Several observational studies have reported that a number of micronutrient deficiencies among HIV-1 infected patients are associated strongly with a faster progression to disease and death [1–5]. This is thought to be either through increased oxidative stress and enhanced viral replication [6–8] or through a reduction in the number of circulating T lymphocytes [5,9–11]. The impact of micronutrient supplementation on surrogate markers of disease stage has been examined in two randomized placebo-controlled trials: in one large trial conducted in Tanzania, multivitamin supplementation during pregnancy led to significantly greater increases in CD4 cell count of 49% and 67% at 6 and 30 weeks postpartum respectively , while in a small trial conducted in Canada, supplementation with vitamins E and C reduced oxidative stress and resulted in approximately 1 log10 copies/ml lower HIV plasma viral load although this did not reach statistical significance . Whether micronutrient supplementation might provide clinical benefit and reduce death rates from HIV/AIDS has not been tested. This is a question of high public health importance as micronutrients are cheap and easily tolerated. We conducted a large randomized placebo-controlled trial of commercially available micronutrients, which were rich in antioxidants, among HIV-infected individuals in Thailand. The aim was to measure the impact of supplementation on mortality and hospital admissions. We also measured impact on CD4 cell count and HIV plasma viral load.
The trial was conducted at the Departments of Dermatology and Obstetrics and Gynaecology, Siriraj Hospital, Mahidol University, Bangkok, Thailand. The hospital has more than 2500 beds and about 1.2 million outpatient visits each year. Management of trial participants was integrated into the routine clinical practices of the hospital. HIV-infected individuals living in the community were contacted through various patients’ representative organizations and invited to come forward for the trial. A few patients attending hospital for routine testing for sexually transmitted infections also volunteered. Individuals were eligible for the trial if they were over 18 years old, had not been taking micronutrients or antiretrovirals in the last 30 days and had a CD4 cell count between 50 × 106 and 550 × 106/l. Their HIV status was confirmed from hospital records or HIV antibody testing was repeated.
The micronutrients comprised a comprehensive mix of vitamins and minerals (Immunace®, Vitabiotics Ltd, London, UK) and corresponded to daily doses of vitamin A 3000 μg, betacarotene 6 mg, vitamin D3 20 μg, vitamin E 80 mg, vitamin K 180 μg, vitamin C 400 mg, vitamin B1 24 mg, vitamin B2 15 mg, vitamin B6 40 mg, vitamin B12 30 μg, folacin 100 μg, panthothenic acid 40 mg, iron 10 mg, magnesium 200 mg, manganese 8 mg, zinc 30 mg, iodine 300 μg, copper 3 mg, selenium 400 μg, chromium 150 μg and cystine 66 mg. The dosages used daily were higher than the recommended daily allowances (RDA) for healthy individuals , the rationale being that HIV-1 infected individuals have greater requirements for micronutrients . Trial participants were asked to take one tablet twice a day after food. Placebo tablets, which were identical in appearance and size, contained dibasic calcium phosphate and were coated with iron oxide. Participants were randomized to one of 10 codes, with five codes corresponding to the micronutrients and five to placebo, using computer generated random numbers with block size 10. The micronutrients and placebo tablets were packaged in small plastic bottles by the manufacturer who held the code. The code was printed on the bottles as part of a six-digit number, the other five being random numbers designed to introduce further blinding of trial participants. The study physicians knew which number corresponded to the code but were kept unaware of whether the bottle contained micronutrients or placebo. Trial participants were provided with pills for 100 days. The randomization was stratified by CD4 cell count of < 200 × 106/l or ≥ 200 × 106/l.
Procedures at baseline
A physical examination was conducted by one of the study physicians. Based on sample size calculations, the first 140 consecutive trial participants were selected for further laboratory testing. Ten millilitres of blood were taken following further written informed consent and tested for HIV plasma viral load and plasma levels of vitamin E and selenium.
