Introduction

Nonalcoholic fatty liver disease (NAFLD) affects approximately 25% of global population, and can proceed to more severe liver diseases such as hepatic cirrhosis or liver cancer. There are no approved drugs for treating such diseases. Thus, lifestyle changes, including dietary modifications, to achieve weight loss, are key to prevent and reduce the progression of NAFLD. Recently, Mediterranean diet, characterized by high intake of vegetables, and the combination of physical exercise, are highly proposed for preventing and managing NAFLD.

A growing body of evidence, however, showed that specific vegetables may have heterogeneous health impacts. For example, higher intake of starchy vegetables, such as potatoes, peas, and corn, were associated with greater weight gain [1]. Also, starchy vegetables might not confer similar health benefits to non-starchy vegetables on multiple health outcomes such as mortality [2], which could be partly due to their high glycemic load (GL) [3] and antioxidant loss during processing [4]. In addition, previous observational studies of vegetable intake and NAFLD are limited with mixed results [5,6,7], and did not distinguish between starchy and non-starchy vegetables. Likewise, current dietary guidelines generally treat all types of vegetables equally [8, 9], highlighting the importance to investigate potentially distinct health effects of different subgroups of vegetables. To the best of our knowledge, no studies to date have comprehensively examined whether intake of specific vegetables might be more beneficial or deleterious for preventing chronic liver disease (CLD). Therefore, we hypothesized that starchy vegetables might be positively, while non-starchy vegetables might be inversely associated with hepatic steatosis or fibrosis.

To test this hypothesis, we assessed the association between specific vegetables (i.e., starchy and non-starchy vegetables and their subgroups) and the prevalence of hepatic steatosis and fibrosis, with the use of a nationwide cross-sectional data from the US National Health and Nutrition Examination Survey (NHANES). To facilitate translations of dietary recommendations, we also conducted substitution analysis to investigate association between the replacement of starchy vegetables with non-starchy vegetables and the odds of steatosis and fibrosis.

Methods

Study population

Participants in the present study were selected from the 2017 to 2018 cycle of the NHANES, in which transient elastography (TE) examination was performed for the first time in the survey. NHANES is a continuous, cross-sectional survey conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention. Every year, a nationally representative sample of approximately 5000 persons of all ages is investigated. Details of the NHANES design and data collection methods have been described previously [10]. Written informed consent is obtained from all participants. The NCHS Research Ethics Review Board approved the NHANES study protocols (Protocol #2011–17; Protocol #2018–01). The NHANES data are publicly available at https://www.cdc.gov/nchs/nhanes/index.htm.

Participants aged ≥ 18 years were included in this analysis. In addition, we excluded individuals if they had missing dietary data (n = 873) or implausible energy intake (< 600 or > 3500 kcal/day for female and < 800 or > 4200 kcal/day for male, n = 221), or did not receive TE examination (n = 70), or had unreliable TE data (n = 522). A total of 4170 eligible participants (2133 female and 2037 male) were included in the analysis (Supplementary Fig. 1).

Dietary assessment

Diet was measured using 24-h dietary recall by trained dietary interviewers. Multiple-pass approaches were used to enhance complete and accurate food recall and decrease the respondent burden. From 1999 to 2002, only single dietary interview was administered in person at the NHANES Mobile Examination Center (MEC). Since 2003, a second interview was performed by telephone 3–10 days after the first recall. In the current analysis, most participants (n = 3662, 87.8%) had two valid 24-h dietary recalls.

Definition of vegetables was based on the US Department of Agriculture Food and Nutrition Database for Dietary Studies (Supplementary Table 1). Starchy vegetables included white potatoes (e.g., baked, boiled, mashed, scalloped, and fried potatoes and potato chips) and other starchy vegetables (e.g., immature peas, lima beans, and corn). Non-starchy vegetables included dark-green vegetables (e.g., raw and cooked broccoli, romaine, and collards), deep-yellow vegetables (e.g., carrots, pumpkin, and winter squash), and other non-starchy vegetables (e.g., tomatoes and lettuce).

