The Cross-Sectional Association Between Ultra-Processed Food Intake and Metabolic Dysfunction-Associated Steatotic Liver Disease

Abstract

Background and Aims:

The prevalence of Metabolic Dysfunction-Associated Steatotic Liver Disease has increased in parallel with a rise in consumption of ultra-processed foods (UPF), but little is known about their association.

Methods:

We cross-sectionally examined associations of UPF with hepatic steatosis and fibrosis in 2,458 (mean age 54 years; 55.9% women) community-dwelling adults who completed vibration-controlled transient elastography and a food frequency questionnaire. Dietary intake was categorized into levels of food processing via the NOVA system. We used multivariable-adjusted logistic regression models to evaluate the association of energy-adjusted UPF intake (per 1-SD unit and by quintile) with clinical hepatic steatosis (Controlled Attenuation Parameter [CAP]≥ 290 dB/m) and fibrosis (Liver Stiffness Measurement [LSM] ≥ 8.2 kPa) and tested for linear trends of UPF intake with CAP and LSM. We adjusted for age, sex, smoking, alcohol intake, physical activity, and intake of minimally processed foods. Additional models adjusted for diet quality index or body mass index (BMI).

Results:

Higher intake of UPF was directly associated with higher odds of hepatic steatosis (Odds Ratio 1.33 [95% Confidence Interval 1.21, 1.46] per standard deviation increase). UPF intake and CAP had a dose-response relation (P_trend <0.001). There were 2.50 times higher odds of hepatic steatosis (Confidence Interval 1.81, 3.45) with a 19.49 (standard error: 3.73) unit increase in CAP (P<0.001) when comparing quintile 5 to quintile 1 of UPF consumption. Higher UPF was not significantly associated with hepatic fibrosis. Adjustment for BMI attenuated the strength of all UPF-hepatic associations.

Conclusions:

UPF consumption was positively associated with hepatic steatosis. Longitudinal studies are needed to assess whether lowering consumption of UPF can decrease odds of hepatic fibrosis.

Introduction:

Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) is present in roughly 30% of adults across the globe, and the prevalence has been increasing over recent years [1]. End-stage liver disease as a sequela of MASLD has become a leading indication for liver transplantation in the United States. The recent nomenclature change from Non-Alcoholic Fatty Liver Disease (NAFLD) to MASLD reflect the growing understanding of the pathophysiological link between liver and cardio-metabolic disease [2]. Individuals with hepatic fibrosis are at particularly high risk for poor cardiometabolic health and cardiovascular disease (CVD)-related death [3,4].

Dietary changes are the cornerstone of the management of MASLD [5]. Recently, there has been an emphasis on how ultra-processed foods (UPF) may affect metabolic health [6]. UPF are defined as formulations of industrialized ingredients that have undergone extensive manufacturing and are characterized by high caloric density, low nutritional quality, and the presence of cosmetic additives and/or substances of rare culinary use [7]. In the United States, UPF constitute >50% of energy consumed and global consumption of UPF is on the rise [8]. Prior studies have observed associations between UPF and increased risk of obesity, type 2 diabetes, and CVD [9,10]. ‘Among the few studies of highly processed foods and MASLD, most focus on individual foods as opposed to the overall consumption of UPF using uniform classification system, and few use imaging to define both steatosis and fibrosis.[11,12].

We evaluated the association of vibration-controlled transient elastography (VCTE)-defined liver steatosis and fibrosis with consumption of UPF in a community cohort of middle-aged adults. We hypothesized that higher consumption of UPF, categorized by the NOVA classification system, is associated with greater odds of liver steatosis and fibrosis.

Methods:

STUDY SAMPLE

The Framingham Heart Study (FHS) is an ongoing prospective cohort study that originated in 1948 [13]. We included participants from the FHS Third Generation, New Offspring Spouse (NOS) and OMNI2 cohorts, which have previously been described [14]. Participants who underwent VCTE and completed a food frequency questionnaire at the third examination cycle (2016–2019) were considered for inclusion. We excluded 202 participants with high alcohol intake (defined as ≥14 and ≥21 drinks/week in women and men, respectively), and 16 participants with missing covariate data. A total of 2,458 participants remained for the final analysis (Supplemental Figure 1). The Boston University Charles River Campus Institutional Review Boards approved the study protocol, and all participants provided written informed consent.

