Randomized trial of vitamin supplements in relation to transmission of HIV-1 through breastfeeding and early child mortality

Background: HIV-1 transmission through breastfeeding is a global problem and has been associated with poor maternal micronutrient status.

Methods: A total of 1078 HIV-infected pregnant women from Tanzania were randomly assigned to vitamin A or multivitamins excluding A from approximately 20 weeks' gestation and throughout lactation.

Results: Multivitamins excluding A had no effect on the total risk of HIV-1 transmission (RR 1.04, 95% CI 0.82-1.32, P = 0.76). Vitamin A increased the risk of transmission (RR 1.38, 95% CI 1.09-1.76, P = 0.009). Multivitamins were associated with non-statistically significant reductions in transmission through breastfeeding, and mortality by 24 months among those alive and not infected at 6 weeks. Multivitamins significantly reduced breastfeeding transmission in infants of mothers with low baseline lymphocyte counts (RR 0.37; 95% CI 0.16-0.85, P = 0.02) compared with infants of mothers with higher counts (RR 0.99, 95% CI 0.68-1.45, P = 0.97; P-for-interaction 0.03). Multivitamins also protected against transmission among mothers with a high erythrocyte sedimentation rate (P-for-interaction 0.06), low hemoglobin (P-for-interaction 0.06), and low birthweight babies (P-for-interaction 0.04). Multivitamins reduced death and prolonged HIV-free survival significantly among children born to women with low maternal immunological or nutritional status. Vitamin A alone increased breastfeeding transmission but had no effect on mortality by 24 months.

Conclusion: Vitamin A increased the risk of HIV-1 transmission. Multivitamin (B, C, and E) supplementation of breastfeeding mothers reduced child mortality and HIV-1 transmission through breastfeeding among immunologically and nutritionally compromised women. The provision of these supplements to HIV-infected lactating women should be considered.

Introduction

HIV infection is a major public health problem in many developing countries, particularly in sub-Saharan Africa and south and southeast Asia [¹]. In many African settings, 15-30% of women attending prenatal care clinics are infected. Between 20 and 45% of children born to HIV-1-infected women become infected during the intrauterine, intrapartum, or breastfeeding periods. Micronutrient supplementation has been proposed as a low-cost intervention for reducing HIV-1 vertical transmission in developing countries where specific antiretroviral and prophylactic drugs are unavailable. In observational studies, inverse associations have been reported between the vitamin status and the risk of HIV disease progression, an advanced stage being a predictor of vertical transmission [²]. Results from some [^3-5], but not all [^6,7], observational studies suggested that low serum levels of vitamin A among HIV-infected pregnant women are associated with a higher risk of vertical transmission of HIV-1. Low plasma vitamin A concentrations in HIV-infected women during pregnancy were also associated with a higher risk of viral shedding in breast milk [⁸]. Given the observational design of such studies, residual confounding by the stage of HIV disease, immunological status, and other variables may provide an explanation for these apparent protective associations [²].

A randomized, double-blind, placebo-controlled trial was conducted to examine the effects of the supplementation of vitamin A or multivitamins (excluding vitamin A) on the vertical transmission of HIV-1 and other health outcomes among pregnant women infected with HIV-1 in Dar es Salaam, Tanzania. We reported previously that multivitamin supplements (excluding vitamin A) resulted in large and significant reductions in the risks of adverse pregnancy outcomes (fetal death, low birthweight, and severe prematurity) [⁹], but had no statistically significant effect on transmission through the intrauterine or intrapartum and early breastfeeding routes ascertained at birth and 6 weeks postpartum, respectively [¹⁰]. In this continued follow-up of the same population, we examined the efficacy of the supplements on transmission through breastfeeding after 6 weeks of age, and on child mortality in the first 2 years of life.

