Daily eating frequency, nighttime fasting duration, and the risk of non-alcoholic fatty liver disease: a cross-sectional study

Abstract

Background

The understanding of daily eating frequency (DEF) and nighttime fasting duration (NFD) is limited. The aim of this research is to investigate the links between DEF, NFD, and non-alcoholic fatty liver disease (NAFLD).

Methods

The research involved 11,153 participants from the National Health and Nutrition Examination Survey (NHANES) conducted between 2005 and 2018. The evaluation of DEF and NFD was conducted through interviews focusing on dietary recalls spanning 24 h. DEF refers to the overall number of times individuals eat throughout the day, whereas NFD indicates the duration between the last and first meal of the day. The diagnosis of NAFLD was established through the application of the US fatty liver index (USFLI). A weighted logistic regression model investigated the connection between DEF, NFD, and NAFLD.

Results

After full adjustment, participants with DEF ≤ 3 times exhibited a 21% higher risk of NAFLD than those with DEF > 4.5 times (OR = 1.21, 95% CI: 1.01–1.45). Similarly, individuals with NFD ≥ 14 h were 26% more likely to develop NAFLD than those with NFD ≤ 10 h (OR = 1.26, 95% CI: 1.04–1.53). The effect of DEF on NAFLD risk was more evident in participants without T2D and with low fibrosis risk, whereas the adverse impact of NFD was particularly pronounced among those younger than 60 years.

Conclusion

DEF below 3 times and NFD exceeding 14 h were significantly linked to a heightened risk of developing NAFLD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12937-025-01213-6.

Introduction

The rise of NAFLD has become a significant contributor to liver-related health issues and fatalities as its occurrence continues to grow worldwide [1]. This phenomenon poses a considerable challenge to public health, especially given the potential for the disease to advance into severe liver disorders like cirrhosis and hepatocellular carcinoma [2]. In recent decades, nutrition research has concentrated mainly on the quantity and quality of food intake, leading to the identification of multiple dietary factors linked to the onset of NAFLD [3–8]. These findings have informed current treatment strategies, including those proposed by the American Association for the Study of Liver Diseases (AASLD), emphasizing achieving negative energy balance through dietary interventions [9].

Recently, emerging research has emphasized the significance of food intake timing in regulating metabolic health [10–12]. Circadian rhythms, which govern the timing of eating and fasting, have been shown to influence liver function [13]. Studies have demonstrated that behaviors such as irregular eating, consuming food late at night, and missing breakfast can elevate the likelihood of developing NAFLD [14–16]. A randomized trial has shown that having meals at inconsistent times (3–9 meals a day) is linked to elevated postprandial LDL cholesterol, higher insulin concentrations, and increased insulin resistance, all of which exacerbate NAFLD, compared to a regular meal pattern (6 meals per day) [17]. Conversely, research suggests that extending nighttime fasting (time-restricted eating) can improve body weight and reduce liver steatosis, potentially benefiting individuals with NAFLD or those at high risk [18]. However, circadian dietary behaviors, including daily eating frequency (DEF) and nighttime fasting duration (NFD), have not been thoroughly investigated in developing NAFLD.

This study aimed to investigate the association of DEF, NFD, with NAFLD risk to enhance our understanding of how circadian rhythms dietary behaviors affect liver wellness and to offer important perspectives on possible nutritional strategies for preventing and managing NAFLD.

Methods

Study population

This study utilized data from the National Health and Nutrition Examination Survey (NHANES), a cross-sectional, nationally representative survey to evaluate health and nutritional conditions [19]. Participants were initially recruited from the 2005–2018 survey cycles, totaling 70,190 individuals. The following groups were excluded: individuals with excessive alcohol consumption (characterized by three or more alcoholic beverages per day for men and two or more for women) or those diagnosed with hepatitis B or C (n = 1,580), individuals under the age of 18 (n = 28,043), participants with incomplete dietary data (n = 3,028), individuals with missing variables for the U.S. fatty liver index (USFLI) calculation (n = 21,310), and those with energy intakes outside the recommended ranges (men: 800–4500 kcal, women: 500–3500 kcal) (n = 918). Additionally, participants with missing covariate data (n = 4,158) were excluded. Finally, 11,153 participants remained for the follow-up analysis (Fig. 1).

