An estimated 620,000 children were infected with HIV in sub-Saharan Africa in 2003, accounting for 89% of the world's total of new infections among children in the same year.1 In the absence of antiretroviral therapy, children in developing countries not only have a 25%-48% risk of acquiring HIV from their infected mothers2 but also are at an increased risk of being born prematurely, with low birthweight, or small for gestational age.3 These adverse pregnancy outcomes are associated with an increased risk of neonatal morbidity and mortality,4 impaired cognitive development,5 and subsequent manifestations of stunting and underweight.6
In the search for modifiable risk factors to improve pregnancy outcomes among HIV-infected pregnant women, multivitamin supplements containing vitamins B, C, and E reduced the risk of low birthweight, fetal death, and severe prematurity in a study conducted in Tanzania.7 Further research on the role of micronutrients in HIV disease indicates that selenium deficiency may be a risk factor for mother-to-child transmission (MTCT) of HIV, which can take place through the in utero, intrapartum, or breast-feeding routes. Selenium deficiency may increase the risk of MTCT through all 3 routes by accelerating HIV disease progression during pregnancy.8,9 It has also been associated with shedding of the virus in the female genital tract and may thus increase the risk of intrapartum transmission.10 In animal studies, low selenium status is linked to an increased risk of mastitis, an inflammatory process in the breast, which is a potent risk factor for MTCT of HIV during the breast-feeding period.11-13
Among populations who are presumably HIV uninfected, low selenium levels have been associated with low birthweight pregnancies14 and fetal death in the form of miscarriage.15 Other studies did not find such associations, and the role of selenium status in pregnancy outcomes thus remains unclear.16,17 To further address the role of selenium status in MTCT of HIV and in pregnancy outcomes, we conducted a cohort study among HIV-infected pregnant women in Tanzania.
Between April 1995 and July 1997, 1078 HIV-infected women were enrolled in a trial to examine the efficacy of vitamin supplements in relation to maternal and child health outcomes. Details of the trial have been described elsewhere.7 In brief, eligible women were between 12-27 weeks' gestation, residents of Dar es Salaam, and free of AIDS. Women were recruited from 4 prenatal clinics in Dar es Salaam, enrolled into the study at Muhimbili Medical Center, and randomly assigned to receive either placebo, vitamin A, multivitamins excluding vitamin A, or multivitamins including vitamin A. None of the treatment arms contained selenium. All women received daily doses of folate and iron and weekly doses of chloroquine phosphate. At delivery, women in the vitamin A group received an additional high-dose vitamin A supplement.
At the baseline visit, women underwent a complete physical examination. Anthropometric measurements included height, weight, and mid-upper arm circumference, while the sociodemographic variables included age and educational level. The stage of HIV disease was assigned based on the World Health Organization (WHO) Staging System.18 Obstetric history was evaluated using information on parity and prior adverse pregnancy outcomes. Women were tested for the presence of malaria, sexually transmitted diseases, intestinal helminths, and protozoan infections.
Plasma selenium and sodium concentrations were measured by neutron activation analysis at the University of Missouri Research Facility, Columbia, MO.19 Each sample's selenium concentration was adjusted by regressing the selenium concentration on the sodium concentration of the same sample. The residuals were obtained from this regression model, and to make them interpretable the median selenium value was added to them. This method adjusts for the degree of dilution of each sample that would otherwise introduce random measurement error. Despite the adjustment, the correlation between unadjusted and adjusted concentrations was high (r = 0.89), and adjusted selenium concentrations were used in all analyses. Absolute CD4 cell counts were measured using the fluorescein-activated cell sorter (FACScount and FACScan systems, Becton-Dickinson, San Jose, CA). Plasma levels of vitamins A and E were determined using reversed-phase high-performance liquid chromatography (HPLC) in a modification of the method of Zaman et al.20 Hemoglobin was measured using either a CBC5 Coulter counter (Coulter Corp., Miami, FL) or the cyanmethemoglobin method with a colorimeter (Corning, Inc., Corning, NY).