VCTE MEASUREMENTS

VCTE (Fibroscan; Echosens, Paris, France) was performed on each participant by a certified operator to obtain values for controlled attenuation parameter (CAP) and liver stiffness measurement (LSM). The protocol for obtaining these measurements has been previously described [4]. We excluded poorly reliable examinations, defined as an interquartile range/median ratio >0.30 when the median LSM is ≥7.1 kPa. We used the cutoff values of LSM ≥ 8.2 and LSM ≥ 13.6 kPa for clinically significant fibrosis and severe fibrosis, respectively, based on previous studies [15,16]. We used cutoffs of CAP ≥ 290 dB/m and CAP ≥ 302 dB/m for any steatosis, and severe steatosis, respectively, based on previous studies [4,15]. Participants with VCTE evidence of steatosis were considered to have MASLD without specifically assessing concomitant metabolic dysfunction or excluding other etiologies for steatotic liver disease based on prior studies demonstrating significant overlap between previously defined non-alcoholic fatty liver disease and MASLD in the FHS population [17].

DIETARY DATA AND CHARACTERIZATION OF ULTRA-PROCESSED FOOD CONSUMPTION

Dietary intake was assessed via the Harvard (Willett) food frequency questionnaire (FFQ), which is a 131-item questionnaire that has been validated in comparison to seven-day dietary records [18]. For each food, participants reported how frequently they consumed specific foods in the previous year on average, on a scale of <1 serving per month to ≥ 6 servings per day. FFQ data was considered invalid when the reported energy intake was <2.5 MJ/d (600 kcal/d) for both men and women, ≥16.7 MJ/d (4000 kcal/d) for women, ≥17.5 MJ/d (4200 kcal/d) for men, or ≥13 food items were left blank.

Reported foods were then categorized by level of processing according to the NOVA classification system [7]. The NOVA Food Classification system was designed by the University of Sao Paulo and categorizes foods into four groups by increasing levels of processing (Supplemental Methods), with the NOVA4 group foods being considered as UPF. One serving of UPF corresponded to common serving sizes, for example, 1 can of sugar-sweetened beverage or 1 oz of potato chips. Servings of UPF were adjusted for energy using the residual method [19]. Energy-adjusted ultra-processed food servings were then categorized into quintiles.

The Healthy Eating Index 2015 (HEI-2015), a 13-component index that serves as a measure of adherence to the Dietary Guidelines for Americans [20], was derived from FFQ data (Supplemental Methods).

COVARIATES

Study participants attended a routine examination between 2016–2019, during which all covariates were assessed in accordance with standard protocols. Body mass index (BMI) was calculated as weight divided by height squared (kg/m²). Waist circumference (WC) was measured at the level of the umbilicus with the participant in a standing position. Diabetes status was defined as HbA1c ≥ 6.5%, or fasting blood glucose ≥ 126 mg/dL or on a targeted medication for diabetes. Current cigarette-users were defined as participants who self-reported smoking at least one cigarette per day in the prior year. Participants self-reported their consumption of beer, wine, liquor, and spirits in the FFQ. We generated a physical activity score using the intensity and time spent for each type of activity assessed by the physical activity questionnaire [21].

