Diet composition impacts the natural history of steatotic liver disease

Abstract

Background:

Metabolic-associated steatotic liver disease (MASLD), caused by insulin resistance and the metabolic syndrome, may result in progressive liver fibrosis. Animal studies suggest that dietary content modulates liver fibrosis progression. Our aim was to identify dietary components and food-related behaviors that may be associated with fibrosis progression and liver-related outcomes in a well-characterized human MASLD cohort.

Methods:

Patients with MASLD who had completed a detailed Lifestyle Survey, including a semiquantitative Food Frequency Questionnaire in the Veterans Health Administration Million Veteran Program, were included. The primary outcome was liver fibrosis progression using the Fibrosis-4 slope; the secondary outcome was time to cirrhosis by ICD9/10 codes. Key baseline covariates included: race/ethnicity, body mass index, diabetes mellitus, AUDIT-C score, and baseline Fibrosis-4 score. Using bootstrapped Elastic Net regression in R, self-reported food intake and scaled nutrient variables of interest associated with the outcomes were identified and then validated using multivariable Generalized Linear Model and Cox models.

Results:

A total of 84,024 individuals with MASLD with nutritional data were included in this study. Median age at MASLD diagnosis was 56 years (IQR 49–63). Frequency of consumption of coffee, tea, vegetables (broccoli, spinach/collard greens), legumes, nuts, modest alcohol, white meat, rice/pasta, dairy, and intakes of specific nutrients including nitrate/vitamin K, caffeine, betaine, amino acids, and beta carotene were associated with reduced fibrosis progression. Consumption of white bread, cookies, breakfast cereals, and specific nutrients such as iron (non-heme), B vitamins, and flavanones were all significantly associated with increased fibrosis progression in MASLD (p<0.05).

Conclusions:

Dietary choices such as intake of processed foods, high-fructose foods, and refined carbohydrates may be associated with MASLD progression, while intake of vegetables, nuts, whole grains, and caffeine may be protective.

INTRODUCTION

Metabolic–associated steatotic liver disease (MASLD) is the most common cause of chronic liver disease, affecting up to 30% of Americans.¹ A subset of patients with MASLD manifests metabolic-associated steatohepatitis (MASH) and are at risk for progressive fibrosis, cirrhosis, HCC, and hepatic decompensation.

MASLD is mediated by insulin resistance, leading to hepatic lipid accumulation. When accompanied by lipotoxicity with resultant inflammation (MASH), a fibrogenic response may result that is modulated by genetic and environmental cofactors.² The primary therapeutic management of MASLD/MASH focuses on weight loss interventions and improvement of insulin resistance.¹,³ Given the importance of obesity and insulin resistance in pathogenesis, there has been interest in identifying dietary factors that contribute to the incidence and progression of MASLD/MASH.⁴ To date, fructose-rich diets have yielded the strongest evidence.⁵ In animal models, fructose-rich diets increase hepatic triglycerides, cholesterol, as well as macro-vesicular and micro-vesicular fat deposits.⁶ In mouse models, unhealthy diet composition has also been shown to drive progression to MASH and liver cancer in obese mice.⁷ Human data relating diet quality and MASLD progression are sparse, but consumption of fructose has been found to be 2-fold to 3-fold higher in patients with MASLD than controls,⁸ and diets with higher levels of sugars are associated with increased hepatic ALT/AST levels.⁹ There is also limited evidence that the Mediterranean diet and coffee consumption may be protective against fibrosis progression in MASLD.¹⁰,¹¹,¹²

This study utilized the Million Veteran Program (MVP) database, a large biorepository with greater than one million veterans enrolled and initiated by the Veterans Health Administration (VHA) Office of Research and Development, to study dietary intake in relation to MASLD. We previously identified a MASLD cohort using a proxy definition of chronically abnormal ALT values without other known causes of liver disease and defined their clinical trajectory as well as genetic polymorphisms associated with MASLD and fibrosis progression¹³,¹⁴; a large fraction of these individuals completed detailed dietary questionnaires.¹⁵ Our objective in the present study was to evaluate the relationship of self-reported habitual dietary intake with fibrosis progression and liver outcomes among patients with MASLD in MVP.

METHODS

Million Veteran Program

The MVP was launched in 2011 by the VHA Office of Research and Development to develop a genetic repository of US Veterans linked with clinical data and has enrolled more than one million veterans.¹³,¹⁶ The ongoing MVP cohort provided written informed consent to participate in the study as previously described,¹³,¹⁶ resulting in a large biorepository linked to results from self-reported baseline and lifestyle survey questionnaires.¹⁵ This work was conducted under the MVP cardiometabolic project (MVP003/028: “Genetics of Cardiometabolic Diseases in the VA Population”), which is approved by the VA Central Institutional Review Board (IRB).

Definition of MASLD cohort with longitudinal evaluation of fibrosis progression

MVP participants with MASLD were identified using a validated algorithm.¹¹ Briefly, patients were included if they manifested 2 abnormal ALT measurements at least 6 months apart over a 2-year period or had 1 inpatient or 2 outpatient ICD10 diagnoses of MASLD or MASH (K76.0, K75.8).¹⁴ Participants with other causes of chronic liver disease diagnoses such as known alcohol use disorder [as determined by ICD9/10 diagnosis and/or Alcohol Use Disorders Identification Test (AUDIT-C) ≥4], viral hepatitis, genetic liver disease, cholestatic liver disease, or autoimmune hepatitis were excluded (n=181,973) (Supplemental Figure S1, http://links.lww.com/HC9/C29). Follow-up began at the first date meeting, one of these criteria; patients with fewer than 5 years of follow-up and/or fewer than 5 observations were excluded from the cohort. The specific subcohort used was also restricted to patients aged 40–65 years at the age of first available Fibrosis-4 (FIB-4) measurement and baseline FIB-4 score ≤2.67, excluding patients likely to have advanced fibrosis at baseline to allow assessment of progression to fibrosis. Patients with advanced/metastatic cancers (cirrhosis comorbidity score 5+0 or 5+1) (n=254) or a diagnosis of immune thrombocytopenic purpura (ITP) (n=380) were also excluded due to the possible effect of cancer and ITP on outcomes and FIB-4 measurements (Supplemental Figure S1, http://links.lww.com/HC9/C29). Additional clinical data were collected from this cohort, including initial FIB-4 score, demographics, cirrhosis comorbidity indices, component laboratory tests for Model for End-Stage Liver Disease—Sodium (MELDNa) scores (creatinine, bilirubin, international normalized ratio, sodium), lipid profile, serum alpha-fetoprotein, and liver function tests.

Nutritional cohort

MVP enrollees who completed a Lifestyle Survey, including a Semi-quantitative Food Frequency Questionnaire (SFFQ) assessing the habitual consumption of specific foods, added sugar, consumption of fried foods, and nutritional supplements over the prior 12 months, were included in the Nutritional Cohort.¹⁵ The Lifestyle Survey included a validated SFFQ, with dietary characteristics including dairy, meat, fruit, vegetables, sweets/other, alcoholic beverages, etc., on a serving/day basis as shown in Supplemental Table S1, http://links.lww.com/HC9/C29. As shown, for data analysis, SFFQ answers were grouped based on average sample size to “Never or less than once a month,” “1–4/month,” “2–6/week,” and “≥1/day.” Dietary assessment was translated to nutrient data on the macronutrient and micronutrient level using the Harvard Food Composition Database and a previously validated method.¹² The energy-adjusted nutritional variables were used to reduce skewness between variable measurements.¹³ Variables with >95% missingness were excluded from the analysis. To reduce the possibility of collinearity, correlation analysis was conducted using Spearman correlation and any variables with C>0.8 were summed if originating from similar food groups (eg, individual amino acids were summed to form total amino acids). Where applicable, total nutrient variables were used as opposed to individual nutrient subsets (eg, total flavonoids vs. individual flavonoids). The full list of variables included in the analysis is summarized in the supplemental information (Supplemental Material, http://links.lww.com/HC9/C29).

