Correlational Associations Between Ambient and Household Particulate Matter Exposure and Body Mass Index Across Childhood and Adolescence in Thailand

Abstract

While previous studies have established inverse relationships between (particulate matter) PM exposure and the body mass index (BMI), this study is the first to demonstrate that the strength of this relationship varies significantly according to the PM source type and age group, particularly across developmental stages. Through a comprehensive 31-year analysis in Thailand that uniquely captured the transition from traditional to modern energy sources, this research investigated the relationship between PM exposure and a high BMI among children and adolescents across different demographic groups, using a correlational analysis of time point data from the Global Burden Disease (GBD) study (1990–2021). The analysis examined the association between a high BMI and two categories of PM exposure—ambient (outdoor) and household (indoor)—through cross-correlation, Spearman correlation, and mixed-effects models. The results reveal a significant inverse relationship between household PM exposure and a high BMI, particularly pronounced in younger age groups (2–9 years), with household PM showing consistently stronger associations compared to ambient PM. Among children aged 2–4 years, household PM exposure exhibited a strong negative correlation with a high BMI without a time lag, suggesting persistent effects of the household air quality on physical development. Conversely, ambient PM exposure showed relatively weaker associations, with only slight positive correlations observed in certain subgroups. Further analysis indicated that decreases in household PM exposure correlated with longitudinal increases in a high BMI, with this relationship diminishing during adolescence. These findings provide crucial insights for targeted public health interventions and offer a model for understanding PM-BMI relationships in other developing nations experiencing similar transitions.

1. Introduction

Childhood and adolescent obesity represents one of the most pressing public health challenges globally, with its increasing prevalence linked to severe long-term health consequences, including cardiovascular disease, diabetes, and a reduced life expectancy [1]. Although several studies have extensively investigated traditional risk factors for obesity, emerging evidence has brought environmental factors, particularly air pollution exposure, to the forefront as significant contributors to obesity and a high BMI among young populations [2]. PM, a primary component of air pollution, originates from diverse sources: ambient sources encompass traffic emissions, industrial activities, and urban development, while household sources include indoor cooking practices and heating systems. While the impacts of PM exposure on respiratory and cardiovascular health are well established [3,4], recent studies have revealed its role in influencing metabolic outcomes, including obesity. This relationship holds particular significance for children, whose developing physiological systems and proportionally higher exposure rates make them especially vulnerable to environmental pollutants [5,6].

A study has demonstrated that ambient PM exposure, primarily originating from urban sources, induces systemic inflammation that may disrupt metabolic regulation and promote weight gain in children [7,8]. The impact of household PM exposure, which stems from indoor activities such as cooking and heating, presents distinct challenges due to the sustained and intimate nature of domestic exposure patterns [9,10]. This concern becomes particularly acute in regions where biomass fuels serve as the primary household energy sources, subjecting children to prolonged periods of elevated exposure and heightening their vulnerability to associated health risks [9,10,11]. While studies have thoroughly examined the effects of PM on respiratory health, the understanding of its influence on the BMI and obesity-related outcomes remains limited, especially regarding cumulative and long-term exposure patterns [12].

The respiratory and cardiovascular consequences of PM exposure in children have been extensively reported through rigorous studies. Previous studies have revealed that prolonged exposure to ambient PM_2.5 significantly impairs lung function development and increases the incidence of respiratory conditions, including asthma, bronchitis, and respiratory infections, particularly in regions experiencing poor air quality [13,14]. From a mechanistic perspective, fine particles penetrate the respiratory system, triggering oxidative stress and inflammatory responses that can culminate in chronic respiratory conditions such as asthma [15,16]. Children’s heightened susceptibility to these effects stems largely from their developing immune systems [15,16]. The cardiovascular implications of PM exposure extend beyond respiratory concerns, with mounting evidence suggesting significant impacts on cardiovascular health during early development [15,16]. A study has established correlations between early-life air pollution exposure and various cardiovascular indicators, including increased arterial stiffness, elevated blood pressure, and early markers of atherosclerosis, in pediatric populations [17,18]. These cardiovascular effects likely arise from systemic inflammation and oxidative stress triggered by PM’s infiltration into the circulatory system [17,18]. Indoor air pollution where biomass fuels are used for cooking present particularly concerning scenarios, as the resulting high concentrations of household PM significantly amplify these health risks [19,20].

Thailand provides an exceptional context for investigating the effects of PM exposure on the adolescent BMI due to its distinctive environmental and demographic characteristics. Recent demographic studies have revealed that approximately 20 million Thai citizens, representing 29% of the total population, are under 19 years of age, making them particularly susceptible to environmental exposure [8,21]. This substantial youth demographic experiences diverse PM exposure patterns influenced by Thailand’s tropical climate, seasonal agricultural activities, and accelerating urbanization [8]. The country’s varied geographical regions demonstrate markedly different PM exposure profiles. The northern regions experience peak PM_2.5 concentrations during February–April, primarily attributed to seasonal agricultural burning practices. In contrast, the central plains, encompassing the Bangkok Metropolitan Region, face persistent elevated PM levels throughout the year due to sustained industrial activities and vehicular emissions [8]. Urban areas consistently exceed the World Health Organization (WHO) guidelines for PM_2.5, with the measured concentrations fluctuating between 25 and 45 μg/m³, with variations that depend on the seasonal conditions and specific locations [22,23]. These distinct regional exposure patterns may serve as valuable models for understanding similar phenomena in other southeast Asian nations undergoing comparable developmental transitions. Indoor exposure patterns in Thailand reflect the country’s current stage of economic development. Approximately 15% of households continue to rely on traditional cooking methods [20], resulting in indoor PM_2.5 concentrations that can reach alarming levels of 150–300 μg/m³ during cooking periods. This exposure particularly affects children, who typically spend 14–16 h daily in indoor environments [24]. The observed exposure profile characterizes the situation in many middle-income countries transitioning from traditional to modern energy sources, suggesting that our findings may have broader applicability to similar developing nations.

The study by Johnson et al. [14] demonstrates that the relationship between PM exposure and the BMI is significantly influenced by both age and gender, with the developmental stage playing a crucial role in the vulnerability. Younger children exhibit heightened susceptibility to PM exposure due to their elevated breathing rates relative to their body size and their still-developing immune and metabolic systems [14]. Gender-specific responses to air pollution have also emerged as an important consideration in understanding the effect of PM on the BMI. These differences appear to stem from a combination of physiological and behavioral factors, resulting in distinct patterns of BMI changes between boys and girls even when exposed to similar PM levels [8,25]. The biological pathways include the activation of systemic inflammatory responses, the generation of oxidative stress throughout the body, and significant alterations in the gut microbiota composition [15,16,17,18,26,27].

While the impacts of PM on respiratory and cardiovascular health are well documented, the evidence linking PM exposure to metabolic outcomes such as a high BMI and obesity remains less robust. This gap is particularly pronounced in the context of cumulative and long-term exposure to both ambient and household PM during critical developmental stages. Furthermore, existing studies have predominantly focused on high-income countries, leaving middle-income nations like Thailand underrepresented despite their unique environmental and demographic characteristics. Although the mechanisms linking PM exposure to the BMI are well reported, there remains a significant gap in our understanding of how both ambient and household PM exposure influences BMI trajectories over time, particularly across different demographic groups [12]. This study addressed these knowledge gaps by examining the correlational relationship between PM exposure and the BMI throughout crucial developmental periods, from early childhood through adolescence. A distinguishing feature of this study lies in its examination of both ambient and household PM sources. This dual consideration allowed us to develop a more nuanced understanding of how different types of PM exposure collectively influence metabolic health across diverse demographic groups. By analyzing these combined effects, we can better understand the total environmental burden on children’s physical development and identify potential critical periods when interventions might be most effective.

2. Materials and Methods

This correlational study examined the relationship between PM exposure (both ambient and household) and a high BMI in Thailand using data from the GBD study from 1990 to 2021 [28]. The analytical framework encompassed three complementary statistical methods, each designed to capture different aspects of the relationship between PM exposure and a high BMI. The first analytical approach utilized cross-correlation analysis to investigate time-lagged relationships between PM exposure and a high BMI. By examining lags ranging from 0 to 3 years, this method enabled us to detect potential delayed effects of PM exposure on high BMI development. This temporal analysis proved particularly valuable given that environmental exposure may take considerable time to manifest in physiological changes, especially during developmental periods. The second component of the analysis employed Spearman’s rank correlation coefficient to evaluate the strength and directionality of monotonic relationships between PM exposure and a high BMI across different age and gender groups, incorporating various time lags to explore possible delayed associations. The third component of the analytical framework involved the implementation of a mixed-effects model. This statistical approach allowed us to examine how PM exposure influences high BMI trajectories over time while accounting for both population-level trends and individual variations. The model incorporated fixed effects to capture the overall impact of PM exposure across the entire study population, while simultaneously including random effects to account for individual variations in baseline high BMI values. This dual consideration of fixed and random effects provided a more nuanced and realistic understanding of the complex relationship between PM exposure and BMI development. All statistical analyses and data processing were performed using Python version 3.11.8.

