Using human genetics to understand the epidemiological association between child obesity at different ages and MASLD

Abstract

Background:

While the relationship between adult obesity and metabolic dysfunction-associated steatotic liver disease (MASLD) is well established, the impact of childhood obesity remains unclear. We aimed to comprehensively characterize the causal effect and genetic relationships of child obesity-related exposures at different ages and MASLD.

Methods:

Using summary statistics from large-scale genome-wide association studies of child obesity-related traits and MASLD in European-ancestry populations, we performed linkage disequilibrium score regression and SUPERGNOVA to quantify overall and local genetic correlations, respectively. Univariable Mendelian randomization (UVMR) analysis infers causal relationships. Genomic structural equation modeling (GenomicSEM) and multivariable Mendelian randomization (MVMR) estimated independent genetic correlations and causal effects of each obesity-related trait on MASLD. A two-step Mendelian randomization analysis investigated the mediating pathway from childhood obesity to MASLD through adult obesity.

Result:

We observed a positive genetic correlation between BMI at age 5 years and MASLD ( $r_g$ = 0.45, P-value = 0.01), and a similar correlation for BMI at age 7 ( $r_g$ = 0.44, P-value = 0.044). A significant negative genetic correlation was identified between birth weight and MASLD ( $r_g$ = −0.23, P-value = 4.4 × 10⁻⁸). UVMR indicated increased MASLD risk was associated with overall obesity (OR_ivw = 1.40, 95% CI = 1.22–1.62) and obesity at age 8 years (OR_ivw = 1.34, 95% CI = 1.01–1.78). Birth weight had a protective effect (OR_ivw = 0.69, 95% CI = 0.61–0.78). The birth weight effect estimate attenuated to null after collectively adjusting for lifestyle factors in MVMR (OR = 1.05, 95% CI = 0.98–1.12). Mediation analysis showed that this masking effect was mediated by adult obesity.

Conclusion:

These findings suggest that the association between childhood obesity at different ages and MASLD may differ, indicating a contribution of childhood obesity to MASLD risk and confirmed that sustained obesity into adulthood is the mediating pathway.

Introduction

Over the past few decades, metabolic dysfunction-associated steatotic liver disease (MASLD), previously known as non-alcoholic fatty liver disease^[1], has become the fastest-growing cause of liver-related mortality worldwide^[2,3]. To reduce the burden of MASLD, interventions targeting weight gain trajectories from infancy and childhood have been proposed, emphasizing a life-course perspective^[2,4,5]. However, evidence identifying the sensitive early-life periods of weight gain for MASLD development is needed to guide prevention strategies.

Previous epidemiological studies consistently report associations between weight in infancy/childhood and later MASLD, but findings on specific risk periods are inconsistent^[6,7]. For birth weight, two cohort studies found a negative association with later MASLD risk^[8,9], while a population-based cohort of 1170 individuals found no association^[10]. Regarding childhood adiposity trajectories, one study suggested faster adiposity accretion in the first 5 years predicts later hepatic fat^[11], but another involving children aged 3 months to 10 years reported positive associations only for BMI at ages 7 and 10 years with MASLD, not at younger ages^[9]. These inconsistencies in identifying critical periods challenge determining optimal intervention timing for MASLD prevention via childhood weight management. Future studies should address confounding and provide stronger causal inferences.

Furthermore, clarifying the role of adult adiposity in the relationship between early-life adiposity and the development of MASLD is a critical consideration for interventions targeting infancy and childhood weight gain. If adult adiposity completely mediates the effect of childhood adiposity on later-life MASLD, the unique value of such interventions could be diminished. A previous study^[12] suggested that the associations between conditional weight gain after age 2 years and MASLD outcomes were attenuated after adjusting for adult BMI. However, in the same study^[12], the relationship between anthropometric measurements at birth and 2 years old and MASLD remained significant after adjusting for adult BMI. The inconsistent results across different periods suggest that the role of adult adiposity in the relationship between childhood adiposity and MASLD may vary depending on sensitive periods, and additional research is needed to clarify these relationships.

