点击显示 收起
1 From the Cancer Prevention Fellowship Program, Division of Cancer Prevention, National Cancer Institute, Bethesda, MD (MFM); the Department of International Health, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD (JHH); Paediatrics and Child Health (PJI) and the Division of Nutrition, Institute of Food, Nutrition and Family Sciences (LCM), University of Zimbabwe, Harare, Zimbabwe; and the Division of Nutritional Sciences, Cornell University, Ithaca, NY (RJS and NVM)
2 Supported by the Canadian International Development Agency (R/C Project 690/M3688), the US Agency for International Development (cooperative agreement number HRN-A-00-97-00015-00 between Johns Hopkins University and the Office of Health and Nutrition of USAID), and a grant from the Bill and Melinda Gates Foundation, Seattle WA (all to ZVITAMBO). Additional support was received from the Nestlé Foundation, the Rockefeller Foundation, and BASF. 3 Reprints not available. Address correspondence to MF Miller, 6116 Executive Boulevard, Suite 404, Bethesda, MD 20892-8336. E-mail: millermeli{at}mail.nih.gov.
ABSTRACT
Background: Anemia is prevalent in infants in developing countries. Its etiology is multifactorial and includes vitamin A deficiency.
Objective: Our primary aim was to measure the effect of maternal or neonatal vitamin A supplementation (or both) on hemoglobin and anemia in Zimbabwean infants. Our secondary aim was to identify the underlying causes of postnatal anemia.
Design: A randomized, placebo-controlled trial was conducted in 14 110 mothers and their infants; 2854 infants were randomly selected for the anemia substudy, of whom 1592 were successfully observed for 8–14 mo and formed the study sample. Infants were randomly assigned within 96 h of delivery to 1 of 4 treatment groups: mothers and infants received vitamin A; mothers received vitamin A and infants received placebo; mothers received placebo and infants received vitamin A; and mothers and infants received placebo. The vitamin A doses were 400 000 and 50 000 IU in the mothers and infants, respectively.
Results: Vitamin A supplementation had no effect on hemoglobin or anemia (hemoglobin <105 g/L) in unadjusted or adjusted analyses. Infant HIV infection independently increased anemia risk >6-fold. Additional predictors of anemia in HIV-negative and -positive infants were male sex and lower total body iron at birth. In addition, in HIV-positive infants, the risk of anemia increased with early infection, low maternal CD4+ lymphocyte count at recruitment, and frequent morbidity. Six-month plasma ferritin concentrations <12 µg/L were a risk factor in HIV-negative but not in HIV-positive infants. Maternal HIV infection alone did not cause anemia.
Conclusion: Prevention of infantile anemia should include efforts to increase the birth endowment of iron and prevent HIV infection.
Key Words: Vitamin A • infants • hemoglobin • HIV • supplementation • Zimbabwe
INTRODUCTION
Anemia is prevalent in infants and young children in developing countries (1). Vitamin A is known to play a role in hematopoiesis, and anemia is a common consequence of vitamin A deficiency (2). Infants are born with essentially no vitamin A stores, and they are dependent on dietary sources of vitamin A. Breast milk is a good source of bioavailable vitamin A, provided the mother is not vitamin A deficient. In developing countries, nearly 20% of postnatal mothers have vitamin A deficiency and inadequate concentrations of breast-milk vitamin A (3), and breastfeeding infants of deficient mothers are at risk of vitamin A deficiency (4, 5).
We conducted a longitudinal cohort study nested within the Zimbabwe Vitamin A for Mothers and Babies Project (ZVITAMBO), a randomized, double-blind, placebo-controlled clinical trial in 14 110 mother-infant pairs enrolled between November 1997 and January 2000 that tested the efficacy of immediate postpartum maternal (400 000 IU) or neonatal (50 000 IU) (or both) vitamin A supplementation (VAS) on several infant and maternal health outcomes (6–8). The World Health Organization (WHO) categorizes Zimbabwe as being at high risk for vitamin A deficiency (9), and, at the time of the study, there were no national VAS programs for women or neonates in Zimbabwe.
We hypothesized that vitamin A deficiency was part of the etiology of anemia in these infants and that VAS (especially when given to the mother) would improve hemoglobin concentration during the second half of the first year of life. We followed 320 infants born to HIV-negative mothers (negative/negative) and 1272 infants born to HIV-positive mothers, of whom 210 became infected during follow-up (positive/positive) and 999 did not (positive/negative).
