The protocol for this trial and supporting CONSORT checklist are available as supporting information; see Checklist S1
and Protocol S1
The study was conducted in northern KwaZulu-Natal Province, South Africa. Median household income in the area is one-half the national average, 80% of the population lives below the national poverty line, and unemployment is 75%.
There is little subsistence farming in this area despite 98% of households being defined as rural, and food insecurity within households is common.
Most infants receive both breast and complementary feeding from early infancy,
and the diet is known to be deficient in micronutrients.
Children were enrolled into the study by nurses at five government primary health care clinics. Study subjects were recruited from children coming to the clinics for routine visits, from households identified as having potentially eligible children based on surveillance conducted by the Africa Centre Demographic Information System, and from children enrolled at birth in an observational study of the determinants of mother-to-child transmission of HIV.
Children eligible for study were four to six months old. Children were excluded from study if they were: less than 60% of median weight-for-age using United States National Center for Health Statistics standards;
had nutritional edema; had received vitamin or micronutrient supplements in the previous month; had diarrhea for more than seven days at the time of study enrollment; or were enrolled in another study of a clinical intervention.
To maintain community blinding to the HIV status of women enrolled in the study, we enrolled children in any cohort which reached targeted enrollment before enrollment was completed in other cohorts, provided them only their initial two weeks of supplement, and did not include them in the analysis.
Ethics and study monitoring
The study was approved by the Ethics Review Committee of the University of KwaZulu-Natal and the Institutional Review Board of Tufts-New England Medical Center. Before initiation the study was presented to the Community Advisory Board of the Africa Centre for their review and advice, and interim reports on study progress were provided to this Advisory Board. The study was also monitored by a Data Monitoring and Safety Board established by the NIH. Written informed consent to participate in the study, and to do HIV testing on both the mother and child, was obtained from all mothers whose children participated in the study.
The three treatment arms were: vitamin A alone; vitamin A plus zinc; and vitamin A, zinc and multiple micronutrients. All supplements were given daily at home from entry into the study until 24 months of age. Vitamin A was given as the comparator regimen, rather than placebo, because the South African Department of Health guidelines recommend routine three-monthly vitamin A supplementation for children and clinics did not consistently provide vitamin A to children. Thus each of the three tablet formulations contained 1250 IU of vitamin A. Two of the three tablet formulations contained 10 mg of zinc as zinc gluconate. The multiple micronutrient tablets were similar to those used in recent international trials of multiple micronutrient supplementation,
and in addition to vitamin A and zinc contained : 0.5 mg each of vitamins B1, B2 and B6; 0.9 µg vitamin B12; 35 mg vitamin C; 5 µg vitamin D; 6 mg vitamin E; 10 µg vitamin K; 0.6 mg copper as cupric gluconate; 150 µg folate; 50 µg iodine; 10 mg iron as ferrous fumerate; and 6 mg niacin. The supplements were manufactured as tablets by Hersil Ltd. (Lima, Peru) and packaged in blister packs of seven tablets. All three formulations were similar in color, taste, appearance and size. Because of a delay in shipment, 243 children enrolled in the study did not receive supplements for 11 weeks in 2005.
To study objective was to determine if zinc, or zinc plus multiple micronutrients, would reduce diarrhea and respiratory disease prevalence. We hypothesized that both zinc, and zinc plus multiple micronutrients, would reduce diarrhea and respiratory disease prevalence and examined their efficacy in a randomized controlled trial in children 6–24 months of age born to both HIV-infected and HIV-uninfected mothers.
The primary study outcome was the percentage of days of diarrhea per child compared between study arms within each of the three cohorts. Secondary diarrheal disease outcomes were measures of the severity of diarrhea (duration of episodes, maximum number of stools during an episode, episodes lasting 14 or more days, diarrhea episodes with blood in the stool, and clinic visits for diarrhea) and the distribution of diarrheal morbidity (proportion of children who ever had diarrhea and number of episodes per child). Respiratory disease outcome measures were the percentage of weeks of upper respiratory symptoms per child, the percentage of children who ever had pneumonia by maternal report, and the percentage of children with pneumonia confirmed by the field worker.
Diarrhea was considered to be present on any day the stool was more frequent or looser than normal or the stool contained water or blood. An episode of diarrhea terminated on the last day of diarrhea that was followed by two consecutive days without diarrhea. Blood was considered present in stool if reported by the mother. Upper respiratory tract infection was defined as the history (or presence during the field worker visit) of cough or runny nose. Because field workers asked about the occurrence of these signs since the last visit, rather than by day, the outcome was defined as prevalent weeks. Pneumonia by maternal report was considered to have occurred if there was a history of either fast breathing or chest in-drawing. Confirmed pneumonia was defined as an elevated respiratory rate at rest measured by the fieldworker using WHO/UNICEF Integrated Management of Childhood Illness guidelines.
This was done separately for each of the three cohorts and was based on percentage of days of diarrhea for each child by cohort and treatment arm. We assumed the percentage of days with diarrhea in HIV-infected children receiving vitamin A alone would be 4.4% (SD±0.9%); for those receiving vitamin A and zinc it would be 3.7%±0.9% (17% less than for those receiving vitamin A alone); and for those receiving multiple micronutrients it would be 3.0%±0.9% (33% less than for those receiving vitamin A alone and 20% less than for those receiving vitamin A and zinc). For HIV-uninfected children we postulated the same differences between treatment groups but with lower diarrhea prevalence rates: 3.3%±0.9% (vitamin A alone); 2.8%±0.9% (vitamin A and zinc); and 2.2%±0.9% (multiple micronutrient group). To determine with a power of 0.8 and a two-sided significance level of .05 if these postulated differences existed, 26 children were required in each of the three arms of the HIV-infected cohort, and 45 children in each of the three arms of the two cohorts of HIV-uninfected children. We expected mortality, migration, and withdrawals to diminish days of observation by 50% in the HIV-infected cohort and 25% in the HIV-uninfected groups. Thus the total sample size required was 156 in the HIV-infected cohort and 180 children in each of the two cohorts of HIV uninfected children.