2.1. Data Used

The data were obtained from the recent GBD study, which provides a comprehensive analysis of 88 risk factors, including PM exposure and a high BMI, across 204 countries and 811 subnational locations during 1990–2021 [28]. The study encompassed all 77 provinces of Thailand, including both urban and rural areas, with data collected annually through the national health surveillance system. The analysis examined these associations across age groups (2–19 years) and biological sex categories (male, female, and combined). The terms “male” and “female” are used to denote the biological sex as documented in the dataset responses to PM exposure. While acknowledging that gender encompasses broader aspects of identity and social constructs, this analysis relied solely on the biological sex classifications available in the data. The study analyzed data from Thailand’s total population, which grew from 56,764,608 in the year 1990 to 66,683,246 in the year 2021. This investigation adhered to rigorous ethical and methodological standards in accordance with international research guidelines. The methodology strictly complied with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) statement, ensuring transparency and reproducibility [28]. The research followed the protocol for the GBD, Injuries, and Risk Factors Study, which establishes standardized procedures for data collection and analysis [28]. Institutional support and oversight were provided by leading organizations, including the Bill & Melinda Gates Foundation, the UK Department of Health and Social Care, and the Norwegian Institute of Public Health, maintaining high ethical standards throughout the study process [28].

The GBD database represents the most comprehensive and current population-level health data available for Thailand, providing standardized metrics for both exposure and health outcomes. The database includes age-standardized high BMI values and PM exposure data, categorized into ambient (outdoor) and household (indoor) sources, with consistent measurement protocols maintained throughout the study period [28]. Data on high BMI values and PM exposure (categorized into ambient and household PM) were obtained.

A high BMI was defined using specific thresholds and standards to ensure comparability across age groups. For individuals under 20 years, a high BMI was determined using the age- and sex-specific cutoff points established by the International Obesity Task Force (IOTF) [29]. These cutoffs correspond to adult BMI values of 25 kg/m² for overweight and 30 kg/m² for obesity at age 18, ensuring alignment with international standards [29]. For adults aged 20 years and older, a high BMI was defined as overweight (25.0–29.9 kg/m²) and obesity (BMI ≥ 30.0 kg/m²), following the WHO criteria [30]. By employing these standardized thresholds, the study ensured consistency in assessing and high BMI prevalence across ages, sexes, and time. The BMI is calculated as the weight in kilograms divided by the square of the height in meters (kg/m²) [30]. The GBD dataset provides age-standardized high BMI values, accounting for growth patterns and demographic differences [28]. This standardization enables robust comparisons across population groups and time periods. High BMI data were derived from over 2288 high-quality sources, including national health surveys, cohort studies, and administrative health records [28]. The data underwent rigorous validation and quality control processes to ensure accuracy [28]. For missing or incomplete data, the GBD framework employed statistical imputation techniques based on the comparative risk assessment modeling approach, further enhancing the data completeness [28]. Uncertainty intervals (UIs) were calculated using 1000 draw-level estimates, providing 95% UIs for each parameter to account for variability in the data and modeling process [28]. The particulate matter exposure assessment focused on PM_2.5, measured as the annual average concentrations in micrograms per cubic meter (µg/m³) [28]. These measurements integrated satellite data with ground-level air quality monitoring station readings to ensure comprehensive regional coverage, following GBD protocols [28].

The GBD dataset characterizes PM exposure through two distinct categories: ambient (outdoor) and household (indoor) PM, with measurements primarily focusing on the PM_2.5 concentration. This dual categorization allows for an understanding of the total particulate matter exposure across different environmental contexts. Ambient PM exposure data represent particulate matter in outdoor environments, primarily originating from sources such as vehicular traffic, urban activities, and industrial emissions [28]. The measurement of ambient PM involved a combination of ground-based monitoring stations and satellite-derived estimates. These measurements were then transformed into population-weighted levels, providing a more accurate representation of human exposure across different regions [4,31,32]. Household PM exposure presents a distinct challenge in measurement and estimation, particularly in low- and middle-income countries where solid fuels remain common for cooking and heating purposes. These indoor environments often experience significantly higher PM concentrations compared to outdoor environments due to the confined nature of living spaces and limited ventilation [11,33]. To capture this indoor exposure accurately, the GBD dataset employs an estimation approach that combines information about household energy sources with data from indoor air quality surveys [9,10].

The age categories were defined as 2–4 years, 5–9 years, 10–14 years, and 15–19 years to assess PM exposure effects across key developmental stages, with participants representing both urban and rural regions of Thailand. As described in [28], this geographic diversity captures varying exposure patterns and sources across different environmental contexts. In urban areas, children encounter ambient PM predominantly from traffic emissions, industrial activities, and construction dust, while their rural counterparts experience significant household PM exposure from the biomass combustion used in cooking and heating. The study design acknowledged that each age category represents a critical growth period with distinct physiological characteristics and exposure patterns. Early childhood (2–4 years) is characterized by rapid growth and metabolic changes, making children particularly vulnerable to environmental exposure [34]. During this period, exposure patterns are heavily influenced by the time spent in home environments and early childhood centers, where the indoor air quality plays a crucial role. Children aged 5–9 years experience steady growth and establish dietary habits that influence future BMI trajectories [35]. Their exposure profile expands to include school environments, where urban students may face elevated PM_2.5 and NO₂ levels from nearby traffic, while rural students might encounter indoor air pollutants from biomass-based cooking facilities. Early adolescence (10–14 years) coincides with puberty and significant hormonal changes, potentially modifying how PM exposure affects weight regulation [36]. During this stage, exposure sources diversify further as adolescents spend more time in various environments, including schools, recreational areas, and local markets. Late adolescence (15–19 years) represents the final maturation stage, where health behaviors solidify and the BMI approaches adult patterns [37]. This age group’s exposure patterns often reflect increasing mobility and varied activity patterns across urban and rural settings. This comprehensive age stratification, combined with the cohort’s geographical diversity, enabled a detailed examination of how PM exposure impacts the BMI across developmental stages and environmental contexts.

2.2. Description of Cross-Correlation

The investigation of time-lagged effects between PM exposure and the development of a high BMI required an analytical approach capable of capturing delayed associations. This study employed cross-correlation analysis [38,39], a useful statistical method specifically designed to examine the relationship between two time series as they shift relative to each other temporally. This analytical technique proves particularly valuable in environmental health studies, as exposure effects may manifest over extended periods rather than immediately. This study implemented time lags ranging from 0 to 3 years, allowing us to capture both the immediate and delayed effects of PM exposure on high BMI development. The positive time lags represent the influence of past PM exposure on future high BMI values, while negative time lags indicate reverse causation, which was not expected in this analysis. The mathematical form of a cross-correlation analysis is expressed by the cross-correlation coefficient ${\rho }_k$ at each lag, k, given by Equation (1):

$${\rho }_k={\displaystyle \frac{\sum _{t=1}^T(P{M}_{t-k}-\overline{PM})(BM{I}_t-\overline{BMI})}{\sqrt{\sum _{t=1}^T{(P{M}_{t-k}-\overline{PM})}^2\sum _{t=1}^T({BM{I}_t-\overline{BMI})}^2}}}$$

(1)

where $P{M}_{t-k}$ represents the PM exposure at time t lagged by k years, $BM{I}_t$ represents the BMI at time t, and $\overline{PM}$ and $\overline{BMI}$ are the mean values of PM and the BMI, respectively, over the observed time period.

2.3. Description of Spearman Correlation

To examine the relationship between PM exposure and a high BMI across different demographic groups, we employed Spearman’s rank correlation coefficient, a statistical method particularly well suited for environmental health studies. This analytical approach offers distinct advantages over traditional parametric methods, especially when dealing with complex environmental data. Spearman’s rank correlation coefficient was chosen for its ability to detect monotonic relationships between variables without requiring assumptions about the precise mathematical form of their relationship. A monotonic relationship exists when variables consistently increase or decrease together, even if that relationship is not strictly linear. This characteristic makes the method especially valuable in environmental health studies, where the relationships between exposure and health outcomes often follow complex, non-linear patterns [40]. The mathematical form of the Spearman correlation coefficient ρ is given by Equation (2):

$$\rho =1-{\displaystyle \frac{6\sum d_i^2}{n(n^2-1)}}$$

(2)

where d_i is the difference between the ranks of each pair of observations (BMI and PM exposure) and n is the number of paired observations.

The Spearman correlation was calculated between high BMI values and PM exposure (both ambient and household) across different age and gender groups, with time lags of 0 to 3 years to capture potential delayed effects of PM exposure on BMI outcomes. A positive Spearman coefficient indicates that higher PM exposure is associated with a higher BMI, while a negative coefficient suggests an inverse relationship.

