1. Introduction
Obesity and overweight are serious health problems and pose significant health risks [1]. Approximately 16% of adults aged 18 years and older worldwide were obese, and the global prevalence of obesity more than doubled from 1990 to 2022 [2]. Obesity-related non-communicable diseases (NCDs) are increasing accordingly. Moreover, it was estimated that obesity was associated with 5 million deaths by 2019 [3]. The COVID-19 pandemic further exacerbated the obesity crisis worldwide, resulting from reduced physical activity and dietary changes, which contributed to increased rates of obesity across all age groups [4, 5]. This trend highlights the interplay between global health crises and lifestyle-related risk factors, underscoring the urgent need for comprehensive public health strategies to address the growing burden of obesity.
Childhood obesity and overweight are also serious public health problems, and their rates are increasing. The percentage of obese children and adolescents aged 5–19 years globally increased from an estimated 2% to 8% between 1990 and 2022 [6]. Like adults, children with obesity are at risk of developing various NCDs, such as hypertension [7], type 2 diabetes [8], dyslipidemia [9], metabolic dysfunction-associated steatotic liver disease [10], asthma [11], and polycystic ovary syndrome [12].
The impact of childhood obesity includes not only physical but also psychosocial problems. Childhood obesity negatively affects school performance and quality of life, often exacerbated by stigma, discrimination, and bullying [13]. These issues increase the risk of developing obstructive sleep apnea [14], depression [15], and low self-esteem [16]. Furthermore, childhood obesity increases the likelihood of obesity in school age, adolescence, and adulthood, leading to long-term health consequences [17, 18]. This progression can be partly explained by the fact that the number of adipocytes, a major determinant of adult fat mass, is primarily set during childhood and adolescence [19]. This prolonged trajectory negatively impacts individual well-being and imposes substantial economic burdens, including increased healthcare costs and productivity losses, affecting families, the healthcare system, and society as a whole [20].
For the targeted and effective prevention of childhood obesity, extensive efforts have been made to identify critical risk factors and develop risk prediction methods that can guide early interventions. In the current review, we focus on the progress and perspectives for preventing future obesity. First, we discuss the risk factors for childhood obesity, including environmental, genetic, and epigenetic factors, with a particular emphasis on Japan-specific research. Next, we discuss the technical and ethical challenges of prediction. Finally, we outline perspectives and remaining issues in the prevention of childhood obesity.
2. Risk Factors for Childhood Obesity
2.1. Structure and Cause
Childhood obesity results from a complex interaction between genetic predispositions and environmental influences, with the impact of each shifting across developmental stages. Obesity in children is broadly categorized as primary (non-syndromic) or secondary (syndromic) [21, 22]. Primary obesity, accounting for approximately 90% of cases, results from an imbalance between energy intake and expenditure, driven by genetic and environmental factors, with the latter being modifiable through prevention strategies [23]. In contrast, secondary obesity is linked to specific medical conditions, including endocrine disorders (e.g., Cushing’s syndrome, insulinoma), genetic syndromes (e.g., Prader-Willi syndrome, Turner syndrome), and hypothalamic injury due to tumor, inflammation, surgery, or radiation. These underlying conditions typically require specialized medical diagnosis and treatment, making them less suitable for broad prevention approaches [23, 24]. Thus, primary obesity is generally the target of prevention efforts.
2.2. Environmental Factors
Environmental influences on childhood body mass index (BMI) are broadly classified as shared and non-shared factors, as shown in historical twin studies [25]. Shared factors—such as family lifestyle and socioeconomic status (SES), parenting practices, diet, and cultural or neighborhood factors—exert a considerable impact on BMI during childhood [23, 26]. As children transition to adulthood, however, the influence of these shared factors often diminishes [27]. Instead, non-shared environmental factors, including individual personality traits, personal experiences, and habits become more significant [28]. Furthermore, such lifestyle habits formed in childhood tend to persist through adolescence into adulthood [29-32]. Therefore, early intervention targeting the environmental factors of preschool children and their caregivers is considered the most effective approach for reducing long-term obesity risk [23].