Methods

Starting in April 1995 and over the following 2 years, we enrolled pregnant women presenting to public antenatal clinics between 12 and 27 weeks' gestation who resided in Dar es Salaam and who intended to stay in the city until after delivery and for at least 1 year thereafter. Details of the study design have been published [^9,10]. In brief, 13% of the women tested positive for HIV-1. Eligible women were randomly assigned in a two-by-two factorial design to receive a daily oral dose of one of four regimens from enrollment and throughout the pregnancy and lactation periods: (i) vitamin A alone (30 mg β-carotene plus 5000 IU preformed vitamin A); (ii) multivitamins excluding vitamin A (20 mg B₁, 20 mg B₂, 25 mg B₆, 100 mg niacin, 50 μg B₁₂, 500 mg vitamin C, 30 mg vitamin E, and 0.8 mg folic acid); (iii) multivitamins including vitamin A in the same doses as above; or (iv) placebo. At delivery, women in groups one and three received an additional oral dose of vitamin A (200 000 IU), whereas women in groups two and four were given a placebo. Active tablets and placebo were identical in size and colour. As per standard prenatal care in Tanzania all women, irrespective of the assigned experimental regimen, were given iron and folate supplements, and prophylactic chloroquine; at 6 months of age, all infants received 100 000 IU vitamin A, and were given double that amount every 6 months thereafter.

In accordance with the World Health Organization and Tanzanian Ministry of Health guidelines, HIV-infected women were provided with information about the benefits and risks associated with various infant feeding options; the decision whether to breast feed or not was ultimately made by the mother. At each monthly visit, women were asked if the child was being breastfed and, if applicable, the age at which the child had stopped breastfeeding.

At baseline, laboratory tests were carried out for sexually transmitted disease, routine urine and stool laboratory examination, and complete blood count. Absolute counts of T cell subsets (CD4, CD8, and CD3) were measured using the FACScount system (Becton-Dickinson, San Jose, CA, USA). Details of the assessment of HIV status in infants have been published [¹⁰]. For diagnosis of HIV infection in infants, we attempted to collect whole blood samples at birth, at 6 weeks, and at 3 monthly intervals thereafter. A child was determined to be HIV-1-infected if either a peripheral blood mononuclear cell specimen tested positive using polymerase chain reaction (PCR) at any point, or a plasma specimen obtained at 18 months of age or older tested positive using enzyme-linked immunosorbent assay (ELISA; and was confirmed by a Western blot test). A child was determined to be uninfected if a blood sample obtained at the last visit had a negative HIV-1 PCR test result, or if a plasma specimen obtained at the last visit had a negative ELISA test result from a child at 18 months or older. Transmission through breastfeeding was defined as infection after 6 weeks of age among those who were not known to be infected at 6 weeks.

Data analyses

We examined the effect of the supplements on total HIV infection (i.e. through all three routes), death within the first 2 years, and the combined endpoints of HIV infection or death. These endpoints were examined among all livebirths as well as among all pregnancies, i.e. the latter included infection or death during the intrauterine period. We also examined the effect of the supplements on transmission through breastfeeding after 6 weeks of age, and mortality during the period between 6 weeks and 24 months of age.

We had previously reported that 1083 HIV-positive women were enrolled in the trial. We retested specimens from all randomly assigned women as part of a quality control procedure, and found that five women were HIV negative; a mechanism for counselling these five women was put in place immediately. Of the remaining 1078 HIV-infected pregnant women who were randomly assigned, three were eventually found not to be pregnant, six died before delivery, and pregnancy outcome was not known for 28. Overall, 985 children were born alive, of whom 898 had at least one specimen for HIV testing. No blood specimen was available for HIV testing at any point during the study for 86 children.

