Chronic liver disease is the leading cause of both morbidity and mortality worldwide, with 2 million deaths from the liver disease each year (1, 2). With the improvement of people’s living conditions, dietary patterns, and lifestyle changes, the incidence of non-alcoholic fatty liver disease (NAFLD) has been on the rise year by year and is now the fastest-growing cause of cirrhosis and liver cancer incidence and its associated mortality, resulting in a serious clinical and socio-economic burden (3–5). NAFLD has become the most prevalent chronic liver disease worldwide in the last two decades, and affects about a quarter of the world’s population (6). In addition to liver involvement, NAFLD is associated with an increased risk of diabetes, cardiovascular disease, cerebrovascular disease, chronic kidney disease, and extrahepatic tumors (7). In 2020, the international consensus proposed a new concept: metabolic associated fatty liver disease (MAFLD), which emphasized the role of metabolic disorders without excluding patients with an intake of alcohol or other chronic liver diseases and differed significantly from the diagnostic criteria for NAFLD (8, 9). In the real world, MAFLD diagnostic criteria are more practical than NAFLD diagnostic criteria for recognizing patients with a fatty liver at a high risk of progressive disease (10). The use of MAFLD criteria is more helpful in identifying and treating fatty liver patients at risk of hepatic fibrosis through non-invasive tests (11, 12). However, the treatment of NAFLD/MAFLD is limited to lifestyle changes, and there is a lack of clear and effective drugs (13).

Current studies have confirmed that the macronutrient and micronutrient composition of food can promote or prevent MAFLD (14–16). MAFLD is considered to be the liver manifestation of metabolic syndrome, and its occurrence is usually associated with chronic exposure to a nutrient-deficient diet (16). Wang et al. (17) showed that flavonoids can block oxidative stress by inhibiting CYP2E1 activity, thus improving insulin resistance, lipid peroxidation, and endoplasmic reticulum stress to prevent and treat NAFLD. Flavonoids are natural polyphenolic compounds that are widely found in plants and can be classified into several subtypes based on the degree of oxidation, mainly including isoflavones, flavonoids, flavanones, flavanols, flavonols, and flavan-3-ols (18). Among them, isoflavone compounds belong to phytoestrogens, mainly including daidzein, glycitein, genistein, biochanin A and formononetin (19). Equol is not a phytoestrogen, but as a metabolite of daidzein, it is sometimes included in the isoflavone class (20). The main dietary sources of human isoflavones are soybeans and soy products, containing daidzein and genistein (21). The chemical structure of daidzein is similar to that of mammalian estrogens and can act by replacing or blocking this hormone and its corresponding receptors, thus making daidzein a drug candidate with dual use (22). This makes daidzein a potential therapeutic option in estrogen-dependent diseases like prostate cancer, breast cancer, diabetes, and cardiovascular disease (23–25). Yamagata et al. (26) showed that daidzein had the potential to prevent metabolic syndrome by affecting hypertension, hyperglycemia, dyslipidemia, and atherosclerosis in patients. Another research surveyed the relationship between dietary isoflavone intake and risk of metabolic disorders in 6786 Chinese adults, which showed that total isoflavones, genistein, and daidzein intake were all negatively associated with NAFLD, hyperlipidemia, and hypertension (27). However, it is necessary for us to study the role of daidzein in MAFLD of different ethnicities. In this study, we prospectively surveyed the relationship between dietary daidzein intake and the incidence of MAFLD, hepatic steatosis, and hepatic fibrosis in participants in the National Health and Nutrition Examination Survey (NHANES) to find a potentially effective treatment for MAFLD.

A total of 1,476 participants were included in this study, with a weighted mean age of 45.24 ± 18.28 years and a weighted ratio of 50% for both males and females. The median level of daidzein intake in our study was 0.18 mg (interquartile range [IQR], 0.02–1.55 mg). Table 1 described the baseline characteristics of participants by daidzein level quartiles. Laboratory findings of participants by quartiles of daidzein intake are shown in Supplementary Table S2. The findings suggest that participants in the third quartile may be younger compared with those in the first quartile of daidzein intake levels (49.00 ± 17.79 years vs. 41.49 ± 19.00 years) and that the daidzein diet intake population concentrated between the ages of 20–65 years. The proportion of daidzein intake was significantly higher among non-Hispanic whites, those with a high school degree or more, those married/cohabiting, and those with higher household income (PIR > 3, i.e., household income above three times the poverty line) than among other populations [e.g., other Hispanics, those below high school level, those unmarried, and those with lower household income (PIR < 1)]. In addition, the proportion of daidzein intake was significantly higher in participants without pre-diabetes, diabetes, and hypertension compared to those with underlying conditions such as these conditions.