Children are unable to make lifestyle choices independently, which is a major difference from the environmental factors affecting adults. This feature of children’s environmental risk factors is reflected in the association between childhood obesity/overweight and family SES indicators, such as parental education, parental employment status, and family-perceived wealth [26]. SES often limits access to health-promoting resources such as nutritious food, safe environments for physical activity, and healthcare support, thereby shaping an obesogenic environment [33, 34]. Thus, family interventions, including lifestyle interventions based on behavioral modification strategies, are recognized as the first line of therapy for childhood obesity [23]. As described below, three key factors—eating habits, physical activity, and sleep—are primary targets for effective interventions and significantly influence future health outcomes [35].
Childhood is a critical period for establishing eating behaviors, including satiety responsiveness, responsiveness to visual and olfactory food stimuli, food fussiness, and food choices, which tend to stabilize in later developmental stages [29, 30]. Therefore, early intervention targeting eating behaviors holds promise. Poor eating habits, including consumption of excessive high-calorie foods and sugary beverages and inadequate fruit and vegetable intake, contribute to obesity [36]. Children’s dietary patterns and eating habits are influenced by parenting styles (such as authoritarian, permissive, or authoritative): authoritative approaches and practices like monitoring and encouragement support healthier food choices [37]. Parents serve as essential role models for healthy eating, shaping children’s habits through shared family meals, consistent meal timing, appropriate portions, exposure to diverse nutritious foods, and promoting healthy snack choices [37].
Physical activity is another environmental factor influencing childhood obesity [4]. Children’s physical activity levels are influenced by a variety of factors, including individual traits, family support, and access to recreational spaces, as well as broader socio-economic and cultural influences [38]. These factors shape both the frequency and intensity of physical activity, which play a key role in energy balance and weight management [38]. As a modifiable factor of energy expenditure, physical activity accounts for about 25% of total energy expenditure and serves as a significant lever in managing the energy balance equation [39]. Excessive screen usage has physical as well as psychological effects on children [40]. Encouraging outdoor activity has been suggested as a means to mitigate the negative impacts of increased screen time and lead to healthier BMI levels [41].
Insufficient sleep is a risk factor for childhood obesity and is becoming increasingly common among children [42]. It is associated with hormonal imbalances, such as increased ghrelin levels, which stimulate appetite, and decreased leptin levels, which suppress satiety, thereby promoting hyperphagia and weight gain [43]. Poor sleep also affects cortisol rhythms, elevating nighttime cortisol levels and flattening the usual diurnal pattern, which impairs insulin sensitivity and increases obesity risk [44]. While most research on sleep-related hormonal imbalances has focused on adults, similar mechanisms likely contribute to childhood obesity [42]. In addition, screen use is strongly associated with the insufficient sleep in children [32], probably due to disrupted melatonin secretion and shifts in circadian rhythm caused by blue light exposure [45, 46].
2.3. Genetic Factors
Genetic factors in obesity have been studied since early genetics, and the number of genes identified has increased with the development of genome-wide association studies (GWAS), which are global methods used to identify associations between traits influenced by numerous common susceptibility variants and genetic loci [21]. Heritability estimates suggest that approximately 30% of the variance in adult BMI is attributable to genetic factors, based on studies conducted primarily on individuals of European ancestry [47]. Research on heritability indicates that genetic influence on BMI is stronger in childhood but tends to decline in adulthood (Fig. 1) [48-50]. This suggests that the influence of genetic predisposition on BMI is modulated by age-related changes in environmental exposures and lifestyle autonomy.
Non-syndromic obesity, which accounts for the majority of childhood obesity cases as previously noted, includes both monogenic and polygenic forms. Monogenic obesity in childhood is variable but it explains less than 1% of cases [22]. It is caused by rare, often highly penetrant variants in single genes involved in key obesity pathogenesis pathways, such as leptin-melanocortin signaling pathway (MC4R, LEP, LEPR, POMC, PCSK1, MC3R, MRAP2, SH2B1, SIM1) [21]. These mutations often result in early-onset, severe obesity by disrupting key pathways involved in energy balance and appetite regulation [21].
