Front. Public HealthFrontiers in Public HealthFront. Public Health2296-2565Frontiers Media S.A.10.3389/fpubh.2026.1773424Original ResearchIdentification of preschool children's physical fitness phenotypes and their association with multidimensional body shape indicators: a data-driven approach with risk prediction modelingChenXiaoxiao1†ConceptualizationFormal analysisMethodologyWriting – original draftWriting – review & editingZhaoDeqiang2†ConceptualizationFormal analysisMethodologyWriting – original draftWriting – review & editingSoftwareZhangAoyu3MethodologyWriting – original draftPanXiang4ConceptualizationFormal analysisWriting – original draftWangChuanmiao1ConceptualizationFormal analysisWriting – original draftChenJiaxin1ConceptualizationFormal analysisWriting – original draftZhangYanfeng2*InvestigationProject administrationResourcesWriting – original draftSchool of Physical Education and Sports Rehabilitation, Jinzhou Medical University, Jinzhou, Liaoning, ChinaChina Institute of Sport Science, Beijing, ChinaSchool of Medicine, Tianshi College of Tianjin, Tianjin, ChinaGraduate School of Health and Sports Science, Juntendo University, Inzai, JapanCorrespondence: Yanfeng Zhang, zhangyanfeng0310@126.com
These authors have contributed equally to this work and share first authorship
Traditional physical fitness assessments in young children often rely on single indicators, which fail to capture integrated ability patterns. The common use of body mass index (BMI) to link morphology and fitness has limited discriminant validity. This study aimed to identify naturally emerging physical fitness profiles among preschool children in Macao using a data-driven approach and to develop a screening tool based on multi-dimensional body shape indicators—beyond BMI—for identifying children at risk of low physical fitness.
Methods
The sample comprised 3,180 children aged 3–6 years from the Macao China Physical Fitness Surveillance and Physical Activity Survey (2010, 2015, 2020). K-means clustering was performed on six standardized physical fitness test scores to identify intrinsic fitness profiles. One-way ANOVA was used to compare body shape indicators across clusters. Using the low-fitness cluster as the dependent variable, logistic regression identified key morphological predictors, from which a composite discriminant index was constructed. Predictive performance was evaluated using receiver operating characteristic (ROC) curve analysis.
Results
Cluster analysis revealed three distinct physical fitness profiles: low, moderate, and high fitness. ANOVA showed significant gradient differences across clusters for all body shape indicators (height, weight, circumferences, etc.). Effect sizes (Cohen's f) ranged from 0.025 to 0.190. Logistic regression identified four core predictors: height, weight, waist circumference, and pelvic width. The composite body shape index demonstrated good discriminative ability for low fitness risk, with an area under the curve (AUC) of 0.779 (95% CI: 0.742–0.800). At the optimal cut-off, sensitivity was 77.7%, and specificity was 66.8%. The index showed consistent performance in sex-stratified analyses (Boys: AUC = 0.87; Girls: AUC = 0.88).
Conclusion
This study confirms that preschool children's physical fitness develops in distinct, internally coherent profiles, closely associated with body morphology. The composite index based on waist circumference and pelvic width—among other measures—proves more effective than BMI alone in identifying children at risk of low fitness. It offers a practical tool for rapid, early screening during routine health checks and supports targeted physical activity interventions.
body shape indicatorscluster analysisphysical fitness phenotypespreschool childrenrisk predictionThe author(s) declared that financial support was received for this work and/or its publication. This work was supported by Joint research project between the Sports Bureau of the Government of the Macao Special Administrative Region of China and the Institute of Sports Science of the State General Administration of Sports (b2117); 2025 Macau Citizen Physical Fitness Monitoring Technical Support Service (B2425).section-at-acceptanceChildren and HealthIntroduction
The preschool period (3–6 years) is a critical window for the development of fundamental motor skills and physical growth. Fitness levels during this stage not only influence current health, activity participation, and quality of life but also have long-term implications for physical conditioning, habitual physical activity, cognitive ability, school readiness, and social development (1–3). Establishing a scientific, valid, and forward-looking physical fitness assessment system for young children is therefore a central issue in early health promotion and holistic development, with significant implications for precise intervention and population health.