Outcomes

The primary outcome was identified as progression of fibrosis using FIB-4 slope, a recently validated assessment of fibrosis progress that correlates with risk of development of cirrhosis, hepatic decompensation and HCC.¹¹ Secondary outcome was development of cirrhosis which was defined by the presence of at least 1 inpatient or 2 outpatient ICD9/10 codes (572.2, 572.5, K70.3x, K74.6x).

Covariates

Covariates, including age, race/ethnicity, body mass index (BMI), tobacco use, alcohol use by the AUDIT-C questionnaire, metabolic syndrome criteria, diabetes mellitus, liver-related laboratory tests (total bilirubin, AST, ALP), and baseline FIB-4 score were obtained as previously described.¹⁴

Statistical analysis and association of outcomes with nutritional variables

Missing clinical data was imputed using the R package MICE (Multivariate Imputation by Chained Equations) using all available covariates with a random forest algorithm.¹⁴ The dataset was randomly split into 70% training and 30% validation sets using the R “sample”. Nutrient variables of interest were identified via Elastic Net regression using a bootstrap approach on the training set using either FIB-4 slope or time to cirrhosis as the dependent variable using the R glmnet package.¹⁷,¹⁸ For all food and beverage intake frequencies, reference comparisons were made to intake categorized as “Never or less than once a month.” Training-identified variables were then validated using multivariable Generalized Linear Model and/or Cox models, including the key clinical covariates with Benjamini–Hochberg correction for multiple comparison. Key baseline covariates included: race/ethnicity, BMI class (Underweight: <18.5, Ideal: 18.5–24.9, Overweight: 25–29.9, Class 1: 30–34.9, Class 2: 35–39.9, Class 3: >40), metabolic syndrome criteria, diabetes mellitus, AUDIT-C score, liver-related laboratory tests (total bilirubin, AST, ALP), and baseline FIB-4 score. Baseline multivariable Generalized Linear Model and Cox models not including nutritional or food frequency data were constructed (Supplemental Tables S2–S5, http://links.lww.com/HC9/C29). For both the primary and secondary outcomes of fibrosis progression and time to cirrhosis, alpha=0 minimized the mean squared error for the model including nutrient variables, and alpha=0.15 minimized the mean squared error for the food frequency questionnaire model (Supplemental Figures S1, 2, http://links.lww.com/HC9/C29).¹⁵,¹⁶

RESULTS

Cohort characteristics

Within MVP, 98,361 individuals met the MASLD phenotype, with applied exclusion criteria as detailed above, among whom 84,024 had available nutritional data with at least 5 years of follow-up [median 15.4 y (IQR 10.8–18.7)] (Supplemental Figure S1, http://links.lww.com/HC9/C29). Participants with complete nutritional data were slightly older, more likely to report White race, and more likely to be former rather than current smokers than patients who did not complete the LS but were otherwise similar (Supplemental Table S6, http://links.lww.com/HC9/C29). There were no significant differences in smoking status, BMI, AUDIT-C score, FIB-4 scores, or baseline labs (total bilirubin, albumin, international normalized ratio, platelet count) between the testing and training datasets (Table 1). Consistent with the US Veteran population, 90% of participants were male, and 80% were White, 10% Black, and 6% Hispanic. Median age at time of entry into the cohort was 56 years (IQR 49–63 y). In all, 96% of participants were overweight and obese, with a median BMI of 32.9 [IQR 29.6–36.9]. Median baseline FIB-4 was 1.02 [IQR 0.75–1.39]. The median FIB-4 slope was 0.032 units/year [IQR 0.010–0.061] (Supplemental Figure S3, http://links.lww.com/HC9/C29). A total of 1571 (2.7%) individuals in the training dataset developed cirrhosis, and 705 (2.8%) individuals in the validation dataset developed cirrhosis over a median of 15.2 [95% CI: 15.1–1.2] years of follow-up (Supplemental Figure S4, http://links.lww.com/HC9/C29). The LS was completed a median of 10.2 years [IQR 5.6–13.2] after initial identification of MASLD, and the median follow-up time after LS completion was 3.0 years [IQR 5.2–7.1].

TABLE 1.

Cohort characteristics

Variable	Overall	Training	Testing	p
N	84,024	58,848	25,176
Age, median [IQR]	56.0 [49.0, 63.0]	56. 0 [49.0, 63.0]	56.0 [49.0, 63.0]	0.36
Sex=Female, N (%)	8486 (10.1)	5930 (10.1)	2556 (10.2)	0.75
Race/Ethnicity, N (%)				0.70
White	67,101 (79.9)	46,943 (79.8)	20,158 (80.1)
Black	8128 (9.7)	5723 (9.7)	2405 (9.6)
Hispanic	4702 (5.6)	3321 (5.6)	1381 (5.5)
Asian	1041 (1.2)	738 (1.3)	303 (1.2)
Other	3052 (3.6)	2123 (3.6)	929 (3.7)
Tobacco, N (%)				0.60
Never smoker	35,502 (42.3)	24,893 (42.3)	10,609 (42.1)
Former smoker	27,423 (32.6)	19,160 (32.6)	8263 (32.8)
Current smoker	19,950 (23.7)	13,972 (23.7)	5978 (23.7)
Unknown smoking status	1149 (1.4)	823 (1.4)	326 (1.3)
BMI, median [IQR]	32.9 [29.6, 36.9]	32.9 [29.6, 37.0]	32.9 [29.7, 36.9]	0.77
BMI class, N (%)				0.29
Ideal (18.5–24.9)	3172 (3.8)	2266 (3.9)	906 (3.6)
Underweight (<18.5)	458 (0.5)	312 (0.5)	146 (0.6)
Overweight (25–29.9)	40,456 (48.1)	28,382 (48.2)	12,074 (48.0)
Obesity class 1 (30–34.9)	12,702 (15.1)	8822 (15.0)	3880 (15.4)
Obesity class 2 (35–35.9)	16,107 (19.2)	11,268 (19.1)	4839 (19.2)
Obesity class 3 (>40)	11,129 (13.2)	7798 (13.3)	3331 (13.2)
Baseline AUDIT-C, N (%)				0.65
0	78,288 (93.2)	54,793 (93.1)	23,495 (93.3)
1	2845 (3.4)	2021 (3.4)	824 (3.3)
2	1676 (2.0)	1176 (2.0)	500 (2.0)
3	1215 (1.4)	858 (1.5)	357 (1.4)
Total bilirubin mg/dL, median [IQR]	0.60 [0.45, 0.80]	0.60 [0.45, 0.80]	0.60 [0.44, 0.80]	0.26
Serum albumin g/dL, median [IQR]	4.20 [3.97, 4.45]	4.20 [3.98, 4.45]	4.20 [3.95, 4.50]	1.00
INR, median [IQR]	1.00 [0.98, 1.10]	1.00 [0.98, 1.10]	1.00 [0.98, 1.10]	0.04
Platelet count/mm3, median [IQR]	233 [198, 274]	233 [198, 274]	233 [198, 274]	0.51
Baseline FIB-4, median [IQR]	1.02 [0.75, 1.39]	1.03 [0.75, 1.39]	1.02 [0.75, 1.39]	0.40

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; BMI, body mass index; FIB-4, fibrosis-4; INR, international normalized ratio.

