Eggs, Dietary Choline, and Nonalcoholic Fatty Liver Disease in the Framingham Heart Study

Abstract

Background

Eggs are rich in bioactive compounds, including choline and carotenoids that may benefit cardiometabolic outcomes. However, little is known about their relationship with nonalcoholic fatty liver disease (NAFLD).

Objectives

We investigated the association between intakes of eggs and selected egg-rich nutrients (choline, lutein, and zeaxanthin) and NAFLD risk and changes in liver fat over ∼6 y of follow-up in the Framingham Offspring and Third Generation cohorts.

Methods

On 2 separate occasions (2002–2005 and 2008–2011), liver fat was assessed using a computed tomography scan to estimate the average liver fat attenuation relative to a control phantom to create the liver phantom ratio (LPR). In 2008–2011, cases of incident NAFLD were identified as an LPR ≤0.33 in the absence of heavy alcohol use, after excluding prevalent NAFLD (LPR ≤0.33) in 2002–2005. Food frequency questionnaires were used to estimate egg intakes (classified as <1, 1, and ≥2 per week), dietary choline (adjusted for body weight using the residual method), and the combined intakes of lutein and zeaxanthin. Multivariable modified Poisson regression and general linear models were used to compute incident risk ratios (RR) of NAFLD and adjusted mean annualized liver fat change.

Results

NAFLD cumulative incidence was 19% among a total of 1414 participants. We observed no associations between egg intake or the combined intakes of lutein and zeaxanthin with an incident NAFLD risk or liver fat change. Other diet and cardiometabolic risk factors did not modify the association between egg intake and NAFLD risk. However, dietary choline intakes were inversely associated with NAFLD risk (RR for tertile 3 compared with tertile 1: 0.69, 95% CI: 0.51, 0.94).

Conclusions

Although egg intake was not directly associated with NAFLD risk, eggs are a major source of dietary choline, which was strongly inversely associated with NAFLD risk in this community-based cohort.

Introduction

The egg is a nutrient-dense food that is an important source of dietary choline, lutein, zeaxanthin, and other beneficial nutrients [1]. However, eggs have been historically considered less healthy due to their dietary cholesterol content, which was thought to adversely impact blood lipids. Lipid-related disorders are not only associated with cardiovascular disease risk but also with increased risks of other disorders, such as nonalcoholic fatty liver disease (NAFLD). In recent years, the American Heart Association [2] and the Dietary Guidelines for Americans [3] no longer provide specific limits on dietary cholesterol intake and recommend egg consumption within the context of heart-healthy patterns.

Evidence on the association between eggs or egg-rich nutrients and liver fat in humans is limited and inconsistent [4,5]. This inconsistency may be partly explained by the fact that eggs are typically consumed with other foods (e.g., red and processed meats, fiber-rich foods) that may have different effects on mechanisms underpinning liver fat accumulation such as blood lipids, diabetes, and inflammation-related pathways [[6], [7], [8]].

Egg yolks are rich in dietary choline, an essential nutrient that is required for normal liver and cognitive function and the synthesis of phospholipids [9]. Choline deficiency has been shown to lead rapidly to fatty liver [10] through several mechanisms [11], and conversely, small human studies have shown that hepatic steatosis could be reversed with intravenous choline supplementation [12]. The evidence on whether habitual choline intake prevents NAFLD in larger observational studies has been limited [13,14]. There has also been insufficient evidence to determine a Recommended Dietary Allowance for choline; thus, the Food and Nutrition Board established a sex-specific Adequate Intake (AI) concentration of 425 mg/d for adult females and 550 mg/d for adult males based on the dose that prevented increases in liver function biomarkers in healthy males (with the AI for women being extrapolated from data among males) [15].

Eggs also contain 2 types of carotenoids—lutein and the lutein isomer, zeaxanthin—although the content of these carotenoids in eggs is not as high as that found in other food sources such as cruciferous vegetables [16]. However, the bioavailability of lutein and zeaxanthin from eggs is higher than that from vegetable sources, most likely due to their fat content [16]. Carotenoids are not endogenously synthesized and must be obtained through diet to meet requirements. As with all carotenoids, lutein and zeaxanthin act as antioxidants, which could aid in the prevention of NAFLD [17]. Previous cross-sectional studies have reported an inverse association between dietary intakes and serum concentrations of lutein and zeaxanthin and NAFLD prevalence [[18], [19], [20]], but prospective studies are lacking.

Given the paucity of data, we aimed to examine the prospective association between intakes of eggs and selected egg-rich nutrients (choline, lutein, and zeaxanthin) and NAFLD risk as well as changes in liver fat in the Framingham Heart Study (FHS) cohorts. We also evaluated the independent and combined associations of eggs or choline with other eating patterns and other cardiometabolic risk factors on NAFLD risk.

Methods

Study population

The FHS is a community-based study initiated in 1948 with the FHS Original cohort of 5209 participants. In 1971, 5124 offspring of the original cohort and their spouses enrolled in the Offspring Study. The Third Generation cohort, children of the Offspring cohort participants, was enrolled in 2002 with 4095 participants. In addition, 2 racially diverse cohorts whose participants lived in the Framingham area but who were unrelated to the Framingham participants were enrolled in the Omni 1 (n = 506) and Omni 2 (n = 410) cohorts in 1994 and 2003, respectively.