In contrast, polygenic obesity, the much more common form of non-syndromic obesity, is caused by the cumulative effects of multiple genetic variants, each contributing relatively small effects. Genes associated with polygenic obesity have mainly been identified in GWASs for adult BMI rather than childhood BMI. More than 700 associated loci have been identified in large-scale GWASs, such as the Genetic Investigation of ANthropometric Traits (GIANT) consortium studies (∼700,000 individuals of European ancestry) [51] and Biobank Japan (173,430 individuals from Japanese populations) [52]. These findings underscore the polygenic nature of obesity but also highlight the limitations of conventional GWASs, which primarily detect common variants and struggle to capture rare variants or non-additive effects [53]. To address this gap and further clarify the genetic architecture of obesity, alternative approaches, including rare variant analysis based on whole-genome sequencing [47], are needed.
Previous studies suggest that part of the composition of genetic factors in polygenic obesity differs between adults and children [54-56]. For example, OLFM4 and HOXB5 appear to be unique to pediatric populations, as shown by GWAS in children, while FTO, MC4R, and TMEM18 are associated with both childhood and adult obesity [54]. Some of these BMI-related genetic factors are involved in different biological processes in adults and children [57]. In addition, genetic influences on childhood BMI vary even across different growth stages in early childhood, particularly during critical periods such as the infant adiposity peak and childhood adiposity rebound [58]. In later childhood the genetic factors associated with adult BMI begin to exert an increasing influence [54, 55]. These findings highlight the complexity of polygenic obesity and its dynamic nature across developmental stages.
Lifestyle choices can either increase or decrease the contribution of genetic factors to the overall risk of obesity [59]. Indeed, the genetic susceptibility to childhood obesity has been shown to increase with lower parental education levels and decrease with healthier behaviors, such as higher fiber intake and reduced screen time [60]. To gain further understanding, gene-environment interaction (GxE) analyses [59]—similar to those used in adult BMI studies—may provide a more comprehensive understanding of the genetic factors underlying childhood obesity risk.
2.4. Epigenetic Factors
Epigenetic modification, which is a chemical modification of DNA and histones without altering the DNA sequence, is influenced by genetic and environmental factors and affects gene expression [61]. DNA methylation, one of the most widely studied forms of epigenetic modification, typically involves the addition or removal of a methyl group at the 5' position of cytosine residues, primarily within CpG dinucleotides. High-throughput technologies like the Illumina Infinium Methylation450K and MethylationEPIC arrays [62] have standardized DNA methylation analysis, enabling cost-effective studies in large-scale birth cohorts [63]. These platforms also support meta-analyses in international consortia, providing insights into the role of epigenetic modifications from prenatal to early postnatal development in long-term health outcomes.
Recent studies suggest that the perinatal environment plays a crucial role through the epigenome in shaping obesity risk. Maternal nutritional factors during pregnancy, including undernutrition, overnutrition, and nutrient imbalance, can induce epigenetic changes including DNA methylation in the developing fetus [64, 65]. Animal studies further demonstrated that maternal diets can epigenetically reprogram key genes regulating energy metabolism, adipogenesis, and central appetite control, thereby influencing offspring susceptibility to obesity [64, 65]. This phenomenon aligns with the framework of Developmental Origins of Health and Disease (DOHaD) theory, which posits that exposures during critical periods, such as fetal development, have long-lasting effects on health outcomes, including obesity [66].
Research in animal models provides strong evidence for transgenerational epigenetic inheritance, where environmental exposures, such as diet or stress, induce epigenetic changes in germ cells that are passed on to subsequent generations. Rodent models have demonstrated that various stressors, such as paternal diet, maternal nurturing behaviors, prenatal exposure to endocrine-disrupting chemicals, and trauma, can induce heritable changes across generations [67-70]. These include altered stress responses, decreased fertility, and behavioral or metabolic abnormalities, mediated by modifications in DNA methylation and non-coding RNAs [67-70]. While these findings highlight the potential mechanisms, the extent to which similar processes occur in humans remains less clear. Human studies, such as the Överkalix study [71], the ALSPAC study [72], and the Dutch Hunger Winter cohort [73], suggest that grandmaternal exposures—such as nutritional status or smoking—can influence obesity-related traits in grandchildren, highlighting the role of transgenerational effects in non-genetic inheritance. However, several confounding factors, such as shared social environments and genetic influences, complicate the interpretation of these findings, highlighting the need for further human studies to clarify the mechanisms of transgenerational epigenetic inheritance and its role in obesity risk [74].
