Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.

130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
British Journal of Nutrition (2006), 95, 762–770 DOI: 10.1079/BJN20051684
q The Authors 2006

The effect of multi-vitamin/mineral supplementation on mortality during

treatment of pulmonary tuberculosis: a randomised two-by-two factorial trial
in Mwanza, Tanzania

Nyagosya Range1, John Changalucha2, Henrik Krarup3, Pascal Magnussen4, Åse B. Andersen5 and
Henrik Friis6*
1
National Institute for Medical Research, Muhimbili Research Station, PO Box 3436, Dar es Salaam, Tanzania
2
National Institute for Medical Research, Mwanza Medical Research Centre, PO Box 1462, Mwanza, Tanzania
3
Department of Clinical Biochemistry, Aalborg University Hospital, PO Box 561, 9100 Aalborg, Denmark
4
DBL Institute for Health Research and Development, Jaegersborg Allé 1D, Charlottenlund, Denmark
5
Department of Infectious Diseases, Rigshospitalet, University of Copenhagen, Blegdamsvej 9, 2100 Copenhagen Ø, Denmark
6
Department of Epidemiology, Institute of Public Health, University of Copenhagen, Øster Farimagsgade 5B, DK-1014
Copenhagen K, Denmark
(Received 2 August 2005 – Revised 11 November 2005 – Accepted 22 November 2005)

Malnutrition is common in pulmonary tuberculosis (TB), and may impair survival. The objective of this study was to assess effects of multi-vita-
min/mineral (MVM) and zinc (Zn) supplementation during TB treatment on mortality. Patients diagnosed with sputum-positive pulmonary TB in
Mwanza, Tanzania, were randomised, using a two-by-two factorial design, to Zn (45 mg) or placebo, and MVM (vitamins A, B, C, D, E, and
selenium and copper) or placebo. Survival status was ascertained at the end of the 8-month TB treatment and supplementation period. Of 499
TB patients, 213 (43 %) had HIV. The mean weight gain at 7 months was 6·88 kg (95 % CI 6·36, 7·41). Zn and MVM combined, but neither
alone (interaction, P¼ 0·03), increased weight gain by 2·37 kg (95 % CI 0·91, 3·83), irrespective of HIV status. Survival status at 8 months was
determined for 422 patients (84·6 %), of which fifty-two (12·3 %) had died. Among fifty-two deaths, there were no effects of MVM (relative
risk (RR) 0·73; 95 % CI 0·43, 1·23) and Zn (RR 0·76; 95 % CI 0·46, 1·28). However, among HIV co-infected patients, marginally significant effects
of both MVM (RR 0·60; 95 % CI 0·34, 1·05) and Zn (RR 0·63, 95 % CI 0·37, 1·08) were seen, and MVM and Zn combined reduced mortality (RR
0·29; 95 % CI 0·10, 0·80; interaction ratio 0·52). In conclusion, supplementation with MVM, including Zn, during treatment of pulmonary TB may
reduce mortality in those co-infected with HIV. A randomised trial of the effect of the combined intervention used in this study should be con-
ducted in a different setting to confirm the finding.

Tuberculosis: HIV: Vitamins: Minerals: Weight: Mortality: Tanzania

The risk of active tuberculosis (TB) disease depends on of HIV to AIDS and death in adults (Fawzi et al. 2004; Jiam-
exposure to Mycobacterium tuberculosis and reduced host ton et al. 2003).
defence. One-third of the world’s population is infected, Despite appropriate anti-TB chemotherapy, the usual
nine million develop clinical disease each year, and two cereal-based diet will not ensure an adequate micronutrient
million die due to the infection (Frieden et al. 2003). HIV intake during convalescence. Micronutrient deficiencies not
infection is considered the strongest single risk factor of devel- only impair host immune functions, but may also affect effi-
oping TB disease after primary infection, and of death during cacy of TB drugs (Thurnham, 2005), and impair weight gain
treatment. which may be important to survival (Paton et al. 2004).
Deficiencies of vitamins and minerals, leading to nutrition- Although nutritional supplementation during the short-course
ally acquired immune deficiency syndrome (Beisel, 2001) may TB treatment seems logistically feasible and affordable, and
be another important determinant of TB infection and disease, may improve treatment outcomes, nutritional assessment and
as well as poor treatment outcomes (Cegielski & McMurray, support is currently not part of the case-management.
2004). For example, it is well-documented that zinc (Zn) sup- We conducted a randomised, placebo-controlled, double-
plementation reduces risk and severity of diarrhoea and pneu- blind, two-by-two factorial trial to assess the effect of daily
monia in children (Bhutta et al. 1999, 2000), and it has been Zn and multi-vitamin/mineral (MVM) supplementation on
shown that micronutrient supplementation reduces progression treatment outcomes. Although we found no beneficial effect

Abbreviations: MVM, multi-vitamin/mineral; RR, relative risk; TB, tuberculosis.