2.4. Mixed-Effects Model

Mixed-effects models were employed to evaluate the association between PM exposure and a high BMI over time. This method was chosen due to its ability to account for repeated measures within individuals and to capture both population-level fixed effects (e.g., overall trends in PM exposure and the BMI) and individual-level random effects (e.g., variations in the baseline BMI across participants). Mixed-effects models were particularly well suited for this study’s longitudinal dataset, which contained repeated cross-sectional measurements from different individuals at each time point, allowing for an analysis that incorporated temporal changes while controlling for within-group variability [41,42]. Time-lagged effects were specifically incorporated to assess delayed associations between PM exposure and the BMI, as environmental exposure often exerts cumulative impacts on health outcomes. Time lags ranging from 0 to 3 years were included based on the prior literature, which indicates that exposure to air pollution can have both immediate and delayed physiological effects, particularly during critical developmental stages. Distributed lag models (DLMs) were considered but deemed less appropriate for this study due to the long time span of the dataset (1990–2021) and the aggregated nature of the PM exposure and BMI data across age groups. Mixed-effects models, by contrast, offer greater flexibility in handling longitudinal data and allow for robust adjustments of time-varying and individual-level effects. The model for this study was structured as follows (3):

$${BMI}_{it}={\beta }_0+{\beta }_1\times {PM}_{it}+{\beta }_2\times {year}_{it}+u_i+{\in }_{it}$$

(3)

where BMI_it represents the BMI of individual i at time ttt, PM_it is the PM exposure level for individual i at time t (categorized as ambient or household), ${year}_{it}$ is a centered variable to account for time effects, β₀ is the fixed intercept, β₁ and β₂ are fixed-effect coefficients representing the impacts of PM exposure and time, respectively, u_i represents the random intercept, capturing individual-level variation in the baseline BMI values, and ϵ_it is the residual error term, assumed to be normally distributed.

The model included both fixed and random effects to control for individual differences in the baseline BMI and account for unobserved heterogeneity in the PM sensitivity across the population. Time effects (modeled through the centered variable Yearit) were included to adjust for temporal changes that may have impacted the BMI independently of PM exposure, such as improvements in nutrition or healthcare access over the study period. The time lags (0 to 3 years) were chosen based on the prior literature suggesting delayed effects of environmental exposure on metabolic outcomes in children, as PM exposure may exert cumulative impacts that manifest over several years [17]. Including these lags allowed the model to detect both immediate and sustained associations between PM and a high BMI, helping to capture a fuller range of exposure effects over time. To be explicit about our study design, the GBD database provides repeated cross-sectional measurements from different individuals at each time point, rather than following the same individuals. While the random intercepts, $u_i$, in our model account for population-level variations in the baseline high BMI, these measurements represent different individuals sampled at each time point within our defined demographic groups. This specification assumes that individual differences in the baseline high BMI are randomly distributed and uncorrelated with PM exposure, allowing the model to isolate the population-level effect of PM on a high BMI while accounting for individual variability. Additionally, the residuals, ϵ_it, were assumed to follow a normal distribution, consistent with standard practice in mixed-effects modeling, to enable the reliable estimation of fixed effects and the error variance.

2.5. Ethical Considerations

This study utilized publicly available data from the GBD database, which are de-identified and aggregated at the population level. As the study did not involve human participants directly or require access to individual-level data, formal ethical approval from an institutional review board (IRB) or ethics committee was not applicable. The research adhered to the ethical standards outlined in the GATHER and respected the principles of data protection and the responsible use of secondary data. Additionally, all analyses were conducted in compliance with the terms and conditions of the GBD database, ensuring the ethical handling and reporting of the findings.

3. Results

3.1. Time Series Analysis

The time series analysis (Figure 1) illustrated the relationship between a high BMI and ambient PM exposure for different age and gender groups, covering the ages 2–19 from 1990 to 2021. Each sub-figure displays the high BMI trends alongside the ambient PM levels. For children aged 2–4, both males (Figure 1a) and females (Figure 1b) showed an increase in high BMI values as the ambient PM levels decreased. However, this inverse relationship was not consistent across the entire time series. Certain time intervals deviated from this pattern, suggesting that additional factors may have contributed to variations in the high BMI trends beyond PM exposure alone. Specifically, the PM levels declined from approximately 30 µg/m³ in the early 1990s to below 10 µg/m³ by 2020, while the high BMI values rose from around 125 to 170 kg/m² for males (Figure 1a) and from 135 to 145 kg/m² for females (Figure 1b). When analyzing both sexes combined (Figure 1c), the high BMI trends aligned with the observed decrease in PM, further suggesting an inverse relationship between PM exposure and a high BMI in this age group. In the 5–9 years age group, a similar temporal inverse relationship between PM exposure and a high BMI was observed, though it appeared more pronounced in females (Figure 1e) than in males (Figure 1d). Female high BMI values increased from around 200 to 230, while male high BMI values increased from approximately 180 to 260 over the study period. For both sexes combined (Figure 1f), the trend mirrored these observations, indicating a potentially stronger influence of ambient PM on a high BMI in females. Among children aged 10–14 years, the relationship between a high BMI and ambient PM exposure became less distinct and showed localized fluctuations. In males (Figure 1g), high BMI values fluctuated between 250 and 290 kg/m² as the PM levels decreased from approximately 20 µg/m³ to 5 µg/m³. A similar pattern was seen in females (Figure 1h), with high BMI values ranging from 230 to 260 kg/m². For both sexes combined (Figure 1i), the high BMI trends appeared inconsistent with the PM changes, suggesting additional factors influencing a high BMI in this age group and weakening the inverse correlation between PM exposure and a high BMI. In adolescents aged 15–19 years, a notable decline in high BMI values was observed during the years 2005–2010, followed by a subsequent increase. Male high BMI values (Figure 1j) remained relatively stable between 115 and 145 kg/m², while female high BMI values (Figure 1k) fluctuated between 145 and 155 kg/m². For both sexes combined (Figure 1l), the high BMI trends did not clearly align with the changes in PM exposure. While the PM levels continued to decline across the study period, the variations in high BMI values suggest localized and age-specific factors, leading to a weaker or negligible relationship in this age group.

Figure 2 shows the time series trends for high BMI values and household PM exposure across age and gender groups, covering the ages of 2–19 years from 1990 to 2021. Household PM exposure declined significantly across all groups, likely due to improvements in the indoor air quality from cleaner cooking technologies and the reduced use of solid fuel. The PM levels decreased from approximately 80 µg/m³ in the early 1990s to below 10 µg/m³ by 2020. For children aged 2–4 years, both males (Figure 2a) and females (Figure 2b) showed high BMI increases as the household PM levels decreased. Male high BMI values rose from approximately 120 to 175 kg/m² (Figure 2a), while female high BMI values increased from around 115 to 155 kg/m² (Figure 2b). For both sexes combined (Figure 2c), the high BMI trends followed a similar pattern, suggesting a possible positive association between reduced household PM exposure and a higher high BMI in young children. In the 5–9 years age group, an inverse relationship between household PM and a high BMI was again observed, with high BMI values increasing as PM exposure declined. This trend was more pronounced in females (Figure 2e), where high BMI values rose from approximately 170 to 240 kg/m², compared to males (Figure 2d), who exhibited a smaller increase from 180 to 260 kg/m². For both sexes combined (Figure 2f), the high BMI trends mirrored these observations, further supporting the association between lower household PM and a higher high BMI during middle childhood. Among children aged 10–14 years, the BMI trends became more variable despite the continued decline in the household PM levels from approximately 60 µg/m³ to below 10 µg/m³. In males (Figure 2g), high BMI values fluctuated between 240 and 320, while in females (Figure 2h), high BMI values ranged from around 220 to 270 kg/m². For both sexes combined (Figure 2i), the high BMI trends appeared inconsistent with the observed decline in household PM, indicating that other factors, such as dietary habits or pubertal changes, may play a stronger role in influencing a high BMI during this developmental stage. For adolescents aged 15–19 years, the relationship between household PM and the BMI was the least apparent. The male BMI (Figure 2j) remained relatively stable between 115 and 140 kg/m², while female high BMI values (Figure 2k) fluctuated between 150 and 160 kg/m². In both sexes combined (Figure 2l), the trends in high BMI values did not directly correspond to the declining PM levels, suggesting that lifestyle changes, increased physical activity, and hormonal fluctuations during late adolescence may overshadow the effects of household PM exposure on a high BMI. Notably, a transient decline in high BMI values during 2005–2010 was observed in some subgroups, particularly adolescents aged 15–19 years (Figure 2j–l), before resuming an upward trajectory. This may reflect temporary changes in dietary patterns, increased physical activity, or regional health interventions during this period, although further investigation is warranted to identify the contributing factors.

