JBI evidence synthesis · 2026
Growth curve models for weight among children up to 2 years: a scoping review.
СтатьяScoping ReviewНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Alves Marta, Serra Sofia, Costa Teresa, Brás-Geraldes Carlos, Papoila Ana Luisa, Heleno Bruno
Аннотация
OBJECTIVE: The aim of this scoping review was to systematically identify and map the literature on statistical and mathematical approaches used to estimate weight growth curves in children under 24 months of age. This review traces the evolution of these models and identifies knowledge gaps in these approaches.
INTRODUCTION: Growth monitoring through standardized growth curves, such as those from the World Health Organization, is essential for identifying growth faltering and emerging risks such as childhood obesity. Early growth models estimated individuals' growth trajectories, followed by fixed effects models estimating mean population growth curves. More advanced mixed effects regression models were later introduced to account for individual variations, followed by even more flexible models that allowed for the estimation of additional parameters (eg, Generalized Additive Models for Location, Scale, and Shape; SuperImposition by Translation and Rotation).
ELIGIBILITY CRITERIA: This review included studies using statistical or mathematical methodologies to estimate weight growth curves for children under 24 months of age. Studies using cohort or cross-sectional designs were included, while case series, reviews, short letters, books, and abstract-only publications were excluded. Only studies published in English, Portuguese, or Spanish were included.
METHODS: This review followed the JBI methodology for scoping reviews. Searches was conducted in 7 databases: PubMed, Scopus, Web of Science Core Collection, SciELO, and LILACS, with theses and dissertations retrieved from ProQuest Dissertations and Theses Citation Index and RCAAP. The search was conducted on July 25, 2024. Two reviewers independently screened titles, abstracts, and full texts, with additional studies identified through the reference lists of included articles. Two reviewers extracted data using a customized data extraction tool. The findings are summarized in tables and in a narrative synthesis.
RESULTS: A total of 4274 records was screened, resulting in the inclusion of 29 studies, published between 1987 and 2023, most employing a cohort study design. Seven studies used mathematical approaches, while 22 used statistical approaches. Based on their structural/non-structural classification, the Berkey-Reed first-order model was the most often applied (n=7), followed by the Count model (n=6). Both the Jenss-Bayley model and mixed effects regression models with fractional polynomials appeared in 4 studies each. All mathematical studies applied structural models, with the Count model being the most frequent, followed by the Berkey-Reed first-order. Among statistical studies, 13 applied non-structural models, 5 used structural models, and 4 used both types. For structural models, the mixed effects Berkey-Reed first-order was the most used (n=4), while for non-structural models, mixed effects fractional polynomials models were the most applied (n=4).
CONCLUSIONS: Both structural and non-structural models incorporate random effects, with the Berkey-Reed first-order and the Jenss-Bayley being the most used structural models, and fractional polynomials and SuperImposition by Translation and Rotation (SITAR) the most used non-structural models. Machine learning techniques, namely, artificial neural networks adapted for longitudinal data while providing interpretable results, may contribute to the development of more robust growth models.
REVIEW REGISTRATION: OSF https://osf.io/95udq.
INTRODUCTION: Growth monitoring through standardized growth curves, such as those from the World Health Organization, is essential for identifying growth faltering and emerging risks such as childhood obesity. Early growth models estimated individuals' growth trajectories, followed by fixed effects models estimating mean population growth curves. More advanced mixed effects regression models were later introduced to account for individual variations, followed by even more flexible models that allowed for the estimation of additional parameters (eg, Generalized Additive Models for Location, Scale, and Shape; SuperImposition by Translation and Rotation).
ELIGIBILITY CRITERIA: This review included studies using statistical or mathematical methodologies to estimate weight growth curves for children under 24 months of age. Studies using cohort or cross-sectional designs were included, while case series, reviews, short letters, books, and abstract-only publications were excluded. Only studies published in English, Portuguese, or Spanish were included.
METHODS: This review followed the JBI methodology for scoping reviews. Searches was conducted in 7 databases: PubMed, Scopus, Web of Science Core Collection, SciELO, and LILACS, with theses and dissertations retrieved from ProQuest Dissertations and Theses Citation Index and RCAAP. The search was conducted on July 25, 2024. Two reviewers independently screened titles, abstracts, and full texts, with additional studies identified through the reference lists of included articles. Two reviewers extracted data using a customized data extraction tool. The findings are summarized in tables and in a narrative synthesis.
RESULTS: A total of 4274 records was screened, resulting in the inclusion of 29 studies, published between 1987 and 2023, most employing a cohort study design. Seven studies used mathematical approaches, while 22 used statistical approaches. Based on their structural/non-structural classification, the Berkey-Reed first-order model was the most often applied (n=7), followed by the Count model (n=6). Both the Jenss-Bayley model and mixed effects regression models with fractional polynomials appeared in 4 studies each. All mathematical studies applied structural models, with the Count model being the most frequent, followed by the Berkey-Reed first-order. Among statistical studies, 13 applied non-structural models, 5 used structural models, and 4 used both types. For structural models, the mixed effects Berkey-Reed first-order was the most used (n=4), while for non-structural models, mixed effects fractional polynomials models were the most applied (n=4).
CONCLUSIONS: Both structural and non-structural models incorporate random effects, with the Berkey-Reed first-order and the Jenss-Bayley being the most used structural models, and fractional polynomials and SuperImposition by Translation and Rotation (SITAR) the most used non-structural models. Machine learning techniques, namely, artificial neural networks adapted for longitudinal data while providing interpretable results, may contribute to the development of more robust growth models.
REVIEW REGISTRATION: OSF https://osf.io/95udq.