Instead of relying solely on large-scale cohort studies, we can now utilize methods from genetic epidemiology to assess the impact of age-specific adiposity during infancy and childhood on MASLD. Additionally, by leveraging genome-wide association studies (GWASs) of age-specific adiposity, we can explore the role of adult adiposity in the potential causal relationship. While existing evidence has suggested a significant genetically predicted association between infancy and childhood adiposity and later-life MASLD using Mendelian randomization (MR)^[13] and linkage disequilibrium score regression (LDSC)^[14], none of the existing studies have systematically investigated the shared and distinct genetic etiology underlying adiposity at different time points in infancy and childhood, and MASLD. Emerging research on MASLD treatment highlights key targets such as the AKAP1/PKA/GPAT1 signaling complex and the immune regulator MAVS, which play important roles in lipid metabolism and steatosis^[15,16]. Building on these insights, a novel strategy has been proposed that simultaneously enhances PPARα activity and inhibits ERRα to restore metabolic balance^[17]. Although these strategies have not yet been directly applied in clinical settings, they highlight promising translational opportunities for MASLD treatment and prevention.

Therefore, in this study, we used the largest GWASs to date on age-specific infancy and childhood obesity to (1) investigate whether genetic links exist between age-specific infancy and childhood adiposity and MASLD; (2) examine which age-specific adiposity in infancy/childhood is sensitive for the development of MASLD; (3) estimate the independent causal effects and/or genetic links of each childhood obesity-related trait, accounting for adult adiposity, tobacco use, alcohol consumption, sleep duration, and physical activity as confounders; and (4) reveal whether causal inference could be mediated by adult adiposity. Figure 1 illustrates the overall design of the study. The work has been reported in line with the STROCSS criteria^[20] (see in Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/F891). In compliance with ethical publishing standards, this study adheres to the TITAN 2025 guidelines for AI transparency in scientific writing^[21].

Figure 1.

Flowchart on the overall study design.

Materials and methods

Data summary

This secondary analysis utilized summary statistics from published GWASs in European-ancestry populations. Summary statistics were obtained from publicly available GWASs for obesity-related traits (including overall childhood BMI^[22], age-specific childhood BMI^[23], birth weight^[24], and adult BMI^[25]), MASLD^[26] and lifestyle factors (including tobacco use^[27], alcohol consumption^[27], sleep duration^[28], physical activity^[29], educational attainment^[30], and household income^[31]). The characteristics of each included dataset are detailed in Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/F891. The description of these data is shown in the Supplemental Digital Content methods, available at: http://links.lww.com/JS9/F890.

HIGHLIGHTS

• We discovered compelling evidence supporting a significant shared genetic basis, both globally and locally, indicating shared biology linking infancy/childhood adiposity and metabolic dysfunction-associated steatotic liver disease (MASLD).

• In the univariable analysis, we identified heterogeneous effects of BMI at different life stages on MASLD. However, after accounting for adult obesity in the multivariable analysis, only the protective effect of birth weight remained significant. Notably, this effect attenuated to null after adjusting for all lifestyle factors.

• We also discovered that the impact of birth weight and overall childhood BMI on MASLD is mediated by adult BMI. This suggests that adult body mass index may serve as a critical mediator in the relationship between early life body metrics and MASLD.

Statistical analyses

Overall and local genetic correlation analysis

We quantified the genome-wide average genetic effect shared between exposures and outcomes using LDSC^[32]. This method relies solely on GWAS summary statistics and assesses shared genetic effects between traits, free from bias introduced by environmental confounders. Genetic correlation estimates ( $r_g$) range from −1 to 1, with −1 indicating a perfect negative genetic correlation and 1 indicating a perfect positive genetic correlation. Global genetic correlations were considered statistically significant at a P-value threshold of less than 0.05.

To determine whether the significant overall genetic correlation between childhood BMI and MASLD is influenced by adult BMI, we utilized genomic structural equation modeling (GenomicSEM)^[33] to perform multivariable LDSC. The R package “GenomicSEM” was employed to fit structural equation models using summary statistics from the GWAS data.