Food behavior related to fibrosis progression

We first identified food-intake recall associated with fibrosis progression using the FIB-4 slope.

Food habits statistically significantly and independently related to increased fibrosis progression in the validation cohorts included: white bread (heavy: ≥1 slice/day, β/median=0.727), cookies (light: 1–4/month, β/median=0.554; heavy: ≥1/day, β/median=1.08), breakfast cereal (moderate: 2–6 cups/week, β/median=0.427; heavy: ≥1 cup/day, β/median=0.854), and hard liquor (heavy: ≥1 drink/day, β/median=1.143) (Table 2). By contrast, Intake of several food groups were significantly associated with reduced fibrosis progression including: broccoli (light: 1–4 ½ cups/month, β/median=−0.472; moderate: 2–6 ½ cups/week, β/median=−0.810; heavy: ≥1 ½ cups/day, β/median=−1.17), spinach/collard greens (light: 1–4 ½ cups/month, β/median=−0.469; moderate: 2–6 ½ cups/week, β/median=−0.841; heavy: ≥1 ½ cups/day, β/median=−1.33), beans/lentils (light: 1–4 ½ cups/month, β/median=−0.359; moderate: 2–6 ½ cups/week, β/median=−0.732), coffee (heavy: ≥1 cups/day, β/median=−0.421), tea (light: 1–4 cups/month, β/median=−0.382), nuts (light: 1–4 oz/month, β/median=−0.470; heavy: ≥1 oz/day, β/median=−0.837), chicken/turkey with/without skin, yogurt (light: 1–4 cups/month, β/median=−0.520; moderate: 2–6 cups/week, β/median=−0.440; heavy: ≥1 cups/day, β/median=−0.627), among others (Table 2). Modest beer intake (1–4 glasses/month) was significant in univariable analysis but did not remain significant in adjusted analysis (Table 2).

TABLE 2.

Food intake is significantly associated with fibrosis progression slope in the validation cohort

			Univariable				Adjusteda
Variable	Level	Frequency	β	β/median FIB-4 slope	SE	p	β	β/median FIB-4 slope	SE	p
Diabetes mellitus			0.006	0.184	0.001	<0.0001	0.005	0.171	0.001	<0.0001
Broccoli (1/2 cup)	None	5651 (22)	REF				REF
	1–4/mo	14,177 (56)	−0.004	−0.595	0.001	0.001	−0.004	−0.472	0.001	0.01
	2–6/wk	5213 (21)	−0.007	−0.926	0.002	0.0000	−0.006	−0.810	0.002	0.0003
	≥1 day	341 (1.0)	−0.009	−1.251	0.005	0.06	−0.009	−1.172	0.005	0.07
Spinach or collard greens (1/2 cup)	None	10,638 (42)	REF				REF
	1–4/mo	11,279 (44)	−0.004	−0.596	0.001	0.0002	−0.004	−0.469	0.001	0.003
	2–6/wk	3152 (12)	−0.008	−1.027	0.002	0.0000	−0.006	−0.841	0.002	0.0005
	≥1 day	313 (2.0)	−0.012	−1.588	0.005	0.014	−0.010	−1.332	0.005	0.039
Beans or lentils (1/2 cup)	None	5890 (23)	REF				REF
	1–4/mo	14,652 (58)	−0.003	−0.356	0.001	0.050	−0.003	−0.359	0.001	0.047
	2–6/wk	4521 (18)	−0.005	−0.700	0.002	0.002	−0.005	−0.732	0.002	0.001
	≥1 day	319 (1.0)	−0.009	−1.260	0.005	0.07	−0.009	−1.235	0.005	0.07
String beans (1/2 cup)	None	4552 (18)	REF				REF
	1–4/mo	15,123 (60)	−0.002	−0.249	0.001	0.21	−0.002	−0.294	0.001	0.137
	2–6/wk	5425 (21)	−0.004	−0.589	0.002	0.012	−0.005	−0.653	0.002	0.005
	≥1 day	282 (1.0)	0.014	1.870	0.005	0.007	0.014	1.825	0.005	0.009
Yams or sweet potatoes (1/2 cup)	None	11,947 (47)	REF				REF
	1–4/mo	11,555 (46)	−0.003	−0.416	0.001	0.007	−0.003	−0.381	0.001	0.013
	2–6/wk	1761 (7.0)	−0.005	−0.611	0.002	0.035	−0.004	−0.576	0.002	0.047
	≥1 day	119 (0.0)	−0.008	−1.029	0.007	0.30	−0.007	−0.909	0.007	0.36
Yellow squash (1/2 cup)	None	14,701 (56)	REF				REF
	1–4/mo	9289 (37)	−0.001	−0.182	0.001	0.24	−0.002	−0.214	0.001	0.16
	2–6/wk	1292 (5.1)	−0.007	−0.878	0.003	0.009	−0.006	−0.814	0.003	0.016
	≥1 day	100 (0.9)	−0.010	−1.384	0.009	0.24	−0.009	−1.267	0.009	0.28
Coffee, not decaffeinated (1 cup)	None	7077 (28)	REF				REF
	1–4/mo	2044 (8.0)	−0.003	−0.425	0.002	0.15	−0.002	−0.252	0.002	0.39
	2–6/wk	3592 (14)	0.000	0.033	0.002	0.89	0.000	0.035	0.002	0.88
	≥1 day	12,669 (50)	−0.003	−0.337	0.001	0.05	−0.003	−0.421	0.001	0.016
Tea, not herbal (1 cup)	None	11,776 (46)	REF
	1–4/mo	6297 (25)	−0.004	−0.524	0.001	0.004	−0.003	−0.382	0.001	0.036
	2–6/wk	3698 (15)	−0.002	−0.332	0.002	0.13	−0.002	−0.206	0.002	0.35
	≥1 day	3611 (14)	−0.003	−0.338	0.002	0.13	−0.002	−0.247	0.002	0.26
Beer (1 glass, bottle, can)	None	16,872 (67)	REF				REF
	1–4/mo	5438 (21)	−0.003	−0.403	0.001	0.027	−0.002	−0.260	0.001	0.16
	2–6/wk	2242 (8.8)	−0.003	−0.440	0.002	0.09	−0.002	−0.268	0.002	0.31
	≥1 day	830 (3.2)	0.000	0.045	0.003	0.91	0.001	0.193	0.003	0.64
Liquor (1 drink or shot)	None	19,470 (77)	REF
	1–4/mo	4143 (16)	−0.004	−0.506	0.001	0.011	−0.002	−0.294	0.001	0.14
	2–6/wk	1316 (5.2)	0.002	0.237	0.002	0.47	0.003	0.430	0.002	0.19
	≥1 day	453 (1.8)	0.008	1.100	0.004	0.044	0.009	1.143	0.004	0.036
Nuts (small packet or 1 oz)	None	5896 (23)	REF				REF
	1–4/mo	12,616 (50)	−0.004	−0.577	0.001	0.002	−0.004	−0.470	0.001	0.011
	2–6/wk	5630 (22)	−0.002	−0.282	0.002	0.19	−0.001	−0.180	0.002	0.41
	≥1 day	1240 (5)	−0.007	−0.913	0.003	0.014	−0.006	−0.837	0.003	0.025
Chicken or Turkey, with skin (4–6 oz)	None	4687 (19)	REF				REF
	1–4/mo	11,539 (46)	−0.002	−0.268	0.002	0.19	−0.001	−0.187	0.002	0.36
	2–6/wk	8407 (33)	−0.004	−0.595	0.002	0.005	−0.003	−0.372	0.002	0.08
	≥1 day	749 (2.0)	−0.013	−1.805	0.003	0.000	−0.011	−1.491	0.003	0.001
Chicken or Turkey, without skin (4–6 oz)	None	2802 (11)	REF				REF
	2–6/wk	9437 (37)	−0.005	−0.611	0.002	0.016	−0.004	−0.532	0.002	0.036
	≥1 day	578 (2.3)	−0.011	−1.436	0.004	0.008	−0.009	−1.211	0.004	0.026
White bread (slice), including pita bread	None	9583 (38)	REF				REF
	1–4/mo	7083 (28)	0.001	0.164	0.001	0.372	0.002	0.235	0.001	0.199
	2–6/wk	6713 (26)	0.000	−0.012	0.001	0.949	0.000	0.007	0.001	0.969
	≥1 day	2003 (8.0)	0.006	0.839	0.002	0.004	0.005	0.727	0.002	0.012
Cookies (1)	None	6170 (24)
	1–4/mo	14,239 (56)	0.004	0.571	0.001	0.001	0.004	0.554	0.001	0.002
	2–6/wk	4149 (16)	0.004	0.523	0.002	0.026	0.003	0.388	0.002	0.098
	≥1 day	824 (4.0)	0.010	1.341	0.003	0.002	0.008	1.080	0.003	0.014
Rice or pasta (1 cup)	None	2079 (8.2)
	1–4/mo	14,996 (59)	−0.003	−0.419	0.002	0.126	−0.003	−0.367	0.002	0.178
	2–6/wk	7833 (31)	−0.006	−0.866	0.002	0.003	−0.005	−0.712	0.002	0.013
	≥1 day	474 (1.8)	−0.010	−1.363	0.004	0.022	−0.008	−1.017	0.004	0.09
Cold breakfast cereal (1 cup)	None	6125 (24)
	1–4/mo	9074 (36)	0.003	0.348	0.001	0.071	0.002	0.324	0.001	0.09
	2–6/wk	8054 (32)	0.004	0.550	0.001	0.005	0.003	0.427	0.001	0.031
	≥1 day	2129 (8.0)	0.008	1.090	0.002	0.000	0.006	0.854	0.002	0.004
Yogurt (1 cup)	None	12,404 (49)
	1–4/mo	7036 (28)	−0.005	−0.615	0.001	0.000	−0.004	−0.520	0.001	0.003
	2–6/wk	4458 (18)	−0.004	−0.516	0.002	0.012	−0.003	−0.440	0.002	0.032
	≥1 day	1484 (5.0)	−0.005	−0.659	0.002	0.039	−0.005	−0.627	0.002	0.049
Butter (pat)	None	7445 (29)
	1–4/mo	7088 (28)	−0.005	−0.674	0.001	0.000	−0.004	−0.534	0.001	0.006
	2–6/wk	7639 (30)	−0.002	−0.259	0.001	0.173	−0.002	−0.214	0.001	0.259
	≥1 day	3210 (13)	−0.002	−0.276	0.002	0.265	−0.002	−0.310	0.002	0.211