A subgroup of participants in the aforementioned 4 cohorts underwent a multidetector computed tomography (CT) scan for the assessment of coronary and aortic calcium; liver fat measures were coincidentally available from the scan [21,22]. Table 1 shows the time period of data collection for these analyses. Two CT scans were collected in the Offspring and Third Generation cohorts (2002–2005 and 2008–2011), and 1 scan each in the 2 Omni cohorts (2008–2011). For these analyses, we used the dietary intake derived from food frequency questionnaires (FFQ) administered at exam 7 (1998–2001) in the Offspring cohort and exam 1 (2002–2005) in the Third Generation cohort. These exams were considered the baseline exams for the analyses of incident NAFLD and liver fat change. For the analyses of prevalent NAFLD, we used the FFQ administered closest in time to the CT scan as shown in Table 1.

TABLE 1.

Timing of data collection for diet and outcomes used in the current analyses in the Framingham Heart Study

	Diet via FFQ	NAFLD status via CT scans
Incident NAFLD and liver fat change analyses
Offspring Cohort	ex7 (1998–2001)	1st scan: 2002–2005
		2nd scan: 2008–2011
Third Generation Cohort	ex1 (2002–2005)	1st scan: 2002–2005
		2nd scan: 2008–2011
Prevalent NAFLD analyses
Offspring Cohort	ex8 (2005–2008)	2nd scan: 2008–2011
Omni Cohort 1	ex3 (2007–2008)	2nd scan: 2008–2011
Third Generation Cohort	ex2 (2008–2011)	2nd scan: 2008–2011
Omni Cohort 2	ex2 (2009–2011)	2nd scan: 2008–2011

CT, computed tomography; ex, exam; FFQ, food frequency questionnaire; NAFLD, nonalcoholic fatty liver disease

Figure 1 shows the inclusion and exclusion criteria for the analyses of incident NAFLD and liver fat change. Of the 9219 Offspring and Third Generation participants, 7634 attended the designated baseline exam (exam 7 for Offspring; exam 1 for Third Generation). We excluded participants with missing (missing ≥12 FFQ food items) or/and invalid baseline dietary data (<600 kcals/d or >4000 kcals/d for females, <600 kcals/d or >4200 kcals/d for males) (n = 965), those missing 1 or both CT scans (n = 4783), and those with excessive alcohol use as >14 drinks/wk for females and >21 drinks/wk for males at baseline and next subsequent exam (n = 155), leaving an eligible sample of 1731 participants. Finally, we excluded 24 participants with extreme or missing intakes for important covariates as follows: extreme intakes (>99.9% percentile) on choline intake (n = 1), red meat (n = 2), vitamin B12 (n = 5), vitamin B6 (n = 3) or betaine (n = 1), and missing information on waist-to-height ratio (n = 9) or type 2 diabetes mellitus or impaired fasting glucose (T2DM/IFG) (n = 3). For the analysis of incident NAFLD, we additionally excluded 293 prevalent cases, yielding a final sample of 1414 participants. For the analyses related to change in liver fat, we excluded 17 participants whose liver phantom ratio (LPR) was below the first percentile, yielding a sample size of 1690 participants for these analyses.

FIGURE 1.

Flowchart of study participants for the analyses related to liver fat changes and incident NAFLD risk. Abbreviations: CT, computed tomography; FFQ, food frequency questionnaire; HTN, hypertension; T2DM/IFG, type 2 diabetes mellitus or impaired fasting glucose, and TEE, total energy expenditure. ∗These criteria were developed by FHS investigators.

For the analyses of prevalent NAFLD (Supplementary Figure 1), we included participants from all 4 cohorts (Offspring, Omni 1, Third Generation, and Omni 2) and used the same exclusion criteria as outlined above, yielding a final sample of 2644. The Boston Medical Center and Boston University Medical Campus Institutional Review Board (protocol code H-43060) approved this study.

Dietary assessment

Diet was assessed using a self-administered 126-item semiquantitative FFQ [23]. Information on egg intake per week was derived from 2 questions—one for consumption of whole eggs and a second for scrambled egg consumption. Choline and combined lutein and zeaxanthin contents from foods were estimated as previously described [24,25]. Total choline intake (mg) was calculated as the sum of choline-contributing compounds (phosphocholine, sphingomyelin, free choline, glycerophosphocholine, and phosphatidylcholine) from foods and supplements. Total lutein and zeaxanthin intakes (mcg) were calculated as the sum of dietary and supplement intakes.

Outcome assessment

Our primary endpoints were cumulative incidence of NAFLD and liver fat change; a secondary endpoint was prevalent NAFLD. Protocols for measuring liver fat have been described in detail previously [21,22]. Participants underwent a multidetector CT scan (LightSpeed Ultra; General Electric Health Care). Liver fat was quantified as the mean Hounsfield units (HU) for 3 regions in the liver and one from a radiographic phantom (for calibration control). The LPR was calculated by dividing the mean liver attenuation (HU) by the referent phantom (HU), multiplied by 100. A lower LPR value reflects more liver fat. We calculated the annualized difference in the LPR between the 2 CT scans by subtracting the initial LPR measure (2002–2005) from the follow-up measure (2008–2011), divided by the time between the 2 CT scan dates for each participant. Here, an increase in the LPR reflects a decrease in liver fat. Lastly, NAFLD was defined as LPR ≤33.0 [21,26]. Cumulative incidence was determined from the 2008–2011 CT scan data after excluding participants with LPR ≤33.0 at the initial CT scan. Prevalent NAFLD was determined from the 2008–2011 CT scan.

