Metabolic Dysfunction-Associated Fatty Liver Disease (MAFLD) is characterised by the co-occurrence of ≥5 % hepatic steatosis and metabolic disorders. This new term supersedes nonalcoholic fatty liver disease, as proposed by an international expert panel in early 2020, emphasising the crucial role of metabolic risk factors in liver steatosis(Reference Eslam, Newsome and Sarin1). Besides liver disease, individuals with MAFLD are prone to extrahepatic diseases such as CVD(Reference Lee, Lee and Kim2), type 2 diabetes mellitus, chronic kidney disease(Reference Wang, Wang and Bu3) and cognitive impairment(Reference Yu, He and Jiang4). Among patients with MAFLD, the primary cause of death is CVD(Reference Ramandi, George and Merat5), with a 65 % increased risk of adverse cardiovascular events and a higher all-cause mortality rate(Reference Kim, Konyn and Sandhu6).
The prevalence of MAFLD has escalated, affecting nearly one-quarter of adults globally and causing significant clinical and economic burdens. Unhealthy lifestyles, such as inadequate physical activity (PA), excessive caloric intake and unbalanced nutrient intake, primarily contribute to this disease(Reference Eslam, Newsome and Sarin1,Reference Eslam, Sanyal and George7) . Targeting the improvement of metabolic disorders rather than organ-specific therapy can enhance the effectiveness of treatments. Because there are currently no approved drugs for MAFLD(Reference Eslam, Sanyal and George7), dietary modifications and increased PA are the initial treatment strategies(Reference Shi and Fan8). Some dietary nutrients that influence fat metabolism, such as sulforaphane and naringin, have gained attention(Reference Li, Xie and Teng9,Reference Guo, Wang and Liu10) . Therefore, exploring dietary protective factors for MAFLD and improving lifestyle habits to reduce MAFLD occurrence are critical.
Although B-scan ultrasonography and CT are widely used in clinical imaging diagnostics, their sensitivity and specificity for diagnosing fatty liver disease are limited. Although magnetic resonance spectroscopy offers an accurate detection of hepatic steatosis, its high cost is a barrier to its widespread use. Fibroscan (Echosens, Paris, France), which uses vibration-controlled transient elastography to obtain controlled attenuation parameters (CAP) liver stiffness measurements, permits non-invasive and quantitative assessments of hepatic steatosis and fibrosis, offering superior sensitivity and specificity to B-ultrasound and CT(Reference Castera, Friedrich-Rust and Loomba11).
Carotenoids, which are phytochemicals of polyphenols abundantly found in various green leafy vegetables, colourful fruits, fungi, algae and bacteria, are primarily composed of β-carotene, α-carotene, lutein, zeaxanthin, β-cryptoxanthin and lycopene(Reference Melendez-Martinez, Mandic and Bantis12). These compounds offer several health benefits, including eye health, cognitive function, cardiovascular system health, bone health, immune function and cancer prevention(Reference Eggersdorfer and Wyss13). As potent antioxidants and anti-inflammatory nutrients, carotenoids decrease malondialdehyde, tumor necrosis factor-α, IL-6 and IL-12 levels while increasing glutathione peroxidase and catalase activity to prevent liver steatosis and hepatocyte apoptosis(Reference Yilmaz, Sahin and Bilen14). Epidemiological studies have shown that individuals consuming higher carotenoid levels, including β-carotene, α-carotene, lutein, zeaxanthin and β-cryptoxanthin, were significantly associated with a decreased risk of nonalcoholic fatty liver disease(Reference Christensen, Lawler and Mares15). However, the beneficial effects of carotenoids on MAFLD prevalence and liver fibrosis require further investigation.
This study utilised a large population from the National Health and Nutrition Examination Survey (NHANES) 2017–March 2020 to explore the association between the five types of carotenoids and MAFLD, as well as liver fibrosis evaluated by vibration-controlled transient elastography. The investigation also explored a dose–response relationship.