Then, we calculated local genetic correlations using SUPERGNOVA^[34] to quantify the genetic similarity of complex traits within specific genomic regions. This method requires only GWAS summary statistics and a reference panel [from the 1000 Genomes Project (1KGP) Phase II, with rare variants (MAF < 5%) filtered out)] as input. The algorithm partitions the genome into 2353 approximately independent regions, each averaging 1.6 centimorgans, using LDetect and with LD estimates based on the 1KGP Phase III data. A significant local signal was identified if the P-value from SUPERGNOVA remained significant after multiple testing corrections (P < 0.05/2353).

Causal relation identification analysis

To investigate the total effect of each childhood obesity-related trait on MASLD, univariable Mendelian randomization (UVMR) was performed to detect putative causal relationships. The inverse-variance weighted (IVW) approach^[35] was conducted as our primary approach. Complementary to IVW, MR-Egger regression^[36], weighted median approach^[37], MR-RAPS^[38], and MR-PRESSO (outlier-corrected)^[39] were performed to examine the robustness and consistency of results under relaxed model assumptions. Additionally, the statistical power of MR was evaluated using an online calculator (https://shiny.cnsgenomics.com/mRnd/)^[40], yielded values of 0.05 across all analyses. To minimize the risk of weak instrument bias, we assessed the strength of the instruments by calculating F-statistics^[41].

Several sensitivity analyses were performed to evaluate the robustness of the findings, including (1) analysis excluding palindromic SNPs with strand ambiguity; (2) analysis excluding pleiotropic SNPs associated with confounding traits based on the GWAS Catalog; and (3) leave-one-out analysis, where each SNP was sequentially removed to detect outliers potentially influencing the MR estimates. We reported the MR analyses following the Strengthening the Reporting of Observational Studies in Epidemiology using MR (STROBE-MR)^[42] (Supplemental Digital Content Table S2, available at: http://links.lww.com/JS9/F891).

To further evaluate whether the causal effects of childhood obesity on MASLD are independent of confounders. We conducted multivariable Mendelian randomization (MVMR)^[43] using adulthood BMI, smoking, drinking, sleep duration, physical activity, education attainment, and household income as confounders. This analysis can only be done on exposures which has some significant SNPs.

In case where specific age groups lacked genome-wide significance SNPs, we applied the latent causal variable model (LCV)^[44], which is a method for inferring genetically causal relationships using GWAS data. To measure causality, we defined the genetic causality proportion ( $\widehat{GCP}$) of age-specific childhood BMI on MASLD. The $\widehat{GCP}$ r ranges from 0 (indicating no genetic causality) to 1 (indicating full genetic causality), considering $\widehat{GCP}0.6$ to be highly effective, high-GCP estimates.

In order to explore potential causal pathways and assess the indirect influence of childhood BMI on MASLD outcomes through risk factors, we performed a mediation analysis using effect estimates from a two-step MR approach. The total effect was determined through a standard univariable MR analysis, serving as the primary MR. To separate the direct and indirect effects, results from the two-step MR were applied, with the Product method used to estimate the beta of the indirect effect and the Delta method employed to calculate the standard error (SE)^[45].

Results

Global and local genetic correlation

We analyzed the genetic overlap between exposure and outcome pairs by examining SNPs across the whole genome. As illustrated in Figure 2, a significant negative genetic correlation was observed between birth weight and MASLD (=−0.23, P-value = 4.4 × 10⁻⁸). Both BMI at age 5 years ( $r_g$ = 0.45, P-value = 0.01) and BMI at age 7 years ( $r_g$ = 0.44, P-value = 0.044) showed a positive genetic correlation with MASLD. Other childhood obesity datasets observed no significance with MASLD.

Figure 2.

Genome-wide genetic correlation between child obesity-related traits and MASLD. r_g refers to the genetic correlation estimates ranging from −1 to 1, with −1 indicating a perfect negative genetic correlation and 1 indicating a perfect positive genetic correlation. CI, confidence interval. P refers to the P-values.