Assessment of potential confounding and effect modification

We assessed a wide range of potential confounding variables at baseline, including age, physical activity, current cigarette smoking (yes/no), alcohol intake (grams per day and number of drinks per day), energy intake (kilocalories per day), food groups, macronutrient intakes including types and amounts of protein and fat, fiber, various vitamins B, total energy expenditure (TEE), multivitamin intake, use of lipid- or glucose-lowering medications, and prevalent cardiometabolic disorders.

The physical activity index [metabolic equivalents (Mets) per hour] was the sum of hours of activity level multiplied by a numerical weight for oxygen consumption for each activity level, as previously described [27]. We determined both a moderate to vigorous physical activity index and a total physical activity index. TEE was calculated based on sex, age, weight, and total physical activity level [28]. Alcohol servings per day were calculated from self-reported intakes of total grams of alcohol from data collected at baseline and subsequent follow-up exams.

The following cardiometabolic risk factors at baseline were explored as potential confounding variables: 8-h fasting triglycerides (TG), HDL [29], and glucose; prevalent T2DM; impaired fasting glucose (IFG) (fasting glucose concentration of 100–125 mg/dL or a nonfasting glucose ≥126 mg/dL) [30]; hypertension (HTN) (30); measures of body fat (BMI and waist-to-height ratio); elevated sex-specific TG:HDL ratio (>3.8 for males, >3.0 for females), and the use of lipid-lowering drugs [31]. Because of natural height loss after the age of 60 y, height was averaged from all available exams at which adults were <60 y of age. For participants aged ≥60 y, we used baseline height in the calculation of BMI and waist-to-height ratio. In these analyses, we substituted for missing information on alcohol, physical activity, fasting TGs, and HDL using the mean of nonmissing data from adjacent exams.

Statistical analysis

Due to the limited intake of eggs during the time in which these data were collected, egg consumption was classified as <1, 1, and ≥2 eggs per week. The intake of dietary choline was adjusted for the participant’s body weight using the residual method. Linear regression models with absolute nutrient intake (grams per day) as the dependent variable and body weight at baseline (at the time of dietary assessment) as the independent variable were used to compute the residuals. Further, we added the residuals from the regression models to the overall median intake value to derive the adjusted choline intakes by body weight. Choline, lutein, and zeaxanthin intakes were classified into tertiles.

Cumulative incidence for NAFLD was calculated as the number of new NAFLD cases occurring during the follow-up period (median of 6 y). We used multivariable modified Poisson regression models to derive incident NAFLD risk ratios (RR) and 95% confidence intervals (CI) associated with intakes of eggs, choline, lutein, and zeaxanthin. We also used multivariable general linear models to calculate adjusted mean levels for annualized liver fat change over 6 y of median follow-up. In secondary analyses, we used logistic regression models to estimate prevalence odds ratio (OR) and 95% CI associated with egg intakes. Finally, we carried out sensitivity analyses after restricting participants to those ages 40–79 y (the overlapping ages in all 4 cohorts). The median intake value within each category for eggs or nutrients was used in each model to compute P values for linear trends.

Confounding was assessed by adding each factor 1 at a time to the age- and sex-adjusted models, then building the model forward by adding each individual confounder singly or together to the model while avoiding collinearity. We retained those variables that altered the age and sex parameter estimate by ∼5% or more. Models for the analyses on egg intakes were adjusted for age, sex, TEE, education level, red meat intake, baseline BMI, T2DM/IFG, and prevalent HTN. Models for choline, lutein, and zeaxanthin were adjusted for age, sex, education level, and baseline waist-to-height ratio. Because of the interaction between choline and other nutrients [32] involved in 1-carbon metabolism, which modulates hepatic inflammatory signaling pathways and gene expression in hepatic steatosis [33], choline-related analyses were additionally adjusted for the intake of betaine, folate, vitamin B12, and vitamin B6. All models for analyses related to liver fat change were additionally adjusted for baseline LPR.

Lastly, we evaluated whether the association between eggs and NAFLD risk were modified by other dietary habits or other cardiometabolic risk factors, including baseline BMI, prevalent T2DM/IFG, or dyslipidemia as measured by the TG:HDL ratio. We repeated similar effect modification analyses for choline intakes. To assess biological interaction on an additive rather than multiplicative scale [34,35], we used Poisson regression models on a ratio scale as previously described [36], allowing us to estimate relative excess risk due to interaction (RERI) for preventive exposures [34]. We calculated the RERI using the regression coefficients and covariance matrix obtained from multivariable modified Poisson regression analyses. The 95% CI for the RERI was calculated using the MOVER statement (method of variance estimates recovery) [37] using SAS statistical software (version 9.4; SAS Institute).

Results

The descriptive characteristics of the study participants at baseline across egg and choline intake categories are shown in Table 2. Participants consuming ≥2 eggs per week and those with high choline intakes (tertile 3) were less frequently female and more likely to hold a college degree compared to those with low intakes. Further, higher egg consumption but not higher choline intakes were positively associated with prevalent HTN, T2DM/IFG, and adiposity at baseline. However, both eggs and dietary choline were inversely associated with the TG:HDL ratio. In terms of diet, participants consuming ≥2 eggs per week had a slightly higher TEE and slightly higher intakes of vegetables, dairy, and red and processed meat compared with those consuming less. Higher intakes of dietary choline were also associated with higher intakes of vegetables, dairy, and red and processed meats as well as fruit. Further, higher (compared with lower) intakes of eggs, as well as dietary choline, were associated with higher intakes of most nutrients. Descriptive characteristics according to tertiles of lutein and zeaxanthin are shown in Supplementary Table 1. As expected, intakes of lutein and zeaxanthin were positively associated with consumption of plant-based foods (such as whole grains, fruits, vegetables, legumes, nuts, and seeds).