Methods
Study population
The study population was obtained from the NHANES, conducted from January 2017 to March 2020, which is a research program designed to assess the health and nutritional status of adults and children in the USA(Reference Ahluwalia, Dwyer and Terry16). The survey cycle yielded 15 560 participants. Participants aged 18 years were excluded (n =5867). Individuals whose dietary recall status in the two interviews was reliable and met the minimum criteria were eligible for further analysis, with 2732 participants being excluded. Among the remaining subjects, 6173 completed the liver elastography examination. One participant with missing median CAP values was excluded from this study. Subsequently, participants with missing values for education level, house income, smoking status and PA were excluded (n =992). Furthermore, eighty-two participants whose MAFLD diagnoses were inconclusive were excluded. The final sample size comprised 5098 participants.
Definition of metabolic dysfunction-associated fatty liver disease
MAFLD was diagnosed based on the presence of hepatic steatosis coupled with one or more of the following criteria: overweight/obesity, type 2 diabetes mellitus or metabolic dysregulation(Reference Eslam, Newsome and Sarin1). Metabolic dysregulation was defined as meeting at least two of the following criteria. (1) waist circumference ≥102 cm in men and ≥88 cm in women, (2) blood pressure ≥130/85 mmHg or specific drug treatment, (3) plasma TAG ≥1·70 mmol/l or specific drug treatment, (4) plasma HDL-cholesterol <1·0 mmol/l for men and <1·3 mmol/l for women or specific drug treatment, (5) prediabetes (i.e. fasting glucose level 5·6–6·9 mmol/l, or HbA1c 5·7 % to 6·4 %), (6) homeostasis model assessment-insulin resistance score ≥2·5 and (7) plasma high-sensitivity C-reactive protein level >2 mg/l. Liver steatosis was defined as a CAP score ≥248 dB/m, while liver fibrosis was confirmed with liver stiffness measurements ≥ 6·3 kPa(Reference Tian, Zhang and Xie17). CAP and liver stiffness measurements were obtained using a FibroScan® model 502 V2 Touch.
Intake of dietary carotenoids
Detailed dietary intake data were obtained from the NHANES participants through two 24-h dietary recall interviews to estimate the consumption of energy, various nutrients and beverages over the preceding day. The first interview was conducted face-to-face at the NHANES mobile examination center, and the second was conducted via telephone 3–10 d later. The average of the two dietary surveys was used for the subsequent analysis. The complex survey design weight from the second dietary interview was applied to all the analyses in this study. Dietary carotenoids, including α-carotene, lycopene, β-carotene, β-cryptoxanthin and lutein+zeaxanthin (LZ), along with total energy and alcohol intake, were extracted from the dietary interviews. The intake of energy and carotenoids underwent logarithmic transformation. The dietary intake of carotenoid was adjusted for total energy intake using an energy-adjusted model(Reference Willett, Howe and Kushi18,Reference Kan, Guo and Yuan19) .
Definition of covariate
The NHANES provides extensive participant data through standardised questionnaires and laboratory tests, including demographic details, lifestyle factors and biochemical indicators. The selected covariates for our analysis included the healthy eating index-2015 (HEI-2015) scores, age, gender, race, education level, house income, PA, smoking and alcohol consumption. HEI-2015 scores were used to assess the alignment of individual dietary intake with the Dietary Guidelines of Americans, thus reflecting the dietary quality of the participants. Detailed methods for calculating HEI-2015 scores have been documented previously(Reference Zhu and Wang20,Reference Shea, Barger and Rogers21) . In summary, data on food pattern equivalents were derived from the United States Department of Agriculture Food Patterns Equivalents Database, and HEI-2015 scores were calculated using the SAS code provided by the National Cancer Institute (https://epi.grants.cancer.gov/hei/sas-code.html). Age was categorised into four groups: 18–29, 30–49, 50–69 and ≥70 years. Race was subdivided into Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black and other races. Education level was categorised as less than high school, high school and more than high school. House income was assessed by poverty income ratio (PIR) and classified as low income (PIR < 1·30), middle income (1·3 ≤ PIR < 3·5) and high income (PIR ≥ 3·5)(Reference Tian, Zhang and Xie17). Smoking status was divided into never smokers (smoked less than 100 cigarettes in life), former smokers (smoked more than 100 cigarettes but did not smoke currently) and current smokers (smoked more than 100 cigarettes and smoke currently)(Reference Zhou, Meng and Deng22). Alcohol consumption was divided into non-drinkers (0 g/d), drinkers (> 0– < 70 g/d in males, > 0–< 56 g/d in females) and heavy drinkers (≥ 70 g/d in males, ≥ 56 g/d in females). PA was classified as low PA (<1MET-h/week), moderate PA (1-48 MET-h/week) and high PA (>48 MET-h/week)(Reference Tian, Xue and Wang23).
