Selenium levels in relation to morbidity and mortality among children ...

European Journal of Clinical Nutrition (2005) 59, 1250–1258

& 2005 Nature Publishing Group All rights reserved 0954-3007/05 $30.00 www.nature.com/ejcn

ORIGINAL COMMUNICATION Selenium levels in relation to morbidity and mortality among children born to HIV-infected mothers R Kupka1,2*, GI Msamanga4, D Spiegelman2,3, N Rifai5,6, DJ Hunter1,2 and WW Fawzi1,2 1 Department of Nutrition, Harvard School of Public Health, Boston, MA, USA; 2Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA; 3Department of Biostatistics, Harvard School of Public Health, Boston, MA, USA; 4Department of Community Health, Muhimbili University College of Health Sciences, Dar es Salaam, Tanzania; 5Department of Laboratory Medicine, Children’s Hospital, Boston, MA, USA; and 6Department of Pathology, Harvard Medical School, Boston, MA, USA

Objective: To examine the relation between selenium status and child mortality and morbidity among children born to HIVinfected mothers. Design: Prospective cohort study. Methods: Study participants were originally part of a trial to study the effect of maternal vitamin supplements on maternal and child health outcomes. Morbidity information was collected during monthly clinic visits until the child reached 24 months of age. Out of 984 livebirths, 806 had morbidity information, and 610 also had data on plasma selenium levels available. Setting: A study clinic at Muhimbili National Hospital, Dar es Salaam, Tanzania, a tertiary-care hospital. Results: The median age at baseline was 10.5 weeks. A total of 117 (19%) of the 610 study children died during follow-up. In a multivariate model, child plasma selenium levels were inversely associated with risk of all-cause mortality (P-value, test for trend ¼ 0.05). Plasma selenium levels were not significantly associated with risk of diarrhea or respiratory outcomes. Conclusions: Among infants born to HIV-infected women in sub-Saharan Africa, selenium status may be important to prevent child mortality. These preliminary findings warrant future reexamination. Sponsorship: This study was supported by the National Institute of Health and Human Development (NICHD R01 32257), and the Fogarty International Center (NIH D43 TW00004).

European Journal of Clinical Nutrition (2005) 59, 1250–1258. doi:10.1038/sj.ejcn.1602236; published online 13 July 2005 Keywords: selenium; HIV disease; infant mortality; diarrhea; pneumonia

Introduction Of the 10.8 million children aged 0–59 months who die every year, 41% of them do so in sub-Saharan Africa, mostly as a consequence of HIV disease and malnutrition (UNICEF; Black et al, 2003). Nutrition intervention studies from subSaharan Africa indicate that vitamin A supplements reduce

*Correspondence: R Kupka, Department of Nutrition, Harvard School of Public Health, 665 Huntington Ave., Boston, MA 02115, USA. E-mail: [email protected] Guarantor: R Kupka. Contributors: RK analyzed and interpreted the data and wrote the initial draft of the manuscript. GIM and WWF are principal investigators of the Tanzania Vitamin Supplementation Trial and contributed to the study design and its implementation. DS provided statistical guidance in data analyses. NR conducted the nutrient analyses. DJH assisted in the design of the Tanzania Vitamin Supplementation Trial and in data interpretation. All coauthors participated in manuscript preparation. Received 27 October 2004; revised 26 April 2005; accepted 25 May 2005; published online 13 July 2005

the risk of child mortality (Fawzi et al, 1999a) and that multivitamins (vitamins B, C, and E) given to women during pregnancy and lactation reduce the risk of child morbidity (Fawzi et al, 2003). It is likely that other micronutrient deficiencies contribute to child morbidity and mortality in sub-Saharan Africa and other settings worldwide. For example, supplements of the trace element zinc significantly reduce the risk of pneumonia and diarrhea among children in developing countries (Bhutta et al, 1999). Little is known on the role of the trace element selenium in child morbidity and mortality, even though selenium is known to function in important physiological processes such as antioxidant defense and T- and B-cell-mediated immunity (McKenzie et al, 1998). Observational studies from the US among HIV-infected children suggest that low selenium levels are associated with impaired anthropometric and immunological status (Miller et al, 1993), faster HIV disease progression (Campa et al, 1999), and increased risk of mortality (Campa et al, 1999). However, it is not clear

Selenium levels in relation to child morbidity and mortality R Kupka et al

1251 whether these findings apply to sub-Saharan Africa, the region most deeply afflicted with the AIDS epidemic. In addition, the relation between child selenium status and diarrhea as well as respiratory infections, leading causes of child morbidity and mortality worldwide (Black et al, 2003), has not been examined. Trials are not yet warranted to examine these associations due to the scarcity of existing information. We therefore employed a prospective cohort design to investigate the association between plasma selenium levels and child mortality as well as morbidity among children born to HIV-infected mothers in Tanzania.