TABLE 2.

Adjusted baseline characteristics of the participants included in the incident analyses across intake categories of eggs and choline in a combined sample from Offspring and Third Generation cohorts

	Egg intake (weekly)			Dietary choline (weight-adjusted mg/day)
	<1	1	≥2	Tertile 1	Tertile 2	Tertile 3
	n = 553	n = 368	n = 493	n = 459	n = 476	n = 479
	%			%
Females	58.6	52.2	46.3	56.6	55.9	45.5
College graduate	44.3	48.9	53.8	38.8	47.3	59.9
Current smokers	8.3	9.0	7.1	9.2	8.0	7.1
Multivitamin users	49.8	52.5	53.3	48.6	53.8	52.6
Prevalent hypertension	23.0	23.1	26.6	25.5	26.3	21.1
Prevalent T2DM / IFG	23.0	28.8	32.1	28.1	26.9	27.9
Elevated TG:HDL ratio	29.5	28.0	27.2	31.6	27.3	26.1
	Means (SE)1			Means (SE)1
Age (y)	51.0 (0.4)	50.8 (0.5)	51.1 (0.4)	52.6 (0.5)	51.0 (0.5)	49.4 (0.5)
Physical activity (METs/h)	14.7 (0.4)	13.8 (0.5)	15.2 (0.5)	13.8 (0.5)	14.4 (0.5)	15.8 (0.5)
BMI (kg/m2)	26.4 (0.2)	26.7 (0.2)	27.7 (0.2)	27.3 (0.2)	27.1 (0.2)	26.4 (0.2)
Waist-to-height ratio	0.552 (0.003)	0.557 (0.004)	0.571 (0.003)	0.566 (0.004)	0.562 (0.003)	0.551 (0.003)
TEE	2570 (23)	2611 (28)	2721 (24)	2581 (25)	2611 (25)	2706 (25)
Foods, srvgs/d	Median (IQR)			Median (IQR)
Eggs per week	0.5 (0.0–0.5)	1.0 (1.0–1.0)	3.0 (3.0–3.0)	0.5 (0.5–1.0)	1.0 (0.5–3.0)	3.0 (1.0–3.0)
Fruits	1.9 (1.1–2.9)	1.9 (1.1–2.8)	2.0 (1.2–2.9)	1.5 (0.8–2.4)	1.9 (1.2–2.7)	2.5 (1.5–3.7)
Vegetables	3.3 (2.4–4.4)	3.3 (2.4–4.4)	3.8 (2.7–5.3)	2.6 (1.9–3.4)	3.5 (2.6–4.4)	4.6 (3.4–6.3)
Dairy	1.6 (1.0–2.5)	1.7 (1.2–2.6)	1.9 (1.3–2.9)	1.2 (0.8–1.7)	1.7 (1.2–2.4)	2.6 (1.6–3.7)
Red and processed meat	0.5 (0.3–0.8)	0.6 (0.4–0.9)	0.9 (0.5–1.3)	0.4 (0.2–0.6)	0.6 (0.4–1.0)	1.0 (0.6–1.4)
Body weight adjusted nutrients/d
Carbohydrates (g)	219 (167–278)	229 (175–286)	245 (184–311)	177 (140–214)	224 (186–269)	298 (250–356)
Protein (g)	76 (60–96)	84 (69–101)	95 (76–116)	60 (52–68)	84 (76–93)	113 (102–129)
Total fat (g)	59 (44–76)	65 (52–82)	76 (58–97)	48 (37–59)	66 (54–79)	89 (73–108)
Saturated fat (g)	20 (15–27)	22 (18–29)	27 (20–34)	17 (13–20)	23 (19–27)	31 (25–39)
Polyunsaturated fat (g)	11 (8–14)	12 (9–14)	13 (10–17)	9 (7–11)	12 (10–14)	16 (13–19)
Monounsaturated fat (g)	22 (16–28)	24 (19–30)	28 (21–36)	17 (13–22)	24 (20–29)	33 (26–41)
Fiber (g)	17 (13–22)	17 (14–22)	19 (15–25)	13 (10–17)	18 (14–22)	24 (19–29)
Choline (mg)	289 (235–358)	323 (262–388)	394 (323–482)	240 (207–264)	330 (309–353)	453 (415–518)
Phosphatidyl choline (mg)	130 (102–168)	157 (126–185)	211 (175–256)	112 (94–131)	166 (146–186)	231 (198–267)
Betaine (mg)	133 (96–198)	151 (103–211)	156 (109–220)	107 (75–156)	150 (106–210)	180 (136–248)
Vitamin B6 (mg)3	2.9 (1.8–4.5)	3.0 (1.9–4.7)	3.2 (2.1–4.7)	2.0 (1.4–3.7)	3.0 (2.0–4.5)	3.9 (2.5–5.4)
Vitamin B12 (mcg) 3	7.6 (4.3–13.5)	9.0 (5.2–14.2)	9.4 (5.9–15.0)	5.4 (3.4–9.9)	8.1 (5.4–12.6)	11.6 (7.5–17.7)
Folate (mcg)3	588 (359–817)	553 (369–871)	630 (418–847)	417 (277–713)	585 (383–816)	754 (518–1009)
Lutein & zeaxanthin (mcg)3	2421 (1552–3543)	2641 (1709–3936)	3227 (2051–4892)	1958 (1196–2758)	2688 (1857–3765)	3948 (2559–5474)
Alcohol (g)4	7.0 (2.6–13.3)	6.2 (2.1–12.5)	8.1 (3.2–16.3)	6.0 (2.0–11.7)	7.1 (2.4–14.0)	9.8 (4.0–16.4)

Abbreviations: BMI, body mass index; d, day; IFG, impaired fasting glucose; IQR, interquartile range; METs, metabolic equivalents; srvgs, servings; T2DM, type 2 diabetes mellitus; TEE, total energy expenditure, and TG:HDL, triglyceride to high density lipoprotein ratio.