The etiology of anemia is multifactorial, which in part explains the limited success of public health programs in combating anemia, especially in infants. Thus, in a secondary analysis, we examined the effect of the following factors on anemia at 1 y of age: maternal and infant HIV status, sex of the infant, iron status including total body iron (TBI) at birth, infant-feeding pattern, acute and frequent morbidity, growth, and maternal hemoglobin concentration. We also examined the CD4+ lymphocyte count at recruitment in HIV-positive mothers and the timing of HIV infection in HIV-positive infants.
SUBJECTS AND METHODS
Study population
The study was carried out in Harare, the capital of Zimbabwe. Harare has a temperate climate with cool, dry, sunny winters and warm, wet summers. The common diet during the time of the study was maize meal porridge served with vegetables, peanuts, and other legumes and sometimes meat. The prevalence of HIV in adult Zimbabweans was 25.8% at the time of the study (10). Malaria is not endemic in Harare because of the high altitude (1500 m), and hookworm infection is rare throughout Zimbabwe.
Recruitment
The ZVITAMBO trial recruited mothers and their infants within 96 h of delivery at maternity clinics and hospitals in Harare. Pairs were eligible if neither the mother nor the infant had an acutely life-threatening condition and if the infant was a singleton with a birth weight of >1500 g. At baseline, the HIV status of mothers was ascertained by 2 separate enzyme-linked immunosorbent assay kits (HIV 1.0.2 ICE; Murex Diagnostics, Edenvale, South Africa; and GeneScreen HIV; Sanofi Diagnostics Pasteur, Johannesburg, South Africa) run in parallel. Duplicate pairs of discordant enzyme-linked immunosorbent assay results were resolved by using Western blot assay (HIV Blot 2.2; Genelabs Diagnostics SA, Geneva, Switzerland) and interpreted by using the manufacturer's guidelines. Mothers who tested negative at baseline were retested for HIV at 6 and 12 mo after delivery. Quality control was monitored by using the controls in the kits, inclusion of an internal quality-control sample on every plate, and participation in the Zimbabwe Ministry of Health's quality-assurance program for HIV testing. Of the mothers, 4496 (32%) were HIV positive at baseline.
Intervention
Mother-infant pairs were randomly assigned to 1 of 4 treatment groups: mothers and infants received vitamin A (Aa); mothers received vitamin A and infants received placebo (Ap); mothers received placebo and infants received vitamin A (Pa); and mothers and infants received placebo (Pp). The dose of vitamin A (as retinyl palmitate) in the mothers was 400 000 IU and that in the infants was 50 000 IU. Treatment and placebo capsules appeared identical, and both contained a soy-oil base with vitamin E as a preservative (50 IU/maternal capsule; 10 IU/infant capsule; Tishcon Corp, Westbury, NY). Treatment assignment was determined by using a computer random-number generator, and treatment assignment was concealed by placing study supplements in sequentially numbered series assigned to subjects' identification numbers. Neither participants nor nurses who administered the capsules or assessed outcomes were aware of treatment group assignment. Lists linking the study number to the treatment were kept in sealed envelopes and encrypted computer files that were not accessible to the Zimbabwe-based study team. Further details on the aims, study design, and recruitment procedures were published previously (6–8).
Written informed consent was obtained from all of the mothers. The ZVITAMBO trial protocol was approved by the Medical Research Council of Zimbabwe, the Committee on Human Research of The Johns Hopkins University Bloomberg School of Public Health, the Medicines Control Authority of Zimbabwe, and the Montreal General Hospital Research Ethics Committee. The anemia substudy was further approved by the first 2 of these agencies.
Study design
For this cohort study, we randomly selected 34% of the infants (7 of every 8 born to HIV-positive mothers and 1 of every 10 born to HIV-negative mothers) enrolled in the ZVITAMBO Study between October 1998 and January 2000 (n = 8301) to form a subsample, as shown in Figure 1. A total of 2319 and 535 infants born to HIV-positive and -negative mothers, respectively, were enrolled into the anemia substudy. Infants were excluded who received a neonatal blood transfusion (n = 5), were born to mothers with indeterminate HIV status at baseline (n = 5), or were missing other baseline information (n = 13). Among those included (n = 2831), 705 had been randomly assigned to group Pp, 708 to group Pa, 721 to group Ap, and 697 to group Aa. In a study clinic or in the home, study midwives followed the others and their infants at 6 wk, 3 mo, and every 3 mo until infant age of 12 mo. At baseline and each follow-up visit, study midwives conducted interviews with the mother and collected maternal blood by venipuncture and infant blood by venipuncture or heelstick (capillary). Acutely ill patients were referred to a study physician or directly to the hospital.