Fig. 1.

Flowchart for screening the study population

Exposure

Participants’ food intake was collected through two 24-hour dietary recall interviews conducted on separate days. The initial interview took place face-to-face, with the follow-up interview occurring 3 to 10 days later via telephone. During the interviews, participants reported the times they consumed food and drink. Eating events were defined as food or drink consumed, with 50 kcal as the conservative threshold [20]. The typical frequency of eating was determined by averaging the number of reported eating instances from both interviews. Additionally, the duration of fasting overnight was determined by subtracting the time of the final meal from 24 h and the duration of the initial meal. To illustrate, should the initial meal of a participant occur at 7 a.m. and the final one at 6 p.m., the night fasting duration would be 24 − 18 + 7 = 13 h.

Main outcome

The outcome variable was NAFLD, assessed using the USFLI [21]. The USFLI is designed to assess NAFLD prevalence in the multiethnic US population within the NHANES, with a USFLI ≥ 30 indicating the presence of NAFLD. The variables included in the calculation were race (coded as 1 for race and 0 for noncompliance), age (years), Log_e GGT (IU/L), waist circumference (cm), Log_e fasting insulin (pmol/L), and Log_e fasting glucose (mg/dL). The formula used to calculate the USFLI was as follows:

1

Assessment of confounders

Potential confounders included age (years), sex (male/female), race (Mexican American/Other Hispanic/Non-Hispanic White/Non-Hispanic Black/Other Race), smoking (yes/no), drinking (yes/no), education (below high school/high school or above), physical activity (yes/no), the ratio of family income to poverty (PIR), sleep duration (hours), shiftwork (yes/no), daily energy intake (kcal), body mass index (BMI) (≤ 30 kg/m²/>30 kg/m²), alternate healthy eating index (AHEI) (healthy/unhealthy), dietary supplements (yes/no), dietary data surveyed on weekend (yes/no), cancer (yes/no), cardiovascular disease (CVD) (yes/no), hypertension (yes/no), type 2 diabetes (T2D) (yes/no), triglycerides(mmol/L), and total cholesterol(mmol/L). The top 2/5 of AHEI is defined as healthy [22]. To minimize the potential inclusion of participants with type 1 diabetes, T2D was defined as diabetes diagnosed after the age of 30 years [23].

Statistical analysis

All analyses adjusted for the complex sampling design of NHANES, incorporating sample weights, stratification, and clustering. Continuous baseline characteristics were illustrated as means ± standard deviations (SD), while categorical variables were summarized using frequencies and percentages. The Kruskal-Wallis test was used to analyze continuous variables and assess baseline differences between the groups. In contrast, the chi-square test was utilized to evaluate variation in categorical variables. To further explore the relationship between DEF, NFD, and NAFLD, we employed restricted cubic splines (RCS) for visualization. RCS was modelled with 5 knots positioned at the 5th percentile, 27.5%, 50%, 72.5%, and the 95th percentile. Analysis of variance (ANOVA) was analyzed to assess whether the correlation between DEF, NFD, and NAFLD was linear or nonlinear. DEF and NFD were categorized into five quintiles. Weighted logistic regression models were employed to assess the associations of these factors with NAFLD while controlling for a range of confounding factors, including age, sex, race, smoking, drinking, education, physical activity, PIR, sleep duration, shiftwork, daily energy intake, BMI, AHEI, dietary supplements, dietary data surveyed on the weekend, cancer, CVD, hypertension, T2D, triglycerides, and total cholesterol. Finally, we stratified the analysis based on age, sex, race, BMI, smoking, physical activity, AHEI, dietary supplements, and T2D. To evaluate the role of liver fibrosis in the associations of DEF and NFD with NAFLD, we further stratified participants by fibrosis risk according to the FIB-4 index [24], with the high-risk group merged into the intermediate-risk group due to small numbers. All statistical analyses were conducted by R version 4.4.1, and a two-sided P-value of < 0.05 was considered statistically significant.

