Association between weight fluctuation and the risk of metabolic dysfunction-associated steatotic liver disease

Abstract

BACKGROUND

The global incidence of metabolic dysfunction-associated steatotic liver disease (MASLD) has increased in recent years. It has already been demonstrated that exercise and weight change are associated with the occurrence of MASLD; however, the association between weight fluctuation caused by different exercise intensities and the risk of MASLD remains to be studied.

AIM

To investigate the impact of weight fluctuation and physical activity intensity on the risk of MASLD prevalence.

METHODS

Data from the National Health and Nutrition Examination Survey database including five cycles from 2009 to 2018 were analyzed. The model included variables such as age, sex, and poverty income ratio. Weighted multivariate logistic regression was used to examine the influence of different weight fluctuation patterns within the two time intervals on the prevalence of MASLD. Nonparametric restricted cubic spline curves were used to analyze the non-linear relationship between net weight change and MASLD prevalence.

RESULTS

Among 3183 MASLD cases, the risk of MASLD increased with age for individuals transitioning from non-obese to obese or maintaining obesity, with odds ratio (OR) changing from 8.91 (95%CI: 7.40–10.88) and 11.87 (95%CI: 9.65–14.60) at 10 years before baseline to 9.58 (95%CI: 8.08–11.37) and 12.51 (95%CI: 9.33-16.78) at 25 years. Stable obesity correlated with age-dependent MASLD prevalence escalation, whereas increased physical activity attenuated MASLD risk in this group, with an OR changing from 13.64 (95%CI: 10.59–17.57) to 6.42 (95%CI: 4.24–9.72). Further analysis of the net weight changes revealed a paradoxical risk elevation with intensified physical activity during different time periods.

CONCLUSION

The risk of MASLD increases in individuals transitioning from non-obese to obese or maintaining obesity. High-intensity physical activity is beneficial for MASLD among individuals with stable obesity.

Key Words: Metabolic dysfunction-associated steatotic liver disease; Metabolic dysfunction-associated steatohepatitis; Weight fluctuation; Body mass index; Net weight; Physical activity intensity

Core Tip: The risk of metabolic dysfunction-associated steatotic liver disease (MASLD) increases with age in individuals transitioning from non-obese to obese or maintaining obesity. High-intensity physical activity is beneficial in reducing the risk of MASLD among individuals with stable obesity. Delineation of the dose-response relationship between weight fluctuation patterns and MASLD prevalence risk will facilitate the development of personalized exercise prescriptions.

INTRODUCTION

Metabolic dysfunction-associated steatotic liver disease (MASLD) is characterized by fat accumulation in the liver, ranging from simple steatotic liver disease to metabolic dysfunction-associated steatohepatitis (MASH), cirrhosis, and liver cancer. The prevalence of MASLD has been increasing worldwide. According to a global epidemiological study, MASLD was the most common cause of chronic liver disease[1]. MASLD has emerged as a pandemic-scale metabolic disorder that currently affects approximately 31.3% of the global population (Figure 1), with prevalence trajectories parallel to the obesity and type 2 diabetes epidemics[2,3]. The prevalence of MASLD has increased significantly in developed countries and it has become one of the leading causes of liver transplantation[4]. Relevant studies have shown that the prevalence of MASLD has significantly increased in Asian populations over the last few decades and is closely related to changes in lifestyle and dietary habits, including physical activity[5].

Figure 1 Recent global prevalence of metabolic dysfunction-associated steatotic liver disease and recent prevalence of metabolic dysfunction-associated steatotic liver disease by continent. Figure 1 was created using ArcMap 10.2. Maps and data were separately sourced from publicly available resources on the websites (Resource and Environmental Science Data Platform, Geoboundaries, etc.) and literature surveys. Figure 1 now incorporates data obtained from recent epidemiological studies on MASLD. MASLD: Metabolic dysfunction-associated steatotic liver disease.

Weight change refers to changes in body weight over a relative period, during which weight can have periodic increases, decreases, or persistent fluctuations, and is commonly assessed by body mass index (BMI) or in combination with waist circumference, fat mass, etc.[6]. Weight gain and obesity are major risk factors for the development and progression of MASLD[7]. Long-term weight fluctuations can trigger metabolic disturbances, inflammation, and increased fat storage, thereby increasing the risk of MASLD. Conversely, scheduled weight loss can improve liver health by reducing fat accumulation and inflammation[8]. Long-term weight gain is strongly associated with an increased prevalence of MASLD and may further exacerbate the development of MASLD through mechanisms such as interference with insulin signaling and fatty acid metabolism pathways. Studies have shown that scheduled weight loss and the maintenance of an ideal body weight can reduce the risk and progression of MASLD[9]. Weight loss and improved lifestyle behaviors such as a healthy diet and moderate exercise can also improve the pathophysiology of MASLD and liver function[10].

Exercise can be beneficial in MASLD by increasing fat oxidation and improving lipid metabolism[11]. Performing moderate to high-intensity physical activity can reduce the possibility of MASLD progression[11,12]. For example, long-term aerobic exercises such as brisk walking, running, and swimming can reduce MASLD fat deposition and are associated with improved liver function[13]. Studies have shown that a combination of aerobic exercise and resistance training is effective in reducing fat content in MASLD and improving insulin resistance[14]. In addition, 150 minutes of moderate-intensity physical activity per week has been shown to significantly reduce the risk of developing MASLD[15]. Given that patients with MASLD often present with complex metabolic comorbidities, reliance solely on BMI fails to accurately reflect specific fat distribution patterns or true metabolic health status. This study focuses on investigating age-related differentials in metabolic mechanisms of exercise intervention for MASLD. Existing research has extensively investigated lifestyle-related risk factors in MASLD pathogenesis. While given MASLD's distinct temporal cumulative effects, the interactive effects of long-term exercise patterns, body weight fluctuations, and aging on disease progression remain inadequately elucidated. Particularly, a critical knowledge gap persists regarding how weight fluctuation mediated by varying exercise intensities modulate MASLD risk across different age strata, which constitutes a pivotal yet underexplored dimension in current scientific inquiry. Our study used the National Health and Nutrition Examination Survey (NHANES) database to conduct an in-depth investigation into how weight change trajectories under varying physical activity intensities influence the prevalence risk, with the aim of providing scientifically robust and clinically actionable evidence for MASLD prevention and intervention strategies, seeking to explore the effects of weight fluctuation and physical activity intensity on the risk of MASLD by analyzing laboratory data from the NHANES database.