Methods Study design and population This study was conducted among children born to HIVpositive women from Dar es Salaam, Tanzania who were participating in a clinical trial to examine the effect of maternal vitamin supplements during pregnancy and lactation on maternal and child health outcomes (Fawzi et al, 1998, 1999b). Trial participants were enrolled and followed up at a study clinic at Muhimbili National Hospital, the main tertiary-care hospital in Tanzania, between April 1995 and August 2003. As part of the vitamin trial, pregnant women who were at 12–27 weeks of gestation were randomized to either a daily dose of vitamin A alone, multivitamins (vitamins B, C, and E) but no vitamin A, multivitamins (vitamins B, C, and E) plus vitamin A, or placebo. No selenium was given. Standard prenatal and childcare services were provided to the women and children irrespective of experimental group. At the time of the study, antiretroviral therapy was not available to study participants. Detailed information about the vitamin trial has been published elsewhere (Fawzi et al, 1998, 2000, 2002). There were 1041 pregnancies with known pregnancy outcome in the vitamin trial, and 984 resulted in livebirths. Among the livebirths, 806 infants had morbidity and mortality information collected. Enrollment into this prospective cohort study was scheduled to occur at 6 weeks, the time when nutritional laboratory investigations were scheduled to be performed. Of the infants with morbidity and mortality information, 610 had valid baseline selenium measurements and therefore constituted the study population. The biochemical nutritional parameters evaluated consisted of plasma levels of selenium, albumin, ferritin, vitamin A, vitamin B12, vitamin E, and cholesterol. A second blood draw for these parameters was scheduled for 6 months. In addition, CD4 cell counts were enumerated at birth and at 6-month intervals thereafter, while child HIV status was determined at birth, 6 weeks, and 3-month periods thereafter. Follow-up continued until a child reached 24 months of age. Information on maternal socioeconomic status was collected through questionnaires during pregnancy. Birthweight was determined by a research midwife and the

duration of pregnancy was calculated based on the duration elapsed since the last menstrual period. Mothers were asked to present to the clinic with their children at 6 weeks after delivery and at monthly intervals thereafter. At each visit, child anthropometric measurements were obtained and mothers were asked to report episodes of child morbidity during the previous months. Specifically, mothers reported the number of days, if any, that the child had signs of diarrheal or respiratory disease during the previous month. Women who missed their clinic visits were contacted via home visits, and neighbors or relatives in the area were consulted to gather information on the vital status of the mother and her child. No morbidity information was collected during home visits. Based on a mother’s perception, diarrhea was defined as Z3 watery stools in a 24-h time period during the previous month, and we inquired whether it was associated with blood or mucus. Episodes of diarrhea during each monthly interval were defined as follows: all intervals with Z1 but o14 days of diarrhea were considered acute diarrhea. Episodes of acute diarrhea with mucus or blood were classified as dysentery; all other episodes (with neither mucus nor blood) were classified as watery diarrhea. Since there were only a small number of episodes of persistent diarrhea (defined as Z14 days of diarrhea), this outcome measure could not be examined. Child mortality was based on mothers’ reports and was captured using verbal autopsy forms. As signs of respiratory infection, we considered fever, cough, difficulty in breathing, chest retractions, and difficulty in eating, drinking, or breastfeeding. A child’s respiratory rate was measured using a stopwatch on the day of the monthly visit. Rapid respiratory rate was diagnosed if the number of breaths per minute was Z50 for infants and Z40 for children older than 1 y. Respiratory infection was classified into four different categories based on the following symptoms: (1) cough alone; (2) cough with rapid respiratory rate on the day of visit; (3) cough and fever; or (4) ‘cough plus’, defined as cough with at least one of the following symptoms: difficulty breathing; chest retractions; or refusal to eat, drink, or breastfeed.

Laboratory methods Nutrient concentrations of the plasma portions of blood samples were determined at the Clinical and Epidemiologic Research Laboratory, Boston Children’s Hospital. Plasma selenium was measured using a graphite-furnace atomic absorption spectrophotometric method with deuterium background correction and a reduced palladium modifier, employing a Perkin-Elmer system (Jacobson & Lockitch, 1988). Plasma vitamins A and E were analyzed using highperformance liquid chromatography using the Shimadzu system (De Leenheer et al, 1979). Vitamin B12 was measured by a competitive magnetic separation assay on the Technico Immuno-1 analyzer (Bayer, Tarrytown, NY, USA), and ferritin European Journal of Clinical Nutrition


1252 was measured by a particle-enhanced immunoturbidimetric assay using the Hitachi 911 analyzer (Roche Diagnostics, Indianapolis, IN, USA) and Kamiya Biomedical reagents (Seattle, WA, USA). Cholesterol was determined with an enzyme and subsequent peroxidase/phenol-4-aminophenazone indicator reaction (Allain et al, 1974), and albumin was measured by an automated dye-binding method using the Hitachi 911 analyzer and Roche Diagnostics reagents (Indianapolis, IN, USA). A child was determined to be HIV 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 ELISA (and was confirmed by a Western blot test). A child was determined to be HIV uninfected if a blood sample obtained at the last visit had a negative HIV PCR test result or if a plasma specimen obtained at the last visit among children 18 months or older had a negative ELISA test result. Infant CD4 cell counts were measured using the FACScount system (Becton-Dickinson, San Jose, CA, USA); the FACScan system was used when the FACScount system indicated that CD4 cell counts were Z2000 cells/mm3.