* Corresponding author: Dr Henrik Friis, fax þ45 35 32 73 83, email [email protected]
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
Supplementation and tuberculosis mortality 763

on sputum conversion (Range et al. 2005), the effects on diagnosis of parasitic infections (World Health Organization,
weight gain and mortality over the full duration of TB treat- 1991), and determination of Hb concentration and leucocyte
ment were also assessed and are reported here. counts. Heights were measured to the nearest 0·1 cm and
weights to the nearest 0·1 kg, with the patient barefoot and
wearing light clothing. The same staff member measured the
Subjects and methods
patient at baseline and follow-up. Patients found anaemic or
New or relapse sputum microscopy-positive TB patients aged infected with Schistosoma spp., intestinal helminths or malaria
15 years or above and resident in the area were consecutively were treated accordingly (World Health Organization, 1995).
recruited from five health facilities in Mwanza and Magu dis- HIV testing was done using two different ELISA. In samples
tricts, Mwanza Region, Tanzania, between August 2001 and found positive, HIV-1 load was determined using a modifi-
July 2002. Patients were given information about the study cation of an in-house developed RT-PCR (Krarup et al.
and offered inclusion after informed consent. TB patients 1998). CD4 cell count was determined by manual immuno-
returning to treatment after defaulting or smear-positive failure cytochemistry (Gomo et al. 2004).
cases, or those with serious TB or other disease unlikely to sur-
vive, and pregnant or lactating women were excluded. The
study was conducted within the framework of the National
Study design and intervention
TB and Leprosy Programme (Ministry of Health, 1991), with
diagnosis, classification, registration and treatment of TB The study was a randomised, placebo-controlled trial, using a
patients done in accordance with recommended standard pro- two-by-two factorial design, to assess the effects of daily sup-
cedures (World Health Organization, 2003). In brief, new TB plementation with a MVM and a Zn tablet. Although sputum
cases were given a combination of four drugs daily under super- conversion and weight gain during the 2-month intensive
vision of a health worker during the initial 2-month intensive phase were the main outcomes on which sample size consider-
phase, while two drugs were self-administered daily by the ations were based as previously reported (Range et al. 2005),
patient, during the 6-month continuation phase. daily Zn and MVM supplementation was continued through-
TB suspects attending in- and out-patient clinics at any of out the 8 months of short-course TB treatment when survival
the recruitment centres were asked to submit three sputum status was assessed. The MVM tablets contained vitamin A
specimens (spot, early morning, spot) for examination of (1·5mg), vitamin B1 (20 mg), vitamin B2 (20 mg), vitamin
acid-fast bacilli using the Ziehl-Neelsen staining technique. B6 (25 mg), vitamin B12 (50 mg), folic acid (0·8 mg), niacin
Those found sputum microscopy-positive were requested to (40 mg), vitamin C (200 mg), vitamin E (60 mg), vitamin D3
submit a morning sputum specimen in a sterile universal (5 mg), selenium (0·2 mg) and copper (5 mg), and Zn tablets
bottle for confirmation at the Zonal TB Reference Laboratory contained 45 mg elementary Zn (Table 1). The rationale for
at Bugando Medical Centre, based on microscopy after fluor- our interventions was the high requirements during TB treat-
ochrome (Auramine O) staining, and culture on Lowenstein ment, when considerable lean body mass is being synthesised.
Jensen solid media (Githui et al. 1993). Only those confirmed As for the vitamins, the doses were based on the multivitamin
to be sputum microscopy- or culture-positive at the Zonal TB supplement proven beneficial among HIV-infected women
Reference Laboratory were included in the trial (Fig. 1). (Fawzi et al. 1999). However, we included more vitamin E
All patients found sputum microscopy-positive at the (60 instead of 30 mg), since higher doses have been shown
recruitment centres were started on TB treatment, and given to improve immune functions (Food and Nutrition Board,
a study number and supplement according to the randomis- 2000, pp. 186–283), and less vitamin C (200 instead of
ation, as described later. Prior to that, the TB Clinical Officers 1000 mg), since there has been concern that high doses may
collected demographic, socio-economic and medical history cause oxidative damage in individuals with high iron stores
data from all patients, examined the patients anthropometri- (Food and Nutrition Board, 2000, pp. 95– 185). However,
cally, and collected stool, urine and blood samples for later we included preformed vitamin A only, since it has been

530 sputum-positive at health facility randomised

to zinc or placebo and MVM or placebo

132 placebo/placebo 133 zinc/placebo 133 MVM/placebo 132 zinc/MVM

7 sputum-negative at 10 sputum-negative at 5 sputum-negative at 9 sputum-negative at

reference laboratory reference laboratory reference laboratory reference laboratory

40 HIV load at 2 months 52 HIV load at 2 months 49 HIV load at 2 months 43 HIV load at 2 months
97 weighed at 7 months 96 weighed at 7 months 101 weighed at 7 months 95 weighed at 7 months
125 followed for survival 123 followed for survival 128 followed for survival 123 followed for survival
4 never seen 0 never seen 4 never seen 2 never seen
12 lost before day 210 6 lost before day 210 17 lost before day 210 14 lost before day 210
18 lost before day 244 11 lost before day 244 28 lost before day 244 20 lost before day 244

Fig. 1. Trial profile. Sputum-positive pulmonary tuberculosis patients were randomised to daily supplementation with zinc or placebo and, independently, to multi-
vitamin/mineral (MVM) or placebo. Those found sputum-negative at the reference laboratory after randomisation were excluded.
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
764 N. Range et al.