3.2. Cross-Correlation

In this section, we describe how the cross-correlation analysis explored the relationship between a high BMI and PM exposure (ambient and household) over time, accounting for different lag periods. This analysis was critical for understanding how past PM exposure may influence high BMI changes at various time points, offering insights into delayed or persistent effects. The time-lagged cross-correlation analysis (

Table 1. Cross-correlation coefficients with 95% confidence intervals and p-values for time-lagged associations between a high BMI and PM exposure (ambient and household) across age groups and genders. Single asterisk (*): indicates that the correlation is statistically significant at the 0.05 level (p < 0.05); double asterisk (**): indicates that the correlation is statistically significant at the 0.01 level (p < 0.01).

Category		Pollution Type	No Lag	p-Values	Lag 1	p-Values	Lag 2	p-Values	Lag 3	p-Values
Aged 2–4 years	Male	Ambient	−0.27 [−0.57, 0.09]	0.1335	−0.19 [−0.51, 0.18]	0.3153	−0.09 [−0.44, 0.28]	0.6199	0.01 [−0.36, 0.38]	0.9512
		Household	−0.83 ** [−0.91, −0.67]	0.0000	−0.81 ** [−0.90, −0.64]	0.0000	−0.79 ** [−0.90, −0.60]	0.0000	−0.76 ** [−0.88, −0.55]	0.0000
	Female	Ambient	0.00 [−0.35, 0.35]	0.9861	0.10 [−0.26, 0.44]	0.5840	0.22 [−0.15, 0.54]	0.2419	0.36 [−0.01, 0.64]	0.0540
		Household	−0.69 ** [−0.84, −0.45]	0.0000	−0.71 ** [−0.85, −0.47]	0.0000	−0.71 ** [−0.85, −0.48]	0.0000	−0.71 ** [−0.86, −0.47]	0.0000
	Both	Ambient	−0.19 [−0.50, 0.17]	0.3086	−0.10 [−0.44, 0.26]	0.5922	−0.00 [−0.36, 0.36]	0.9822	0.11 [−0.27, 0.46]	0.5784
		Household	−0.79 ** [−0.89, −0.60]	0.0000	−0.78 ** [−0.89, −0.59]	0.0000	−0.77 ** [−0.88, −0.57]	0.0000	−0.75 ** [−0.88, −0.53]	0.0000
Aged 5–9 years	Male	Ambient	−0.24 [−0.54, 0.12]	0.1834	−0.17 [−0.49, 0.20]	0.3703	−0.10 [−0.44, 0.27]	0.6005	−0.05 [−0.41, 0.32]	0.7777
		Household	−0.69 ** [−0.84, −0.45]	0.0000	−0.69 ** [−0.84, −0.45]	0.0000	−0.69 ** [−0.84, −0.44]	0.0000	−0.68 ** [−0.84, −0.42]	0.0000
	Female	Ambient	−0.30 [−0.59, 0.05]	0.0920	−0.26 [−0.56, 0.11]	0.1611	−0.22 [−0.54, 0.15]	0.2325	−0.20 [−0.53, 0.18]	0.2886
		Household	−0.46 ** [−0.70, −0.14]	0.0077	−0.52 ** [−0.74, −0.20]	0.0027	−0.56 ** [−0.76, −0.25]	0.0014	−0.58 ** [−0.78, −0.27]	0.0011
	Both	Ambient	−0.29 [−0.58, 0.06]	0.1058	−0.23 [−0.54, 0.14]	0.2204	−0.17 [−0.50, 0.20]	0.3575	−0.14 [−0.48, 0.24]	0.4711
		Household	−0.60 ** [−0.79, −0.32]	0.0003	−0.62 ** [−0.80, −0.35]	0.0002	−0.64 ** [−0.81, −0.36]	0.0002	−0.64 ** [−0.81, −0.36]	0.0002
Aged 10–14 years	Male	Ambient	0.16 [−0.20, 0.48]	0.3954	0.14 [−0.22, 0.47]	0.4402	0.10 [−0.27, 0.44]	0.6083	0.01 [−0.36, 0.38]	0.9545
		Household	−0.39 * [−0.65, −0.04]	0.0293	−0.38 * [−0.65, −0.03]	0.0335	−0.38 * [−0.65, −0.02]	0.0396	−0.36 [−0.64, 0.00]	0.0526
	Female	Ambient	−0.01 [−0.36, 0.34]	0.9537	−0.04 [−0.38, 0.32]	0.8514	−0.11 [−0.45, 0.26]	0.5558	−0.23 [−0.55, 0.14]	0.2207
		Household	−0.35 * [−0.62, −0.00]	0.0492	−0.39 * [−0.66, −0.04]	0.0287	−0.42 * [−0.68, −0.07]	0.0202	−0.43 * [−0.69, −0.08]	0.0187
	Both	Ambient	0.08 [−0.28, 0.41]	0.6756	0.06 [−0.30, 0.41]	0.7473	−0.00 [−0.36, 0.36]	0.9991	−0.10 [−0.45, 0.27]	0.5889
		Household	−0.37 * [−0.64, −0.03]	0.0359	−0.39 * [−0.65, −0.04]	0.0300	−0.40 * [−0.66, −0.05]	0.0283	−0.40 * [−0.67, −0.04]	0.0321
Aged 15–19 years	Male	Ambient	0.04 [−0.32, 0.38]	0.8426	0.09 [−0.27, 0.43]	0.6266	0.10 [−0.27, 0.45]	0.5812	0.05 [−0.32, 0.41]	0.7945
		Household	−0.57 ** [−0.77, −0.27]	0.0007	−0.48 ** [−0.72, −0.16]	0.0057	−0.38 * [−0.65, −0.02]	0.0383	−0.26 [−0.57, 0.12]	0.1717
	Female	Ambient	0.13 [−0.23, 0.46]	0.4795	0.15 [−0.22, 0.48]	0.4199	0.13 [−0.24, 0.47]	0.4974	0.04 [−0.33, 0.40]	0.8449
		Household	−0.18 [−0.50, 0.18]	0.3166	−0.11 [−0.44, 0.26]	0.5727	−0.02 [−0.38, 0.34]	0.9159	0.08 [−0.29, 0.44]	0.6707
	Both	Ambient	0.08 [−0.28, 0.42]	0.6643	0.13 [−0.24, 0.46]	0.4981	0.13 [−0.24, 0.47]	0.5030	0.06 [−0.32, 0.41]	0.7732

) examined the relationship between the BMI and PM exposure (ambient and household) across different age and gender groups, providing insights into how prior PM exposure’s correlation with the BMI was generally weak and statistically insignificant across most groups. For example, in males aged 2–4 years, the correlation coefficients ranged from −0.27 with no lag to 0.01 with a 3-year lag, with all p-values being above 0.1, indicating no significant association. Similarly, in females aged 2–4 years, ambient PM showed no significant correlation, with coefficients ranging from 0.00 with no lag to 0.36 with a 3-year lag (p = 0.0540). These results suggest that ambient PM has a limited impact on the BMI in this age group. Across other age groups, similar trends were observed. For example, in males aged 10–14 years, the ambient PM correlations ranged from 0.16 with no lag to 0.01 with a 3-year lag, all of which were statistically insignificant (p > 0.39). This indicates that ambient PM exposure has a minimal and inconsistent influence on a high BMI overall.

In contrast, household PM exposure showed consistently stronger and more significant negative correlations with the BMI, particularly in younger children. For males aged 2–4 years, high household PM had a significant negative correlation with all lags, with coefficients ranging from −0.83 with no lag (p < 0.001) to −0.76 with a 3-year lag (p < 0.001). This indicates a persistent inverse relationship between household PM and a high BMI in this group. Similarly, females aged 2–4 years showed strong negative correlations with household PM, with coefficients of −0.69 with no lag to −0.71 with a 3-year lag, all highly significant (p < 0.001). When analyzing both sexes combined in the 2–4 years age group, the household PM correlations remained significant across all lags, further supporting the sustained effect of household PM on a high BMI in young children. For children aged 5–9 years, the pattern was similar. Males exhibited strong negative correlations with household PM exposure with all lags, with coefficients such as −0.69 with no lag (p < 0.001) and −0.68 with a 3-year lag (p < 0.001). Females also showed significant negative correlations, ranging from −0.46 with no lag (p < 0.01) to −0.58 with a 3-year lag (p < 0.01). These results suggest a sustained adverse effect of household PM on a high BMI during early and middle childhood. In older age groups (10–14 and 15–19 years), the relationship between household PM exposure and a high BMI weakened. For males aged 10–14 years, significant negative correlations were observed only with no lag (−0.39, p = 0.0293) and a 1-year lag (−0.38, p = 0.0335), but the effect diminished with longer lags. Similarly, females in this age group showed significant correlations only with longer lags, such as −0.42 with a 2-year lag (p = 0.0202) and −0.43 with a 3-year lag (p = 0.0187). These findings indicate that while household PM may still influence a high BMI in older children, the relationship is less consistent compared to that of younger age groups. For adolescents aged 15–19 years, the correlation between household PM and a high BMI became weaker and often insignificant. For males, household PM showed a significant correlation with no lag (−0.57, p = 0.0007) and a 1-year lag (−0.48, p = 0.0057), but this effect diminished with 2-year (−0.38, p = 0.0383) and 3-year (−0.26, p = 0.1717) lags. In females aged 15–19 years, no significant correlations were observed with any lag, with coefficients ranging from −0.18 with no lag to 0.08 with a 3-year lag (all p > 0.31). This suggests that other factors, such as hormonal changes and lifestyle behaviors, may overshadow the effect of household PM exposure on a high BMI during late adolescence.