To comprehensively examine DNA methylation and its influence on childhood obesity, epigenome-wide association studies (EWAS) have been conducted. Unlike the broader DOHaD framework, which focuses on the impact of early-life exposures, EWAS focuse on identifying specific DNA methylation patterns associated with obesity traits. These epigenetic patterns not only provide insights into obesity-related phenotypes and metabolic traits but also show potential as surrogate markers for environmental factors influencing obesity risk. Previous EWAS have identified several candidate DNA methylation sites and genes associated with childhood obesity, as cataloged in resources like the EWAS Atlas [75]. This repository highlights numerous CpG sites and associated genes potentially linked to obesity-related traits, further expanding the scope of known epigenetic markers [76]. However, Vehmeijer et al. suggested that these previously reported DNA methylations may simply reflect obesity rather than act as causal factors. Therefore, there is still little evidence supporting a causal role for epigenetic factors in childhood obesity [77]. The small sample size and the diversity of probes on methylation microarrays may have been a limitation in these previous EWAS for childhood obesity [78]. There appears to be a need for technological innovations to reduce the cost of the high-throughput method, such as sequencing-based methylation analysis.
2.5. Local Situation and Risk Factors in Japan
The health risks associated with obesity vary significantly across populations due to differences in ethnicity and geographic location [79]. Unlike their Caucasian counterparts, Asian children face greater health risks in even mild obesity, largely due to a higher propensity for visceral fat accumulation [80]. Although data on obesity-related health risks for Asians are limited [79], as shown below, research on Japanese populations could be helpful in understanding the interaction between genetic and environmental factors in obesity, such as a study on obesity in Western populations of Asian descent.
In Japan, childhood obesity is primarily diagnosed using the Percentage of Overweight (POW), a modified weight-for height index that reflects time-dependent changes during a child’s growth [24]. For children aged 6 to 17, a POW between –20% and 20% is the normal range, ≥20% indicates mild obesity, ≥30% moderate obesity, and ≥50% severe obesity [24]. For younger children, the normal range is ±15% POW to account for their rapid and variable growth patterns. A POW of 20% (120% of the standard weight) is roughly equivalent to the 90th BMI percentile in children of average height and weight [81]. Besides POW, clinically available BMI percentile curves based on national survey data (1978–1981, updated in 2000) serve as standardized references for Japanese children [82, 83]. Fig. 2A illustrates trends in obesity prevalence among children aged 6 to 17 years over time. Between 1977 and 2000, the prevalence of overweight children increased, then remained stable or slightly decreased until 2021. However, since the onset of the COVID-19 pandemic in 2020, obesity rates have begun to rise again, continuing the upward trend observed since 2006 when the current calculation method was introduced (Fig. 2A). Furthermore, Fig. 2B suggests that the critical periods for developing obesity in Japanese children occur during early childhood, highlighting the importance of focusing interventions on preschool-aged children [84].
As shown in Fig. 3, the prevalence of childhood obesity in Japan exhibits a regional pattern similar to that seen in adults [85]. The incidence of childhood obesity in Japan tends to be high in the Tohoku region and Hokkaido (northeast Japan), Okinawa (southernmost Japan), as well as in Kyushu and parts of Shikoku (southern Japan) (Fig. 3). Evidence suggests a positive correlation between high salt consumption and BMI; the traditionally high salt intake in Tohoku—originally adopted to withstand harsh winters—could potentially contribute to higher BMI in this region [86, 87]. Other childhood obesity risk factors in Japan include behaviors such as breakfast skipping [88, 89], decreased physical activity related to screen time [5], and environmental factors like regional walkability [90]. Income biases due to SES are also important determinants of obesity risk [91]. These risk factors may be influenced by diverse lifestyles and other environmental factors, similar to those observed in Western countries.
3. Risk Prediction
The significant studies are summarized in Table 1, which outlines key milestones and methodologies in the timeline of obesity risk prediction research.
