Dietary Fat Intake Is Differentially Associated with Risk of Paroxysmal Compared with Sustained Atrial Fibrillation in Women^,,

Abstract

Background: Dietary fats have effects on biological pathways that may influence the development and maintenance of atrial fibrillation (AF). However, associations between n–3 (ω-3) polyunsaturated fatty acids and AF are inconsistent, and data on other dietary fats and AF risk are sparse.

Objectives: We examined the association between dietary fatty acid (FA) subclasses and risk of incident AF and explored whether these associations differed for sustained and paroxysmal AF.

Methods: We conducted a prospective cohort study in 33,665 women ≥45 y old without cardiovascular disease (CVD) and AF at baseline in 1993. Fat intake was estimated from food frequency questionnaires at baseline and in 2004. Incident AF was confirmed by medical records through October 2013. AF patterns were classified according to the most sustained form of AF within 2 y of diagnosis. Cox proportional hazards models with the use of a competing risk model approach estimated the RR.

Results: Over 19.2 y, 1441 cases of incident AF (929 paroxysmal and 467 persistent/chronic) were confirmed. Intakes of total fat and FA subclasses were not associated with risk of AF. Saturated fatty acids (SFAs) and monounsaturated fatty acids (MUFAs) were differentially associated with AF patterns. The RR for a 5% increment of energy from SFAs was 1.47 (95% CI: 1.04, 2.09) for persistent/chronic and 0.85 (95% CI: 0.66, 1.08) for paroxysmal AF (P-difference = 0.01). For MUFAs, the RR for a 5% increment was 0.67 (95% CI: 0.46, 0.98) for persistent/chronic and 1.03 (95% CI: 0.78, 1.34) for paroxysmal AF, although the difference between patterns was not significant (P-difference = 0.07).

Conclusions: Dietary fat was not associated with risk of incident AF in women without established CVD or AF. High SFA and low MUFA intakes were associated with greater risk of persistent or chronic, but not paroxysmal, AF. Improving dietary fat quality may play a role in the prevention of sustained forms of AF. The Women’s Health Study was registered at clinicaltrials.gov as NCT00000479.

Introduction

Atrial fibrillation (AF)⁹ is the most common cardiac arrhythmia in clinical practice. At present, an estimated 5.1 million Americans suffer from AF, and the incidence is projected to surpass 12 million by 2050 (1). The clinical consequences of AF are significant; individuals with AF are at greater risk of thromboembolic stroke, congestive heart failure (CHF), cognitive dysfunction, and mortality (2–4). Current treatment options have limited long-term success rates and come with significant risks (5, 6), and patients who go on to develop sustained forms of AF (persistent/permanent) are often less amenable to treatment (7, 8). Recent data also suggest that patients with sustained forms of AF have higher rates of cardiovascular disease (CVD), morbidity, and mortality (2, 3). Therefore, strategies that focus on the prevention of AF and AF progression are needed to reduce the overall burden of morbidity and mortality from AF. Prevention in women is of particular concern, because the absolute number of women with AF surpasses that among men after age 75 (9), and women with AF have higher mortality and stroke rates than do their male counterparts (10).

Dietary fats have effects on cellular electrophysiology and structural remodeling, which have been implicated in the development and maintenance of AF (11). Previous studies of FAs and incident AF have focused mainly on n–3 PUFAs and have yielded conflicting results (12–19). In a recently conducted large multicenter randomized clinical trial, the PREDIMED (PREvención con DIeta MEDiterránea) study, a Mediterranean diet supplemented with extra-virgin olive oil was protective against incident AF compared with a low fat diet, whereas a Mediterranean diet supplemented with nuts was not (20). Therefore, dietary modification with other fat subclasses besides n–3 PUFAs may play a role in the prevention of AF. Current examinations of other fat subclasses are limited to circulating blood concentrations from one cohort (19, 21), which are not highly correlated with dietary intake. Further, prior studies have not assessed whether FAs differ with respect to the development of sustained (persistent/permanent) rather than paroxysmal AF.

Therefore, to address these gaps, we prospectively examined the association between subclasses of FA intake and risk of AF within the Women’s Health Study (WHS). Further, we explored whether these associations differed for sustained (persistent/permanent) vs. paroxysmal AF patterns.

Methods

Study population.

The WHS (22) is a completed randomized clinical trial that examined the effects of low-dose aspirin, β-carotene, and vitamin E in the primary prevention of CVD and cancer in 39,876 female health care professionals in the United States. Women were ≥45 y old and free of CVD, cancer, and other chronic diseases at random assignment, which began in 1993. Randomized treatment ended on 31 March 2004, and subsequently women were invited to participate in continued observational follow-up. Information on baseline characteristics and study outcomes were collected through mailed questionnaires at baseline, 6 mo, and every 12 mo subsequently. Written informed consent was obtained from all participants and the study was approved by the institutional review board of Brigham and Women’s Hospital, Boston.

Ascertainment of dietary fat.

At baseline, 39,310 women (98.6%) completed a 131-item FFQ, and 82.8% of these women also completed a second FFQ administered in 2004. Participants were asked how often, on average, they had consumed a specified portion size for each food item over the past year. The FFQs also included questions to distinguish type of fat or oil used while cooking and brand and type of margarines. We calculated average fat intake by multiplying the frequency of consumption of each item by its fat content, and summing across all foods. All nutrient values, including FAs, were obtained from the Harvard University Food Composition Database, which includes values from the USDA, food manufacturers, and independent academic sources. In order to address the limitations of assessing trans fats from food databases, the Biomarker Laboratory at Harvard University conducted analyses of the trans fat content of a variety of foods in 1991, 2000, 2002, and 2007. Analyzed foods included butter, margarines, cooking fats, commercially prepared bakery and snack items, chocolate, and fast food French-fried potatoes. In previous validation studies, the FFQ has been shown to provide a reasonable measure of FA intake when compared with dietary records (23). The correlation between fat intake assessed by FFQ (average in 1980, 1984, and 1986) and 1 wk diet records (average of 1980 and 1986) was r = 0.83 for total fat and r = 0.95 for SFAs (23).