Bold value are statistically significant P < 0.05.

^a Model adjusted for race/ethnicity, BMI class, number of metabolic syndrome risk factors, diabetes mellitus, AUDIT-C baseline, total bilirubin, AST, ALP, and baseline FIB-4.

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; FIB-4, fibrosis-4.

Nutrient intake related to fibrosis progression

Repeating this approach for specific nutrients, iron from non-heme sources (β/median=0.049), B vitamins (β/median=0.041), and flavanones (β/median=0.038) were significantly associated with fibrosis progression (p<0.05) after adjustment for key covariates (Table 3 and Figure 1A) but did not meet significance after adjustment for multiple comparisons. Caffeine (β=−0.064), betaine (β=−0.064), nitrate/vitamin K (β=−0.069), total amino acids (β/median=−0.059), beta carotene (β/median=−0.042), and iron from heme sources (β/median=−0.036) were significantly associated with reduced FIB-4 slope (p<0.05); however only nitrate/vitamin k, caffeine, betaine, and total amino acids remained significant after Benjamini–Hochberg correction (Table 3 and Figure 1A).

TABLE 3.

Nutrients significantly associated with fibrosis progression in the validation cohort

	Univariable				Adjusteda
Variable	β/SD	β/SD/median FIB-4 slope	SE	p	β/SD	β/SD/median FIB-4 slope	SE	p	p Benjamini–Hochberg
Diabetes mellitus	0.006	0.184	0.001	<0.0001	0.005	0.171	0.001	<0.0001
Nitrate/vitamin K	−0.002	−0.072	0.001	0.0001	−0.002	−0.069	0.001	0.0001	0.005
Caffeine	−0.002	−0.067	0.001	0.0001	−0.002	−0.064	0.001	0.0002	0.005
Betaine	−0.002	−0.069	0.001	0.0001	−0.002	−0.064	0.001	0.0003	0.005
Total amino acids	−0.002	−0.058	0.001	0.001	−0.002	−0.059	0.001	0.001	0.010
Iron (non-heme)	0.002	0.062	0.001	0.0004	0.002	0.049	0.001	0.006	0.052
Beta carotene	−0.001	−0.037	0.001	0.037	−0.001	−0.042	0.001	0.018	0.129
Total B vitamins	0.002	0.050	0.001	0.005	0.001	0.041	0.001	0.020	0.129
Total flavanones	0.001	0.044	0.001	0.015	0.001	0.038	0.001	0.035	0.202
Iron (heme)	−0.001	−0.038	0.001	0.033	−0.001	−0.036	0.001	0.045	0.228

Note: Other nutrients suggested by Elastic Net but not validated include total furocoumarin, sucrose, alcohol, matairesinol, potassium, acrylamide, calcium, fructose+glucose, total flavanols, zinc, aspartame, vitamin D, copper, total cis-polyunsaturated fatty acids, AOAC fiber+phytosterols, total isoflavones, vitamin C, phosphorus, starch, manganese, total cis-monounsaturated fatty acids, total anthocyanidins, coumestrol, cholesterol, gluten, lycopene, total flavones, sodium, total tocopherols, total theaflavin and polymers proanthocyanidins, magnesium, alpha carotene, maltose, total saturated+unsaturated+conjugated linoleic fatty acids, vitamin E, and retinol activity equivalents.

Bold value are statistically significant P < 0.05.

^a Model adjusted for race/ethnicity, BMI class, number of metabolic syndrome risk factors, diabetes mellitus, AUDIT-C baseline, total bilirubin, AST, ALP, and baseline FIB-4.

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; BMI, body mass index; FIB-4, fibrosis-4.

FIGURE 1.

Forest plot of validated nutrients identified by Elastic Net regression at alpha=0. (A) Validated nutrients associated with fibrosis progression. All variables are statistically significant at p<0.05 with Benjamini–Hochberg correction. Negative beta coefficients are associated with reduced FIB-4 slopes, and positive beta coefficients with increased FIB-4 slopes. (B) Validated nutrients associated with time to cirrhosis. All variables are statistically significant at p<0.05 with Benjamini–Hochberg correction. HR <1 is associated with reduced risk of cirrhosis, and HR >1 is associated with increased risk of cirrhosis.

