower risk of HF after adjustment for potential confounders. This lower risk was observed in all of the models, regardless of the inclusion or exclusion of any potential confounders or mediating factors. In contrast, no association was found between plasma trans 16:1 or trans 18:1 fatty acids and risk of HF. In a secondary analysis, we did not find any association between dietary TFAs and incident HF or CAD after an average follow-up of ∼10 y.

Fat is one of the most challenging dietary components to accurately measure, making it difficult to identify associations between dietary fat and health. Although many previous studies have examined the role of fat consumption on CAD risk, their conclusions are potentially hampered because assessment of dietary fat by commonly used methods is limited by recall biases and measurement errors (26). Consequently, biomarkers are increasingly being used in nutritional epidemiology to assist in dietary measurement and to deal with problems inherent in self-reported intakes (27).

Although many previous studies have focused on the relation of TFAs with CAD and intermediate phenotypes, little is known about the association of TFAs with HF in the literature. In earlier related studies, higher plasma concentrations of trans 18:2 fatty acids were associated with a higher risk of primary cardiac arrest (28) and fatal ischemic heart disease among older adults (29). In rodent models of HF, however, a diet high in fat and low in carbohydrate prevented the development and progression of HF compared with low-fat/high-carbohydrate diets (30, 31). Although underlying mechanisms for such protection are unclear, it is increasingly being recognized that the type of fats consumed appears to be far more relevant for cardiometabolic health than the proportion of energy consumed from total fat (32, 33).

A significant amount of investigation has been done on the relation of fat consumption with CVD (34, 35). Some studies have reported increased risk of type 2 diabetes (36), obesity (11), dyslipidemia (12), and hypertension (13) with TFAs. However, other studies did not find deleterious effects of TFAs with diabetes (14), dyslipidemia (37), and CAD (16). Few of the above-mentioned studies evaluated individual TFAs. Various types of TFAs are produced during partial hydrogenation of vegetable and seed oil. These include trans isomers of oleic (trans 18:1) and linoleic (trans 18:2) acids. A third major group—palmitoleic acid (trans 16:1)—is usually produced by bacterial metabolism of unsaturated fatty acids in ruminant animals and found in beef, lamb, and dairy products (38). Divergent reports suggest the importance of studying individual TFAs because their effects may be heterogeneous.

The fact that 18:1 trans isomers have biophysical characteristics (39) different than 18:2 trans isomers encourages the hypothesis that these 2 classes of TFAs may actually have different associations with HF risk. trans 18:2 Fatty acids are preferentially incorporated in the sn-2 (middle) position of phospholipids where PUFAs are typically found. In contrast, trans 18:1 fatty acids are incorporated in the sn-1 position where saturated fats are usually found. This differential incorporation within cell membranes can alter the composition of cell membrane phospholipid bilayer and potentially affect membrane properties, resulting in subsequent changes in physiologic mechanisms.

At this point, with limited evidence, we cannot be certain that the observed association between plasma trans 18:2 fatty acids and a lower risk of HF is causal. Antiinflammatory properties of trans 18:2 fatty acids (40) may partially explain the observed inverse relation of plasma trans 18:2 fatty acids with HF. Future studies are warranted for confirmation and to elucidate underlying pathways.

trans 18:1 Isomers have been reported to be associated with foods commonly made with partially hydrogenated vegetable oils, including biscuits, chips, popcorn, margarine, fried foods, and bakery foods. Linoleic acid isomers (trans 18:2) have been reported to be mainly associated with bakery foods, whereas ruminant nutriments are major sources of trans 16:1 fatty acid isomers (41).

Nutrients derived from FFQs are widely used in epidemiologic studies despite the potential for measurement errors and recall biases. We examined the relation of dietary TFAs (assessed by FFQ) with incident HF and CAD in our study. Contrary to previous reports, we found no association between dietary TFAs and incident HF. This null finding was also observed across individual TFAs (ie, 16:1, 18:1, and 18:2; data not shown). Our finding of no association between dietary TFAs and incident CAD is consistent with 2 previously completed studies in male cohorts (42, 43). In contrast, a prospective study in women (44) reported a 27% (95% CI: 3%, 56%) higher risk of CAD among women in the highest quintile of dietary TFA consumption compared with the lowest quintile. Differences in study design, subject characteristics, and other factors may partially explain the divergent results. The average age of participants in that study was ∼46 y (compared with a mean age of 66 y in our cohort). In addition, whereas the study by Hu et al (44) included only women, the present study consisted exclusively of men. Finally, whereas Hu et al (44) analyzed probable and confirmed cases of CAD, the present study included only confirmed cases of CAD.

Our study has some limitations. The measurement of plasma phospholipid TFA concentrations was performed on baseline blood samples and therefore may not fully represent habitual patterns of dietary TFA consumption. Also, whereas plasma concentrations of TFAs are biomarkers of dietary TFA consumption, it is important to note that there are other nondietary factors that may influence measured fatty acid biomarkers, for example, metabolism, endogenous synthesis, and sample handling/storage, and diseases (diabetes, malabsorption). In addition, participants in our study were all middle-aged male physicians and were mostly white; hence, our findings cannot be generalized to the overall US population. The assessment of dietary fat intake by using FFQs is limited by recall bias and measurement errors. Furthermore, despite a reliable positive predictive value of self-reported HF against review of medical records, we cannot exclude misclassification of HF cases in our study. Because subjects were unaware of their plasma or dietary TFA levels, such misclassification of the outcome would be nondifferential and is more likely to bias the results toward the null. Although we did not correct for multiple testing, our findings of an inverse association of trans 18:2 fatty acids with HF would remain significant on Bonferroni adjustment at an α of 0.017 (3 exposures and 3 different tests).

Strengths of this study include the prospective nested case-control design, matching on potential confounders, use of biomarkers in exposure assessment, long follow-up time, and availability of data on several potential confounding factors.

In conclusion, our data are consistent with a lower risk of HF with plasma trans 18:2 but not trans 16:1 or trans 18:1 fatty acids among male physicians. Dietary TFAs were not related to incident HF or CAD risk. If confirmed in future studies, our data on plasma trans 18:2 fatty acids would emphasize the importance of heterogeneity and pleotropic effects of TFAs on CVD.