MATERIALS AND METHODS

Study group

The NHANES uses a complex, multi-stage, probability sampling design to collect representative health data from the United States population. The survey has been ongoing since 1999, and survey data have been released periodically over a two-year cycle. Data were obtained via face-to-face interviews, mobile medical examinations, and laboratory tests. The current study used data from the National Health and NHANES database covering five cycles from 2009 to 2018. The exclusion criteria included (1) Missing weight data or physical activity intensity data; (2) Missing outcome (MASLD) variables; and (3) Pregnant participants (during pregnancy, female patients exhibited physiological changes, such as increased metabolic demands and heightened hepatic burden; these pregnancy-specific adaptations differ from those observed in non-pregnant individuals).

Weight fluctuation assessment

In the baseline questionnaire, participants reviewed their weight data 10 years before baseline and at 25 years of age. Height and weight were measured at baseline at a Mobile Examination Center (MEC). BMI values, calculated as weight (kg) divided by the square of height, were recorded at three time points (baseline, 10 years before baseline, and 25 years of age) and noted separately as BMI baseline, BMI 10 prior, and BMI 25. Obesity was defined as. The four patterns of weight fluctuation in early adulthood (25 years of age to 10 years before baseline) and in mid-to-late adulthood (10 years before baseline to baseline) were defined as follows: stable non-obesity (all BMI 30, the reference group), obesity to non-obesity (to), non-obesity to obesity (to), and stable obesity (all BMI 30). Considering the extensive timeframe of data collected from survey respondents, there is a dearth of research examining the patterns of long-term weight change. Furthermore, the natural trajectory of weight fluctuation is difficult to summarize and analyze because of the myriad influences of geography, cultural practices, economic status, and genetic predispositions. In light of these complexities, the present study draws upon the extant literature concerning short-term randomized controlled trials on weight reduction[16,17], classifying the net weight changes of participants into five distinct categories arranged in ascending order of change magnitude: weight change < 2.5 kg (reference group), weight loss ≥ 2.5 kg, weight gain 2.5-9.9 kg, weight gain 10-19.9 kg, and weight gain ≥ 20.0 kg. This stratification offers a refined framework for analyzing patterns in long-term weight alterations, thereby enhancing the integrity and academic rigor of such investigations.

MASLD definition

The core diagnostic criteria for MASLD require the following three components: (1) Evidence of hepatic steatosis documented by imaging techniques, blood biomarkers, or liver histology; (2) Presence of at least one cardiometabolic risk factor, defined as: Overweight/obesity (BMI ≥ 23 kg/m² for Asian populations or ≥ 25 kg/m² for non-Asian populations, or elevated waist circumference), type 2 diabetes, hypertension (≥ 130 mmHg/85 mmHg or on antihypertensive therapy), hypertriglyceridemia (≥ 150 mg/dL or on lipid-lowering treatment), or low high-density lipoprotein-cholesterol (men < 40 mg/dL, women < 50 mg/dL); and (3) Exclusion of alternative etiologies including excessive alcohol consumption (men < 30 g/day, women < 20 g/day), viral hepatitis, drug-induced liver injury, and other competing causes of hepatic steatosis. These criteria collectively establish a positive diagnostic framework emphasizing metabolic dysfunction while maintaining compatibility with coexisting liver conditions[18,19]. It comprises two entities: (1) MASL; and (2) MASH. In this NHANES database study, MASLD was defined based on the United States Fatty Liver Index (USFLI) and NHANES data[20] using the following formula:

USFLI = e^{(-0.8073 × nonHispanic Black people + 0.3458 × Mexican American + 0.0093 × age + 0.6151 × ln (γ-Glutamyl transpeptidase) + 0.0249 × waist + 1.1792 × ln (insulin + 0.8242 × ln (glucose) - 14.7812))}/1 + e^{(-0.8073 × nonHispanic Black people + 0.3458 × Mexican American + 0.0093 × age + 0.6151 × ln (γ-Glutamyl transpeptidase) + 0.0249 × waist + 1.1792 × ln (insulin + 0.8242 × ln (glucose) -14.7812))} × 100

The Fatty (Steatotic) Liver Index (FLI) (SLI) is a non-invasive assessment model for hepatic steatosis based on clinical parameters including BMI, waist circumference, gamma-glutamyl transferase, triglyceride levels etc. Its derivative version, United States FLI (SLI), incorporates optimized calibration for metabolic characteristics and disease phenotypes in multi-ethnic American populations, significantly enhancing diagnostic accuracy in heterogeneous cohorts. In epidemiological investigations (e.g., NHANES-based cohort or cross-sectional studies), threshold-based classification is commonly employed for case identification, with FLI > 30 defining high-risk individuals for fatty liver. This study utilized the NHANES database to investigate MASLD. Given the ethnic heterogeneity of the study population and region-specific metabolic profiles, the United States FLI—validated for localized applicability—was adopted as the diagnostic criterion. Participants with United States FLI values > 30 were enrolled in the MASLD case group. As the data used in this study were sourced from the NHANES database, no additional external validation of this formula could be conducted. Based on the recommended values from previous studies, MASLD was defined as a USFLI of > 30.