² Includes moderate and vigorous physical activity.

¹Means were adjusted for age and sex, except age adjusted for sex only, and TEE not adjusted at all.

³Nutrients were not body weight adjusted.

⁴Among drinkers.

Table 3 shows the overall and sex-specific associations of egg consumption with the cumulative incidence of NAFLD over 6 y of follow-up. After adjusting for age, sex, TEE, education, red meat intakes, baseline BMI, and prevalent T2DM/IFG and HTN, higher egg consumption was not associated with NAFLD risk in females (P-trend = 0.62) or males (P-trend = 0.36). In secondary analyses pooling data from the 4 FHS cohorts (Supplementary Table 2), we also found no association between egg consumption and prevalent NAFLD odds (P-trend = 0.23).

TABLE 3.

Risk ratios for incident NAFLD associated with egg intake categories in females and males in a combined sample from Offspring and Third Generation cohorts

Weekly egg intakes	Cases/n	Cumulative incidence %		RR (95% CI)1	RR (95% CI)2
All participants
<1	92/553	16.6		1.00 (Ref.)	1.00 (Ref.)
1	71/368	19.3		1.13 (0.85–1.49)	1.08 (0.82–1.43)
≥2	101/493	20.5		1.12 (0.87–1.44)	1.00 (0.77–1.30)
P-trend				0.50	0.85
Females
<1	37/324	11.4		1.00 (Ref.)	1.00 (Ref.)
1	28/192	14.6		1.27 (0.80–2.00)	1.34 (0.87–2.08)
≥2	40/228	17.5		1.43 (0.95–2.15)	1.19 (0.78–1.80)
P-trend				0.11	0.62
Males
<1	55/229	24.0		1.00 (Ref.)	1.00 (Ref.)
1	43/176	24.4		1.03 (0.73–1.45)	0.94 (0.66–1.32)
≥2	61/265	23.0		0.94 (0.69–1.29)	0.86 (0.62–1.18)
P-trend				0.62	0.36

Abbreviations: BMI, body mass index; HTN, hypertension; NAFLD, nonalcoholic fatty liver disease; Ref, reference; T2DM/IFG, type 2 diabetes mellitus or impaired fasting glucose, and TEE, total energy expenditure.

¹Model adjusted for age, sex (in all participants model only), and TEE.

²Model additionally adjusted for education level, red meat intakes, baseline BMI, prevalent T2DM/IFG, and HTN.

We also examined the overall and sex-specific associations between selected egg-rich nutrients (choline, lutein, and zeaxanthin) and incident NAFLD risk (Table 4). After adjusting for age, sex, education, and baseline adiposity, participants in tertile 3 of body weight-adjusted choline intakes had a 28% lower NAFLD risk (95% CI: 0.55, 0.94) than those in tertile 1. The sex-stratified results suggest that there was a statistically significant inverse trend among males (P-trend = 0.0048) but not females (P-trend = 0.65). Males with high choline intakes (tertile 3 compared with tertile 1) had a statistically significant 32% lower risk of NAFLD (95% CI: 0.48, 0.95). Additional adjustments for nutrient intakes involved in one-carbon metabolism did not change the results. Finally, in Table 4, we observed no association between lutein and zeaxanthin intakes and incident NAFLD risk.

TABLE 4.

Risk ratios for incident NAFLD associated with egg-rich nutrient intake tertiles in females and males in a combined sample from Offspring and Third Generation cohorts

	Cases/n	Cumulative incidence %		RR (95% CI)1	RR (95% CI)2	RR (95% CI)3
	Body weight-adjusted choline intake (mg)
All participants
T1	104/459	22.7		1.00 (Ref.)	1.00 (Ref.)	1.00 (Ref.)
T2	88/476	18.5		0.83 (0.64–1.07)	0.84 (0.65–1.08)	0.83 (0.63–1.08)
T3	72/479	15.0		0.65 (0.49–0.86)	0.72 (0.55–0.94)	0.69 (0.51–0.94)
P-trend				0.0022	0.0156	0.0189
Females
T1	39/245	15.9		1.00 (Ref.)	1.00 (Ref.)	1.00 (Ref.)
T2	37/249	14.9		0.96 (0.64–1.46)	1.10 (0.74–1.63)	1.07 (0.71–1.60)
T3	29/250	11.6		0.77 (0.49–1.22)	0.86 (0.56–1.31)	0.82 (0.51–1.31)
P-trend				0.27	0.65	0.62
Males
T1	64/217	29.5		1.00 (Ref.)	1.00 (Ref.)	1.00 (Ref.)
T2	54/230	23.5		0.81 (0.59–1.11)	0.79 (0.58–1.09)	0.79 (0.56–1.11)
T3	41/223	18.4		0.64 (0.45–0.91)	0.68 (0.48–0.95)	0.67 (0.46–1.00)
P-trend				0.0018	0.0048	0.0092
	Lutein & zeaxanthin intakes (mcg)
All participants
T1	93/471	19.8		1.00 (Ref.)	1.00 (Ref.)	—
T2	84/451	18.6		0.96 (0.74–1.25)	0.99 (0.75–1.30)
T3	87/492	17.7		0.93 (0.72–1.21)	0.99 (0.76–1.28)
P-trend				0.61	0.93
Females
T1	36/253	14.2		1.00 (Ref.)	1.00 (Ref.)	—
T2	33/239	13.8		1.00 (0.64–1.54)	1.08 (0.71–1.66)
T3	36/252	14.3		1.01 (0.66–1.55)	1.08 (0.71–1.65)
P-trend				0.85	1.00
Males
T1	58/215	27.0		1.00 (Ref.)	1.00 (Ref.)	—
T2	45/213	21.1		0.77 (0.55–1.08)	0.76 (0.53–1.07)
T3	56/242	23.1		0.85 (0.62–1.17)	0.88 (0.64–1.20)
P-trend				0.63	0.83