Sensitivity analysis

In order to affirm the reliability of the findings, three sensitivity analyses were performed. First, we imputed missing covariates using the mice R package to minimize potential bias from listwise deletion, ensuring that all available information was included and enhancing the generalizability of the results. Second, we used unweighted logistic regression to assess the link between DEF, NFD, and NAFLD. This approach excluded sample weights to assess whether the associations remained consistent when the complex NHANES survey design was not accounted for. Third, we analyzed NAFLD, which is defined by the fatty liver index (FLI), to reduce potential misclassification due to the USFLI [25]. Since the FLI index incorporates BMI, we excluded this covariate from the regression model to avoid over-adjustment.

Result

Baseline characteristics

In this study, we compared demographic characteristics, lifestyle factors, health outcomes, and biomarker profiles between the non-NAFLD and NAFLD cohorts (Table 1). The results indicated that individuals in the NAFLD cohort were older, were primarily male, and showed a more excellent representation of individuals from non-Hispanic white backgrounds. Furthermore, the NAFLD cohort exhibited a greater tendency towards smoking and a reduced likelihood of alcohol consumption relative to the non-NAFLD cohort. Additionally, individuals with NAFLD showed lower levels of education, poorer dietary quality, and a greater incidence of obesity. In terms of health outcomes, the NAFLD cohort displayed a significantly higher prevalence of cancer, CVD, hypertension, and T2D, as well as elevated triglyceride levels. Additionally, participants in the NAFLD group reported a lower DEF and longer NFD compared to the non-NAFLD cohort. With respect to liver fibrosis risk assessed by FIB-4, participants with NAFLD were more likely to fall into the intermediate- and high-risk categories compared with those without NAFLD.

Table 1.

Baseline characteristics of participants

Variables	Non-NALFD	NAFLD	P-value
N	7344	3809
Age (years)	48.14 (0.32)	54.31 (0.34)	< 0.001
Male (%)	3137 (42.50)	2067 (54.37)	< 0.001
Race (%)
Mexican American	781 (10.35)	861 (21.21)	< 0.001
Other Hispanic	677 (9.05)	388 (10.04)
Non-Hispanic White	3328 (46.36)	1832 (49.75)
Non-Hispanic Black	1694 (22.84)	442 (11.47)
Other Race	864 (11.40)	286 (7.53)
Smoking (%)	2984 (40.69)	1810 (47.66)	< 0.001
Drinking (%)	858 (11.56)	297 (7.68)	< 0.001
High school or above (%)	6870 (93.66)	3346 (88.07)	< 0.001
Physical activity%)	3548 (47.27)	1779 (45.68)	0.154
PIR	2.71 (0.03)	2.47 (0.04)	< 0.001
Sleep duration (hours)	7.07 (0.02)	7.05 (0.03)	0.753
Shiftwork (%)	308 (3.81)	143 (3.52)	0.462
BMI > 30 kg/m2 (%)	1591 (21.71)	2660 (69.67)	< 0.001
Daily energy intake (kcal/d)	1974.61 (9.65)	2007.25 (12.53)	0.051
Healthy AHEI (%)	3155 (43.56)	1306 (34.60)	< 0.001
Dietary supplement (%)	3183 (43.79)	1590 (42.12)	0.136
Dietary data surveyed on weekend (%)	3839 (55.03)	1812 (50.28)	< 0.001
Cancer (%)	604 (8.50)	464 (12.65)	< 0.001
CVD (%)	523 (7.29)	529 (14.43)	< 0.001
Hypertension (%)	2085 (29.13)	1943 (51.70)	< 0.001
T2D (%)	694 (9.64)	1315 (34.95)	< 0.001
Triglycerides (mmol/L)	1.20 (0.02)	1.95 (0.03)	< 0.001
Total cholesterol (mmol/L)	4.97 (0.02)	5.01 (0.02)	0.068
Daily eating frequency (times)	4.26 (0.02)	4.12 (0.02)	< 0.001
Nighttime fasting duration (hours)	12.62 (0.03)	12.78 (0.05)	0.003
FIB-4
Low risk (< 1.3)	5135 (69.02)	2461 (64.17)	0.001
Intermediate risk (1.3–2.67)	1950 (27.56)	1163 (31.36)
High risk (≥ 2.67)	239 (3.42)	163 (4.47)

