|Home | About | Journals | Submit | Contact Us | Français|
Background.We assessed morbidity rates during short intervals that accompanied weaning and cumulative mortality among HIV-exposed, uninfected infants enrolled in the postexposure prophylaxis of infants in Malawi (PEPI-Malawi) trial.
Methods.Women were counseled to stop breastfeeding (BF) by 6 months in the PEPI-Malawi trial. HIV-uninfected infants were included in this analysis starting at age 6 months. Breastfeeding and morbidity (illness and/or hospital admission and malnutrition [weight-for-age Z-score, ≤2]) were assessed during age intervals of 6–9, 9–12, and 12–15 months. BF was defined as any BF at the start and end of the interval and no breastfeeding (NBF) was defined as NBF at any time during the interval. The association of NBF with morbidity at each mutually exclusive interval was assessed using Poisson regression models controlling for other factors. Cumulative mortality among infants aged 6–15 months with BF and NBF was assessed using an extended Kaplan–Meier method.
Results.At age 6 months, 1761 HIV-uninfected infants were included in the study. The adjusted rate ratios for illnesses and/or hospital admission for NBF, compared with BF, was 1.7 (P < .0001) at 6–9 months, 1.66 (P = .0001) at 9–12 months, and 1.75 (P = .0008) at 12–15 months. The rates of morbidity were consistently higher among NBF infants during each age interval, compared with BF infants. The 15 months cumulative mortality among BF and NBF children was 3.5% and 6.4% (P = .03), respectively.
Conclusions.Cessation of BF is associated with acute morbidity events and cumulative mortality. Prolonged BF should be encouraged, in addition to close monitoring of infant health and provision of support services.
Optimizing the survival of HIV-uninfected infants born to HIV-infected mothers (ie, HIV-exposed infants) is a major challenge in sub-Saharan Africa, where prevalence of HIV infection remains high among reproductive-age women . The detrimental effects of early weaning to prevent HIV transmission have been shown in multiple studies from Malawi, Uganda, and Botswana [2–5]. In these studies, the risk of severe gastroenteritis and mortality associated with gastroenteritis were higher among HIV-exposed but uninfected infants who were weaned early (at 4–6 months of age), compared with those who breastfed (BF) for longer periods. Additional adverse effects of nonbreastfeeding (NBF) or early weaning among HIV-uninfected African children include acute malnutrition, such as marasmus and kwashiorkor , and growth faltering [6, 7]. Infants who stopped BF also did not show increases in HIV-free survival (ie, both remaining alive and free of HIV infection) [8, 9]. How to balance the risks and benefits of BF of HIV-uninfected children of HIV-infected mothers in resource-constrained settings of sub-Saharan Africa where BF alternatives are not safe has been extensively discussed [10–12]. The recent World Health Organization (WHO) recommendations on infant feeding and HIV emphasize infant HIV-free survival .
Despite the evidence that NBF increases infant morbidity and mortality in resource-constrained settings, our understanding of the timing of these adverse events is limited. For clinical management and program planning, it will be valuable to know immediate and long-term effects of weaning whenever this practice is initiated. We hypothesize that cessation of BF to prevent transmission of HIV leads to early acute morbidity that occurs during the process of weaning or immediately after weaning. However, infant mortality associated with weaning is a late effect and occurs after the weaning process is completed. The aim of this study was to determine, among HIV-exposed, uninfected Malawian infants, morbidity rates in short follow-up intervals (~3 months) that encompass weaning and to assess cumulative rates of mortality over time. These infants were originally enrolled in the extended antiretroviral Post-exposure Prophylaxis of Infants in Malawi (PEPI-Malawi) trial.
The PEPI-Malawi study was a randomized clinical trial conducted in Blantyre, Malawi, during 2004–2009 to assess efficacy of extended infant antiretroviral prophylaxis to reduce postnatal HIV transmission. The details of this trial are reported elsewhere . In brief, infants were randomized to 3 arms: control regimen, in which infants received oral single-dose nevirapine (NVP) at birth and twice-daily oral zidovudine (ZDV) for 1 week; control regimen followed by extended daily NVP to age 14 weeks; and control regimen followed by extended daily NVP plus ZDV to age 14 weeks. During the intervention, infants (and their mothers) were seen at birth and at ages 1, 3, 6, 9, and 14 weeks. After 14 weeks, follow-up visits were at age 6 months and every 3 months thereafter to age 24 months.