Assessments of covariates

Sex, age, race/ethnicity, education, marital status, income, physical activity, and smoking were collected during household interviews using standardized questionnaires. Height, body weight, and alcohol consumption data were assessed in the MEC. The ratio of family income to poverty was used to measure family income. We used metabolic equivalent tasks (METS)-hours/week to estimate the amount of physical activity. Body mass index (BMI) was calculated as the weight in kilograms divided by the square of height in meters (kg/m2). Hepatitis B virus (HBV) infection was determined by the presence of hepatitis B surface antigen, while hepatitis C antibody or ribonucleic acid being positive indicate hepatitis C virus (HCV) infection. Diabetes was defined as self-reported diagnosis of diabetes, fasting glucose level ≥ 126 mg/dL, or a hemoglobin A1c (HbA1c) level ≥ 6.5%. Alcohol drinking status including never drinking, ever drinking, light drinking, moderate drinking, and heavy drinking was defined according to the method by the US Centers for Disease Control and Prevention [11].

Ascertainments of hepatic steatosis and fibrosis

Vibration-controlled TE was conducted in the MEC by trained health technicians, using the FibroScan® model 502 V2 Touch equipped with a medium or extra-large probe. Consistent with previous studies [10, 12], we used TE-derived controlled attenuation parameter (CAP) to define S1, S2, and S3 hepatic steatosis, with cut-off values of median CAP ≥ 274, 290, and 302 dB/m, respectively. Simultaneously, liver stiffness measurement (LSM) was derived to define significant fibrosis, advanced fibrosis, and cirrhosis, with cut-off values of median LSM ≥ 8.2, 9.7, and 13.6 kPa, respectively [10, 12]. TE examinations were considered as valid only if more than 10 LSMs were obtained after a fasting time no less than 3 h, with an interquartile range (IQR) to median ratio less than 30%. By comparing FibroScan® for the detection of steatosis and fibrosis against biopsy, the area under the receiver operating characteristic curves ranged from 0.70 to 0.89 among patients with NAFLD, and probe type and steatosis seemed not to affect LSM [12].

Statistical analysis

All analyses incorporated appropriate sampling weights, stratification, and clustering of the complex sampling design in the present study to ensure nationally representative estimates. The prevalences of hepatic steatosis and fibrosis were standardized by 2020 US Census population [10]. Consistent with previous study [2], we defined the standard serving size as 80 g for all types of vegetables, and categorized vegetable intake into 3 groups (i.e., 0, 0–1, and ≥ 1 servings/day). The multivariable logistic regression model was used to calculate the odds ratios (ORs) and 95% confidence intervals (CIs). We also presented the ORs per 1-standard deviation (SD) increment of each vegetable intake. Model 1 was adjusted for age. Model 2 was further adjusted for sex, total energy intake, race/ethnicity, education, marital status, ratio of family income to poverty, physical activity, smoking, alcohol consumption, diabetes, HBV, HCV, BMI, and intake of whole grains, refined grains, fruits, nuts, tea, coffee, fruit juices, and sugar sweetened beverages. Selection of these covariates was based on observed incomparability of characteristics of participants (Table 1) and the previously identified risk factors for the outcomes. A missing value indicator was created for each covariate in the models if possible. Linear trend was tested by treating each exposure as a continuous variable in the models. We used restricted cubic spline to evaluate the potential non-linear relationships between vegetable consumption and odds of hepatic steatosis and fibrosis.

Table 1 Age-adjusted characteristics of participants according to individual vegetable intake in NHANES (2017–2018)
Full size table

Intakes of total starchy vegetables and total non-starchy vegetables were dichotomized into “no” (i.e., no consumption) and “yes” (i.e., > 0 servings/day), resulting in four mutually exclusive groups (yes–no, no–no, yes–yes, and no–yes). These four groups were then evaluated in relation to the levels of plasma markers of glucose metabolism, lipid metabolism, and inflammation markers. The results of the multivariate linear regression analysis were presented as the percentage differences in markers comparing starchy and non-starchy vegetable consumers in different categories to the reference group [i.e., non-starchy vegetables (yes)–starchy vegetables (no)], expressed as [exp(β) – 1] × 100%. All marker concentrations were transformed using the natural logarithm to improve normality.