Physical activity levels and covariates

The physical activity levels of the participants were assessed using the Global Physical Activity Questionnaire, which converts total physical activity time into a metabolic equivalent task (MET) (minute/week). After consulting the international expert practice recommendations for MASLD[21], we categorized the physical activity intensity into the following three groups based on the conversion of metabolic equivalents: (1) Low intensity (< 600 MET minute/week); (2) Moderate-intensity (600–1200 MET minute/week); and (3) High intensity (> 1200 MET minute/week) (150 minutes of moderate-intensity exercise per week or 75 minutes of high-intensity exercise per week is approximately equivalent to engaging in 600 MET minute/week).

Based on the findings of previous studies, various covariates were used, including age, sex (male, female), race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, other), marital status (married, other), education level (dropping out of high school or lower, high school graduate, drop out of college, college graduate, or higher), and household income poverty ratio level (0–1.0, 1.1–3.0, > 3.0). Alcohol consumption was classified as never drinking, abstaining, or drinking alcohol. Smoking status was categorized as never smoked, abstinent, or smoking. Disease history was considered separately for oncology (yes/no), diabetes mellitus (yes/no), and cardiovascular disease (yes/no), which included any of the following: (1) Congestive heart failure; (2) Coronary heart disease; (3) Angina pectoris; (4) Myocardial infarction; (5) Stroke; and (6) Hypertension.

Statistical analysis

In complex sample designs, sampling weights are used to adjust the representativeness of the sample. Each sample unit is assigned a weight to reflect its representation in the overall population. Clustering refers to dividing the population into several groups during the sampling process to improve sampling efficiency. Stratification involves dividing the population into different subpopulations (strata) based on specific variables and then drawing samples from each stratum to enhance the representativeness of the sample. The Rao-Scott χ² test is a statistical method designed to account for the complexities of survey sampling designs, such as stratification and clustering, when testing the independence of variables in a contingency table. This test is particularly useful for analyzing data from complex surveys like NHANES, where the sampling design can significantly affect the distribution of the test statistics. Weighted linear regression (WLR) is an extension of traditional linear regression that assigns different weights to each observation to address heteroscedasticity. This method is particularly useful when the reliability and importance of different observations vary, such as in survey data where certain samples may be over-sampled. By adjusting weights, WLR can reflect the true proportion of these samples in the overall population.

Appropriate sampling weights, clusters, and strata were used in this study. Data on demographic characteristics were presented according to weight fluctuation patterns in mid-to-late adulthood, with continuous variables expressed as weighted means and standard errors, and categorical variables expressed as frequencies and weighted percentages. Continuous variables were compared across weight fluctuation patterns using weight-adjusted (weighted) linear regression, whereas categorical variables were analyzed using the Rao–Scott χ² test to identify baseline differences.

Weighted multivariate logistic regression was used to examine the effect of different weight fluctuation patterns on the prevalence of MASLD over two-time intervals as well as the relationship between the net weight change group and the risk of MASLD prevalence. The nonlinear relationship between net weight change and the risk of MASLD was assessed using nonparametric restricted cubic spline curves. The NHANES used a complex multistage probability sampling design comprising four hierarchical stages: (1) Counties; (2) Segments; (3) Households; and (4) Individuals. This methodology inherently generates unequal selection probabilities across participants and introduces nonindependent sampling across stages. Therefore, analytical procedures must incorporate three critical structural parameters provided in the official dataset: (1) Sample weights (accounting for differential selection probabilities); (2) Strata [reflecting primary sampling units (PSUs)]; and (3) PSUs. These parameters were pre-calculated and distributed using the NHANES datasets.

For detailed computational methodologies, refer to the official NHANES analytical guidelines at https://wwwn.cdc.gov/nchs/nhanes/analyticguidelines.aspx#estimation-and-weighting-procedures. When pooling data across multiple NHANES survey cycles (2009–2010 through 2017–2018, comprising five biennial cycles), investigators must recalibrate the sampling weights to account for combined-cycle analyses. For the MEC-collected variables, the adjusted composite weight was calculated as adjusted weight (WTMEC_adj) = (1/5) × WTMEC2YR, where WTMEC2YR represents the original two-year MEC examination weight. Subsequent analytical procedures in Sequence Analysis and Statistics (SAS) require the systematic integration of three design elements: (1) WTMEC_adj; (2) Stratification variables; and (3) PSUs (SDMVPSU). These parameters must be explicitly declared in the survey procedures (such as PROC SURVEYMEANS/SURVEYREG) using the WEIGHT, STRATA, and CLUSTER statements. For comprehensive implementation guidance, consult https://wwwn.cdc.gov/nchs/nhanes/analyticguidelines.aspx#analytic-guidelines. In survival analysis, the risk function generally refers to the instant-aneous risk rate at a given time point for an individual who has not yet experienced the target event (such as MASLD diagnosis) to develop the event at that specific moment.

Sensitivity analysis was performed to test the robustness of the results. First, participants with a history of tumors, diabetes, or cardiovascular disease were excluded to reduce the potential for reverse causality due to serious illness. Subsequently, participants aged < 45 years were excluded to distinguish between the time points of early and mid-to-late adulthood. Finally, underweight participants were excluded to minimize the potential effect of being underweight on the results. Following methodological adjustments, comprehensive statistical analyses were performed on the remaining study population using recalibrated analytical protocols.