PIR poverty income ratio, BMI body mass index, AHEI alternative healthy eating index, CVD cardiovascular disease, T2D type 2 diabetes;

Continuous variables were expressed as means ± standard deviations (SD), while categorical variables were presented as frequencies and percentages

The Kruskal-Wallis test was applied to compare continuous variables, and the chi-square test was used for the evaluation of differences in categorical variables

Dose‒response relationship of DEF and NFD with NAFLD

Our analysis observed a significant inverse relationship between DEF and NFD (r = −0.26, p < 0.001, Supplementary Fig. 1). The relationship between these two variables and NAFLD risk was visualized by the RCS curve (Fig. 2). The results revealed that a lower DEF was related to a higher risk of NAFLD (p = 0.020), while an extended NFD was linked to a higher NAFLD risk (p = 0.006). However, no evidence was found for a nonlinear association between DEF or NFD and NAFLD (both p for nonlinear > 0.05).

Fig. 2.

Dose–response association for the DEF and NFD and NAFLD A DEF and NAFLD; B NFD and NAFLD; Model adjusted for age, sex, race, smoking, drinking, education, physical activity, PIR, sleep duration, shiftwork, BMI, daily energy intake, AHEI, dietary supplements, dietary data surveyed on weekend, cancer, CVD, hypertension, T2D, triglycerides, and total cholesterol PIR poverty income ratio, BMI body mass index, AHEI alternative healthy eating index, CVD cardiovascular disease, T2D type 2 diabetes;

Association of DEF and NFD with NAFLD

Based on the dose-response relationship between DEF and NFD with NAFLD, we defined the fifth quintile (more than 4.5 times) as the reference group. Participants in the lowest quintile (≤ 3 times per day) showed an increased risk of NAFLD in all models. In the fully adjusted model, the risk was 21% higher (OR = 1.21, 95% CI: 1.01 to 1.45). Furthermore, when comparing participants in the shortest NFD group (≤ 10 h) to those in the longest NFD group (≥ 14 h), a significantly higher risk of NAFLD was observed. After full adjustment, the analysis indicated a 26% higher risk for the latter group (OR = 1.26, 95% CI: 1.04 to 1.53) (Fig. 3).

Fig. 3.

Association of DEF and NFD with NAFLD A DEF and NAFLD; B NFD and NAFLD; Model 1, adjusted for age, sex, race Model 2, further adjusted for smoking, drinking, education, physical activity, PIR, sleep duration, shiftwork, BMI Model 3, further adjusted for daily energy intake, AHEI, dietary supplements, dietary data surveyed on weekend, cancer, CVD, hypertension, T2D, triglycerides, and total cholesterol PIR poverty income ratio, BMI body mass index, AHEI alternative healthy eating index, CVD cardiovascular disease, T2D type 2 diabetes;

Stratified analysis

In the stratified analysis of this study (Supplementary Tables 1–2), significant interactions were observed for DEF with T2D status (P for interaction = 0.002) and FIB-4 index (P for interaction = 0.008). The association between eating frequency and NAFLD risk was evident among participants without T2D (OR = 1.27, 95% CI: 1.03–1.56), but not among those with T2D. Similarly, among participants with low fibrosis risk (FIB-4 < 1.3), the lowest eating-frequency group showed a significantly higher risk of NAFLD (OR = 1.41, 95% CI: 1.12–1.77), whereas no significant association was observed in those with intermediate/high fibrosis risk. For NFD, a significant interaction was detected with age (P for interaction = 0.045). Among participants aged < 60 years, those in the highest quintile of NFD (≥ 14 h) had a significantly increased risk of NAFLD (OR = 1.54, 95% CI: 1.19–1.99), whereas no significant association was observed in participants aged ≥ 60 years. No other significant interactions were detected across the examined subgroups (all P for interaction > 0.05). Given prior evidence implicating aging in the development of NAFLD [26], we conducted additional stratified analyses in the NFD model using alternative age cutoffs at 30, 40, and 50 years. None of these thresholds yielded significant interactions (all P for interaction > 0.05), suggesting that age-related effect modification of the NFD–NAFLD association was not evident at these earlier cut points (Supplementary Table 3).