Table 1 Major advances in risk prediction models for childhood obesity
| Year |
Researcher |
Category |
Features & Methods |
Key Outcomes & Limitations |
| 2007 |
Frayling et al. |
Genetic |
Early GWAS study identifying the association between the FTO gene and BMI |
Highlighted the importance of genetic factors, laying the foundation for subsequent genetic studies. |
| 2009 |
Zhang et al. |
ML |
Regression model, ML (decision trees, Naive Bayes, etc.) |
Improved prediction accuracy using limited variables like sex, birth weight, and BMI. |
| 2012 |
Morandi et al. |
Integrated |
Multifactor model (39 SNPs + environmental factors) |
Adding genetic factors resulted in only marginal improvements in prediction accuracy. |
| 2015 |
Dugan et al. |
ML |
ML model using more than 80 variables |
Achieved 89% sensitivity and 85% accuracy, but diversity of data remains a challenge. |
| 2015 |
Shah et al. |
Integrated |
Integration of MRS (Methylation Risk Scores) + PGS |
Demonstrated MRS outperformed PGS in predicting adult BMI. |
| 2018 |
McCartney et al. |
Integrated |
Combined PGS and MRS model |
Improved predictive performance for adult BMI and emphasized the potential of integrating environmental factors. |
| 2019 |
Khera et al. |
Genetic |
PGS (2.1 million SNPs) + Bayesian approach |
Demonstrated potential applicability of adult BMI PGS for predicting childhood obesity and identifying high-risk groups. |
| 2020 |
Hammond et al. |
ML |
LASSO regression + gradient boosting |
Leveraged electronic health records to enhance robustness and managed variable redundancy effectively. |
| 2021 |
Hüls et al. |
Genetic |
Validation of Khera et al.’s PGS in pediatric cohorts |
Explained 11% of BMI variance and 9% of waist circumference variance in European children by adult BMI PGS. |
| 2023 |
Mondal et al. |
ML |
Simplified model using Random Forest (birth weight, sex, height, etc.) |
Achieved 89% accuracy with a minimal set of variables. Limitations: Lack of external validation and limited consideration of genetic and environmental diversity. |
| Planned |
Our Project |
Integrated |
Integrated PGS + MRS model |
Aims to develop MRS tailored for pediatric populations, focusing on regional specificity. |
Studies are categorized into genetic studies, machine learning (ML) approaches, and integrated models. Genetic studies focused on identifying obesity-related genetic factors, including polygenic scores (PGS), which predict risk based on genetic predisposition. ML employed advanced computational techniques, such as random forest and gradient boosting, to enhance prediction accuracy by analyzing environmental factors and clinical data to identify complex patterns in obesity risk. Integrated models combined multiple approaches, including PGS, methylation risk scores (MRS), and environmental factors, to improve comprehensiveness and precision in risk prediction.
Various methods have been proposed to predict the risk of childhood obesity. One simple and conventional approach is prediction based on logistic or linear regression model without incorporating genetic factors [92-97]. Although several models have been proposed for predicting childhood obesity using variable environmental factors such as smoking exposure, SES, and breastfeeding, these have shown little improvement over simple prediction based on BMI [94, 95]. In contrast, a model using longitudinal growth data from infancy or BMI information from later childhood showed significantly improved predictions [96, 97]. Additionally, an approach incorporating parental information and birth weight in the models was also effective in improving prediction accuracy [92-95, 97]. However, logistic and linear regression models are limited in their ability to account for complex interactions and non-linear patterns, which can reduce their predictive accuracy [98].
3.2. Machine Learning Approaches with Environmental Factors
Early research by Zhang et al. highlighted the potential of machine learning in risk prediction, showing that their models outperformed conventional logistic regression in accuracy for predicting childhood obesity [99]. These models were designed to predict childhood obesity at age 3 using a limited set of variables, including sex, standard deviation score (SDS) for birth weight, SDS length, SDS BMI at various stages (birth, six weeks, eight months, or two years), and weight gain. Multiple machine learning algorithms, including decision trees, Naïve Bayes, and support vector machines, were employed in the development process [99]. Similarly, Dugan et al. modeled childhood obesity up to 10 years of age using >80 variables, including demographics, length, weight, BMI at various ages, and questionnaire data collected by 2 years of age [100]. Their model achieved a sensitivity of 89% and an accuracy of 85%, demonstrating notable predictive performance.
Hammond et al. employed more advanced machine learning models, such as LASSO regression and gradient boosting, leveraging extensive data from existing electronic health records and census data [101]. By incorporating redundancy in the variables, their model demonstrated robustness against incomplete data, resulting in a highly resilient predictive model.