Current fitness assessments typically rely on isolated performance measures (4–6). While useful for evaluating specific motor skills, such approaches fail to capture the multidimensional, synergistic, and heterogeneous nature of overall physical fitness and cannot systematically address the fundamental question of what typical integrated fitness development patterns exist among children (7). This limits holistic understanding and hinders personalized guidance. In parallel, BMI remains the most widely used morphological indicator internationally due to its simplicity. However, its validity is particularly challenged in children aged 3–6 years, a period of rapid and non-uniform changes in body composition (8). Numerous studies note that BMI cannot distinguish between lean mass and fat mass or reflect fat distribution (e.g., central adiposity) (9–11). Evidence suggests that body composition (muscle-to-fat ratio) and fat distribution may be more closely linked to motor competence and metabolic health than overall adiposity (12, 13). Thus, relying solely on BMI provides a fragmented and limited link between morphology and comprehensive physical ability, offering insufficient basis for differentiated health promotion and fitness intervention (14).
To address these gaps, prior research has often correlated single fitness tests with isolated anthropometrics or focused on long-term growth indices. However, a significant methodological gap exists in objectively identifying holistic fitness phenotypes from multi-test data and subsequently linking these phenotypes to a practical, multidimensional morphological screening tool suitable for immediate risk assessment. This study aims to fill this gap. Specifically, this study aimed to: (1) identify naturally emerging physical fitness phenotypes among preschool children in Macao using data-driven clustering; (2) examine the association between these phenotypes and multidimensional body shape indicators; (3) develop and validate a composite body shape index for screening children at risk of low physical fitness. The application of unsupervised machine learning (cluster analysis) represents a substantive methodological advance by shifting the analytical focus from comparing children on isolated tests to identifying their inherent, integrated ability patterns. This approach is grounded in the recognition that physical fitness is a latent construct best revealed through pattern recognition in multi-dimensional data.
MethodsParticipants
Data were drawn from the Macao China Physical Fitness Surveillance and Physical Activity Survey (2010, 2015, 2020), which used stratified random cluster sampling with one-year age groups (15). The study included 3,180 children aged 3–6 years. Parents or guardians provided written informed consent, and the study was approved by the Ethics Committee of the China Institute of Sport Science (CISS-2019-06-07). While combining data from three waves increases sample size and generalizability within the Macao context, we acknowledge the potential for unmeasured cohort or period effects as a study limitation. Data collection followed standardized national protocols with trained examiners to ensure consistency, although formal inter-rater reliability metrics were not recorded for all waves.
Measures
Height and weight were measured following the China Physical Fitness Standards for Preschool Children (CPFS-preschool) manual, and BMI was calculated. Physical fitness was assessed using six tests: standing long jump, tennis ball throw, 10-m shuttle run, 15-m obstacle run, sit-and-reach, and balance beam walk, administered according to CPFS-preschool guidelines (16). Raw scores were standardized (Z-scores) by age and sex. All raw scores were converted to age- and sex-specific Z-scores based on the entire combined sample to control for developmental differences prior to clustering, aligning with our aim to identify fitness patterns independent of normal age/sex variation.
Data analysisClustering analysis.
To identify latent fitness profiles, K-means clustering was performed on the six standardized fitness Z-scores. The choice of K-means was justified by its efficiency with large samples, its common use for phenotype discovery, and the standardized, continuous nature of our variables. We addressed key assumptions and stability: (1) variables were standardized to equalize scales; (2) the Euclidean distance metric was used; (3) the algorithm was run with 50 random initializations to minimize the impact of initial centroid placement and achieve a stable solution; (4) cluster stability was assessed by calculating the average silhouette width (0.41), indicating fair structure. The optimal number of clusters (k = 3) was determined using the elbow method and scree plot of within-cluster sum of squares. To support the suitability of the data for structure detection, we report the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (0.82) and Bartlett's test of sphericity (p < 0.001). We clarify that these tests indicate the data's suitability for multivariate analysis (like PCA used for visualization) rather than directly validating the cluster solution.
Group comparisons.
One-way ANOVA was used to compare body shape indicators across the three clusters. Assumptions of homogeneity of variance and normality were tested using Levene's test and Shapiro-Wilk test on residuals, respectively. For indicators where Levene's test was significant (p < 0.05), Welch's ANOVA was applied, and this is specified in the results. Effect sizes are reported using Cohen's f. Post-hoc comparisons used Tukey's HSD test.
Predictive modeling.