Sensitivity analysis

The first sensitivity analysis, imputing missing covariates using the mice R package, showed consistent results. This suggests that imputing the missing data did not substantially alter the connection between DEF, NFD, and NAFLD, thereby strengthening the robustness of the findings. The second sensitivity analysis, conducted without adjusting for the complex NHANES survey design, also revealed that the links among DEF, NFD, and NAFLD remained significant. This implies that the observed relationships are stable and not affected by the survey’s design complexity. Lastly, the third sensitivity analysis, which redefined NAFLD using the FLI index, confirmed that the ongoing association between eating frequency, fasting duration, and NAFLD risk persisted, further supporting the influence of these factors on NAFLD (Supplementary Tables 4–6).

Discussion

For the first time, this extensive cross-sectional analysis investigated the relationship between DEF, NFD, and NAFLD. The results showed that the DEF below 3 times a day and the NFD over 14 h were independently associated with an increased risk of NAFLD. This connection remained significant after adjusting for established cardiovascular and metabolic risk factors.

In stratified analyses, we further observed that the association between DEF and NAFLD was modified by T2D status and fibrosis risk. These findings suggest that the detrimental impact of infrequent eating may be more pronounced in metabolically healthier individuals and at earlier stages of liver injury, but may be masked once overt metabolic disease or advanced fibrosis has developed. For NFD, a significant interaction was detected with age: prolonged fasting (≥ 14 h) increased NAFLD risk in participants younger than 60 years but not in older adults. Additional sensitivity analyses applying alternative age thresholds (30, 40, and 50 years) did not reveal significant interactions, indicating that the modifying effect of age appears most evident around the 60-year cutoff. This finding suggests that younger and middle-aged individuals may be more susceptible to the harmful effects of prolonged fasting windows, possibly due to higher metabolic activity and greater dietary variability in earlier life stages, whereas in older adults, age-related metabolic changes and the cumulative impact of comorbidities may obscure or attenuate the influence of eating patterns. Taken together, these results imply that midlife may represent a critical window in which eating patterns exert stronger effects on NAFLD risk, and targeted dietary interventions in this age group could yield substantial preventive benefits.

Disruption of circadian rhythms is a critical factor in the development of NAFLD. The body’s biological clock is composed of a central clock in the suprachiasmatic nucleus of the hypothalamus, along with peripheral clocks found in various organs [27, 28]. The central clock modulates peripheral clocks by modulating the body’s overall circadian rhythm [29]. As the primary metabolic organ, the liver is crucial in regulating lipids and cholesterol and the metabolism of bile acids processes sensitive to circadian disturbances [29–31]. Previous studies have found that eating time resets the liver clock but does not affect the central clock rhythm [32]. Our study shows that DEF below three times and nighttime fasting over 14 h are linked to an elevated probability of developing NAFLD. This association might be intricately associated with disruptions in the circadian rhythm.

Studies have found that eating more frequently reduces obesity rates and cholesterol levels compared to individuals who eat fewer than three meals daily [33, 34]. This aligns with our findings. A clinical study has shown that irregular eating patterns significantly contribute to the onset of insulin resistance [35]. Insulin secretion has a significant circadian rhythm, and under normal circumstances, insulin levels gradually increase during the day and gradually decrease at night to adapt to food intake and metabolic needs [36]. However, irregular meal patterns, especially low DEF, may break this biological rhythm and interfere with the regular secretion pattern of insulin, causing metabolic diseases such as insulin resistance and fat metabolism disorders. In addition, meal frequency changes can influence the gut microbiome’s makeup [37]. Emerging research has shown that the gut microbiome can cause metabolic disorders by disrupting the circadian rhythms of the epithelium and liver [38, 39]. It has indicated that the dysregulation of the intestinal microbiota has significant associations with insulin resistance, inflammatory response and obesity [40]. Our research found that a lower DEF is linked to NAFLD, possibly due to the disruption in gut microbial balance and resulting metabolic issues.