Statistical methods Baseline selenium levels were categorized into quartiles, using the following cutoff levels (in mg/l): r57.0, 57.1–66.4, 66.5–77.5, and Z77.6 for quartiles 1 (lowest) to 4 (highest), respectively. For study participants who had a second sample available, categorizations for the second time point were based on the distribution of the second samples, yielding the cutoffs (in mg/l) r58.9, 59.0–69.0, 69.1–80.5, and Z80.6 for quartiles 1 (lowest) to 4 (highest), respectively. Prematurity was defined as duration of pregnancy o37 weeks, while low birthweight was defined as birthweight o2500 g. The Kruskal–Wallis test (Kruskal, 1952) was used to compare means of continuous baseline variables across quartiles of baseline plasma selenium levels; the w2 test (Colton, 1974) was constructed to compare binary baseline variables. Incidence rates were calculated for the all-cause mortality outcome measure and each morbidity outcome measure. Cox’s proportional-hazards models were fitted to model the association between plasma selenium levels and mortality (Cox, 1972). A counting process data structure (Andersen & Gill, 1982) was used to take into account the time-varying nature of selenium measurements and other covariates. Child age (in weeks) was chosen as the time variable in these models. Generalized estimating equations (Proc Genmod in SAS) (Diggle et al, 1994) were used to compare incidences of morbidity across time-varying plasma selenium levels. In these models, a working binomial distribution with a log link function was used. The time variable in the model corresponded to child age in weeks. To accommodate nonlinear trends of morbidity rates over time, we used restricted cubic European Journal of Clinical Nutrition

spline models for child’s age with four knots (Stone & Koo, 1985; Devlin & Weeks, 1986). Multivariate mortality and morbidity analyses were adjusted for time-varying HIV infection status; analyses were also performed that stratified by HIV-infected and HIVuninfected person-time. Both stratified and unstratified multivariate models were adjusted for child age, multivitamin treatment regimen, baseline CD4 cell count, and time-varying measurements of weight-for-age, plasma albumin, plasma ferritin, plasma vitamin E (cholesterol adjusted), and plasma vitamin A. The following categorizations were chosen for these variables: CD4 count (median; o1300 cells/mm3), weight-for-age (z-score; o1.645), plasma albumin (o3.5 g/dl), plasma ferritin (o200 mg/l), plasma vitamin E (o5 mg/ml), and plasma vitamin A (o10 mg/dl). Missing values for CD4 cell count and weight-for-age were represented with a missing indicator term in multivariate models (Miettinen, 1985). There were no missing values for other covariates in multivariate models. Only 2% of infants were no longer breastfeeding at the time of the nutrient measurements were taken; hence, confounding by breastfeeding status is unlikely to be substantial and we did not include breastfeeding status in multivariate models. All P-values reported are two-sided. SAS software version 8.2 (SAS Institute Inc., Cary, NC, USA) was used for all statistical analyses.

Ethical clearance Written consent for the parent trial and subsequent nutrient analyses was obtained from the study women before enrollment into the parent trial. 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.

Results Of the 610 children in the study, 262 had only one plasma selenium measurement available (median time point (s.d.) ¼ 6.0 weeks (7.6)), while 348 also had a second sample available (median time point (s.d.) ¼ 26.0 weeks (3.8)). The median selenium concentration (25th, 75th percentile) increased from 66.4 (57.1, 77.5) mg/l at the first time point to 69.1 (59.1, 80.5) mg/l at the second time point. The morbidity experience of children with and without selenium measurements was not significantly different for any of the outcome measures (PZ0.05). Child and maternal baseline characteristics are shown in Table 1. Among the study children, plasma albumin concentrations were lowest in the lowest quartile of plasma selenium, while ferritin concentrations were highest in the lowest two quartiles. Plasma vitamin A levels were positively associated with plasma selenium levels. Plasma vitamin E levels also differed


1253 Table 1 Baseline characteristics of 610 children aged 6 weeks who were born to HIV-infected pregnant women, stratified by quartiles of baseline plasma selenium levels Quartile of baseline plasma selenium level: median (25th, 75th percentile)

Child characteristic Albumin (g/dl) Ferritin (mg/l) Vitamin A (mg/dl) Vitamin B12 (pg/ml) Vitamin E (mg/ml)c CD4 cell count (cells/mm3)d Weight-for-age (z-score)e HIV infection at 6 weeks (%) Low birthweight (%)f Prematurity (%)g Male gender (%)h Maternal characteristics EducationZsecondary school (%) Money spent on food (TSh)I Trial regimen (%) Multivitaminsj Vitamin A

1 (lowest) 51.3 (45.1–54.4) mg/l

2 62.6 (59.9–64.7) mg/l

3 71.4 (68.7–74.3) mg/l

4 (highest) 86.4 (81.4–95.0) mg/l

P-value a

3.1 (0.3)b 244 (213) 12.2 (4.6) 508 (267) 6.5 (2.8) 1249 (576) 0.33 (1.22) 12 10 21 49

3.2 (0.3) 239 (171) 13.7 (5.7) 454 (201) 7.1 (2.9) 1312 (599) 0.06 (0.98) 25 8 22 54

3.2 (0.4) 197 (145) 14.3 (4.8) 446 (227) 7.2 (2.7) 1518 (743) 0.02 (1.04) 11 11 23 51

3.2 (0.4) 204 (158) 14.9 (5.1) 490 (260) 6.6 (2.9) 1436 (701) 0.06 (1.09) 16 12 24 46

o0.0001 0.05 o0.0001 0.24 0.01 0.02 0.12 0.002 0.69 0.93 0.59

10 487 (212)

16 523 (257)

11 520 (258)

13 536 (271)