FIGURE 1.. Flow of participants throughout the trial. NA, not available; Hb, hemoglobin.
Anthropometric measurements
Birth weight was measured to the nearest 5 g by using an electronic scale (model 727; Seca Inc, Hanover, MD) in the labor unit of the hospital and transcribed from medical records. The same type of scale was used for follow-up weights measured in the clinic, but measurement of weights on home visits used a Salter scale (Salter, Fairfield, NJ) that was precise to the nearest 100 g. Length was measured to the nearest millimeter by using a stadiometer (Seca or Shorr Industries, Columbia, MD) in the clinic and to the nearest centimeter by using a "Measure Mat" (Seca) on home visits. Weight and linear growth velocity were calculated by using the difference between the current weight or length and the birth measurement.
Laboratory assays
Blood samples from infants were collected into clotting tubes and EDTA-containing tubes. Hemoglobin was measured on the same day by using the HemoCue hemoglobinometer (HemoCue, Mission Viejo, CA), and daily quality control of the instrument was performed. Hemoglobin data were unavailable if the infant missed the study visit or if the blood sample was insufficient as a result of the mother's refusal for her infant to give blood, inadequate sample volume, sample clotting, or failure of the laboratory to measure hemoglobin before the end of the day on which it was drawn. Infant HIV status was ascertained by using a prototype qualitative DNA polymerase chain reaction assay (Roche Molecular Systems Inc, Branchburg, NJ; 11). We measured ferritin concentration in plasma (or serum when plasma was unavailable) by using an enzyme immunoassay (Ramco Laboratories Inc, Houston, TX). The interassay CV was 9.6% and 9.2% at 70 µg/L and 300 µg/L, respectively, and the intraassay CV was 4.0% and 4.7%, respectively.
Infant feeding
A detailed infant-feeding history was elicited at each visit. Mothers were asked if and how often they were breastfeeding their infant. If the infant was not being breastfed, the infant's age at weaning and the reason for weaning were recorded. A list of 22 liquid and solid foods commonly fed to young Zimbabwean infants was created. At all visits through 6 mo, the mother was asked if each of these foods had ever been fed to her infant. Feeding of any other, unspecified foods was also elicited. Infants were classified as exclusively breastfeeding (EBF) if they had been fed nothing but breast milk and Western medicines. This data-collection method is similar to and the EBF definition is consistent with recent WHO recommendations (12). Iron-rich foods included in the questionnaire were eggs, fish, and poultry (listed as a single item); red meat; and infant formula, which we assumed was iron-fortified.
Infant morbidity
A 7-d morbidity history was elicited at all follow-up visits, and a morbidity score was calculated as the total number of visits at which the infant reportedly had diarrhea, fever, or cough divided by the total number of follow-up visits completed by the pair. The chronic morbidity score was dichotomized as < or 25% on the basis of relations observed in exploratory analyses. When the morbidity score was included in multivariate models, we adjusted for its denominator (ie, the total number of follow-up visits completed).
Iron status
Iron status was assessed by TBI at birth and the plasma ferritin concentration at 6 mo. Depleted storage iron was defined as a plasma ferritin concentration < 12 µg/L (13). We calculated TBI at birth as the sum of 2 components, hemoglobin iron (HbI) and body storage iron (BSI), according to a method used previously in infants (14–21).