An allocation list was prepared using computer-generated random numbers and a block size of six. Assignment to the three treatment arms was done separately for three cohorts of children stratified by HIV status of child and mother: HIV-infected children and mothers; HIV-uninfected children of HIV-infected mothers; and HIV-uninfected children of HIV-uninfected mothers.
Randomization–allocation concealment and implementation
The manufacturer prepared numbered packs of tablets corresponding to the allocation list. Children enrolled in the study were assigned by a study physician to one of the three study cohorts after results of the HIV tests became available. The physician then allocated the next pack of tablets from the blocks assigned to that cohort to the participant.
Investigators, study staff and participants were blind to the treatment assignments.
At enrollment study subjects had a physical examination and medical history taken by the study nurse. Weight was measured to a precision of 0.1 kg using an electronic digital scale and recumbent length to a precision of 0.1 cm using an infantometer. For determination of anthropometric centiles National Center for Health Statistics standards were used.
Mothers and children had a blood sample taken for determination of HIV infection status if their status was not already known and documented, and for determining haemoglobin concentration. A two-week supply of tablets were provided at enrolment, and the mothers instructed on administering tablets to their child by crushing it and mixing it with food, and assuring that the child ate all the food the supplement was mixed in.
Trained fieldworkers visited the children at home weekly during the study. They supplied sufficient tablets to ensure a two week supply was on hand. Any tablets remaining from the previous supply were counted and recorded. At each home visit field workers observed the preparation and administration of a tablet, and waited for 30 minutes to observe any adverse effect.
At each visit field workers administered a standardized questionnaire that focused on diarrheal and respiratory morbidity using locally validated terms.
If the mother reported diarrhea then the field worker asked for each day since the last visit if the stool was looser or more frequent than normal, contained water or blood, and the number of stools on the worst day of diarrhea. Mothers were asked if since the last visit the child had a runny nose or cough, difficult or fast breathing, or chest in-drawing. Field workers measured the respiratory rate at rest in any child with current or reported signs or symptoms of respiratory distress.
If the mother or another informant familiar with the child was not available during the scheduled home visit two further attempts were made to visit that week, following which the next planned weekly visit was made. Field workers collected information on events that occurred a maximum of 28 days before the visit.
Mothers and infants were requested to return for clinic visits at 7, 8, 9, 12, 15, 18, 21 and 24 months of age to assess growth and morbidity. Ill children were referred to the study nurses at the clinics, or for more serious illness to the district hospital. When antiretroviral therapy for children became available in clinics in January 2005 all children with HIV infection in the study were referred for evaluation.
Supervisors assessed reliability of field worker assessments by conducting unscheduled visits and comparing findings with interviews conducted the same day by field workers
HIV testing of children was done using a quantitative HIV RNA assay (Nuclisens HIV-1 QT, Organon Teknika or Nuclisens EasyQ HIV-1, Biomerieux, Boxtel, The Netherlands). Maternal HIV status was determined using two ELISA tests (first Vironostika HIV-1 Microelisa system and then Uni-Form II plus O if the first test was positive, both Biomèrieux) each of which had to be positive for HIV infection to be diagnosed. Hemoglobin determination was done at the clinic by study nurses using a portable HemoCue system (Angelholm, Sweden).
Data were entered into an electronic data base using an optical form scanning system (TeleForm version 7.1 Cardiff, Vista CA, USA) and were analyzed using SPSS version 13.0 (SPSS Inc., Chicago, IL, USA) and EpiInfo version 3.3.2 (Centers for Disease Control and Prevention, Atlanta, GA, USA).
Analyses were done on a locked dataset using an analysis plan, and tests of significance, that were determined before the analyses were conducted. Analyses were initially conducted using coded treatment groups with the analysts blinded to the actual treatment.
Analysis of all outcomes was on an intention-to-treat basis–all participants for whom there were observations on study outcomes were included in the analysis. In addition to comparing baseline characteristics between children in the treatment groups, we also examined features during study that could have confounded study outcome. These were rates of drop-out and death, proportion of missing information, dietary pattern, adherence to treatment, and intake of other vitamin and micronutrient preparations. Any baseline or during treatment characteristics that differed significantly between treatment arms was tested for its association with the study outcome using Spearman's rank correlation test.
The primary outcome–percentage of days of diarrhea per child by cohort and treatment group-was not normally distributed as determined by the Kolmogorov-Smirnov test, and a Kruskall-Wallis test was used to determine the overall significance of difference between treatment arms. For secondary outcomes and for baseline characteristics the significance of differences for continuous variables was tested using one-way ANOVA if the variable was normally distributed or the Kruskall-Wallis test if not normally distributed. Differences in proportions were tested with the χ2
test. Comparisons between individual groups were made only if the overall group comparison found a significant difference. To determine the effect size for the primary outcome, we determined the difference between medians, and the 95% confidence intervals for those differences, using the binomial method.
All tests of differences or association were two-sided, and were considered significant if the P value was ≤ 0.05.
We also tested for any bias in assignment to treatment cohort (on the assumption that blinding may not have been completely effective) using the Berger-Exner test.