Maternal sociodemographic information was collected at enrollment. As part of the PEPI-Malawi trial, women were counseled to stop BF at 6 months. BF history was obtained using a structured questionnaire that included information on whether a woman was still BF at the scheduled visit. Infant frequency of reported illnesses and hospital admissions and weight and height measurements were collected at each follow-up visit. The variable “illness” included positive responses from 2 separate questions: if the infant was admitted to hospital and/or if infant was taken to a clinic for an illness since the last visit. As part of the clinical service provided during the conduct of the PEPI study, all infants received Pneumocystis carinii (now Pneumocystis jiroveci) pneumonia (PCP) trimethoprim-sulfamethoxazole (TMP-SMX) prophylaxis. TMP-SMX prophylaxis started at age 6 weeks (daily ¼ adult tablet 400 mg/80 mg for infants < 6 months and ½ the adult tablet for infants ≥6 months). HIV-uninfected infants stopped TMP-SMX prophylaxis at age 9 months if they stopped BF at 6 months. If the infant was HIV uninfected and did not stop BF, TMP-SMX prophylaxis was continued until 3 months after cessation of BF. HIV-infected infants would continue TMP-SMX prophylaxis for life. Information on TMP-SMX prophylaxis was collected at each visit. Infant death was ascertained from maternal reports and hospital records. Infant HIV testing was performed at each visit, as described elsewhere . The PEPI study was approved by institutional review boards at the College of Medicine, University of Malawi, Johns Hopkins University Bloomberg School of Public Health, and the Centers for Disease Control and Prevention.
In the current study, infants BF at 14 weeks and HIV uninfected at 6 months were included. The analysis started at age 6 months and included infants known to be BF or NBF. Women with inconsistent BF history (eg, reported BF after an earlier NBF history) were entirely excluded from the analysis (n = 26). Infants who became HIV infected after 6 months were censored. The outcomes of interest were infant morbidity and mortality. Morbidity includes illnesses and/or hospital admissions and malnutrition (as defined below). Illnesses and/or admissions (ever had illness or admitted to hospital since last visit) were obtained from case report forms based on maternal history at the scheduled visits of 6, 9, 12, and 15 months. The analysis included 3 mutually exclusive age intervals (6–9, 9–12, and 12–15 months); no further analysis beyond age 15 months was considered, because BF frequency after 15 months was low. Malnutrition was defined as weight-for-age Z-score ≤2, based on WHO 2006 growth curves . For morbidity analyses, events during the interval between visits were considered.
Both BF and NBF and the outcomes of interest were assessed during the same interval (6–9, 9–12, and 12–15 months). The group of infants labeled as BF was reported as BF at both the start and end of the interval. The group of infants labeled as NBF included infants of women who reported NBF at the beginning and end of the interval and infants of women who switched from BF to NBF during the interval (the date when a woman switched from BF to NBF is not known, but would have occurred sometime during the interval). These 2 groups of NBF infants were combined and identified as the NBF group in this analysis, because the characteristics of these women were similar for the majority of variables. In addition, our interest was assessment of the effect of NBF, compared with BF, on morbidity and mortality. Comparison of various demographic and clinical factors by BF and NBF during each age interval was performed. Statistical significance (at P ≤ .05) was determined using χ2 test for unadjusted binary variables, Student t test for unadjusted continuous variables, and log rank test for time to event analyses.
Morbidity (having at least 1 event in the interval—that is, no repeated observations in the same interval) rates among BF and NBF infants were estimated using number of events and person-years (PY)of follow-up during each of the age intervals 6–9, 9–12, and 12–15 months (rates estimated dividing number of events by total PY times 100). Mortality analysis included cumulative death probability during 6–15 months of age with use of an extended Kaplan–Meier estimator method  that considered BF and NBF at the beginning of the interval as a time-varying covariate and grouped the infant into denominator and numerator of the BF and NBF Kaplan–Meier risk set according to this value. The association of NBF with outcomes at each mutually exclusive 3-month interval was assessed using Poisson regression models for morbidity and Cox proportional hazards models for mortality with a time-varying NBF covariate. These models controlled for infant extended antiretroviral prophylaxis, maternal HIV disease clinical stage, and infant TMP-SMX prophylaxis (with the exception of infant extended antiretroviral prophylaxis, the other variables were time-varying covariates). Maternal plasma viral load and CD4 cell count measurements were not available for all visits beyond baseline; we therefore used HIV clinical stage at each visit to control for maternal disease stage.
Table 1 shows the characteristics of the 2 groups of women (BF and NBF) and their infants at each infant age interval: the characteristics of the BF group (column 3 vs 5 vs 7) remained approximately comparable during each age interval (ie, remained stable over time). Likewise, the characteristics of NBF group (column 4 vs 6 vs 8) remained approximately comparable and showed no change between intervals. In each age interval, however, there were few statistically significant differences between the BF and NBF group (Table 1).
