Body height and mass of children and adolescents in Poland: age-specific centile distributions from the 2024–2025 nationwide survey

Anna Pastuszak

https://orcid.org/0000-0003-3786-4833
Institute of Sport – National Research Institute, Warsaw, Poland

Janusz Dobosz

https://orcid.org/0000-0002-2425-4932
Institute of Sport – National Research Institute, Warsaw, Poland
Józef Piłsudski University of Physical Education in Warsaw, Poland

Dorota Sadowska

https://orcid.org/0000-0002-9823-7991
Institute of Sport – National Research Institute, Warsaw, Poland

Introduction

Data on the growth and development of younger generations constitute a reliable indicator of the health and nutritional status of societies. They can be used to monitor the overall social and economic situation of the populations and to identify at-risk groups for potential interventions (WHO, 1995). Measurements of body height and body mass are among the fundamental tools used to assess physical development and nutritional status at the population level. Studies of children and adolescents aged 5–19 years from more than 190 countries showed that between 2000 and 2025, the pre­valence of obesity increased from 3% to 9.4% and exceeded underweight pre­valence globally except in Sub-Saharan Africa and South Asia (UNICEF, 2025). Findings from nationwide studies in Poland (Jarosz et al., 2016; Mazur, 2024) confirm these global trends. Recent international evidence also indicates unfavourable trends in excessive body mass among children and adolescents worldwide, including in Europe (NCD-RisC, 2024).

Growth and maturation parameters are widely used as indicators of intergenerational changes and of the biological effects of living conditions at the population level. Previous studies have documented long-term changes in body height, body mass, and maturation timing in children and adolescents in Poland (Gomula et al., 2021; Kułaga et al., 2011), as well as in many other European countries (Eiholzer et al., 2025; Júlíusson et al., 2013; Pop et al., 2021; Roelants et al., 2009; Saari et al., 2011) and globally, particularly in industrialized regions (Cole & Mori, 2018; NCD-RisC, 2020). More recent reports, however, suggest that the pace of these changes may differ from that observed in earlier decades (Kryst et al., 2022; Pastor-Fajardo et al., 2020; NCD-RisC, 2020). These observations support the need for updated empirical data describing the current somatic characteristics of younger populations.

The health status and physical development of children and adolescents are shaped by multiple biological, environmental, and social factors, which should be considered when interpreting population-level somatic data (Bielicki, 1986; Charzewski et al., 2003; Kozieł, Nowak-Szczepańska, & Gomuła, 2014). For this reason, nationwide empirical material collected according to a standardized protocol is of particular importance for developing up-to-date centile-based descriptions of basic somatic traits.

In many countries, measurements of basic somatic traits serve as growth references used to support the assessment of growth and developmental status in children and adolescents. These cross-sectional studies are typically updated every 10–15 years, as illustrated by research conducted in Finland (Saari et al., 2011), Norway (Júlíusson et al., 2013), Switzerland (Eiholzer et al., 2025), Romania (Pascanu et al., 2016; Pop et al., 2021), China (Li et al., 2009), and Poland (Kułaga et al., 2011; Palczewska & Niedźwiedzka, 2001). In more than 140 countries lacking up-to-date data for assessing growth in school-aged children and adolescents, WHO growth references are used (de Onis et al., 2007). In Europe, WHO growth references have been adopted in several countries, including Portugal, Albania, and Moldova. In Poland, the most recent nationwide reference studies based on a random sample of children and adolescents were conducted between 2007 and 2012 within the OLAF and OLA projects, and their findings were published as the Polish 2010 and 2012 reference systems for school-aged and preschool children, respectively (Kułaga et al., 2011; 2013; 2015).

The data analysed in the present paper were collected as part of the project “Centile Charts for Body Height, Body Weight, and Body Mass Index in Children and Adolescents in Poland”, funded by the Ministry of Science and Higher Education under the Science for Society II programme. The project was designed to obtain a nationwide empirical dataset for the development of up-to-date centile distributions of basic somatic traits in school-aged children and adolescents. Field examinations were conducted in 2024–2025 in schools across all regions of Poland, using a standardized student selection procedure based on the assumptions previously applied in the OLAF project, along with a standardized anthropometric measurement protocol.

Against the background of the broader objectives of the project, the present paper focuses on a specific scope of data analysis and result presentation. The aim of this paper is to present the empirical material underlying the centile estimation of body height and body mass in Polish children and adolescents, including the relationship between the age and territorial structure of the study sample and that of the nationwide population, as well as to describe the method used to develop centile distributions and to present the results of centile modelling for the analysed somatic traits.