The results from the cross-correlation analysis revealed that household PM exposure consistently exhibited stronger and more significant negative correlations with a high BMI, particularly in younger children. For example, in males aged 2–4 years, household PM showed a significant negative correlation with no lag (−0.83, p < 0.001), which persisted with a 3-year lag (−0.76, p < 0.001). Similarly, females in this age group displayed strong negative correlations across all lags (−0.69 to −0.71, p < 0.001). In contrast, ambient PM exposure demonstrated weak and mostly insignificant correlations with a high BMI across all age and gender groups, except for females aged 2–4 years, who exhibited a significant positive correlation with longer lags (e.g., 0.62 with a 3-year lag, p = 0.0003). These findings suggest that household PM has a more immediate and lasting impact on a high BMI, while ambient PM exposure has minimal influence, with occasional exceptions.

3.3. Spearman Correlation

In this section, we discuss how the Spearman correlation analysis assessed the monotonic relationship between PM exposure and a high BMI across ages, genders, and lag times. Unlike the cross-correlation analysis, which examined temporal dynamics, this method provided a straightforward measure of the overall strength and direction of the association.

Table 2. Spearman correlation coefficients with 95% confidence intervals and p-values for the time-lagged associations between a high BMI and PM exposure (ambient and household) across age groups and genders. Single asterisk (*): indicates that the correlation is statistically significant at the 0.05 level (p < 0.05); double asterisk (**): indicates that the correlation is statistically significant at the 0.01 level (p < 0.01).

Category		Pollution Type	No Lag	p-Values	Lag 1	p-Values	Lag 2	p-Values	Lag 3	p-Values
Aged 2–4 years	Male	Ambient	−0.23 [−0.54, 0.13]	0.2035	−0.12 [−0.45, 0.25]	0.5339	0.02 [−0.34, 0.38]	0.9062	0.19 [−0.19, 0.52]	0.3361
		Household	−0.63 ** [−0.80, −0.35]	0.0001	−0.59 ** [−0.78, −0.30]	0.0005	−0.55 ** [−0.76, −0.23]	0.0018	−0.50 ** [−0.73, −0.16]	0.0061
	Female	Ambient	0.31 [−0.04, 0.60]	0.0818	0.44 * [0.10, 0.69]	0.0139	0.54 ** [0.22, 0.75]	0.0021	0.62 ** [0.33, 0.81]	0.0003
		Household	−0.58 ** [−0.77, −0.29]	0.0005	−0.57 ** [−0.77, −0.26]	0.0009	−0.55 ** [−0.76, −0.23]	0.0018	−0.52 ** [−0.74, −0.19]	0.0039
	Both	Ambient	−0.01 [−0.35, 0.34]	0.9778	0.12 [−0.25, 0.45]	0.5353	0.25 [−0.12, 0.56]	0.1814	0.41* [0.05, 0.67]	0.0291
		Household	−0.64 ** [−0.81, −0.37]	0.0001	−0.60 ** [−0.79, −0.32]	0.0003	−0.56 ** [−0.77, −0.25]	0.0012	−0.52 ** [−0.74, −0.18]	0.0041
Aged 5–9 years	Male	Ambient	−0.10 [−0.43, 0.26]	0.5941	0.02 [−0.34, 0.37]	0.9074	0.11 [−0.26, 0.45]	0.5513	0.17 [−0.21, 0.50]	0.3795
		Household	−0.70 ** [−0.84, −0.46]	0.0000	−0.68 ** [−0.84, −0.44]	0.0000	−0.66 ** [−0.82, −0.39]	0.0001	−0.63 ** [−0.81, −0.34]	0.0003
	Female	Ambient	−0.30 [−0.59, 0.05]	0.0954	−0.19 [−0.51, 0.17]	0.2979	−0.08 [−0.43, 0.29]	0.6646	0.01 [−0.36, 0.37]	0.9656
		Household	−0.58 ** [−0.77, −0.29]	0.0005	−0.63 ** [−0.80, −0.35]	0.0001	−0.63 ** [−0.81, −0.35]	0.0002	−0.61 ** [−0.80, −0.32]	0.0004
	Both	Ambient	−0.22 [−0.53, 0.14]	0.2345	−0.08 [−0.43, 0.28]	0.6506	0.03 [−0.34, 0.38]	0.8877	0.10 [−0.27, 0.45]	0.5880
		Household	−0.67 ** [−0.83, −0.42]	0.0000	−0.66 ** [−0.82, −0.40]	0.0001	−0.64 ** [−0.82, −0.37]	0.0001	−0.62 ** [−0.80, −0.33]	0.0003
Aged 10–14 years	Male	Ambient	0.34 [−0.01, 0.62]	0.0562	0.35 [−0.00, 0.63]	0.0521	0.31 [−0.06, 0.60]	0.0971	0.20 [−0.18, 0.53]	0.3056
		Household	−0.36 * [−0.63, −0.02]	0.0412	−0.35 [−0.63, 0.00]	0.0516	−0.34 [−0.62, 0.02]	0.0674	−0.30 [−0.60, 0.07]	0.1077
	Female	Ambient	0.13 [−0.23, 0.46]	0.4765	0.09 [−0.27, 0.43]	0.6167	0.03 [−0.34, 0.38]	0.8803	−0.05 [−0.41, 0.33]	0.8134
		Household	−0.42 * [−0.67, −0.08]	0.0181	−0.45 * [−0.69, −0.11]	0.0115	−0.47 ** [−0.71, −0.13]	0.0086	−0.48 ** [−0.72, −0.14]	0.0078
	Both	Ambient	0.26 [−0.09, 0.56]	0.1467	0.26 [−0.10, 0.56]	0.1590	0.22 [−0.15, 0.54]	0.2417	0.13 [−0.25, 0.47]	0.5095
		Household	−0.34 [−0.62, 0.01]	0.0568	−0.35 [−0.63, 0.00]	0.0507	−0.37 * [−0.64, −0.01]	0.0459	−0.35 [−0.64, 0.02]	0.0609
Aged 15–19 years	Male	Ambient	0.17 [−0.19, 0.49]	0.3648	0.22 [−0.15, 0.53]	0.2438	0.20 [−0.17, 0.52]	0.2882	0.12 [−0.26, 0.46]	0.5447
		Household	−0.39 * [−0.65, −0.05]	0.0265	−0.33 [−0.61, 0.03]	0.0689	−0.26 [−0.57, 0.11]	0.1622	−0.18 [−0.51, 0.20]	0.3427
	Female	Ambient	0.33 [−0.02, 0.61]	0.0678	0.28 [−0.08, 0.58]	0.1256	0.19 [−0.18, 0.51]	0.3175	0.05 [−0.33, 0.41]	0.8115
		Household	−0.18 [−0.50, 0.18]	0.3143	−0.10 [−0.44, 0.26]	0.5805	−0.02 [−0.38, 0.34]	0.9247	0.08 [−0.30, 0.43]	0.6826
	Both	Ambient	0.25 [−0.10, 0.55]	0.1594	0.28 [−0.09, 0.57]	0.1320	0.23 [−0.14, 0.55]	0.2119	0.13 [−0.25, 0.47]	0.5030
		Household	−0.35 * [−0.62, −0.00]	0.0485	−0.29 [−0.58, 0.07]	0.1174	−0.21 [−0.53, 0.16]	0.2576	−0.13 [−0.47, 0.25]	0.5046

