Risk of bias in included studies
Vitamin A/beta-carotene in adults with HIV infection
Allocation concealment was adequate in three of the six trials supplementing adults with vitamin A or beta-carotene, and unclear in the remaining three, which did not describe the method of randomisation. Placebo was given to participants, and treatment providers and assessors were blinded to the treatment assignments in all but one trial (Constans 1996). The duration of the follow-up periods ranged from 1 to 12 months and loss to follow-up from 3% to 31%. Intention-to-treat analyses were performed in two trials (Baeten 2002a, Coodley 1996).
There was adequate allocation concealment. Placebo was used and field researchers were blinded to treatment assignments. Forty-six of 400 (11.5%) women did not return for follow-up, and the median time to follow-up was 42 days (range 32 to 445). Participants lost to follow-up had more advanced HIV-1 disease than those who completed the study, and were more likely to be vitamin A deficient. Analyses were on an intention-to-treat basis. A multivitamin arm was added after 100 participants had been enrolled but was not included in the reported analysis.
This was a brief report on a small pilot study (n=52) in a hospital clinic setting of selenium or beta-carotene or no supplements. The method of randomization was not stated, and it was not stated whether outcome assessors were blinded. It was reported that the beta-carotene recipients had more advanced disease at baseline than the other participants. Four of 52 participants (7.7%) were lost to follow-up at 12 months.
The method of allocation concealment was adequate in this crossover trial. Placebo was used and blinding of patients and assessors was performed.
The method of randomization was not stated and the adequacy of allocation concealment is therefore unclear. Placebo was used and assessors were blind to the treatment assignments. Twenty-two of 72 participants (31%) were lost to follow-up at three months. Results were analysed on an intention-to-treat basis.
The method of randomization was not stated. Placebo was used and assessors were blind to the treatment assignments. Only one of the 40 participants was lost to follow-up at two months. An intention-to-treat analysis was not performed.
Allocation concealment was adequate and placebo was used in this double-blind trial. Ten of the 120 participants (8.3%) were lost to follow-up at one month. An intention-to-treat analysis was not performed.
Other micronutrients in adults with HIV infection
The method of randomization was not described in two of the four trials of other micronutrients to adults, and the adequacy of allocation concealment was judged adequate in only one trial (Jiamton 2003). Placebo was used in all trials. The duration of the follow-up periods ranged from four weeks to 12 months and loss to follow-up from 16% to 28%. Intention-to-treat analyses were conducted in two of the trials (Allard 1998a; Jiamton 2003).
Randomization was by means of a random-number table, but the method of allocation concealment was not clear from the trial report. Placebo was used and outcome assessors were blinded. Nine of 49 participants (18.4%) had baseline and follow-up measurements (one or two months) but were unable to keep their three-month appointments. An intention-to-treat analysis was performed.
This trial was poorly reported and contained several errors. The method of randomisation was not stated, and the adequacy of allocation concealment is therefore unclear. Placebo was used and assessors were blind to the treatment assignments. Data regarding hospital admissions were not available for 73 of 259 participants (28%); excluded participants were reportedly not significantly different from the 186 included patients with respect to socio-demographic and disease status. No intention-to-treat analysis was performed.
Randomization was computer-generated and allocation concealment was adequate. Placebo was used and study physicians were blinded to the treatment assignments. Seventy-nine (16%) of 481 trial participants were lost to follow-up at 48 weeks and 23 (5%) died. An intention-to-treat analysis was performed.
The method of randomization was not described, and the adequacy of allocation concealment is therefore unclear. Placebo capsules were not identical to the micronutrient capsules but were not identifiable by any markings. Twenty-nine of 135 participants (21%) were lost to follow-up at four weeks, and sixty (44%) at 12 weeks. No intention-to-treat analysis was performed.
Multivitamins in pregnant and lactating women with HIV infection
A trial among 1078 pregnant and lactating women in urban Tanzania (Fawzi 1998) was designed to examine the effects of daily supplements of vitamin A or multivitamins (B, C, and E) or both on disease progression and on vertical transmission of HIV. A 2x2 factorial design was used to maximize the use of limited resources. Randomization was performed in blocks of 20 participants, and concealment of allocation (pre-labelled bottles) was adequate. Placebo was used and assessors were blinded. All children received a large dose of vitamin A every six months because of evidence of its protective effect against mortality. Vitamin A was dropped from two of the regimens in September 2000 and replaced with placebo because it was found to be associated with increased vertical transmission of HIV.