We evaluated various FA subclasses, including SFAs, MUFAs, trans fat, and n–3 and n–6 PUFAs. Because of potential different biological effects of individual n–3 PUFAs (24), we explored separate associations for intermediate-chain α-linolenic acid (ALA) (18:3n–3), long-chain EPA (20:5n–3), and DHA (22:6n–3). Furthermore, we explored intake of individual SFAs based upon their differential associations with CVD risk (25, 26).

In secondary analysis, we also explored the association between adherence to a Mediterranean-style diet and risk of AF. We quantified adherence using the alternate Mediterranean Diet, which was previously adapted from the Trichopolou score for the American population (27). Specifically, the alternate Mediterranean Diet score includes 9 components: high intake of vegetables, fruits, nuts, whole grains, legumes, and fish; ratio of MUFA to SFA; moderate intake of alcohol; and low intake of red and processed meat. For all components except red and processed meat and alcohol, women received 1 point for intake above the median. Women received 1 point for intake below the median of red and processed meat intake and 1 point for moderate alcohol intake (0.5–1 drink/d). The possible range for the Mediterranean diet score was 0–9 points, with higher scores representing greater resemblance to the Mediterranean-style diet.

Ascertainment of AF.

Details about the confirmation of AF have been reported previously (28). Women were asked to report diagnoses of incident AF on the questionnaire at baseline, 48 mo, and annually thereafter. Beginning on 19 September 2006, women who reported an incident AF event on at least one annual questionnaire were sent an additional questionnaire to confirm the AF episode and collect additional information. Additionally, we requested permission to obtain their medical records, including available electrocardiograms, rhythm strips, 24-h electrocardiogram monitoring, and testing regarding cardiac structure and function. For all deceased participants who reported AF during the trial and extended follow-up period, we contacted family members to obtain consent and additional relevant information. Medical records were reviewed by study physicians and an incident AF event was confirmed if there was electrocardiogram evidence of AF or if a medical report clearly indicated a personal history of AF. We defined the date of onset of AF as the earliest date in the medical records when AF documentation was believed to have occurred.

Additionally, we classified AF by types according to the 2006 AHA/American College of Cardiology/European Society of Cardiology Guidelines (8). Paroxysmal AF was defined as self-terminating within 7 d, persistent AF was defined as requiring cardioversion or lasting ≥7 d, and permanent AF was defined as lasting >1 y and/or resulting in attempts to convert rhythm being abandoned. We classified the AF pattern according to the most severe form of AF (permanent > persistent > paroxysmal) within 2 y of initial AF onset to characterize AF type in women consistently throughout the study. We combined persistent and permanent AF into sustained AF, as the difference between groups was primarily determined by physician decision not to pursue cardioversion (4).

Population for analysis.

Of the original cohort, 10.8% opted out of the observational follow-up in 2004. These women were excluded from the main analysis because their AF diagnosis and pattern could not be confirmed. However, we performed a sensitivity analysis using self-reported AF events in all women as the main outcome variable to ensure that exclusion of these women did not significantly alter our results. We also excluded women with a history of AF (n = 769) or a confirmed cardiovascular event (n = 10) before study entry, as well as women who did not complete the FFQ (n = 144) or who had invalid FFQ data (reported implausible total energy intake <600 or >3500 kcal/d, or left >70 items blank; n = 964), leaving 33,665 women for our main analyses.

Statistical analysis.

Because the temporal relation and critical window of dietary fat intake for AF risk is unknown, we modeled fat intake in 2 ways. Our primary analyses focused on baseline FFQ, which assumes a longer induction period for dietary fat on AF risk. In alternative analyses, we updated fat intake with the second FFQ in 2004, and estimated the cumulative average of fat intake over follow-up (29). This method assumes the most recent diet plays a stronger role on AF risk. In the updated analyses, baseline fat intake predicted AF incidence from the baseline questionnaire until the second FFQ, whereas the average of fat intake from the 2 FFQs predicted incidence from the second FFQ until the end of follow-up. Furthermore, to avoid time-dependent confounding by these intermediate events, we did not update FA information for women who developed incident hypertension, high cholesterol, diabetes, CVD, or CHF before the return of the second FFQ (29). To assess the relation between individual FAs and FA subclasses, we calculated partial Spearman correlation coefficients (r) adjusted for age.

Each woman contributed person-months of follow-up from the date of return of the baseline questionnaire to the first occurrence of AF, date of death, or date of the most recent morbidity information collection date (1 October 2013), whichever came first. Women for whom AF pattern could not be characterized were censored from the analysis at the time of their AF diagnosis (n = 45). We modeled fat intake as a percentage of total energy. To simulate a “substitution” for carbohydrates, we created a multivariable nutrient density model by adjusting for total energy intake and percent of energy from protein. In analyses of FA subclasses, we additionally adjusted for the percentage of energy from other FA subclasses (29). We grouped women into quintiles according to the distribution of fat intake for each FA subtype.

We used Cox proportional hazards models to estimate HRs as estimates of the RR of total AF. The assumption of proportional hazards was not violated. In our primary analysis, with the use of only baseline diet information, we controlled for smoking, BMI, height, alcohol, exercise, education, race, randomization groups (β-carotene, vitamin E, and aspirin), systolic blood pressure, and diagnosis of hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication) and physician-diagnosed diabetes all at baseline. In models in which we used cumulative average of dietary fat, we also updated information on smoking, BMI, alcohol, exercise, and systolic blood pressure assessed at the time the second FFQ was completed and included them as time-varying covariates in our model. For both analyses, we further adjusted for incident diagnosis of diabetes, high blood pressure, high cholesterol, CVD, and CHF during follow-up in separate models to assess potential mediation of the dietary fat–AF pathway by these risk factors. We tested for linear trend by assigning the median value of fat intake to each quintile and modeling this variable as a continuous variable.

Next, we evaluated whether the association between dietary fat and risk of AF differed by AF subtype in Cox proportional hazards models stratified by paroxysmal and nonparoxysmal AF with the use of a competing risk model approach detailed by Lunn and McNeill (30). This method provides separate associations between a risk factor and its relative hazards for the 2 AF types simultaneously under a proportional hazards assumption while assuming different associations between each model variable and paroxysmal and nonparoxysmal AF. To test formally for differences in the association between the 2 outcomes, we ran a reduced model in which each FA was forced to have a single effect across both outcomes, whereas the effects of all variables were allowed to be different. We then used a likelihood ratio test comparing the full competing risk model with the reduced models. All reported P values are 2-sided and P < 0.05 is considered statistically significant. Analyses were performed with the use of SAS software, version 9.2.