We used leave-one-out model [13] to assess the associations between replacement of 5% of energy from starchy vegetables with equivalent amounts of energy from non-starchy vegetables and odds of hepatic steatosis and fibrosis. In sensitivity analysis, we repeated analysis using different cut-off values to define hepatic steatosis (CAP ≥ 288 dB/m for any steatosis) [14] and fibrosis (LSM ≥ 10.1 kPa for advanced fibrosis) [15]. We also repeated analysis within persons who did not have hepatitis B and/or C. In subgroup analysis, we stratified analyses by age, sex, race/ethnicity, marital status, education level, ratio of family income to poverty, smoking status, alcohol consumption, physical activity, BMI, or diabetes. We used Wald test to examine whether interaction terms between these variables and exposures were statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC).

Results

Characteristics of participants

Of the 4170 participants (mean age 49.4 years; SD 18.3 years), 1436 (age-standardized prevalence 33.8%) were diagnosed with hepatic steatosis (CAP ≥ 290 dB/m), and 255 (5.6%) with advanced fibrosis (LSM ≥ 9.7 kPa). Potato and other non-starchy vegetable intakes contribute substantially to total vegetable intake, with proportions of energy of 57% from white potatoes, 29% from other non-starchy vegetables, 6% from other starchy vegetables, and both 4% from dark-green and deep-yellow vegetables (Supplementary Fig. 2). Participants with higher consumption of starchy vegetables were more likely to be non-Hispanic white and current smokers, had lower family income to poverty ratio, and were less physically active. These trends were reversed for the non-starchy vegetables (Table 1).

Vegetables and steatosis and fibrosis

Compared with no consumption, daily consumption of one serving of total starchy vegetables or more was associated higher prevalence of steatosis and advanced fibrosis with ORs of 1.42 (95% CI 1.15–1.76, ptrend = 0.035) and 1.68 (95% CI 1.06–2.68, ptrend = 0.001), respectively (Table 2). For the same comparison, higher consumption of potatoes was associated with higher odds of steatosis (OR = 1.30, 95% CI 1.03–1.64, ptrend = 0.043) and fibrosis (OR = 1.72, 95% CI 1.16–2.54, ptrend = 0.001). Other starchy vegetable intake was not associated with steatosis or fibrosis. Restricted cubic spline did not support a non-linear association between starchy vegetables and odds of hepatic steatosis or fibrosis (all pnon-linearity > 0.05, Fig. 1).

Table 2 Association between starchy vegetable intake and hepatic steatosis and fibrosis in NHANES (2017–2018)
Full size table
Fig. 1
figure 1

Dose–response relationship between starchy vegetable consumption and hepatic steatosis (CAP ≥ 290 dB/m) (a, b) and fibrosis (LSM ≥ 9.7 kPa) (c, d) in NHANES (2017–2018). CAP controlled attenuation parameter, CI confidence interval, LSM liver stiffness measurement, NHANES National Health and Nutrition Examination Survey, OR odds ratio. Covariates adjusted in the models were the same as those in model 2 in Table 2 (see footnote in Table 2). Solid lines indicate estimates, and dashed lines depict 95% CI

Full size image

Participants with higher intake of total non-starchy vegetables (OR = 0.73, 95% CI 0.49–1.06, ptrend = 0.008), or dark-green vegetables (OR = 0.63, 95% CI 0.40–1.00, ptrend = 0.010) had lower odds of steatosis (Table 3). However, deep-yellow vegetables (ptrend = 0.984) and other non-starchy vegetables (ptrend = 0.052) did not show any significant associations with steatosis. A non-significant association between total or any subgroups of non-starchy vegetable intake and prevalence of fibrosis was found.