Table 2 shows morbidity rates at each age interval among the BF and NBF groups. During each interval the rates of illness and/or hospital admission and malnutrition were consistently higher in the NBF group than in the BF group. The confidence intervals (CIs) did not overlap for the morbidity outcomes (illness and/or hospital admission and malnutrition) for all age intervals, with the exception of malnutrition rate in the age group 9–12 months. The differences in morbidity rates between BF and NBF groups were still maintained if infants of women who switched from BF to NBF during the interval were excluded and the analysis was limited to infants of women who were NBF throughout the interval. For example, the hospital admission and/or illness rates were: 6–9-month interval, 51.33 cases/100 PY (95% CI, 39.61–65.42 cases/100 PY) in BF infants and 91.94 cases/100 PY (95% CI, 64.04–127.87 cases/100 PY) in NBF infants; 9–12 months interval, 52.84 cases/100 PY (95% CI, 37.75–71.95 cases/100 PY) in BF infants and 95.55 cases/100 PY (95% CI, 84.53–107.61 cases/100 PY) in NBF infants; and 12–15 months interval, 30.23 cases/100 PY (95% CI, 16.92–49.87 cases/100 PY) in BF infants and 68.57 cases/100 PY (95% CI, 59.46–78.68 cases/100 PY) in NBF infants.
The adjusted rate ratios from the Poisson multivariable models for association of NBF with morbidity are shown in Table 3. After controlling for extended antiretroviral infant prophylaxis, infant TMP-SMX prophylaxis, and maternal HIV disease stage, NBF was significantly (always P ≤ .0008) associated with higher risk of illness and/or hospital admission at each of the infant age intervals (~1.7-fold higher risk in NBF group, compared with BF group). NBF was also associated with significantly (P < .05) higher risk of malnutrition, except during age interval 9–12 months (P = .07). Another factor significantly (always P < .005) associated with lower risk of illness and/or hospital admission during each interval was receipt of infant TMP-SMX prophylaxis. Similarly, infant TMP-SMX prophylaxis was significantly (P < .005) associated with lower risk of malnutrition during age intervals 6–9 and 9–12 months. Advanced maternal HIV disease (stage 3 or 4) was associated (P < .01) with increased risk of infant illness and/or hospital admission, compared with disease stage 1 or 2 during the interval 12–15 months.
Figure 1 shows KM mortality curves, and Table 4 shows the association of NBF with mortality, adjusting for other factors with use of a proportional hazard model. Figure 1 shows no differences in mortality during the period 6–12 months. The differences in mortality became more apparent later after the first year; NBF infants showed higher mortality during the second year, compared with BF infants. At 15 months, the cumulative rate of mortality was 6.4% in the NBF group and 3.5% in the BF group. In the multivariable model, NBF was significantly (P = .04) associated with higher mortality after adjusting for antiretroviral extended prophylaxis, maternal HIV disease, and infant TMP-SMX prophylaxis. Likewise, advanced maternal HIV disease stage was associated with significantly higher mortality after adjusting for other factors.
The findings from this analysis show strong associations between NBF and morbidity at each age interval. Rates of illnesses and/or admissions and malnutrition were consistently higher in the NBF group than in the BF group; the adjusted rate ratios for illnesses and/or hospital admission were significantly higher in the NBF group than in the BF group; similarly, the adjusted rate ratios of malnutrition were consistently higher in the NBF group than in the BF group. These findings support our original hypothesis and suggest that the adverse effects of weaning occur quickly and manifest as acute events resulting in clinic visit or hospital admission. These complications occurred despite continuous counseling about BF and provision of care for women and children in the PEPI-Malawi study .
We also estimated cumulative mortality at 6–15 months of age. The cumulative mortality rate at 15 months in the NBF group was approximately twice the rate in the BF group (6.4% vs 3.5%). However, these rates were similar during 6–12 months of age and substantially higher in the NBF group during 12–15 months of age. Although there were no deaths after 12 months in the BF group, it should be noted that the denominator becomes small after 12 months (Figure 1). After adjusting for other factors, NBF was significantly associated with higher hazard of death. The association of cessation of NBF with mortality appears to be substantially higher than the baseline rate among BF infants starting in the second year. These results agree with findings from another study that showed that early cessation of BF is associated with increased mortality that extends to the second year among HIV-uninfected African children born to HIV-infected mothers . Other studies that examined interval-specific effect of BF were mainly focused on the effect of early exclusive BF on morbidity .