Results

Over 20 y of follow-up (median: 19.2; IQR: 17.9, 19.7 y), we documented 1441 cases of incident AF, of which 929 (64.5%) remained paroxysmal and 467 (32.4%) were persistent/permanent AF within 2 y of initial onset. Common patterns in characteristics were seen across intake of all fat subclasses (Table 1). Women with a higher intake of any FA had a higher intake of total fat and total energy and higher BMI, and were more likely to smoke and to have hypertension at baseline. Furthermore, women with a high fat intake were less likely to have a graduate degree or consume ≥2 alcoholic drinks/d. Women with a higher intake of SFAs and MUFAs were less likely to have hypercholesterolemia at baseline.

TABLE 1.

Population characteristics of 33,665 women from the Women’s Health Study by quintiles of dietary fat intake at baseline¹

	SFAs		MUFAs		PUFAs		Trans fat
	Q1	Q5	Q1	Q5	Q1	Q5	Q1	Q5
Total, n	6693	6682	6707	6677	6723	6729	6712	6701
Median intake, % energy	7.2	13.4	7.9	14.5	4.1	7.6	0.56	1.91
Range intake, % energy	<8.2	>12.2	<9.0	>13.3	<4.6	>6.9	<0.73	>1.59
Age, y	55 ± 0.09	54 ± 0.08	55 ± 0.09	54 ± 0.08	54 ± 0.08	55 ± 0.09	55 ± 0.09	54 ± 0.08
BMI, kg/m2	24.8 ± 0.06	27.1 ± 0.06	25.0 ± 0.06	27.0 ± 0.06	25.3 ± 0.06	26.5 ± 0.06	25.0 ± 0.06	27.0 ± 0.06
Height, cm	165 ± 0.10	165 ± 0.10	165 ± 0.10	165 ± 0.10	165 ± 0.10	165 ± 0.10	165 ± 0.10	165 ± 0.10
Current smoker	72	20	6	19	9	14	6	16
Education
Bachelor’s degree	22	19	23	20	22	22	23	19
Master’s degree or doctorate	24	14	24	14	22	16	25	14
Caucasian	84	86	85	86	86	87	85	86
Hypertension3	22	26	21	26	21	24	21	26
High cholesterol4	30	25	28	26	26	27	27	27
Diabetes	2	3	2	3	2	3	2	3
Total fat, % energy	22.3 ± 0.04	37.5 ± 0.04	21.8 ± 0.03	38.2 ± 0.03	24.5 ± 0.06	35.2 ± 0.06	23.9 ± 0.06	35.2 ± 0.06
SFAs	6.9 ± 0.01	13.9 ± 0.01	7.4 ± 0.02	13.0 ± 0.02	9.1 ± 0.03	11.0 ± 0.03	7.9 ± 0.03	12.0 ± 0.03
MUFAs	8.1 ± 0.02	14.0 ± 0.02	7.6(0.01	14.9 ± 0.01	9.0 ± 0.03	13.1 ± 0.03	8.5 ± 0.02	13.5 ± 0.02
PUFAs	5.1 ± 0.02	6.2 ± 0.02	4.6 ± 0.02	6.9 ± 0.02	4.0 ± 0.01	8.0 ± 0.01	5.1 ± 0.02	6.5 ± 0.02
Trans fat	0.7 ± 0.01	1.6 ± 0.01	0.7 ± 0.01	1.7 ± 0.01	0.9 ± 0.01	1.4 ± 0.01	0.6 ± 0.003	2.0 ± 0.003
Total energy, kcal/d	1681 ± 6.5	1748 ± 6.5	1660 ± 6.5	1771 ± 6.5	1687 ± 6.5	1731 ± 6.5	1711 ± 6.5	1746 ± 6.5
Alcohol >2 drinks/d	5	3	4	3	5	3	5	2

¹Values are means ± SEMs or percentages, unless otherwise indicated. With the exception of age, all characteristics are age-standardized. Q, quintile.

²Frequency of category, all such values (percentages).

³Defined as SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication.

⁴Defined as total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication.

Total fat, major fat classes, and risk of AF.

In isocaloric models, there was no significant association between total fat and risk of total incident AF events in age-adjusted and multivariable models (Table 2). The multivariable RR of total AF for a 5% energy increase in total fat intake was 0.97 (95% CI: 0.92, 1.02). Similarly, there were no significant associations between intake of any type of fat—SFAs, MUFAs, trans fat or PUFAs—and risk of combined incident AF events. However, when we examined paroxysmal vs. sustained AF events separately, intake of SFAs and MUFAs both was significantly associated with risk of sustained AF, but not paroxysmal AF (Table 3), even after controlling for other AF risk factors and fat classes. The multivariable RR for a 5% energy increase in SFAs was 1.47 (95% CI: 1.04, 2.09) for sustained AF, compared with 0.85 (95% CI: 0.66, 1.08) for paroxysmal AF (P-difference = 0.01). Alternatively, MUFAs were associated with a lower risk of sustained AF. The multivariable RR for a 5% energy increase in MUFAs was 0.67 (95% CI: 0.46, 0.98) for sustained AF, compared with 1.03 (95% CI: 0.78, 1.34) for paroxysmal AF, although the difference in estimates did not quite reach statistical significance (P-difference = 0.07). Further adjustment for potential mediating factors, such as incident hypertension, high cholesterol, diabetes, CVD, and CHF, did not appreciably alter the magnitude of association. The association between PUFAs and trans fat did not differ by AF type (P-difference = 0.74 and 0.84, respectively).

TABLE 2.