Another finding of this study is that NFD (especially more than 14 h) correlates with a heightened likelihood of developing NAFLD. Previous studies have emphasized the positive impact of prolonging the duration of nocturnal fasting [41, 42]. However, the role of the NFD may be complex. Several observational studies have indicated that skipping breakfast can increase body weight, insulin resistance and NAFLD [43–45]. Excessive fasting at night aggravates the accumulation of fat in the liver by affecting the rhythmicity of lipid metabolism in the liver, resulting in fatty liver disease. Under normal circumstances, the lipid metabolism process of the liver is regulated by biological clock genes, especially the expression of genes like PER, CLOCK, and BMAL1, which are essential for metabolic processes [46]. However, extended nighttime fasting can disrupt the rhythmic expression of these biological clock genes, leading to disturbances in the liver’s circadian rhythm [47]. This disruption interferes with the daily regulation of lipid metabolism in the liver, and promoting the occurrence of NAFLD [48, 49]. Excessive duration of nighttime fasting may also affect the secretion of other hormones related to fat metabolism. For instance, growth hormone (GH) is an important hormone that regulates fat metabolism. Prolonging the fasting time at night may alter the growth hormone secretion’s circadian rhythm, thereby weakening the liver’s GH-mediated lipid metabolism [50, 51]. Moreover, glucocorticoid synthesis is closely regulated by peripheral clock genes [52]. Disruption of hormone levels may exacerbate liver fat accumulation, thereby facilitating the development of NAFLD [53].

The main advantage of this research is based on a representative large dataset and high-quality dietary data. Furthermore, the sensitivity analysis results revealed a consistent relationship between the DEF and the NFD, which were stably associated with NAFLD. However, there are several limitations in the study. Firstly, due to its cross-sectional design, it cannot establish causal relationships between dietary habits and NAFLD. Further longitudinal studies are necessary to confirm these findings. Secondly, evaluating dietary habits via 24-hour dietary recall may be affected by recall and information bias. Thirdly, the research sample mainly represents the American population, which may not fully explain racial and cultural differences, limiting the universality of these findings globally. Finally, although various confounding factors have been controlled, some potential biases may still exist and require further investigation.

Conclusion

This study’s findings indicate that a DEF below three times and a NFD over 14 h are linked to a greater likelihood of developing NAFLD. Future investigations should delve deeper into the potential pathways connecting circadian rhythm in dietary habits and the onset of NAFLD.

Abbreviations

DEF

Daily eating frequency

NFD

Nighttime fasting duration

NAFLD

Non-alcoholic fatty liver disease

NHANES

National health and nutrition examination survey

USFLI

US fatty liver index

AASLD

American association for the Study of liver diseases

PIR

Ratio of family income to poverty

BMI

Body mass index

AHEI

Alternate healthy eating index

CVD

Cardiovascular disease

T2D

Type 2 diabetes

FLI

Fatty liver index

Authors’ contributions

Y.L. conceived the study design. L.Y., X.T. and X.L. did the statistical analysis. L.Y., and Y.L. re-analyzed the results. L.Y., X.T., F.L. and Y.G. wrote the manuscript. Y.L. contributed to the revision of the manuscript. All authors provided critical revisions of the draft and approved the submitted draft. L.Y., and Y.L. are the guarantor of this work and is responsible for the integrity of the data and the accuracy of the data analysis.

Funding

This project was funding by the National Natural Science Foundation of China (award 62372142).

Data availability

The datasets generated and/or analyzed during the current study are available in the https://wwwn.cdc.gov/nchs/nhanes/Default.aspx.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the NCHS Ethics Review Board. The participants provided written informed consent to participate in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.