An important secondary finding in this study was that TMP-SMX prophylaxis to prevent PCP of HIV-exposed, uninfected children was significantly associated with lower frequency of illnesses and/or admissions at each interval (and similarly early malnutrition) after simultaneously controlling for BF and other factors. PCP prophylaxis was provided to HIV-exposed BF infants as part of clinical care and was not part of the main PEPI trial in Malawi. Although this finding is observational and should be interpreted with caution, it reveals that TMP-SMX has benefits and potentially could prevent other infections that could lead to illnesses and hospital admissions among HIV-uninfected children. Similar benefits of reductions in hospital admissions, antibiotic prescriptions, and death have been reported from other countries in Africa among HIV-infected children [19, 20]. In addition, TMP-SMX taken by HIV-infected adults was shown to reduce morbidity and mortality among other HIV-uninfected family members . In Zambia, a study among HIV-infected children reported reductions in morbidity and mortality despite high background TMP-SMX resistance . A clinical trial (the PROMISE study) is planned to start soon to assess whether extended TMP-SMX prophylaxis can reduce morbidity and mortality after BF cessation among HIV-exposed, uninfected children (clinical trial NCT01061151).
A potential bias in our study is censoring of infants who became HIV infected during the study follow-up period, because most of these infants were in the BF group. Therefore, it is likely that the BF group would appear to be healthier than the NBF group, and some of the differences in morbidity and mortality may have been over-estimated. However, the rates of HIV infection during 6–15 months of age were low. Another limitation common to BF studies is how to address the issue of reverse causality  (ie, the association between NBF and increased infant morbidity and mortality is caused by other factors [eg, infant illness] that may have lead to cessation of BF or switching the child to NBF). Therefore, the outcome of illness or death is the cause of NBF instead of being the result of NBF. There is no general agreement on how to address this potential bias. In our study, some of the analytic approaches and comparisons that we performed suggest that the effect of this bias could be minimal. In the morbidity analyses, we examined the association of disease frequency and malnutrition with BF and NBF at 3-month intervals. The results at each interval were consistent after adjusting for other risk factors. Excluding from the analysis infants of women who switched during the interval from BF to NBF did not change the differences in morbidity rates. For the mortality analysis, we used an extended Kaplan–Meier analysis approach, which addressed the exposure to BF as a time-varying covariate when assessing its association with mortality. In the extended Kaplan–Meier curves, participants contributed to different curves, representing hypothetical cohorts at different times during follow-up [16, 24]. We compared the characteristics of women at each age interval whether they remained BF or NBF (Table 1). The characteristics of the BF and NBF women remained stable over the study duration, with almost no change between intervals, suggesting that there were no major changes in underlying characteristics of women or children at each age interval because of excessive movement or mixing between the groups of BF and NBF women. Of note, there were more women with low CD4 cell count (<250 cells/mm3) in the NBF group than in the BF group (Table 1). However, CD4 cell count was determined at enrollment. Other variables that we assessed at each visit, such as HIV disease stage, were not significantly different between the BF and NBF women. Misclassification of self-reported BF history and some of the reported morbidity events (eg, reported general illnesses and hospital admissions) are potential weaknesses of this study.
How to protect the infant from complications related to NBF and keep the child HIV free is a dilemma in the settings where replacement feeding is not safe. BF provides several protective factors, meets nutritional requirements, and is critical for survival in developing countries, especially during the first months of life, as was shown in a large multi-country study . More options are now available to prevent mother-to-child transmission of HIV, ranging from extended infant prophylaxis to use of maternal antiretrovirals for prevention [14, 26]. More-potent drugs and regimens that start during pregnancy can lower further rates of transmission to levels seen in developed countries, as was recently demonstrated in a clinical trial conducted in Botswana . Making maximum use of current successful interventions to prevent HIV infection in children should provide the opportunity to prevent non–HIV infection–related morbidity and mortality among children.
Weaning is associated with acute and late serious adverse outcomes among HIV-exposed, uninfected children. Therefore, prolonged BF to ≥12 months should be encouraged, consistent with the recent WHO 2009 recommendations . With introduction of maternal antiretrovirals for both prophylaxis and treatment, BF duration could be extended. The impact of these changes on child health needs to be evaluated. In settings similar to Malawi, where the background morbidity and mortality are high , inexpensive preventive strategies, such as TMP-SMX prophylaxis of HIV-exposed, uninfected children during and after weaning, should be considered.
We thank the mothers and children who participated in the PEPI-Malawi study and the nursing and technical staff in Malawi for their excellent collaboration throughout this study.
The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the US Centers for Disease Control and Prevention or the National Institutes of Health. Use of trade names is for identification purposes only and does not constitute endorsement by the US Centers for Disease Control and Prevention, National Institutes of Health, or the Department of Health and Human Services.
Financial support.This work was supported by a Cooperative Agreement from the Centers for Disease Control and Prevention and the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health (U50 PS022061-05; U50/CC0222061).
Potential conflicts of interest.All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed in the Acknowledgments section.