Multivariable RR (95% CI) of incident atrial fibrillation by quintiles of FA intake in 33,665 women from the Women’s Health Study¹

	Quintiles of FAs
	Q1	Q2	Q3	Q4	Q5	P- trend
Median of total fat, % energy	22.3	26.9	30	33.2	37.8
Range of total fat, % energy	<25.0	25.0–28.4	28.5–31.4	31.5–35.1	>35.1
Total, n	6702	6810	6778	6710	6665
Cases, n	277	302	285	284	293
Basic model 12	1.0 (ref)	1.08 (0.92, 1.28)	1.04 (0.88, 1.23)	1.07 (0.91, 1.27)	1.17 (0.99, 1.38)	0.09
Multivariable model 13	1.0 (ref)	1.06 (0.89, 1.25)	0.98 (0.83, 1.16)	0.93 (0.78, 1.10)	0.95 (0.80, 1.13)	0.29
Multivariable model 13 + intermediary factors4	1.0 (ref)	1.06 (0.90, 1.25)	0.99 (0.83, 1.17)	0.94 (0.80, 1.12)	0.97 (0.82, 1.16)	0.42
Median of SFAs, % energy	7.2	8.9	10.1	11.4	13.4
Range of SFAs, % energy	<8.2	8.2–9.5	9.6–10.7	10.8–12.2	>12.2
Total, n	6693	6795	6766	6729	6682
Cases, n	276	303	302	279	281
Basic model 25	1.0 (ref)	1.22 (1.01, 1.46)	1.29 (1.05, 1.59)	1.25 (0.99, 1.57)	1.31 (1.02, 1.69)	0.15
Multivariable model 26	1.0 (ref)	1.16 (0.96, 1.40)	1.18 (0.95, 1.46)	1.14 (0.90, 1.45)	1.08 (0.83, 1.41)	0.98
Multivariable model 26 + intermediary factors4	1.0 (ref)	1.18 (0.97, 1.42)	1.23 (0.99, 1.52)	1.18 (0.93, 1.49)	1.16 (0.90, 1.51)	0.64
Median of MUFAs, % energy	7.9	9.8	11.2	12.5	14.5
Range of MUFAs, % energy	<9.0	9.0–10.5	10.6–11.8	11.9–13.3	>13.3
Total, n	6707	6805	6781	6695	6677
Cases, n	293	279	300	273	296
Basic model 25	1.0 (ref)	0.86 (0.71, 1.05)	0.90 (0.73, 1.13)	0.81 (0.63, 1.04)	0.87 (0.65, 1.15)	0.33
Multivariable model 26	1.0 (ref)	0.90 (0.74, 1.09)	0.93 (0.74, 1.16)	0.81 (0.62, 1.05)	0.89 (0.66, 1.20)	0.42
Multivariable model 26 + intermediary factors4	1.0 (ref)	0.89 (0.73, 1.08)	0.92 (0.74, 1.15)	0.79 (0.61, 1.03)	0.86 (0.65, 1.16)	0.29
Median of trans fat, % energy	0.58	0.85	1.09	1.38	1.91
Range of trans fat, % energy	<0.73	0.73–0.97	0.98–1.22	1.23–1.59	>1.59
Total, n	6712	6743	6744	6765	6701
Cases, n	318	268	295	257	303
Basic model 25	1.0 (ref)	0.83 (0.70, 0.99)	0.94 (0.78, 1.12)	0.83 (0.68, 1.01)	1.01(0.82,1.24)	0.51
Multivariable model 26	1.0 (ref)	0.83 (0.70, 1.00)	0.93 (0.77, 1.12)	0.79 (0.65, 0.97)	0.94 (0.76, 1.16)	0.85
Multivariable model 26 + intermediary factors4	1.0 (ref)	0.81 (0.68, 0.97)	0.91 (0.76, 1.09)	0.77 (0.63, 0.94)	0.92 (0.75, 1.14)	0.76
Median of total PUFAs, % energy	4.1	5	5.6	6.4	7.6
Range of total PUFAs, % energy	<4.6	4.6–5.2	5.3–5.9	6.0–6.9	>6.9
Total, n	6723	6739	6780	6694	6729
Cases, n	264	277	309	275	316
Basic model 25	1.0 (ref)	1.07 (0.90, 1.28)	1.17 (0.98, 1.40)	1.06 (0.88, 1.27)	1.17 (0.97, 1.42)	0.15
Multivariable model 26	1.0 (ref)	1.05 (0.88, 1.26)	1.14 (0.95, 1.36)	1.00 (0.83, 1.21)	1.12 (0.92, 1.36)	0.38
Multivariable model 26 + intermediary factors4	1.0 (ref)	1.07 (0.90, 1.27)	1.15 (0.96, 1.37)	1.02 (0.84, 1.23)	1.13 (0.93, 1.38)	0.31

¹Q, quintile; ref, reference.

²Includes protein (percentage of energy), total calories, and age.

³Adjusted for age, protein (percentage of energy), total calories, smoking, BMI, height, alcohol, exercise, education, race, randomization group (β-carotene, vitamin E, and aspirin), systolic blood pressure, and diagnosis of hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), and diabetes, all at baseline.

⁴Includes diagnosis of hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), diabetes, cardiovascular disease, or congestive heart failure during follow-up.

⁵Includes variables in basic model 1 plus SFAs, MUFAs, total PUFAs, and trans fat in the same model.

⁶Includes variables in multivariable model 1 plus SFAs, MUFAs, total PUFAs, and trans fat in the same model.

TABLE 3.

Multivariable RR of incident paroxysmal and persistent/chronic AF by quintiles of FA intake in 33,665 women from the Women’s Health Study¹