57 45

58 55

51 48

42 55

0.40 0.73 0.02 0.22

a

Based on Kruskal–Wallis test for continuous variables and w2 test for binary variables. Mean (s.d.). c Adjusted for cholesterol concentration of the sample. d N ¼ 442. e N ¼ 593. f N ¼ 569. Defined as birthweight o2500 g. g N ¼ 569. Defined as gestational age at birth o37 weeks. h N ¼ 608. i N ¼ 551. Defined as average household income (in Tanzanian shillings) spent on food per person per day, as estimated by dividing the total amount of money spent on food, by the number of household members. j Vitamins B, C, and E. b

significantly across quartiles of plasma selenium, but not plasma B12 levels. CD4 cell counts were positively associated with selenium concentrations. Weight-for-age z-scores were lowest in the lowest quartile of plasma selenium, while the proportion of children infected with HIV at 6 weeks was highest in the second-to-lowest quartile of plasma selenium. The gender distribution and the proportions of low birthweight and premature births did not vary across levels of plasma selenium. Among the maternal characteristics, socioeconomic indicators (educational level, money spent on food) did not vary across plasma selenium levels. The proportion of women receiving vitamin A as part of the parent trial was similar across quartiles of plasma selenium, whereas there were differences for the proportion of women receiving multivitamins.

Mortality Over the 24 months of follow-up, 117 (19%) of the 610 children died (Table 2). The median time of death (25th, 75th percentile) was 40 (23, 67) weeks. In an age-adjusted Cox’s proportional-hazards model, plasma selenium levels

were inversely related to risk of all-cause mortality (P-value, test for trend ¼ 0.0003) (Table 2). Children in the lowest quartile had a 134% increased risk of mortality (rate ratio (RR) ¼ 2.34, 95% CI ¼ 1.36–4.04) compared to those in the highest quartile. Adjusting for confounding factors attenuated this association (P-value, test for trend ¼ 0.05). In the multivariate model, children in the lowest quartile of plasma selenium had a 62% increased risk of mortality (RR ¼ 1.62, 95% CI ¼ 0.93–2.84) compared to children in the highest quartile. In stratified analyses, the association between selenium status and mortality did not significantly differ by child HIV status over the 24-month observation period (P-value, test for interaction ¼ 0.76). Elevated point estimates were generally noted for decreasing concentrations of plasma selenium, but none of the estimates were statistically significant.

Morbidity There was no association between child selenium status and episodes of diarrheal morbidity in univariate and European Journal of Clinical Nutrition


1254 Table 2

Association between plasma selenium levels and mortality among 610 children born to HIV-infected mothers, overall and stratified by HIV status Age-adjusted model b Mortality rate (cases/person weeksa)

RR (95% CI) c

4 (highest)f 3 2 1 (lowest)

19/12940 22/12258 35/11700 41/11607

1.00 1.20 (0.65,2.22) 1.98 (1.13,3.46) 2.34 (1.36,4.04)

0.0003

1.00 1.12 (0.60–2.09) 1.39 (0.78–2.47) 1.62 (0.93–2.84)

0.05

Stratified by HIV status HIV infected Selenium quartile 4 (highest) Selenium quartile 3 Selenium quartile 2 Selenium quartile 1 (lowest)

13/2173 19/2216 27/3053 33/2748

1.00 1.46 (0.72,2.96) 1.46 (0.75,2.83) 2.03 (1.07,3.86)

0.02

1.00 1.39 (0.67–2.88) 1.32 (0.67–2.63) 1.62 (0.83–3.15)

0.15

6/10767 3/10042 8/8647 8/8860

1.00 0.53 (0.13,2.12) 1.64 (0.57,4.72) 1.61 (0.56,4.64)

0.19

1.00 0.56 (0.14–2.25) 2.01 (0.68–5.94) 1.43 (0.49–4.18)

0.28

Total mortality Selenium quartile Selenium quartile Selenium quartile Selenium quartile

HIV uninfected Selenium quartile Selenium quartile Selenium quartile Selenium quartile

4 (highest) 3 2 1 (lowest)

P-value, test for trend d

Multivariate model e

RR (95% CI)

P-value, test for trend

a

Person weeks ¼ sum of weeks contributed by study participants at risk of death in each stratum over the follow-up period. Adjusted for age (in weeks). From Cox’s proportional-hazards models with a counting data structure. d Based on median plasma selenium concentration in each quartile of plasma selenium. e Adjusted for age (in weeks), multivitamin treatment regimen, baseline CD4 cell count (o1300 cells/mm3), and time-varying measurements of weight-for-age (zscore, o1.645), plasma albumin (o3.5 g/dl), plasma ferritin (o200 mg/l), plasma vitamin E (o5 mg/ml), and vitamin A (o10 mg/dl). Multivariate models for ‘Total Mortality’ also adjusted for time-varying HIV infection status. f Cutoffs for first selenium measurement (in mg/l): quartile 4, Z77.6; quartile 3, 66.5–77.5; quartile 2, 57.1–66.4; quartile 1, r57.0. Cutoffs for second selenium measurement, if available (in mg/l): quartile 4, Z80.6; quartile 3, 69.1–80.5; quartile 2, 59.0–69.0; quartile 1, r58.9. b c