STATISTICAL ANALYSIS

MULTIVARIABLE LOGISTIC REGRESSION MODELS

We used multivariable logistic regression models to examine the relation of UPF intake with our primary outcomes of hepatic steatosis and fibrosis. We modeled energy-adjusted UPF intake as a continuous variable (per 1 standard deviation unit, approximately 2.3 servings of UPF) and by quintiles (independent variables) with hepatic steatosis (CAP ≥ 290 dB/m), hepatic fibrosis (LSM ≥ 8.2 kPa) and hepatic steatosis with fibrosis as binary outcomes (dependent variables, separate model for each independent and dependent variable). We additionally examined the association of UPF intake (continuous and by quintile) with continuous CAP and LSM scores. The median value of UPF within each quintile was used in a linear regression model test for a linear trend. Age- and sex-adjusted least squared mean values of CAP and LSM are displayed by UPF quintiles to depict linear trends. We examined age and sex-adjusted models for the relationship between UPF intake and hepatic outcomes. Our primary model adjusted for age, sex, physical activity, smoking, alcohol intake and intake of NOVA1 foods. In a second model, we adjusted for the HEI-2015, an index of overall diet quality, in place of NOVA1. In a third model, we adjusted for BMI, which may be in the biological pathway between UPF and hepatic phenotypes. We additionally ran a model that adjusted for waist circumference in place of BMI. Finally, we tested for the presence of an interaction in the relation between UPF consumption and hepatic phenotypes by age, sex, BMI, presence of diabetes, and HEI-2015 scores (separate models for each).

SECONDARY ANALYSIS

In a secondary analysis, we used multivariable logistic regression models (models 1–3 above) to evaluate the association of UPF intake (1-standard deviation (SD) unit) with severe hepatic steatosis and severe fibrosis (separate models for each).

Further, food intake at a given time can be considered a compositional exposure. Therefore, to estimate the relative association of substituting highly processed food with unprocessed or minimally processed food, we modeled replacing 1-SD (~2.3 servings) of NOVA4 intake with an equal unit of NOVA1 intake on the outcome of hepatic steatosis.

SENSITIVITY ANALYSIS

Some food items could belong to more than one NOVA category and may be susceptible to misclassification bias [22]. We assigned these foods to their lowest potential NOVA category in primary analyses. Hence, in sensitivity analysis we examined models assigning these foods to the highest NOVA group they could belong to (Supplemental Table 1).

Further, we completed a sensitivity analysis examining the association of UPF consumption with hepatic steatosis and fibrosis after excluding an additional 101 participants with prevalent CVD.

Results:

PARTICIPANT CHARACTERISTICS

The participant characteristics for the total sample and by quintile of UPF consumption are presented in Table 1. Prevalence of hepatic steatosis, fibrosis, and steatosis with fibrosis was 27.3%, 8.5% and 4.6%, respectively. The median age in our sample was 54, and 56% were women. The average BMI was 28.2 kg/m². High (Q5) consumers of UPF were more likely to be male, cigarette-users, and have type 2 diabetes mellitus (T2D) in χ² comparisons (all p<0.002) and have a higher BMI,lower HEI scores, and lower alcohol consumption than the lowest UPF consumers (Q1) in T-test comparisons (unequal variances, Satterthwaite p<0.001 for all). A description of dietary characteristics across quintiles of UPF intake is in Supplemental Table 2. Compared to the lowest quintile, participants in the highest quintile of UPF consumption consumed a greater percentage of energy from carbohydrates, grams of trans-fatty acids, milligrams of sodium, and grams of added sugar, but consumed less dietary fiber and alcohol.

Table 1.

Characteristics of Framingham Heart Study Third Generation participants by quartiles of ultra-process food intake