Statistical analysis
All statistical analyses were performed using R software (4·2·2). The ‘survey’ and ‘rms’ packages facilitated adjustments for the complex survey design to ensure national representativeness of the results. Continuous variables were presented as weighted mean ± se and weighted median (P25, P75) and categorical variables as unweighted frequencies (weighted percentages). To discern the basic characteristic differences between subjects with and without MAFLD, weighted student’s t-tests or χ 2 tests were employed. The logistic regression model examined the association between MAFLD and each of the five carotenoids (β-cryptoxanthin, β-carotene, α-carotene, LZ and lycopene). The first model was unadjusted; the second model was adjusted for age, gender and HEI-2015 scores and the third model was further adjusted for education level, race, smoking status, drinking status, PA and PIR. Carotenoid intake was also categorised based on quartile cutoffs and analyzed using logistic regression. OR and 95 % CI were calculated. Non-linear relationships were explored using a restricted cubic spline (RCS) with three (10th, 50th and 90th percentiles) or four knots (5th, 35th, 65th and 95th percentiles), with knot selection guided by the minimisation of Akaike’s Information Criterion unless specified otherwise(Reference Wei, Min and Song24). RCS analyses were executed by ‘svyglm’ and ‘rcs’ functions in the fully adjusted model, determining P for overall and P for non-linear via Wald tests and calculating 95 % CI using bootstrap methods. Statistical significance was defined as a two-tailed P value of <0·05.
Several sensitivity analyses were performed to assess the robustness of the proposed model. Initially, the diagnostic criterion for hepatic steatosis was modified to a CAP score ≥274, and the analysis was performed again. The combined effect of the five carotenoids on MAFLD was evaluated using weighted quantile sum regression (WQS), with the contribution of each carotenoid to the overall effect assessed. Carotenoid weights were obtained by bootstrapping 500 times in a training dataset and were tested using a validation dataset. The WQS index was calculated for all participants based on the weights of the five carotenoids. We primarily focused on the negative direction in the WQS analysis to avoid the reversal paradox(Reference Li, Xiao and Wu25). We then performed RCS analysis adjusted for age, gender, education level, race, smoking status, drinking status, PA, PIR and HEI-2015, using the WQS index and MAFLD as the basis, while incorporating NHANES weights. Finally, subgroup analyses accounting for age, gender, alcohol consumption, smoking status and PA, but not considering interaction, were conducted to assess whether dietary carotenoids exerted specific effects on certain population subgroups.
Results
Basic characteristics of the study participants
This study included 5098 participants, of which 55·5 % were diagnosed with MAFLD based on the baseline characteristics outlined in Table 1. Subjects without MAFLD exhibited higher HEI-2015 scores, α-carotene, β-carotene and LZ intakes and significantly lower energy consumption. Additionally, participants with MAFLD were more likely to be older, male, smokers, less physically active and have lower education levels. Finally, the subjects with MAFLD were more susceptible to liver fibrosis.
Baseline characteristics of participants included in study according to the MAFLD
MAFLD, metabolic dysfunction-associated fatty liver disease; PA, physical activity; HEI-2015, health eating index-2015; LZ, lutein+zeaxanthin.
* Data are represented as n (unweighted) (weighted proportions %) for categorical variables and as weighted mean (se) and weighted median (P25, P75) for continuous variables.
† Weighted t test and weighted χ 2 test were used to explore the difference between groups.