The results presented in this paper provide an empirical basis for further work on updated reference systems for basic somatic traits in Polish school-aged children and adolescents. Additional analyses based on the same dataset will be presented separately.

Materials and methods

The research project was reviewed by the Research Ethics Committee operating at the Institute of Sport – National Research Institute (approval No. KEBN-24-100-AP). Field examinations were conducted in 405 schools across all regions of Poland between May 2024 and June 2025. The school selection procedure was based on the sampling framework previously applied in the OLAF project (Kułaga et al., 2011), which was designed to approximate the population structure of children and adolescents in Poland. The original OLAF school list served as the starting point for sample construction in the present study, and approximately 76% of the schools included in the current survey had also participated in the OLAF project. Some schools from the original list could not be retained, mainly because of changes in the educational system following the abolition of lower secondary schools. To preserve, as far as possible, the assumptions of the original sampling framework and territorial and environmental comparability, these schools were replaced by the nearest schools enrolling students in the relevant age range.

Student selection followed a multistage systematic sampling procedure. At the school level, one class group from each educational level was selected. Within the selected class group, every third student was chosen from the class register, separately for girls and boys. After the candidates had been identified, written informed consent was obtained from the parents or legal guardians of minors and from students who were of legal age. In the absence of consent, the next student on the list was invited to participate, and the procedure was continued until the target number of participants had been reached or the list had been exhausted. At each educational level within a participating school, the plan was to include, within the selected class group, at least 3 students of each sex in primary schools and at least 6 students of each sex in post-primary schools, as well as 2 reserve students in each case.

The study had a cross-sectional design and was conducted in school settings according to a standardized measurement protocol. Measurements were performed by a team of 20 experienced anthropologists, each responsible for examinations in a designated region of the country. Before the start of data collection, all examiners underwent training in measurement techniques and error identification procedures.

The exact age of each participant on the day of examination was calculated as the difference between the examination date and the date of birth, expressed in days. For descriptive purposes, age was converted into decimal years (dividing the number of days by 365.25), whereas the continuous age variable used in modelling was expressed in months (age in years × 12).

Anthropometric techniques

Anthropometric measurements were performed during morning hours on a single day in each school among students selected for the study who were present on the examination day. This approach helped reduce within-day variation in the recorded values. Body height was measured in duplicate with a GPM anthropometer. Measurement error was calculated separately for each examiner. The Type A evaluation of standard uncertainty yielded values ranging from 0.01 to 0.03 cm, indicating very high repeatability of body height measurements. Participants were measured barefoot, with heels together, standing upright, shoulders and hips aligned, arms hanging freely, and the head positioned in the Frankfurt plane. Height was recorded to the nearest 0.1 cm. Body mass was measured in light underwear to the nearest 0.05 kg using a certified electronic medical scale (Charder MS6110).

Exclusion criteria

The exclusion criteria were consistent with those applied in the previous project (Kułaga et al., 2011) and included major postural abnormalities, genetic syndromes (e.g., Down syndrome, Turner syndrome), cancer, and other chronic diseases affecting growth. Additional exclusion criteria included previous growth hormone treatment or a diagnosis of growth hormone deficiency; systemic steroid therapy for conditions such as asthma or arthritis, or following renal transplantation; diabetes, cerebral palsy, cystic fibrosis, renal disease, congenital adrenal hyperplasia, congenital heart defects associated with impaired physical fitness; and thyroid hormone supplementation. Children wearing a plaster cast that could not be removed for measurement were also excluded, as its presence could affect measurement accuracy.

Statistical analysis

Body height and body mass were analysed using generalized additive models for location, scale, and shape (GAMLSS). This framework generalizes the LMS method (Cole, 1988; Cole & Green, 1992) by allowing the modelling of not only location and scale parameters but also distributional skewness (ν) and kurtosis/heavy tails (τ), which typically improves the fit to anthropometric data, particularly during puberty (Rigby & Stasinopoulos, 2005). GAMLSS-based approaches are widely used in the development of growth references and normative centile charts (e.g., Ortega et al., 2023; Tomkinson et al., 2018; WHO, 2006).

Data preparation and modelling assumptions

Data were initially recorded on paper forms and subsequently entered into a computer database (Excel). Standard quality control procedures were applied to identify and correct missing values, implausible records, duplicates, and outliers. The sample size (approximately 12,000 observations per sex, ages 6–19 years) was within the range discussed as adequate for constructing reference centiles across childhood and adolescence (Cole, 2021). Considering age expressed on a continuous scale in years, the density of observations across age supported the chosen degree of smoothing and the use of model diagnostics based on Q-tests and worm plots (Pan & Cole, 2004; Rigby & Stasinopoulos, 2005; Stasinopoulos et al., 2017).