Logistic regression was performed with the low-fitness cluster as the dependent variable (1 = low, 0 = moderate/high) and all body shape indicators as potential predictors. Model building followed a purposeful selection approach. Key demographic covariates (age, sex, survey year) were examined. Age and sex were not included as their effects were already statistically controlled for via Z-score standardization of the fitness outcomes used to define the dependent variable. Including them again in the prediction model could lead to over-adjustment. Survey year was tested as a covariate but was not a significant predictor (p = 0.43) and did not improve model fit (Likelihood Ratio Test p = 0.41), thus it was excluded for parsimony. Multicollinearity was assessed using variance inflation factors (VIF < 5). A composite Body Shape Index was derived from the final model's coefficients.
Validation and comparison
The index's discriminative performance was evaluated using ROC curve analysis. The area under the curve (AUC) with 95% confidence interval (derived from 1,000 bootstrap samples) was reported. Sensitivity and specificity at the optimal cut-off (Youden's index) were calculated. Internal validity was assessed via 10-fold cross-validation, reporting the mean cross-validated AUC. To formally test the claimed superiority over BMI, a separate logistic regression model using only BMI as a predictor was fitted, and its ROC curve was compared to the composite index's ROC curve using DeLong's test for two correlated ROC curves. Additionally, analyses were stratified by sex to explore potential differences, with ROC curves reported for boys and girls separately. All analyses used α = 0.05 and were performed in R (version 4.3.1).
ResultsSample characteristics
The final sample included 3,180 children (1,903 boys, 1,277 girls) across three survey years (Table 1).
Sample distribution by year, age, and sex.
Year
Age
Boys
Girls
Total
3
136
72
208
2010
4
173
108
281
5
175
98
273
6
95
73
168
3
142
91
233
4
170
119
289
2015
5
205
139
344
6
89
82
171
3
209
132
341
2020
4
189
125
314
5
222
160
382
6
98
78
176
Total
1,903
1,277
3,180
Identification of physical fitness phenotypes
K-means clustering on the six fitness Z-scores yielded a stable 3-cluster solution (Figures 1, 2). The clusters were characterized as: Cluster 1 (Low Fitness, n = 412), Cluster 2 (Moderate Fitness, n = 1,200), and Cluster 3 (High Fitness, n = 1,568) (Table 2). The clustering solution explained 39.3% of the total variance (R2 = 0.393). Principal Component Analysis (PCA) for visualization showed clear separation of the three clusters in the space of the first two principal components (cumulative variance explained = 61.1%) (Figures 3, 4).
Elbow method for optimal clusters.
Line graph titled “Elbow Method for Optimal Clusters” plots number of clusters k on the x-axis and total within sum of squares on the y-axis, showing a sharp decrease until k equals three, where a dashed vertical line indicates the optimal cluster count.
Scree plot.
Bar chart titled Scree Plot displaying percentage of explained variances for six dimensions, with values decreasing from 44.3 percent for dimension one to 7.6 percent for dimension six, illustrating variance explained by principal components.
Physical fitness cluster characteristics.
Cluster
n
Standing long jump
Tennis ball throw
Sit and reach
10 m shuttle run
Balance beam
Continuous jump
1
412
−1.24 ± 0.62
−1.00 ± 0.54
0.13 ± 0.87
−1.58 ± 1.12
−2.65 ± 1.25
−1.60 ± 1.47
2
1,200
−1.04 ± 0.58
−0.84 ± 0.51
0.28 ± 0.89
−0.99 ± 0.77
−0.22 ± 0.59
−0.83 ± 0.96
3
1,568
0.00 ± 0.64
−0.09 ± 0.72
0.02 ± 0.94
0.13 ± 0.52
0.15 ± 0.61
0.25 ± 0.51
PCA biplot.
Scatter plot showing clustered results of a principal component analysis of a physical test with three groups: green circles for cluster one on the left, purple triangles for cluster two in the center, and orange squares for cluster three on the right. Horizontal axis is labeled Dim1 with forty-four point three percent variance, and vertical axis is Dim2 with sixteen point eight percent variance. Black dashed lines cross at the origin.
PCA loading.