Table 2 showed the relationship between MAFLD incidence and daidzein intake evaluated by three univariate and multivariate binary logistic regression models. In the crude model, there was a negative linear correlation between daidzein intake and MAFLD incidence (OR for Q4 versus Q1 was 0.63, 95% CI: 0.46–0.86, p for trend was 0.0024). Multivariate adjusted model II also revealed a significant negative association between daidzein intake and the incidence of MAFLD with an OR of 0.98 (0.97, 0.99) and a value of p of 0.0039. Similarly, participants in the highest quartile had an OR of 0.65, 95% CI: 0.46–0.86, compared to those in the lowest quartile of daidzein intake, and statistical significance remained. In addition, we observed a dose–response relationship between daidzein intake and MAFLD incidence, and MAFLD incidence decreased with increasing daidzein intake independent of gender (Figures 2A, 3A; Table 3).

Table 2 also showed the relationship between CAP, APRI, LSM, FIB-4, NFS, FLI, HSI, and daidzein intake by three different linear regression models. We found that CAP was negatively related to daidzein intake in the crude model. Similar results were present in the model I (after adjusting for age, sex, and race, β = −0.40, 95% CI: −0.66 to −0.14, p = 0.0024) and model II (after adjusting for age, sex, race, education level, marital status, PIR, smoking, and alcohol use, β = −0.37, 95% CI: −0.63 to −0.12, p = 0.0046). Stratified by quartiles of daidzein intake levels, trend analysis of the association between daidzein intake and CAP remained significant (p for trend = 0.0054). We also found a negative correlation between HSI, FLI, NFS, and daidzein intake, and the detailed results were shown in Table 2. In model I, the β values between daidzein intake and HSI, FLI, and NFS were −0.06 (95% CI: −0.10 to −0.03, p = 0.0012), −0.24 (95% CI: −0.37 to −0.11, p = 0.0003), −0.01 (95% CI: −0.01 to 0.00, p = 0.0005), respectively. In model II, the β values between daidzein intake and HSI, FLI, and NFS were −0.06 (95% CI: −0.10 to −0.02, p = 0.0027), −0.22 (95% CI: −0.35 to −0.09, p = 0.0008), −0.01 (95% CI, −0.01to 0.00, p = 0.0006), respectively. LSM was negatively associated with daidzein intake, but p > 0.05, with no significant statistical differences. The correlation between APRI, FIB-4, and daidzein intake was not strong (although p < 0.05, all β values were 0).

In Figures 2, 3 and Table 3, we further assessed the dose–response relationships between each outcome variable and daidzein intake by generalized additive modeling and smoothed curve fitting. With this dose–response relationship, we also used a log-likelihood ratio-based test to assess the presence of a saturation threshold effect and used a two-step recursive method to determine the inflection point of the smoothing curve. In Table 3, if the log-likelihood ratio test p > 0.05, it indicates a linear correlation between daidzein intake and the outcome variable, and the curve inflection point is not significant, referring to the results of Model I. If p < 0.05, it indicates a curvilinear relationship between daidzein intake and the outcome variable, and the inflection point is significantly present, referring to the results of Model II. Therefore, Table 3 and Figure 2 show a linear relationship between MAFLD, CAP, LSM, FIB-4, NFS, FLI, HSI, and daidzein intake, except for APRI, where the presence of two inflection points is a non-linear relationship. In addition, we performed interaction tests for gender and found that the value of ps for all interactions after adjusting for variables were >0.05, which indicated that the relationship between daidzein intake and each outcome variable was not significantly dependent on gender (Figure 3; Table 3). Figure 4 further stratifies by age, sex, race, education level, marital status, PIR, alcohol use, and smoking to determine the independent relationship between MAFLD prevalence and daidzein intake. We observed that higher daidzein intake was related to lower MAFLD prevalence among those aged 20–65 years, married/cohabiting, with less than high school level, with PIR between 1 and 3, and who never smoked (Figure 4).