Gestational age was based on the woman's recollection of the date of her last menstrual period. Infant birthweight was measured to the nearest 10 g on a standard beam balance immediately after birth by a research midwife. Low birthweight was defined as birthweight <2500 g, preterm birth was defined as delivery before 37 weeks' gestation, and small for gestational age was defined as birthweight <10th percentile for gestational age using the standards of Brenner et al.21 Fetal death was defined as either miscarriage (delivery before 28 weeks' gestation) or stillbirth (delivery of a dead baby at or after 28 weeks' gestation). An “adverse pregnancy” endpoint was created among livebirths, which used composite information on small for gestational age, birthweight <2500 g, and preterm birth.
Infant blood samples were collected at birth (range, birth-21 days), at 6 weeks (range, 22-49 days), and every 3 months thereafter. HIV infections were based on a positive peripheral blood mononuclear cell specimen using the polymerase chain reaction (PCR) for children <18 months of age; samples were tested using the Amplicor HIV-1 detection kit (Roche Diagnostic Systems, Branchburg, NJ). At 18 months or later, HIV infections were based on a positive enzyme-linked immunosorbent assay that was confirmed by a Western blot. Babies who were HIV positive at birth were most likely infected in utero, while those who were HIV negative at birth but HIV positive at 6 weeks were probably infected during the intrapartum period or as a result of breast-feeding in the first 4 weeks of life. Children who were first found to be HIV infected after 6 weeks were assumed to be infected through breast-feeding. Children's vital status was tracked over the first 2 years by means of regular clinic visits or home visits in case a scheduled visit to the clinic was missed. A composite endpoint “HIV positive or dead at birth” was defined based on whether an infant survived the first 21 days of life and remained free of HIV infection during this time, while an endpoint “HIV positive or dead in the first 24 months” was used to denote whether a child survived the first 24 months of life and was free of HIV infection.
Of the 1078 HIV-infected women who were randomized in the original trial, there were 690 singleton pregnancies with known outcomes and plasma selenium levels (Fig. 1). We excluded 20 of these 690 pregnancies in the study with ≤1% missing data for the following variables: hemoglobin, history of adverse pregnancy outcome, mid-upper arm circumference, and WHO HIV stage. The missing indicator method was used for covariates with missing values >1%.22 Of the remaining 670 pregnancies, 611 resulted in livebirths and 59 in fetal deaths. Among the livebirths, 561 were HIV negative at birth. To enter analyses examining the risk of HIV transmission by 6 weeks, babies had to be HIV uninfected at birth. At 6 weeks, a specimen determining HIV status was available for 307 babies. For HIV transmission between 6 weeks and 24 months, 432 children were not known to be HIV infected by 6 weeks and had HIV testing thereafter. Among these, 92 were found to be HIV infected after 6 weeks. In this group, HIV status was not known for 35 children at 6 weeks, even though they had a negative test at birth. We conducted sensitivity analyses including and excluding these 35 children and obtained consistent results. Therefore, we present results from analyses using all 92 children.
The Kruskal-Wallis test23 was used to assess the statistical significance of differences between the levels of continuous baseline variables across tertiles for plasma selenium levels, while the χ2 test24 was constructed to assess the significance of differences in the distribution of categorical variables. For all pregnancy outcomes and HIV transmission at birth and at 6 weeks, relative risks and 95% CIs were obtained from binomial regression models with a natural logarithm link function.25 For mortality and HIV transmission outcomes after 6 weeks, Cox proportional hazards models were fit.26 A counting process data structure27 was used to stratify mortality analyses over the 24-month follow-up period by HIV infected and uninfected person-time. In all models, plasma selenium levels were modeled as tertiles and a test for trend was constructed using an ordinal variable that assumed the median selenium concentration of each tertile. Based on prior scientific knowledge and baseline distributions of variables across tertiles of selenium,28 we included the following maternal baseline variables in multivariate models: mid-upper arm circumference, WHO HIV disease stage, CD4 count, plasma vitamin A, plasma vitamin E, hemoglobin concentration, history of adverse pregnancy outcomes, and treatment regimen.
The study protocol was approved by the College 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.