There was monthly follow-up at a study clinic or at home if necessary for a minimum of 18 months. Women were questioned about signs of HIV-related complications, and weight, height, and mid-upper arm circumference were determined. The WHO stage of HIV was assessed on the basis of the history and physical examination. Verbal autopsy techniques by means of standardized interviews, review of medical records or both were used to approximate causes of death, which were then ascertained in a blinded fashion.
Analyses were on an intention-to-treat basis, but it was not clear whether this applied to all-cause mortality as well. Data were reportedly censored at the time of death when cause of death could not be ascertained or was deemed to be not AIDS-related (100 out of 343 deaths from all causes over the entire duration of the study), but the basis for allocation of 'AIDS-related deaths' was unclear. Fifty-four of 1078 participants (27 from each group) were lost to follow-up by the time of delivery and were excluded from the analysis of birth outcomes. Subsequent losses to follow-up (other than deaths) were not described.
Vitamin A in children with HIV infection
Placebo was used in all four trials of vitamin A supplements given to children, and all were reported as double-blind. The method of randomization was not stated in one of the trials, allocation concealment was adequate in one trial, and intention-to-treat analyses were conducted in two trials.
The study population in this trial comprised 118 offspring, 28 of whom were HIV-infected, of HIV-infected women. A table of random numbers was used for randomization of the children, but the adequacy of allocation concealment was unclear. Placebo was used and investigators were blinded to the treatment group of the participants. One-third of the participants were lost to follow-up by 18 months. An intention-to-treat analysis was performed.
There was monthly morbidity recall; multiple episodes of the same condition in a single month were counted as one episode (per 100 months). Diarrhoea was defined as four or more loose watery stools per day; persistent diarrhoea was defined as episodes lasting >= seven days. Upper respiratory tract infection was defined as the presence of one or more of the following: rhinitis, throat, or ear infection, cough. Lower respiratory tract infection was defined as presence of cough with one or more of the following: rapid breathing, chest indrawing, crackles, or wheezing.
The study population in this trial was 687 children aged six months to five years hospitalized for pneumonia. HIV-infected children comprised fewer than a tenth (n=58). Block randomization using pre-numbered bottles was carried out, and concealment of allocation was adequate. Placebo was used and assessors were blind to the treatment assignments.
One-quarter of participants were lost to follow-up at four months, and results were analyzed on an intention-to-treat basis. Cause of death was ascertained by means of a review of the hospital records and home verbal autopsy questionnaires. There was bi-weekly morbidity recall. Acute diarrhoea was defined as < 14 days of watery or dysenteric stools, classified as either mild or severe. No data were provided on persistent diarrhoea among HIV-infected children. Respiratory infection was defined either as cough alone, cough and fever, or cough with rapid respiratory rate (>=50 breaths per minute for infants and >40 breaths per minute for children > one year) on the day of the visit.
Randomization was computer-generated and allocation concealment was adequate. Placebo was used and the trial was reported as double-blind. Only one of the 59 participants was lost to follow-up at one month. No intention-to-treat analysis was performed.
The method of randomization was not stated and the adequacy of allocation concealment is therefore unclear. Placebo was used and assessors were blind to the treatment assignments. One of the 75 participants was lost to follow-up at two months. Children who received vitamin A were more immuno-suppressed than those in the placebo group.
Effects of interventions
Vitamin A/ Beta-carotene in adults with HIV infection
There was no overall effect of either vitamin A or beta-carotene on mortality, morbidity, viral load, or on CD8 cell counts in the six trials of vitamin A or beta-carotene that were included in the review. There were significant increases in serum levels in all the beta-carotene supplemented groups.
A one-year pilot study (n=52) of three treatments (beta-carotene, selenium, and no supplements) found no improvements in mortality, morbidity (opportunistic infections), or CD4 counts for either treatment regimen (Constans 1996). Two trials of a single high dose of vitamin A, one among 40 women of reproductive age (Humphrey 1999), and the other among 120 intravenous drug users (Semba 1998), found no effects up to two months on CD4 counts or viral load. The baseline prevalence of vitamin A deficiency (< 30 microgram/dl) was under 20%, and mean serum levels remained unchanged at four weeks.
A trial of vitamin A daily for six weeks to 400 Kenyan women with a high prevalence (59%) of vitamin A deficiency at baseline reported a slight increase in CD4 counts at follow-up, but this effect was not retained in multivariate analysis. There was no effect on CD8 counts or viral load. There was a significant increase in serum levels of vitamin A at six weeks in the supplemented group (median: 29.4 versus 26.8 microgram/dl; p<0.03), but not among a severely deficient (< 20 microgram/dl) sub-group (Baeten 2002a).