Although our earlier findings suggest that certain childhood obesity traits may either increase or decrease the risk of MASLD, their independent effects remain unclear. To clarify this, we performed a series of GenomicSEM analyses to determine whether the overall genetic correlation between childhood obesity and MASLD will influence by later obesity in adult (Fig. 3). Notably, after adjusting for adult BMI in the GenomicSEM, the effect of BMI at age 5 years on MASLD diminished to non-significance ( $r_g$ = −0.138, P-value = 0.278), as did the effect of BMI at age 7 years ( $r_g$ = −0.212, P-value = 0.145), suggesting the effect of childhood BMI in 5 or 7 years on MASLD is influenced by adult obesity. The inverse relationship between birth weight and MASLD slightly strengthened and remained statistically significant after controlling for adult BMI ( $r_g$ = −0.277, P-value = 1.27 × 10⁻⁸). Besides, the negative effect of the overall child BMI dataset on MASLD became significant after adjusting for adulthood BMI ( $r_g$ = −0.319, P-value = 0.036).

Figure 3.

The result of Genomic structural equation models(GenomicSEM) between child obesity-related traits and MASLD. The figure displays the path diagram (with standard errors in parentheses) for the models adjusting the adult BMI. For example, the model means after adjusting the adult BMI genome-wide genetic correlation between birth weight and MASLD. * P-value <0.05, ** P-value <0.01, *** P-value <0.001. (a) The result of GenomicSEM between birth weight and MASLD after adjusting Adult BMI; (b) the result of GenomicSEM between BMI at 5 years and MASLD after adjusting Adult BMI; (c) the result of GenomicSEM between BMI at 7 years and MASLD after adjusting Adult BMI; (d) the result of GenomicSEM between overall BMI and MASLD after adjusting Adult BMI.

Six genomic regions exhibiting significant local genetic covariance between childhood BMI and MASLD were identified. Three of these regions were associated with birth weight, one with BMI at age 1 year, one with BMI at age 8 years, and one in the overall childhood BMI-MASLD analysis. No regions with a covariance between child BMI at other specific ages and MASLD were identified. All estimates remained significant after multiple testing corrections (P < 0.05/2353). Detailed information for each region is provided in Supplemental Digital Content Figure S1, available at: http://links.lww.com/JS9/F890.

Causal relation identification

The basic characteristics of each GWAS dataset and IVs are shown in Supplemental Digital Content Table S3, available at: http://links.lww.com/JS9/F891 and Supplemental Digital Content Table S4, available at: http://links.lww.com/JS9/F891. All exposures had F-statistics greater than 50, indicating strong instruments.

As shown in Table 1, using UVMR IVW approach, we observed a 40% increased risk of MASLD associated with genetically predicted overall child BMI (OR = 1.40, 95% CI = 1.22–1.62). This effect remained consistent for BMI at age 8 years (OR = 1.34, 95% CI = 1.01–1.78). For birth weight, there was strong evidence of a statistically significant protective effect against MASLD (OR = 0.69, 95% CI = 0.61–0.78). Similar findings were noted for genetically predicted BMI at 8 months, which was associated with a decreased risk of MASLD (OR = 0.87, 95% CI = 0.77–0.98). These estimates remained stable across different methods, including MR-Egger regression, weighted median, MR-RAPS, and MR-PRESSO (See Supplemental Digital Content Table S5, available at: http://links.lww.com/JS9/F891). For certain childhood BMI ages without available IVs, we applied the LCV model using GWAS data to identify genetically causal relationships. Only BMI at age 7 years showed a statistically significant effect on MASLD ( $\widehat{GCP}$ = 0.76, P-value = 3.98 × 10⁻⁵).

Table 1.