	Total n=2458	UPF Q1 n=491	UPF Q2 n=492	UPF Q3 n=492	UPF Q4 n=492	UPF Q5 n=491	Q1 vs Q5* P value
Age, years	54.4 ± 9.1	54.8 ± 8.8	53.7 ± 9.2	53.9 ± 9.2	54.2 ± 9.1	55.6 ± 9.2	0.15
Men, n (%)	1083 (44%)	192 (39%)	193 (39%)	217 (44%)	240 (49%)	241 (49%)	0.002
Current smoking, n (%)	139 (6%)	12 (2%)	28 (6%)	31 (6%)	28 (6%)	40 (8%)	<0.001
Physical Activity Index	33.9 ± 5.4	33.8 ± 5.0	34.2 ± 5.0	33.2 ± 5.2	34.3 ± 5.8	33.9 ± 5.9	0.86
Body mass index, kg/m2	28.2 ± 5.6	26.7 ± 5.2	27.3 ± 4.9	28.1 ± 5.3	28.9 ± 5.7	29.8 ± 6.0	<0.001
Diabetes, n (%)	201 (8%)	29 (6%)	23 (5%)	29 (6%)	42 (9%)	78 (16%)	<0.001
NOVA Categories, svg/day: Unprocessed and/or Minimally processed (NOVA 1)	14.4 ± 5.6	18.9 ± 5.9	14.7 ± 4.5	13.1 ± 4.8	12.5 ± 4.6	12.6 ± 5.3	<0.001
Processed culinary ingredients (NOVA 2)	1.9 ± 2.0	2.1 ± 2.2	1.9 ± 1.9	1.9 ± 1.8	1.9 ± 2.1	1.8 ± 2.0	0.04
Processed (NOVA 3)	2.3 ± 1.4	2.7 ± 1.6	2.2 ± 1.2	2.1 ± 1.2	2.2 ± 1.4	2.3 ± 1.4	<0.001
Ultra-processed (NOVA 4)	5.4 ± 3.0	3.5 ± 1.7	3.9± 1.7	4.7 ± 1.7	6.0 ± 1.9	9.3 ± 3.2	<0.001
Total energy intake, kcal/day	1891 ± 627	2112 ± 625	1760 ± 552	1736 ± 567	1832 ± 608	2008± 685	0.01
Alcohol, drinks/week	8.2 ± 8.8	9.7 ± 9.1	9.1 ± 8.4	8.1 ± 8.6	7.5 ± 8.3	6.9 ± 9.4	<0.001
2015 HEI score	76.7 ± 11.1	83.2 ± 8.6	79.9 ± 9.0	76.1 ± 10.0	73.7 ± 10.5	70.8 ± 12.4	<0.001
Controlled attenuation parameter (CAP), dB/m	258.0 ± 55.5	245.6 ± 50.2	251.3 ± 52.1	258.0 ± 55.0	263.7 ± 57.3	271.4 ± 59.1	<0.001
Liver Stiffness Measure (LSM), kPa	5.7 ± 3.8	5.7 ± 4.1	5.6 ± 2.3	5.5 ± 3.0	5.5 ± 2.3	6.4 ± 6.0	0.04
Hepatic fibrosis (LSM ≥ 8.2 kPa)	208 (9%)	39 (8%)	40 (8%)	39 (8%)	32 (7%)	58 (12%)	0.054
Hepatic steatosis (CAP ≥ 290 dB/m)	672 (27%)	88 (18%)	106 (22%)	129 (26%)	155 (32%)	194 (40%)	<0.001
Hepatic fibrosis and steatosis (LSM ≥ 8.2 kPa & CAP ≥ 290 dB/m)	113 (5%)	19 (4%)	19 (4%)	17 (4%)	22 (5%)	36 (7%)	0.03

Values are mean + SD or frequency and percent. Servings (svg) of ultra-processed foods (UPF) were adjusted for energy intake using the residual method. Hepatic steatosis defined by controlled attenuation parameter (CAP) cutoff ≥ 290 dB/m. Hepatic fibrosis is defined by liver stiffness measurement (LSM) cutoff ≥ 8.2 kPa. Kcal, kilocalories; 2015 HEI, healthy eating index.

^*Comparisons are via two-sample T-Tests for continuous variables and chi-squared tests for binary variables.

ASSOCIATIONS OF UPF INTAKE WITH HEPATIC PHENOTYPES

Table 2 lists the covariate-adjusted associations of UPF intake and hepatic phenotypes and age and sex adjusted models are presented in Supplemental Table 3. In our primary model, a 1-SD (~2.3 servings/day) increase in UPF intake was associated with 1.33 times increased odds of hepatic steatosis (95% CI: 1.21–1.46). Compared to individuals consuming the lowest amount of UPF (Q1), those consuming the highest amount (Q5) had 2.5 times higher odds of hepatic steatosis (95% CI: 1.81–3.45). We observed similar results when adjusting for the 2015 HEI. Results were attenuated but remained significant when adjusting for BMI. In our primary model, the association between UPF intake and hepatic fibrosis was not statistically significant (OR 1.15, 95% CI: 0.99–1.32). The association between UPF consumption (continuous, 1-SD unit ) and combined odds of hepatic steatosis with fibrosis was significant in the primary model and with adjustment for HEI-2015 but attenuated to the null when adjusting for BMI. Adjusting for waist circumference yielded similar results as adjusting for BMI (data not shown).