Principal component analysis (PCA) biplot shows vectors representing six physical test variables: tennis ball throw, shuttle run, standing long jump, continuous jump, balance beam, and sit and reach. Axes Dim1 and Dim2 explain 44.3 percent and 18.8 percent of variance, respectively. Color gradient indicates contribution strength of each variable.Association between fitness phenotypes and body shape
All nine body shape indicators differed significantly across the three clusters (all p < 0.001). Standard ANOVA was used for height, sit height, weight, chest circumference, hip circumference, shoulder breadth, and foot length. Welch's ANOVA was applied for waist circumference and pelvic width due to significant heteroscedasticity (Levene's test p < 0.05). As shown in Table 3, a clear gradient was observed: the High Fitness cluster had the most developed body shape metrics, followed by the Moderate and Low Fitness clusters. Effect sizes (Cohen's f) ranged from small (0.025 for waist circumference) to medium-large (0.190 for height).
Comparison of body shape indicators across physical fitness clusters.
Indicator
Cluster
Mean ±SD
F
Cohen
p
Height (cm)
1
98.10 ± 4.42
371.971
0.190
< 0.001
2
100.25 ± 4.76
3
105.24 ± 5.01
Sit height (cm)
1
56.30 ± 2.50
237.131
0.130
< 0.001
2
57.37 ± 2.63
3
59.61 ± 2.93
Weight (kg)
1
14.92 ± 2.07
155.278
0.125
< 0.001
2
15.59 ± 2.16
3
17.16 ± 2.55
Chest circumference (cm)
1
51.46 ± 2.86
63.433
0.089
< 0.001
2
51.85 ± 2.76
3
53.15 ± 3.06
Waist (cm)
1
48.40 ± 3.56
38.543
0.025
< 0.001
2
48.40 ± 3.56
3
49.84 ± 4.02
Hip circumference (cm)
1
53.28 ± 3.92
87.505
0.052
< 0.001
2
53.81 ± 3.87
3
55.96 ± 4.21
Shoulder breadth (cm)
1
21.94 ± 1.41
121.192
0.071
< 0.001
2
22.30 ± 1.40
3
23.12 ± 1.41
Width of pelvic (cm)
1
16.38 ± 1.27
60.978
0.132
< 0.001
2
16.40 ± 1.23
3
16.96 ± 1.15
Foot length (cm)
1
15.53 ± 1.23
152.711
0.088
< 0.001
2
15.77 ± 1.01
3
16.50 ± 1.05
(Table content as per original, with added 'Test' column specifying ANOVA/Welch and 'Cohen's f' column).
Development of the composite body shape index
Multivariate logistic regression (low fitness cluster as reference) identified four significant predictors: height (β = −1.156, p < 0.001), weight (β = −0.412, p = 0.028), waist circumference (β = 0.493, p < 0.001), and pelvic width (β = 0.225, p = 0.004). The derived body shape index was:
BSI = −2.376 – 1.156 × Height (cm) – 0.412 × Weight (kg) + 0.493 × Waist Circumference (cm) + 0.225 × Pelvic Width (cm). Higher values of waist circumference and pelvic width increased the odds of low fitness, while greater height and weight decreased it. The model showed no significant multicollinearity (all VIFs < 2.5) and adequate fit (Hosmer-Lemeshow test p = 0.553) (Table 4).
Overall summary of the logistic regression model for predicting poor physical fitness risk.
Predictor
β co-efficient (SE)
Odds ratio (OR)
95% CI
p-value
Height (cm)
−1.156 (0.215)
0.315
0.207–0.480
< 0.001
Weight (kg)
−0.412 (0.187)
0.662
0.459–0.955
0.028
Waist circumference (cm)
0.493 (0.098)
1.637
1.351–1.983
< 0.001
Pelvic width (cm)
0.225 (0.078)
1.252
1.075–1.459
0.004
Hosmer-Lemeshow: χ2 = 6.85, p = 0.553, R2 = 0.065
Predictive performance and validation
The ROC curve for the BSI is shown in Figure 5. The apparent AUC was 0.779 (95% CI: 0.742–0.800). The 10-fold cross-validated mean AUC was 0.765 (SD = 0.021), indicating robust internal validity. At the optimal cut-off score of −2.003, sensitivity was 77.7% and specificity was 66.8%. Further analysis regarding gender differences revealed that the composite body shape index demonstrated good predictive performance in both boys and girls. In the subsample of boys, the area under the curve (AUC) of the BSI for predicting low physical fitness risk was 0.87 (95% CI: 0.84–0.90), while in the subsample of girls, the AUC was 0.88 (95% CI: 0.85–0.91). These results suggest that within the current preschool-aged population of 3–6 years, the body shape index developed in this study shows consistent applicability and strong predictive ability across sexes. Nevertheless, if sex-specific models were to be constructed separately, the regression coefficients might differ, which presents a potential direction for refinement in future research. The comparative analysis and sex-stratified results are visually presented in Figure 6 through ROC curves.