Causal association between child obesity-related traits and MASLD

Exposure	Two-sample Mendelian randomizationa					Latent causal variable model (LCV)
	nSNP	F-statistics	OR (95% CI)	Intercept (SE)	P	GCP (SE)	ρg (SE)	PLCV
Birth weight	117	57.02	0.68 (0.58–0.81)	−0.38 (0.09)	7.17 × 10−6***	-	-	-
6 weeks BMI	-	52.25	-	-	-	0.32 (0.15)	−0.18 (0.06)	0.005**
3 months BMI	5	40.57	0.98 (0.88–1.10)	−0.02 (0.06)	0.723	-	-	-
6 months BMI	8	57.18	1.12 (0.93–1.35)	0.11 (0.10)	0.244	-	-	-
8 months BMI	6	45.94	0.87 (0.77–0.98)	−0.14 (0.06)	0.017*	-	-	-
1 year BMI	5	66.55	1.15 (0.92–1.43)	0.14 (0.11)	0.219	-	-	-
1.5 years BMI	4	45.66	0.92 (0.52–1.65)	−0.08 (0.30)	0.786	-	-	-
2 years BMI	-		-	-	-	0.05 (0.40)	0.05 (0.03)	0.12
3 years BMI	-		-	-	-	0.05 (0.40)	−0.01 (0.05)	0.794
5 years BMI	-		-	-	-	0.13 (0.23)	0.42 (0.06)	0.631
7 years BMI	-		-	-	-	0.76 (0.17)	0.44 (0.07)	3.981 × 10−5***
8 years BMI	2	101.21	1.34 (1.01–1.78)	0.29 (0.15)	0.04*	-	-	-
Overall child BMI	19	85.59	1.40 (1.22–1.62)	0.34 (0.07)	2.21 × 10−6***	-	-	-

The statistical power of MR was evaluated using an online calculator (https://shiny.cnsgenomics.com/mRnd/), with all results calculated as 0.05.

- have zero matched SNPs.

SE is standard error.

GCP is the posterior mean estimated genetic causality proportion (range from 0 to 1), with posterior standard error

P_LCV is the P value for the null hypothesis of no partial genetic causality.

ρ_g is the estimated genetic correlation.

^aIVW is displayed in two-sample MR.

^*<0.05, **<0.01, ***<0.001.

Sensitivity analyses excluding pleiotropic SNPs or palindromic SNPs, as well as the leave one-out analysis, showed similar findings, demonstrating the robustness of our results (Supplemental Digital Content Table S5, available at: http://links.lww.com/JS9/F891, Supplemental Digital Content Table S6, available at: http://links.lww.com/JS9/F891).

We conducted a series of MVMR to examine whether the causal effects of birth weight were influenced by childhood and adulthood BMI (Fig. 4, Models 1–3). Notably, the effect of birth weight on MASLD still significant in MVMR after adjusting for overall childhood BMI (OR = 0.62, 95% CI = 0.53–0.75) as well as adjusting for adulthood BMI (OR = 0.61, 95% CI = 0.49–0.74), and both (OR = 0.60 95% CI = 0.49–0.74), suggesting that the effect of birth obesity on MASLD is not influenced by childhood obesity and adulthood obesity. Furthermore, the association persisted after additional adjustments for socioeconomic factors, including educational attainment and household income (Fig. 4, Models 4–6). When considering lifestyle factors such as tobacco use, alcohol consumption, sleep duration, and physical activity individually, the effect of birth weight on MASLD remained significant (Fig. 4, Models 7–10). However, after adjusting for all lifestyle factors collectively, the impact of birth weight on MASLD was attenuated to null (OR = 1.05, 95% CI = 0.98–1.12) (Fig. 4, Model 11).

Figure 4.

Multivariable Mendelian randomization between birth weight and MASLD adjusted for confounders. Models 1– 3 for the MVMR accounting for the overall child BMI and adult BMI, Models 4–6 for the MVMR adjusted for education attainment and household income, Models 7–11 for the MVMR adjusted for lifestyle factors. OR refers to odds ratio. CI, confidence interval. P refers to the P-values derived from IVW intercepts.