Table 2.

Associations of ultra-processed food intake (continuous and by quintile) with hepatic steatosis, hepatic fibrosis, and hepatic fibrosis and steatosis (binary)

	UPF 1-SD n=2458 OR (95% CI)	UPF Q1 n=491 OR (95% CI)	UPF Q2 n=492 OR (95% CI)	UPF Q3 n=492 OR (95% CI)	UPF Q4 n=492 OR (95% CI)	UPF Q5 n=491 OR (95% CI)
Hepatic steatosis	672 cases	88 cases	106 cases	129 cases	155 cases	194 cases
Model 1	1.33 (1.21, 1.46)	-	1.24 (0.90, 1.72)	1.51 (1.09, 2.08)	1.84 (1.33, 2.54)	2.50 (1.81, 3.45)
Model 2	1.28 (1.16, 1.40)	-	1.21 (0.88, 1.67)	1.40 (1.02, 1.93)	1.66 (1.21, 2.28)	2.16 (1.57, 2.97)
Model 3	1.14 (1.01, 1.28)	-	1.15 (0.79, 1.68)	1.27 (0.87, 1.85)	1.37 (0.94, 2.00)	1.62 (1.10, 2.38)
Hepatic fibrosis	208 cases	39 cases	40 cases	39 cases	32 cases	58 cases
Model 1	1.15 (0.99, 1.32)	-	1.01 (0.63, 1.61)	0.95 (0.58, 1.54)	0.72 (0.43, 1.20)	1.33 (0.83, 2.14)
Model 2	1.14 (0.99, 1.31)	-	1.02 (0.64, 1.63)	0.96 (0.60, 1.54)	0.72 (0.43, 1.20)	1.32 (0.83, 2.11)
Model 3	1.01 (0.87, 1.18)	-	0.96 (0.59, 1.55)	0.84 (0.51, 1.39)	0.56 (0.33, 0.95)	0.94 (0.57, 1.54)
Hepatic fibrosis and steatosis	113 cases	19 cases	19 cases	17 cases	22 cases	36 cases
Model 1	1.24 (1.04, 1.48)	-	0.93 (0.48, 1.81)	0.76 (0.38, 1.52)	0.89 (0.46, 1.74)	1.43 (0.76, 2.71)
Model 2	1.21 (1.02, 1.45)	-	0.95 (0.49, 1.83)	0.76 (0.38, 1.50)	0.87 (0.45, 1.69)	1.37 (0.73, 2.56)
Model 3	0.97 (0.79, 1.19)	-	0.89 (0.43, 1.82)	0.61 (0.29, 1.30)	0.55 (0.26, 1.16)	0.74 (0.37, 1.51)

Multivariable logistic regression models for ultra-processed food intake with hepatic steatosis, hepatic fibrosis, and hepatic fibrosis and steatosis. Hepatic steatosis defined by controlled attenuation parameter (CAP) cutoff ≥ 290 dB/m. Hepatic fibrosis is defined by liver stiffness measurement (LSM) cutoff ≥ 8.2 kPa.

Model 1 adjusted for age, sex, PAI, smoking, alcohol intake, NOVA1.

Model 2 adjusted for age, sex, PAI, smoking, alcohol intake, 2015 HEI scores.

Model 3 adjusted for age, sex, PAI, smoking, alcohol, NOVA1, BMI.

PAI, physical activity index; HEI, healthy eating index; BMI, body mass index.

In the primary model, each 1-SD unit increase in UPF intake was associated with a 6.38 (SE 1.17) unit increase in CAP score (P<0.001) and a 0.02 (SE 0.01) unit increase in LSM score (P =0.004, Table 3). Results were attenuated when adjusting for BMI. We observed a dose-response relation between UPF consumption and CAP (Figure 1, Supplemental Table 4, P_trend <0.001). However, we did not observe a linear association between UPF intake and LSM (P_trend = 0.14, Figure 1, Supplemental Table 4). There were no significant interactions in the relation between UPF intake and hepatic phenotypes by age, sex, BMI, diabetes status, or diet quality.

Table 3.