HbI was calculated by using the following equation:
RESULTS
Study participants
We compared baseline characteristics of treatment groups stratified by maternal HIV status in Table 1. The number of infants born to HIV-positive mothers (n = 2298) reflects the oversampling of this group in our study design; those infants represent 81.1% of the total sample. There were no significant differences between treatment groups with respect to any of the maternal or infant characteristics at recruitment except an imbalance in the sex ratio of infants born to HIV-negative (P = 0.010) and -positive (P = 0.053) mothers, and we adjusted for the sex of the infant in subsequent analyses of the effect of VAS. Of infants born to HIV-positive mothers, 343 (20.4%) were positive for HIV on polymerase chain reaction at recruitment or by 6 wk of age, and another 115 infants (6.8%) became positive on polymerase chain reaction between 6 wk and the completion of follow-up. Hemoglobin concentrations at the 9- or 12-mo visit were available for 1592 infants (56.2% of total sample; see Figure 1). The median age at the completion of follow-up was 9 mo (range: 8–14 mo). The follow-up rate did not differ across treatment groups (P = 0.36 for the 3-df chi-square test).
View this table:
TABLE 1. Baseline characteristics of Zimbabwean infants according to treatment group and maternal baseline HIV status1
Effect of VAS
Overall in infants, the mean (±SD) hemoglobin concentration was 104.0 ± 17.5 g/L, and 46.3% of infants were anemic (hemoglobin <105 g/L). Neither maternal nor infant VAS administered in the immediate postpartum period had a significant effect on hemoglobin concentration or on the proportion of anemia after adjustment for the sex of the infant (Figure 2). The HIV status of the mother or infant did not modify the association between VAS and hemoglobin concentration or anemia—ie, there was no maternal or infant HIV status x VAS interaction.
FIGURE 2.. Effect of maternal and neonatal vitamin A supplementation on the risk of anemia (hemoglobin < 105 g/L) and hemoglobin concentration in Zimbabwean infants at 8–14 mo of age (n = 1592). Pp, mother and infant received placebo; Pa, mother received placebo and infant received vitamin A; Ap, mother received vitamin A and infant received placebo; Aa, mother and infant received vitamin A. Odds ratios, proportions, and means were adjusted for infant sex by using logistic regression analysis to estimate risk of anemia and linear regression to estimate hemoglobin concentration. , The proportion of those with anemia; , the mean hemoglobin concentration; whiskers, 95% CIs. The risk of anemia did not differ significantly between treatment groups and the reference Pp group, as indicated by odds ratios with 95% CIs that overlap 1.0, and there were no significant differences in mean hemoglobin concentrations between treatment groups.
Effect of HIV infection
Hemoglobin concentrations and the proportion of anemia did not differ between HIV-negative infants born to HIV-positive and those born to HIV-negative mothers (negative/negative compared with positive/negative; Table 2). However, mean hemoglobin concentrations were significantly lower and anemia was more common in HIV-positive infants than in negative/negative or positive/negative infants (P < 0.0001 for all tests). In unadjusted analysis, infant HIV infection increased the risk of anemia nearly 5-fold (odds ratio: 4.46; 95% CI: 3.32, 5.99). TBI at birth did not differ significantly between groups. The distribution of 6-mo plasma ferritin concentrations was shifted to higher values with maternal and infant HIV infection. Because serum ferritin is an acute-phase protein, we posit that higher ferritin concentrations do not indicate better iron status in HIV-exposed (positive/negative) and -infected (positive/positive) infants, but they do suggest the presence of inflammation.
View this table:
TABLE 2. Indicators of anemia and iron status in Zimbabwean infants by maternal and infant HIV infection1
Other determinants of postnatal anemia
The predictors of anemia differed according to infant HIV infection status. The determinants for HIV-negative infants were male sex, lower TBI at birth, and 6-mo plasma ferritin < 12 µg/L (Table 3). Reported fever and cough during the preceding 7 d and frequent morbidity were associated with slightly lower hemoglobin values and with no difference in anemia prevalence. In HIV-positive infants, risk of anemia increased with male sex, lower TBI, frequent morbidity, early HIV infection, and low maternal CD4+ lymphocyte count at recruitment (Table 4). Plasma ferritin concentrations <12 µg/L were associated with a risk of anemia in HIV-negative infants twice that in HIV-positive infants (P < 0.001), but it was not associated with anemia in HIV-positive infants (P = 0.13). We suggest that the lack of association in HIV-positive infants is an artifact of elevated ferritin concentrations during infection. A protective effect of EBF was observed in both HIV-negative and -positive infants, but it was not significant in the stratified analysis. In the total sample combining HIV-negative and -positive infants, the association was significant (adjusted odds ratio: 0.57; 95% CI: 0.35, 0.91; P = 0.012).