Since birth, obesity is likely to persist into childhood and adulthood, and BMI in these later stages may mediate the negative effect on MASLD. We conducted a two-step MR analysis to explore the mediating pathway from birth BMI to MASLD through later-life obesity. In the first step, genetic instruments for birth weight were used to estimate the causal effect of birth weight on potential mediators. We identified causal relationships, with increased birth weight associated with higher childhood BMI (β = 0.34, SE = 0.04, P-value = 4.42 × 10⁻²²) and adulthood BMI (β = 0.11, SE = 0.03, P-value = 0.0002). In the second step, we applied multivariable MR to estimate the direct effect of birth weight on MASLD, accounting for the mediators. Causal evidence was found after adjusting for childhood BMI (β = 0.28, SE = 0.06, P-value = 2.21 × 10⁻⁶) or adult BMI (β = 0.77, SE = 0.05, P-value = 2.90 × 10⁻⁴⁰). We then calculated the indirect effect of birth weight on MASLD via BMI in later life, with the mediation effect of childhood and adult BMI being 0.10 (SE = 0.03, 95% CI = 0.05–0.14) and 0.09 (SE = 0.02, 95% CI = 0.04–0.13), respectively. The total effects are shown in Figure 5a, b. Additionally, we considered overall childhood BMI as the exposure, adult BMI as the mediator, and MASLD as the outcome (Figure 5c). We estimated the indirect effect of overall childhood BMI on MASLD via adult BMI and found a mediation effect of 0.42 (SE = 0.07, 95% CI = 0.28–0.55). The detailed calculation process for the two-step MR is presented in Supplemental Digital Content Table S7, available at: http://links.lww.com/JS9/F891.

Figure 5.

The result of mediation mendelian randomization. *P-value <0.05, **P-value <0.01, ***P-value <0.001, β1 represents the regression coefficients for the association between birth weight and mediators; β2 represents the regression coefficients for the association between mediators and MASLD; in parentheses are the SE (standard error) values; the total effect between birth weight and MASLD (without the adjustment for mediators); and the indirect effect between birth weight and MASLD (taking into account the adjustment for mediators). (a) The potential mediating role of overall child BMI in the association between birth weight and MASLD; (b) The potential mediating role of adult BMI in the association between birth weight and MASLD; (c) The potential mediating role of adult BMI in the association between overall child BMI and MASLD.

Discussion

This study examined age-specific infancy and childhood obesity-related traits in the development of MASLD, utilizing data from the largest GWAS conducted for each trait and applying innovative genetic methods, including genetic correlation and a comprehensive MR framework. We found strong evidence of a significant shared genetic foundation, both globally and locally, highlighting shared biological mechanisms linking early-life adiposity with MASLD. Univariable analysis revealed heterogeneous effects of BMI across different life stages. Then, after adjusting for adult obesity in multivariable analysis, the effect of birth weight on MASLD remained significant. When accounting for some lifestyle factors, the impact of birth weight on MASLD was attenuated to null. Additionally, we performed a mediation analysis to explore potential mediators.

A negative genetic and causal correlation was identified between birth weight/BMI and MASLD, while a positive correlation was observed in later years (BMI at ages 5, 7, and 8 years). This consistent pattern was evident in both genetic and causal associations. Previous studies have shown that children with lower birth weight tend to have a significantly higher percentage of total body fat by ages 7–8 years, compared to those with higher birth weight^[46]. Infants with low birth weight (LBW) are likely to experience catch-up growth later on, and studies have shown that catch-up growth can lead to obesity^[47–49] and even MASLD^[50] in later life. These results are in line with observational studies or systematic reviews^[51–53], as well as previous evidence from MR studies^[13,54], which reported that LBW and high BMI around age 7 years were risk factors for MASLD. These results provide valuable insights for healthcare professionals to implement targeted interventions for different age groups, facilitating the early prevention of MASLD.