Associations of ultra-processed food intake (continuous and by quintile) with Controlled Attenuation Parameter and Liver Stiffness Measure (continuous)

	UPF 1-SD n=2458		UPF Q1 n=491	UPF Q2 n=492		UPF Q3 n=492		UPF Q4 n=492		UPF Q5 n=491
Controlled Attenuation Parameter	Beta (SE)	P	Referent	Beta (SE)	P	Beta (SE)	P	Beta (SE)	P	Beta (SE)	P
Model 1	6.38 (1.17)	<0.001	-	5.20 (3.47)	0.13	9.55 (3.57)	0.01	13.65 (3.65)	<0.001	19.49 (3.73)	<0.001
Model 2	5.28 (1.15)	<0.001	-	4.82 (3.41)	0.16	7.89 (3.47)	0.02	11.29 (3.53)	0.001	15.94 (3.65)	<0.001
Model 3	0.48 (0.96)	0.62	-	1.60 (2.79)	0.57	2.96 (2.87)	0.30	3.07 (2.95)	0.30	3.53 (3.04)	0.25
Liver Stiffness Measure *	Beta (SE)	P	Referent	Beta (SE)	P	Beta (SE)	P	Beta (SE)	P	Beta (SE)	P
Model 1	0.02 (0.01)	0.004	-	0.005 (0.02)	0.84	−0.02 (0.02)	0.39	−0.02 (0.02)	0.51	0.04 (0.03)	0.09
Model 2	0.02 (0.01)	0.01	-	0.01 (0.02)	0.69	−0.02 (0.02)	0.44	−0.01 (0.02)	0.55	0.04 (0.02)	0.10
Model 3	0.01 (0.01)	0.49	-	−0.01 (0.02)	0.79	−0.04 (0.02)	0.09	−0.05 (0.02)	0.049	−0.01 (0.02)	0.83

Multivariable logistic regression models for ultra-processed food intake with controlled attenuation parameter and liver stiffness measure.

^*log transformed

Model 1 adjusted for age, sex, PAI, smoking, alcohol intake, NOVA1.

Model 2 adjusted for age, sex, PAI, smoking, alcohol intake, 2015 HEI scores.

Model 3 adjusted for age, sex, PAI, smoking, alcohol, NOVA1, BMI.

PAI, physical activity index; HEI, healthy eating index, BMI, body mass index.

Figure 1. Age and sex adjusted least squared means of controlled attenuation parameter (CAP) and liver stiffness measure (LSM) across quintiles of ultra-processed food intake.

SECONDARY ANALYSES

We observed an increased odds of severe steatosis with higher UPF consumption (OR 1.31 per SD-increase, 95% CI: 1.18–1.44, Supplemental Table 5). This association was attenuated when adjusting for BMI. The association between UPF intake and severe fibrosis was not statistically significant (OR 1.29 [95% CI 1.00, 1.67], Supplemental Table 5). In a multivariable food substitution analysis, we observed that substituting 1-SD (~2.3 servings) of UPF (NOVA 4 group) with unprocessed (NOVA 1) foods was associated with 13% decreased odds of hepatic steatosis (Supplemental Table 5).

SENSITIVITY ANALYSIS

Re-assigning foods that may be susceptible to misclassification bias to their highest conceivable NOVA group did not significantly alter their associations with hepatic phenotypes (Supplemental Table 7). Further, associations between UPF intake and hepatic steatosis were similar when excluding individuals with prevalent CVD (Supplemental Tables 8–9).

Discussion:

In this community cohort study of middle-aged adults, we observed a cross-sectional, dose-dependent association between the consumption of UPF and the prevalence of VCTE-defined clinical hepatic steatosis, but not fibrosis. Adjusting for BMI largely attenuated our results, implying that general adiposity should be evaluated as a mediator in the relation between UPF and the pathogenesis of MASLD. In our model, replacing one daily serving of UPF (NOVA 4) with the equivalent in unprocessed or minimally processed (NOVA 1) foods was associated with 13% lower odds of hepatic steatosis. Our findings may have implications for dietary guidelines and counseling for patients at risk for MASLD.