Food behavior related to the time to cirrhosis

Food behaviors identified that were significantly associated with decreased risk of cirrhosis were light or moderate wine (light: 1–4 4 oz/month, HR 0.66 [0.53–0.83]; moderate: 2–6 4 oz/week, HR 0.26 [0.15–0.47]), light or moderate beer (light: 1–4 4 oz/month, HR 0.75 [0.61–0.92], moderate: 2–6 4 oz/week, HR 0.71 [0.51–0.97]), light use of hard liquor (light: 1–4 drinks/month, HR 0.74 [0.59–0.94]), coffee (heavy: ≥1 cup/day, HR 0.81 [0.68–0.96]), yogurt (moderate: 2–6 cups/week, HR 0.71 [0.56–0.89]), and nuts (heavy: ≥small packet or 1 oz/day, HR 0.62 [0.38–0.99]). Nutrients associated with higher risk of cirrhosis included intake of cheese (light: 1 slice or 1 oz 1–4/month, HR 1.55 [1.05–2.29]; moderate: 2–6/month, HR 1.65 [1.12–2.43]; heavy: ≥1/day, HR 1.67 [1.08–2.58]), fruit drinks (moderate: 2–6 drinks/week, HR 1.32 [1.06–1.65]), white bread (moderate: 2–6 slices/week, HR 1.22 [1.01–1.48]; heavy: ≥1 slice/day, HR 1.45 [1.11–1.89]), cake (light: 1–4 slices/month, HR 1.17 [1.00–1.36]), hot dogs (light: 1–4/month, HR 1.27 [1.06–1.51]), and ready-made pie (light: 1–4 slice/month, HR 1.17 [1.01–1.37]) among others (Table 4). Other foods that reached significance in univariable analysis but not adjusted analysis include: carbonated beverages (regular and diet), broccoli, spinach/collard greens, peaches/apricots/plums, food fried at home, hamburger, homemade pie, French fried potatoes, and processed meats (Table 4).

TABLE 4.

Food intake significantly associated with time to cirrhosis in the validation cohort

