Growth faltering due to breastfeeding cessation in uninfected children born to HIV-infected mothers in Zambia

Abstract

Background: The effect of breastfeeding on growth in HIV-exposed infants is not well described.

Objective: The objective was to evaluate the effect of early breastfeeding cessation on growth.

Design: In a trial conducted in Lusaka, Zambia, HIV-infected mothers were randomly assigned to exclusive breastfeeding for 4 mo followed by rapid weaning to replacement foods or exclusive breastfeeding for 6 mo followed by introduction of complementary foods and continued breastfeeding for a duration of the mother's choice. Weight-for-age z score (WAZ), length-for-age z score (LAZ), and weight-for-length z score (WLZ) and the self-reported breastfeeding practices of 593 HIV-uninfected singletons were analyzed. Generalized estimating equations were used to adjust for confounders.

Results: WAZ scores declined precipitously between 4.5 and 15 mo. The decline was slower in the breastfed infants. At 9, 12, and 15 mo, mean WAZs were, respectively, -0.74, -0.92, and -1.06 in infants who were reportedly breastfed and were -1.07, -1.20, and -1.31 in the weaned infants (P = 0.003, 0.007, and 0.02, respectively). No differences were observed past 15 mo. Breastfeeding practice was not associated with LAZ, which declined from -0.98 to -2.24 from 1 to 24 mo. After adjustment for birth weight, maternal viral load, body mass index, education, season, and marital and socioeconomic status, not breastfeeding was associated with a 0.28 decline in WAZ between 4.5 and 15 mo (P < 0.0001). During the rainy season, not breastfeeding was associated with a larger WAZ decline (0.33) than during the dry season (0.22; P for interaction = 0.02).

Conclusions: Early growth is compromised in uninfected children born to HIV-infected Zambian mothers. Continued breastfeeding partially mitigates this effect through 15 mo. Nutritional interventions to complement breastfeeding after 6 mo are urgently needed. This trial was registered at clinicaltrials.gov as NCT00310726.

FIGURE 1

Flow chart detailing the number of randomized subjects excluded from the analysis, the reason for their exclusion, and the distribution of included subjects by randomization group and actual breastfeeding (BF) status at 5 mo.

FIGURE 2

Mean (±SE) weight-for-age z score among 593 uninfected children born to HIV-infected mothers at each follow-up visit by actual breastfeeding (BF) practice at the time of each visit. Significant differences between groups at a single visit using the t test are indicated with asterisks (P = 0.003, 0.007, and 0.02 at 9, 12, and 15 mo, respectively). Sample sizes are presented below the graph, and the mean is plotted only when the sample size in groups exceeded 10.

FIGURE 3

Mean (±SE) length-for-age z score among 593 uninfected children born to HIV-infected mothers at each follow-up visit by actual breastfeeding (BF) practice at the time of each visit. No significant differences were observed (t test). Sample sizes are presented below the graph, and the mean is plotted only when the sample size in groups exceeded 10.

FIGURE 4

Mean (±SE) weight-for-length z score among 593 uninfected children born to HIV-infected mothers at each follow-up visit by actual breastfeeding (BF) practice at the time of each visit. Significant differences between groups at a single visit are indicated by asterisks (P = 0.05, 0.0005, and 0.02 at 9, 12, and 15 mo, respectively; t test). Sample sizes are presented below the graph, and the mean is plotted only when the sample size in groups exceeded 10.

FIGURE 5

Mean (±SE) weight-for-age z scores of uninfected children born to HIV-infected mothers at each follow-up visit by actual breastfeeding (BF) practice at the time of each visit during the rainy (A) and dry (B) seasons. Significant differences between groups at a single visit are indicated by asterisks (dry season, 6 mo: P = 0.03; rainy season, 9, 12, and 15 mo: P = 0.004, 0.005, and 0.05 respectively; t test). The interaction between BF practice and season in the generalized estimating equation regression model that assessed the association between weight-for-age z score and BF practice at time of visit for children aged 4.5–15 mo was significant using the z statistic (P = 0.02) after age, birth weight, maternal viral load, BMI, marital status, education, and socioeconomic status were controlled for. Sample sizes are presented below each graph, and the mean is plotted only when the sample size in groups exceeded 10.