DO MODIFIABLE HOUSEHOLD BEHAVIORS AND EARLY INFANT FEEDING PRACTICES CONTRIBUTE TO VARIATIONS IN INFANT LINEAR GROWTH? EVIDENCE FROM A BIRTH COHORT IN DHAKA, BANGLADESH

Abstract

BACKGROUND Numerous postnatal risk factors have been associated with child length/height in cross-sectional studies in low- and middle-income countries. However, there have been few longitudinal studies of the effects of modifiable risk factors on postnatal linear growth during discrete developmental windows of infancy. OBJECTIVES We aimed to assess associations between modifiable household behaviours and conditional growth from birth to 1 year of life. DESIGN/METHODS We conducted a longitudinal cohort study using data from women and their infants (n=1162 pairs) in the Maternal Vitamin D and Infant Growth trial in Dhaka, Bangladesh. Infant length was measured tri-monthly from birth to 12 months, and infant feeding patterns were ascertained at weekly visits from 0 to 6 months of age. Confounder-adjusted associations of selected modifiable household factors (i.e., household air quality, sanitation/hygiene) or early infant feeding with change in length-for-age z-score (LAZ) were estimated in five intervals: birth to 3 months, 3 to 6 months, birth to 6 months, 6 to 12 months and birth to 12 months. In primary analyses, the outcome was conditional growth in LAZ (cLAZ) in each interval, derived as model residuals from regression of end-interval LAZ on initial LAZ. Effect estimates were expressed as mean difference in cLAZ (95% confidence interval) between the exposed versus referent group. RESULTS LAZ was symmetrically distributed, with mean (± standard deviation) LAZ of -0.95 (± 1.02) at birth and -1.00 (± 1.04) at 12 months. In multivariable-adjusted linear regression models, indicators of household air quality and sanitation/hygiene were not significantly associated with cLAZ in any interval. No breastfeeding and partial breastfeeding (versus exclusive breastfeeding), and any infant formula use (versus no formula use) were associated with slower growth in the 0–3 month interval: -0.11 (95% CI: -0.20, -0.02), -0.30 (95% CI: -0.52, -0.08), and -0.13 (95% CI: -0.22, -0.05), respectively, but not in later intervals. Several non-modifiable factors (maternal height, paternal education, and household wealth) were associated with cLAZ and LAZ in multivariable models. CONCLUSION Compared to international standards, the length distribution of infants in Dhaka, Bangladesh was harmonically shifted down at birth and throughout the first year of life, suggesting that observed infant length deficits relative to international norms were primarily caused by ubiquitous factors. Infant feeding practices explained some between-child variation in linear growth in the early postnatal period (0–3 months). Behaviors related to cooking or sanitation/hygiene were not related to infant linear growth trajectories.