A previous study from China supported a negative association between daidzein intake in isoflavones and the incidence of NAFLD (27). However, a number of growing evidence has supported that MAFLD diagnostic criteria are more useful than NAFLD for recognizing patients with a fatty liver at high risk of progressive disease (10–12). It was also necessary to study the association of daidzein intake with MAFLD in different ethnic groups. We found a relatively high proportion of daidzein intake among non-Hispanic white participants who had a high school degree or higher, were married/cohabiting, had higher household income, and did not have underlying conditions such as prediabetes, diabetes, and hypertensive disease. We also observed that the incidence of MAFLD decreased with increasing levels of daidzein intake and was not affected by gender, and more specifically, it is likely that this relationship was more significant among those aged 20–65 years, married/cohabiting, with education below high school level, with a PIR between 1 and 3 and never having smoked. We also found that CAP, HSI, FLI, and NFS were negatively associated with daidzein intake, but the correlation between daidzein intake and LSM, APRI, and FIB-4 was not strong.

The previous 2001–2004 NHANES study showed that income and race were associated with differences in dietary intake, particularly among individuals from low-income households and non-Hispanic blacks, who consumed less fruits, vegetables, grains, legumes, dairy, etc. (48). Another study from 2011 to 2016 NHANES data demonstrated the association of flavonoid intake in the US population with higher socioeconomic status (49). Our findings were similar, revealing a higher daidzein intake among non-Hispanic white, higher household income participants. Daidzein is protective against certain diseases related to estrogen regulation like prostate cancer, breast cancer, diabetes, and cardiovascular diseases (23–25). In addition, studies have shown that daidzein has the potential to prevent metabolic syndromes such as cardiovascular disease and diabetes (26), which may well explain the lower risk of diabetes and hypertension in participants with a high percentage of daidzein intake in this study. In mice, daidzein may ameliorate insulin resistance in obesity via direct modulation of hepatic ab initio adipogenesis and insulin signaling, and alter adipocyte metabolism to indirectly control obesity and reduce NAFLD by regulating adipokine expression through PPAR γ (50, 51). Another study also showed that daidzein improved lipopolysaccharide (LPS)-induced damage to hepatocytes by inhibiting inflammation and oxidative stress in mice (52). The MAFLD definition emphasizes more metabolic disorders and hepatic steatosis. Thus, the above-mentioned basic research further explains the negative correlation between daidzein intake and the prevalence of MAFLD.

A prospective cohort study from Guangzhou, China, included 2,694 participants and assessed the association between dietary flavonoid intake and NAFLD status using a food frequency questionnaire with face-to-face interviews, and revealed that higher flavonoid intake was related to a lower risk of progression of NAFLD in an elderly Chinese population (53). Our study also showed that the higher the daidzein intake in the middle-aged and elderly population, the more significant the reduction in the incidence of MAFLD. Another study from Tehran showed that education level and marital status were related to cardiovascular risk factors and dietary intake (54), and married women in the Hong Kong research had significantly higher intakes of vegetables, soy products, and fish than single women (55). Smoking was also considered an independent risk factor for poor prognosis in various chronic liver diseases (56–58). The result of a recent nationally representative cohort study from Thailand showed that smoking also increases the risk of all-cause mortality in patients with NAFLD, and this association was more pronounced in women with NAFLD (59). These studies also explained well that increased daidzein intake in the nonsmoking population in our study reduced the occurrence of MAFLD.

The pathophysiology of NAFLD/MAFLD has evolved from a “first strike” characterized by increased hepatic fat to a “second strike” consisting of adipokines, inflammatory cytokines, oxidative stress, and mitochondrial dysfunction; and the “multiple strikes” hypothesis consisting of insulin resistance, inflammation, lipotoxicity, cytokine imbalance, innate immune activation and microbiota disorders in the background of genetic and environmental factors (60–62). However, the accumulation of hepatic fat caused by insulin resistance remains the first and central striking factor (63). The current diagnosis of MAFLD is the basis of the presence of hepatic steatosis, which can be assessed clinically by non-invasive (blood biomarkers and imaging) and invasive means (liver biopsy) (64). Liu et al. (65). evaluated the accuracy of five commonly non-invasive hepatic steatosis algorithms such as FLI, HSI, Visceral Adiposity Index (VAI), NAFLD-liver fat score (NAFLD-LFS), and Steato text (ST) in the diagnosis of MAFLD using NHANES III, and the result showed that FLI had the highest diagnostic performance in the diagnosis of MAFLD. In this study, we also used three non-invasive liver fat assessment methods such as FLI, HSI, and CAP assess the relationship between daidzein intake and MAFLD, and the results showed that FLI, HSI, and CAP were negatively associated with daidzein intake, implying that daidzein may improve hepatic steatosis.