Association between the five carotenoids and metabolic dysfunction-associated fatty liver disease
Logistic regression analysis was used to assess the relationship between these five carotenoids and MAFLD (Table 2). In the unadjusted model, β-carotene, α-carotene and LZ were inversely associated with the occurrence of MAFLD. After adjusting for factors including age, gender, education level, race, smoking status, drinking status, PA, PIR and HEI-2015, only β-carotene (OR = 0·94, 95 % CI: 0·89, 0·99) and α-carotene (OR = 0·96, 95 % CI: 0·91, 1·00) maintained a significant association with MAFLD occurrence, indicating that per 1 log mcg/d increase in β-carotene and α-carotene was associated with 6 % and 4 % lower risk for MAFLD occurrence, respectively, although the associations were weak. Subsequent analysis utilising quartile divisions of the dietary intake of the five carotenoids revealed that being in the highest quartile of carotenoid intake was associated with a 35 % lower risk for MAFLD for β-carotene (β-carotene: OR = 0·65, 95 % CI: 0·44, 0·97).
Logistic regression results for the association between dietary carotenoids and MAFLD
MAFLD, metabolic dysfunction-associated fatty liver disease; LZ, lutein+zeaxanthin; OR, OR; 95 % CI, 95 % CI; HEI-2015, health eating index-2015.
* Model 1 was unadjusted model.
† Model 2 adjusted for age, gender and HEI-2015.
‡ Model 3 was adjusted for age, gender, race, education level, smoking status, drinking status, physical activity, poverty income ratio and HEI-2015.
These results suggest a non-linear association between carotenoid intake and the risk of MAFLD. Hence, to examine the dose–response relationship, RCS was incorporated into the fully adjusted logistic regression model. The results, displayed in Fig. 1, indicated a decrease in the MAFLD occurrence corresponded to dietary intakes of α-carotene (P overall = 0·030 and P non-linear = 0·390, Fig. 1(a)), β-carotene (P overall = 0·008 and P non-linear = 0·050, Fig. 1(b)), β-cryptoxanthin (P overall = 0·013 and P non-linear = 0·007, Fig. 1(c)), lycopene (P overall = 0·178 and P non-linear = 0·101, Fig. 1(d)) and LZ (P overall = 0·013 and P non-linear = 0·080, Fig. 1(e)) that surpassed approximately 0.29 mg, 1·05 mg, 0.15 mg, 0.83 mg and 1·91 mg, respectively.
The dose–response relationships between the dietary intake of five carotenoids and MAFLD using restricted cubic spline regression model. The model was adjusted for age, gender, race, education level, smoking status, drinking status, physical activity, poverty income ratio and HEI-2015 scores. The intake of carotenoids was log-transformed. HEI-2015, health eating index-2015; LZ, lutein+zeaxanthin; MAFLD, metabolic dysfunction-associated fatty liver disease.
Association between the five carotenoids and fibrosis
On the basis of the fully adjusted logistic regression model, no linear associations were identified between the five carotenoids and liver fibrosis (see online supplementary material, Supplemental Table S1). However, an interesting non-linear dose–response relationship was observed between β-cryptoxanthin and MAFLD (see online supplementary material, Supplemental Fig. S1), in which the occurrence of liver fibrosis rapidly increased when dietary β-cryptoxanthin intake exceeded 0·04 mg.
Sensitivity analysis
We re-evaluated the association between carotenoids and MAFLD using a CAP ≥ 274 diagnostic criterion for fatty liver disease, as suggested by prior studies. No carotenoids were associated with the occurrence of MAFLD in the fully adjusted model (see online supplementary material, Supplemental Table S2). However, RCS analysis suggested that the dietary intake of β-carotene (P overall = 0·030 and P non-linear = 0·209) and lycopene (P overall = 0·015 and P non-linear = 0·005) had a significant dose–response relationship with MAFLD occurrence (see online supplementary material, Supplemental Fig. S2).
The WQS index, representing the combined effect of the five carotenoids calculated using WQS regression, was negatively correlated with MAFLD development (OR = 0·95, 95 % CI: 0·90, 1·00, P = 0·037). The non-linear relationship between this index and the risk of MAFLD was explored by RCS (see online supplementary material, Supplemental Fig. S3(a)), with β-carotene identified as the largest contributor to this combined effect (see online supplementary material, Supplemental Fig. S3(b)).