Table 1. Composition and doses of vitamins and min- The profile for recruitment, randomisation and trial partici-
erals in the two interventions pants is shown in Fig. 1. Trial participants were asked to come
to the recruitment centre after 2, 5 and 7 months. Collection of
Dose
blood was repeated at the 2-month visit, and anthropometric
Multi-vitamin/mineral assessment and collection of sputum were repeated at all
Vitamin A 1·5mg follow-up visits. Those not seen at the clinic during scheduled
Vitamin B1 20 mg follow-up visits were followed up by the study team using
Vitamin B2 20 mg
Vitamin B6 25 mg
addresses given during registration. In cases where the patient
Vitamin B12 50 mg had died, the date of death was obtained from relatives. In the
Folic acid 0·8 mg analysis of the effect of supplementation on mortality, a
Niacine 40 mg person was considered to have died if death occurred within
Vitamin C 200 mg the 8-month supplementation, i.e. not later than 244 d after
Vitamin D3 5 mg
Vitamin E 60 mg start of treatment and supplementation, and to have survived
Selenium (as methionine) 0·2 mg if examined at the 7-month visit or known to be alive later.
Copper (as sulphate) 5 mg Cause of death was not ascertained.
Zinc The sample size of 500 was initially based on sputum cul-
Zinc (as gluconate) 45 mg
ture conversion and weight gain as the primary outcomes, as
previously reported (Range et al. 2005). However, assuming
a 10 % cumulative mortality during the 8 months of treatment
postulated that previously observed adverse effects of vitamin we would be able to detect a reduction to 5 % or less with
A interventions could be due to provitamin carotenoids (Friis, 80 % power and 95 % confidence, while allowing for 10 %
2005). We also added vitamin D, since vitamin D deficiency loss to follow-up.
has been suggested as a risk factor of TB (Wilkinson et al. Permission to conduct the study was granted by the Ethics
2000). The minerals selenium and copper were added, due Committee of the National Institute for Medical Research in
to their importance to the immune system (Beisel, 2001). Tanzania, and the Danish Central Medical Ethics Committee
Due to the importance of Zn to immune functions, and to sus- also approved the study. Permission was also obtained from
ceptibility to respiratory and other infections, and to synthesis the Regional Medical Officer and the Council Medical Offi-
of lean body mass (Friis & Sandström, 2001), we assessed its cers in Mwanza Region. Informed oral consent was obtained
effect separately, using a two-by-two factorial design. The from all study participants. Pre-test HIV counselling was
dose of Zn (45 mg) was considered appropriate, given the given to all, and post-test counselling was given to those
low bioavailability of Zn in populations subsisting on a who wanted to know their HIV results. Antiretroviral treat-
cereal-based diet (Friis & Sandström, 2001). Placebo tablets ment was not available.
were identical in colour, shape and size to the corresponding
white Zn and green MVM tablets. All tablets were manufac-
Statistical analysis
tured by Almega (Ringsted, Denmark). Tablets were packed
in identical white plastic containers and delivered labelled Two-sample t test and one-way ANOVA were used to test for
with four different alphabetical letters. Containers with differences in means, and x2-test to test for differences in pro-
white tablets (Zn and placebo) were labelled either T or H, portions between groups. Assessment of the effect of the study
and containers with green tablets (MVM and placebo) were interventions on mortality was done prior to breaking the
labelled either X or S. Before recruitment of patients, all the code. The analysis was based on intention-to-treat, and done
letters on the containers were replaced with study serial num- with and without stratification for HIV status. Multiple logistic
bers from 1 to 550 based on the computer-generated random regression analysis was used to estimate the effects of the
sequences, using permuted blocks of four. The five recruit- study interventions while controlling for potential baseline
ment centres assigned consecutive identity numbers, using differences, and to assess for interactions. Binreg was used
different starting numbers, to patients being recruited. Each to compute relative risks (RR), rather than odds ratios, using
participant was assigned two containers labelled with the cor- Stata version 8.2 (StataCorp LP, College Station TX, USA).
responding identity number: one containing white (Zn or pla- Survival analysis using Cox regressions was also done, with
cebo) and the other green (MVM or placebo) tablets enough follow-up time to either the date of death, the date last to
for the full 8-month supplementation. Containers were kept follow-up or end of supplementation period. P values below
in the health facilities where the participants were attending 0·05 were considered significant.
TB treatment. The codes for the MVM and Zn tablets
remained in a sealed envelope, and were only broken after
Results
completion of the initial data analysis.
The same health personnel administering TB drugs during Of 530 patients found positive based on detection of tubercle
the intensive phase ensured that the participant also received bacilli in sputum on microscopy and randomised to the two
the trial tablets under direct observation. At the monthly study interventions, thirty-one could not be confirmed positive
visits during the continuation phase, each participant was based on re-examination at the reference laboratory and were
given two plastic bags, appropriately labelled with the excluded from the study (Fig. 1). Thus, 499 confirmed
patient’s identity number, containing thirty white (Zn or pla- sputum-positive cases were included in the trial. The mean
cebo) and thirty green (MVM or placebo) tablets to take at age was 35·4 (range 15 –85) years, and 41·1 % were women.
home along with TB drugs. Most of the participants were recruited at Sekou Toure
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
Supplementation and tuberculosis mortality 765