Abbreviations: NAFLD, nonalcoholic fatty liver disease; Ref, reference, and T, tertile.

¹Model adjusted for age and sex (in all participants model only).

²Model additionally adjusted for education level and baseline waist-to-height ratio.

³Model additionally adjusted for methyl-donor-related nutrient intakes (vitamins B6 and B12, folate, and betaine).

Figure 2 shows the associations between intakes of eggs (Panel A), choline (Panel B) and lutein and zeaxanthin (Panel C) and annualized liver fat changes over the study follow-up in females or males. There were no statistically significant associations between the intakes of eggs, choline, or lutein and zeaxanthin with liver fat change (P > 0.05 for all).

FIGURE 2.

Annualized liver fat changes (SE) associated with egg and egg-rich nutrients in females and males in a combined study sample of Offspring and Third Generation. A: Liver fat changes associated with egg intake. B: Liver fat changes associated with body weight-adjusted choline intakes. C: Liver fat changes associated with lutein and zeaxanthin intakes. Analyses for egg intakes were adjusted for age, sex (in all participants’ model only), TEE, education level, red meat intakes, baseline BMI, prevalent T2DM/IFG, HTN, and baseline LPR. Analyses for egg-rich nutrients were adjusted for age, sex (in all participants’ model only), education level, baseline waist-to-height ratio, and baseline LPR. The scale on the y-axis represents the amount of LPR decline (SE); e.g., 0.001 annualized liver fat increase indicates 0.001 unit decrease in LPR. Abbreviations: HTN, hypertension; LPR, liver phantom ratio; T2DM/IFG, type 2 diabetes or impaired fasting glucose, and TEE, total energy expenditure.

In Figure 3, we examined the independent and combined associations between eggs or and other risk factors with incident NAFLD risk and evaluated effect modification on an additive scale. Overall, we found that the association between egg intake and NAFLD risk was not modified by any of these cardiometabolic risk factors or other dietary factors (P values for the RERI >0.05 for all factors). It was also apparent that egg consumption was not adversely associated with NAFLD risk in any of these models, even among those with higher baseline adiposity (BMI ≥25kg/m², Panel A), dyslipidemia (Panel B) or prevalent T2DM/IFG (Panel C). Further, there was no modification of the association between egg intake and NAFLD risk by other dietary factors (Panels D-F).

FIGURE 3.

Independent and combined associations of egg intake categories with other risk or diet factors on incident NAFLD risk in a combined sample of Offspring and Third Generation cohorts. Each panel indicates a different risk factor as follows: (A) BMI; (B) elevated TG:HDL; (C) T2DM/IFG; (D) fiber intake; (E) dairy intake; and (F) red and processed meat intake. Analyses were adjusted for age, sex, TEE (except in the analyses for BMI), education level, red meat intakes (except in the analyses for red meat), baseline BMI (except in the analyses for BMI, lipids, and diabetes) and prevalent T2DM/IFG (except in the analyses for T2DM/IFG) and hypertension. Additive interaction was estimated through RERI for risk factors, and none reached statistical significance. Abbreviations: NAFLD, nonalcoholic fatty liver disease; RERI, relative excess risk due to interaction; T2DM/IFG, type 2 diabetes mellitus or impaired fasting glucose; TEE, total energy expenditure, and TG:HDL, triglyceride to high density lipoprotein ratio.

The associations of dietary choline and incident NAFLD modified by other risk factors are shown in Figure 4. Here, we found no effect modification on an additive scale by baseline BMI (Panel A), physical activity (Panel B), or other dietary factors (Panel C–E) (P values for RERI >0.05 for all factors). BMI itself was strongly associated with NAFLD risk. Nonoverweight individuals (Panel A) with lower choline intakes had a 65% lower risk (95% CI: 0.23, 0.54) than overweight individuals with lower choline intakes. In contrast, nonoverweight participants with higher choline intakes had an 80% lower NAFLD risk (95% CI: 0.12, 0.34) than overweight individuals with lower choline intakes. Higher choline intake alone (among overweight participants) still had a 21% lower NAFLD risk (95% CI: 0.63, 1.00). Finally, for most other risk factors, those participants with higher intakes of dietary choline and other beneficial risk factors (i.e., higher physical activity, higher fruit and vegetable intakes, and higher dietary fiber intakes) had the lowest NAFLD risk than those in the referent category.

FIGURE 4.