presents the time-lagged Spearman correlation analysis between the BMI and both ambient and household PM exposure across age and gender groups. Overall, household PM exposure showed stronger and more consistent associations with the BMI than ambient PM exposure, particularly in younger children. For ambient PM exposure, correlations with the BMI were generally weak and mostly insignificant across age and gender groups. An exception was observed in females aged 2–4 years, where there were significant positive correlations with 1-, 2-, and 3-year lags (0.44, 0.54, and 0.62, respectively; p = 0.0139, 0.0021, and 0.0003). This suggests a delayed positive relationship where higher past PM levels may have been correlated with a greater high BMI in this group. For most other age groups, the correlations remained close to zero. For example, in males aged 2–4 years, the coefficients ranged from −0.23 (no lag; p = 0.2035) to 0.19 (3-year lag; p = 0.3361), indicating no significant relationship between ambient PM exposure and a high BMI. Similarly, in children aged 5–9 years, ambient PM correlations were weak and insignificant for both males and females, with coefficients close to zero across all lags. In contrast, household PM exposure showed consistently strong and significant negative correlations with a high BMI, particularly in younger age groups. Among males aged 2–4 years, household PM exposure had highly significant negative correlations across all lags, with coefficients ranging from −0.63 (no lag; p < 0.001) to −0.50 (3-year lag; p = 0.0061). Similarly, females in the same age group showed significant negative correlations, ranging from −0.58 with no lag (p = 0.0005) to −0.52 with a 3-year lag (p = 0.0039). When analyzing both sexes combined in the 2–4 years category, the pattern persisted, with correlations ranging from −0.64 (no lag; p < 0.001) to −0.52 (3-year lag; p = 0.0041). These results suggest that higher household PM exposure was consistently associated with a lower high BMI in young children, possibly due to increased indoor exposure during early developmental periods. In the 5–9 years age group, the relationship between household PM and a high BMI remained significant. For males, correlations ranged from −0.70 with no lag (p < 0.001) to −0.63 with a 3-year lag (p = 0.0003). Females showed similarly strong negative correlations, with coefficients ranging from −0.58 (no lag; p = 0.0005) to −0.61 (3-year lag; p = 0.0004). When both sexes were combined, the household PM correlations remained consistently significant across all lags, further emphasizing the sustained impact of household PM on a high BMI during childhood. In older age groups (10–14 and 15–19 years), the influence of household PM exposure on a high BMI weakened. Among males aged 10–14 years, a significant negative correlation was observed only with no lag (−0.36, p = 0.0412), with weaker and insignificant correlations with longer lags. For females in this age group, significant negative correlations were found with 1-, 2-, and 3-year lags (−0.45, −0.47, and −0.48, respectively; all p < 0.05). These results suggest that household PM exposure may still influence a high BMI during early adolescence, but the relationship diminishes as children age. In adolescents aged 15–19 years, the relationship between household PM and the BMI was weaker and less consistent. For males, a significant correlation was observed only with no lag (−0.39, p = 0.0265), while correlations with longer lags were insignificant. Females in this age group showed no significant correlations between household PM and a high BMI across all lags, with coefficients close to zero. These findings indicate that other factors, such as lifestyle changes, physical activity, and hormonal influences, may play a more dominant role in determining the BMI during late adolescence.

The results from the Spearman correlation analysis corroborate the findings from the cross-correlation analysis, highlighting the stronger relationship between household PM exposure and a high BMI. For example, in males aged 2–4 years, household PM showed significant negative correlations with all lags, with coefficients ranging from −0.63 (no lag; p = 0.0001) to −0.50 (3-year lag; p = 0.0061). In females aged 5–9 years, household PM also exhibited strong negative correlations, such as −0.58 with no lag (p = 0.0005) and −0.61 with a 3-year lag (p = 0.0004). On the other hand, the ambient PM exposure correlations remained weak and insignificant for most groups, reinforcing the limited impact of ambient PM on a high BMI across this population.

3.4. Mixed-Effects Model Analysis

In this section, we describe how we used the mixed-effects model to examine the long-term relationship between PM exposure and a high BMI while accounting for repeated measures within individuals and controlling for covariates such as age, gender, and socioeconomic factors. This approach allowed for a more robust evaluation of how PM exposure impacts a high BMI over time. The mixed-effects model analysis (Figure 3 and Figure 4) explored trends in high BMI values alongside ambient and household PM exposure across different age and gender groups from 1990 to 2021, as shown and quantified in

Table 3. Mixed-effects model analysis of relationship between high BMI and ambient PM exposure (1990–2021).

Age	Group	High BMI-PM Corr.	R2	High BMI Range	PM Range	Trend Description
2–4 years	male	−0.27	0.83	122.9–170.0	35.0–10.9	BMI: Increasing; PM: Decreasing
	female	0	0.7	124.6–141.8	25.6–11.7	BMI: Increasing; PM: Decreasing
	both	−0.19	0.8	247.5–311.8	60.6–22.6	BMI: Increasing; PM: Decreasing
5–9 years	male	−0.24	0.65	183.6–256.3	20.9–11.3	BMI: Increasing; PM: Decreasing
	female	−0.3	0.39	198.4–228.5	20.8–9.5	BMI: Increasing; PM: Decreasing
	both	−0.29	0.54	382.0–484.8	41.7–20.8	BMI: Increasing; PM: Decreasing
10–14 years	male	0.16	0.33	247.6–293.4	6.5–4.5	BMI: Increasing; PM: Decreasing
	female	−0.01	0.24	227.5–258.1	7.2–4.9	BMI: Increasing; PM: Decreasing
	both	0.08	0.28	475.0–551.6	13.7–9.3	BMI: Increasing; PM: Decreasing
15–19 years	male	0.04	0.35	113.7–141.8	5.1–3.9	BMI: Increasing; PM: Decreasing
	female	0.13	0.07	145.2–156.1	6.5–4.9	BMI: Increasing; PM: Decreasing
	both	0.08	0.25	258.9–297.9	11.6–8.8	BMI: Increasing; PM: Decreasing

Notes: 1. High BMI-PM Corr: correlation between high BMI and PM exposure; 2. R²: model fit quality (higher values indicate better fit); 3. High BMI Range: observed high BMI values’ range over study period; 4. PM Range: PM concentration range (μg/m³); 5. Trend Description: overall direction of high BMI and PM changes.

and

Table 4. Mixed-effects model analysis of relationship between high BMI and household PM exposure (1990–2021).

Age	Group	High BMI-PM Corr.	R2	High BMI Range	PM Range	Trend Description
2–4 years	male	−0.83	0.7	122.9–170.0	42.2–0.6	BMI: Increasing; PM: Decreasing
	female	−0.69	0.5	124.6–141.8	46.3–0.9	BMI: Increasing; PM: Decreasing
	both	−0.79	0.62	247.5–311.8	88.5–1.6	BMI: Increasing; PM: Decreasing
5–9 years	male	−0.69	0.49	183.6–256.3	32.2–0.8	BMI: Increasing; PM: Decreasing
	female	−0.46	0.31	198.4–228.5	32.5–0.6	BMI: Increasing; PM: Decreasing
	both	−0.6	0.41	382.0–484.8	64.6–1.4	BMI: Increasing; PM: Decreasing
10–14 years	male	−0.39	0.15	247.6–293.4	10.1–0.3	BMI: Increasing; PM: Decreasing
	female	−0.35	0.15	227.5–258.1	11.0–0.3	BMI: Increasing; PM: Decreasing
	both	−0.37	0.14	475.0–551.6	21.1–0.7	BMI: Increasing; PM: Decreasing
15–19 years	male	−0.57	0.38	113.7–141.8	7.9–0.3	BMI: Increasing; PM: Decreasing
	female	−0.18	0.07	145.2–156.1	10.1–0.3	BMI: Increasing; PM: Decreasing
	both	−0.45	0.25	258.9–297.9	18.0–0.6	BMI: Increasing; PM: Decreasing

Notes: 1. High BMI-PM Corr: correlation between high BMI and PM exposure; 2. R²: model fit quality (higher values indicate better fit); 3. High BMI Range: observed high BMI values’ range over study period; 4. PM Range: PM concentration range (μg/m³); 5. Trend Description: overall direction of high BMI and PM changes.

.

For the ambient PM exposure relationship, the analysis revealed age-specific patterns. In younger age groups (ages of 2–4 and 5–9 years), both males and females exhibited a inverse association between ambient PM exposure and a high BMI, with correlation coefficients ranging from −0.27 to −0.30. The 2–4 years age group showed a particularly strong model fit (R² = 0.83 for males, 0.70 for females), with high BMI values increasing from 122.9 to 170.0 kg/m² for males and 124.6 to 141.8 kg/m² for females as the PM levels decreased from 35.0 to 10.9 μg/m³. This suggests that improved air quality might be associated with an increased high BMI in early childhood, though the moderate correlation strengths indicated other influential factors. For children aged 10–14, the relationship weakened substantially, with correlation coefficients near zero (0.16 for males, −0.01 for females) and a lower model fit quality (R² = 0.33 and 0.24, respectively). While ambient PM continued to decrease (6.5 to 4.5 μg/m³), high BMI values showed consistent increases across both genders (247.6 to 293.4 kg/m² for males, 227.5 to 258.1 kg/m² for females), suggesting that other factors, particularly developmental changes, may overshadow PM exposure effects during this period. For adolescents aged 15–19 years, the mixed-effects model identified a noticeable dip in high BMI values during 2005–2010 across both ambient and household PM exposure analyses (Figure 3 and Figure 4). The high BMI trends for males, females, and combined groups exhibited a transient decline before resuming an upward trajectory. While this decline did not correspond to any observable changes in the PM exposure trends, it may reflect behavioral changes, environmental influences, or modeling variations during this period. Factors such as dietary shifts, increased physical activity, or temporary health interventions targeting adolescents could have contributed to this deviation, underscoring the complexity of BMI trends during late adolescence. Despite this decline, the overall trend across the study period remained upward, aligning with findings from global analyses.