Procedures at follow-up
Trial participants were asked to return to hospital at any time that they were unwell and routinely every 12 weeks. A physical examination and pill counts were conducted at each visit and counselling on the need for taking study medications and on eating a balanced diet was provided. Data were collected by study physicians on the use of antiretrovirals, cotrimoxazole and other prophylaxis. CD4 cell count testing was repeated at 24 and 48 weeks. Testing for HIV-1 viral load and levels of micronutrients in plasma were repeated at 48 weeks. Transport fare was provided to trial participants at each visit. Those who failed to return for a scheduled appointment were followed up by telephone and letters.
Between hospital appointments, trial participants were telephoned at home every 4 weeks by trained nursing staff to ascertain their general health status. Trial participants who did not have access to a private phone were given telephone cards and asked to telephone the research team at these times. All reported admissions and diagnoses of AIDS defining illnesses were verified by contacting the hospital that the individual had visited. Deaths were ascertained from the next of kin, details of whom had been collected from trial participants, and verified from death certificates. Causes of deaths were taken from the trial participant's death certificate, which was usually completed by a general practitioner if the individual died at home, or from hospital records if the individual died in hospital.
Recruitment began on 8 March 2000 and ended on 16 January 2001. Follow-up of trial participants ended on 30 September 2001. After 48 weeks of follow-up, at the request of trial participants, we provided a locally available mix of micronutrients. From the end of the follow-up, trial participants were provided with the micronutrients used in the trial, a supply which was to be provided for life should the trial demonstrate potential benefit.
Sample handling and laboratory methods
CD4 cell count was estimated by FACScan flowcytometry (Becton-Dickinson, Rutherford, New Jersey, USA) from whole blood stored in EDTA medium. The presence of HIV antibodies were tested using two different enzyme immunoassays (AxSYM HIV1/2 gO, Abbott GmbH Diagnostika, Weisbaden, Germany and Vironostika HIV Uniform II Plus O, Organon Teknika, Boxtel, the Netherlands), with confirmatory testing by Western blot (HIV-BLOT 2.2, Genelabs Diagnostics, Singapore).
Blood samples from trial participants in the viral load studies were stored at 4°C and sent to the laboratory within 4 h. Fresh whole blood was separated into plasma and cellular fractions, aliquoted and stored at −80°C. HIV viral load was tested using the NucliSens assay (Organon Teknika), which has a limit of detection of 40 copies/ml.
Plasma levels of vitamin E (α-tocopherol) were measured by high performance liquid chromatography (HPLC) (Waters, Millipore Co., Billerica, Massachusetts, USA)  and selenium by atomic absorption (Avanta model E PAL3000, GBE Scientific Equipment, Victoria, Australia). Vitamin E deficiency was categorized as < 11.6 μmol/l [16,17]. Plasma levels of selenium indicating deficiency have not been defined but normal levels among HIV-uninfected adults in the USA are between 1.5 and 1.7 μmol/l; deficiency was defined as < 1.5 μmol/l .
The study was approved by the ethics committees of the London School of Hygiene and Tropical Medicine, Siriraj Hospital, and the Ministry of Public Health, Thailand.
Statistical power and analysis
Assuming that 20% of trial participants would die in the placebo group over a period of 48 weeks , then a trial with 200 individuals per arm would have had just under 80% power to detect a reduction in mortality by half in the supplementation group. A trial of this size was also estimated to have more than 90% power to detect as statistically significant a difference of 20% or more in mean CD4 cell count between the placebo and supplementation group assuming that the mean CD4 cell count in the placebo group would be 250 (SD, 150) cells/μl [12,18]. Assuming that the mean viral load would increase by 0.5 log10 copies per ml (SD, 0.4) in the control group , then 50 individuals per group would have provided more than 90% power to detect as statistically significant a mean viral load increase of 0.2 log10 copies/ml or less in the supplementation group (equivalent to a twofold difference in increase in mean viral load between the two groups). In order to allow for losses to follow-up, an attempt was made to recruit 500 individuals to the trial, of whom the first 140 would also be enrolled into the viral load study.
A plan of analysis was decided before the code was broken. This included examining the impact of micronutrients according to baseline CD4 cell count. Continuous data were compared between groups using t tests if the data were distributed approximately normally or by the Wilcoxon test. Categorical data were compared by the χ2-test. Survival analysis was by Poisson and Cox regressions with follow-up time calculated for each individual from the date of randomization to either the date of the final follow-up, the date of death or the date that the individual was lost to follow-up. Statistical significance was assessed using likelihood ratio tests. All statistical tests were two-sided.