	Quintiles of FAs
	Q1	Q2	Q3	Q4	Q5	P-trend	P-difference2
Median of total fat, % energy	22.3	26.9	30	33.2	37.8
Range of total fat, % energy	<25.0	25.0–28.4	28.5–31.4	31.5–35.1	>35.1
Total, n	6702	6810	6778	6710	6665
Paroxysmal AF
Cases, n	179	193	186	190	181
Basic model 13	1.0 (ref)	1.07 (0.88, 1.32)	1.05 (0.86, 1.29)	1.11 (0.90, 1.36)	1.11 (0.90,1.37)	0.30
Multivariable model 14	1.0 (ref)	1.07 (0.87, 1.31)	1.00 (0.81, 1.23)	0.99 (0.80, 1.22)	0.94 (0.75, 1.17)	0.41
Multivariable model 14 + intermediary factors5	1.0 (ref)	1.06 (0.87, 1.31)	1.00 (0.81, 1.24)	1.00 (0.81, 1.23)	0.95 (0.76, 1.18)	0.48
Persistent/chronic AF							0.72
Cases, n	92	104	83	89	99
Basic model 13	1.0 (ref)	1.13 (0.85, 1.49)	0.92 (0.69, 1.24)	1.02 (0.76, 1.37)	1.21 (0.91, 1.61)	0.33
Multivariable model 14	1.0 (ref)	1.07 (0.80, 1.42)	0.85 (0.62, 1.15)	0.84 (0.62, 1.13)	0.93 (0.69, 1.26)	0.31
Multivariable model 14 + intermediary factors5	1.0 (ref)	1.08 (0.81, 1.44)	0.85 (0.63, 1.16)	0.86 (0.64, 1.16)	0.97 (0.72, 1.77)	0.48
Median of SFAs, % energy	7.2	8.9	10.1	11.4	13.4
Range of SFAs, % energy	<8.2	8.2–9.5	9.6–10.7	10.8–12.2	>12.2
Total, n	6693	6795	6766	6729	6682
Paroxysmal AF
Cases, n	179	213	187	181	169
Basic model 26	1.0 (ref)	1.26 (1.00, 1.58)	1.12 (0.87, 1.46)	1.10 (0.83, 1.47)	1.04 (0.75, 1.43)	0.67
Multivariable model 27	1.0 (ref)	1.21 (0.96, 1.52)	1.04 (0.80, 1.36)	1.03 (0.77, 1.38)	0.89 (0.64, 1.23)	0.19
Multivariable model 27 + intermediary factors5	1.0 (ref)	1.22 (0.97, 1.54)	1.06 (0.81, 1.39)	1.05 (0.78, 1.40)	0.94 (0.68, 1.30)	0.32
Persistent/chronic AF							0.01
Cases, n	93	81	102	87	104
Basic model 26	1.0 (ref)	1.05 (0.75, 1.46)	1.54 (1.07, 2.21)	1.51 (1.01, 2.26)	2.05 (1.32, 3.16)	0.00
Multivariable model 27	1.0 (ref)	0.98 (0.69, 1.38)	1.41 (0.97, 2.05)	1.37 (0.91, 2.09)	1.70 (1.08, 2.67)	0.03
Multivariable model 27 + intermediary factors5	1.0 (ref)	1.01 (0.72, 1.42)	1.50 (1.04, 2.16)	1.41 (0.93, 2.13)	1.81 (1.16, 2.83)	0.01
Median of MUFAs, % energy	7.9	9.8	11.2	12.5	14.5
Range of MUFAs, % energy	<9.0	9.0–10.5	10.6–11.8	11.9–13.3	>13.3
Total, n	6707	6805	6781	6695	6677
Paroxysmal AF
Cases, n	189	178	192	183	187
Basic model 26	1.0 (ref)	0.86 (0.67, 1.09)	0.94 (0.72, 1.24)	0.93 (0.68, 1.26)	1.00 (0.70, 1.42)	0.87
Multivariable model 27	1.0 (ref)	0.89 (0.70, 1.14)	0.95 (0.72, 1.26)	0.91 (0.66, 1.26)	1.01 (0.70, 1.45)	0.86
Multivariable model 27 + intermediary factors5	1.0 (ref)	0.88 (0.69, 1.13)	0.95 (0.72, 1.26)	0.90 (0.66, 1.24)	0.97 (0.68, 1.40)	0.97
Persistent/chronic AF							0.07
Cases, n	97	96	96	82	96
Basic model 26	1.0 (ref)	0.89 (0.64, 1.25)	0.80 (0.54, 1.18)	0.60 (0.38, 0.95)	0.61 (0.37, 1.01)	0.03
Multivariable model 27	1.0 (ref)	0.95 (0.67, 1.34)	0.82 (0.55, 1.23)	0.60 (0.38, 0.96)	0.62 (0.37, 1.05)	0.04
Multivariable model 27 + intermediary factors5	1.0 (ref)	0.93 (0.66, 1.31)	0.82 (0.55, 1.22)	0.60 (0.38, 0.95)	0.62 (0.37, 1.04)	0.04
Median of trans fat, % energy	0.58	0.85	1.09	1.38	1.91
Range of trans fat, % energy	<0.73	0.73–0.97	0.98–1.22	1.23–1.59	>1.59
Total, n	6712	6743	6744	6765	6701
Paroxysmal AF
Cases, n	199	177	195	164	194
Basic model 26	1.0 (ref)	0.89 (0.72, 1.11)	1.00 (0.80, 1.26)	0.85 (0.66,1.09)	1.03 (0.80, 1.33)	0.60
Multivariable model 27	1.0 (ref)	0.90 (0.73, 1.13)	1.00 (0.80, 1.26)	0.80(0.62, 1.04)	0.97 (0.75,1.26)	0.91
Multivariable model 27 + intermediary factors5	1.0 (ref)	0.90 (0.72, 1.11)	0.99 (0.79, 1.25)	0.80 (0.62, 1.03)	0.96 (0.74, 1.25)	0.90
Persistent/chronic AF							0.84
Cases, n	109	84	92	86	96
Basic model 26	1.0 (ref)	0.75 (0.55, 1.01)	0.85 (0.62, 1.16)	0.82 (0.58, 1.15)	0.95 (0.67, 1.36)	0.85
Multivariable model 27	1.0 (ref)	0.72 (0.53, 0.99)	0.81 (0.58, 1.12)	0.77 (0.55, 1.10)	0.87 (0.60, 1.26)	0.75
Multivariable model 27 + intermediary factors5	1.0 (ref)	0.70 (0.51, 0.95)	0.80 (0.58, 1.10)	0.74 (0.53, 1.05)	0.84 (0.59, 1.22)	0.65
Median of total PUFAs, % energy	4.1	5	5.6	6.4	7.6
Range of total PUFAs, % energy	<4.6	4.6–5.2	5.3–5.9	6.0–6.9	>6.9
Total, n	6723	6739	6780	6694	6729
Paroxysmal AF
Cases	166	172	204	189	198
Basic model 26	1.0 (ref)	1.05 (0.85, 1.31)	1.22 (0.98, 1.52)	1.14 (0.90, 1.43)	1.15 (0.90, 1.46)	0.24
Multivariable model 27	1.0 (ref)	1.05 (0.84, 1.31)	1.20 (0.96, 1.50)	1.10 (0.87, 1.39)	1.13 (0.88, 1.45)	0.34
Multivariable model 27 + intermediary factors5	1.0 (ref)	1.06 (0.85, 1.33)	1.22 (0.98, 1.53)	1.12 (0.89, 1.42)	1.14 (0.89, 1.46)	0.29
Persistent/chronic AF							0.74
Cases	91	95	96	80	105
Basic model 26	1.0 (ref)	1.10 (0.82, 1.48)	1.10 (0.81, 1.50)	0.96 (0.69, 1.34)	1.24 (0.89, 1.73)	0.35
Multivariable model 27	1.0 (ref)	1.06 (0.79, 1.44)	1.05 (0.77, 1.43)	0.90 (0.64, 1.26)	1.11 (0.79, 1.57)	0.79
Multivariable model 27 + intermediary factors5	1.0 (ref)	1.07 (0.79, 1.44)	1.05 (0.77, 1.43)	0.90 (0.65, 1.26)	1.13 (0.81, 1.58)	0.70