We used proportional hazard models [¹¹] to investigate the effects of the supplements on time to HIV infection, death, or the combined endpoint of time to HIV infection or death. Cumulative rates were determined by the Kaplan-Meier method [¹²]. Follow-up time for the transmission analysis was the lower of the time to cessation of breastfeeding and the midpoint between the last sample considered (which was the first positive sample or the last negative sample) and the sample before that. Mortality follow-up ended at 24 months of age or at the last time survival was ascertained, whichever was less. We examined whether the effect of one treatment (multivitamins or vitamin A) was modified by the other or by categories of birthweight (< 2500 g, ≥ 2500 g) and gestational age at birth (< 34 weeks, ≥ 34 weeks), as well as within categories of the following baseline maternal characteristics (dichotomization level): total lymphocyte counts (1340 cells/μl), CD4 cell counts (350 cells/μl), CD8 cell counts (565 cells/μl), hemoglobin concentration (8.5 g/dl), plasma vitamin A concentration (20 μg/dl), plasma vitamin E concentration (9.7 μg/ml), erythrocyte sedimentation rate (ESR) (81 mm/h), and mid-upper arm circumference (23 cm). Conventional cutpoints were used when available (CD4 cell count, hemoglobin, vitamin A); otherwise a priori selected cutoff points were used: the median (CD8 cell counts and vitamin E) or quartile values (lymphocyte count and ESR). Likelihood ratio tests were used to assess the statistical significance of the interactions. We present results in which P for interaction was ≤ 0.20 (except for CD4 cell count, which we present because of general interest). In further analyses, we examined the effect of the supplements on mortality within strata of HIV-1 infection at birth, 6 weeks, and defining HIV infection as a time-varying covariate. For the latter analyses, children whose last specimen was negative for HIV-1 and were subsequently lost to follow-up were considered to have remained negative for the next 3, 6, 12 or 24 months; the results were essentially the same in all four cases, and we present the findings assuming they remained negative for 24 months.

The sample size of the study was calculated for the study to have more than 90% power to detect a 30% effect on the risk of transmission, assuming a two-sided type I error of 0.05 and a background rate of transmission of 30%, and including a potential loss to follow up of 20%. All P values reported are two-sided; statistical significance in this study is defined as P < 0.05. A Data Safety and Monitoring Board reviewed study progress and the interim analyses of primary endpoints. In September 2000, the Data Safety and Monitoring Board recommended that the vitamin A arm of the study be dropped. Although all children of index pregnancies had stopped breastfeeding at that time, there was concern about subsequent pregnancies in women still randomly assigned to vitamin A.

The study protocol was approved by the Research and Publications Committee of Muhimbili University College of Health Sciences, the Ethical Committee of the National AIDS Control Program of the Tanzanian Ministry of Health, and the Institutional Review Board of the Harvard School of Public Health.

Results

Women assigned to the various treatment groups whose children had HIV testing were similar with respect to the baseline characteristics shown in Table 1 as well as other variables, including marital status, and spouse's age and education. Compliance was evaluated as the number of tablets absent from the returned bottles divided by the total number of tablets the individual should have taken. We previously reported that compliance was high (median ∼90%) before delivery and when examined up to 30 weeks' postpartum [¹⁰]. Median compliance was 89% at 12 months, 84% by 18 months, and 84% by 24 months. Breastfeeding was almost universal in this study population (> 99%).

Table 1

We obtained the final HIV status for 580 children: 268 were positive, 312 had a final negative status, defined as having had a negative PCR test at least 6 weeks after the cessation of breastfeeding or a negative ELISA test taken at 18 months of age or older and at least 6 months after the cessation of breastfeeding. The remaining 318 were HIV negative at the last specimen that was examined, although that sample did not meet the criteria for final HIV status defined above. Of these, 73 children had died before 24 months of age; for others, a specimen was unavailable because of travel out of the study area, and in some cases, because of the refusal of mothers to permit phlebotomy. The number of specimens obtained from children who had a non-final HIV-uninfected status was on average 2.67 specimens (median 3, SD 1.3), and the age at the last specimen tested was 41 weeks on average (median 26, SD 43). The availability of blood specimens over time did not differ by treatment arm. Of the 984 live births, survival status was known at 12 months for 925 (94%) and at 24 months for 912 (93%).

Multivitamins had no effect on the overall risk of transmission of HIV-1 [relative risk (RR) 1.04, 95% confidence interval (CI) 0.82-1.32; P = 0.76] (Table 2). Total child mortality by 24 months of age, including fetal deaths, was lower among the multivitamin group compared with those who did not receive multivitamins (RR 0.82, 95% CI 0.66-1.02; P = 0.08). Multivitamins had non-significant protective effects on death and HIV-free survival by 24 months.

Table 2

Among all livebirths, a significantly higher risk of HIV-1 transmission was observed in the vitamin A arm compared with the no vitamin A arm (RR 1.38, 95% CI 1.09-1.76; P = 0.009) (Table 2, Fig. 1). Of the children whose mothers did not receive vitamin A, 15.4% were infected by 6 weeks, compared with 21.2% of those whose mothers received vitamin A. At 6 months, the cumulative incidences were 22.4 and 28.1%, respectively, and at 24 months they were 33.8 and 42.4%, respectively. Vitamin A supplements had no effect on mortality or HIV-free survival. We did not observe evidence of an interaction between the effect of vitamin A and multivitamins for any of the endpoints examined.

Fig. 1

Overall, multivitamins had a small and non-significant effect on transmission through breastfeeding (RR 0.85, 95% CI 0.61-1.19; P = 0.34) (Table 3). The effect of multivitamins on mortality was somewhat stronger although not statistically significant (RR 0.78, 95% CI 0.50-1.21; P = 0.26). Among children born to women in the lowest quartile of total lymphocyte counts at enrollment, the RR of transmission was 0.37 (95% CI 0.16-0.85; P = 0.02) when comparing mothers who received multivitamins with those who did not receive multivitamins. The corresponding RR among mothers in the higher lymphocyte group was 0.99 (95% CI 0.68-1.45; P = 0.97, P for interaction 0.03). Protection against transmission was also observed among children born to mothers with high ESR (P for interaction 0.06), or low hemoglobin (P for interaction 0.06), and among low birthweight babies (P for interaction 0.04).

Table 3

Children born to women with low maternal lymphocyte counts and low plasma vitamin E levels at baseline had better chances of survival up to 24 months if the mothers received multivitamins. Therefore, among women in the lowest quartile of lymphocyte counts, the RR of child mortality was 0.30 (95% CI 0.10-0.92; P = 0.04); among women with higher lymphocyte counts multivitamins were without an effect on mortality (RR 0.96, 95% CI 0.58-1.60; P = 0.89, P for interaction 0.05). Multivitamins also reduced mortality among children born to mothers with low plasma vitamin A (P for interaction 0.03) or low vitamin E levels at baseline (P for interaction 0.008). Similar benefits of multivitamins were observed for the HIV-free child survival outcome among women who had low baseline CD8 cell counts (P for interaction 0.09), low baseline lymphocyte counts (P for interaction 0.01), low baseline hemoglobin (P for interaction 0.11), or low baseline vitamin E levels (P for interaction 0.12).

Multivitamins were associated with a reduction in mortality among children who were HIV negative at birth (RR 0.76, 95% CI 0.54-1.07; P = 0.11), but not among children who were HIV infected at birth (RR 0.90, 95% CI 0.49-1.67; P = 0.74) (Table 4). When using HIV status as a time varying covariate through the period of follow-up, multivitamins were associated with reduction in mortality after 6 weeks of age among HIV-uninfected children (RR 0.59, 95% CI 0.33-1.05; P = 0.07), but had no effect among HIV-infected children (RR 0.93, 95% CI 0.65-1.33; P = 0.68).

Table 4

Vitamin A had no effect on mortality by 24 months (RR 1.03, 95% CI 0.66-1.62; P = 0.89), but increased breastfeeding transmission (RR 1.33, 95% CI 0.95-1.86; P = 0.10) and the risk of HIV infection or death (RR 1.27, 95% CI 0.93-1.72; P = 0.14). These effects of vitamin A were not modified by any of the characteristics examined (P for interaction ≥ 0.15 in all cases). There was no effect of vitamin A on mortality within groups of children stratified by HIV status at birth or at 6 weeks (Table 4).

Discussion

We observed that multivitamins (vitamins B, C, and E) had a modest and not statistically significant reduction in the overall risks of HIV infection through breastfeeding, mortality by 24 months, or the combined risks of infection and mortality among children. Children born to women who were in relatively poorer nutritional or immunological conditions at baseline benefited significantly with respect to these endpoints. Maternal multivitamin supplementation also reduced child mortality among HIV-uninfected children.

To our knowledge, this is the first randomized study to examine the effect of maternal multi-micronutrient supplements on infant health. Multivitamins could affect breastfeeding transmission of HIV through a number of mechanisms. Multivitamin supplementation of breastfeeding mothers may enhance the systemic immune function of the mother, reducing the rate of clinical, immunological, or virological progression [²]. In longitudinal studies among homosexual and bisexual men [^13-18], higher intakes of micronutrients (B vitamins, C and E) and lower plasma levels of vitamins B₁₂ and E were associated with a slower progression to AIDS, and a lower risk of mortality. In a study from Canada [¹⁹], large daily doses of antioxidant vitamins C and E resulted in a significant reduction in viral load. The above studies were carried out among adults in developed countries, and are limited by their observational design. In the same trial [⁹], we previously reported that multivitamin supplements substantially increased levels of T cell subsets; however, the long-term clinical relevance of this finding has yet to be determined.

Antioxidant vitamin status may be associated with a reduced risk of mastitis, an inflammation of breast tissues that is associated with a higher viral load in breast milk and a greater risk of vertical transmission [^20-25]. The provision of multivitamins (including B vitamins, and vitamins C and E) to mothers in our study population may also have resulted in improved milk quality, including the concentrations of micronutrients and immunological properties [²³]. Improved micronutrient intake by babies may enhance their developing immune system, and hence reduce their risk of morbidity and mortality [^24-28].

Multivitamins (not including vitamin A) were particularly beneficial for children of women with proxies for advanced HIV disease (low lymphocyte count, severe anemia, and high ESR) or poor nutritional status (low plasma vitamins A or E) and for low birthweight babies. Low plasma levels of vitamin A may not necessarily mean vitamin A deficiency. Levels could be reduced as part of the acute phase response to infection even in the presence of adequate liver stores of vitamin A, and could be a marker of HIV disease stage [²].

We observed more cases of HIV-1 infection among children of mothers in the vitamin A arm compared with those who did not receive vitamin A. The increased occurrence of infection in the vitamin A group was unexpected. Data from two other randomized, placebo-controlled trials from Malawi [²⁹] and South Africa [³⁰] found no effect of vitamin A on transmission. There are, however, differences between our trial and the latter two trials. In both trials, the supplements were given during the antenatal period only, whereas in the Tanzania trial supplementation continued during the antenatal and breastfeeding periods. Although in South Africa, vitamin A supplements included both preformed vitamin A (5000 IU) and β-carotene (30 mg) in doses identical to ours, the Malawi trial supplement contained 10 000 IU of vitamin A. Results from earlier observational studies suggested that plasma vitamin A concentrations were inversely associated with the risk of transmission [^3-5], although other studies of similar design found no relationship [^6,7]. In-vitro studies also provided conflicting results regarding the possible effects of vitamin A. Vitamin A was associated with increased [^31,32] or decreased [^33-35] replication of HIV-1. Understanding the effect of vitamin A supplements on HIV viral load in human studies should be informative. This question was examined in three randomized, placebo-controlled studies from the United States [^36,37] and South Africa [³⁸]. A single large dose of 200 000 IU of vitamin A was not associated with a change in HIV-1 viral load in the two studies from the United States [^36,37]. In the South African trial [³⁸], there was no effect on the viral load of daily doses of vitamin A and β-carotene supplements in doses identical to ours, which were provided to pregnant women between 28 and 32 weeks' of gestation and delivery. Whereas these findings provide some reassurance about the safety of vitamin A supplements, the doses and periods of follow-up in the latter trials were not similar to the Tanzania trial. We intend to examine the efficacy of supplements on viral load in the Tanzania trial.

It is not possible to examine whether the observed adverse effect was caused by preformed vitamin A or the β-carotene parts of the supplement. High doses of β-carotene under highly oxidative conditions may result in eccentric cleavage oxidation breakdown products. These products may have adverse effects by interfering with retinoic acid availability at the tissue level [³⁹]. However, one cannot exclude the possibility that preformed vitamin A contributed to the observed increased transmission. Results from two prospective observational studies raised concerns about the potential adverse effects of vitamin A in HIV infection. In a study among HIV-1-negative Kenyan men, lower plasma vitamin A levels were associated with a decreased risk of HIV-1 seroconversion [⁴⁰]. In another prospective study among HIV-infected men in the United States, a U-shaped relationship was observed between dietary vitamin A intake and the risks of progression to AIDS [²³] and mortality [²⁴], suggesting that men with higher or lower levels of intake were at a higher risk compared with men who consumed moderate amounts of vitamin A. It was suggested that vitamin A supplements may increase the risk of transmission of HIV-1 by enhancing the differentiation of myeloid and lymphoid cells, which is associated with an increased expression of CCR5, receptors that increase the susceptibility to HIV-1 infection [⁴⁰].

An alternative explanation to the findings of vitamin A supplements and HIV transmission may be that the supplements prolonged the survival of HIV-uninfected children who were then exposed to HIV-infected breast milk for a longer period. We did observe a small and non-significant decrease in child mortality risk in the vitamin A arm, particularly among HIV-uninfected children. In a meta-analysis of nine large community-based trials [⁴¹], the periodic supplementation of children resulted in approximately 23% reduction in all-cause mortality. These findings were confirmed in another recent study trial from Tanzania [⁴²]. The relatively modest effect of maternal vitamin A supplements on child survival observed in the current trial may be attributable to the fact that all children received vitamin A at 6 months of age and at 6-monthly intervals thereafter, as per the standard of care, irrespective of whether their mothers were in the vitamin A or no vitamin A group.

Vitamin A supplements had no effect on pregnancy outcomes, or on maternal CD4 cell counts or hemoglobin levels during pregnancy in the same study population [⁹]. Vitamin A supplements may be beneficial to HIV-uninfected pregnant women, as has been shown by the significant reduction in maternal mortality among women from Nepal [⁴³]. However, the possible increased risk of HIV-1 infection we reported may mean that individuals offered vitamin A might also need to receive HIV-1 counselling and testing. This may not be cost-effective to pursue unless such a counselling program is already in place as part of a strategy to reduce mother-to-child transmission. Our findings should not raise concerns about ongoing child programs that provide vitamin A supplements starting at 6 months of age, particularly in settings in which early weaning at 4-6 months is encouraged to minimize transmission through breastfeeding.

Conclusion

Although it is difficult to base policy on the results of a single study, our data give little encouragement for the widespread use of vitamin A supplements in HIV-positive pregnant women. The possibility that maternal supplements of vitamin A increase the risk of transmission of HIV requires further investigation. In light of the protective effects of multivitamins other than vitamin A observed on fetal loss, low birthweight, and severe prematurity [⁹], multivitamin supplementation of HIV-positive pregnant women is more widely practised in antenatal programs in a number of countries. The findings in this paper suggest that the supplementation of breastfeeding mothers with multivitamins may also be beneficial among immunologically and nutritionally compromised women. It may be difficult to target such an intervention only to women who are in these categories, as the added burden of screening for these characteristics will not be practical. Given the observed protective effects in reducing child mortality and the transmission of HIV-1 through breastfeeding, it is important to consider the provision of these supplements to all HIV-positive women during the lactation period.

Acknowledgements

The authors are grateful to Hoffmann-La Roche for donating the raw material that was used for preparing the vitamin and placebo tablets. They would also like to thank the mothers and children, field teams including nurses, midwives, supervisors, laboratory staff, and administrative staff who made the study possible. The authors greatly appreciate input from the following colleagues: Illuminata Ballonzi, Josephine Ballati, Beth Chaplin, Jenny Coley, Ellen Hertzmark, Heavengton Mshiu, Willy Urassa, Eduardo Villamor, Walter Willett and all other members of the Harvard-Tanzania collaboration; members of the Data Safety and Monitoring Board: Graham Colditz, Kenneth McIntosh, Fatma Mrisho, Marcello Pagano (Chair), and Richard Platt; Dr Anne Willoughby at the National Institutes of Child Health and Human Development and Dr Ken Bridbord at Fogarty International Center of the National Institutes of Health for their valuable support. They also thank Muhimbili Medical Centre, Muhimbili University College of Health Sciences, and the National AIDS Control Program in Dar es Salaam for their institutional support.

Sponsorship: This work was supported in part by the National Institute of Child Health and Human Development (NICHD R01 32257), and the Fogarty International Center (NIH D43 TW00004).