All the analyses were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC, United States) and the R software version 4.2.1 (R Core Team, Auckland, United States). Statistical significance was defined as a two-tailed P value less than 0.05.

RESULTS

Baseline characteristics

A total of 8191 participants with a mean age of 55.49 years (range: 36-80 years) were included in the analysis. Most participants were female (51.28%), non-Hispanic Caucasians (73.05%), married (64.84%), educated to college graduation level or higher (31.66%), and had a household income poverty ratio of 3.0 or higher (54.57%). More than half of the participants had never smoked (51.75%), but 76.43% drank alcohol (Table 1). Notably, the observed proportion of 76.43% in this cohort was derived through the rigorous application of MASLD diagnostic criteria, which mandated the exclusion of patients with clinically significant alcohol consumption. Specifically, alcohol intake thresholds were defined according to international guidelines (male: < 30 g/day; female: < 20 g/day). Crucially, the final analytical cohort exclusively comprised MASLD-confirmed cases without a significant history of alcohol exposure, as validated using standardized alcohol use quantification tools. Notably, 62.68% of the participants had low physical activity intensity, and the overall BMI of the participants showed an increasing trend from 24.16 at age 25 years to 27.53 at 10 years before baseline and 29.16 at baseline, with a mean weight change from 10 years before baseline of 4.56 kg. A total of 3183 participants were diagnosed with MASLD, most of whom experienced a non-obese to obese weight change from 10 years before baseline to baseline, with a general weight gain of 2.5–10 kg (Table 2). A retrospective analysis of patient weight from early adulthood to baseline revealed that the number of patients with MASLD who went from a non-obese to an obese status was much greater, with most patients having a weight gain of > 20 kg. The methodological framework of this study, encompassing participant eligibility criteria, exposure variable considerations, final cohort composition, and analytical strategies, is presented schematically in Figure 2.

Figure 2 Flow chart. Metabolic dysfunction-associated steatotic liver disease development, total number of participants, exclusion criteria, final study group, incorporation of variables, and purpose of statistical analysis. HCC: Hepatocellular carcinoma; MASH: Metabolic dysfunction-associated steatohepatitis; MASL: Metabolic dysfunction-associated steatotic liver; MASLD: Metabolic dysfunction-associated steatotic liver disease; NHANES: National Health and Nutrition Examination Survey.

Table 1 Baseline demographic characteristics, n (%).

Baseline characteristics	Overall population (n = 8191)	Patterns of weight change1				P value
		Stabilizing non-obesity (n = 4565)	Obesity to non-obesity (n = 491)	Non-obesity to obesity (n = 1348)	Stabilizing obesity (n = 1787)
Age (year)	55.49 (0.22)	55.49 (0.29)	59.40 (0.75)	52.75 (0.42)	56.53 (0.31)	< 0.0001
Gender						0.0003
Male	4036 (48.72)	2346 (48.60)	282 (57.10)	541 (43.28)	867 (50.97)
Women	4155 (51.28)	2219 (51.40)	209 (42.90)	807 (56.72)	920 (49.03)
Race						< 0.0001
Non-Hispanic White	3777 (73.05)	2184 (74.81)	219 (72.59)	540 (66.61)	834 (73.30)
Non-Hispanic Black	1568 (9.19)	720 (7.22)	99 (9.87)	326 (12.80)	423 (11.58)
Mexican	1134 (6.49)	529 (5.10)	94 (8.87)	233 (9.53)	278 (7.30)
Other Hispanics and mestizos	1712 (11.27)	1132 (12.87)	79 (8.66)	249 (11.06)	252 (7.83)
Marital status						0.0197
Married	4840 (64.84)	2783 (66.83)	277 (63.84)	759 (61.31)	1021 (62.45)
Else	3348 (35.16)	1782 (33.17)	214 (36.16)	588 (38.69)	764 (37.55)
Mean body mass index (kg/m2)
25 years old	24.16 (0.07)	22.37 (0.06)	26.24 (0.30)	24.07 (0.14)	28.432 (0.16)	< 0.0001
10 years before baseline	27.53 (0.11)	24.13 (0.06)	32.77 (0.13)	26.97 (0.08)	35.64 (0.18)	< 0.0001
Baseline	29.16 (0.12)	25.02 (0.07)	27.56 (0.13)	33.53 (0.12)	37.23 (0.19)	< 0.0001
Net weight change (kg)	4.56 (0.20)	2.50 (0.13)	-14.84 (0.49)	18.34 (0.41)	4.61 (0.47)	< 0.0001
Level of education3						< 0.0001
High school dropout or below	2110 (16.56)	1112 (15.45)	153 (20.13)	370 (18.82)	475 (16.95)
Graduate from high school	1868 (22.34)	1030 (21.77)	118 (22.59)	313 (24.08)	407 (22.50)
Drop out of college	2203 (29.43)	1107 (25.95)	147 (35.14)	413 (33.00)	536 (34.57)
University graduate or above	2003 (31.66)	1311 (36.83)	73 (22.14)	251 (24.09)	368 (25.98)
Household income poverty ratio3						< 0.0001
0-1.0	1434 (11.48)	742 (10.41)	104 (15.60)	265 (13.44)	323 (11.85)
1.1-3.0	3119 (33.95)	1659 (31.50)	220 (42.09)	518 (36.14)	722 (36.82)
> 3.0	2982 (54.57)	1797 (58.09)	123 (42.32)	444 (50.42)	618 (51.33)
Smoking status3						< 0.0001
Never smoked	4243 (51.75)	2356 (51.87)	222 (45.38)	717 (51.10)	948 (53.48)
Cessation	2373 (29.91)	1264 (28.02)	153 (33.58)	379 (29.91)	577 (34.00)
Smoking	1569 (18.34)	941 (20.11)	115 (21.05)	251 (18.99)	262 (12.52)
Drinking status3						0.0037
Never	1068 (10.63)	571 (9.76)	73 (14.25)	179 (10.96)	245 (11.73)
Cessation	1134 (12.95)	567 (11.44)	62 (13.69)	189 (13.61)	316 (16.18)
Drinking	5416 (76.43)	3077 (78.80)	325 (72.06)	884 (75.44)	1130 (72.10)
Sports						< 0.0001
Low intensity	5424 (62.68)	2921 (59.41)	337 (65.11)	907 (63.90)	1259 (69.78)
Medium intensity	1119 (15.71)	680 (17.12)	63 (16.31)	169 (15.46)	207 (12.01)
High intensity	1648 (21.62)	964 (23.47)	91 (18.57)	272 (20.64)	321 (18.21)
History of the tumor						0.0001
None	7222 (87.32)	4018 (87.01)	408 (82.00)	1236 (91.74)	1560 (86.15)
Have	962 (12.68)	543 (12.99)	81 (18.00)	112 (8.26)	226 (13.85)
History of diabetes						< 0.0001
None	6667 (87.64)	4071 (94.05)	313 (71.79)	1094 (87.10)	1189 (74.57)
Have	1276 (12.36)	401 (5.95)	158 (28.21)	205 (12.90)	512 (25.43)
History of cardiovascular disease						< 0.0001
None	4195 (55.18)	2771 (65.39)	189 (40.24)	638 (49.15)	597 (36.40)
Have	3996 (44.82)	1794 (34.61)	302 (59.76)	710 (50.85)	1190 (63.60)