View this table:
TABLE 3. Predictors of hemoglobin concentration and anemia from multivariate adjusted linear and logistic regression models in 8–14-mo-old HIV-negative Zimbabwean infants1
View this table:
TABLE 4. Predictors of hemoglobin concentration and anemia from multivariate adjusted linear and logistic regression models in 8–14-mo-old HIV-positive Zimbabwean infants1
Overall, the strongest determinants of postnatal hemoglobin concentration and anemia prevalence were TBI at birth, infant HIV infection, and sex of the infant, and the effect of these determinants appeared to be additive (Figure 3). There was a strong dose-response relation between the quartile of TBI at birth and the risk of postnatal anemia (P for trend < 0.0001), and this relation persisted across HIV and sex groups. In a combined multivariate model for all infants, the adjusted odds ratio for infant HIV infection was 6.43 (95% CI: 4.40, 9.40; P < 0.001). Male HIV-positive infants in the lowest quartile of TBI at birth had the lowest predicted mean hemoglobin concentration of 86.4 (95% CI: 81.2, 91.5) g/L and the highest prevalence of anemia (96.5%, adjusted). Female HIV-negative infants in the highest quartile of TBI at birth had the highest predicted mean hemoglobin of 112.9 (95% CI: 110.4, 115.4) g/L and the lowest prevalence of anemia (18.9%, adjusted).
FIGURE 3.. Adjusted prevalence (95% CIs) of anemia (hemoglobin < 105 g/L) in Zimbabwean infants by infant HIV status, sex, and total body iron (TBI) at birth. In the TBI quartiles—236, 209–235, 181–208, and <181 mg—there were, respectively, 110, 120, 129, and 109 HIV-negative females; 145, 120, 125, and 109 HIV-negative males; 27, 26, 28, and 22 HIV-positive females; and 26, 29, 27, and 25 HIV-positive males. Prevalence values were adjusted for birth weight, vitamin A supplementation group, postnatal age, common morbidity, duration of exclusive breastfeeding, and 6-mo ferritin values. Values in HIV-negative infants also were adjusted for recent fever and cough, and values in HIV-positive infants also were adjusted for the time of HIV infection and maternal CD4+ lymphocyte count. The 3-factor interaction between infant HIV status, sex, and TBI quartile and all combinations of 2-factor interactions were not significant in logistic regression analysis. The main effects of infant HIV status, sex, and TBI quartile were significant (P < 0.01).
By maternal report, breastfeeding initiation was nearly 100%, and 99% of infants in this study were breastfed for 6 mo, but only 9.0% were exclusively breastfed for 6 wk. Solid foods were introduced to approximately one-quarter of the infants by 6 wk of age and to more than one-half by 3 mo, although very rarely did these foods include rich sources of iron (eg, meat, poultry, eggs, fish, or iron-fortified infant formula). The common infant foods were juice, porridge made from maize meal, and cooking oil. At the 6-mo visit, only 15% of infants consumed iron-rich foods during the preceding 7 d, and this variable was not related to hemoglobin concentration or anemia.
Weight and linear growth velocity were not retained in the final multivariate model for HIV-negative or -positive infants. However, in a separate analysis at 6 mo of age, in which anemia was defined more specifically as iron deficiency anemia (hemoglobin < 105 g/L and plasma ferritin < 12 µg/L), significantly greater weight (P < 0.01) and linear (P = 0.01) growth velocities were significantly associated with anemia even after adjustment for several potential confounding factors. Maternal hemoglobin concentrations measured at recruitment (within 96 h of delivery) or concurrently in a subsample of mothers (n = 534) did not affect infant hemoglobin or anemia.
DISCUSSION
Our findings confirm that anemia in late infancy is a public health problem in Zimbabwe, where it affects more than one-third of HIV-negative infants and three-quarters of HIV-positive infants. As far as we know, this is the first study to show the effects of high-dose neonatal and maternal VAS on the hemoglobin concentration and anemia in later infancy and to compare the effect of iron reserves at birth with the effects of other postnatal risk factors. VAS did not significantly improve hemoglobin concentrations. Infant HIV infection increased the risk of anemia 6-fold, and TBI at birth remained a strong independent predictor of anemia even when considered jointly with other postnatal risk factors in both HIV-infected and -uninfected infants.