Characteristics of trial participants at recruitment
Eight-hundred individuals were screened and 481 who returned after the initial screening and were eligible were invited to join the trial; none refused (Fig. 1). Table 1 shows their characteristics, which were similar between the two arms of the trial. Most individuals were asymtomatic at the time of recruitment: at the initial screening 95% had come to hospital for the sole purpose of joining the trial.
Plasma viral load was above the detectable level in all and was distributed similarly between the two arms of the trial. Ten (9%) were deficient in vitamin E. Forty-four (34%) had selenium levels below the normal level for US populations.
Pill counts and changes in plasma levels of micronutrients over time
At the final follow-up visit, all 193 trial participants in the micronutrients arm and 185 of 186 in placebo arm who were known to be alive returned to clinic with the boxes containing their pills; 83% and 85% respectively had taken ≥ 75% of the prescribed number of pills (P = 0.6). The differences in the earlier scheduled visits were similar (data not shown). At the final follow-up, plasma levels of vitamin E and selenium were significantly higher in the micronutrients than in the placebo group: the differences in the mean increases from baseline [95% confidence interval (CI)] were 10.7 (7.0–14.3) n = 44; P < 0.001) and 0.16 (0–0.34) (n = 54; P = 0.04) μmol/l respectively.
A total of 137 minor adverse events were reported: 64 in the micronutrients arm and 73 in the placebo. The most common complaints were dizziness (n = 23), drowsiness (n = 13), nausea (n = 16), headache (n = 15) and rash (n = 10). Twenty-three individuals in the micronutrients arm and none in the placebo reported discoloration of urine (P < 0.0001).
By the end of the study, 79 (16%) trial participants were lost to follow-up, 23 (5%) had died and 379 (79%) were known to have survived. Their median CD4 counts at recruitment [interquartile range (IQR)] were, respectively, 206 (124, 355), 82 (68, 222) and 271 (152, 384) cells/μl; these differences are highly significant. (P < 0.0001). Among those lost to follow-up, the median (IQR) CD4 counts/μl were 201 (111, 330) and 209 (141, 380) in the micronutrients and placebo groups respectively (P = 0.7).
Fifteen deaths occurred in hospital and eight at home. Five (22%) deaths were attributed to cryptococcal meningitis, five (22%) to pulmonary tuberculosis, one (4%) to both cryptococcal meningitis and pulmonary tuberculosis, one (4%) to Pneumocystis carinii pneumonia and the remainder to a variety of different HIV-related causes.
Table 2 shows the mortality rates according to supplementation group. Eight deaths occurred in the micronutrients group and 15 in the placebo group (mortality hazard ratio, 0.53; 95% CI, 0.22–1.25; P = 0.1). The result was virtually unchanged after adjusting for age, sex, and CD4 cell count at baseline (mortality hazard ratio, 0.52; 95% CI, 0.22–1.23; P = 0.1).
The mortality rate was significantly lower in the micronutrients arm among those whose CD4 cell counts were either < 200 × 106 or < 100 × 106/l, the hazard ratios being 0.37 (95% CI, 0.13–1.06; P = 0.052) and 0.26 (95% CI, 0.07–0.97; P = 0.03) respectively. There was no significant difference in mortality rates between the micronutrients and placebo groups among trial participants with higher CD4 cell counts.
Admissions to hospital
Thirty-six trial participants were admitted to hospital on at least one occasion, 16 in the micronutrients arm and 20 in the placebo arm. The rate of first admissions did not differ significantly between the two groups either overall or when stratified by baseline CD4 cell count category: the hazard ratios (95% CI) of admissions in the micronutrients compared to the placebo group were 0.76 (0.40–1.47; P = 0.4) overall, 0.69 (0.30–1.58; P = 0.4) for those with CD4 cell count < 200 × 106/l and 0.86 (CI 0.29–2.6; P = 0.8) for those with CD4 cell count ≥ 200 × 106/l.