In this study, high selenium levels were marginally associated with higher vitamin A levels and lower vitamin E and hemoglobin concentrations (Table 1). Previous low birth-weight and preterm pregnancies were least common among women in the middle tertile of selenium. Gestational age and educational level did not vary across tertiles of selenium.
The incidences of small for gestational age, preterm birth, and low birthweight were 12, 22, and 10%, respectively, giving rise to a total incidence of adverse pregnancy outcomes of 34% (Table 2). Fetal death was the result of 9% of all pregnancies (Table 3). HIV infection was diagnosed in 8% of children at birth, while 19% of children either died or became infected in the first 21 days of life.
Of children who were HIV uninfected at birth, 16% were infected by 6 weeks. Between the 6-week time point and the end of follow-up, 21% of children negative at 6 weeks got infected. The risk of total HIV transmission over the follow-up period was 34%. Among all livebirths, 24% died during the 24-month follow-up period and 45% either died or became HIV infected. These incidences obtained from women with selenium measurements are comparable to incidences obtained from the entire cohort.7,9,29,30
Low plasma selenium levels were not significantly associated with the risk of preterm birth or low birthweight. Low levels of plasma selenium were related to a decreased risk of small for gestational age (P value, test for trend = 0.03); compared with the highest tertile of plasma selenium, infants born to women in the lowest tertile of plasma selenium had a 44% decreased risk of being small for gestational age (RR = 0.56; 95% CI = 0.32-0.97). However, this association may have been biased due to the strong association between selenium levels and fetal death. Statistical adjustment of this survival bias31 slightly attenuated the association between selenium levels and occurrence of infants small for gestational age, yielding a 36% decreased risk (RR = 0.64; 95% CI = 0.35-1.15) among infants born to women in the lowest tertile of plasma selenium, compared with infants born to women in the highest tertile. Selenium levels were not associated with the endpoint “adverse pregnancy outcome,” defined as the composite of small for gestational age, birthweight <2500 g, and preterm birth.
While plasma selenium levels were not related to risk of in utero transmission of HIV, they were inversely related to risk of transmission through the intrapartum and early breast-feeding period (P value, test for trend = 0.01). Compared with the highest tertile, the risks were elevated by 79% (95% CI = −20%-295%) and 151% (95% CI = 19%-430%) in the middle and lowest tertiles, respectively. Selenium levels were not significantly associated with risk of breast-feeding HIV transmission (P value, test for trend = 0.15).
Risk of fetal death was significantly elevated in the lowest tertile of plasma selenium (RR = 1.94; 95% CI = 1.08-3.49) relative to the highest (Table 3). When all child deaths (including fetal deaths) during the first 24 months were considered, selenium levels were inversely related to risk of death (P value, test for trend = 0.03). This association was slightly weakened when fetal deaths were excluded (P value, test for trend = 0.15). Child HIV status throughout follow-up was not a statistically significant modifier of the association between selenium levels and child death, even though lower selenium levels were associated with higher risk estimates among children who remained HIV negative during follow-up. Selenium levels were not related to HIV-free survival during the first 24 months of life.
We found that infants born to women with low plasma selenium levels were at increased risks of fetal death, child deaths in the first 2 years, particularly among HIV-uninfected children, and HIV transmission during the intrapartum and early breast-feeding period. We also noted an apparent positive association between maternal selenium levels and risk of small for gestational age.
The effect of maternal selenium status for preterm birth has not previously been examined among HIV-infected women, and its role among HIV-uninfected women remains controversial. In studies from Northern Ireland and the United States among presumably HIV-negative women, selenium concentrations taken at birth did not differ between women giving birth to term or preterm babies.17,32 Selenium levels at delivery were significantly lower among preterm parturients in one study from Poland33; in a second study from the same country, selenium levels were higher among preterm parturients, but this difference was not statistically significant.34 In our study, selenium status was not related to risk of preterm birth, but it is difficult to compare findings from the European and US studies with the findings presented in this paper. This is because our study population was HIV infected, differed in terms of ethnic as well as lifestyle factors, and had higher selenium status (as evidence by higher mean selenium levels), possibly due to higher dietary intakes of selenium. Furthermore, since none of the prior studies adjusted for confounding factors, the results obtained may have been biased by the failure to account for other determinants of risk.