1·14
24·1
12·7
2·5
Hospital (47·7 %), and the rest from Bugando Medical Centre

SD

338
(16·6 %), Magu Hospital (18·0 %), and Buzuruga (10·2 %) and
Butimba (7·4 %) Health Centres. The prevalence of HIV infec-

3·72
Mean

106·4
36·1
18·3
tion was 42·7 % (213), and 34·7 % (173) were found to excrete

423
MVM þ Zn (n 123)
Schistosoma mansoni and 20·3 % (101) hookworm eggs, and

Table 2. Baseline characteristics of 499 pulmonary tuberculosis patients randomised to zinc or placebo supplementation, and to multi-vitamin/mineral (MVM) or placebo supplementation†
4·6 % had malaria parasitaemia. As seen in Table 2, baseline

7
48

38

25
19
75
48

28
19

45
n
equivalence was achieved with respect to age, BMI, Hb, and

Prevalence (%)
prevalence of heavy culture intensity and HIV, although
there seemed to be more women in the placobo/placebo group.

39·0

30·9

20·7
15·7
62·0
39·0

22·8
5·7
15·5

36·6
As seen from the trial profile (Fig. 1), viral load and CD4
counts were determined after 2 months, weight at 7 months,
and survival up to the end of treatment and supplementation
at 8 months.

0·98
11·8

21·1
2·5
SD

275
Viral load and CD4 count

4·02
Mean

34·9

105·4
18·0
At the 2-month follow-up examination, viral load and CD4

363
MVM þ placebo (n 128)
count data were determined in 184 (86·4 %) of the 213 HIV-
infected participants (Table 3). The mean viral load increase

51

59

7
49

12

38
17
72

22
38
n
was 0·23 log10(geq/l) (95 % CI 0·01, 0·46). There was no sig-

Prevalence (%)
nificant effect of MVM (2 0·22 log10(geq/l); 95 % CI 2 0·67,
0·23; P¼0·35) or Zn supplementation (0·27 log10(geq/l),

39·8

46·1
38·3

9·4

29·9
13·4
56·7

17·2
5·5
29·6
95 % CI 2 0·18, 0·72; P¼0·24) (interaction, P¼0·37), and
no effect of any combination of the two interventions. Zn sup-
plementation alone was associated with a 0·54 log10(geq/l)
(95 % CI 20·08, 1·15; P¼0·08) higher increase. Similarly,

1·19
13·0
2·5
24·5
there were neither effects of Zn (18 cells/ml; 95 % CI 2 101,
SD

331
137) nor MVM (46 cells/ml; 95 % CI 2 72, 165) on CD4
counts (interaction, P¼0·22), and no effect of any combi-

3·83
Mean

35·2
17·8
102·5
Intervention group

nation of the two interventions (Table 3).

422
Placebo þ Zn (n 123)

41

58

4
45

17

17
12
85

49
24
Weight gain
n

At the 7-month follow-up examination, 389 (80·0 %) of the

Prevalence (%)

499 patients were weighed. The mean weight gain was

33·3

47·2
36·6

13·8

16·0
10·1
71·4

39·8
19·5
3·3
6·88 kg (95 % CI 6·36, 7·41). However, an interaction was
found between Zn and MVM supplementation (P¼0·03).
The effect of each combination of Zn and MVM was therefore
compared to placebo, based on the use of dummy variables in
1·20

multiple linear regression analysis (Table 4). As seen, those

11·7
2·7
25·2
SD

295

receiving both Zn and MVM supplementation had a 2·37 kg

(95 % CI 0·91, 3·83; P¼0·002) greater weight gain than
† For details of procedures, see pp. 763 –764; ‡Among HIV-infected only.
3·90
Mean

those receiving placebo and placebo, whereas neither Zn nor

35·3
18·7
104·6
Placebo þ placebo (n 125)

403

MVM alone had any effect. After adjustment for sex, age,
HIV status and heavy culture intensity, the estimated effects
39

48
9

41

5
63

27
14
74

27
n

of Zn and MVM supplementation combined was 2·63 kg

(95 % CI 1·18, 4·09; P, 0·001), whereas there were no effects
Prevalence (%)

of Zn or MVM alone.
31·2

38·4
7·2
50·4

23·1
12·0
63·3

33·1
21·8
4·0

Survival status
As seen in the trial profile (Fig. 1), survival status could not be
ascertained at any time-point after commencement of treat-
Viral load (log10(geq/l))
CD4 count (cells/ml)‡

ment and supplementation for ten (2·0 %) of the 499 study par-
Schistosoma mansoni

ticipants. Forty-nine (9·8 %) were lost to follow-up before

210 d post-treatment, and 77 (15·4 %) before 244 d post-treat-
Culture intensity
Alcohol drinking

ment. The median observation time among those lost to full

HIV infection
Moderate

Hookworm
BMI (kg/m2)

follow-up was 197 d (interquartile range 121–221 d). Hence,

Female sex
Age (years)

Heavy
Infections

Malaria
Smoking

Low

422 (84·6 %) of the 499 study participants were known

Hb (g/l)

either to have died within 244 d post-treatment or to be alive

after that. There were no differences between the 422 followed
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
766 N. Range et al.