¹AF, atrial fibrillation; Q, quintile; ref, reference.

²P value for the difference in the association between paroxysmal and sustained types of AF.

³Adjusted for age, protein (percentage of energy), and total calories.

⁴Adjusted for age, protein (percentage of energy), total calories, smoking, BMI, height, alcohol, exercise, education, race, randomization group (β-carotene, vitamin E, and aspirin), systolic blood pressure, and diagnosis of hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), and diabetes, all at baseline.

⁵Includes diagnosis of incident hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), diabetes, cardiovascular disease, and congestive heart failure during follow-up.

⁶Adjusted for basic model 1 plus SFAs, MUFAs, total PUFAs, and trans fat in the same model.

⁷Adjusted for multivariable model 1 plus SFAs, MUFAs, total PUFAs, and trans fat in the same model.

These associations were not explained by adherence to the Mediterranean diet. The RR of AF comparing women with a Mediterranean score ≥7 (out of 9) points to ≤2 points was 1.05 (95% CI: 0.84, 1.31) for total AF, 0.96 (95% CI: 0.73, 1.25) for paroxysmal AF, and 1.20 (95% CI: 0.81, 1.79) for sustained AF (Supplemental Table 1). However, the Mediterranean diet score was not correlated strongly with intake of MUFAs (r = –0.01; P = 0.09) or SFAs (r = −0.23; P < 0.001).

When we updated dietary fat intake with the use of information from the second FFQ, the association between SFAs and MUFAs and risk of sustained AF was attenuated and no longer statistically significant (Supplemental Table 2); however, the associations for SFAs remained statistically different for sustained vs. paroxysmal AF (P-difference = 0.01). Consistent with the analysis with the use of baseline-only diet, there were no significant associations between intake of other fat subclasses and risk of total, persistent, or sustained AF. Thus, we restricted subsequent analyses to the fat intake from the baseline FFQ only.

Polyunsaturated subclasses: n–3 and n–6 PUFAs and risk of AF.

When we explored PUFA subclasses, we found no significant association between intake of total n–6 (P-trend = 0.21) or n–3 (P-trend = 0.70) PUFAs and risk of incident AF, and these associations did not vary by AF subtype (P-difference = 0.69 for n–6 and 0.89 for n–3 PUFAs) (Supplemental Table 3). When the individual n–3 PUFAs were examined separately, EPA and DHA were not associated with incident AF or AF patterns (Supplemental Table 4). There was a trend toward an inverse association between intake of ALA and total incident AF and sustained AF, but this did not reach statistical significance (P-trend = 0.07 for both endpoints) (Supplemental Table 4).

Individual SFAs and risk of AF.

The median intake of SFAs accounted for 10% of total energy and the primary individual SFAs were the long-chain SFAs—16:0 (x = 6.8% total energy) and 18:0 (x = 2.4% total energy) (Supplemental Table 5). Because of low intake, we combined 12:0 and 14:0 SFAs (12:0 + 14:0 = 1.4% total energy) and short-to-medium–chain SFAs (4:0 − 10:0 = 0.5% total energy). The correlations among the individual SFAs can be found in Supplemental Table 5.

There was no significant association between intake of individual SFAs and risk of incident total AF, whereas the association between individual SFAs and sustained AF varied by chain length (Table 4). Higher intake of 4:0 − 10:0, 12:0 + 14:0, and 16:0 were associated with greater risk of sustained AF, and this association differed significantly from that observed with paroxysmal AF (P-difference < 0.01). In contrast, intake of 18:0 was not associated with risk of either paroxysmal (P-trend = 0.10) or sustained (P-trend = 0.54) AF.

TABLE 4.

Multivariable RR of total, paroxysmal and persistent/chronic AF by quintiles of short, medium and long-chain SFA intake among 33,665 women from the Women’s Health Study¹