Table 3 Association between non-starchy vegetable intake and hepatic steatosis and fibrosis in NHANES (2017–2018)
Full size table

Replacing 5% of energy from total starchy vegetables (OR = 0.65, 95% CI 0.44–0.97) or potatoes (OR = 0.65, 95% CI 0.43–0.97) with dark-green vegetables was associated with lower odds of steatosis (Fig. 2). However, substituting total or any specific non-starchy vegetable for starchy vegetables was not associated with prevalence of fibrosis (data not shown). Compared with those who consumed non-starchy vegetables but not starchy vegetables, participants consumed starchy vegetables but not non-starchy vegetables had higher high-sensitivity C-reactive protein (hs-CRP) and lower total cholesterol. No significant differences were observed for other markers (Supplementary Table 2).

Fig. 2
figure 2

Association between replacement of 5% of energy from total starchy vegetables and potatoes with equivalent amount of energy from non-starchy vegetables and odds of hepatic steatosis (CAP ≥ 290 dB/m) in NHANES (2017–2018). CAP controlled attenuation parameter, CI confidence interval, NHANES National Health and Nutrition Examination Survey, OR odds ratio. Covariates adjusted in the models were the same as those in model 2 in Table 2 (see footnote in Table 2)

Full size image

Sensitivity and subgroup analysis

In sensitivity analysis, the results were not appreciably changed when using different cut-off values to define steatosis and advanced fibrosis (Supplementary Table 3), or restricting analysis within persons who did not have hepatitis B and/or C (Fig. 3 and Supplementary Fig. 3). In subgroup analysis, no differential associations between each specific vegetable intake and odds of liver diseases were found according to age, sex, race/ethnicity, education level, marital status, ratio of family income to poverty, smoking status, alcohol drinking, physical activity, BMI, or diabetes (Fig. 3 and Supplementary Fig. 3).

Fig. 3
figure 3

Subgroup analysis of associations between per 1-SD increase of intake of total starchy vegetables and potatoes and odds of hepatic steatosis (CAP ≥ 290 dB/m) in NHANES (2017–2018). BMI body mass index, CAP controlled attenuation parameter, CI confidence interval, HBV hepatitis B virus, HCV hepatitis C virus, METS metabolic equivalent tasks, NHANES National Health and Nutrition Examination Survey, OR odds ratio. Covariates adjusted in the models were the same as those in model 2 in Table 2 (see footnote in Table 2). Of note, the variables examined in this figure were not adjusted. Light physical activity was defined as participants with physical activity less than 8.3 METS-h per week, and moderate to vigorous activity was defined as participants who had a physical activity of 8.3 METS-h per week or more. We excluded participants with any missing values in covariates

Full size image

Discussion

In this nationwide cross-sectional study, we generally showed that starchy and non-starchy vegetables had an opposite association with prevalence of hepatic steatosis and fibrosis. Specially, intake of total starchy vegetables or potatoes was associated with high odds of hepatic steatosis and advanced fibrosis. Conversely, higher intake of total non-starchy or dark-green vegetables was associated with lower odds of steatosis among US adults. In addition, substituting 5% of energy from dark-green vegetables for total starchy vegetables or potatoes was associated with reduced odds of hepatic steatosis. These findings did not support the current dietary guidelines [8, 9] and federal nutrition assistance programs’ positions on treating all types of vegetables equally without considering their potentially differential nutritional properties and health effects.

Observational studies of vegetables and CLD risk are limited and have yielded inconsistent results. A prospective cohort study among Korean adults showed that total vegetable intake was associated with a lower risk of NAFLD [7], whereas a cross-sectional study in Japan found no association between total vegetables and odds of NAFLD [5]. This discrepancy could be partly due to the different study design (cohort vs. cross-sectional study). Among two case–control studies in China, one reported a lower [6] and another reported a higher [16] odds of NAFLD associated with potato intake. A possible reason for this discrepancy could be due to differences in study populations (faculty members and staff vs. participants who received annual physical examinations). Among three existing studies on the association between non-starchy vegetables and NAFLD, two reported an inverse [17, 18] and one reported a non-significant [5] association. In addition, in a cross-sectional study of overweight Latino youth, higher non-starchy vegetable intake was associated with lower liver fat deposition [19], which was generally consistent with our findings.