Receiver operating characteristic (ROC) curve for the composite body shape index in predicting the low physical fitness phenotype.
ROC curve for a body shape model with sensitivity on the y-axis and specificity on the x-axis, and the model’s AUC value is 0.779, indicating moderate discrimination ability.
Receiver operating characteristic (ROC) curves of the composite Body Shape Index for predicting low physical fitness risk, stratified by sex. The solid line represents the ROC curve for boys (AUC = 0.87), and the dashed line represents the ROC curve for girls (AUC = 0.88). The diagonal gray line indicates the reference line of no discrimination (AUC = 0.50). The close and high AUC values in both sexes demonstrate the strong and consistent predictive ability of the BSI across genders in this population.
ROC curve comparing model performance for body shape classification by sex, with lines for boys and girls. Area under the curve values are 0.87 for boys and 0.88 for girls. Sensitivity is on the y-axis and specificity on the x-axis.Comparison with BMI and sex-stratified analysis
To validate the superiority of the constructed Body Shape Index (BSI), we conducted a direct comparison with the traditional BMI. Separate logistic regression models, using only age- and sex-standardized BMI as the predictor, were fitted for boys, girls, and the combined sample to predict membership in the low physical fitness phenotype. The discriminative performance of BMI was limited in all groups. The area under the curve (AUC) was 0.52 (95% CI: 0.47–0.57) for boys, 0.51 (95% CI: 0.46–0.56) for girls, and 0.54 (95% CI: 0.50–0.58) for the total sample (Figure 7). These results indicate that BMI alone has only marginal discriminative ability, with AUC values not meaningfully exceeding the chance level of 0.5, in identifying children at risk of low physical fitness in this study population.
Receiver operating characteristic (ROC) curves of BMI for predicting low physical fitness phenotype in preschool children. Results are shown for the total sample and stratified by sex. The AUC (95% Confidence Interval) values were 0.54 for the total sample, 0.52 for boys, and 0.51 (0.46–0.56) for girls. The diagonal gray line indicates the reference line of no discrimination (AUC = 0.5).
Line chart showing ROC curves for BMI classification by gender, with separate lines for boys (blue, area under curve 0.52), girls (green, 0.51), and all participants (red, 0.54), indicating poor predictive performance.
To formally test the difference in diagnostic performance between the composite BSI and the BMI-only model, we performed DeLong's test for two correlated ROC curves using the total sample. The results are summarized in Table 5. The BSI demonstrated a significantly larger AUC (0.779) compared to the BMI-only model for the total sample (0.54), with a difference of 0.239 (Z = 10.15, p < 0.001). This provides statistical confirmation of the superior discriminant validity of the proposed body shape index over the traditional BMI metric.
Comparison of predictive performance between BSI and BMI using DeLong‘s Test.
Model
AUC (95% CI)
Sensitivity
Specificity
DeLong‘s test (vs. BSI)
Body shape index (BSI)
0.779 (0.742–0.800)
77.7%
66.8%
Reference
BMI-only (Total sample)
0.54 (0.50–0.58)
61.5%
58.3%
Z = 10.15, p < 0.001
AUC, Area Under the Curve; CI, Confidence Interval. Sensitivity and specificity are reported at the optimal cut-off point (Youden's index) for each model. The p-value indicates the significance of the difference in AUC between the two models.
Discussion
This study achieved its 3-fold aim. First, using a data-driven clustering approach with enhanced methodological reporting, we identified three distinct, naturally emerging physical fitness phenotypes among preschool children. Second, we established strong gradient associations between these phenotypes and a range of body shape indicators. Third, and most innovatively, we developed and validated a composite Body Shape Index (BSI) that demonstrates good performance in screening for low fitness risk, with statistical evidence supporting its discriminant advantage over the traditional BMI metric in our sample.
The core methodological contribution lies in the sequential application of unsupervised learning (for phenotype discovery) and supervised learning (for index development), bridging the gap between holistic pattern identification and practical screening tool creation. The gradient in body shape across fitness clusters underscores the integrated nature of physical development. The logistic regression model provides a quantitative link, revealing that after accounting for overall body size (height and weight), specific circumferential, and skeletal breadth measures (waist circumference and pelvic width) are independently associated with fitness risk (17). This suggests that body proportions and fat distribution may be more sensitive morphological correlates of integrated fitness in young children than body size alone (18, 19), which aligns with and extends critiques of BMI, as it does not differentiate these dimensions (20, 21).
