BMI Trajectory Analysis in Children: Review Maps Methods and Patterns

Researchers have published a scoping review of how scientists use latent class trajectory analysis to study body mass index changes in children and adolescents, offering the most detailed map yet of a method that is increasingly central to pediatric obesity research. The review, published in Discover Pediatrics, systematically examined 24 studies conducted between 2007 and 2025 across ten countries, cataloguing the statistical models, software tools, evaluation metrics and trajectory patterns that define this growing field.

The team, led by Diwen Yu and Li Li, followed the scoping review framework of Arksey and O’Malley. They searched eight databases, including PubMed, Web of Science, Embase, The Cochrane Library and four Chinese databases, covering literature from database inception to June 17, 2025. Two researchers independently screened titles, abstracts and full texts, with disagreements resolved by discussion with a third researcher. From 1,243 retrieved articles, 585 were removed by automated deduplication and 605 were excluded on title and abstract screening, leaving 24 studies for full analysis.

The included studies were cohort investigations with sample sizes ranging from 338 to 29,576 participants, and follow-up periods spanning 2 to 27 years. Eighteen of the 24 studies included more than 1,000 participants, and the median sample size across all studies was 1,744 cases. The research was conducted in China, the United States, France, the Netherlands, Canada, Australia, Italy, England, Portugal and Germany, with the United States contributing eight studies and China four.

On the methods side, the group-based trajectory model, or GBTM, dominated the field, appearing in 58.3 percent of studies. The growth mixture model, or GMM, was used in 25 percent, while the latent class growth model appeared in three studies and the latent growth mixture model in one. Software choices were similarly concentrated: Mplus was used in 29.2 percent of studies, SAS with the Proc Traj package in 20.8 percent, and Stata with the Traj package in 20.8 percent, while three studies used the lcmm package in R and four did not report their software at all.

Model evaluation practices showed considerable variation. The Bayesian Information Criterion was the most frequently used metric, appearing in 87.5 percent of studies, followed by the Average Posterior Probability at 58.3 percent, with all studies that reported a posterior probability threshold using 0.7 as the cutoff. Studies used between two and seven evaluation metrics, incorporating tools such as the Akaike Information Criterion, the Bootstrapped Likelihood Ratio Test, the Lo-Mendell-Rubin test, entropy and minimum group size requirements. One study did not specify its evaluation approach, and only two studies reported adherence to an eight-step framework for latent trajectory modeling proposed by Hannah Lennon and colleagues in 2018.

The number of BMI trajectories identified across the included studies ranged from 2 to 7, with four trajectories being the most common result, reported by 54.2 percent of studies. Most studies modeled either BMI directly, BMI Z-scores, or BMI standard deviation scores, and the authors note that the choice of metric matters: BMI Z-scores are suited to specific age ranges in childhood, while BMI SDS may better capture weight variation across broader age spans up to age 20.

Despite differences in naming conventions across the 88 trajectory labels found in the literature, the review identified six core trajectory types. A stable healthy weight trajectory, representing roughly 40 to 70 percent of children in most studies, was reported by more than 80 percent of the included research and showed the highest consistency. Other recurring patterns included persistent overweight or obesity from early age, early-onset obesity emerging in preschool or early childhood, delayed-onset weight gain accelerating around puberty, BMI decline that may reflect the normal physiological dip before adiposity rebound, and rapid BMI increase, which the authors associate with the highest long-term risks of obesity and metabolic disorders.

The review also highlighted why methodological choices matter. Citing work by Herle and colleagues on the ALSPAC cohort, the authors note that GBTM identified five trajectory types where GMM identified only three to four on the same data, because GBTM assumes zero within-group variance and may split individual deviations into extra categories, while GMM treats part of the variability as within-group random effects. The authors argue that the wide variation in reported trajectory numbers, which they state ranged from 2 to 9 across research reports, cannot be explained by population differences alone and likely reflects differing model assumptions, inconsistent terminology and incomplete model-fit reporting. They recommend reporting both GBTM and GMM results, following the GRoLTS checklist for reporting, and specifying software versions and packages.

Among the associated factors reported in the included studies, the review found consistent gender differences: girls’ trajectories diverged earlier than boys’, with girls more likely to show early-onset obesity and boys more likely to show late-onset patterns. One study reported that 64.1 percent of females versus 21.5 percent of males fell into a high-growth-rate group, while another found 46.5 percent of females versus 65.7 percent of males in a low-stability group. Childhood obesity was associated with precocious puberty, particularly in girls, and with maternal obesity, smoking, stress, sedentary behavior and dietary habits. Family environment factors, including weaker emotional self-regulation, lower parental attachment and more frequent parent-child conflict, were also linked to BMI trajectories, as were low parental income, low parental education, low socioeconomic status and rural residence.

The authors caution that their review has limitations. No formal methodological quality or bias risk assessment was performed on the included studies, the heterogeneity of BMI measures complicates comparisons, and differences in follow-up duration, measurement frequency and regional representativeness may affect the pooled description. They conclude that latent class trajectory analysis can display longitudinal BMI trends and identify relevant influencing factors, and that the findings may inform early weight management research and practice, but they emphasize that the clinical utility of trajectory classifications requires validation in prospective implementation studies, and that future research should standardize model evaluation criteria and reporting to improve comparability between studies.

Subject of Research: Latent class trajectory analysis of childhood and adolescent BMI

Article Title: A review of the scope of BMI latent class trajectory analysis in children and adolescents

Article References: Yu, D., Yang, X., Gu, Y., & Li, L. (2026). A review of the scope of BMI latent class trajectory analysis in children and adolescents. Discover Pediatrics, 1(1), Article 9. https://doi.org/10.1007/s44571-026-00007-3

Image Credits: AI Generated

DOI: 10.1007/s44571-026-00007-3

Keywords: BMI trajectories, children, adolescents, latent class trajectory analysis, group-based trajectory model, growth mixture model, obesity, scoping review, pediatrics, weight management, model evaluation, longitudinal analysis