Independent and combined associations of body weight-adjusted choline intake categories with other risk or diet factors on incident NAFLD risk in a combined sample of Offspring and Third Generation cohorts. Each panel indicates a different risk factor as follows: (A) BMI, (B) physical activity, (C) fruit and vegetable intake, (D) fiber intake, and (E) legumes, nuts, and seeds intake. Analyses were adjusted for age, sex, and education level. Analyses for fruits and vegetables, fiber, and legumes, nuts and seeds were additionally adjusted for TEE. Moderate and vigorous physical activity was classified as low (Q1-Q2) and high (Q3-Q5) METs per hour. Additive interaction was estimated through RERI for risk factors, and none reached statistical significance. Abbreviations: MV, moderate to vigorous; METs, metabolic equivalents; NAFLD, nonalcoholic fatty liver disease; RERI, relative excess risk due to interaction; and TEE, total energy expenditure.

Discussion

In current analyses, habitual egg intakes alone or combined with other eating patterns were not associated with either incident or prevalent NAFLD or liver fat changes over a median of 6 y of follow-up. We observed no adverse associations between egg intakes and NAFLD risk, even among those with other cardiometabolic risk factors, including greater adiposity, dyslipidemia, or T2DM/IFG. Lastly, intake of dietary choline, but not lutein and zeaxanthin, was strongly associated with a lower incidence of NAFLD risk, especially in males.

Our findings add to the limited number of epidemiologic [38,39] and clinical [5] studies on the link between egg intake and liver fat. For example, a prior cross-sectional analysis using 14,369 participants from the NHANES showed that higher egg intakes (tertile 3 compared with 1) were positively associated with prevalent NAFLD odds after adjustment for sociodemographic and lifestyle factors. However, additional adjustment for cardiometabolic risk factors attenuated this adverse association toward the null, a finding that is consistent with our own [38]. Similarly, a randomized controlled trial that included participants without diabetes or taking lipid-lowering medication showed that consumption of one extra egg per day was not associated with biomarkers for liver function and inflammation, compared with a habitual egg consumption of 1–2 eggs/wk [5]. In the current analyses, the range of egg intake was limited (median: 1, IQR: 0.5–3 eggs/wk), possibly due to the diet policy in the United States at the time to reduce egg intake out of concerns of dietary cholesterol and as a result this may partially explain the lack of association in the current analyses.

The effects of egg consumption on lipid metabolism—an integral part of NAFLD pathophysiology—have been very mixed in the literature. Two meta-analyses showed that higher egg consumption was associated with an increase in LDL:HDL concentrations [40], especially in longer-duration trials [41]. However, studies also have shown that egg consumption led to a higher concentration of larger LDL and HDL particles, which are considered less prone to oxidation and thus reduce the likelihood of atherosclerosis [[42], [43], [44]]. In addition, egg intakes were found not to be associated with elevated TG or very low density lipoprotein (VLDL) levels and, in fact, some studies have found a beneficial association with plasma apolipoprotein (Apo) A1 (in females), whereas others found an adverse association with ApoB100 levels [41]. Although the evidence on egg consumption and lipid metabolism is inconsistent, cumulative evidence shows that egg intake is not adversely associated with dyslipidemia-related diseases, including cardiovascular disease [2,45] and diabetes risk [46], 2 important comorbidities for NAFLD.

Eggs are a rich source of dietary choline, an essential nutrient for maintaining muscle and liver function [47]. In these analyses, the average choline intake was 331 mg/d (314 mg/d in females and 353 mg/d in males), which is below the AI [47] but similar to previously reported intakes in other US cohorts [13,48]. Here, we showed that dietary choline was inversely associated with a higher risk of new onset of NAFLD. Overall, our findings are in line with the limited number of previous analyses [13,14]. For example, in a cross-sectional analysis of 20,643 participants in NHANES, higher choline intakes were inversely associated with a validated liver index that predicts NAFLD—the fatty liver index (FLI) [13]. Further, in a Chinese cohort, higher choline intakes (mainly derived from eggs and soy products) were also associated with a reduced risk of NAFLD, especially in females [14].

The exact mechanisms of how dietary choline protects against liver fat accumulation are not clear; however, several plausible mechanisms are proposed to affect the pathophysiology of NAFLD [11]. Choline metabolism has 2 major fates: phosphorylation to phosphatidylcholine or oxidation to betaine [11]. Phosphatidylcholine is one of the most abundant phospholipids of mammalian cell membranes, and it is also necessary to form the monolayers of lipoproteins and export TGs through VLDLs from the liver [49]. Lower levels of phosphatidylcholine could lead to increases in the curvature of membranes in the endoplasmic reticulum, which in turn could activate de novo lipogenesis, leading to increased TG synthesis in the liver [50]. Lower phosphatidylcholine levels could also disrupt the export of TGs from the liver via VLDL, leading to hepatic TG accumulation, a central mechanism in hepatic steatosis [51]. Lastly, decreased phosphatidylcholine membrane concentrations secondary to choline-deficient diets could lead to mitochondrial dysfunction through disruption of mitochondrial bioenergetics and fatty acid β-oxidation [52]. Choline’s other fate, oxidation to betaine, is also implicated in the etiology of NAFLD; mutations in genes that convert choline to betaine and other metabolites important to methylation reactions result in fatty liver in animals [53,54] and may be associated with NAFLD in humans [55].