¹Weight fluctuation patterns in mid-to-late adulthood (10 years before baseline to baseline).

²It provides a statistical breakdown based on the patterns of weight fluctuation from 10 years before baseline to baseline itself (as indicated in the annotations of Table 1). The weight fluctuation patterns pertained to two time points, with the endpoint being body mass index (BMI) at baseline. "Stable obesity" does not require that all three time points have a BMI greater than 30, and there is no restriction on the BMI at the age of 25, so it is possible for this value to be < 30.

³During baseline data analysis, covariate missingness was identified within the dataset. These missing values were not generalized into separate categories for subsequent weighting calculations.

It provides the baseline characteristics of the analyzed population, summarizing and counting the four patterns of weight change. We also analyzed the number and percentage of people with the four patterns of weight fluctuation under different variable conditions. This was only a portion of the baseline data analyzed in the statistical analysis (mid-to-late adulthood).

Table 2 Effect of weight fluctuation and physical activity intensity on the risk of metabolic dysfunction-associated steatotic liver disease prevalence.

Weight fluctuation assessment	Overall population		Low-intensity physical activity		Medium-intensity physical activity		High-intensity physical activity
	MASLD	OR (95%CI)	MASLD	OR (95%CI)	MASLD	OR (95%CI)	MASLD	OR (95%CI)
Weight fluctuation patterns
Mid-to-late adulthood
Stabilizing non-obesity	924	1	614	1	121	1.00 (0.71-1.40)	189	0.85 (0.64-1.13)
Obesity to non-obesity	159	1.21 (0.85-1.72)	113	1.21 (0.78-1.86)	19	0.90 (0.37-2.18)	27	1.27 (0.60-2.66)
Non-obesity to obesity	836	8.97 (7.40-10.88)1	571	9.46 (7.35-12.19)3	98	9.39 (5.85-15.07)3	167	6.24 (4.37-8.91)3
Stabilizing obesity	1264	11.87 (9.65-14.60)1	896	13.64 (10.59-17.57)3	145	11.65 (7.07-19.20)3	223	6.42 (4.24-9.72)3
Early adulthood
Stabilizing non-obesity	1051	1	702	1	140	0.96 (0.69-1.35)	209	0.84 (0.65-1.09)
Obesity to non-obesity	32	0.56 (0.26-1.22)	25	0.33 (0.15-0.71)	0	0.00 (0.00-0.00)	7	0.99 (0.33-2.94)
Non-obesity to obesity	1685	9.58 (8.08-11.37)1	1185	10.29 (8.43-12.55)	193	10.03 (6.44-15.61)	307	6.11 (4.48-8.32)
Stabilizing obesity	415	12.51 (9.33-16.78)1	282	16.14 (11.00-23.70)	50	10.05 (5.75-17.56)	83	5.74 (3.27-10.11)
Net weight change
Mid-to-late adulthood
≤ -2.5 kg	617	0.67 (0.51-0.90)	432	0.65 (0.46-0.93)	70	0.58 (0.33-1.02)	115	0.52 (0.32-0.85)
-2.5 to 2.5 kg	435	12	305	14	56	0.85 (0.45-1.60)	74	0.74 (0.43-1.26)
2.5-10 kg	846	1.37 (1.09-1.73)	570	1.39 (1.01-1.90)	109	1.11 (0.68-1.80)	167	0.98 (0.65-1.48)
10-20 kg	728	2.03 (1.57-2.62)2	485	1.92 (1.32-2.78)	96	2.47 (1.47-4.15)	147	1.20 (0.72-2.01)
> 20 kg	557	4.78 (3.37-6.76)2	402	4.93 (3.17-7.65)	52	3.10 (1.38-6.96)4	103	3.69 (2.10-6.46)4
Early adulthood
≤ -2.5 kg	163	0.97 (0.54-1.74)	119	1.02 (0.50-2.10)	16	0.62 (0.18-2.13)	28	0.72 (0.32-1.62)
-2.5 to 2.5 kg	111	1	73	1	12	0.71 (0.30-1.69)	26	0.86 (0.42-1.76)
2.5-10 kg	405	1.36 (0.92-2.02)	265	1.48 (0.87-2.51)	53	1.05 (0.53-2.09)	87	0.92 (0.50-1.70)
10-20 kg	885	2.49 (1.71-3.61)	571	2.27 (1.37-3.77)	119	2.56 (1.52-4.32)	195	2.04 (1.25-3.36)
> 20 kg	1619	5.73 (3.66-8.98)	1166	5.89 (3.41-10.17)	183	5.33 (2.88-9.89)	270	3.53 (1.96-6.34)

¹Odds ratio (metabolic dysfunction-associated steatotic liver disease prevalence risk) for two patterns of weight change, non-obesity to obesity or stable obesity, 10 years before baseline and at 25 years old, compared with stabilizing non-obesity.