Subgroup analyses were conducted to investigate the effects of carotenoids on different population groups, revealing a predominantly negative relationship between carotenoids and MAFLD occurrence in the 50–69 age group (see online supplementary material, Supplemental Table S3, Fig. S4). An increase in the dietary intake of carotenoids was more likely to be associated with a decreased MAFLD occurrence in females (see online supplementary material, Supplemental Table S4, Fig. S5). Sex-specific carotenoids, β-carotene (P overall = 0·027 and P non-linear = 0·036), lycopene (P overall = 0·003 and P non-linear = 0·001) and LZ (P overall < 0·001 and P non-linear = 0·011), were identified, which only had a dose–response relationship with MAFLD in females. In the smoking status stratified analysis, only an increase in α-carotene (P overall < 0·001 and P non-linear = 0·022) intake was associated with a decreased occurrence of MAFLD in current smokers, whereas LZ potentially acted as a risk factor (see online supplementary material, Supplemental Table S5, Fig. S6). Moreover, when participants were divided into drinkers and non-drinkers, carotenoids exhibited no association with the occurrence of MAFLD in either group according to logistic regression (see online supplementary material, Supplemental Table S6). However, RCS analysis revealed that increases in β-carotene (P overall = 0·015 and P non-linear = 0·086), β-cryptoxanthin (P overall = 0·039 and P non-linear = 0·021) and LZ (P overall < 0·001 and P non-linear < 0·001) were significantly associated with reduced occurrence of MAFLD in non-drinkers (see online supplementary material, Supplemental Fig. S7). Finally, subgroup analysis revealed that carotenoids were not significantly associated with MAFLD occurrence in individuals with low or moderate PA (see online supplementary material, Supplemental Table S7, Fig. S8).
Discussion
MAFLD is characterised by liver lesions accompanied by one or more metabolic disorders. MAFLD has become the most common liver disease worldwide, and its development is mainly due to unhealthy lifestyles and dietary habits, posing a serious threat to public health(Reference Fu, Wang and Qin26). Previous studies have indicated that polyphenols can mitigate MAFLD through the regulation of the gut microbiota(Reference Tang, Yuan and Li27,Reference Tan, Hu and Gong28) . In this study, we investigated the correlation between the five types of carotenoids and MAFLD using cross-sectional data from the NHANES 2017–March 2020. The results showed a weak association between β-carotene and MAFLD. After categorising the dietary intake of carotenoids into quantiles, higher quantile intakes of β-carotene and α-carotene were significantly associated with a reduced occurrence of MAFLD in the unadjusted model, with only β-carotene maintaining this association in the fully adjusted model. We also uncovered a dose–response relationship between carotenoid intake and the occurrence of MAFLD. In conclusion, our results suggest a protective effect of a higher intake of carotenoids against MAFLD.
RCS results indicated that the risk of MAFLD occurrence hardly changed or even increased as carotenoid intake increased below a certain threshold. However, when intake exceeded this value, these carotenoids became protective factors. These findings might be partially explained by the anti-inflammatory and anti-oxidant properties of carotenoids(Reference Arruda, Souza and Siqueira29,Reference de Souza Mesquita, Casagrande and Santamarina30) . Certain provitamin As, such as β-cryptoxanthin, β-carotene and α-carotene, can be converted to vitamin A in the body, thereby improving liver fat accumulation and metabolic disorders(Reference Blaner31,Reference Clugston32) . Moreover, apocarotenoids, especially non-retinoid cleavage products of carotenoids, can also interact with retinoic acid receptors(Reference Harrison, dela Sena and Eroglu33) and exert beneficial effects on the prevention of abdominal adiposity(Reference Harrison and Quadro34). Finally, carotenoids, as polyphenols, might also regulate the gut microbiota to improve the MAFLD phenotype(Reference Tang, Yuan and Li27,Reference Tan, Hu and Gong28) . Consistent with our results, epidemiological data have revealed that higher serum carotenoids are inversely associated with the prevalence and progression of nonalcoholic fatty liver disease(Reference Xiao, Chen and Zeng35).