The failure to find a significant increase in hemoglobin in response to VAS may largely be attributed to iron deficiency and is consistent with other research in iron-deficient children. Bloem et al (30) did not observe an increase in hemoglobin concentrations at 2 and 4 mo after supplementation [110 mg vitamin A (retinol equivalent) + 40 mg vitamin E] of Thai children selected on the basis of the presence of anemia (hemoglobin < 7.5 mmol/L). Similarly, Khan et al (31) did not find a significant difference in hemoglobin concentrations in 4–8-y-old Pakistani children 6 wk after their receipt of a single high dose (200 000 IU) of vitamin A. Both the Thai and Pakistani children had low iron stores, as evidenced by low serum ferritin concentrations, and the authors attributed the lack of hemoglobin response after VAS to iron deficiency. In a community-based study in West Java, Indonesia, vitamin A fortification of monosodium glutamate for 5 mo significantly increased hemoglobin concentrations, but a lack of further improvement in hemoglobin after a additional 6 mo of vitamin A fortification was likely due to iron deficiency (32). It is also possible that the single postpartum vitamin A doses administered in the current study to the mother (400 000 IU), the infant (50 000 IU), or both were insufficient to affect hematopoiesis 8–14 mo after supplementation. A multicountry study sponsored by the WHO (33) indicated that the previously recommended supplementation schedule for young infants [25 000 IU given with each diphtheria, pertussis, and tetanus vaccination at 6, 10, and 14 wk] was inadequate to improve or even sustain adequate vitamin A status in infants aged > 6 mo, and it was recommended that doses be doubled to 50 000 IU at each visit for a diphtheria, pertussis, and tetanus injection (34). Even the maternal dose, which increases the vitamin A content of breast milk (35, 36) and likely transfers more vitamin A than the neonatal dose in small daily increments transfers to the breastfeeding infant, may be inadequate to improve hemoglobin concentrations in later infancy (5).
In the current study, the consumption of iron-rich foods during the first 6 mo of life was rare and had no effect on hemoglobin concentrations or risk of anemia. Given the strong and persistent effect of TBI at birth on infant anemia, we were interested in comparing the relative amounts of iron from the estimated birth endowment (TBI) and from exogenous dietary sources absorbed during the latter half of infancy. Using data from the literature, we estimated the total intake and absorption of dietary iron from 6 to 11 mo of age. We assumed that breast-milk consumption was typical for partially breastfed infants aged 6–8 mo (660 mL/d) and 9–11 mo (616 mL/d) in developing countries, that the iron concentration in these infants was 0.30 mg/L (37), and that 50% of the iron in human milk was absorbed (25). We then considered that the breastfeeding and complementary feeding practices in our study sample were comparable to those reported in mothers and their infants in rural Malawi (38) and estimated that the median intake of iron was 1.2 mg/d with 5.5% bioavailability (yielding 0.07 mg/d absorbed) in those aged 6–8 mo and 2.8 mg/d with 7.4% bioavailability (0.20 mg/d absorbed) in those aged 9–11 mo. According to these estimates, total iron absorbed from breast milk and complementary foods over the 6-mo period was 42 mg, roughly one-fifth of the estimated birth endowment of iron (212 mg). For children who are still breastfeeding in the second half of the first year of life, complementary foods need to have a very high iron density if they are to meet the need for iron. Gibson et al (39) evaluated the iron density of 23 traditional, unfortified complementary food mixtures used in developing countries and concluded that none realized the desired density, not even those including animal sources of iron (heme iron). Similar conclusions were reached in a study of rural Bangladeshi infants, when linear programming was used to design a complementary food diet that meets multiple nutrient needs and minimizes cost (40).
To our knowledge, ours is the first study to link exclusivity of early breastfeeding to later anemia. Foods given before 6 mo tend to displace breast milk even when nursing frequency is maintained (41, 42) and to interfere with iron absorption from breast milk. The most common weaning foods consumed by young infants in this study were cooking oil, fruit juice, and a thin, maize-flour porridge, which contribute little other than energy and may decrease the infant's appetite for breast milk; moreover, maize is high in dietary phytate, a known inhibitor of iron absorption (43). Displacement of breast milk reduces the intake of the antiinfective components of human milk and increases exposure to pathogens if there is a scarcity of safe water and facilities for safe preparation and storage of foods. The early introduction of complementary foods has been associated with a greater risk of diarrheal morbidity (44, 45) and gastrointestinal infection (46), and it may lower hemoglobin in a nonspecific way during infection—leading to the alleged anemia of infection, a subset of the anemia of chronic disease (47). Although we included an indicator of morbidity in our multivariate model, the variable was unlikely to capture the full effect of repeated illness on anemia.