Compared to prior work focusing on single test correlations or long-term growth indices, our study shifts the application scenario toward immediate, cross-sectional risk screening (16, 22, 23). The BSI's discriminative ability (AUC = 0.779) and its statistically significant superiority over BMI (AUC = 0.54; p < 0.001 per DeLong's test) highlight its potential as a more informative morphological indicator in this context. Furthermore, the sex-stratified analysis revealed that the BSI performed consistently well for both boys (AUC = 0.87) and girls (AUC = 0.88), supporting its internal generalizability within the preschool age range of our study population. Taken together, these findings suggest that the BSI, based on simple anthropometric measures, could serve as a promising preliminary screening tool to identify children who may benefit from more comprehensive fitness assessments.
However, our interpretations are tempered by the study's design and methodological choices. We acknowledge that the cross-sectional data preclude causal inferences between morphology and fitness phenotypes. Furthermore, the decision not to include age, sex, and survey year as covariates in the final predictive model was based on statistical and conceptual grounds (i.e., prior control via Z-score standardization for age/sex, and non-significance of survey year), which we argue prevents over-adjustment. Nonetheless, this choice, along with the potential for unmeasured cohort effects from combining survey waves, should be considered when interpreting the model's coefficients and generalizing findings. Critically, the external validity of the BSI requires confirmation in independent and diverse populations, and its predictive utility over time warrants longitudinal investigation before it can be recommended for clinical or public health practice. The current findings primarily offer a methodological advancement and a proof-of-concept for an integrated, morphology-based screening approach.
Conclusion
This study identified three distinct physical fitness phenotypes among preschool children using a data-driven clustering approach. The derived composite Body Shape Index (BSI), integrating height, weight, waist circumference, and pelvic width, demonstrated good and consistent discriminative ability for identifying children at risk of low fitness (AUC = 0.779), significantly outperforming BMI alone. While promising as a practical screening tool for rapid risk stratification during routine health checks, its cross-sectional design and single-population origin necessitate future longitudinal and external validation. The methodological integration of unsupervised and supervised learning offers a novel framework for translating holistic fitness patterns into actionable screening indices.
Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Ethics statement
The studies involving humans were approved by China Institute of Sport Science. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation in this study was provided by the participants' legal guardians/next of kin.
Author contributions
XC: Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. DZ: Conceptualization, Formal analysis, Methodology, Software, Writing – original draft, Writing – review & editing. AZ: Methodology, Writing – original draft. XP: Conceptualization, Formal analysis, Writing – original draft. CW: Conceptualization, Formal analysis, Writing – original draft. JC: Conceptualization, Formal analysis, Writing – original draft. YZ: Investigation, Project administration, Resources, Writing – original draft.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
ReferencesChopraG. Development during the preschool years. In:ChopraG, editor. Systems Approach to Early Childhood Development: Bridging Theory, Policy and Practice. Singapore: Springer Nature Singapore (2025). p. 49–74.TanJSYLimCBS. The development of gross motor skills in children: insights from the motor assessment test for children. J Exp Child Psychol. (2025) 256:106255. doi: 10.1016/j.jecp.2025.10625540220729DappLCGashajVRoebersCM. Physical activity and motor skills in children: a differentiated approach. Psychol Sport Exerc. (2021) 54:101916. doi: 10.1016/j.psychsport.2021.101916CapioCMEguiaKF. Movement skills, perception, and physical activity of young children: a mediation