			Univariable		Adjusteda
Variable	Level	Frequency	HR (95% CI)	p	HR	p	C
Diabetes present			2.27 (1.95–2.64)	<0.0001	1.67 (1.41–1.97)	<0.0001	0.751
Wine (4 oz)	None	19,017 (75)
	1–4/mo	4692 (19)	0.62 (0.49–0.77)	<0.0001	0.66 (0.53–0.83)	0.0004	0.758
	2–6/wk	1254 (5.0)	0.43 (0.27–0.70)	0.001	0.26 (0.15–0.47)	<0.0001
	≥1 daily	419 (1.0)	0.90 (0.51–1.59)	0.71	0.82 (0.45–1.48)	0.51
Beer (1 glass, bottle, can)	None	16,872 (67)
	1–4/mo	5438 (21)	0.72 (0.59–0.88)	0.001	0.75 (0.61–0.92)	0.005	0.754
	2–6/wk	2242 (8.8)	0.61 (0.44–0.84)	0.003	0.71 (0.51–0.97)	0.034
	≥1 daily	830 (3.2)	0.64 (0.39–1.04)	0.07	0.72 (0.44–1.18)	0.19
Liquor (1 drink or shot)	None	19,470 (77)
	1–4/mo	4143 (16)	0.69 (0.55–0.87)	0.002	0.74 (0.59–0.94)	0.012	0.754
	2–6/wk	1316 (5.2)	0.66 (0.44–0.98)	0.04	0.75 (0.50–1.11)	0.14
	≥1 daily	453 (1.8)	0.85 (0.48–1.51)	0.59	0.85 (0.47–1.51)	0.57
Coffee, not decaffeinated (1 cup)	None	7077 (28)
	1–4/mo	2044 (8.0)	0.80 (0.59–1.09)	0.16	0.87 (0.64–1.18)	0.36	0.752
	2–6/wk	3592 (14)	0.92 (0.73–1.15)	0.48	0.93 (0.74–1.17)	0.56
	≥1 daily	12,669 (50)	0.79 (0.67–0.94)	0.008	0.81 (0.68–0.96)	0.015
Carbonated beverage with sugar, eg, Coke, Pepsi	None	13,742 (54)
	1–4/mo	6502 (26)	0.81 (0.67–0.97)	0.021	0.88 (0.73–1.06)	0.17	0.751
	2–6/wk	3343 (13)	0.77 (0.60–0.98)	0.033	0.94 (0.73–1.20)	0.59
	≥1 daily	1795 (7.0)	0.90 (0.67–1.22)	0.50	1.07 (0.79–1.45)	0.68
Low-calorie carbonated beverage, eg, Diet Coke	None	12,991 (51)
	1–4/mo	5106 (20)	1.12 (0.92–1.36)	0.26	0.90 (0.73–1.10)	0.28	0.751
	2–6/wk	4027 (16)	1.33 (1.08–1.62)	0.006	1.02 (0.83–1.25)	0.86
	≥1 daily	3258 (13)	1.26 (1.01–1.57)	0.039	0.97 (0.77–1.21)	0.79
Cheese (1 slice or 1 oz)	None	1496 (6.0)
	1–4/mo	8903 (35)	1.12 (0.79–1.59)	0.51	1.55 (1.05–2.29)	0.028	0.753
	2–6/wk	12,245 (48)	1.22 (0.87–1.71)	0.25	1.65 (1.12–2.43)	0.012
	≥1 daily	2738 (11)	1.29 (0.87–1.90)	0.20	1.67 (1.08–2.58)	0.020
Skim or low-fat milk (8 oz)	None	9448 (37)
	1–4/mo	3897 (16)	0.82 (0.64–1.04)	0.10	0.65 (0.50–0.84)	0.001	0.751
	2–6/wk	5842 (23)	1.08 (0.89–1.31)	0.43	1.00 (0.83–1.21)	0.99
	≥1 daily	6195 (24)	1.07 (0.88–1.29)	0.49	0.96 (0.79–1.17)	0.70
Hawaiian Punch, lemonade, and other fruit drinks	None	14,581 (58)
	1–4/mo	6699 (26)	0.89 (0.75–1.07)	0.22	0.94 (0.78–1.13)	0.51	0.751
	2–6/wk	2906 (11)	1.20 (0.96–1.50)	0.11	1.32 (1.06–1.65)	0.015
	≥1 daily	1196 (5.0)	1.08 (0.76–1.54)	0.66	1.27 (0.89–1.81)	0.19
White bread (slice), including pita bread	None	9583 (38)
	1–4/mo	7083 (28)	1.18 (0.97–1.42)	0.09	1.09 (0.90–1.32)	0.36	0.752
	2–6/wk	6713 (26)	1.24 (1.03–1.50)	0.025	1.22 (1.01–1.48)	0.036
	≥1 daily	2003 (8.0)	1.52 (1.16–1.98)	0.002	1.45 (1.11–1.89)	0.006
Cake (slice)	None	12,462 (49)
	1–4/mo	12,053 (48)	1.20 (1.03–1.39)	0.019	1.17 (1.00–1.36)	0.046	0.753
	2–6/wk	822 (3.2)	0.98 (0.62–1.53)	0.92	1.01 (0.65–1.60)	0.95
Broccoli (1/2 cup)	None	5651 (22)
	1–4/mo	14,177 (56)	0.77 (0.65–0.92)	0.004	0.87 (0.73–1.04)	0.11	0.752
	2–6/wk	5213 (21)	0.76 (0.61–0.94)	0.012	0.83 (0.66–1.03)	0.09
	≥1 daily	341 (1.0)	0.44 (0.18–1.08)	0.07	0.49 (0.20–1.18)	0.11
Peas or lima beans (1/2 cup)	None	6723 (26)
	1–4/mo	14,472 (57)	0.98 (0.82–1.18)	0.86	1.05 (0.87–1.26)	0.62	0.753
	2–6/wk	4000 (16)	1.28 (1.03–1.60)	0.028	1.40 (1.12–1.75)	0.003
	≥1 daily	187 (1.0)	0.86 (0.36–2.10)	0.75	0.99 (0.41–2.41)	0.98
Spinach or collard greens (1/2 cup)	None	10,638 (42)
	1–4/mo	11,279 (44)	0.82 (0.70–0.96)	0.014	0.93 (0.79–1.09)	0.37	0.751
	2–6/wk	3152 (12)	0.85 (0.67–1.08)	0.18	0.99 (0.77–1.26)	0.92
	≥1 daily	313 (2.0)	0.87 (0.45–1.69)	0.68	1.03 (0.53–2.00)	0.93
Whole milk (8 oz)	None	17,596 (69)
	1–4/mo	3550 (14)	1.23 (1.00–1.51)	0.046	1.21 (0.98–1.50)	0.07	0.753
	2–6/wk	2551 (10)	1.23 (0.97–1.55)	0.09	1.33 (1.05–1.69)	0.016
	≥1 daily	1685 (7.0)	1.04 (0.77–1.40)	0.80	1.16 (0.86-1.56)	0.34
Yogurt (1 cup)	None	12,404 (49)
	1–4/mo	7036 (28)	0.97 (0.82–1.15)	0.72	1.03 (0.86–1.22)	0.77	0.752
	2–6/wk	4458 (17)	0.76 (0.60–0.94)	0.014	0.71 (0.56–0.89)	0.003
	≥1 daily	1484 (6.0)	0.87 (0.63–1.21)	0.41	0.95 (0.68–1.32)	0.74
Other fruits, fresh, frozen, and canned (1/2 cup)	None	4682 (18)
	1–4/mo	10,869 (43)	1.25 (1.00–1.55)	0.050	1.29 (1.03–1.60)	0.024	0.753
	2–6/wk	7725 (31)	1.25 (0.99–1.57)	0.06	1.23 (0.97–1.54)	0.08
	≥1 daily	2106 (8.0)	1.14 (0.83–1.58)	0.41	1.15 (0.84–1.59)	0.39
Peaches, apricots, and plums (1 fresh, 1/2 cup canned)	None	10,166 (40)
	1–4/mo	11,244 (44)	0.98 (0.83–1.16)	0.82	0.95 (0.80–1.12)	0.54	0.752
	2–6/wk	3417 (14)	1.25 (1.01–1.55)	0.039	1.08 (0.87–1.35)	0.48
	≥1 daily	555 (2.0)	1.31 (0.83–2.06)	0.25	1.26 (0.80–1.99)	0.32
Hamburger (1 patty)	None	2648 (10)
	1–4/mo	16,731 (66)	1.35 (1.03–1.78)	0.031	1.21 (0.92–1.60)	0.17	0.751
	2–6/wk	5840 (23)	1.44 (1.07–1.94)	0.016	1.20 (0.89–1.62)	0.23
	≥1 daily	163 (1.0)	2.35 (1.12–4.92)	0.023	1.92 (0.91–4.02)	0.08
Hot dog (1)	None	8104 (32)
	1–4/mo	14,892 (59)	1.41 (1.18–1.68)	0.0001	1.27 (1.06–1.51)	0.008	0.753
	2–6/wk	2280 (9.0)	1.33 (1.00–1.75)	0.050	1.18 (0.89–1.57)	0.24
	≥1 daily	106 (0.0)	1.28 (0.41–4.02)	0.67	1.18 (0.38–3.69)	0.78
Nuts (small packet or 1 oz)	None	5896 (23)
	1–4/mo	12,616 (50)	0.88 (0.74–1.05)	0.16	0.99 (0.82–1.18)	0.87	0.753
	2–6/wk	5630 (22)	0.76 (0.61–0.94)	0.013	0.85 (0.68–1.06)	0.14
	≥1 daily	1240 (5)	0.52 (0.32–0.82)	0.006	0.62 (0.38–0.99)	0.045
Pie, ready-made (slice)	None	14,476 (57)
	1–4/mo	10,217 (40)	1.20 (1.03–1.39)	0.019	1.17 (1.01–1.37)	0.038	0.754
	2–6/wk	655 (2.6)	1.33 (0.86–2.04)	0.19	1.44 (0.93–2.21)	0.098
	≥1 daily	34 (0.4)	1.06 (0.15–7.55)	0.95	0.80 (0.11–5.68)	0.821
French fried potatoes (4 oz)	None	6066 (24)
	1–4/mo	15,768 (62)	1.13 (0.94–1.35)	0.21	1.06 (0.88–1.28)	0.515	0.752
	2–6/wk	3361 (13)	1.31 (1.03–1.67)	0.029	1.27 (1.00–1.62)	0.051
	≥1 daily	187 (1.0)	1.24 (0.51–3.02)	0.64	1.40 (0.58–3.42)	0.455
Processed meats (piece or slice)	None	5876 (23)
	1–4/mo	13,499 (53)	1.25 (1.03–1.51)	0.024	1.12 (0.93–1.36)	0.240	0.753
	2–6/wk	5477 (22)	1.24 (0.99–1.56)	0.06	1.13 (0.90–1.42)	0.298
	≥1 daily	530 (2.0)	1.73 (1.08–2.77)	0.021	1.54 (0.96–2.46)	0.071

Bold value are statistically significant P < 0.05.

^a Model adjusted for race/ethnicity, BMI class, number of metabolic syndrome risk factors, diabetes mellitus, AUDIT-C baseline, total bilirubin, AST, ALP, and baseline FIB-4.

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; BMI, body mass index.

Nutrient intake related to time to cirrhosis

Sodium (HR 1.13 [1.05–1.21]), total unsaturated/saturated/conjugated linoleic fatty acids (HR 1.14 [1.06–1.22]), iron from non-heme sources (HR 1.11 [1.04–1.17]), and bergapten/furocoumarin (HR 1.08 [1.01–1.16]) were associated with increased risk of cirrhosis; all variables except bergapten/furocoumarin remained significant with Benjamini–Hochberg correction (Table 5 and Figure 1B). Intake of secoisolariciresinol (HR 0.86 [0.79–0.93]), caffeine (HR 0.85 [0.78–0.93]), and manganese (HR 0.92 [0.85–0.99]) was associated with reduced risk of cirrhosis (Table 5 and Figure 1B). Only secoisolariciresinol and caffeine remained significant with Benjamini–Hochberg correction.

TABLE 5.