Table 3. Effects of zinc and multi-vitamin/mineral (MVM) supplementation on HIV viral load (log10(geq/l)) and CD4 count (cells/ml) during first 2 months
of treatment of pulmonary tuberculosis patients among 213 HIV co-infected patients†

Intervention group

Placebo þ placebo (n 48) Placebo þ Zn (n 58) MVM þ placebo (n 59) MVM þ Zn (n 48)

Mean 95 % CI Mean 95 % CI Mean 95 % CI Mean 95 % CI

Viral load
Baseline 3·90 3·53, 4·27 3·83 3·52, 4·15 4·02 3·76, 4·27 3·72 3·38, 4·05
2 months 4·10 3·67, 4·54 4·28 3·86, 4·71 4·14 3·73, 4·55 3·85 3·46, 4·25
Change‡ 0·08 2 0·40, 0·55 0·62 0·20, 1·03 0·16 2 0·22, 0·56 2 0·01 2 0·47, 0·45
Difference§ – 0·54 2 0·08, 1·15 0·09 2 0·27, 0·55 2 0·09 2 0·71, 0·54
CD4 count
Baseline 460 351, 569 406 327, 485 365 289, 442 460 349, 570
2 months 403 309, 569 422 331, 512 363 281, 444 423 318, 529
Change‡ 2 104 2 243, 34 29 2 113, 96 19 89, 127 2 45 2 175, 85
Difference§ – 96 2 67, 259 124 2 45, 292 60 2 113, 233

† Data at follow-up available for 184 participants. For details of procedures, see pp. 763– 764.
‡ Change (2 months 2 baseline), t test.
§ Difference (each active group 2 (placebo þ placebo)). Linear regression analysis, interaction between interventions, P¼0·10 (viral load) and P¼ 0·18 (CD4 counts). None
were significant.

up and the seventy-seven lost to follow-up with respect to (the RR of MVM supplementation among those given Zn
proportion with heavy culture intensity and prevalence of divided by that of those given placebo) was 0·67 for all
HIV co-infection (Table 5), but the mean viral load among patients combined, and 0·52 for HIV-infected patients only.
HIV-infected participants was lower among those lost to The effects of each combination of MVM and Zn supplemen-
follow-up. Furthermore, slightly fewer of those lost to tation are therefore presented (Table 6). Overall, neither Zn
follow-up were allocated to Zn/placebo and more to pla- alone, MVM alone, nor Zn and MVM combined had effects
cebo/MVM. However, the mean observation time among on mortality. For HIV-infected patients only, neither Zn (RR
those lost to follow-up was not significantly different between 0·75; 95 % CI 0·40, 1·42) nor MVM (RR 0·73; 95 % CI
the four groups (P¼0·65). 0·38, 1·43) alone had effects on mortality, whereas MVM
Of the 422 (84·6 %) patients for which survival status at 8 and Zn combined considerably reduced mortality (RR 0·29;
months was known, fifty-two (12·3 %) had died: ten (4·1 %) 95 % CI 0·10, 0·80). Similar results were obtained using survi-
of the 241 without HIV and forty-two (23·2 %) of the 181 val analysis (Fig. 2).
with HIV co-infection. The median survival time was 106 d These effect estimates did not change after adjustment for
(interquartile range 65 –164 d). sex, age, alcohol intake or smoking in logistic regression ana-
Of the fifty-two deaths, sixteen occurred in the placebo/pla- lyses, but adjustment for heavy culture intensity slightly
cebo, fifteen in the placebo/Zn, thirteen in the MVM/placebo increased the independent effect of Zn supplementation (RR
and eight in the MVM/Zn group (P¼0·42). Of the forty-two 0·55; 95 % CI 0·31, 0·95; P¼0·03) and reduced the independent
deaths in HIV-infected patients, fourteen were in the pla- effect of MVM (RR 0·64; 95 % CI 0·37, 1·12; P¼0·12). With
cebo/placebo, thirteen in the placebo/Zn, eleven in the similar adjustments, MVM and Zn together were found to
MVM/placebo and four in the MVM/Zn group (P¼0·07). reduce mortality (RR 0·28; 95 % CI 0·10, 0·77; P¼0·01),
Although there was no statistically significant interaction whereas neither Zn (RR 0·66; 95 % CI 0·34, 1·28) nor MVM
between Zn and MVM supplementation, the interaction ratio (RR 0·80; 95 % CI 0·41, 1·55) alone had effects.