Our results are biologically plausible. First, starchy vegetables generally have higher glycemic index (GI), and may thereby raise glucose levels faster than non-starchy vegetables, and lead to increased insulin response [20]. Consistently, we observed higher levels of fasting glucose and HbA1c among participant who consumed starchy vegetables than among those who consumed non-starchy vegetables, despite no statistical significance. Previous studies showed that diets or foods with high GI or GL levels were associated with hepatic steatosis and/or fibrosis [10, 21,22,23]. Also, in our previous study, higher dietary insulinemic potential was associated with higher prevalence of both hepatic steatosis and fibrosis [10].

Second, oxidative stress is an important mediator in CLD pathogenesis and progression. Green leafy vegetables can reduce oxidative stress because they contain several antioxidant nutrients, minerals, phytochemicals, flavonoids, phenols, or dietary fiber [18, 20]. For example, an animal study showed that dietary nitrate, which was found in high levels in green leafy vegetables, could attenuate the progression of hepatic steatosis [24]. Although potatoes are also rich in antioxidant ingredients, such as ascorbic and chlorogenic acids [25], these antioxidants would be significantly reduced in cooking treatments and peeling [26]. Likewise, the antioxidant activity of other starchy vegetables, such as peas, may be markedly decreased during the canning process [4]. Additionally, long-term intake of acrylamide-containing potato chips may induce oxidative stress in humans by activating leukocytes and increasing the generation of reactive oxygen radicals [27].

Third, chronic inflammation has been implicated in the pathogenesis of CLD. Although intake of pigmented potatoes, such as yellow and purple-flesh potatoes, can reduce inflammation, white potato consumption may increase the plasma CRP concentration [28]. A study of 1158 participants used reduced rank regression with CRP and Interleukin 6 (IL-6) as response variables to derive a pro-inflammatory dietary pattern, and found that potato was one of the pro-inflammatory foods [29]. In addition, intake of potato chips may increase IL-6 and CRP concentrations, possibly because acrylamide contained in potato chips can lower the reserves of glutathione, an important cellular antioxidant [27]. However, antioxidants contained in non-starchy vegetables, such as polyphenols and dietary nitrate, have anti-inflammatory effects [7, 18]. Our results also suggested that intake of starchy vegetables was associated with higher hs-CRP level compared with non-starchy vegetable.

Fourth, obesity is an independent risk factor for CLD. Previous studies found that high potato intake [30] was positively, whereas increased intake of non-starchy vegetables [1] was inversely associated with waist circumference and weight gain. As mentioned above, starchy vegetables have higher energy content from starch and higher GL than non-starchy vegetables [1, 20]. Therefore, the positive association between starchy vegetables and weight-related outcomes may be caused by energy and fat deposition promoted by the higher insulin response. Conversely, non-starchy vegetables have higher water and fiber content, which can promote satiety and assist in weight management. Moreover, increased intake of green leafy vegetables which are rich in nitrate conveys an anti-obesity effect. Therefore, weight gain associated with starchy vegetable intake may also be a potential mechanism for the development of hepatic steatosis and fibrosis.

Strengths of our study include a large nationally representative sample of US adults and a valid noninvasive TE detection to measure hepatic steatosis and fibrosis [12]. However, several limitations should be noted. First, measurement error was unavoidable for self-reported diet and other lifestyle factors. In addition, two 24-h recalls may not well capture long-term dietary habits, although the NHANES design used several methods, such as the dietary sampling weight and multiple-pass method, to reduce measurement error and improve estimate of dietary intake. Second, we were unable to consider cooking methods in the analysis, which may lead to confounding bias, because previous studies suggested that different cooking/processing methods might influence nutrient content and GI value of potatoes [25]. Third, cross-sectional design in the current study cannot establish causal relationship. Fourth, despite the nationally representative sample in the current study, the generalizability of our results to other populations, such as Asia–Pacific populations, is limited, given the differences in food composition and cooking or processing method across regions or countries.

In summary, our results suggest that total starchy vegetable and potato intake are associated with higher, whereas total non-starchy and dark-green vegetable intake are associated with lower odds of steatosis and/or fibrosis, and provide further food-specific guidance for preventing CLD. Given the cross-sectional design in the current study, cohort studies or clinical trials from diverse populations that carefully consider cooking/processing methods of vegetables are warranted to validate our findings.