analysis. Pediatr Int. (2021) 63:442–7. doi: 10.1111/ped.1443632786106MiguelesJHSolis-UrraP. Assessment of physical activity in children and adolescents. In:García-HermosoA, editor. Promotion of Physical Activity and Health in the School Setting. Cham: Springer Nature Switzerland (2024). p. 89–105.KeDMaimaitijiangRShenSKishiHKurokawaYSuzukiK. Field-based physical fitness assessment in preschool children: a scoping review. Front Pediatr. (2022) 10:2022. doi: 10.3389/fped.2022.93944235989998ValentiniNC. Motor skill assessment in children and adolescents. In:García-HermosoA, editor. Promotion of Physical Activity and Health in the School Setting. Cham: Springer Nature Switzerland (2024). p. 133–63.BellKAWagnerCLPerngWFeldmanHAShypailoRJBelfortMB. Validity of body mass index as a measure of adiposity in infancy. J Pediatr. (2018) 196:168–74. doi: 10.1016/j.jpeds.2018.01.02829551311RoySMFieldsDAMitchellJAHawkesCPKellyAWuGD. Body mass index is a better indicator of body composition than weight-for-length at age 1 month. J Pediatr. (2019) 204:77–83. doi: 10.1016/j.jpeds.2018.08.00730268397LiuLBanCJiaSChenXHeS. Association of predicted fat mass, predicted lean mass and predicted percent fat with diabetes mellitus in Chinese population: a 15-year prospective cohort. BMJ Open. (2022) 12:e058162. doi: 10.1136/bmjopen-2021-05816235672066SweattKGarveyWTMartinsC. Strengths and limitations of BMI in the diagnosis of obesity: what is the path forward?Curr Obes Rep. (2024) 13:584–95. doi: 10.1007/s13679-024-00580-138958869YuPCHsuCCLeeWJLiangCKChouMYLinMH. Muscle-to-fat ratio identifies functional impairments and cardiometabolic risk and predicts outcomes: biomarkers of sarcopenic obesity. J Cachexia Sarcopenia Muscle. (2022) 13:368–76. doi: 10.1002/jcsm.1287734866342ShaoYWangNShaoMLiuBWangYYangY. The lean body mass to visceral fat mass ratio is negatively associated with cardiometabolic disorders: a cross-sectional study. Sci Rep. (2025) 15:3422. doi: 10.1038/s41598-025-88167-139870802WellsJCKTreleavenPColeTJ. BMI compared with 3-dimensional body shape: the UK national sizing survey. Am J Clin Nutr. (2007) 85:419–25. doi: 10.1093/ajcn/85.2.41917284738ChoiSMDongHLeiSMTomkinsonGRLangJJCadenas-SanchezC. A 15-year decline in physical fitness among children and adolescents from the macao special administrative region (2005–2020). J Phys Act Health. (2025) 28:1–9. doi: 10.1123/jpah.2025-053241151568LiRLiRLiuMZhaoHDengPZhuJ. Field-based physical fitness assessment in preschool children in China. BMC Public Health. (2024) 24:2722. doi: 10.1186/s12889-024-20237-x39375647WangLZouWWangYKohDMunsif Bin Wan PaWAGaoR. The impact of preschool children's physical fitness evaluation under self organizing maps neural network. Sci Rep. (2025) 15:1461. doi: 10.1038/s41598-025-85725-539789314ZhaoRLiXWangJZhangLGaoZ. Evaluation of physical fitness and health of young children aged between 3 and 6 based on cluster and factor analyses. BMC Public Health. (2024) 24:420. doi: 10.1186/s12889-024-17660-538336673Rico-GonzálezMArdigòLPRamírez-ArroyoAPGómez-CarmonaCD. Anthropometric influence on preschool children's physical fitness and motor skills: a systematic review. J Funct Morphol Kinesiol. (2024) 9:95. doi: 10.3390/jfmk902009538921631WebsterEKSurIStevensARobinsonLE. Associations between body composition and fundamental motor skill competency in children. BMC Pediatr. (2021) 21:444. doi: 10.1186/s12887-021-02912-934629074FreyGCChowB. Relationship between BMI, physical fitness, and motor skills in youth with mild intellectual disabilities. Int J Obes. (2006) 30:861–7. doi: 10.1038/sj.ijo.080319616404408ChouYYing HuBWinslerAWuHGreenburgJKongZ. Chinese preschool children's physical fitness, motor competence, executive functioning, and receptive language, math, and science performance in kindergarten. Child Youth Serv Rev. (2022) 136:106397. doi: 10.1016/j.childyouth.2022.106397Cadenas-SanchezCIntemannTLabayenIPeinadoABVidal-ContiJSanchis-MoysiJ. Physical fitness reference standards for preschool children: the PREFIT project. J Sci Med Sport. (2019) 22:430–7. doi: 10.1016/j.jsams.2018.09.22730316738
Edited by: Antonella Muscella, University of Salento, Italy
Reviewed by: Alexander Plakida, Odessa National Medical University, Ukraine
Agnieszka Wasiluk, Józef Piłsudski University of Physical Education in Warsaw, Poland