Given the current state of treatment for NAFLD/MAFLD, daidzein may be a potential therapeutic option. Recent research has demonstrated that soy protein concentrates and related isoflavones play an important role in lowering blood lipids, reducing hepatic steatosis, and improving the symptoms of NAFLD (66, 67). A cross-section study of 17,685 US adults also showed a negative association between the consumption of flavonoids and FLI (68). A randomized controlled trial from Iran found that 8 weeks of soy milk consumption and a low-calorie diet showed significant improvements in blood pressure and insulin resistance-related indicators in patients with NAFLD (69). Canadian food guidelines also recommend that the diet of soy foods or supplements may be a beneficial strategy to reduce the burden of disease and prevalence of MAFLD (70).

NAFLD can develop from simple steatosis to NASH, progressive hepatic fibrosis, cirrhosis, hepatocellular carcinoma, and liver failure (71, 72). NAFLD is related to increased mortality from hepatic and cardiovascular events, in which advanced fibrosis is considered an essential predictor of prognosis in NAFLD patients (72, 73). Liver biopsy is the golden standard for diagnosis of NAFLD fibrosis but is limited by the invasive nature of liver biopsy. Therefore, non-invasive assessments such as VCTE, point shear wave elastography (pSWE), magnetic resonance elastography (MRE), FIB-4, NFS, APRI, and other tests are used for screening fibrosis in NAFLD (74). In this study, we assessed the relationship between daidzein intake and hepatic fibrosis using four methods, including FIB-4, NFS, APRI, and LSM, and the findings revealed a negative correlation between NFS and daidzein intake levels, but no strong between daidzein intake and LSM, APRI, and FIB-4. There is insufficient evidence for the ability of daidzein to alleviate hepatic fibrosis. A previous study reported a significant inhibitory effect of high-dose soy isoflavones on thioacetamide-induced hepatic fibrosis in rats, which may be related to the inhibition of hepatic stellate cell activation and proliferation (75). The future may require us to investigate the association between daidzein intake and MAFLD-associated hepatic fibrosis in greater depth.

The advantages of our study are the use of representative national NHANES data, the multi-stage stratified sampling and the collection of relevant information by trained technical staff, the relatively large size of the sample included studies and the adjustment for multiple potential confounders to improve the reliability of the findings. However, our study also has some limitations. First, the cross-sectional study design of the study could not conclude whether there was a causal association between daidzein intake and MAFLD; therefore, further validation in a prospective cohort study is necessary. Second, the diagnosis of MAFLD in this study was defined as hepatic steatosis based on a CAP ≥248 dB/m measured by VCTE rather than by liver biopsy, which inevitably leads to a biased diagnosis of MAFLD. Also, non-invasive assessment methods do not perfectly predict the presence of hepatic steatosis and fibrosis in liver biopsies. Third, in our study, we tried to incorporate as many confounding factors as possible that were relevant to the study results, but we still could not completely rule out the possibilities of other confounding factors causing bias in the conclusions. Finally, due to limitations in the design of the USDA Food and Nutrient Dietary Study Database (FNDDS), it does not have data addressing the relationship between daidzein intake and calories, fat, protein, and dietary fiber. Therefore, we also do not know whether the daidzein intake is associated with a reduced intake of calories, fat, and protein. We will further investigate the relationship between daidzein intake and energy, lipid and protein metabolism, and dietary fiber, to elaborate more deeply on the relationship between daidzein intake and MAFLD.

In conclusion, we found that the prevalence of MAFLD decreased with increasing daidzein intake and that CAP, HSI, and FLI were negatively correlated with daidzein intake, suggesting that daidzein intake may have improved hepatic steatosis. Therefore, dietary patterns of soy food or supplement consumption may be a beneficial strategy to reduce the disease burden and prevalence of MAFLD. However, the correlation between daidzein intake and hepatic fibrosis indicators such as LSM, APRI, and FIB-4 was not strong, and we may need to investigate the association between daidzein intake and MAFLD-related hepatic fibrosis more deeply in the future.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ZY, YS, and QH conceptualized the study idea and conducted the interpretation, manuscript writing, and final approval. DG, XH, and FH performed the data analysis and collection, as well as the linguistic polishing. All authors contributed to the article and approved the submitted version.

This work was supported by Jingzhou Science and Technology Bureau Plan Project (2021CC28-04 to ZY) and the Natural Science Foundation of Hubei Province (2019CFB567 to YS).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.