Nutrients significantly associated with time to cirrhosis in the validation cohort

	Univariable		Adjusteda
Variable	HR (95% CI)	p	HR (95% CI)	p	C	p Benjamini–Hochberg
Diabetes present	2.27 (1.95–2.64)	<0.0001	1.67 (1.41–1.97)	<0.0001	0.751
Sodium	1.17 (1.09–1.25)	0.0000	1.13 (1.05–1.21)	0.001	0.739	0.01
Total unsaturated+saturated+conjugated linoleic fatty acids	1.16 (1.09–1.24)	0.0000	1.14 (1.06–1.22)	0.000	0.738	0.005
Secoisolariciresinol	0.85 (0.78–0.92)	0.0001	0.86 (0.79–0.93)	0.000	0.738	0.005
Caffeine	0.84 (0.77–0.92)	0.0001	0.85 (0.78–0.93)	0.000	0.738	0.005
Iron (non-heme)	1.12 (1.06–1.19)	0.0001	1.11 (1.04–1.17)	0.001	0.740	0.011
Cholesterol	1.11 (1.04–1.18)	0.002	1.05 (0.98–1.12)	0.20	0.737	0.45
Manganese	0.91 (0.84–0.98)	0.017	0.92 (0.85–0.99)	0.036	0.738	0.26
Betaine	0.91 (0.84–0.99)	0.021	0.94 (0.86–1.02)	0.12	0.736	0.35
Total cis-monounsaturated fatty acids	1.08 (1.01–1.16)	0.032	1.03 (0.95–1.11)	0.45	0.737	0.70
Aspartame	1.07 (1.00–1.14)	0.043	1.03 (0.96–1.11)	0.38	0.737	0.66
Phosphorus	1.07 (1.00–1.15)	0.061	1.03 (0.96–1.11)	0.44	0.736	0.70
Alcohol	0.92 (0.85–1.01)	0.076	0.93 (0.86–1.01)	0.08	0.737	0.34
Total flavols	0.93 (0.86–1.01)	0.086	0.93 (0.86–1.02)	0.11	0.737	0.34
Nitrate+vitamin K	0.93 (0.85–1.01)	0.088	0.94 (0.87–1.03)	0.19	0.737	0.44
Bergapten/furocoumarin	1.04 (0.97–1.11)	0.296	1.08 (1.01–1.16)	0.033	0.736	0.26
AOAC fiber+phytosterols	0.94 (0.87–1.01)	0.098	0.93 (0.86–1.00)	0.051	0.737	0.30

Note: Variables suggested by Elastic Net that did not validate include retinal activity equivalents, sucrose, gluten, total theaflavin and polymers proanthocyanidins, coumestrol, beta cryptoxanthin, fructose+glucose, total amino acids, betaine+choline, copper, zinc, beta carotene, starch, lycopene, iron (heme), potassium, lactose, vitamin C, alpha carotene, magnesium, total tocopherols, total B vitamins, matairesinol, total anthocyanidins, vitamin D, vitamin E, total cis-polyunsaturated fatty acids, total flavanone, total isoflavones, calcium, acrylamide, total flavones, and maltose.

Bold value are statistically significant P < 0.05.

^a Model adjusted for race/ethnicity, BMI class, number of metabolic syndrome risk factors, diabetes mellitus, AUDIT-C baseline, total bilirubin, AST, ALP, and baseline FIB-4.

Abbreviations: AUDIT-C, Alcohol Use Disorders Identification Test; BMI, body mass index.

Mediation analysis

To isolate direct effects of food behavior and nutrients from indirect effects mediated by other exposures, we performed a mediation analysis (Supplemental Tables S7, S8, http://links.lww.com/HC9/C29) for diabetes mellitus, a marker of liver function, total bilirubin, and a marker of inflammation, AST, sodium, fatty acids, secoisolariciresinol, caffeine, and non-heme iron were associated with fibrosis progression independently of diabetes, bilirubin, and AST, with partial mediation of the effects of secoisolariciresinol, caffeine, and non-heme iron through diabetes mellitus. The associations of coffee intake, nuts, and caffeine with fibrosis progression and time to cirrhosis were independent of diabetes, but associations with carbonated beverages and fruit drinks appeared fully mediated by diabetes mellitus. Secoisolariciresinol was associated with time to cirrhosis independent of diabetes but with effects partially mediated by total bilirubin and AST, suggesting anti-inflammatory effects.

DISCUSSION

This study identifies several nutrients and food-intake behaviors that appear to be associated with the progression of MASLD and affirms several clinical variables previously associated with progression. Two outcomes were evaluated: fibrosis progression via FIB-4 slope and time to cirrhosis diagnosis. Food-related behaviors such as consumption of cookies and processed foods (white bread, breakfast cereals) appeared to be associated with increased fibrosis progression, while intake of vegetables (broccoli, spinach/collard greens, yams/sweet potatoes, yellow squash, string beans), grains (rice/pasta), nuts, legumes, poultry, tea/coffee, and yogurt appeared to be protective against fibrosis progression. Regarding the nutrient-specific findings, nitrate/vitamin K, betaine, caffeine, and amino acids were associated with reduced fibrosis progression, while no nutrients were significantly associated with progression of fibrosis after correction for multiple comparisons.

Consumption of nuts, coffee, skim milk, yogurt, and moderate alcohol (wine/beer/liquor) was associated with reduced risk of diagnosis with cirrhosis. By contrast, consumption of sweets (pie, cake, fruit drinks), processed foods (white bread, hot dogs), dairy (cheese, whole milk), and starchy vegetables (peas/lima beans) was associated with an increased risk of cirrhosis diagnosis. Nutrients such as secoisolariciresinol and caffeine were associated with decreased risk; sodium, total unsaturated + saturated + conjugated linoleic fatty acids, and iron from non-heme sources were associated with increased risk. Most of these effects were direct, with modest mediation by diabetes.

The results identified in this study are consistent with existing literature⁴,¹⁹ while providing new insights on outcomes in a longitudinally evaluated MASLD cohort. The general pattern of food types appearing to show benefit are those associated with the Mediterranean diet (MD) and/or EAT-Lancet diet,⁴ including daily consumption of vegetables, whole grains, nuts, and fiber. The MD is high in unsaturated fats and low in refined sugars such as fructose. Further, the American Heart Association (AHA) recommends reduced egg and cholesterol intake, reduced sodium, and limiting refined carbohydrates and saturated fats consistent with the MD.²⁰ Here, we show that nutrient intake of nitrate/vitamin K and betaine appears to be protective against fibrosis progression in MASLD, with secoisolariciresinol being associated with reduced risk of developing cirrhosis. Dietary intake of nitrate and vitamin K is found in highest concentrations in leafy greens and cruciferous vegetables, and betaine in beets, spinach, and whole grains. Secoisolariciresinol, a type of plant lignan, comes from nuts/seeds, whole grains, legumes, and vegetables. Total unsaturated/saturated/conjugated linoleic fatty acids were associated with increased risk of cirrhosis; while dietary unsaturated fats are generally recommended by the MD/AHA, this grouped variable also includes saturated fats and conjugated linoleic fatty acids (commonly found in meat and dairy products), both of which should be reduced in the AHA guidelines and MD, respectively.²¹ Similarly, sodium was found to be associated with increased risk of cirrhosis. Finally, while amino acids, associated with decreased fibrosis risk, are typically associated with animal-based protein sources, they are also found in high concentrations in plant-based sources such as soy products, grains, legumes, and nuts. While the MD suggests a reduction of red meats and limited intake of poultry, the diet encourages frequent seafood and other plant-based sources of proteins, which could explain the effect observed with amino acid intake.