Adult obesity is a well-established risk factor for MASLD in observational studies^[55,56], which likely plays a significant role in the association between weight gain during infancy and childhood and MASLD. We found that the genetic correlation between birth weight and MASLD remained significant even after adjusting for adult BMI, while the effect of BMI at ages 5 and 7 years diminished to null. This suggests that late childhood obesity and adult obesity are strongly linked, likely due to environmental factors – such as exposure to high-calorie foods and limited opportunities for physical activity during childhood-that are difficult to change and may persist into adulthood, leading to sustained obesity. In order to understand whether birth weight, childhood obesity and adult obesity constitute a continuous process. we conducted mediation analysis to investigate whether childhood obesity and adult obesity serve as mediating factors in the relationship between LBW and the occurrence of MASLD. Our two-step MR analysis for mediation revealed that adult obesity has a substantial mediating effect, obscuring the impact of birth weight on MASLD. The same result was obtained when childhood obesity was considered as the mediating variable. This will enable us to intervene in childhood and adulthood obesity to block this pathway and prevent MASLD in individuals born with LBW. Furthermore, we found that adult BMI also plays a significant mediating role in the relationship between childhood BMI and MASLD, which could be due to individuals who were obese in childhood maintaining that weight into adulthood, rather than any unique risk posed by childhood adiposity itself. These results emphasize the intricate interaction of obesity-related traits in childhood across various stages of life. By exploring obesity at different points in time, we provide valuable insights into the complex relationship between childhood obesity and MASLD, thereby enhancing our understanding of this multifaceted connection.

In our study, the relationship between birth weight and MASLD attenuated to null after adjusting for lifestyle factors. This aligns with some existing evidence. Emerging evidence highlights the protective effects of healthy lifestyle factors against MASLD. Chinese cohort studies demonstrate that combined healthy behaviors (adequate sleep, regular exercise, alcohol/smoking abstinence, and balanced diet) may reduce MASLD risk by over 90%^[57,58]. Furthermore, longitudinal data indicate that sustained lifestyle adherence significantly decreases mortality among MASLD patients^[59]. Socioeconomic determinants also play crucial roles, as NHANES analyses reveal elevated MASLD risks associated with low household income and educational level^[60]. These sociodemographic factors are additionally linked to fetal growth patterns and childhood obesity development^[61,62]. Notably, our findings reveal persistent associations between birth weight and MASLD risk even after adjusting for sociodemographic confounders. However, several unmeasured factors could influence these relationships. Our analysis did not account for early-life environmental exposures or childhood central obesity measures due to data availability constraints. The developmental origins of health and disease hypothesis states that environmental exposures spanning the perinatal period through birth can impact disease outcomes in later life^[63]. This framework is strongly supported by evidence demonstrating maternal influences on birth weight – particularly through modifiable factors like gestational nutrition and maternal BMI^[64,65]. A review of 72 prospective cohort studies indicated that central fatness was positively and significantly associated with a higher all-cause mortality risk, with general and central obesity^[66]. Furthermore, emerging evidence suggests central obesity independently correlates with arterial stiffness, whereas the impact of general obesity appears contingent on metabolic health status^[67]. Although these findings underscore the importance of body fat distribution in metabolic risk, detailed GWAS for childhood central obesity (like waist-hip ratio) are currently unavailable. These gaps highlight the need for future studies integrating life-course exposure assessments with detailed adiposity phenotyping.

Our findings have implications for clinicians to understand the intervention. First, clinicians should recognize that mitigating MASLD risk requires lifelong weight management strategies rather than isolated interventions, with particular attention to individuals with LBW who necessitate sustained promotion of healthy weight trajectories from childhood through adulthood. This includes establishing early nutritional guidance for LBW children through regular dietary assessments and personalized meal plans that restrict added sugars and trans fats throughout life^[68] and integrating structured physical activity interventions^[69], such as daily moderate-to-vigorous exercise programs should be implemented starting in school-aged children to counteract sedentary behaviors across the lifespan. Furthermore, enhanced metabolic monitoring is warranted in LBW adolescents, including biennial liver ultrasonography and routine biochemical profiling (e.g., ALT, AST, and GGT) to detect subclinical hepatic dysfunction. These targeted strategies not only address the lifestyle-mediated pathways linking LBW to MASLD but also align with preventive frameworks for mitigating long-term metabolic risks^[70]. By integrating nutritional, behavioral, and surveillance components into standardized clinical protocols, healthcare providers can empower LBW individuals to attenuate their inherent predisposition to MASLD and related cardiometabolic complications.