The literature on the association between selenium status and low birthweight and small for gestational age is also scant. In 2 European studies among presumably HIV-uninfected women, maternal plasma selenium levels were not associated with low birthweight pregnancy outcomes.33,35 In a study from Zaire among HIV-untested women, selenium levels at delivery were lower among women delivering babies <2500 g compared with those delivering babies with higher birthweights.14 Existing studies therefore cannot explain the decreased risk of small for gestational age associated with low maternal levels observed in our study. It is possible that the association with small for gestational age is partially due to the association between selenium status and fetal death: since children born to women with low selenium levels had an increased risk of fetal death, a disproportionate number of children in this group may have died instead of being born small for gestational age. In fact, adjusting for this survival bias31 diluted the association between selenium levels and risk of small for gestational age. Nevertheless, it is difficult to exclude a potentially harmful effect of high maternal selenium status on risk of small for gestational age, and this issue deserves further study.
Women with low selenium levels were at an increased risk of passing HIV to their children during the intrapartum period. In a study from Kenya, women who were selenium deficient (plasma levels <85 μg/L ) were nearly 3 times more likely to shed HIV in the genital tract compared with women with adequate selenium levels (≥85 μg/L).10 Since viral shedding in the genital tract is a potent risk factor for intrapartum HIV transmission,36 the Kenyan findings may help explain the association between maternal selenium levels and risk of intrapartum HIV transmission observed in our study. Low selenium levels may also be a risk for breast-feeding transmission of HIV by facilitating the development of mastitis.13,37 However, we did not find a significant association between selenium levels and breast-feeding HIV transmission in this study.
Associations between low serum selenium status and fetal death in the form of miscarriage have been described in the literature among presumably HIV-uninfected women.15,38 Selenium is a cofactor for the antioxidant enzyme glutathione peroxidase and poor selenium status may lead to oxidative damage of DNA and cell membranes.39 Oxidative damage during embryonal and fetal development may be a pathway leading to miscarriage and fetal death and may thus help explain the increased risk of fetal death observed in this study.15,40,41 Previous work in this study cohort described an association between low plasma selenium levels and decreased weight gain during pregnancy,42 which may be a mechanism leading to an increased risk of fetal death.30,43 It is also possible that the increased risk of fetal death in this study was due to an increased risk of HIV infection in utero, but this issue cannot be addressed, as we did not assess HIV status among fetal deaths.
We observed an association between low maternal plasma selenium levels and an increased risk of child mortality in the first 24 months of life. Low maternal selenium status may lead to lower breast milk selenium concentrations44 and may impair child selenium status and immune system development.45 Low maternal selenium levels were somewhat more indicative of an increased mortality risk among children who remained free of HIV than among children who became infected. A possible explanation is that HIV infection overrides the adverse consequences of low child selenium status, conferring a benefit only among HIV-uninfected children. Nevertheless, a study among HIV-infected children from the United States found an association between low selenium levels and an increased risk of mortality.8
In this study, selenium concentrations were adjusted for the degree of dilution of the sample. Adjusted results were generally stronger than unadjusted results, possibly because adjustment lowered the degree of random measurement error. Plasma selenium levels have been shown to reflect dietary selenium intake among apparently healthy populations.46 It is possible, however, that plasma selenium levels in our study were depressed as a result of pathologic processes that may also be determinants of pregnancy outcomes and risk of MTCT of HIV, such as inflammation and poor immunologic status. In this regard, it is reassuring that adjusting for proxies for these underlying pathologies, such as vitamin A levels, CD4 cell count, and erythrocyte sedimentation rate (data not shown) did not substantially change the results; however, as in every observational study, it is possible that some confounding variables remained unmeasured.
The value of plasma selenium levels in defining selenium deficiency remains uncertain. While previous studies on selenium status and HIV disease have used a criterion of deficiency of <85 μg/L, there is evidence that higher plasma levels are needed for optimal biologic function.47 In addition, selenium supplementation increased immune function in a study population with mean selenium levels >130 μg/L, implying that these levels did not optimize immune function.48 In our study population, selenium levels were relatively high (range, middle tertile = 114-131 μg/L) but the inverse associations with risk of fetal death, child death, and intrapartum HIV transmission illustrate that such levels may not have been adequate to optimize HIV-related pregnancy and child health outcomes.