²Odds ratio (metabolic dysfunction-associated steatotic liver disease prevalence risk) for two patterns of net weight change, weight gain of 10-20 kg and weight gain greater than 20 kg, 10 years before baseline to baseline, compared with weight fluctuations of less than 2.5 kg.

³Odds ratio (metabolic dysfunction-associated steatotic liver disease prevalence risk) of different exercise intensities in two weight change patterns, non-obesity to obesity or stable obesity, 10 years before baseline, compared with stabilizing non-obesity.

⁴10 years before baseline to baseline, the odds ratio (metabolic dysfunction-associated steatotic liver disease prevalence risk) was higher for high-intensity exercise than for moderate-intensity exercise, in the group that net weight change > 20 kg, compared with net weight change within 2.5 kg.

Table 2 presents the effects and correlations between the three exercise intensity levels and the prevalence of metabolic dysfunction-associated steatotic liver disease across distinct population groups. The analysis stratified participants into four distinct body weight fluctuation patterns and five net weight change categories, with temporal differentiation between mid-to-late adulthood and early adulthood observations. MASLD: Metabolic dysfunction-associated steatotic liver disease; OR: Odds ratio.

Weight fluctuation patterns and risk of MASLD

The association between weight fluctuation patterns and MASLD risk at each time point was assessed using stable non-obese individuals as the reference group. Participants in the non-obese to obese and stable obese groups had an increased risk of MASLD, with odds ratio (OR) of 8.97 (95%CI: 7.40–10.88) and 11.87 (95%CI: 9.65–14.60) at 10 years before baseline and 9.58 (95%CI: 8.08–11.37) and 12.51 (95%CI: 9.33–16.78) at 25 years old, respectively (Table 2). For net weight change, a 10–20 kg increase from 10 years before baseline to baseline increased the risk of MASLD by 103%, and a weight gain of > 20 kg increased the risk of MASLD by 4.78 times that of the reference group compared with the “within 2.5 kg” group, which had the smallest number of MASLD patients (Table 2). The OR for weight change from age 25 years to 10 years baseline were 2.49 (95%CI: 1.71–3.61) and 5.73 (95%CI: 3.66–8.98) for increases of 10–20 kg and > 20 kg, respectively (Table 2).

Effects of physical activity intensity on weight fluctuation and risk of MASLD

In the overall population analysis, the reference group was defined as having an OR of 1. For subgroup analyses stratified by physical activity intensity, a unified reference group was established as either "low-intensity physical activity practitioners with stable non-obese status" (for weight trajectory patterns) or "low-intensity physical activity practitioners exhibiting absolute weight changes within the range of –2.5 to +2.5 kg" (for net weight change analyses). For instance, among middle-to-late adulthood individuals demonstrating a transition pattern from non-obese to obese status, those engaged in high-intensity physical activity showed an adjusted OR of 6.24 (95%CI: 4.18–9.31), indicating a 6.24-fold increased risk of MASLD development compared to the reference group of stable non-obese individuals maintaining low-intensity physical activity. Statistical significance was determined through the simultaneous evaluation of the OR magnitude and its 95%CI, with associations deemed significant only when both values were above 1 (positive association) or below 1 (protective effect). The omission of P values in OR reporting emphasizes the precision of the effect size estimation over binary significance thresholds, aligned with contemporary epidemiological reporting standards. Further assessment of the effect of physical activity intensity on the association between weight fluctuation and MASLD risk revealed that the risk of MASLD in the non-obese to obese and stable obese groups was significantly reduced with increasing physical activity intensity in both early and mid-to-late adulthood compared to the stable non-obese low-intensity physical activity group. The MASLD risk decreased with an increase in physical activity intensity in the non-obese to obese population 10 years before baseline, with an OR trend of 9.46 (95%CI: 7.35–12.19), 9.39 (95%CI: 5.85–15.07), and 6.24 (95%CI: 4.37–8.91) for low-intensity, moderate-intensity, and high-intensity physical activity, respectively (Table 2). The ORs for MASLD risk in the stably obese population were 13.64 (95%CI: 10.59–17.57), 11.65 (95%CI: 7.07–19.20), and 6.42 (95%CI: 4.24–9.72) for low-intensity, moderate-intensity, and high-intensity physical activity, respectively (Table 2). A similar trend was observed in weight fluctuations in early adulthood (Table 2).