Recently, efforts have also been made to simplify models for practical clinical application by reducing the number of variables. For example, Mondal et al. developed a model using a minimal set of variables—birth weight and height, sex, gestational age, and data from a single well-child visit—achieving 89% accuracy in predicting obesity in children under age 5 [102]. This model employed data augmentation techniques to virtually supplement missing data and generate synthetic subjects, optimizing the training dataset and ensuring high performance even with limited health data or imbalanced datasets [102].
The interpretability of machine learning models is a critical issue, particularly for clinicians who need transparent and understandable models to make informed decisions in clinical settings [98]. While machine learning models were once criticized for their “black-box” nature, recent advancements in explainable AI (XAI) technology, such as Shapley Additive Values and Local Interpretable Model-agnostic Explanations, have significantly improved their transparency [103]. The XAI tools may allow researchers and clinicians to visualize the impact of individual variables on outcomes, and their application has advanced particularly in the research fields of early disease detection and personalized medicine [104, 105]. Leveraging these advancements in machine learning, improved interpretability has the potential to facilitate development of more effective and widely applicable models for predicting childhood obesity.
3.3. Polygenic Scores (PGS)
Progress in GWAS has led to accelerated development of risk prediction models using genetic factors that affect obesity. A landmark discovery was made by Frayling et al. (2007), who identified an association between the FTO gene and BMI [106], sparking a wave of genetic studies focused on markers of childhood obesity [107]. Subsequent studies led to the development of predictive models incorporating a limited number of obesity-related SNPs in simple weighted models [108, 109].
Morandi et al. constructed a prediction equation using Northern Finland Birth Cohort data, combining traditional risk factors such as parental BMI, birth weight, and maternal gestational weight gain with 39 obesity-related polymorphisms. However, the inclusion of genetic factors resulted in only marginal improvements in predictive accuracy [108].
A breakthrough came from Khera et al. (2019), who demonstrated that PGS developed using large-scale GWAS data from European adults could also be applied to children [53]. Compared to previous models that used fewer than 100 SNPs, Khera et al. used approximately 2.1 million SNPs and applied a Bayesian approach (LDPred) [110] to generate a more sophisticated PGS [53]. Although their study primarily focused on adults, they observed that individuals in the highest PGS deciles exhibited weight divergence starting from early childhood. Moreover, their study highlighted an important comparison between polygenic and monogenic obesity: individuals in the top 1.6% of PGS had a BMI increase of 4.1 kg/m2, comparable to those with monogenic obesity due to rare MC4R variants [53]. Despite the similar effect size, individuals with high PGS were approximately ten times more prevalent (1.6%) than carriers of rare MC4R variants (0.14%), understanding the significance of PGS in the early identification of pediatric obesity risk [53].
Hüls et al. (2021) later validated the approach in a European pediatric cohort, where it explained 11% of BMI variance and 9% of waist circumference variance [60]. However, as discussed below, the accuracy of the major models may be reduced depending on the population, as most existing PGS models have been constructed using data primarily from individuals of European ancestry [53].
3.4. Integrating Epigenetic Data with PGS in Obesity Risk Prediction
Methylation Risk Scores (MRS), an extension of PGS to CpG methylation data, have been recognized as a promising complement tool to PGS, addressing its limitations in capturing environmental and prenatal factors [111]. MRS derived from blood DNA in adults predicted BMI, prenatal maternal smoking, and smoking status independently of PGS and outperformed PGS in predicting these traits [112]. Furthermore, the use of a comprehensive model combining both PGS and MRS was shown to improve the predictive performance for adult BMI [113, 114]. However, it has also been found that methylation profiles developed based on adult data are less effective at predicting BMI during adolescence. This suggests that environmental influences may intensify with age or that different epigenetic factors may be involved in predicting BMI in children and adolescents. The combination of PGS and MRS may be a promising approach, but large-scale EWAS studies in pediatric obesity would be needed to construct MRS for pediatric BMI. To address this, we are conducting DNA methylation profiling analyses using cord blood DNA, collected as part of the Tohoku Medical Megabank Project’s Three-Generation Cohort Study, to develop MRS tailored for pediatric BMI and to elucidate the epigenetic mechanisms underlying early-life obesity risk.