Selenium is currently included in a prenatal supplement that is being evaluated in pilot effectiveness programs in the developing world.49 In the absence of data on the efficacy of selenium to improve birth outcomes, it should not be assumed that administering selenium to women without evidence of selenium deficiency will be safe and beneficial. Selenium supplementation may, however, help reduce the risk of intrapartum HIV transmission among women who do not receive antiretroviral therapy. Further examination of these outcomes in observational studies and possibly in randomized trials is warranted.
1. UNAIDS. AIDS Epidemic Update
. Geneva: UNAIDS; 2003.
2. Dabis F, Msellati P, Dunn D, et al. Estimating the rate of mother-to-child transmission of HIV
: report of a workshop on methodological issues Ghent (Belgium), February 17-20, 1992. The Working Group on Mother-to-Child Transmission of HIV
3. Brocklehurst P, French R. The association between maternal HIV
infection and perinatal outcome: a systematic review of the literature and meta-analysis. Br J Obstet Gynaecol
4. Ashworth A. Effects of intrauterine growth retardation on mortality and morbidity in infants and young children. Eur J Clin Nutr
5. Grantham-McGregor SM. Small for gestational age, term babies, in the first six years of life. Eur J Clin Nutr
6. Martorell R, Ramakrishnan U, Schroeder DG, et al. Intrauterine growth retardation, body size, body composition and physical performance in adolescence. Eur J Clin Nutr
7. Fawzi WW, Msamanga GI, Spiegelman D, et al. Randomised trial of effects of vitamin supplements on pregnancy
outcomes and T cell counts in HIV
-1-infected women in Tanzania
8. Campa A, Shor-Posner G, Indacochea F, et al. Mortality risk in selenium
-positive children. J Acquir Immune Defic Syndr
9. Fawzi W, Msamanga G, Renjifo B, et al. Predictors of intrauterine and intrapartum transmission of HIV
-1 among Tanzanian women. AIDS
10. Baeten JM, Mostad SB, Hughes MP, et al. Selenium
deficiency is associated with shedding of HIV
-1-infected cells in the female genital tract. J Acquir Immune Defic Syndr
11. Ndiweni N, Field TR, Williams MR, et al. Studies on the incidence of clinical mastitis and blood levels of vitamin E and selenium
in dairy herds in England. Vet Rec
12. Semba RD, Neville MC. Breast-feeding, mastitis, and HIV
transmission: nutritional implications. Nutr Rev
13. John GC, Nduati RW, Mbori-Ngacha DA, et al. Correlates of mother-to-child human immunodeficiency virus type 1 (HIV
-1) transmission: association with maternal plasma HIV
-1 RNA load, genital HIV
-1 DNA shedding, and breast infections. J Infect Dis
14. Arnaud J, Preziosi P, Mashako L, et al. Serum trace elements in Zairian mothers and their newborns. Eur J Clin Nutr
15. Barrington JW, Lindsay P, James D, et al. Selenium
deficiency and miscarriage: a possible link? Br J Obstet Gynaecol
16. Dobrzynski W, Trafikowska U, Trafikowska A, et al. Decreased selenium
concentration in maternal and cord blood in preterm compared with term delivery. Analyst
17. Mask G, Lane H. Selected measures of selenium
status in full-term and preterm neonates, their mothers and nonpregnant women. Nutr Res
18. World Health Organization. Proposed “World Health Organization staging system for HIV
infection and disease”: preliminary testing by an international collaborative cross-sectional study. The WHO International Collaborating Group for the Study of the WHO Staging System. AIDS
19. Mason M, Morris J, Spate V, et al. Comparison of whole blood, plasma, and nails as monitors for the dietary intake of selenium
. J Radioanal Nucl Chem
20. Zaman Z, Fielden P, Frost PG. Simultaneous determination of vitamins A and E and carotenoids in plasma by reversed-phase HPLC in elderly and younger subjects. Clin Chem
21. Brenner WE, Edelman DA, Hendricks CH. A standard of fetal growth for the United States of America. Am J Obstet Gynecol
22. Miettinen O. Theoretical Epidemiology
. New York: John Wiley & Sons; 1985.