Household PM exposure demonstrated notably stronger relationships, especially in younger age groups. For the ages of 2–4 years, the correlation coefficients were substantially higher (−0.83 for males, −0.69 for females) compared to those of ambient PM exposure, with a good model fit (R² = 0.70 and 0.50, respectively). The analysis showed dramatic decreases in the household PM levels (42.2 to 0.6 μg/m³ for males, 46.3 to 0.9 μg/m³ for females) concurrent with BMI increases, suggesting a more direct influence of the indoor air quality on early childhood growth patterns. This inverse relationship persisted but weakened with age. In the 5–9 age group, the correlations remained moderate to strong (−0.69 for males, −0.46 for females), while the 10–14 years age group showed further attenuation (−0.39 for males, −0.35 for females) with a lower model fit quality (R² = 0.15). By adolescence (15–19 years), the relationship varied by gender, with males showing stronger correlations (−0.57) compared to females (−0.18), though both groups maintained an inverse pattern as the household PM levels decreased to near-zero concentrations.

The model results indicate a significant negative association between household PM exposure and a high BMI, particularly in early childhood. For instance, among children aged 2–4 years, a 1-unit increase in household PM exposure was associated with a 0.45-unit decrease in high BMI values (p < 0.01), even after adjusting for confounding variables. Conversely, ambient PM exposure showed no significant relationship with a high BMI in most age groups, aligning with the findings from the correlation analyses. In older age groups (e.g., 15–19 years), the effect of household PM diminished, suggesting that other factors, such as lifestyle behaviors and hormonal changes, become more influential during adolescence. The transient decline in a high BMI observed during 2005–2010 may highlight the need for further research to examine localized or temporary factors influencing BMI trends in adolescents, such as public health campaigns, economic transitions, or temporal shifts in the environmental or social determinants of health.

4. Discussion

This correlational study explored associations between ambient and household PM exposure and a high BMI across children and adolescents (aged 2–19) from 1990 to 2021 in Thailand. The findings demonstrate a pronounced inverse relationship between household PM exposure and the BMI in younger children (ages 2–9), with weaker and less consistent associations observed for ambient PM. These relationships attenuated as children transitioned into adolescence, highlighting the age-dependent effects of PM exposure on the BMI.

Thailand’s environmental and economic context evolved significantly during the study period (1990–2021), marked by rapid industrialization, urbanization, and changing patterns of agricultural activity [8,43]. The current economic structure, dominated by manufacturing (33% of the GDP) and services (58% of the GDP) [44], creates distinct exposure environments that vary by both the location and demographic groups. Manufacturing activities, concentrated in industrial estates, contribute significantly to ambient PM levels in urban areas, particularly affecting adolescents who spend time in these regions [45]. The temporal evolution of PM exposure reflects Thailand’s development trajectory. Household PM levels have shown dramatic improvement, declining from approximately 80 µg/m³ in the early 1990s to near-zero levels by 2020 [20], primarily due to improvements in indoor cooking technologies and reduced reliance on solid fuels. However, the ambient PM levels have decreased more gradually, reflecting ongoing urban and industrial emissions [22]. Age-specific exposure patterns remained a crucial consideration throughout this period. Young children (aged 2–4) experience primarily residential exposure, spending 14–16 h at home [46,47], where PM levels are influenced by both household activities and the infiltration of outdoor pollution. School-age children (5–14 years) face a combination of residential and educational environment exposure, with PM levels affected by both local traffic and industrial emissions [46]. Adolescents (15–19 years) experience the most diverse exposure patterns due to their increased mobility across urban environments [48]. These patterns are further modified by Thailand’s agricultural sector (9% of the GDP) [44], particularly through seasonal burning practices. Agricultural burning creates temporal spikes in the PM levels, especially affecting rural and peri-urban areas, where many school-age children reside [49]. While the overall air quality improved over the study period, seasonal agricultural burning and urban pollution hotspots remained key contributors to PM exposure in specific regions [19].

The results highlight the stronger and more consistent negative association between household PM exposure and a high BMI in younger children. For children aged 2–4 years, significant inverse correlations were observed across multiple lag times, suggesting lasting impacts of household PM exposure on growth and weight gain. For example, males in this age group exhibited a correlation of −0.83 (p < 0.001) with no lag, which persisted with similar magnitudes with 1-, 2-, and 3-year lags. These findings align with previous research linking household PM exposure to impaired growth outcomes, including stunting and underweight, in children from low- and middle-income countries [19,20,33,50]. The chronic inflammation, oxidative stress, and respiratory infections associated with PM exposure are likely mechanisms contributing to this relationship. Inflammatory markers such as cytokines can interfere with nutrient absorption and utilization, while oxidative stress leads to cellular damage and impaired metabolic efficiency [16]. Conversely, the associations between ambient PM exposure and a high BMI were weaker and less consistent. In younger children (aged 2–9 years), ambient PM exposure showed moderate inverse correlations, particularly in females aged 2–4, but these relationships were not as robust or persistent as those observed for household PM. This discrepancy may reflect the cumulative exposure differences between household and ambient PM. Young children spend the majority of their time indoors, where exposure to household PM is more prolonged and intense. In contrast, ambient PM exposure is influenced by outdoor activities and environmental factors, which may vary substantially across individuals and regions [13,14].

As children aged, the associations between PM exposure and a high BMI weakened, particularly during adolescence (ages 10–19). For instance, household PM exposure in males aged 10–14 exhibited a correlation of −0.39 (p < 0.05) with no lag, which diminished further with increasing lag times. By late adolescence (ages 15–19), household PM exposure was only weakly associated with a high BMI in males (correlation = −0.57, p < 0.01) and was generally insignificant in females. This attenuation may reflect the increasing influence of other factors on a high BMI during adolescence, including hormonal changes, the diet, physical activity, and social behaviors [17,25]. Hormonal shifts during puberty, such as increased levels of growth hormone and sex steroids, significantly influence the body composition, fat distribution, and metabolic processes, potentially overshadowing the effects of PM exposure [7]. Adolescents also experience greater mobility and spend more time outdoors, reducing the household PM exposure and diversifying their exposure patterns to ambient PM [46]. The stronger associations observed in younger children highlight their heightened vulnerability to PM exposure due to their developing physiology and metabolic systems. Chronic exposure to household PM can impair growth through mechanisms such as systemic inflammation, oxidative stress, and respiratory infections, which collectively reduce nutrient absorption and metabolic efficiency [16]. For example, oxidative stress caused by PM exposure generates reactive oxygen species (ROS), leading to cellular damage and reduced energy availability for growth [15]. Additionally, PM exposure has been linked to gut microbiota dysbiosis, which disrupts nutrient absorption and the energy balance [27]. These mechanisms are particularly relevant in early childhood, a critical period for physical growth and metabolic development. In contrast, the weaker associations observed in adolescents may reflect the complex interplay between environmental exposure and physiological changes during puberty. Growth spurts and hormonal activity during adolescence contribute to rapid changes in the body composition, which may dilute the effects of PM exposure on the BMI [51]. Moreover, lifestyle factors such as increased physical activity, dietary changes, and reduced time spent indoors further modulate the impact of PM exposure on metabolic outcomes [48,52].

The analysis of the high BMI trends across age groups revealed a general increase over the study period (1990–2021), consistent with global patterns reported in the GBD study [28]. However, localized fluctuations were observed, particularly among adolescents aged 15–19 years, where the BMI appeared to decline slightly after 2005–2010 before resuming an upward trajectory. This trend differs from the GBD study-reported continuous rise in the high BMI prevalence across all demographic groups and time periods. Such discrepancies may arise from differences in the methodologies or regional factors. The report from the GBD 2021 study integrates global datasets and employs comprehensive modeling approaches to estimate the high BMI prevalence, reporting an annual increase of 1.8% in metabolic risk exposure, including a high BMI, between 1990 and 2021 [28]. In contrast, the trends observed in this study were derived from population-specific longitudinal data that capture both national and local dynamics, potentially highlighting variability not fully represented in global models. Localized declines in the BMI during 2005–2010 may reflect transient changes in adolescent behaviors or environmental influences, such as shifts in the diet, physical activity, or public health interventions [53]. During specific time periods, localized declines in the BMI among adolescents can emerge from a complex interplay of behavioral adaptations and environmental changes. These temporary shifts often reflect both individual-level behavioral modifications and broader societal influences that collectively impact adolescent weight trajectories [36]. The behavioral changes typically manifest through alterations in physical activity patterns, where adolescents might temporarily engage more frequently in structured exercise, organized sports, or active recreational activities. These changes could stem from school-based initiatives, community programs, or shifting social norms that temporarily prioritize physical fitness [48,54]. Dietary behaviors during these periods often undergo simultaneous transformations. Adolescents might experience modifications in their eating patterns due to various factors, including increased nutritional awareness, changes in food availability, or shifts in family dietary habits [35]. These dietary adjustments could involve the reduced consumption of high-calorie foods, an increased intake of fruits and vegetables, or more structured meal timing, all of which can contribute to temporary BMI reductions [34]. Environmental influences play a crucial role in facilitating these behavioral changes. The physical environment, including the availability of recreational spaces, safe walking paths, or sports facilities, can temporarily enhance opportunities for physical activity [52]. Additionally, social environments, such as school settings or community spaces, might implement temporary programs or policies that promote healthier lifestyle choices. These environmental modifications can create supportive conditions that encourage and sustain healthier behaviors among adolescents [45]. Public health interventions during these periods often serve as catalysts for change through multiple mechanisms. These interventions might include educational campaigns that raise awareness about healthy lifestyle choices, policy changes that modify food environments in schools or communities, or structured programs that provide resources and support for maintaining healthy behaviors [37]. The effectiveness of these interventions often depends on their intensity, duration, and the degree of community engagement they generate [55]. However, the transient nature of these changes highlights the challenges of maintaining long-term behavioral modifications. As environmental support or intervention programs evolve or diminish, adolescents may gradually revert to previous behavioral patterns [8]. This phenomenon underscores the importance of understanding not just how to initiate positive changes in adolescent health behaviors, but also how to sustain these improvements over time through consistent environmental support and ongoing intervention strategies [2,8]. Additionally, modeling artifacts or the smoothing techniques applied in our mixed-effects models may have contributed to these patterns. Mixed-effects models estimate fixed and random effects to account for variability across subgroups, but this can introduce artifacts, such as transient declines in the BMI, particularly when combined with smoothing techniques or imputation for missing data [56]. These methods, while robust, may amplify short-term deviations or fluctuations, especially in subgroups like adolescents aged 15–19 years, where the sample sizes or data availability might be more limited. Despite these subgroup-specific variations, the overarching trend aligns with the GBD framework, showing a steady increase in the high BMI prevalence over the three-decade period. This consistency underscores the reliability of the overall findings and their alignment with global trends [28].

