As prevention of mother-to-child HIV-1 transmission (PMTCT) programs decrease the numbers of HIV-1-infected infants, it remains important to improve growth in HIV-1-exposed, uninfected (EU) infants. To determine the growth rate and predictors of growth faltering in breast-fed and formula-fed EU infants, growth analyses [weight-for-age (WAZ), weight-forlength (WLZ), and length-for-age (LAZ) Z-scores] were conducted by using data from a randomized feeding trial in HIV-1-infected women in Kenya. Growth faltering in EU infants was compared based on randomization to breastfeeding (BF) or formula feeding (FF) using Cox proportional hazards regression models. Linear mixed-effects models determined rate and cofactors of length growth. Among 338 EU infants, 164 (49%) were breast-fed and 174 (51%) formula-fed. In both arms, growth declined steadily during follow-up. By 2 y, 29% of children were underweight (WAZ < -2), 18% were wasted (WLZ < -2), and 58% were stunted (LAZ < -2), with no differences by feeding arm. Higher maternal education (y) and taller stature (cm) were associated with a decreased risk of underweight and stunting [underweight: adjusted HR (aHR) = 0.90 (95% CI: 0.83, 0.99), P = 0.03, and aHR = 0.92 (95% CI: 0.87, 0.97), P = 0.002; and stunting: aHR = 0.91 (95% CI: 0.85, 0.97), P = 0.003, and aHR = 0.96 (95% CI: 0.92, 0.99), P = 0.02, respectively]. Diarrhea was associated with an increased risk of wasting [aHR = 2.26 (95% CI: 1.11, 4.62), P = 0.03]. In multivariate analyses, FF was associated with slower declines in length velocity [0.24 LAZ/y (95% CI: 0.06, 0.43), P = 0.009]. Despite being uninfected, HIV-1-exposed infants showed frequent growth faltering, suggesting the need for vigilance in recognizing stunting within PMTCT programs. The slower rate of decline in length growth with FF may reflect benefits of micronutrients. Because BF is the best option for HIV-1-infected mothers in resource-limited settings, nutritional interventions should be examined for their impact on growth in EU breast-fed infants. © 2012 American Society for Nutrition.
We conducted secondary analysis of data collected from the Breastfeeding and HIV Transmission Study in Nairobi, Kenya, from November 1992 to July 1998, as described elsewhere (11, 12). Pregnant HIV-1–positive women were eligible to participate if they were Nairobi residents, had access to municipal-treated water, and were willing to be randomly assigned to BF or FF. This trial was conducted prior to the availability of antiretroviral prophylaxis for PMTCT in Africa; thus, none of the women received ARV. Of 425 women enrolled in the study, 212 were randomly assigned to the BF arm and 213 to the FF arm. There were 401 live-born singletons and firstborn infants, 197 in the BF arm and 204 in the FF arm. Follow-up information and a confirmed HIV PCR negative result at the first test were available for 164 infants in the BF arm and 174 in the FF arm (Supplemental Fig. 1). The study was approved by the institutional review boards of the University of Washington and the University of Nairobi, and all women provided informed consent. Per Kenyan guidelines at the time, mothers who were randomly assigned to BF were counseled to continue BF up to 2 y and beyond. Demonstrations of safe formula preparation in addition to free infant formula were given to the formula arm. Women were counseled to introduce complementary feeding at 6 mo of age. Study staff provided guidance on the optimal use and preparation of readily available household foods as complementary foods. Infants showing early signs of growth faltering were provided with formula as an adjunct to complementary foods. A complete physical examination was conducted in infants within 48 h of birth, at wk 2 and 6, then monthly until 12 mo, and quarterly until 24 mo to detect signs of HIV-1 infection and monitor growth and development. Weight was measured to the nearest 0.05 kg by using a beam balance scale, and recumbent length was measured to the nearest cm by using a length board. At each visit, information on duration and amount of breast-milk exposure was ascertained through maternal self-report. Infant blood samples were collected, and an HIV-1 DNA PCR assay was conducted to test for HIV-1 status at birth, at wk 6 and 14, and every 3 mo thereafter, as previously described (12). Maternal sociodemographic and behavioral characteristics were evaluated at enrollment through questionnaires. At 32 wk gestation, maternal plasma HIV-1 RNA viral load was measured by using a Gen-Probe HIV-1 RNA assay, and absolute CD4 cell count was obtained by using monoclonal antibodies and flow cytometry. Mortality and cause of death were determined by hospital record or verbal autopsy. Z-scores were used to standardize anthropometric measurements. The Z-score quantifies how many SD a child’s anthropometric value varies from the mean (Z-score = 0 or 50th percentile) value of a child of the same age and sex in a reference population. The 2006 WHO Child Growth Standards (13) were used to calculate WAZ, LAZ, and WLZ. This analysis included EU infants defined as HIV-1 antibody negative by PCR assay. Excluded infants included the following: 1) those testing HIV-1 PCR positive at birth or infants who never had a negative HIV-1 test result and 2) infants with no follow-up information after birth. Duration of follow-up was defined as age at last HIV-1–negative test result or age at last regular follow-up visit, whichever was less. Two analytical approaches were used. First, we conducted comparisons based on randomization to BF and FF by using intent-to-treat analysis. Second, because the randomized clinical trial had substantial noncompliance with the assigned feeding modality, we compared feeding as practiced: infants who ever breast-fed were compared with those who never breast-fed. Pearson’s chi-square test and Fisher’s exact test were used to compare categorical variables, and the Mann-Whitney U test was used to compare continuous variables. We considered confounding factors based on a priori knowledge of causal influences with growth. Birth weight, birth length, and feeding modality are known to influence growth and thus were included in all multivariate models. In addition, the following variables were considered potential confounding factors: maternal age, maternal height, education, gravidity, marital status, maternal HIV-1 viral load and CD4 count, diarrheal episode in the month preceding the study visit, and early introduction of other foods. Confounding factors were included in the final multivariate model if they were associated with growth (P < 0.10) in bivariate analysis and/or influenced the other coefficients upon inclusion into the model if P ≥ 0.10. Kaplan-Meier survival curves were used to compare time to growth faltering (Z-score of < −2) within the first 2 y of life. Specifically, growth faltering was subdivided and defined by the following parameters: underweight as a WAZ of < −2; wasting as a WLZ of < −2; and stunting as an LAZ of < −2. The log-rank test was used to compare differences between time-to-event curves. Cox proportional hazards regression models were used to estimate the risk of growth faltering for each of 3 anthropometric indicators independently, and RRs and 95% CI were calculated for covariates of interest including BF vs. FF arm. Infants who were lost to follow-up or who died were censored at the last known visit date. Infants who seroconverted to HIV-1 were censored at the last HIV-1–negative PCR test. We explored interactions between maternal height and infant birth size, modeling maternal height and birth weight and length as continuous variables, on risk of growth faltering. Loess curves were used to plot growth profiles based on feeding modality. Linear mixed-effects models with unstructured correlation and random intercepts and slopes for time were used to assess the rate of change in length growth and to determine the influence of BF on growth profiles. We modeled length as a continuous variable and calculated change in yearly LAZ in each level of the potential confounding factor by using an interaction term between the covariate and time since birth. The multivariate model included the main effect and interaction term with time for each covariate. Details of the confounding factors were considered, and model construction was similar to those described above. Statistical analyses were conducted by using SPSS 18 (SPSS, Inc.) and STATA 11.0 (StataCorp).