Slightly different results were obtained when exploring the effect of physical activity intensity on the association between net weight change and MASLD risk. When net weight change was considered a categorical variable, an increase in physical activity intensity increased the risk of MASLD. For example, compared with patients with a range of weight change of 2.5 kg for low-intensity physical activity, the OR for high-intensity physical activity in patients with a weight gain of > 20 kg was 3.69 (95%CI: 2.10–6.46), which was higher than that for moderate-intensity physical activity [3.10 (95%CI: 1.38–6.96)] (Table 2). However, analysis of the nonlinear association between net weight change and MASLD risk showed that for the weight gain of 5–15 kg, the risk of MASLD in mid-to-late adulthood was very close between low-intensity and medium-intensity physical activity, whereas for other values of net weight change, the risk of MASLD showed a trend of “low-intensity physical activity > medium-intensity physical activity > high-intensity physical activity” (Figure 3). In early adulthood, the risk of MASLD was lowest for high-intensity physical activity; however, after a weight gain of > 10 kg, the risk of MASLD was slightly higher for medium-intensity physical activity than for low-intensity physical activity. That’s probably because in the NHANES database, certain individuals may not have received or adhered to appropriate exercise prescriptions during the inter-survey period. Following weight gain, the decline in their physical and metabolic capacities may have exceeded expectations, rendering the initial exercise interventions not only ineffective but potentially detrimental to health. This also suggests that a weight gain of 10 kg may serve as a significant metabolic health tipping point. The lowest prevalence risk across all weight gain patterns is associated with high-intensity exercise, which may be linked to the heterogeneity of fat metabolism. Some individuals, despite being prone to weight gain, tend to accumulate fat in the hips and legs rather than in the visceral region. This heterogeneity of fat metabolism is likely related to long-term lifestyle habits and gene phenotypes. Conducting cohort studies or metabolomic analyses on individuals with high body weight but high exercise intensity may yield additional insights.

Figure 3 Arc diagram illustrates the association of exercise intensity and body fluctuation with metabolic dysfunction-associated steatotic liver disease risk based on odds ratio analysis. A: Early adulthood; B: Mid-to-late adulthood. OR: Odds ratio.

Sensitivity analysis

The results of the sensitivity analyses showed that high-intensity physical activity reduced the risk of MASLD in all three sensitivity analysis groups (no medical history, age ≥ 45 years, and non-underweight) in both non-obese and obese populations (Table 3). This result suggests that a certain level of physical activity intensity is beneficial for reducing the risk of developing MASLD in obese populations, which is consistent with the results of previous studies.

Table 3 Effect of weight fluctuation patterns and physical activity intensity on the risk of metabolic dysfunction-associated steatotic liver disease prevalence.

Weight fluctuation patterns	Overall population	Low-intensity sports	Medium-intensity sports	High-intensity sports
	OR (95%CI)	OR (95%CI)	OR (95%CI)	OR (95%CI)
No medical history
Stabilizing non-obesity	1	1	0.72 (0.42-1.21)	0.57 (0.36-0.90)
Obesity to non-obesity	0.76 (0.33-1.77)	1.10 (0.40-3.06)	0.26 (0.04-1.68)	0.10 (0.01-0.74)
Non-obesity to obesity	9.90 (6.79-14.45)	8.20 (5.23-12.84)	9.13 (4.44-18.78)	7.31 (3.84-13.92)
Stabilizing obesity	11.64 (7.63-17.76)	13.85 (8.24-23.29)	9.77 (4.63-20.65)	4.05 (2.08-7.90)
Age ≥ 45 years
Stabilizing non-obesity	1	1	0.96 (0.65-1.40)	0.86 (0.63-1.19)
Obesity to non-obesity	1.39 (0.94-2.07)	1.31 (0.82-2.09)	0.99 (0.36-2.70)	1.78 (0.78-4.04)
Non-obesity to obesity	8.39 (6.47-10.88)	9.20 (6.70-12.63)	9.21 (5.14-16.51)	4.47 (2.81-7.09)
Stabilizing obesity	11.96 (9.32-15.34)	12.68 (9.56-16.83)	16.42 (9.03-29.88)	6.37 (3.95-10.27)
Not underweight
Stabilizing non-obesity	1	1	1.00 (0.70-1.42)	0.90 (0.67-1.22)
Obesity to non-obesity	1.23 (0.86-1.77)	1.24 (0.79-1.93)	0.92 (0.37-2.27)	1.36 (0.65-2.83)
Non-obesity to obesity	8.96 (7.35-10.92)	9.55 (7.37-12.38)	9.55 (5.86-15.55)	6.38 (4.40-9.25)
Stabilizing obesity	11.97 (9.71-14.76)	13.99 (10.81-18.11)	11.82 (7.17-19.48)	6.57 (4.34-9.95)

Table 3 presents the comparative epidemiological risk profiles of metabolic dysfunction-associated steatotic liver disease prevalence across the distinct exercise intensity gradients, focusing specifically on the sensitivity analysis-excluded subpopulations. OR: Odds ratio.

DISCUSSION

The present study analyzed data from a large and nationally representative sample of NHANES 2009–2018 data and showed that the risk of MASLD increased with age in non-obese to obese and stably obese participants. Previous large-scale epidemiological studies have shown that in the adult population, weight typically stabilizes after peaking in youth but may increase with midlife and old age[22]. Weight gain in the middle and old age stages may be associated with a decrease in metabolic rate, loss of muscle mass, and changes in lifestyle and dietary habits[23]. Physical activity is an important strategy for reducing the risk of MASLD in obese patients. Exercise may improve liver function and reduce fat deposition and inflammation by increasing fat oxidation, improving insulin resistance, and promoting weight loss[24], thereby reducing the risk of MASLD in obese patients. As frequent weight fluctuations may increase the risk of MASLD, long-term adherence to moderate aerobic exercise may reduce the adverse effects of weight fluctuations on MASLD[25], leading to good liver health. The current study analyzed the effects of a combination of different age stages, weight changes, and physical activity intensities on the risk of MASLD. Combined with sensitivity analyses, our findings demonstrated that high-intensity physical activity reduced the risk of MASLD in non-obese to obese and stably obese participants.