Impact on CD4 cell count and plasma viral load
Median CD4 cell count at the final follow-up and the mean fall in CD4 cell count from baseline did not differ significantly between the micronutrients and placebo groups overall or among those with baseline CD4 cell < 200 × 106/l or ≥ 200 × 106/l (P > 0.3 in each case) (Table 3).
Plasma viral load at 48 weeks did not differ significantly between the micronutrients and placebo arms, the mean log10 copies per ml being 4.4 (95% CI, 4.1–4.7; n = 55) and 4.5 (95% CI, 4.3–4.8; n = 57) respectively (P = 0.4).
In this randomized trial, the death rates among HIV-infected adults taking multiple micronutrient supplementation were lower than among those taking placebo. Statistically significant differences were observed among those whose CD4 cell counts at recruitment were < 200 × 106 or < 100 × 106/l, with 63% and 74% lower mortality rates respectively when compared to placebo. Based on these figures, 10 individuals with CD4 cell count < 200 × 106/l and four with CD4 cell count < 100 × 106/l would need to be supplemented for 1 year to prevent the death of one individual over this time. Mortality rates were similar between the two arms of the trial among those who had higher CD4 cell counts at enrolment although the power for this comparison was very low. The differences in the rates of hospital admissions between the micronutrients and placebo arms did not reach statistical significance. The confidence intervals of theses estimates and of those relating to survival were wide. These findings need to be reproduced in other settings. Micronutrients are relatively affordable, with for example, brands produced locally in Thailand costing about 1 $ US dollar for a 1-month supply, and have a good safety profile, so that even a modest benefit may have high public health implications.
Our hypothesis was based on evidence reported in observational studies showing associations between micronutrient deficiency and poorer survival and from two randomized trials of multiple micronutrients, which examined the impact on surrogate markers of disease stage; ours is the first to show a clinical benefit – a reduction in deaths – as a result of multiple micronutrient supplementation. One of the previous randomized trials of surrogate markers was very small . The other was large with a sample size of more than 250 pregnant women per group  but the women were relatively healthy with median CD4 cell count of > 400 × 106/l. Our study, which comprised men and women living in an urban community, showed clinical benefit among those with CD4 cell count < 200 × 106/l.
Our study also highlights the need to measure impact against clinical endpoints rather than on surrogate markers as the beneficial effect would have been missed if only surrogate markers had been measured. Discordance between clinical and biomedical endpoints has been demonstrated previously in, for example, a large-scale trial of zidovudine monotherapy, in which no clinical benefit was demonstrated despite a significant difference in CD4 cell count [19,20]. Interestingly, our trial did not show an impact on CD4 cell count or plasma viral load suggesting a protective mechanism that may operate independently of these parameters. It is possible that micronutrients may have improved the immune response by improving the T-lymphocyte function or one or more components of the innate immune system. For example, retinoids may regulate several elements of the immune response  while vitamin E supplementation seems to improve indices of cell-mediated immunity without increasing the number of CD4 cells . Zinc deficiency has been associated independently with an increased risk of mortality of HIV-1 infected homosexuals  while in vitro treatment with zinc of peripheral blood mononuclear cells led to an enhancement of lymphoproliferative immune response and apoptosis inhibition . Selenium supplementation is associated with an improvement of T-cell function and selenoproteins are thought to be important for activated T-cell function and the control of the immune response . Antioxidants such as vitamin E and selenium can also decrease inflammatory responses both by downregulating inflammatory cytokine production and by limiting tissue damage resulting from inflammatory responses . Further research is needed to elucidate the potential mechanism(s) for increased survival with micronutrient supplementation.