Selection of the distribution family

To select the most appropriate distribution family, we applied a multi-stage comparative procedure based on the information criteria AIC (Akaike, 1998), BIC (Schwarz, 1978), GAIC6 and GAIC15 (Rigby & Stasinopoulos, 2005). Four distribution families commonly used in anthropometric modelling were considered: NO, BCCG, BCPE and BCT. For each outcome, five independent random subsamples comprising 33% of the dataset were drawn (subset validation). Within each subsample, we computed information-criterion deltas (ΔAIC, ΔBIC, ΔGAIC6, ΔGAIC15) as differences relative to the best model in that draw (i.e., the lowest value of a given metric) (Burnham & Anderson, 2002). The deltas were then min–max normalized within each metric and draw:

where j denotes the evaluated distribution family (NO, BCCG, BCPE, BCT).

Based on these values, a Weighted Information-Criteria Index (WICI) was computed for the BCCG, BCPE and BCT families:

WICI_j = 0,35 × Norm(∆AIC)_j + 0,35 × Norm(∆BIC)_j + 0,15 × Norm(∆GAIC6)_j + 0,15 × Norm(∆GAIC15)_j

where j denotes the evaluated family (BCCG, BCPE, BCT). WICI was restricted to BCCG/BCPE/BCT to prevent extremely large deltas for NO from dominating the min–max scaling and reducing the resolution of comparisons among shape-parameter families.

The NO family was treated as a baseline model and was allowed to be selected in each draw if it minimized the total sum of deltas ΔAIC+ΔBIC+ΔGAIC6+ΔGAIC15. In each of the five subsamples, the best family was defined as the one with the lowest WICI (or NO in the case of the baseline rule).

The final choice of distribution family followed a majority rule: we selected the family identified as best in at least three out of five draws. This threshold corresponds to the classical majority rule used in resampling procedures and reduces the impact of individual outlying draws (Efron & Tibshirani, 1993). If no family met this criterion, a fallback procedure was applied by selecting the family with the lowest mean WICI across draws. This approach was intended to ensure stability and reproducibility of the selected distribution family across the entire age range.

All analyses were performed using R (version 4.4.1; R Core Team) with the packages gamlss, gamlss.dist, and gamlss.add. The analytical workflow was implemented in custom modular R code, and reproducibility was ensured through the use of configuration files and a fixed random seed.

Full GAMLSS model estimation

Final models were fitted using GAMLSS, with the distribution family selected as described above. Age (in months) was modelled as a continuous covariate using penalized B-splines (pb) for the location parameter (μ) and, where applicable, for the scale and shape parameters (σ, ν, τ), yielding smooth age-specific trajectories in line with recommendations for modelling anthropometric traits (Cole & Green, 1992; Rigby & Stasinopoulos, 2005).

Model validation and diagnostics

Model fit was evaluated using diagnostic tools commonly recommended for GAMLSS (Rigby et al., 2019; van Buuren & Fredriks, 2001). Local diagnostics included Q-statistics and worm plots. Global diagnostics comprised residual Q–Q plots, residuals versus fitted values, residuals versus age, and residual histograms. We assessed variance stability, potential residual trends, and overall agreement between the residual distribution and the assumed theoretical model, both for the full sample and across age intervals.

Construction of centile charts

Based on the final model, the parameters μ, σ, ν and τ were predicted on an age grid with a step of 0.5 months. Selected centiles (3^rd, 5^th, 10^th, 15^th, 25^th, 50^th, 75^th, 85^th, 90^th, 95^th, and 97^th) were then computed as:

C_p = q_Dist (p; μ, σ, ν, τ)

where q_Dist denotes the quantile function of the selected distribution family (e.g. qBCCG, qBCPE, qBCT). Results were stored in tabular form with columns corresponding to age, C3, C5, C10, C15, C25, C50, C75, C85, C90, C95, and C97, which served as the basis for plotting the centile curves.

Results

A total of 23,988 students were measured (11,880 boys and 12,108 girls). There were 145 children under 6 years of age (65 boys and 80 girls) and 110 participants older than 18 years (63 boys and 47 girls, respectively). The numbers of participants by age category are presented in Table 1.