4. Perspectives and Remaining Problems
The development of personalized obesity prevention strategies using advanced risk prediction models to identify higher-risk individuals offers a promising approach to addressing the childhood obesity epidemic. In preventive medicine, Geoffrey Rose’s strategy distinguishes between two key approaches to public health interventions: the population-based approach and the high-risk approach [115]. While population-based interventions can improve overall public health, the high-risk approach, which targets individuals most susceptible to obesity, offers a more tailored solution to address the specific needs of at-risk groups. A 2024 Cochrane Review concluded that population-based interventions for childhood obesity had very limited effects; dietary-only interventions were largely ineffective, and combined diet and physical activity interventions showed only marginal benefits in younger children and none in adolescents [116, 117]. Many of these interventions were school-based education programs, highlighting the challenge of achieving lasting behavioral change through education alone. Consequently, there has been a shift toward reevaluating the high-risk approach, which aims to more effectively reduce risk factors in these individuals and address the societal and systemic factors that contribute to high-risk groups.
The “first 1,000 days,” the period from conception to age two, is increasingly recognized as a critical window for long-term health outcomes, as this is when adipose tissue development and metabolic programming occur [118]. Evidence from systematic reviews highlights that interventions during this period, such as improving maternal nutrition and addressing excessive gestational weight gain, maternal smoking, and stress during pregnancy, may reduce the risk of obesity later in life [119]. Existing evidence on obesity prevention interventions has largely focused on individual-level behavior changes, while social and systemic approaches remain underdeveloped [120]. However, population-wide interventions, such as policy changes and private sector collaborations, pose challenges for traditional research methodologies, as they are not easily subjected to randomization and blinding. Future research must establish rigorous evaluation frameworks to assess the effectiveness of such interventions [120].
Personalized obesity prevention strategies hold significant promise; however, many established models still lack sufficient external validation, raising concerns about their effectiveness across diverse ethnic, socioeconomic, and geographic contexts [121]. This is particularly concerning in regions like Asia, where obesity rates are rising rapidly but research and resources remain limited [122]. To address these gaps, research platforms like the Japan Birth Cohort Consortium (JBiCC) and iMETHYL provide region-specific data essential for Asian populations [123, 124]. The JBiCC, a network of birth cohorts across Japan, was established to standardize evaluation metrics and conduct nationwide meta-analyses on key social issues, enabling genetic analysis and tracking of environmental and lifestyle risk factors across life stages—from birth through adolescence to adulthood—while reflecting Japan’s regional and socioeconomic diversity [123]. Similarly, iMETHYL’s DNA methylation references support the identification of epigenetic markers specific to Japanese populations, further strengthening early obesity prevention research in Asian ethnic groups.
Another problem is the lack of a standardized definition for childhood obesity, which makes it difficult to identify at-risk children and compare studies on a global scale. Childhood obesity is commonly defined based on BMI-for-age percentiles (BMI%), as established by the WHO [125], the International Obesity Taskforce (IOTF) [126], and the US Centers for Disease Control and Prevention (CDC) [127]. In contrast, some Asian countries, including Japan [24] and India [128], utilize Z scores of BMI to establish their own standards. Especially for those under two years old, the thresholds for diagnosing obesity in young children remain unclear, with the WHO Child Growth Standards being one of the few references available [126, 127, 129]. A universally accepted definition of childhood obesity is crucial for developing global prevention strategies and ensuring that research findings are comparable across different settings.
Finally, the ethical and technical challenges associated with interventions following risk prediction should be addressed. Group interventions in schools and other settings where obese children are easily identifiable can raise ethical concerns, as they may unintentionally exacerbate obesity-related stigma, including societal biases that perceive obese children as lacking self-control, being lazy, or personally responsible for their condition [13, 23]. Such stigma can lead to exclusion, reduced self-esteem, and social isolation among obese children [13]. In addition, interventions that target personal lifestyle choices, such as diet and physical activity, also raise ethical concerns regarding individual autonomy [130]. There is a delicate balance between promoting public health and respecting the right of individuals and families to make decisions about their own lives.
https://orcid.org/0000-0001-7902-925X
https://ror.org/02f57qw79 