The existing literature on UPF consumption and MASLD is evolving and heterogeneous. Many studies investigate specific nutrients or food categories, but fewer have used a uniform system such as NOVA to classify foods [23–25]. To add further complexity, studies vary in their methods of measuring hepatic phenotypes, and few have assessed fibrosis. Prior studies have used sonographic features [11], computed tomography [25] and magnetic resonance imaging [26] to define hepatic steatosis, while others rely on serological indices [27–29] or ICD codes [12]. Studies using serological tests have suggested associations of UPF intake with steatosis and fibrosis [27–29]. However, the use of serum tests alone has low specificity for hepatic fibrosis, among other limitations [29,30]. Our study is one of few to use the NOVA system and VCTE to quantify both hepatic steatosis and fibrosis.

A cross-sectional analysis in the National Health and Nutrition Examination Survey found that high consumption of UPF (>900g/day) was associated with hepatic steatosis with or without clinically significant fibrosis [31]. In alignment with this study, a prospective cohort study in a Chinese population observed an association between higher UPF consumption (>55.7g/1000kcal per day) and hepatic steatosis based on sonographic features, but did not specifically assess fibrosis (9). Another prospective study conducted in the UK Biobank observed that higher UPF consumption (per 100g) was modestly associated with hepatic steatosis and fibrosis as captured by ICD codes [12].

We observed a small association between UPF intake and continuous LSM (Table 3) but not with clinical hepatic fibrosis (Table 2). This may partly be due to the small number of participants with hepatic fibrosis in our sample (8.46%). Fibrogenesis is a dynamic process involving the synthesis and degradation of the extracellular matrix and is affected by genetic and environmental factors. Prior work has suggested associations between diet quality and VCTE-defined hepatic fibrosis [32]. Given that fibrosis is a strong prognostic factor of long-term clinical outcomes in MASLD [33], further longitudinal studies are warranted to clarify its association with UPF intake.

Diet quality is likely a key factor between UPF intake and MASLD. Numerous studies have linked overall diet quality to MASLD [32,34–36]. UPF are largely characterized as having an overall poor nutrient density. Conversely, some UPF are key sources of food components (e.g., whole grains), which have favorable associations with metabolic health [37]. Our results remained robust after adjusting for the 2015-HEI. In contrast, our findings were attenuated after adjusting for BMI. This agrees with prior reports that adiposity may mediate the relationship between UPF and MASLD risk [12,28]. BMI is known to be an independent risk factor for MASLD. However, it is well known that liver steatosis can develop without obesity, which is considered in the defining criteria of MASLD (2). While anthropometric measures of adiposity may explain some of the relation between MASLD and UPF consumption, there are likely other contributing factors.

Our study’s strengths include using the NOVA classification system to define the level of food processing. The NOVA system provides a uniform method that allows for increased reproducibility and comparison between studies. Another strength is our use of VCTE to define both hepatic steatosis and fibrosis. Limitations to consider include the predominantly white cohort, which limits the generalizability of our findings. Moreover, our cross-sectional study cannot indicate causality or account for the time burden of UPF exposure, which could impact hepatic phenotypes. Moreover, FFQs are susceptible to measurement error and recall bias. The NOVA system is also subject to several limitations, including misclassifying foods into incorrect NOVA categories. However, we completed a discordant food analysis to consider subjectivity in classification, and our results were similar. There is heterogeneity among UPF, and our study did not examine individual foods.

In conclusion, we demonstrate cross-sectional associations between UPF consumption with hepatic steatosis, which were attenuated when adjusting for BMI. Reducing the consumption of UPF and replacing them with unprocessed/minimally processed foods may reduce the odds of MASLD. Larger, longitudinal studies are warranted to investigate the association between UPF and MASLD, emphasizing progression to fibrosis.

Abbreviations:

MASLD

Metabolic Dysfunction-Associated Steatotic Liver Disease

UPF

Ultra-Processed Foods

CAP

Controlled Attenuation Parameter

LSM

Liver Stiffness Measurement

FHS

Framingham Heart Study

VCTE

Vibration-Controlled Transient Elastography

FFQ

Food Frequency Questionnaire

BMI

Body mass index

WC

Waist circumference

CVD

Cardiovascular Disease

ICD

International Statistical Classification of Diseases