23. Kruskal W. A nonparametric test for the several sample problem. Ann Math Stat
24. Colton T. Statistics in Medicine
. Boston: Little, Brown & Co.; 1974.
25. Wacholder S. Binomial regression in GLIM: estimating risk ratios and risk differences. Am J Epidemiol
26. Cox D. Regression models and life tables. J R Stat Soc
27. Andersen P, Gill R. Cox's regression model counting process: a large sample study. Ann Stat
28. Rothman K, Greenland S. Modern Epidemiology
. 2nd ed. Philadelphia: Lippincott Williams & Wilkins; 1998.
29. Fawzi WW, Msamanga G, Hunter D, et al. Randomized trial of vitamin supplements in relation to vertical
transmission of HIV
-1 in Tanzania
. J Acquir Immune Defic Syndr
30. Dreyfuss ML, Msamanga GI, Spiegelman D, et al. Determinants of low birth weight among HIV
-infected pregnant women in Tanzania
. Am J Clin Nutr
31. Bang H, Spiegelman D. Estimating treatment effects in studies of perinatal transmission of HIV
32. Wilson DC, Tubman R, Bell N, et al. Plasma manganese, selenium
and glutathione peroxidase levels in the mother and newborn infant. Early Hum Dev
33. Dobrzynski W, Trafikowska U, Trafikowska AP, et al. Decreased selenium
concentration in maternal and cord blood in preterm compared with term delivery. Analyst
34. Wasowicz W, Wolkanin P, Bednarski M, et al. Plasma trace element (Se, Zn, Cu) concentrations in maternal and umbilical cord blood in Poland: relation with birth weight, gestational age, and parity. Biol Trace Elem Res
35. Mask G, Lane H. Selected measures of selenium
status in full-term and preterm neonates, their mothers and nonpregnant women. Nutr Res
36. Mofenson LM. Epidemiology and determinants of vertical HIV
transmission. Semin Pediatr Infect Dis
37. Semba RD, Neville MC. Breast-feeding, mastitis, and HIV
transmission: nutritional implications. Nutr Rev
38. Al-Kunani AS, Knight R, Haswell SJ, et al. The selenium
status of women with a history of recurrent miscarriage. Br J Obstet Gynaecol
39. Flohe L. Glutathione peroxidase. Basic Life Sci
40. Ursini F, Bindoli A. The role of selenium
peroxidases in the protection against oxidative damage of membranes. Chem Phys Lipids
41. Sinclair AJ, Barnett AH, Lunec J. Free radicals and antioxidant systems in health and disease. Br J Hosp Med
42. Villamor E, Msamanga G, Spiegelman D, et al. Pattern and predictors of weight gain during pregnancy
-1-infected women from Tanzania
. J Acquir Immune Defic Syndr
43. Kramer MS. Determinants of low birth weight: methodological assessment and meta-analysis. Bull World Health Organ
44. Kumpulainen J, Salmenpera L, Siimes MA, et al. Selenium
status of exclusively breast-fed infants as influenced by maternal organic or inorganic selenium
supplementation. Am J Clin Nutr
45. Dylewski ML, Mastro AM, Picciano MF. Maternal selenium
nutrition and neonatal immune system development. Biol Neonate
46. Longnecker MP, Stram DO, Taylor PR, et al. Use of selenium
concentration in whole blood, serum, toenails, or urine as a surrogate measure of selenium
47. Neve J. New approaches to assess selenium
status and requirement. Nutr Rev
48. Kiremidjian-Schumacher L, Roy M, Wishe HI, et al. Supplementation with selenium
and human immune cell functions. II. Effect on cytotoxic lymphocytes and natural killer cells. Biol Trace Elem Res
49. Shrimpton R, Schultink W. Can supplements help meet the micronutrient needs of the developing world? Proc Nutr Soc