Our study has several notable strengths. To our knowledge, this is the first study to leverage summary genetic associations to explore the causal links between age-specific adiposity during infancy and childhood and MASLD. Additionally, we utilized the largest available GWAS for MASLD to ensure sufficient statistical power and further examined multiple subtypes of childhood obesity, enhancing the age relevance of this investigation. Furthermore, our comprehensive analytical approach was grounded in the core assumptions of the MR model. The F-statistics of the instruments ranged from 26.75 to 358.50, confirming the use of strong instruments and fulfilling the “relevance” assumption. Although the “exchangeability” assumption in standard MR may be violated due to pleiotropic effects of genetic variants, we applied multivariable MR to minimize confounding from adult obesity and mediation analysis based on the results of two-step MR. Finally, we employed various methods to derive robust causal estimates, including sensitivity analyses such as MR-Egger, weighted median, MR-PRESSO, MR-RAPS, excluding pleiotropic or palindromic SNPs, and leave-one-out analysis. However, it is important to recognize several limitations. First, our analysis was restricted to European ancestry, limiting the generalizability of our findings to other ancestry groups. Future studies should investigate these associations in diverse non-European cohorts. Secondly, the linear assumption of the two-sample MR approach precluded exploration of potential non-linear relationship, such as the U-shaped association observed in traditional epidemiological studies^[71,71] Thirdly, previous studies have demonstrated that discrepancies between observed BMI and genetically predicted BMI are associated with metabolic disease risk^[71]. However, our analyses utilized publicly available GWAS summary statistics, which lack the individual-level phenotypic data required to calculate observed BMI. Future one-sample MR studies using an individual-level dataset could employ semiparametric methods to address this limitation. Besides, future studies should prioritize integration of multi-omics data (transcriptomic, proteomic, and metabolomic) to systematically map mediating pathways. This foundational work should be complemented by mechanistic investigations using experimental models to resolve the molecular cascades linking early growth patterns to hepatic lipid metabolism.

Conclusions

In summary, this study deepens our understanding of the previously reported associations between age-specific adiposity during infancy and childhood and MASLD, providing evidence of genetic correlations and potential causal effects. Our findings also suggest that obesity in later life may serve as a critical link between early adiposity and the development of MASLD. Moreover, by improving modifiable lifestyle factors such as smoking, drinking, physical activity, and sleep behaviors, it is possible to reduce the risk of MASLD in later years. This underscores the importance of preventive measures and early interventions during childhood to alter the course of the disease.

Supplementary Material

js9-112-1425-001.xlsx ^{(91.9KB, xlsx)}

js9-112-1425-002.docx ^{(403.8KB, docx)}

Ethical approval

All data used in this study were de-identified publicly available data; therefore, no ethical approval was required for this study.

Consent

Not applicable.

Sources of funding

This work was supported in part by the National Natural Science Foundation Regional Innovation and Development Joint Fund (U23A202600) and the Sichuan Science and Technology Program (2023NSFSC0652).

Author contributions

Concept and design: Q.G., C.X., and Z.L. Acquisition, analysis, or interpretation of data: Q.G., L.Z., J.C., Z.S., and C.Y. Drafting of the manuscript: Q.G., C.X., and Z.L. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: Q.G. and C.X. Obtained funding: Z.L. Administrative, technical, or material support: C.Y. and Z.L. Supervision: C.X. and Z.L.

Conflicts of interest disclosure

The authors report no potential conflicts of interest.

Research registration unique identifying number (UIN)

Not applicable. Our study does not involve direct human participant research but rather utilizes reproduced materials from other sources. We thank the participants and investigators involved in the GWASs used in this study.

Guarantor

Zhenmi Liu and Chenghan Xiao.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

The data generated in this study and a data dictionary (Supplemental Digital Content Table S1, available at: http://links.lww.com/JS9/F891) are provided in the Supplementary Information. GWAS summary statistics for overall childhood BMI are available here (GWAS Catalog: https://ftp.ebi.ac.uk/pub/databases/gwas/summary_statistics/GCST90002001-GCST90003000/GCST90002409/ from the article with PubMed ID of 33045005). The adult BMI (https://portals.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_files), the GWAS data of birth weight (https://egg-consortium.org/birth-weight-2019.html), the age-specific childhood BMI dataset (PubMed ID: 35315439), and MASLD traits (PubMed: 36280732) are described in the relevant articles.