multivariate models (Table 3). Essentially the same results were obtained for the ‘acute diarrhea’ outcome measure, as well as the ‘dysenteric’ and ‘watery diarrhea’ outcome measures (results not shown). When the respiratory outcome measures were examined, slightly increased risks in the lowest three quartiles were observed for the ‘cough and rapid respiratory rate’ but not for the ‘cough’ outcome measure. However, none of the point estimates were statistically significant. The risks observed for the outcome measures ‘cough and fever’ and ‘cough plus’ were essentially the same as for the ‘cough’ outcome measure (results not shown). In subsequent secondary analyses, the data were stratified by HIV-infected and HIV-uninfected person-time (Table 4). In multivariate analyses among the HIV-infected group, the risk of diarrhea was slightly elevated in the lowest quartile of plasma selenium (RR ¼ 1.36, 95% CI ¼ 0.94–1.95). Subclassifying diarrhea into ‘acute’, ‘watery’, or ‘dysenteric’ diarrhea did not yield results different from the overall diarrhea outcome measure (results not shown). For the ‘cough and rapid respiratory rate’ outcome measure, a nonsignificantly elevated risk was noted in the lowest quartile of selenium concentrations (RR ¼ 1.60, 95% CI ¼ 0.55–4.69), compared to the highest quartile. No such risk elevations were observed for the ‘cough’ outcome measure, which gave results similar European Journal of Clinical Nutrition

to the ‘cough and fever’ and ‘cough plus’ outcome measures (results not shown). In the HIV-uninfected group, low plasma selenium levels were not associated with increased risks for any of the diarrheal outcome measures and any of the cough-related outcome measures, except for ‘cough and rapid respiratory rate’, for which nonsignificantly elevated risks were observed in the lowest three quartiles.

Discussion We found that plasma selenium levels were weakly inversely related to risk of mortality among children born to HIVinfected mothers in Tanzania. While the lowest levels of plasma selenium levels were associated with an elevated risk of the ‘cough and rapid respiratory rate’ outcome measure among both HIV-infected and -uninfected children, as well as an elevated risk of diarrhea among HIV-infected children, there were no statistically significant associations between selenium levels and any morbidity outcome measures. The association between plasma selenium levels and mortality was similar among the overall cohort and among children who acquired HIV over the follow-up period. These


1255 Table 3 Multivariate associations between plasma selenium levels and risk of diarrheal and respiratory outcomes among 610 children born to HIVinfected mothers Multivariate model a Outcome measure

Morbidity rate (episodes/childyear)

Relative risk b (95% CI)

P-value, test for trend c

Diarrhea Selenium Selenium Selenium Selenium

quartile quartile quartile quartile

4 (highest)d 3 2 1 (lowest)

1.56 1.40 1.46 1.52

1.00 0.90 (0.73–1.10) 1.00 (0.82–1.22) 0.99 (0.80–1.23)

0.94

Cough Selenium Selenium Selenium Selenium



4.83 4.48 4.62 4.78

1.00 0.97 (0.87–1.08) 0.98 (0.87–1.10) 1.00 (0.89–1.12)

0.94

Cough and rapid respiratory rate Selenium quartile 4 (highest) Selenium quartile 3 Selenium quartile 2 Selenium quartile 1 (lowest)

0.10 0.14 0.14 0.16

1.00 1.34 (0.77–2.31) 1.27 (0.71–2.29) 1.44 (0.83–2.51)

0.23

a

Adjusted for age (in weeks, using restricted cubic splines with four knots), time-varying HIV infection status, multivitamin treatment regimen, baseline CD4 cell count (o1300 cells/mm3), and time-varying measurements of weight-for-age (z-score, o1.645), plasma albumin (o3.5 g/dl), plasma ferritin (o200 mg/l), plasma vitamin E (o5 mg/ml), and vitamin A (o10 mg/dl). b From generalized estimating equations using a working binomial distribution and a log link function. c Based on median selenium concentration in each quartile of plasma selenium. d Cutoffs for first selenium measurement (in mg/l): quartile 4, Z77.6; quartile 3, 66.5–77.5; quartile 2, 57.1–66.4; quartile 1, r57.0. Cutoffs for second selenium measurement, if available (in mg/l): quartile 4, Z80.6; quartile 3, 69.1–80.5; quartile 2, 59.0–69.0; quartile 1, r58.9.

Table 4 Multivariate associations between plasma selenium levels and incidence of diarrhea and respiratory outcomes among 610 children, stratified by HIV status HIV infected

Outcome measure

HIV uninfected

Morbidity rate (episodes/ childyear)

Relative risk (95% CI) a,b

P-value, test for trend c

Morbidity rate (episodes/ childyear)

Relative risk (95% CI)

P-value, test for trend

Diarrhea Selenium Selenium Selenium Selenium


4 (highest)d 3 2 1 (lowest)

2.22 1.42 1.77 2.72

1.00 0.70 (0.44–1.12) 0.93 (0.63–1.37) 1.36 (0.94–1.95)

0.09

1.40 1.39 1.35 1.15

1.00 1.00 (0.79–1.25) 1.01 (0.79–1.29) 0.82 (0.64–1.05)

0.19

Cough Selenium Selenium Selenium Selenium



5.33 4.68 5.23 6.10

1.00 0.88 (0.68–1.13) 1.02 (0.82–1.26) 1.10 (0.90–1.35)

0.13

4.70 4.44 4.40 4.38

1.00 0.99 (0.88–1.12) 0.94 (0.82–1.08) 0.94 (0.82–1.07)

0.33

0.16 0.18 0.15 0.29

1.00 1.21 (0.41–3.56) 0.90 (0.33–2.46) 1.60 (0.55–4.69)

0.49

0.09 0.13 0.13 0.12

1.00 1.54 (0.83–2.83) 1.56 (0.77–3.16) 1.36 (0.69–2.69)