In a crossover trial among 21 American outpatients (Coodley 1993), high doses of beta-carotene supplements for one month (180mg/day) increased the CD4 counts and CD4/CD8 ratios compared to placebo. A later extended evaluation by the same author, where beta-carotene and multivitamins were administered for three months, found no effect on CD4 or CD8 counts, nor on measures of body weight or quality of life (measured by Karnofsky scores) (Coodley 1996).
Minor adverse events, such as the discoloration of skin, were reported in a beta-carotene supplemented group (Coodley 1993). Humphrey et al reported adverse events (nausea, headache, vomiting, diarrhoea, and fever) in a vitamin A trial, but the incidence did not differ significantly from the placebo group (Humphrey 1999).
Other micronutrients in adults with HIV infection
Among 135 adults with persistent diarrhoea in Zambia, a daily oral supplement of multi-micronutrients (vitamins A, C, and E, selenium, zinc) for two weeks, in addition to albendazole, had no effects on mortality or on CD4 cell counts after four weeks. No differences in diarrhoeal morbidity (time with diarrhoea following completion of treatment), body mass index (BMI), Karnofsky scores or vitamin A levels were observed during a 12-week follow-up period. At baseline, 63% of participants had vitamin A concentrations compatible with vitamin A deficiency (< 20 microgram/dl), and 55% were classified vitamin E deficient (< 500 microgram/dl). Vitamin A levels did not increase significantly in supplemented patients compared with placebo (Kelly 1999).
One small (n=49) Canadian trial of large daily supplements of vitamins E and C for three months reduced measures of oxidative stress (serum lipid peroxides, serum malondialdehyde, and breath pentane) but had no significant effect on HIV viral load over three months. No significant differences in the number of new or recurrent AIDS-defining, HIV-associated or other infections (bronchitis, line sepsis, otitis media) were found during a six-month follow-up period. The vitamin E and C serum levels at three months were significantly higher in the treatment group, (mean (+/- SD) micromol/l): Vitamin E 30(21.9) versus 1.3 (5.3), p<0.005; Vitamin C 33.3 (27) versus 6 (24.5), p< 0.005), but no change was noted in the levels of vitamin A, carotenoids, zinc and selenium (Allard 1998a).
In a trial of 481 HIV-infected individuals in Thailand, daily supplementation with a comprehensive mix of vitamins and minerals at dosages up to 20 times higher than the recommended daily allowance had no effect on mortality of all enrolled participants. A trend towards a reduced death rate in those with CD4 counts <200x106/l at baseline was reported after 48 weeks of follow-up (mortality hazard ratio = 0.37; p= 0.052) and a significant effect in individuals with a CD4 count less than 100 x106/l, where mortality was reduced by 74% (HR = 0.26; 95%CI: 0.07-0.97; p=0.03). CD4 cell counts, viral load or rates of hospital admissions were not affected (Jiamton 2003).
Burbano 2002 found that daily selenium supplements for 12 months among drug users (n = 186) had no significant effect on the numbers of patients admitted for all conditions (RR = 0.89; 95% CI: 0.58-1.36). Selenium did reduce the absolute risk of admissions for opportunistic infections, HIV-related conditions and psychiatric disorders by 18.6% (RR = 0.40; 95% CI: 0.21-0.75), and the mean number of admissions per patient with these conditions (WMD = 0.24; 95% CI:-0.46, -0.02). Significantly fewer patients in the selenium group (RR = 0.53; 95% CI: 0.35-0.81) experienced a decline in CD4 counts greater than 50 cells/mm3, and fewer had lowered serum levels of selenium (< 135 microgram/L) at 12 months (RR = 0.53; 95% CI: 0.42-0.67).
Reported side effects of supplementation in the above trials included discoloration of urine, nausea, drowsiness, dizziness, headache, rash, and epigastric discomfort, but no significant differences apart from the urine discoloration were detected between treatment groups (Allard 1998a; Jiamton 2003). Kelly et al (Kelly 1999) reported no adverse events and the other trials made no mention of them.
