Anthropometric prediction models of body composition in 3 to 24month old infants: a multicenter international study

Accurate assessment of body composition in infants is crucial for evaluating early growth and identifying potential nutritional risks. The early years of life are foundational for long-term health, with nutritional status impacting the risk of non-communicable diseases later in life. Several factors influence nutritional status during infancy, including birth weight, feeding practices, dietary nutrient composition, genetics, and environmental factors. Anthropometry, while commonly used, has limitations in assessing body composition. Weight-for-length and BMI-for-age, often employed, don't differentiate between fat mass (FM) and fat-free mass (FFM). Furthermore, anthropometric cut-offs are often based on population distribution rather than biological relevance, and reference standards may be inaccurate for specific ethnic groups, leading to underestimation of overnutrition and overestimation of undernutrition. While body composition is a superior indicator of nutritional status, it's not always practical for widespread use. Anthropometric approaches are easy, quick, inexpensive, portable, and minimally invasive, making them valuable tools. This study aims to develop anthropometry-based prediction equations for body composition in infants from 3 to 24 months old across diverse socioeconomic and ethnic backgrounds, improving upon existing equations that often lack generalizability to various populations.

Numerous anthropometry-based prediction equations exist for assessing body composition in infants. Studies have shown that combining anthropometric measures (e.g., length, weight-for-length, skinfold thickness), gestational age, sex, and ethnicity improves prediction accuracy. However, many of these equations are not generalizable beyond the populations they were developed in, and equations developed in multi-ethnic populations are rare. The lack of widely applicable equations highlights the need for this study, which aims to address this gap by developing and validating equations for a diverse population.

This observational, longitudinal, prospective, multinational study, the Multi-center Infant Body Composition Reference Study (MIBCRS), followed infants from birth to 24 months in countries representing diverse socioeconomic settings (lower-middle, upper-middle, and high-income). The study adhered to international ethical guidelines and obtained ethical approvals. Body composition was assessed using deuterium dilution (DD), while anthropometric data (weight, length, triceps skinfold thickness (TSFT), subscapular skinfold thickness (SSFT)) were collected at 3, 6, 9, 12, 18, and 24 months (3–24-month cohort) or at 6 months (test cohort). Standardized anthropometry protocols were used, with regular standardization sessions to maintain accuracy. Quality control measures were implemented for both DD and anthropometry data collection. Data analysis was performed using the REDCap system. The 3–24-month cohort was randomly split into training (two-thirds) and validation (one-third) datasets. An independent cohort (3–6 months) served as the test dataset. Linear mixed modeling was used to generate sex-specific prediction equations for FM and FFM, using length, weight-for-length, TSFT, SSFT, and South Asian ethnicity as predictors. Models incorporated linear splines with knots at 9 and 18 months to account for non-linear age effects. The models were validated internally (on the validation dataset) and externally (on the test dataset). Several error metrics were used to evaluate the accuracy of the prediction equations, including root mean squared error (RMSE), root mean square percentage error (RMSPE), mean absolute error (MAE), and mean absolute percentage error (MAPE). Sensitivity analyses were conducted to assess model misspecification and evaluate the impact of alternative predictor variables and transformations.

Sex-specific equations for FM and FFM were developed for three age categories (3–9 months, 10–18 months, 19–24 months). RMSE was similar across training, validation, and test datasets for FFM in both boys and girls. For FM, RMSPE and MAPE were higher in the validation dataset compared to the test dataset. RMSE for the South African test data showed good agreement with validation data for FFM, while Australia and India showed less agreement. Length and weight-for-length were positively associated with FM and FFM. Skinfold thicknesses were positively associated with FM but negatively associated with FFM. The association between age and FM/FFM was non-linear. Sensitivity analysis showed that the linear spline model provided a good fit for the data. The equations showed better prediction accuracy for FFM compared to FM.

This study successfully generated anthropometry-based prediction equations for FM and FFM in infants across diverse populations, providing a valuable tool for assessing body composition in resource-limited settings. The equations performed well across internal and external validation, demonstrating good generalizability. The better prediction accuracy for FFM suggests that calculating FM by subtracting FFM from total weight may be more accurate than directly predicting FM. The inclusion of ethnicity as a predictor variable is crucial, recognizing existing ethnic variation in body composition. The study's strength lies in its large, multi-ethnic, and diverse sample, encompassing various socioeconomic backgrounds. The results support the importance of utilizing equations developed on similar populations to maximize accuracy. While the equations demonstrate good applicability, the sensitivity analysis highlighted some limitations that need to be further studied.

This study provides novel anthropometry-based prediction equations for FM and FFM in infants aged 3–24 months, stratified by sex and age group. The equations show good prediction accuracy, particularly for FFM, and are applicable across diverse populations. Future research could focus on refining these equations using larger datasets, incorporating additional anthropometric measures, and exploring other factors influencing body composition during infancy.

The study's primary limitation is the potential for residual confounding, despite adjusting for several factors. While the sample size is substantial, it may not fully capture the diversity of all global populations. Further, the reliance on anthropometry may not accurately capture all aspects of body composition, potentially underestimating or overestimating certain components. Additionally, the study's equations might require adjustments for specific subpopulations with extreme anthropometric values or other conditions.