A number of factors need to be taken into account when interpreting our findings. The death rate was lower than expected which reduced the power of the trial. There are probably a number of explanations for this. Trial participants were selected from the community and were required to visit hospital twice before recruitment; sicker trial participants would not have been able to do this. Ascertainment of deaths may have been incomplete. Those lost to follow-up were sicker than those who remained in the trial (although the baseline median CD4 cell counts among those lost to follow-up did not differ between the micronutrient and placebo arms). Improved counselling and care were provided for all trial participants, including access to co-trimoxazole, which has been shown to reduce mortality by more than 40% in some settings . A proportion of our trial participants may also have had access to antiretroviral drugs. Zidovudine has been available in pharmacies in Bangkok at a much reduced price for many years and soon after the trial began access to cheaper combinations of antiretroviral drugs increased considerably both from supplies sold at subsidised rates by developed countries and from drug combinations produced locally. We collected data on antiretroviral use, which showed that just 10 trial participants, five in the micronutrient arm and five in the placebo arm, reported taking antiretrovirals at some stage. However, other trial participants may have been reluctant to admit to antiretroviral usage since this was an exclusion criteria for recruitment and trial participants may have perceived that taking antiretrovirals might lead to their exclusion from the trial. We also measured mean corpuscular volume of erythrocytes at the last visit as a marker of intake of zidovudine and this exceeded 100 fl in just 10 (5%) trial participants in the micronutrient arm and 13 (7%) in the placebo arm, suggesting that few patients were using this drug. It seems unlikely that many trial participants were on combinations of highly active antiretroviral therapy since none had undetectable plasma viral load at 48 weeks and the three-drug combinations remain prohibitively expensive for the vast majority. Overall, although the death rate was lower than expected, we do not see any reason to suggest bias favouring one arm of the trial over the other.
Our data suggest that adverse effects were as common among placebo as among micronutrient recipients and that compliance as measured by pill counts was high and similar in both arms. It is plausible that trial participants in both arms took additional micronutrients since, during the course of the trial, treatments which essentially comprised a mix of micronutrients were marketed widely as cures for HIV. Many of our trial participants asked us for information about these and some may have sought and tried these therapies. It is also possible that groups of trial participants mixed their pills so that all may get some perceived benefit. However, all of 23 trial participants who reported discolouration of urine were in the micronutrients group suggesting that the mixing of pills between trial participants was unlikely. We also measured plasma levels of vitamin E and selenium at the end of the follow-up in order to assess compliance; both were significantly higher in the micronutrient than in the placebo arm at the end of the trial.
Whether the same or a greater effect would be seen in other settings such as in Africa where nutritional deficiency may be greater is not clear. Much of the evidence linking micronutrient deficiency with poor outcomes in HIV-infected individuals is from well-nourished populations in developed countries but such data cannot be generalized to African settings. Larger trials of micronutrient supplementation among those with CD4 cell count < 200 × 106/l in Africa are now a priority as in this part of the world it seems few people are likely to have access to appropriately monitored highly active antiretroviral therapy in the immediate future. Larger trials among individuals with higher CD4 cell counts are also of interest because if micronutrient supplementation can provide a small amount of benefit and defer the initiation of highly active antiretroviral therapy, this could have high public health importance in Africa.
In summary, our data show a significantly lower death rate as a result of micronutrient supplementation among trial participants who had CD4 cell counts < 200 × 106/l. The data suggest a novel mechanism independent of CD4 T-lymphocyte numbers. Larger trials in different settings are required to examine whether this is true as micronutrients may prove to be a simple cost-effective intervention for millions of HIV infected people living in developing countries.
The authors thank the nursing and administrative staff of the departments of Dermatology and Obstetrics_Gynaecology at Siriraj Hospital and Prof. B. Greenwood, Prof. P. Smith, Prof. H. Whittle and Dr H. Friis for reviewing the manuscript, Dr F. Cowan for discussions on study design and review of the manuscript and Dr S. Kaye for help and discussions on laboratory procedures.
Sponsorship: Supported by the Nestle Foundation. S. J. is supported by a strategic grant in Epidemiology from the Medical Research Council, UK. The micronutrients were supplied by Vitabiotics Ltd, London. The sponsors of the trial had no role in the study design, data collection, data analysis, data interpretation or the writing of the report.
1. Semba RD, Tang AM. Micronutrients and the pathogenesis of human immunodeficiency virus infection.Br J Nutr
2. Friis H, Goma E, Michaelson. Micronutrient interventions and the HIV pandemic.
In Micronutrients and HIV infection
. Edited by Friis H. London: CRC Press; 2002:219–246.