Lastly, dietary choline metabolism has been found to affect the gut microbiome, which plays a key role in the development of NAFLD. A study showed that replenishment of choline deficiency could lead to decreases in the abundance of Gammaproteobacteria species, which are known to affect susceptibility to the development of hepatic steatosis [56]. Further, gut microbes are responsible for metabolizing choline to trimethylamine (TMA), which is further converted to trimethylamine-N-oxide (TMAO) in the liver [57]. TMAO may be a risk factor for NAFLD [58]. However, it is important to note that choline from eggs is mostly present in the form of phosphatidylcholine, which is readily absorbed in the upper gastrointestinal tract, resulting in less available choline in the lower gastrointestinal tract where gut microbes are present to convert it to TMA [57].

In addition to choline, egg yolks contain some amount of 2 important highly bioavailable carotenoids, lutein and zeaxanthin. Generally, there is consensus that diets rich in carotenoids have beneficial effects on hepatic steatosis due mainly to their antioxidant capacity [17]. Studies in animal models of nonalcoholic steatohepatitis (NASH) showed that lutein and zeaxanthin may improve several risk factors of hepatic steatosis, including abdominal fat, hepatic TG content, and hepatic insulin signaling factors [59] or liver fibrosis [60]. Some evidence in humans, although limited, has shown that higher plasma levels or higher dietary intakes of lutein and zeaxanthin were cross-sectionally inversely associated with NAFLD prevalence [19,20]. In contrast, our data failed to find an association between intakes of these 2 carotenoids and liver fat which may be due to the inadequacy of the current FFQ to estimate intakes accurately. Prior validation studies have shown only low to moderate correlations between plasma levels and FFQ intakes of lutein and zeaxanthin in US individuals [61,62].

The current study has some important strengths, starting with its prospective design with a median of 6 y of follow-up (for the incident NAFLD analyses) and the careful assessment of several confounding factors. A limitation of this study is that diet was derived from self-reported questionnaires; thus, we cannot rule out recall bias and measurement error. Energy intake, in particular, is not well measured with FFQ. However, because dietary choline is derived from energy-dense food sources (e.g., meats, fish, eggs, and milk) [63], it is important to consider the need to adjust for some measure of energy intake. Due to the limitations of the FFQ, we chose to adjust our choline intakes for body size (i.e., weight and BMI) using the residual method.

No validation study has yet been done to support the validity of estimated choline intakes from the Harvard FFQ used in the current study against a reference method. However, prior analyses in the Offspring cohort supported the predictive validity of estimated choline intakes from the current FFQ [63].

There are also other limitations and strengths in this study that relate to the assessment of liver fat. Although we adjusted for a variety of factors, we cannot rule out residual confounding by other unmeasured (or imperfectly measured) factors. Although CT scans are better correlated with the gold standard liver biopsy than other methods, they have been previously reported to be insensitive to mild liver steatosis and fibrosis [64]. However, the prevalence of hepatic fibrosis in the Third Generation cohort is low [65] (data on fibrosis in Offspring cohort are not available). Further, the use of available repeated CT-liver fat measures enabled us to calculate cumulative incidence after excluding prevalent NAFLD, minimizing the possibility of reverse causation. We were also able to calculate changes in liver fat over time. Both calculations are more clinically meaningful and advantageous compared with the calculation of NAFLD prevalence at a single time point, which has been used in most prior studies. In the current analyses, we were not able to exclude individuals with secondary causes of liver disease (e.g., hepatitis C, medication-induced hepatitis) due to the lack of availability of the data. Lastly, the FHS consists mainly of White Caucasians, which may compromise the generalizability of these results to other racial populations.

In conclusion, this study suggests that consuming 2 or more eggs per week was not adversely associated with the development of NAFLD over a median of 6 y of follow-up. Further, the intakes of lutein and zeaxanthin were not associated with liver fat change or NAFLD risk. However, dietary choline intake was inversely strongly associated with incident NAFLD risk. Eggs are just single source of dietary choline, and the current results suggest that total choline may be of more importance than choline from eggs alone. Eggs are an inexpensive source of several important nutrients, and the current findings support their inclusion in a healthy diet pattern.

Author contributions

The authors’ responsibilities were as follows – LLM, IY: designed the analysis; IY: analyzed the data; LLM, IY: wrote the manuscript; IY, MTL, PFJ, AB, RTP, LLM: participated in the interpretation of the results and editing of the manuscript; and all authors: read and approved the final manuscript.

Funding

The Framingham Heart Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with Boston University (Contract No. N01-HC-25195, HHSN268201500001I, and 75N92019D00031). Funding to support the Omni cohort recruitment, retention, and examination was provided by NHLBI Contract N01-HC-25195, HHSN268201500001I, and 75N92019D00031, as well as NHLBI grants R01-HL070100, R01-HL076784, R01-HL-49869, and U01-HL-053941. Funding support for the Framingham Food Frequency Questionnaire dataset in the Offspring cohort was provided by ARS Contract #53- 3k06-5-10, ARS Agreement #’s 58-1950-9-001, 58-1950-4-401, and 58-1950-7-707. Funding support for the Framingham Food Frequency Questionnaire dataset in Third Generation was provided by ARS Contract #53- 3k06-5-10, ARS Agreement #’s 58-1950-9-001, 58-1950-4-401, 58-1950-7-707 and 58-1950-0- 014, and NIH grant P01AG031093. This work was additionally supported by the Egg Nutrition Center (grant # 9550305562). The funders had no role in the design, analysis, or writing of this article.

Data availability

Data described in the manuscript and code books will be made available upon request, pending approval by the Framingham executive committee. For additional information please see the Framingham Heart Study website at https://www.framinghamheartstudy.org/

Conflict of interest statement

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: