Observational studies have suggested that low serum vitamin levels are associated with increased mother-to-child transmission (MTCT) of HIV and increased preterm delivery. We aimed to determine the efficacy of vitamins on the prevention of MTCT and preterm delivery by systematically reviewing the available randomized controlled trials [RCTs]. We conducted systematic searches of 7 electronic databases. We extracted data from the RCTs independently, in duplicate.
We included 4 trials in our review. Of the three trials on Vitamin A, two suggested no difference in MTCT, while the third and largest trial (n = 1078) suggested an increased risk of MTCT (Relative Risk 1.35, 95% Confidence Interval [CI], 1.11–1.66, P = 0.009). Two of the vitamin A trials addressed the impact of supplementation on pre-term delivery; one suggested a benefit (RR 0.65, 95% CI, 0.44–0.94) and the other no difference. All three vitamin A trials found no significant effect on infant mortality at 1 year. Of the two trials that looked at multivitamin use, only one addressed the prevention of MTCT, and found a non-significant RR of 1.04 (95% CI, 0.82–1.32). Two of the multivitamin trials found no significant effects on pre-term delivery. The single multivitamin trial examining children's mortality at 1 year yielded a non-significant RR of 0.91 (95% CI, 0.17–1.17).
Randomized trials of vitamins to prevent MTCT have yielded conflicting results without strong evidence of benefit and have failed to exclude the possibility of harm.
Keywords:HIV; Vitamins; Vitamin A; Mother-to-child transmission; Preterm delivery
In Africa, 55% of HIV-1-positive adults are women, most of childbearing age . Data from antenatal clinics show that in several parts of southern Africa, more than 30% of pregnant women are infected with HIV-1. The fastest growth has been in South Africa, where the prevalence of infection in adults increased from 5% in 1990, to over 25% in 2002 . Mother-to-child transmission (MTCT) of HIV-1 can occur during pregnancy, delivery, and post-partum through breastfeeding. In observational cohort studies, the cumulative rates of transmission are between 25% and 45% of all children born to HIV-1-infected mothers in Africa compared with 10–30% in wealthier countries . This difference is greatly but not totally accounted for by the risk of postnatal transmission in populations in which breastfeeding is common.
MTCT is responsible for 5–10% of the total of new HIV infections in many developing countries, with more than 500,000 children being infected each year . In many industrialized countries, the introduction of antiretroviral (ARV) drugs for the prevention of MTCT has dramatically reduced rates of transmission among non-breastfeeding mothers. Improvement is evident as more women enter pregnancy while on combination ARV therapy [2,3]. The limited access to ARV's throughout Africa has, however, led to a search for cheaper alternatives. Observational studies demonstrating an association between low biochemical and dietary levels of micronutrients and MTCT have fueled the hypothesis that micro-nutrient supplementation, particularly with Vitamin A and multivitamin combinations, may reduce vertical transmission [4-9].
Vitamin supplementation may reduce vertical transmission through either intrapartum or breastfeeding routes by reducing HIV viral load in lower genital tract secretions and in breast milk, respectively . Other potential therapeutic mechanisms include improved placental and lower genital tract integrity , and improved fetal and newborn gastrointestinal immunity . Investigators have undertaken several randomized trials addressing the impact of vitamin supplementation on MTCT. In order to determine the effectiveness of these treatments in preventing MTCT and pre-term delivery, we conducted a systematic review of these randomized trials. In addition, we addressed the effect of Vitamin A and multivitamins on childhood mortality.
With the aid of an information specialist, we (EM, PW) performed a systematic, all language search of the following electronic databases independently, in duplicate: MedLine (1966- January 2005), AMED (1985- January 2005), AltHealthWatch (1990- January 2005), CinAhl (1982- January 2005), Embase (1980- January 2005), and the Cochrane Library (2004, issue 2). We supplemented this search by reviewing reference sections of relevant articles, and by searching for unpublished trials on the National Research Register (UK) (October 1998- January 2005) and Clinicaltrials.gov (February 2000- January 2005).
Selection of abstracts
Two of us (EM, PW) independently evaluated the abstracts of retrieved articles. Eligible studies met the following criteria: (1) were original randomized controlled trials examining HIV+ patients using either Vitamin A or multivitamin treatment during pregnancy; (2) examined the outcomes of MTCT or pre-term delivery. We excluded any previous analyses of the same trial in our meta-analysis and used the most recent data available . Kappa scores reflected chance-adjusted inter-observer agreement in the study identification process.
Pre-specified quality criteria included: methods of randomization, allocation concealment, blinding status of patients and assessors, use of placebo, informed consent, a priori sample size estimations, use of intention-to-treat, and sources of funding. In addition, we contacted the study authors for clarification of study methods.
Assessing the quality of trials included in a systematic review is important in determining trial validity, potential for introducing bias and heterogeneity and exploring subgroup analysis. Quality assessment was performed independently, in duplicate (EM, PW). Quality assessment items were used as a priori explanations of heterogeneity.
We extracted data independently, in duplicate (EM, PW) . Data abstraction sheets were developed and piloted amongst the group (EM, PW, DS) to determine outcomes of interest and reproducibility.
We determined agreement between reviewers using the kappa statistic. We report on study sample size and dosing using descriptive data. Outcomes measured were the number of live births, not number of pregnancies. Our primary endpoint for meta-analysis of MTCT was children's infection status at the latest time point reported. In order to provide a best-estimate of treatment effects, we conducted a meta-analysis. Our primary endpoint for the meta-analysis of pre-term delivery was pre-term delivery defined as <37 weeks. We also determined childhood mortality at 1 year where reported. All outcomes were treated as dichotomous outcomes  and the appropriate relative risks (RR) and applicable 95% confidence intervals [CI] were determined. We calculated RR from raw data, when provided. Pooled analysis of relative risk was conducted using a random effects model. We tested for heterogeneity using the Zalen test and the I2 test . A priori explanations of heterogeneity included quality assessment items, design, and length of follow-up. StatsDirect was used for all meta-analytic procedures (StatsDirect, Copyright 1993–2004, Manchester).
Figure 1 displays the yield of our systematic searches. Of 27 clinical trial abstracts that appeared relevant, we examined 16 full text articles. Four trials met inclusion criteria and are included in this systematic review. κ for initial decisions on the inclusion of studies was 0.9, suggesting near-perfect agreement. Three [17-19] studies met our inclusion criteria of examining mother-to-child transmission of HIV and three studies [13,17,20] met our inclusion for pre-term delivery. Three studies examined the role of vitamin A for prevention of MTCT [17,18] and 1 study  examined the role of both vitamin A and a multivitamin using a 2 × 2 factorial design. We identified 1 unpublished and unreported study , from which however no results could be obtained.
Figure 1. Flowchart depicting study selection and inclusion/exclusion.
Details for each of the trials can be found in table 1 (additional file 1) with regards to: the intervention; standards of care for all participants; the number of mothers randomized and the gestational period in which they were enrolled; the number of live births; compliance; outcomes measured; and results. The table is split into 2 sections to reflect details with reference to vitamin A supplementation alone or a combination of multivitamins. A brief overview of the studies found is provided here.
Fawzi et al published, in several analyses of the same factorial trial assessing the impact of vitamin A, and of multivitamins excluding Vitamin A, on vertical transmission of HIV-1 and child mortality on pregnant women in Dar es Salaam, Tanzania (n = 1078) . HIV+ women presenting at antenatal clinics between 12 and 27 weeks of pregnancy were randomized to receive (i) vitamin A alone or matching placebo and, (ii) multivitamins excluding vitamin A or matching placebo. 985 children were born alive from the total sample with 898 having at least one specimen for HIV testing. Of these, 268 tested positive for HIV-1 at 6 weeks of age. Details for the earlier analyses in 1998 are also provided in Table 1 in the section on vitamin A.
In a parallel group randomized trial in Durban, South Africa, Coutsoudis et al randomized 728 pregnant HIV infected women to either placebo (n = 360) or vitamin A retinyl palmitate + B-carotene, with additional vitamin A at delivery(n = 368) . Data on HIV infection at 3 months were available for 502 children of the total 661 live births.
Another parallel group trial by Kumwenda et al randomized 697 HIV infected pregnant women in Blantyre, Malawi to an intervention where vitamin A was added to their supplements (n = 340), or placebo (n = 357) . There were a total of 622 live births (including 14 pairs of twins), however, 32 infants died to prior to 6 weeks of age, making HIV status undeterminable.
As described above, the factorial trial of Fawzi et al (2002)  performed examined the impact of both vitamin A and multivitamins of MTCT on infant mortality. The characteristics and results from the earlier analysis by Fawzi et al (1998) are also listed in table 1 . In a subgroup analysis, not listed in table 1, multivitamin supplementation reduced death and prolonged HIV-free survival in women with low maternal immunological and nutritional status (RR of death 0.30, 95% CI, 0.10–0.92).
Friis et al (2004)  conducted a parallel randomized trial of micronutrients versus placebo. They examined a subgroup of pregnant women with HIV infection (n = 360) enrolled between the period of 22 and 36 weeks gestation (active group n = 189, control group n = 171). The study was hampered by not examining infant HIV infections or reporting specific number of births by HIV group.
Three studies described sequence generation [13,18,20] Two reported allocation concealment [18,20]. Only 1 study described who was blinded . Four studies reported obtaining informed consent [13,17,18,20]. Five studies reported an a priori sample size estimation [13,18-20,22] and 4 reported analysis by intention-to-treat [13,17,19,20]. All studies disclosed the sources of funding.
The combined RR of vitamin A for prevention of MTCT yielded a RR of 1.05 (95% CI, 0.78–1.41, p = 0.2, I2 = 75%, heterogeneity P = 0.01) (figure 2). The impressive variability in results is reflected in the largely non-overlapping confidence intervals between the two studies that suggested no difference between treatment and control, and the Fawzi study that suggested harm. Two trials examined the protection of vitamin A for pre-term delivery and yielded a non-significant pooled RR of 0.85 (95% CI, 0.53–1.37, P = 0.5, I2 = 77%, heterogeneity P = 0.03) (figure 3). Three trials examined the role of maternal vitamin A supplementation on children's mortality at 1 year. The pooled RR was 1.05 (95%CI, 0.88–1.27, P = 0.5, I2 = 0%, heterogeneity P = 0.8).
This single trial by Fawzi et al. examining a multivitamin for prevention of MTCT yielded a non-significant RR of 1.04 (95% CI, 0.82–1.32) (figure 2). The single trial examining maternal multivitamin intake on children's mortality at 1 year yielded a non-significant RR of 0.91 (95% CI, 0.17–1.17). Two trials examined the role of multivitamins for prevention of pre-term delivery. The combined RR yielded a non-significant RR of 0.88 (95% CI, 0.73–1.06, P = 0.1, I2 = 0%, heterogeneity P = 0.8) (figure 3).