Table 4. Effects of zinc and multi-vitamin/mineral (MVM) supplementation on weight gain during treatment of pulmonary tuberculosis among 389
patients weighed at 7 months

Intervention group

Placebo þ placebo (n 97) Placebo þ Zn (n 96) MVM þ placebo (n 101) MVM þ Zn (n 95)

Mean 95 % CI Mean 95 % CI Mean 95 % CI Mean 95 % CI

Baseline 50·2 48·5, 51·9 51·6 49·6, 53·6 50·6 49·1, 52·2 50·7 49·3, 52·2
7 months 57·9 55·8, 60·0 56·2 54·5, 58·0 57·2 55·6, 58·7 59·4 57·8, 61·0
Change‡ 6·28 5·26, 7·30 6·07 5·01, 7·13 6·58 5·76, 7·39 8·65 7·42, 9·88
Difference§ – 2 0·21 2 1·66, 1·25 0·30 2 1·14, 1·73 2·37* 0·91, 3·83

Mean value was significantly different from the placebo þ placebo group: *P¼ 0·002.
† For details of procedures, see pp. 763– 764.
‡ Change (7 months 2 baseline), t test.
§ Difference (each active group 2 (placebo þ placebo)). Linear regression analysis, interaction between interventions, P¼ 0·03.
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
Supplementation and tuberculosis mortality 767

Table 5. Baseline characteristics of the 422 patients followed up for survival status at 8 months compared to 77 not followed up†

Survival status

Followed up (n 422) Not followed up (n 77)

Prevalence (%) n Mean 95 % CI Prevalence (%) n Mean 95 % CI P

Female sex 40·8 172 36·1, 45·5 42·9 33 31·6, 54·2 0·73k
Age (years) 35·6 34·5, 37·0 34·1 31·1, 37·0 0·30{
BMI (kg/m2) 18·2 18·0, 18·5 18·1 17·6, 18·4 0·69{
Hb (g/l) 104·7 102·4, 107·0 104·9 99·9, 109·9 0·93{
Alcohol drinking 33·2 140 28·7, 37·7 37·7 29 26·6, 48·7 0·44k
Smoking 10·9 46 8·1, 14·3 14·3 11 7·4, 24·1 0·39k
Infections
Culture intensity‡ 0·38k
Low 23·4 96 17·8 13
Moderate 12·2 50 16·4 12
Heavy 62·8 258 65·8 48
HIV infection 42·9 181 38·1, 47·6 41·6 32 30·3, 52·8 0·83k
Viral load (log10(geq/l))§ 3·95 3·79, 4·11 3·43 2·98, 3·88 0·02{
CD4 count (cells/ml)§ 406 356, 457 490 389, 590 0·19{
Schistosoma mansoni 34·7 146 30·1, 39·4 35·1 27 24·5, 46·8 0·94k
Hookworm 19·5 82 15·8, 23·6 24·7 19 15·6, 35·8 0·30k
Malaria parasitaemia 4·5 19 2·7, 7·0 5·1 4 1·4, 12·8 0·79k
Study intervention*
Placebo/placebo 25·4 107 23·4 18 0·04k
Zn/placebo 26·5 112 14·3 11
Placebo/MVM 23·7 100 36·4 28
Zn/MVM 24·4 103 26·0 20

MVM, multi-vitamin/mineral.
*Distribution between intervention groups was significantly different between those followed up and those not followed up.
† For details of procedures, see pp. 763– 764.
‡ Twenty-two were only microscopy-positive: seven were culture-negative, and culture data were missing for fifteen.
§ Among HIV-infected only.
k x2 test.
{ t test.

Discussion are needed (Harries et al. 2001). TB patients have consider-

able weight loss and vitamin and mineral deficiencies at the
The deteriorating TB treatment outcome in the wake of the time of diagnosis (Kennedy et al. 1996; Karyadi et al. 2000;
HIV pandemic is a major concern (UNAIDS, 2004; World Mugusi e.t al. 2003). Despite appropriate TB treatment, it is
Health Organization, 2004), and new adjunctive interventions unlikely that a patient will be able to regain normal body

Table 6. The effects of different combinations of zinc and multi-vitamin/mineral (MVM) supplementation on cumu-
lative mortality by 8 months among 422 pulmonary tuberculosis patients†

Intervention group

Placebo þ placebo Placebo þ Zn MVM þ placebo MVM þ Zn

All (n 422) n 107 n 112 n 100 n 103

No. of deaths (n 52) 16 15 13 8
Mortality (%) 15·0 13·4 13·0 7·8
RR (95 % CI)‡ – 0·90 (0·47, 1·72) 0·87 (0·44, 1·71) 0·52 (0·23, 1·16)
HIV-negative (n 241) n 65 n 60 n 55 n 61
No. of deaths (n 10) 2 2 2 4
Mortality (%) 3·2 3·3 3·6 6·6
RR (95 % CI)§ – 1·08 (0·16, 7·45) 1·18 (0·17, 8·12) 2·13 (0·40, 11·21)
HIV-positive (n 181) n 42 n 52 n 45 n 42
No. of deaths (n 42) 14 13 11 4
Mortality (%) 33·3 25·0 24·4 9·5
RR (95 % CI)k – 0·75 (0·40, 1·42) 0·73 (0·38, 1·43) 0·29* (0·10, 0·80)

RR, relative risk.