Socioeconomic status has been significantly associated with the occurrence and severity of MASLD, with a lower socioeconomic status associated with a higher risk of developing MASLD[26,27]. Generally, people with low levels of education, low income, and occupational status are more likely to be at a high risk of obesity, metabolic syndrome, and other deficient factors associated with MASLD, which increase the risk of the disease[28]. Higher socioeconomic status is basically associated with enhanced health consciousness and increased accessibility to premium fitness facilities, giving higher income population more opportunities to engage in moderate or high intensity exercise. In addition, lower socioeconomic status may lead to fewer healthcare resources and poorer disease management, which has implications for the treatment and management of patients with MASLD[29,30]. Moreover, adverse lifestyle factors, such as high-sugar and high-fat diets, lack of physical activity, and a sedentary lifestyle, are strongly associated with the occurrence and progression of MASLD[31]. The adoption of healthy dietary patterns, such as the Mediterranean diet, healthy plant-based diets, low-fat diet, and high-fiber diet can reduce the risk of developing MASLD[32,33].

Given the complex relationship between weight fluctuation and MASLD prevalence risk, it is imperative to further investigate the association between weight fluctuation and social determinants. With the rapid improvement in material living standards over the past century, socioeconomic status has always been a fundamental determinant of midlife weight fluctuation. The primary contributors to weight fluctuation/gain among middle-aged populations underlyingly stem from chronic energy surplus caused by excessive caloric intake, particularly in developed countries such as the United States where hyperpalatable, energy-dense foods demonstrate higher availability. While in developing countries such as China, lower economic level often means lower health awareness and inadequate self-health management, which can also lead to greater midlife weight fluctuations and increases. Secondly, behavioral modifications induced by social role transitions across the life course can’t be ignored. The physiological decline in the metabolic rate during midlife coincides with social role stabilization (e.g., career establishment and shifting family responsibilities), which collectively promote sedentary lifestyle tendencies. The prevalent hedonistic orientation and substantial reduction in physical activity levels around retirement age establish behavioral foundations for weight gain. Finally, the reinforcement effect of social conformity merits emphasis. When peer groups predominantly adopt dietary indiscretion and sedentary behaviors, the implicit pressure from group norms substantially weakens individual motivation for healthy behavioral modifications. This collective sedentary trend synergizes with age-related metabolic changes, exacerbating midlife weight gain.

Potential confounders, including socioeconomic status and lifestyle factors, were considered and applied to weighted multivariate logistic regression in the current study. To test the robustness of the results, comorbidities were excluded from the inclusion criteria to minimize the potential reverse causality due to serious diseases. In addition, participants under the age of 45 years were excluded to distinguish between the time points of early adulthood (25 years old, up to 10 years before baseline) and mid-to-late adulthood (10 years before and up to baseline). The effects of being underweight on the results were excluded.

Although our study conducted separate cross-sectional analyses of weight change, relative weight change, and net weight change across different periods, notably, the weight data may have been influenced by recall bias, as the NHANES database relies on participants' self-reported information regarding their weight history. We also explored many covariates, including age, sex, race/ethnicity, marital status, education level, household income poverty ratio, and physical activity intensity. As these covariates were only available at baseline, those that changed over time were not used to capture possible confounders over time. Finally, the term risk was used solely to reflect the association between the variables and the prevalence of MASLD. Therefore, this should not be interpreted directly as representing the absolute risk of MASLD development. Owing to the lack of data at different time points, we did not assess the association between other obesity-related indicators (such as waist circumference and adiposity) and MASLD, and there were no additional data on changes in body weight over these time intervals. Further studies are required to investigate the relationship between dynamic weight fluctuation and MASLD.

CONCLUSION

In summary, this study showed that, compared with stable non-obese populations, the MASLD risk increases in non-obese to obese and stably obese participants. Although elevated physical activity intensity is beneficial in reducing the risk of MASLD in a stable obese population, because of the different net weight changes and age stages, the risk of MASLD may be approximately or slightly higher for medium-intensity physical activity than for low-intensity physical activity. While enhancing physical activity intensity potentially confers benefits to MASLD risks among individuals with stable obesity, weight fluctuation patterns and heterogeneity across different life cycle stages may diminish the differential efficacy of moderate-intensity and low-intensity physical activities in MASLD risk reduction across distinct cohorts. In certain scenarios (Table 2, early adulthood, weight change assessment > 10 kg), moderate-intensity physical activity was correlated with a marginally elevated MASLD risk. Consequently, personalized weight management and exercise intervention protocols should be developed in clinical practice. For the clinical prevention and management of MASLD, a three-tiered risk-stratified management framework is recommended: (1) Stable non-obese population (risk class I); (2) Stable obese population (risk class II); and (3) Progressive obesity population (risk class III). Evidence-based exercise interventions should be designed differently, and high-intensity interval training is recommended as the primary intervention modality for risk class II and I individuals. For subjects with significant weight fluctuations (fluctuation amplitude > 10 kg), a dynamic monitoring system should be established using personalized metabolic regulation protocols that are initiated when preset thresholds are breached. In addition, young patients should focus on maintaining lean body mass, whereas middle-aged patients require prioritized quantitative control of the visceral fat area. A comprehensive analysis elucidating the mechanistic interactions between cyclical weight fluctuation and structured exercise protocols will establish a conceptual framework essential for guiding future research priorities in precision weight management. Further research is warranted to elucidate the specific effects of different exercise intensities on the risk of MASLD across various traits. A thorough examination of the disparities in risk thresholds and contributing factors between moderate-intensity and high-intensity physical activity may provide a scientific foundation for personalized health promotion strategies informed by age and genetic predispositions.