3. Baum MK, Shor-Posner G, Lai S, Zhang G, Lai H, Fletcher MA, et al
. High risk of HIV-related mortality is associated with selenium deficiency.J Acquir Immune Defic Syndr Hum Retrovirol
4. Tang AM, Graham NM, Saah AJ. Effects of micronutrient intake on survival in human immunodeficiency virus type 1 infection.Am J Epidemiol
5. Kupka R, Fawzi W. Zinc nutrition and HIV infection.Nutr Rev
6. Allard JP, Aghdassi E, Chau J, Salit I, Walmsley S. Oxidative stress and plasma antioxidant micronutrients in humans with HIV infection.Am J Clin Nutr
7. Schwarz KB. Oxidative stress during viral infection: a review.Free Radic Biol Med
8. Treitinger A, Spada C, Verdi JC, Miranda AF, Oliveira OV, Silveira MV, et al
. Decreased antioxidant defence in individuals infected by the human immunodeficiency virus.Eur J Clin Invest
9. Semba RD. Vitamin A and immunity to viral, bacterial and protozoan infections.Proc Nutr Soc
10. Semba RD. The role of vitamin A and related retinoids in immune function.Nutr Rev
11. Moriguchi S, Muraga M. Vitamin E and immunity.Vitam Horm
12. Fawzi WW, Msamanga GI, Spiegelman D, Urassa EA, McGrath N, Mwakagile D, et al
. Randomised trial of effects of vitamin supplements on pregnancy outcomes and T cell counts in HIV-1-infected women in Tanzania.Lancet
13. Allard JP, Aghdassi E, Chau J, Tam C, Kovacs CM, Salit IE, et al
. Effects of vitamin E and C supplementation on oxidative stress and viral load in HIV-infected subjects.AIDS
14. Health Do. Dietary Reference Values: A Guide.
London: HMSO; 2000.
15. Miller KW, Lorr NA, Yang CS. Simultaneous determination of plasma retinol, alpha-tocopherol, lycopene, alpha-carotene, and beta-carotene by high-performance liquid chromatography.Anal Biochem
16. Gibson R. Principles of Nutritional Assessment.
Oxford: Oxford University Press; 1990.
17. Huffman S, Rasmusson E, Newman V, O'Gara C. Breastfeeding and the environment. A policy brief from Wellstart International.Birth Gaz
18. Jaffar S. The natural history of HIV-1, HIV-2 and dual infections in sub-Saharan Africa
(PhD thesis). London: London School of Hygiene and Tropical Medicine; 2001.
19. Concorde: MRC/ANRS randomised double-blind controlled trial of immediate and deferred zidovudine in symptom-free HIV infection. Concorde Coordinating Committee.Lancet
20. Walker AS, Peto TE, Babiker AG, Darbyshire JH. Markers of HIV infection in the Concorde trial. Concorde Co-ordinating Committee.QJM
21. Semba RD. The role of vitamin A and related retinoids in immune function.Nutr Rev
22. Meydani SN, Meydani M, Blumberg JB, Leka LS, Siber G, Loszewski R, et al
. Vitamin E supplementation and in vivo immune response in healthy elderly subjects. A randomized controlled trial.JAMA
23. Lai H, Lai S, Shor-Posner G, Ma F, Trapido E, Baum MK. Plasma zinc, copper, copper:zinc ratio, and survival in a cohort of HIV-1-infected homosexual men.J Acquir Immune Defic Syndr
24. Neves I, Jr., Bertho AL, Veloso VG, Nascimento DV, Campos-Mello DL, Morgado MG. Improvement of the lymphoproliferative immune response and apoptosis inhibition upon in vitro treatment with zinc of peripheral blood mononuclear cells (PBMC) from HIV+ individuals.Clin Exp Immunol
25. Rayman MP. The importance of selenium to human health.Lancet
26. Grimble RF. Nutritional modulation of immune function.Proc Nutr Soc
27. Wiktor SZ, Sassan-Morokro M, Grant AD, Abouya L, Karon JM, Maurice C, et al
. Efficacy of trimethoprim-sulphamethoxazole prophylaxis to decrease morbidity and mortality in HIV-1-infected patients with tuberculosis in Abidjan, Cote d'Ivoire: a randomised controlled trial.Lancet