Significantly different from the placebo þ placebo group: *P¼0·016.
† For details of procedures, see pp. 763– 764.
‡ Test for interactions: MVM £ Zn, P¼ 0·75; MVM £ HIV, P¼0·02; Zn £ HIV, P¼0·03; MVM £ Zn £ HIV, P¼ 0·45.
§ Test for interactions: MVM £ Zn, P¼ 0·66.
k Test for interactions: MVM £ Zn, P¼ 0·30.
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
768 N. Range et al.

1·0
up, so the risk of selection bias seems to be small, and not
able to explain the findings. Both patient and investigator
were blinded as to the treatment allocation, and data analysis
Proportion alive

0·9
was done prior to breaking the code, based on intention-to-
treat. Compliance was not systematically assessed, but the
0·8 supplement was given together with the TB drugs, i.e. under
direct observation during the first 2 months and by self-admin-
0·7 istration during the last 6 months.
This is the first reported randomised trial on the effects of
MVM supplementation on mortality among TB patients. How-
0·6 ever, the effects seen among HIV-infected individuals are in
0 2 4 6 8 accordance with recent data from randomised trials among
Months after supplementation HIV-infected individuals in Thailand and Tanzania. The trial
from Thailand showed that a daily MVM supplement may
Fig. 2. Kaplan – Meier survival plot by intervention group in 213 HIV co-
infected pulmonary tuberculosis patients. —, multi-vitamin/mineral (MVM) þ
reduce mortality among HIV-infected individuals with low
zinc; ·····, MVM þ placebo; – – , zinc þ placebo; –- –, placebo þ placebo. CD4 cell counts. The supplement containing twenty-one vita-
For details of procedures, see pp. 763– 764. mins and minerals in multiples of RDA was given daily to 481
HIV-infected individuals with CD4 counts between 50 and
550 (106 cells/l; Jiamton et al. 2003). After 48 weeks, seventy-
weight and vitamin and mineral status on a typical diet based nine (16 %) were lost to follow-up and twenty-three (5 %) had
on cereals, tubers or legumes, which has low content and bioa- died. The death rate was lower in those allocated to MVM (mor-
vailability of vitamins and minerals (Ramakrishnan & Huff- tality hazard ratio 0·53; 95 % CI 0·22, 1·25), but only signifi-
man, 2001). The vitamin and mineral deficiencies may cantly so among those with CD4 counts below 100 (0·26; 95 %
impair full restoration of lean body mass and immune func- CI 0·07, 0·97). There was no effect of the intervention on viral
tions, and survival. load. Similarly, a recent randomised, two-by-two factorial trial
We therefore assessed the effect of adjunctive Zn and MVM among 1078 pregnant HIV-infected women from Dar es
supplementation given throughout the 8 months of TB treat- Salaam, Tanzania, assessed the effect of a daily vitamin A and
ment. As previously reported, we found no beneficial effect multivitamin supplement given throughout pregnancy and for
on the primary outcome, sputum conversion which was several years after. While vitamin A supplementation increased
assessed over the 2-month intensive phase (Range et al. mother-to-child HIV transmission, multivitamin supplemen-
2005). However, effects on weight gain up to 7-month tation increased maternal CD4 cell counts (Fawzi et al. 1999),
follow-up, and all-cause mortality during the full 8 months reduced mother-to-child HIV transmission in subgroups
of treatment and supplementation were also assessed. (Fawzi et al. 2002), and also reduced viral load and risk of
There were no overall reductions in mortality with Zn and AIDS and AIDS-related deaths among the mothers themselves
MVM supplementation. With only ten deaths, the effect of (Fawzi et al. 2004).
the study interventions could not be assessed among TB The supplement we used was similar to the supplement
patients not infected with HIV. However, among HIV co- given to pregnant HIV-infected women in Dar es Salaam
infected patients, both Zn and MVM supplementation were (Fawzi et al. 1999), but with more vitamin E and less vita-
associated with roughly a halving of the risk of death, min C, but with addition of vitamin D and preformed vita-
although not statistically significant, but those receiving the min A, as well as selenium and copper. We were
combined supplement had a significant 70 % reduction in particularly interested in the mineral Zn, since it is known
risk of death. Due to the deviating interaction ratios it may to be essential to the immune system (Shankar & Prasad,
be considered justified to report the effect of the combined 1998) and host defence to respiratory tract and other infec-
interventions despite the lack of statistical interaction between tious diseases (Bhutta et al. 1999, 2000). Therefore, and
interventions, since power to detect interactions are often lack- due to the concern that Zn may increase progression of
ing in factorial trials (McAlister et al. 2003). HIV in US patients (Tang et al. 1993, 1996), we assessed
Survival status could only be ascertained for 85 % of the the effect of Zn separately, using a two-by-two factorial
study participants for the full 8-month treatment period. design. The dose of Zn (45 mg) was considered appropriate,
Nevertheless, the median observation time for those lost to given the high anti-nutrient content of a typical diet in devel-
follow-up was above 6 months, and they were similar to oping countries, and the high requirements during synthesis
those followed up with respect to age, sex, body composition, of lean body mass.
and the proportion with heavy culture intensity and HIV co- As in the trial in Bangkok, but in contrast to that in Dar es
infection, although the mean HIV viral load among infected Salaam, the effect of our intervention was apparently not
participants was lower. Among those lost to follow-up, how- mediated by reduced viral replication, although viral load
ever, a greater proportion was allocated to MVM and a was only measured over the first 2 months. In contrast, the
lower proportion to Zn. If the mortality rate among those effect on survival was accompanied by considerable effects
lost to follow-up was the same irrespective of treatment allo- on weight gain. Interestingly, neither Zn nor MVM alone
cation, then the effect of MVM would be overestimated and had any effects on weight gain, whereas the combined inter-
the effect of Zn supplementation underestimated. The pro- vention increased weight gain by almost 2·5 kg. This is plaus-
portion receiving both Zn and MVM was not different ible, since several nutrients are required for the synthesis of
among those lost to follow-up compared to those followed new tissues (Golden, 1992).
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
Supplementation and tuberculosis mortality 769