Evidence suggests that PM exposure impacts the BMI through several interrelated pathways. One of the primary mechanisms is systemic inflammation and oxidative stress. PM exposure triggers the release of pro-inflammatory cytokines, such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-α), which disrupt metabolic homeostasis [2]. Chronic inflammation has been linked to insulin resistance and altered lipid metabolism, both of which can contribute to weight gain and increased adiposity [57]. Additionally, PM-induced oxidative stress generates ROS and malondialdehyde (MDA), a key biomarker of lipid peroxidation, leading to mitochondrial dysfunction and an impaired energy balance [16]. Elevated MDA levels have been associated with cellular damage, metabolic disturbances, and impaired growth regulation, particularly in developing children [19,39]. These effects are particularly pronounced during critical developmental periods, such as early childhood and adolescence, when metabolic systems are still maturing [58]. Another important pathway involves endocrine disruption. PM contains endocrine-disrupting chemicals (EDCs), such as polycyclic aromatic hydrocarbons (PAHs) and heavy metals, which interfere with hormone signaling pathways [4]. EDCs can dysregulate appetite control by altering the leptin and ghrelin levels, promoting increased caloric intake and fat storage [59]. Furthermore, PM exposure may disrupt thyroid hormone function, leading to changes in the basal metabolic rate and growth regulation [60]. Beyond endocrine effects, PM also influences neuroinflammatory pathways, which are critical for appetite regulation and the energy balance. Pollutants can cross the blood–brain barrier and activate neuroinflammatory responses in the hypothalamus, a region of the brain that regulates hunger and metabolism [61]. This interference may impair hormonal signaling, such as insulin and leptin signaling, contributing to an increased risk of obesity, especially during neurodevelopmentally sensitive periods [51]. Emerging evidence also points to gut microbiota dysbiosis as a mechanism linking PM exposure to the BMI. PM has been shown to alter the composition of gut microbiota, reducing microbial diversity and promoting an imbalance in gut bacteria [15]. This dysbiosis can impair the energy harvest, increase gut permeability, and exacerbate systemic inflammation, all of which contribute to obesity [33,62]. Children exposed to high levels of PM, particularly from biomass fuel combustion in rural areas, are at greater risk of microbiota disruption due to prolonged exposure to airborne pollutants [14]. However, beyond these biological mechanisms, the findings also suggest the influence of environmental, dietary, and genetic factors. For example, children in urban areas may experience different exposure compared to those in rural areas, where biomass combustion is more prevalent [63,64]. Additionally, dietary behaviors—such as the caloric intake, nutrient deficiencies, or unhealthy food environments—could modify the relationship between PM exposure and a high BMI [15]. For instance, diets high in processed foods may exacerbate oxidative stress and MDA accumulation, compounding the metabolic effects of PM exposure. Moreover, a genetic predisposition may influence the susceptibility to PM-induced inflammation, as certain polymorphisms in cytokine or antioxidant genes have been linked to variability in inflammatory responses [61,64]. Finally, chronic PM exposure can lead to hypoxia, or reduced oxygen delivery to tissues, due to alveolar inflammation and impaired lung function [63]. Hypoxia, in turn, activates hypoxia-inducible factors (HIFs), which stimulate lipogenesis and promote fat accumulation [17]. These multifactorial interactions underscore the complexity of the PM-BMI relationship and highlight the need for further research to disentangle the contributions of the environment, diet, and genetic factors in shaping these outcomes.

This study has several limitations that may influence the generalizability and interpretation of the findings. Firstly, a high BMI is a broad metric that does not fully capture the diverse health impacts of PM exposure. Future research should include additional health metrics, such as respiratory and cardiovascular outcomes, as well as markers of systemic inflammation or metabolic dysregulation, to provide a more comprehensive assessment of PM’s health effects. The socioeconomic status (SES) is another critical factor that influences both PM exposure and the BMI. While this study acknowledges the SES as a potential confounder, it did not explicitly quantify or analyze its effects. A lower SES is often associated with greater household PM exposure due to reliance on biomass fuels and limited access to cleaner technologies, as well as a poorer diet quality and restricted access to health resources. Including the SES as a variable in future analyses could elucidate its role in modulating the observed associations. Additionally, the reliance on GBD data introduces certain constraints. Although the GBD dataset is a robust and comprehensive source, the accuracy of PM exposure estimates may vary across time periods due to changes in the monitoring methods and data quality, particularly in earlier years. Moreover, household PM exposure was estimated from survey data, which may be subject to recall bias and reporting inaccuracies. The use of population-weighted PM values, while practical for large-scale analyses, may also mask localized variations in exposure that could influence the results. The study’s temporal analysis highlighted a transient decline in a high BMI among adolescents aged 15–19 years during 2005–2010, which deviated from the broader upward trend. While this anomaly may reflect behavioral changes, public health interventions, or environmental shifts during this period, the lack of direct data on these factors limits our ability to draw definitive conclusions. Future research should explore the mechanisms behind such localized deviations, potentially through longitudinal cohort studies or regional surveys that capture behavioral and environmental dynamics. In terms of future work, there is a need for more nuanced datasets that integrate multiple health outcomes, high-resolution exposure data, and detailed demographic information. Incorporating biomarkers of exposure and effects, such as particulate-bound toxins or inflammatory markers, could enhance the mechanistic understanding of PM’s impact on the BMI. Furthermore, extending this research to other middle-income countries with similar socioeconomic and environmental profiles would help validate the findings and assess their applicability in diverse settings. Finally, studies employing advanced modeling techniques, such as spatial analysis or machine learning, could improve exposure estimates and provide more precise assessments of PM’s health impacts at the community level.

5. Conclusions

This correlational study examined the association between ambient and household PM exposure and a high BMI across different age and gender groups among children and adolescents aged 2 to 19 years. The results provide strong evidence of an inverse relationship between household PM exposure and the BMI in younger children (aged 2–9 years), with the relationship weakening during adolescence (ages of 10–19 years). This suggests that early-life exposure to household PM, likely due to indoor pollution from biomass combustion and cooking fuels, may negatively impact growth trajectories. In contrast, ambient PM exposure demonstrated weaker and more inconsistent associations with the BMI, likely reflecting lower cumulative exposure levels and differences in the pollutant composition. This study also identified age- and gender-specific trends, including a transient decline in the BMI among adolescents aged 15–19 years during 2005–2010, which may have been influenced by behavioral changes, environmental factors, or public health interventions. These findings underscore the critical role of indoor air quality in shaping the childhood BMI, highlighting the need for targeted interventions, such as cleaner household cooking technologies and ventilation improvements, to reduce exposure risks. However, some limitations should be considered. The study relied on population-level data and did not include the individual-level socioeconomic status (SES), dietary habits, or physical activity, which may also influence BMI trends. Additionally, while the study demonstrates significant associations, the causal mechanisms remain uncertain and warrant further investigation. Future research should employ longitudinal cohort studies with direct exposure measurements and detailed socioeconomic data to better understand the interplay between air pollution, nutrition, and the BMI. Comparative studies across different socioeconomic and environmental settings could also provide deeper insights into how contextual factors shape these relationships. Addressing these gaps will be essential in developing evidence-based policies to mitigate air pollution’s impact on child and adolescent health.