Overall project coverage of the population of children and adolescents in Poland (Project/Poland; as of 31 December 2024, Statistics Poland (GUS)) within the school-age range (6–18 years) was approximately 0.45%. Excluding the boundary age groups (6 and 18 years), coverage was relatively stable and ranged from approximately 0.41% to 0.57% depending on age and sex (Table 2).

For comparisons of the age structure of the study sample with population data from Statistics Poland (GUS), reference age was defined as the number of completed years as of 31 December 2024. In contrast, statistical analyses were based on age at examination (calculated from the date of birth and the measurement date) as a continuous variable. To assess differences arising from these two age definitions, we compared the numbers of participants by age category obtained using both methods (Table 2) and calculated the agreement between age categories, and the extent of ±1-year shifts (Table 3). The percentages for δ=−1 and δ=+1 describe the net balance of shifts in age-group aggregation resulting from the use of different reference points. As a summary measure, we used the difference between the proportion shifted to an older category and that shifted to a younger category, which was 1.22 percentage points in boys (17.02% − 15.80%) and 1.95 percentage points in girls (17.86% − 15.92%).

To verify the comparability of the study sample with voivodeship-level school-age populations (Table 4), we used the reference age as of 31 December 2024 and restricted analyses to ages 6–19 years, consistent with the population data range and the school-age period. Analyses of test outcomes were based on age at examination.

Overall, project coverage of the population of children and adolescents across voivodeships (Project/Poland) was 0.43% and was consistent across most regions (approximately 0.35–0.57%; Table 4). Higher coverage was observed in Świętokrzyskie (0.92%) and Opolskie (0.69%), whereas the lowest coverage was found in Mazowieckie (0.26%) and Śląskie (0.27%). Differences in coverage between voivodeships will be accounted for through the application of weighting factors to reduce the potential impact of unequal regional coverage on the development of reference data.

The results for body height and body mass measurements were presented as centile charts shown against raw-data scatterplots and supplemented with tables of centile values, as this form of presentation was considered more appropriate for GAMLSS modelling and better reflected the empirical distribution of the analysed somatic traits than descriptions based solely on classical descriptive statistics. Figures 1,2 and Table 5,6 present individual measurements of body height, whereas Figures 3,4 and Table 7,8 show body mass measurements for boys and girls examined in the project. The points represent raw observations, whereas the lines depict selected centile curves (P3–P97) estimated using GAMLSS models over the age range 6–18.5 years, providing a descriptive summary of the study population. Corresponding model parameter estimates and selected centile values are provided in the accompanying tables.

Discussion

The tradition of research on the physical development of children and adolescents in Poland dates back to the 1930s, when the Department of Anthropology of the Central Institute of Physical Education in Warsaw conducted the first large-scale project investigating the development and physical fitness of Polish youth (Mydlarski, 1934). This line of research was continued in the 1950s, with the aim of determining the extent of deterioration in the physical development of the post-war generation of children and adolescents in Poland (Trześniowski, 1961). Subsequent nationwide representative studies on the physical development and fitness of schoolchildren were carried out in 1979 (Trześniowski, 1990), 1989 (Przewęda & Trześniowski, 1996), 1999 (Przewęda & Dobosz, 2005, 2007), and 2009 (Dobosz, 2012; Dobosz et al., 2015). At the same time, detailed somatic measurements of adults, school-aged children, and adolescents throughout the country were conducted by anthropologists from the Committee of Anthropology of the Polish Academy of Sciences in Wrocław, established to implement the Polish Anthropological Nationwide Field Investigations project. Over a period of 60 years, five editions of this project were completed (Bielecki et al., 2012; Gomula et al., 2021; Kozieł et al., 2014).

At the beginning of the present century, population reference scales for infants as well as school-aged children and adolescents in Poland were still based on studies of Warsaw populations (Palczewska & Niedźwiedzka, 2001). The most recent nationwide studies based on a random sample of children and adolescents were conducted between 2007 and 2012 within the OLAF and OLA projects, and their results were presented as the Polish 2010 and 2012 reference systems for school-aged children and adolescents and for preschool children, respectively (Kułaga et al., 2011, 2013, 2015).

The present study has a cross-sectional design and lies at the intersection of the social and medical sciences. Its particular strengths include both the large sample size and the nationwide coverage. In order to obtain a sample that reflected as closely as possible the structure of the population of children and adolescents in Poland, the study was conducted in a substantial proportion of the schools previously included in the OLAF project (Kułaga et al., 2011). The sampling methodology also drew on the assumptions adopted in the third, fourth, and fifth editions of the nationwide anthropological surveys of children and adolescents, which were carried out in the same areas of the country (Kozieł et al., 2014).