	Quintiles of FAs
	Q1	Q2	Q3	Q4	Q5	P-trend	P-difference2
Median of 4:0–10:0, % energy	0.23	0.35	0.4	0.56	0.77
Range of 4:0–10:0, % energy	<0.30	0.30–0.39	0.40–0.49	0.50–0.63	>0.63
Total, n	6633	6687	6795	6826	6724
Total AF
Cases, n	287	279	287	293	295
Multivariable model3	1.0 (ref)	0.96 (0.81, 1.13)	0.96 (0.81, 1.14)	1.00 (0.84, 1.18)	1.03 (0.86, 1.22)	0.71
Multivariable model3 + Intermediary factors4	1.0 (ref)	0.96 (0.81, 1.14)	0.97 (0.82, 1.15)	1.01 (0.85, 1.20)	1.06 (0.89, 1.26)	0.47
Paroxysmal AF
Cases, n	197	193	190	182	167
Multivariable model3	1.0 (ref)	0.94 (0.76, 1.15)	0.92 (0.75, 1.13)	0.88 (0.71, 1.08)	0.84 (0.67, 1.04)	0.08
Multivariable model3 + Intermediary factors4	1.0 (ref)	0.95 (0.78, 1.16)	0.93 (0.76, 1.14)	0.90 (0.73, 1.11)	0.86 (0.69, 1.07)	0.13
Persistent/chronic AF							< 0.001
Cases, n	80	79	88	104	116
Multivariable model3	1.0 (ref)	1.05 (0.76, 1.45)	1.09 (0.79, 1.50)	1.37 (1.00, 1.86)	1.58 (1.16, 2.16)	0.001
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.04 (0.75, 1.42)	1.12 (0.81, 1.53)	1.36 (1.00, 1.85)	1.59 (1.17, 2.16)	0.001
Median of 12:0 + 14:0, % energy	0.83	1.12	1.35	1.6	2.11
Range of 12:0 + 14:0, % energy	<0.99	0.99–1.23	1.24–1.46	1.47–1.79	>1.79
Total, n	6705	6713	6798	6781	6668
Total AF
Cases, n	281	298	283	293	286
Multivariable model3	1.0 (ref)	1.08 (0.91, 1.28)	1.02 (0.86, 1.22)	1.12 (0.93, 1.34)	1.10 (0.91, 1.33)	0.48
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.08 (0.91, 1.28)	1.05 (0.88, 1.25)	1.12 (0.94, 1.34)	1.14 (0.94, 1.37)	0.29
Paroxysmal AF
Cases, n	197	196	177	189	170
Multivariable model3	1.0 (ref)	0.98 (0.80, 1.21)	0.89 (0.71, 1.10)	0.99 (0.79, 1.23)	0.89 (0.71, 1.13)	0.33
Multivariable model3 + Intermediary factors4	1.0 (ref)	0.99 (0.81, 1.22)	0.90 (0.73, 1.12)	1.00 (0.80, 1.24)	0.93 (0.73, 1.17)	0.49
Persistent/chronic AF							0.008
Cases, n	78	92	97	97	103
Multivariable model3	1.0 (ref)	1.31 (0.95, 1.79)	1.37 (0.99, 1.89)	1.49 (1.07, 2.08)	1.65 (1.18, 2.32)	0.01
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.29 (0.94, 1.77)	1.41 (1.02, 1.93)	1.47 (1.06, 2.04)	1.67 (1.19, 2.33)	0.008
Median of 16:0, % energy	4.8	5.9	6.7	7.5	8.8
Range of 16:0, % energy	<5.5	5.5–6.3	6.4–7.0	7.1–8.0	>8.0
Total, n	6693	6815	6767	6703	6687
Total AF
Cases, n	280	300	293	287	281
Multivariable model3	1.0 (ref)	1.10 (0.91, 1.34)	1.11 (0.89, 1.39)	1.13 (0.88, 1.45)	1.05 (0.79, 1.39)	0.85
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.11 (0.92, 1.35)	1.13 (0.90, 1.42)	1.16 (0.91, 1.49)	1.12 (0.84, 1.48)	0.79
Paroxysmal AF
Cases, n	181	208	182	185	173
Multivariable model3	1.0 (ref)	1.13 (0.89, 1.43)	0.99 (0.75, 1.31)	1.01 (0.74, 1.38)	0.88 (0.62, 1.25)	0.25
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.14 (0.90, 1.45)	1.00 (0.76, 1.32)	1.03 (0.76, 1.41)	0.93 (0.66, 1.32)	0.41
Persistent/chronic AF							0.03
Cases, n	95	83	96	94	99
Multivariable model3	1.0 (ref)	0.98 (0.69, 1.40)	1.28 (0.86, 1.90)	1.47 (0.95, 2.29)	1.62 (0.98, 2.65)	0.08
Multivariable model3 + Intermediary factors4	1.0 (ref)	0.99 (0.70, 1.39)	1.31 (0.88, 1.93)	1.48 (0.95, 2.29)	1.69 (1.04, 2.75)	0.05
Median of 18:0, % energy	1.5	2	2.4	2.7	3.3
Range of 18:0, % energy	<1.8	1.8–2.1	2.2–2.5	5.6–3.0	>3.0
Total, n	6718	6818	6731	6750	6648
Total AF
Cases, n	288	304	281	307	261
Multivariable model3	1.0 (ref)	1.08 (0.89, 1.31)	1.03 (0.82, 1.29)	1.12 (0.88, 1.44)	0.90 (0.67, 1.20)	0.27
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.09 (0.90, 1.32)	1.05 (0.84, 1.31)	1.14 (0.89, 1.47)	0.95 (0.71, 1.26)	0.44
Paroxysmal AF
Cases, n	191	201	174	201	162
Multivariable model3	1.0 (ref)	1.02 (0.81, 1.30)	0.89 (0.67, 1.18)	1.00 (0.74, 1.36)	0.76 (0.53, 1.09)	0.1
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.03 (0.82, 1.31)	0.89 (0.68, 1.18)	1.01 (0.74, 1.37)	0.79 (0.55, 1.12)	0.14
Persistent/chronic AF							0.15
Cases, n	92	94	97	93	91
Multivariable model3	1.0 (ref)	1.15 (0.82, 1.62)	1.31 (0.88, 1.96)	1.36 (0.87, 2.13)	1.29 (0.78, 2.15)	0.54
Multivariable model3 + Intermediary factors4	1.0 (ref)	1.16 (0.83, 1.62)	1.34 (0.91, 1.99)	1.38 (0.89, 2.14)	1.39 (0.84, 2.28)	0.38

¹AF, atrial fribrillation; Q, quintile; ref, reference.

²P value for the difference in the association between paroxysmal and sustained types of AF.

³Multivariable model includes MUFAs, total PUFAs, and trans fat in the same model, and additionally is adjusted for age, protein (percentage of energy), total calories, smoking, BMI, height, alcohol, exercise, education, race, randomization groups (β-carotene, vitamin E, and aspirin), systolic blood pressure, and diagnosis of hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), and diabetes, all at baseline.

⁴Intermediary factors include diagnosis of incident hypertension (SBP ≥140 and DBP ≥90 mm Hg or use of blood pressure–lowering medication), high cholesterol (total cholesterol ≥240 mg/dL or use of cholesterol-lowering medication), diabetes, cardiovascular disease, and congestive heart failure during follow-up.