Maternal HIV infection without infant infection did not increase the risk of anemia. Similarly, in a clinic-based study in New York City (48), 94 of 100 symptomatic HIV-infected children were anemic (hemoglobin < 110 g/L), but there was no difference in the prevalence of anemia in 22 HIV-uninfected infants born to HIV-infected mothers (31.8%) and 25 HIV-negative infants born to HIV-negative intravenous drug–using mothers (32.0%). The latter 2 groups had similar results on a wide range of hematologic tests, giving no evidence of suppressed hematopoiesis in infants whose mothers are HIV infected.
Additional risk factors for anemia that were specific to HIV-infected infants included repeated illness, early HIV infection, and low maternal CD4+ lymphocyte counts. A high prevalence of anemia has been observed in HIV-infected children who are symptomatic or hospitalized (48–51). The anemia associated with repeated illness may in part be attributed to the anemia of infection. Hemoglobin concentrations in infants infected with HIV have been associated with progression to AIDS (52), and infants infected earlier in life are likely to have more advanced HIV disease. A low maternal CD4 count is a well-known risk factor for mother-to-child transmission of HIV during the intrauterine, intrapartum, and breastfeeding periods (53), and indeed lower maternal CD4 concentrations were associated with earlier infant infection (data not shown). However, the maternal CD4 count remained an independent predictor of infant anemia after adjustment for timing of infection. Thus, the infant anemia associated with sicker mothers may be operating through pathways that are not directly biological (eg, infant care).
The most important application of these findings is as a guide to the development of appropriate interventions to control anemia in Zimbabwean infants. VAS did not affect hemoglobin concentrations, probably because dietary iron was limited. Our results provide a strong basis for improving TBI at birth as a strategy to control anemia. Infants born before term or with low birth weight have smaller absolute amounts of iron and are prone to early postnatal iron deficiency (54), and we previously showed that maternal hemoglobin was a strong linear predictor of TBI at birth (55). We hypothesize that improving nutritional status of pregnant women to prevent maternal anemia and the delivery of low-birth-weight infants may be an essential component of public health efforts to prevent anemia during infancy. We need to find safe and efficacious ways to improve maternal iron status in the context of HIV infection. Boosting the endogenous prenatal source of iron will delay the occurrence of infantile anemia, but, when the birth endowment of iron is exhausted, additional intakes of iron will be essential. Interventions are needed to include greater amounts of bioavailable fortificant iron or heme iron in a form that the infant can chew, eg, powdered red meat (56). More intensive strategies such as medicinal iron may be necessary, but we emphasize that questions about iron supplementation during HIV infection remain to be answered (57). Our findings suggest that the promotion of EBF in the early months of life may be an additional means of controlling anemia. Clearly, HIV infection is an important cause of anemia, and specific interventions to reduce mother-to-child transmission of HIV are urgent. These interventions include antiretroviral therapy, safe delivery practices, and counseling and support with respect to infant-feeding methods.
ACKNOWLEDGMENTS
Members of the Zimbabwe Vitamin A for Mothers and Babies Project (ZVITAMBO) Study Group, in addition to the named authors, are Henry Chidawanyika, Agnes Mahomva, Florence Majo, Edmore Marinda, Michael Mbizvo, Lawrence Moulton, Kuda Mutasa, Mary Ndhlovu, Robert Ntozini, Ellen Piwoz, Lidia Propper, Philipa Rambanepasi, Andrea Ruff, Naume Tavengwa, Brian Ward, Lynn Zijenah, Claire Zunguza, and Partson Zvandasara; principal investigators are Kusum Nathoo and Jean Humphrey.
MFM designed and collected data for this specific study, performed the statistical analyses, and wrote the first draft of the manuscript under the guidance of RJS and JHH. JHH is the principal investigator of the ZVITAMBO and is responsible for fund raising, study design, and protocol development. PJI and LCM are co-investigators of the ZVITAMBO trial and provided oversight during data collection. NVM participated in data collection. RJS, JHH, and PJI contributed to the editing and revision of the manuscript. None of the authors had any financial or personal conflict of interest.
REFERENCES