Characterizing early child growth patterns of height-for-age in an urban slum cohort of Bangladesh with functional principal component analysis

Abstract

Background: Early childhood is a critical stage of physical and cognitive growth that forms the foundation of future wellbeing. Stunted growth is presented in one of every 4 children worldwide and contributes to developmental impairment and under-five mortality. Better understanding of early growth patterns should allow for early detection and intervention in malnutrition. We aimed to characterize early child growth patterns and quantify the change of growth curves from the World Health Organization (WHO) Child Growth Standards.

Methods: In a cohort of 626 Bangladesh children, longitudinal height-for-age z-scores (HAZ) were modelled over the first 24 months of life using functional principal component analysis (FPCA). Deviation of individual growth from the WHO standards was quantified based on the leading functional principal components (FPCs), and growth faltering was detected as it occurred. The risk factors associated with growth faltering were identified in a linear regression.

Results: Ninety-eight percent of temporal variation in growth trajectories over the first 24 months of life was captured by two leading FPCs (FPC1 for overall growth and FPC2 for change in growth trajectory). A derived index, adj-FPC2, quantified the change in growth trajectory (i.e., growth faltering) relative to the WHO standards. In addition to HAZ at birth, significant risk factors associated with growth faltering in boys included duration of breastfeeding, family size and income and in girls maternal weight and water source.

Conclusions: The underlying growth patterns of HAZ in the first 2 years of life were delineated with FPCA, and the deviations from the WHO standards were quantified from the two leading FPCs. The adj-FPC2 score provided a meaningful measure of growth faltering in the first 2 years of life, which enabled us to identify the risk factors associated with poor growth that would have otherwise been missed. Understanding faltering patterns and associated risk factors are important in the development of effective intervention strategies to improve childhood growth globally.

Trial registration: ClinicalTrials.gov Identifier: NCT02734264 , registered 22 March, 2016.

Keywords: Anthropometry; Functional data analysis; Growth faltering; Longitudinal growth; Stunting.

Fig. 1

Individually fitted curves of height (Panels a and b) and height-for-age z-score (HAZ) (Panels c and d) from birth to age 24 months for boys (light blue, n = 270) and girls (pink, n = 225). The mean curves are shown in solid blue line for boys and red line for girls. The World Health Organization (WHO) growth standards are shown as the solid black lines in Panels a and b and as solid horizontal black lines in Panels c and d for the WHO “pseudo children” who were growing along the WHO percentiles, separately for boys and girls

Fig. 2

The two leading functional principal components (FPCs) in boys (black) and girls (grey). The FPC1 in both boys and girls (Panel a) were negative and monotonically decreased over time, reflecting further deviation of individual growth patterns from the mean curve. The FPC2 (Panel b) changed signs approximately at 12 months for boys and 14 months for girls, indicating the changes in growth trajectory. For boys, the FPC1 and FPC2 accounted for 93 and 6% of the variation among fitted height-for-age z-score (HAZ) curves. For girls, the FPC1 and FPC2 accounted for 96 and 3% of the variation among fitted HAZ curves

Fig. 3

Functional principal component analysis (FPCA) results for boys (Panels a and c) and girls (Panels b and d). The mean curves for height-for-age z-score (HAZ) are shown as solid lines, and the changes from the mean curves for FPC1 (Panels a and b) and FPC2 (Panels c and d) are shown in “+++” and “---” when 2•SD (standard deviation) of FPC scores are added to (the “+++” curve) or subtracted from (the “---” curve)

Fig. 4

Histograms of estimated FPC1 scores (solid bars in Panels a and b), original FPC2 scores (clear bars in Panels c and d) and adj-FPC2 scores (solid bars in Panels c and d) in boys and girls. The FPC1 scores for WHO “pseudo children” (growing along WHO percentiles) are shown as dashed lines in Panels a and b, whereas most study children were below the WHO 50^th percentile. Positive adj-FPC2 scores in Panels c and d quantified the degree of growth faltering. Abbreviations: adj-FPC2, adjusted FPC2; FPC, functional principal component; WHO, World Health Organization