Furthermore, the food-related behaviors identified in this study show general trends toward dietary choices that align with the MD as well. Intake of nuts, vegetables, and whole grains was associated with protection, and consumption of high-fructose foods (cookies/cake/pie), refined carbohydrates (white bread, cereals), processed foods (hot dogs), and dairy (cheese, whole milk) was associated with progression. Of note, yogurt was found to be associated with protection in both outcomes; consumption of moderate amounts of yogurt is recommended in the MD. These results also tended to be dose-dependent, with many variables showing stronger effects with increased intake of the food. These results overall suggest that a diet enriched in vegetables/nuts, with less ingestion of starches, fructose, and refined carbohydrates, may be hepatoprotective in MASLD.

One nutrient identified in this study does not align as closely with the suggested diets outlined in the MD. We found that intake of iron from non-heme sources was associated with increased risk of cirrhosis. Iron from non-heme sources is found in the highest concentrations in leafy vegetables. Notably, however, the SFFQ included supplementary sources of vitamins; patients who are prescribed supplementary iron may be sicker at baseline, which suggests that the association seen with iron from non-heme sources and progression may be mediated by confounding by indication.

Several food intakes did not align as expected with the Mediterranean Diet. The MD suggests moderate intake of poultry²⁰; here, there was a protective effect in fibrosis progression identified with moderate and heavy poultry intake. Individuals who eat daily poultry, however, may have more limited intake of other animal-based protein sources such as red meat, which should be even further limited in the MD. “Other fruit” includes any fruit that is not apples/pears, bananas, oranges, and peaches/apricots/plums. This suggests the negative effect seen with intake of “other fruit” could be mediated by intake of high glycemic index fruits, such as canned fruits, watermelon, or pineapple.

The pathophysiology of MASLD progression is linked to several factors, including insulin response, oxidative stress, and chronic inflammation.²² The significance of our findings of nutrients and food-intake behaviors is supported by Zhu et al²³ who found that foods with higher dietary insulin responses have been shown to increase risk of liver steatosis and fibrosis; high glycemic index foods are shown to specifically increase hepatic fat and glycogen stores,²⁴ ultimately contributing to fibrosis. Low glycemic index foods identified in this study, such as vegetables, nuts, and whole grains, are protective against these mechanisms. Conversely, high glycemic index foods, including white bread, cereals, cookies, and cake/pie, were found to be associated with progression.

Several specific nutrients identified in this study have compelling mechanistic evidence supported by prior research. We identified intake of caffeine as a nutrient and consumption of coffee to be associated with both protection against fibrosis progression and risk of cirrhosis. The protective effect of caffeine has been replicated in many other studies, likely due to the role of caffeine in the reduction of hepatic fibrosis.²⁵,²⁶ In another study, non-starchy, dark-green vegetables, including broccoli, were shown to be beneficial for the prevention of liver disease and were associated with decreased odds of steatosis; the opposite effect was shown for starchy vegetables such as peas/lima beans.²⁷ These effects were likely mediated by dietary insulin response. Betaine has also been shown to play a role in regulating genes that contribute to inflammation, fatty acid oxidation, lipogenesis, and insulin resistance.²⁸ Further, new evidence supports supplemental intake of betaine in the reduction of transaminase elevations and hepatic fat seen in a MASLD cohort mouse model—an effect possibly mediated through improvement of insulin resistance.²⁹,³⁰ Finally, secoisolariciresinol has also been found to protect against hepatic steatosis and insulin resistance, and induce improved lipid metabolism in mouse models.³¹

Finally, while other nutrients have mechanistic plausibility to support progression or protection in MASLD, alcohol use has more varied results. Patients with heavy alcohol use, as determined by AUDIT-C score ≥4, were excluded from the study, suggesting that the association identified here relates to moderate alcohol use in individuals without AUD. In all, 70% of individuals surveyed indicated that they had decreased their alcohol use over the past 5 years. The effect here could be mediated by alcohol reduction rather than the current level of alcohol intake. Of note, hard liquor use was positively associated with FIB-4 progression.

There was limited overlap of foods from the FFQ and nutrients associated with both fibrosis progression and time to cirrhosis. The variables that validated for both outcomes included yogurt, nuts, coffee, and caffeine as being protective, while white bread was associated with worse outcomes. Other nutrients/foods that validated in both outcomes, but did not reach significance, included nitrate/vitamin K, iron from non-heme sources, betaine, broccoli, and spinach/collard greens. We believe that the limited overlap of variables is related to a low number of cirrhosis events in the study population.

This study is the first to link quantitative dietary nutrients and food-related behaviors with the progression of MASLD with extended clinic follow-up. The Million Veteran Program is the nation’s largest biorepository in the veteran population, with over multiple decades of clinical monitoring, highlighting the unique strength of the longitudinal cohort studied. Limitations include generalizability as the cohort studied is comprised primarily of white, male, and middle-aged veterans. Further, data for the time to cirrhosis outcome is limited by a low number of events observed in the total cohort. The nutrients and food-related behaviors were estimated based on 12-month food recalls, which are affected by patient memory and validity. Additionally, while the study focused on food behaviors that are more likely to be lifelong, this SFFQ was performed one time and did not capture temporal dietary variation. Further, physical activity levels not accounted for in this study may play a significant role in mediating diet effects observed in this population. An individual’s gut microbiota may also play an important role in mediating dietary effects.³² FIB-4 slope as a measure of fibrosis progression has been validated to correlate with clinical events,¹⁴ but is a surrogate marker and not a direct measurement of fibrosis. Finally, as many foods contribute to individual nutrient intake, it is difficult to suggest which food groups are most beneficial based on nutrient data alone; there are also many foods that were not surveyed in this study that may have undetermined impacts on MASLD.

This study evaluated nutrition as it is associated with the natural history of MASLD, with the ultimate goal of providing guidance to patients about specific food-related behavior changes that could aid in the reduction of MASLD progression and negative outcomes. Overall, the findings of this study provide further support for the MD in protection against progression in MASLD, specifically in veteran cohorts. This study also suggests a specific positive impact of caffeine/coffee, yogurt and nuts and a negative impact of white bread on both fibrosis progression and time to cirrhosis. As MVP has robust genetic data, future directions could also include correlating the nutritional data with patient genetic information.

Supplementary Material

hc9-9-e0754-s001.docx ^{(2.9MB, docx)}

DATA AVAILABILITY STATEMENT

The summary datasets that support the findings of this study are available from the corresponding author, David E. Kaplan, upon reasonable request. A data use agreement would be required for the sharing of de-identified individual subject data.

AUTHOR CONTRIBUTIONS

Lauren E. Callans: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis. David E. Kaplan: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis. Kerry L. Ivey: study concept and design; analysis and interpretation of data; drafting of the manuscript; and critical revision of the manuscript for important intellectual content. Kyong-Mi Chang: study concept and design; acquisition of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; obtained funding, technical, or material support; study supervision.

FUNDING INFORMATION

This research is based on data from the Million Veteran Program (MVP), Office of Research and Development, Veterans Health Administration and was supported by MVP000 as well as award #I01-BX003362 entitled MVP003/028: Genetics of Cardiometabolic Diseases in the VA Population (Kyong-Mi Chang). David E. Kaplan is supported by the Veterans Administration [I01-CX002010, I01-CX002337, I01-CX002542]. This publication does not represent the views of the Department of Veterans Affairs, the U.S. Food and Drug Administration, or the U.S. Government. We thank all MVP participants and research team members for their contributions. The content is the responsibility of the authors alone and does not necessarily reflect the views of or imply endorsement by the U.S. Government.

CONFLICTS OF INTEREST

David E. Kaplan advises and received grants from AstraZeneca. He consults for Roche. He received grants from Gilead, Bausch, and Exact Sciences. The remaining authors have no conflicts to report.