0.34

Cough and rapid respiratory rate Selenium quartile 4 (highest) Selenium quartile 3 Selenium quartile 2 Selenium quartile 1 (lowest) a

Adjusted for age (in weeks, using restricted cubic splines with four knots), time-varying HIV infection status, multivitamin treatment regimen, baseline CD4 cell count (o1300 cells/mm3), and time-varying measurements of weight-for-age (z-score, o 1.645), plasma albumin (o3.5 g/dl), plasma ferritin (o200 mg/l), plasma vitamin E (o5 mg/ml), and vitamin A (o10 mg/dl). b From generalized estimating equations using a working binomial distribution and a log link function. c Based on median selenium concentration in each quartile of plasma selenium. d Cutoffs for first selenium measurement (in mg/l): quartile 4, Z77.6; quartile 3, 66.5–77.5; quartile 2, 57.1–66.4; quartile 1, r57.0.Cutoffs for second selenium measurement, if available (in mg/l): quartile 4, Z80.6; quartile 3, 69.1–80.5; quartile 2, 59.0–69.0; quartile 1, r58.9.

European Journal of Clinical Nutrition


1256 risk elevations are lower, however, than risk elevations reported in a US study among 24 HIV-infected children, in which low plasma selenium levels (r85 mg/l) were associated with a six-fold increased mortality risk (Campa et al, 1999). The fact that our study population was considerably younger at enrollment (median age 10.5 weeks vs 4.2 y in the US study) may partly account for the differences observed. Furthermore, in contrast to our study population, children in the US study predominantly had symptomatic HIV disease at enrollment and the increased oxidative stress associated with advanced HIV disease may increase the importance of adequate selenium status (de Martino et al, 2001). Lastly, we were able to control for several additional nutrients, which is of relevance since nutrient deficiencies generally coexist. In fact, the association between selenium concentrations and risk of mortality was moderately attenuated when other nutrient (and nonnutrient) terms were added to the ageadjusted univariate models for the overall cohort and for children who acquired HIV over the follow-up period. The potential benefits of selenium for mortality among children born to HIV-infected mothers may be mediated by optimizing immune function and delaying HIV disease progression once infection has occurred. Selenium supports cell-mediated immunity by increasing T-cell proliferation and function (Kiremidjian-Schumacher et al, 1990; Roy et al, 1990). T-cell function is also relevant for humoral immunity due to the interplay between T and B cells in the immune response (McKenzie et al, 1998). Selenium may be beneficial during HIV disease through its effect on the body’s cytokine profile, which mediates the type of immune response mounted. It is thought that HIV disease progression is facilitated by a shift from a T helper 1 (Th1) cytokine pattern, which promotes defense against intracellular microbes, to a T helper 2 (Th2) cytokine pattern, which downregulates the Th1 response and favors defense against helminth and arthropod infections (Clerici & Shearer, 1994). Experimental evidence shows that selenium promotes a Th1 response, and may thus help thwart HIV disease progression (Baum et al, 2000). Selenium is an essential cofactor of the antioxidant enzyme glutathione peroxidase and is thus crucial in the defense against potentially damaging reactive oxygen species (Flohe, 1988). In mouse models, selenium deficiency increases virulence of Coxsackie and influenza viruses, presumably due to oxidative damage to the genetic material (Beck et al, 2003). Findings from these RNA viruses may be of relevance for the RNA-virus HIV (Beck et al, 1995). In T-lymphocytes that are latently infected with HIV, impaired activity of the selenium-dependent antioxidant enzyme glutathione peroxidase increases replication of the virus (Sappey et al, 1994). Furthermore, there is evidence that HIV may encode its own selenoproteins, which may be a mechanism for increased HIV expression in the presence of selenium deficiency (Taylor et al, 1997). The presence of reactive oxygen species may contribute to the etiology of diarrheal disease by promoting hypersecreEuropean Journal of Clinical Nutrition

tion in the small intestine (Lindley et al, 1994; Nieto et al, 2000). It is possible that selenium deficiency contributes to this pathway by failing to detoxify reactive oxygen species. Inadequate selenium status may also increase risk of childhood diarrhea by weakening immunity and thus increasing susceptibility to etiologic agents of childhood diarrhea. For example, impaired antibody responses tend to increase susceptibility to bacterial infections such as Escherichia coli (Abbas & Lichtman, 2003a). Animal studies indicate that selenium is important for resistance to dysentery in swine (Teige et al, 1982). Rotavirus infection, a common cause of childhood diarrhea, may be facilitated by impaired antibody and cytolytic T-lymphocyte defense mechanisms due to selenium deficiency (Abbas & Lichtman, 2003b). HIV disease may potentiate the adverse consequences of impaired selenium status, as illustrated by an increased (albeit nonsignificantly) risk of diarrhea among the HIV-infected children with lowest plasma selenium levels. However, the importance of nutrient status for neonatal and child immunity is complicated by the presence of passive immunity provided by maternal antibodies and the naturally depressed T-cell function in neonates (Lewis & Wilson, 2001; Abbas & Lichtman, 2003b). We observed a weakly increased risk for ‘cough and rapid respiratory rate’ among children with low plasma selenium levels. However, there was low statistical power to detect statistically significant associations. Among HIV-infected adult drug users, low plasma selenium levels are associated with increased risk of pulmonary infection caused by mycobacteria (Shor-Posner et al, 2002). Low plasma selenium levels among presumably HIV-uninfected infants with very low birthweight (birthweight o1500 g) are associated with poor respiratory morbidity, such as chronic lung disease and bronchopulmonary dysplasia (Darlow et al, 1995). These results may partly explain the elevated risk of ‘cough and rapid respiratory rate’ among children with low plasma selenium levels in this cohort. The observational design employed in this study is limited by the fact that it cannot exclude the possibility of unmeasured confounding. However, we attempted to address this concern by carefully measuring and subsequently controlling for important confounding factors. There is also concern that plasma selenium levels may be depressed due to underlying acute disease and/or inflammation even though tissue levels are adequate, in which case selenium levels would represent a marker of these underlying pathologies rather than a causal factor for mortality and morbidity (Hawker et al, 1990; Sattar et al, 1997; Nichol et al, 1998). Evidence from randomized trials could therefore add important additional information on the causality of purported associations. The described associations between plasma selenium levels and mortality and morbidity end points among children born to HIV-infected mothers may apply to setting with similar prevalences of nutritional deficiencies, morbidity patterns and where infant mortality rates are similar to the