The RCS results also revealed that higher intake of β-cryptoxanthin and lycopene was positively and negatively associated with the occurrence of liver fibrosis, respectively. However, some researchers have reported that only a higher dietary intake of α-carotene among the five types of carotenoids is positively associated with liver fibrosis(Reference Liu, Shen and Chen36). This discrepancy could be attributed to differences in the study populations, as our study was based on a larger sample size. In addition, many studies have demonstrated the hepatoprotective properties of lycopene(Reference Huang, Hsu and Chang37,Reference Ibrahim, Althagafy and Abd-Alhameed38) , consistent with our findings.
Retinol metabolism occurs in the hepatic microsomes of cytochrome P450s, which metabolise alcohol, drugs and xenobiotic pollutants(Reference Leo and Lieber39). Consequently, the toxicity of vitamin A and pro-vitamin A carotenoids can be intensified in individuals simultaneously exposed to alcohol, drugs and some pollutants, potentially leading to hepatotoxicity. Furthermore, an increase in hepatic retinol levels may be positively associated with liver fibrosis(Reference Yamane, Tanaka and Marumo40). These findings could partially explain why vitamin A exacerbated liver fibrosis during CCl4 treatment in a previous rat model study(Reference Knook, Bosma and Seifert41) and why β-cryptoxanthin increased the occurrence of liver fibrosis in our analysis.
Evidence suggests that smoking can induce a shift in carotenoids towards pro-oxidants and affect the cellular response to smoking(Reference Palozza, Simone and Mele42). Two large epidemiological studies have also reported that β-carotene supplementation was associated with a higher risk of lung cancer among smokers(Reference Blumberg and Block43,Reference Hennekens, Buring and Manson44) . These studies showed that the interaction between smoking and carotenoids could diminish or even eliminate their beneficial effects, which aligns with our findings.
The stratified analysis in our study also suggested that PA might influence the effects of carotenoids. PA is known to enhance blood antioxidant capacity, and several studies have found a positive association between higher intake of carotenoids and higher levels of PA(Reference Rousseau, Margaritis and Arnaud45–Reference Carlsohn, Rohn and Mayer48). We hypothesised that PA might amplify the effects of carotenoids and increase their utilisation, implying that higher carotenoid intake is positively associated with a lower occurrence of MAFLD in populations with more PA.
This study had several notable strengths. The dietary intake of carotenoids was adjusted for dietary energy intake according to the energy-adjusted model. Additionally, the combined effect of the five carotenoids was considered using WQS regression. Finally, the data were sourced from the NHANES to ensure accuracy and reliability.
However, our study also had several limitations. First, the causal relationship explored in a cross-sectional study might not be confirmed, necessitating further cohort studies and clinical trials. Second, some potential confounders might not have been considered. Third, our study population was from the USA, which might limit the applicability of our findings to other populations. Finally, although vibration-controlled transient elastography possesses the merits of high efficiency and a non-invasive approach to determining liver steatosis and fibrosis, liver biopsy remains the gold standard and the cut-off value of CAP and liver stiffness measurements a subject of controversy.
Conclusion
This study suggests that a higher dietary intake of carotenoid is associated with a lower occurrence of MAFLD. However, the association between certain carotenoids and the occurrence of MAFLD may vary among individuals with different age, gender and lifestyles. These findings suggest that dietary carotenoid intake may play an important role in preventing MAFLD. Further prospective studies are needed to validate these findings.
Acknowledgements
We thank the participants in the NHANES study.
Financial support
This research received no external funding.
Conflict of interest
None.
Authorship
J.Y. and P.G. conceived and advised on all aspects of the study. J.Y. and P.G. analyzed data and wrote the manuscript. P.G. and J.Y. supervised all aspects of the study. All authors discussed and commented on the manuscript.
Ethics of human subject participation
All data were obtained from secondary sources and available publicly. No protocol approval was necessary. The datasets analyzed for present study can be found in the NHANES database https://www.cdc.gov/nchs/nhanes.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980024001502
Open access