Discussion

In this prospective study, dietary fat was not significantly associated with risk of incident AF in women without established CVD or AF at baseline. However, SFAs and MUFAs were associated with the development of sustained forms, but not paroxysmal forms of AF, and these associations were independent of clinical risk factors for AF. Specifically, high intake of SFAs and low intake of MUFAs was associated with greater risk of persistent or chronic AF within 2 y after initial AF diagnosis, suggesting that these fats may impact the maintenance rather than initiation of AF. Trans fats and PUFAs were not significantly associated with risk of total, paroxysmal, or sustained AF.

In this study, high intake of MUFAs was associated with lower risk of sustained AF, but not paroxysmal AF. Previous studies have not reported on MUFA intake and risk of AF directly; however, a Mediterranean diet, in which MUFAs often predominates, may be protective against AF. In a small case-control study in Italy, adherence to a Mediterranean diet was lower in AF patients than in healthy controls (31). Furthermore, in the aforementioned PREDIMED study, a Mediterranean diet arm enriched with extra-virgin olive oil significantly reduced the risk of AF (RR: 0.62; 95% CI: 0.45–0.85) compared with a low-fat diet (20). In Mediterranean diet arm enriched with mixed nuts, the reduction in AF risk was not significant (RR vs. low-fat control, 0.89; 95% CI: 0.65–1.20). MUFAs intake increased more in the group with olive oil (+3.1% of energy) than in that with mixed nuts (+1.9% of energy) (32). Therefore, in context with findings from the current study, olive oil may reduce the risk of sustained forms of AF through its MUFA content, although other bioactive compounds in olive oil cannot be excluded (33).

The inverse association between MUFAs and AF was not explained by higher scores on the Mediterranean diet score. However, women with greater adherence to the Mediterranean diet in our study population did not have greater intake of MUFAs which may explain inconsistencies with the previous studies. Adherence to the Mediterranean diet in US populations such as the WHS often does not reflect the traditional Mediterranean diet in other parts of the world or the dietary pattern tested in the PREDIMED trial. Additional research in populations that more closely follow a traditional Mediterranean diet will be needed to assess the impact of a Mediterranean-style diet on AF incidence and persistence.

Greater intake of total SFAs was associated with greater risk of sustained, but not paroxysmal, AF. This significant positive association was consistent across individual SFAs ranging from 4 to 16 carbons; however, intake of 18:0 (stearic acid) was not associated with AF risk. Individual SFAs have different biological effects (25, 26, 34). In the Cardiovascular Health Study, a higher plasma concentration of 16:0 was associated with greater risk, whereas longer-chain saturates, including plasma 18:0, 20:0, 22:0, and 24:0, were associated with a lower risk of AF (21). Although circulating concentrations of endogenously metabolized fats, such as SFAs are not reflective of dietary intake, these studies suggest that individual SFAs may have different associations with AF risk.

The significant relations between SFAs and MUFAs and AF were restricted to sustained forms of AF; therefore, long-term intake of these fat subclasses may influence pathways related to AF maintenance rather than initiation. AF leads to electrical and structural remodeling, and self-promoting reentrant circuits may develop and foster the maintenance of AF (35, 36). SFAs may promote cardiac structural remodeling through increased apoptosis in myocytes (37) and electrical remodeling through direct proarrhythmic effects (38). Additionally, diets high in SFAs may increase blood pressure, whereas diets high in MUFAs may reduce blood pressure (39).

Although numerous studies have reported the influence of n–3 PUFAs on structural (40) and electrical remodeling (41), n–3 PUFA intake was not associated with risk of AF in the current study, which is consistent with previous observational studies (12–16). Our data suggested a potential, but not significant (P-trend = 0.07), inverse association between ALA and risk of total AF and sustained AF. In previous studies conducted in older adults and middle-aged men, ALA was not associated with incident AF (17, 19). Therefore, additional studies are needed to elucidate the role of ALA in AF prevention, particularly sustained forms of AF.

Although AF remains paroxysmal for the majority of patients (42, 43), sustained forms of AF are associated with more severe complications and poorer prognoses. Prevention of AF progression could lead to a substantial reduction in AF-related complications. Current known predictors of AF progression include age, hypertension, stroke, chronic obstructive pulmonary disease, heart failure, obesity, and elevated glycated hemoglobin values (4, 42, 44). These findings for dietary fat are novel and provide a potential modifiable lifestyle strategy for the prevention of AF progression. Furthermore, these associations were stronger when we included a long-term, rather than a more recent measure of diet, highlighting the fact that dietary fat may modify risk during earlier stages in the development of the substrate for AF. Additional work is needed to understand the role of dietary fat modification in the long-term prognosis of AF patients.

Strengths of the present study include the prospective design with long follow-up, the large number of adjudicated AF cases, and the classification of AF pattern based on accepted clinical guidelines (8). However, important study limitations require discussion. We likely missed asymptomatic episodes of AF in this free-living population. Furthermore, misclassification of AF pattern may have occurred because of the lack of continuous electrocardiogram monitoring and restriction of AF patterns to within 2 y of diagnosis. Self-reported data are likely measured with some degree of error, although such error leads to nondifferential misclassification, which would underestimate the true association. Moreover, these female health professionals are knowledgeable about health-related topics, and data were collected using validated questionnaires (23, 45). The generalizability of results from this population of primarily Caucasian, middle-aged, female health professionals to other populations is limited, including to populations with greater n–3 PUFA intake (16). However, this homogeneity reduces potential confounding by factors that may associate with healthy diet and lifestyle choices. Nevertheless, we cannot exclude the potential for residual confounding by factors we were unable to adjust for, such as sleep apnea, thyroid disease, or family history of AF. We simulated the “substitution” of carbohydrates with intake of FAs. To estimate the true effect of an isocaloric substitution of SFAs or MUFAs for carbohydrates, a feeding study is required. Furthermore, because of the simultaneous changes in fat and carbohydrate intake in our substitution models, we cannot rule out the potential for a carbohydrate effect on AF risk. Finally, we cannot rule out that nominally significant findings may be due to chance in the setting of multiple comparisons.

In conclusion, FA intake was not associated with incident AF, but intake of SFAs was positively and MUFAs was inversely associated with risk of persistent/chronic AF, but not paroxysmal AF. Improving dietary fat quality may play a role in the prevention of sustained forms of AF, which are often less amenable to treatment and associated with higher rates of morbidity.