1257 160 per 1000 live births that we experienced in our trial cohort. The associations reported may warrant additional investigation in future epidemiological studies in similar settings.

References Abbas AK & Lichtman AH (2003a): Immunity to microbes. In Cellular and Molecular Immunology pp 345–367. Philadelphia: WB Saunders Company. Abbas AK & Lichtman AH (2003b): Effector mechanisms of humoral immunity. In Cellular and Molecular Immunology pp 318–344. Philadelphia: WB Saunders Company. Allain CC, Poon LS, Chan CS, Richmond W & Fu PC (1974): Enzymatic determination of total serum cholesterol. Clin. Chem. 20, 470–475. Andersen P & Gill R (1982): Cox’s regression model counting process: a large sample study. Ann. Statist. 10, 1100–1120. Baum MK, Miguez-Burbano MJ, Campa A & Shor-Posner G (2000): Selenium and interleukins in persons infected with human immunodeficiency virus type 1. J. Infect. Dis. 182 (Suppl 1), S69– S73. Beck MA, Levander OA & Handy J (2003): Selenium deficiency and viral infection. J. Nutr. 133, 1463S–1467S. Beck MA, Shi Q, Morris VC & Levander OA (1995): Rapid genomic evolution of a non-virulent coxsackievirus b3 in seleniumdeficient mice results in selection of identical virulent isolates. Nat. Med. 1, 433–436. Bhutta ZA, Black RE, Brown KH, Gardner JM, Gore S, Hidayat A, Khatun F, Martorell R, Ninh NX, Penny ME, Rosado JL, Roy SK, Ruel M, Sazawal S & Shankar A (1999): Prevention of diarrhea and pneumonia by zinc supplementation in children in developing countries: pooled analysis of randomized controlled trials. Zinc investigators’ collaborative group. J. Pediatr. 135, 689–697. Black RE, Morris SS & Bryce J (2003): Where and why are 10 million children dying every year? Lancet 361, 2226–2234. Campa A, Shor-Posner G, Indacochea F, Zhang G, Lai H, Asthana D, Scott GB & Baum MK (1999): Mortality risk in selenium-deficient HIV-positive children. J. Acquir Immune Defic. Syndr. 20, 508–513. Clerici M & Shearer GM (1994): The Th1–Th2 hypothesis of HIV infection: new insights. Immunol. Today 15, 575–581. Colton T (1974): Statistics in Medicine. Boston, MA: Little, Brown & Co. Cox D (1972): Regression models and life tables. J. R. Stat. Soc. 34, 187–220. Darlow BA, Inder TE, Graham PJ, Sluis KB, Malpas TJ, Taylor BJ & Winterbourn CC (1995): The relationship of selenium status to respiratory outcome in the very low birth weight infant. Pediatrics 96, 314–319. De Leenheer AP, De Bevere VO, De Ruyter MG & Claeys AE (1979): Simultaneous determination of retinol and alpha-tocopherol in human serum by high-performance liquid chromatography. J. Chromatogr. 162, 408–413. de Martino M, Chiarelli F, Moriondo M, Torello M, Azzari C & Galli L (2001): Restored antioxidant capacity parallels the immunologic and virologic improvement in children with perinatal human immunodeficiency virus infection receiving highly active antiretroviral therapy. Clin. Immunol. 100, 82–86. Devlin T & Weeks B (1986): Spline Functions for Logistic Regression Modeling. Cary, NC: SAS Institute, Inc. Diggle P, Liang K & Zeger S (1994): Analysis of Longitudinal Data. London, UK: Oxford University Press. Fawzi WW, Mbise RL, Hertzmark E, Fataki MR, Herrera MG, Ndossi G & Spiegelman D (1999a): A randomized trial of vitamin A supplements in relation to mortality among human immunodeficiency virus-infected and uninfected children in Tanzania. Pediatr. Infect. Dis. J. 18, 127–133.