Multivitamins in pregnant and lactating women with HIV infection
In a large Tanzanian trial in pregnant and lactating women (Fawzi 1998), multivitamins, mainly vitamins B, C and E, at doses up to 22 times recommended daily intakes, resulted in significant reductions over the entire period of follow-up (mean = 61.4 months for survival, median = 60 months for stage of disease) in the risk of progression to stage 4 disease or mortality from AIDS-related causes (Reported Hazard Ratio = 0.71; 95% CI: 0.51-0.98); in the risk of progression to stage 4 disease (Reported HR = 0.50; 95% CI: 0.28-0.90); and in the risk of progression to stage 3 disease or higher (Reported HR = 0.72; 95% CI: 0.58-0.90). Multivitamins also reduced maternal viral load and all signs of HIV-related complications, and raised maternal CD4 and CD8 counts, compared to no multivitamins (vitamin A only or placebo). The effects of vitamin A only were smaller and generally not significantly different from placebo. Adding vitamin A to the multivitamin regimen reduced the benefit with regard to all maternal outcomes. There was a modest improvement in average gestational weight gain during the third trimester in the multivitamin group (mean difference compared to no multivitamins = 304 g; 95% CI: 17- 590 between week 27 and delivery), and a reduced risk of low rate of weight gain (<= 100g per week) (RR = 0.73; 95% CI: 0.58- 0.93).
Multivitamins reduced adverse pregnancy outcomes, including foetal deaths by 39% (RR = 0.61; 95% CI: 0.39-0.94), low birth weights (LBWs) by 45% (RR = 0.55; 95% CI: 0.38-0.81), severe preterm births (<34 weeks) by 39% (RR = 0.61; 95% CI: 0.38-0.95), and those who were small for gestational age (SGA) at birth by 43% (RR = 0.57; 95% CI: 0.40-0.83). Mean birthweight was significantly higher in infants born HIV-negative to mothers in the supplemented group (mean difference = 94 g, p = 0.02 in multivariate analysis).
In women with low immunological or nutritional status, significant reductions were reported in child mortality by 24 months: 70% (RR = 0.30; 95% CI: 0.1-0.92) in women in the lowest quartile of lymphocyte counts (<1340/ mm3); 60% (RR = 0.40; 95% CI: 0.17-0.98) in women with low baseline vitamin A (<20 microgram/dl); and 69% (RR = 0.31; 95% CI: 0.13-0.73) in women with low baseline vitamin E (<9.7 micromol/l). No effects were seen when these outcomes were analysed according to CD4 and CD8 counts.
Maternal supplementation during breastfeeding reduced the risk of diarrhoea up to 24 months in all children by 17% (RR = 0.83; 95% CI: 0.71-0.98), and it raised their CD4 counts (WMD = 153.00 cells/microl; 95% CI: 67.6- 238.4) by six months of age.
Vitamin A in children with HIV infection
Large doses of vitamin A given intermittently to children aged 6-59 months admitted with pneumonia to a Tanzanian hospital (n = 687) reduced all-cause mortality by 49% (RR = 0.51 ; 95% CI: 0.29-0.90) and deaths from diarrhoea by 93% (RR = 0.07 ; 95% CI: 0.00-0.49) during a two-year follow-up period. In 58 HIV-infected children (9% of the total), all-cause mortality was reduced by 63% (RR = 0.37; 95% CI: 0.17-0.84) and the AIDS-related deaths by 68% (RR = 0.32; 95% CI: 0.1-0.99 compared to placebo. A marginal reduction in cough and rapid respiratory rate was reported (RR = 0.54; 95% CI: 0.24-1.20), but there was no effect on the risk of acute diarrhoea. After four months of supplementation, the mean difference in attained height between HIV-infected children under 18 months of age and controls was 2.8 cm (95% CI: 1.0-4.6) (Fawzi 1999).
Supplementation at 3-month intervals of all children born to a population of 118 HIV-infected women in South Africa, reduced all-cause morbidity by a third (OR = 0.69; 95% CI: 0.48-0.99) during 18 months of follow-up post-partum. In 28 children who were known to be HIV-infected at birth, diarrhoea-associated morbidity was reduced by 38% (RR = 0.62; 95% CI: 0.39-0.98) (Coutsoudis 1995).
Vitamin A given to 59 HIV-infected American children before influenza vaccination did not enhance serologic responses, but did dampen the increase in the HIV viral load 14 days after immunisation (p = 0.02) (Hanekom 2000). In 75 children with AIDS attending a South African children's hospital, vitamin A supplementation was associated with increased CD4 counts at four weeks (p = 0.03) (Hussey 1996).
Adverse effects of vitamin A in children included vomiting and bulging fontanelle (Coutsoudis 1995); an increased risk of cough and rapid respiratory rate, and an increased risk of diarrhoea in normally nourished (RR = 1.3 ; 95% CI: 1.06-1.79) and growth-stunted children (RR = 1.84 ; 95% CI: 1.16-2.90) (Fawzi 1999).
Other micronutrients in children with HIV infection
There have been no studies published to date of the effect of supplements such as zinc, given alone, or in combination with vitamin A or other micronutrients, in HIV-infected children with acute or persistent diarrhoea.