An additional strength of the data collected in the project, and a guarantee of the quality of the measurements of basic anthropometric traits, is that they were performed by a small team of 20 anthropologists experienced in conducting this type of research. The Type A standard uncertainty of measurement ranged from 0.01 to 0.03 cm. In previous nationwide studies, anthropometric measurements were performed by large teams of trained students, representatives of the medical community (Kozieł et al., 2014; Kułaga et al., 2011, 2013), or physical education teachers (Dobosz, 2012; Dobosz et al., 2015; Przewęda & Dobosz, 2005, 2007; Przewęda & Trześniowski, 1996; Trześniowski, 1961, 1990).

In the analyses of study outcomes, age at examination was used, calculated on the basis of the full date of birth and the date of measurement. In contrast, analyses of participant counts and comparisons of the sample age structure with nationwide data (Główny Urząd Statystyczny, 2025) were based on reference age, defined as completed years of age as of 31 December 2024, in accordance with the definition used in official statistics. Differences between these two age definitions were assessed on the basis of shifts in counts and flow indicators across age categories (Tables 2 and 3). In the age range of 6–19 years, these differences were small (1.22 and 1.95 percentage points) and resulted mainly from shifts between adjacent categories (±1 year), which indicates that the use of reference age does not materially distort the description of the sample age structure in population comparisons. This approach justifies relating the study results to official statistics using the age categories applied by Statistics Poland (GUS).

At the same time, the analysis of counts by age category (6–18 years) revealed both underrepresentation and overrepresentation of some age groups relative to the nationwide population. Lower relative counts were observed among boys and girls aged 6 and 18 years, whereas higher relative counts were found among 14-year-old boys and among 10- and 14-year-old girls. Even more pronounced differences concerned the territorial structure of the sample. The highest relative share of study participants was recorded in the Świętokrzyskie voivodeship (0.92%), which was approximately 3.5 times higher than the corresponding values observed in the Śląskie and Mazowieckie voivodeships. These differences do not automatically justify the conclusion that the obtained results fail to reflect the distribution of the analysed somatic traits in a large nationwide sample. They do, however, indicate the need for further analyses and the application of statistical weights to adjust the sample structure to the population structure. This step appears necessary for the development of reliable reference scales for body height and body mass in the contemporary Polish population, as well as for assessing the agreement between the unweighted characteristics presented in this paper and the estimates obtained after weighting-based correction of the sample structure.

In Poland, multiple reference systems for the assessment of physical development continue to be used in parallel, including older charts based on Warsaw data (Palczewska & Niedźwiedzka, 2001) and more recent nationwide references developed within the OLAF/OLA studies. Kułaga and Kotowska (2025) showed that the nationwide references from 2010 and 2012, developed using the LMS method accounting for distributional skewness, provide computationally consistent results near key diagnostic cut-off points, such as the 3rd centile for body height and the 95th centile for BMI, whereas older Warsaw references may lead to inconsistencies in the estimation of centiles and Z-scores.

Against this background, the approach used in our project represents a further methodological step forward. GAMLSS models generalize the LMS framework and allow the distribution family to be selected on the basis of information criteria and goodness-of-fit diagnostics, including worm plots and Q-statistics. This makes it possible to achieve a better fit of the model to the empirical shape of the distributions of somatic traits. In addition, the short period of data collection re­duces the risk of within-study secular trends. The variation in cohort and voivodeship coverage identified in the present study provides a basis for the construction of weights and for adjusting the sample structure to the population structure, which will facilitate the development of updated nationwide reference data for the assessment of growth and development in school-aged children and adolescents.

It should be emphasized that the centile modelling results presented in this paper were based on a large nationwide sample; however, the analyses were performed without the application of weighting procedures correcting for differences between the structure of the study sample and that of the national population. Therefore, the resulting curves should currently be interpreted primarily as a description of the empirical distribution of the analysed somatic traits in the study sample. Although the sample size within particular age ranges was sufficient for stable modelling using the GAMLSS approach, a fuller alignment of the estimates with the structure of the nationwide population will require the application of weighting procedures in further analyses.

Conclusions

The present study, based on a cross-sectional design, is characterized by a large sample size and nationwide coverage. The high quality of the measurements of basic anthropometric traits in children and adolescents was ensured by their collection within a short period by a small team of experienced anthropologists. The methods used for data analysis are consistent with contemporary standards of statistical analysis. The results presented here provide a basis for the development of updated reference systems for basic somatic traits in school-aged children and adolescents aged 6–18 years.