Fawzi WW, Msamanga G, Hunter D, Urassa E, Renjifo B, Mwakagile D, Hertzmark E, Coley J, Garland M, Kapiga S, Antelman G, Essex M & Spiegelman D (2000): Randomized trial of vitamin supplements in relation to vertical transmission of HIV-1 in Tanzania. J. Acquir. Immune Defic. Syndr. 23, 246–254. Fawzi WW, Msamanga GI, Hunter D, Renjifo B, Antelman G, Bang H, Manji K, Kapiga S, Mwakagile D, Essex M & Spiegelman D (2002): Randomized trial of vitamin supplements in relation to transmission of HIV-1 through breastfeeding and early child mortality. AIDS 16, 1935–1944. Fawzi WW, Msamanga GI, Spiegelman D, Urassa EJ & Hunter DJ (1999b): Rationale and design of the Tanzania vitamin and HIV infection trial. Control Clin. Trials 20, 75–90. Fawzi WW, Msamanga GI, Spiegelman D, Urassa EJ, McGrath N, Mwakagile D, Antelman G, Mbise R, Herrera G, Kapiga S, Willett W & Hunter DJ (1998): Randomised trial of effects of vitamin supplements on pregnancy outcomes and T cell counts in HIV-1infected women in Tanzania. Lancet 351, 1477–1482. Fawzi WW, Msamanga GI, Wei R, Spiegelman D, Antelman G, Villamor E, Manji K & Hunter D (2003): Effect of providing vitamin supplements to human immunodeficiency virus-infected, lactating mothers on the child’s morbidity and CD4 þ cell counts. Clin. Infect. Dis. 36, 1053–1062. Flohe L (1988): Glutathione peroxidase. Basic Life Sci. 49, 663–668. Hawker FH, Stewart PM & Snitch PJ (1990): Effects of acute illness on selenium homeostasis. Crit. Care Med. 18, 442–446. Jacobson BE & Lockitch G (1988): Direct determination of selenium in serum by graphite-furnace atomic absorption spectrometry with deuterium background correction and a reduced palladium modifier: age-specific reference ranges. Clin. Chem. 34, 709–714. Kiremidjian-Schumacher L, Roy M, Wishe HI, Cohen MW & Stotzky G (1990): Selenium and immune cell functions I. Effect on lymphocyte proliferation and production of interleukin 1 and interleukin 2. Proc. Soc. Exp. Biol. Med. 193, 136–142. Kruskal W (1952): A nonparametric test for the several sample problem. Ann. Math. Stat. 23, 525–540. Lewis D & Wilson C eds (2001): Developmental immunology and role of host defenses in fetal and neonatal susceptibility to infection. In Infectious Diseases of the Fetus and Newborn Infant eds Remington R & Klein S, pp 25–138. Philadelphia, PA: WB Saunders Company. Lindley KJ, Goss-Sampson MA, Muller DP & Milla PJ (1994): Lipid peroxidation and electrogenic ion transport in the jejunum of the vitamin E deficient rat. Gut 35, 34–39. McKenzie RC, Rafferty TS & Beckett GJ (1998): Selenium: an essential element for immune function. Immunol. Today 19, 342–345. Miettinen O (1985): Theoretical Epidemiology. New York: John Wiley & Sons. Miller T, Orav E, McIntosh K & Lipshultz S (1993): Is selenium deficiency clinically significant in pediatric HIV infection? Gastroenterology 104, A746. Nichol C, Herdman J, Sattar N, O’Dwyer PJ, O’Reilly DSJ, Littlejohn D & Fell G (1998): Changes in the concentrations of plasma selenium and selenoproteins after minor elective surgery: further evidence for a negative acute phase response? Clin. Chem. 44, 1764–1766. Nieto N, Lopez-Pedrosa JM, Mesa MD, Torres MI, Fernandez MI, Rios A, Suarez MD & Gil A (2000): Chronic diarrhea impairs intestinal antioxidant defense system in rats at weaning. Dig. Dis. Sci. 45, 2044–2050. Roy M, Kiremidjian-Schumacher L, Wishe HI, Cohen MW & Stotzky G (1990): Selenium and immune cell functions. II. Effect on lymphocyte-mediated cytotoxicity. Proc. Soc. Exp. Biol. Med. 193, 143–148. Sappey C, Legrand-Poels S, Best-Belpomme M, Favier A, Rentier B & Piette J (1994): Stimulation of glutathione peroxidase activity decreases HIV type 1 activation after oxidative stress. AIDS Res. Hum. Retroviruses 10, 1451–1461.



1258 Sattar N, Eatock F, Fell GS & O’Reilly D (1997): Selenium: an acutephase reactant? Ann. Clin. Biochem. 34, 437–439. Shor-Posner G, Miguez MJ, Pineda LM, Rodriguez A, Ruiz P, Castillo G, Burbano X, Lecusay R & Baum M (2002): Impact of selenium status on the pathogenesis of mycobacterial disease in HIV-1infected drug users during the era of highly active antiretroviral therapy. J. Acquir. Immune Defic. Syndr. 29, 169–173. Stone CJ & Koo CY (1985): Additive splines in statistics. Proc. Stat. Comp. Sect. Am. Statist. Assoc. 45–48.


Taylor EW, Bhat A, Nadimpalli RG, Zhang W & Kececioglu J (1997): HIV-1 encodes a sequence overlapping env gp41 with highly significant similarity to selenium-dependent glutathione peroxidases. J. Acquir. Immune Defic. Syndr. Hum. Retrovirol. 15, 393–394. Teige J, Tollersrud S, Lund A & Larsen HJ (1982): Swine dysentery: the influence of dietary vitamin E and selenium on the clinical and pathological effects of Treponema hyodysenteriae infection in pigs. Res. Vet. Sci. 32, 95–100. UNICEF. Child Mortality Statistics, http://www.childinfo.org/cmr2.html (Accessed 25 Oct 2004).