Vitamin and mineral supplementation may considerably transmission of HIV-1 through breastfeeding and early child mor-
increase survival during treatment of sputum-positive TB tality. AIDS 16, 1935 –1944.
patients co-infected with HIV. Nevertheless, the finding Fawzi WW, Msamanga GI, Spiegelman D, et al. (2004) A random-
should be confirmed, because the effect is only seen on sec- ized trial of multivitamin supplements and HIV disease progression
and mortality. N Engl J Med 351, 23– 32.
ondary outcomes in a subgroup of patients, and because com-
Food and Nutrition Board (2000) Vitamin C. Dietary Reference
pliance data are not available. Also, the effects may differ Intakes for Vitamin C, Vitamin E, Selenium and Carotenoids.
between different settings, due to interactions with other nutri- Washington, DC: National Academy Press.
ents and co-infections. The effect should also be assessed Frieden TR, Sterling TR, Munsiff SS, Watt CJ & Dye C (2003)
among sputum-positive TB patients without HIV co-infection, Tuberculosis. Lancet 362, 887–899.
and among sputum-negative pulmonary TB patients. Friis H (2005) Micronutrients and HIV Infection: A Review of Cur-
The cost is around US$8 for the full supply of supplements, rent Evidence. www.who.int/nutrition
which can easily be delivered to the patients together with TB Friis H & Sandström B (2001) Zinc and HIV infection. In Micronu-
drugs. If proven beneficial, then MVM supplementation could trients and HIV Infection, pp. 159–181 [H Friis, editor]. Boca
be a feasible and cost-effective adjunctive intervention during Raton, FL: CRC Press.
Githui W, Kitui F, Juma ES, Obwana DO, Mwai J & Kwamanga D
TB treatment.
(1993) A comparative study on the reliability of the fluorescence
microscopy and Ziehl-Neelsen method in the diagnosis of pulmon-
ary tuberculosis. East Afr Med J 70, 263– 266.
Golden MHN (1992) Nutritional deficiency as a cause of growth fail-
Acknowledgements
ure. In Human Growth: Basic and Clinical Aspects, pp. 175– 182
The study was funded by the Danish International Develop- [M Hernández and J Argente, editors]. Amsterdam: Elsevier
ment Assistance through the Council for Development Science Publishers B.V.
Research and DBL Institute for Health Research and Develop- Gomo E, Vennervald BJ, Ndhlovu P, Kaestel P, Nyazema N & Friis
H (2004) Predictors and reference values of CD4 and CD8 T lym-
ment.The authors would like to thank all staff from health
phocyte counts in pregnancy: a cross sectional study among HIV
facilities, laboratories, and the National Institute for Medical negative women in Zimbabwe. Cent Afr J Med 50, 10 –19.
Research, Mwanza Centre for their valuable contribution to Harries AD, Hargreaves NJ, Kemp J, Jindani A, Enarson DA, Maher
the study, and the patients for participating. N. R., P. M., D & Salaniponi FM (2001) Deaths from tuberculosis in sub-
A. B. A. and H. F. developed the protocol. N. R. was respon- Saharan African countries with a high prevalence of HIV-1.
sible for the conduct of the trial and for data management, Lancet 357, 1519– 1523.
under supervision of J. C., P. M., A. B. A. and H. F.; H. K. Jiamton S, Pepin J, Suttent R, Filteau S, Mahakkanukrauh B, Han-
was responsible for the viral load determinations. H. F. shaoworakul W, Chaisilwattana P, Suthipinittharm P, Shetty P &
suggested the study, analysed and interpreted the data and Jaffar S (2003) A randomized trial of the impact of multiple micro-
wrote the first version of the manuscript. All authors contrib- nutrient supplementation on mortality among HIV-infected indi-
viduals living in Bangkok. AIDS 17, 2461– 2469.
uted to the final version of the manuscript. There were no con-
Karyadi E, Schultink W, Nelwan RH, Gross R, Amin Z, Dolmans
flicts of interests. The funding source had no influence on WM, van der Meer JW, Hautvast JG & West CE (2000) Poor
study design; collection, analysis and interpretation of data; micronutrient status of active pulmonary tuberculosis patients in
in the writing of the report; and in the decision to submit Indonesia. J Nutr 130, 2953– 2958.
the paper for publication. Kennedy N, Ramsay A, Uiso L, Gutmann J, Ngowi FI & Gillespie SH
(1996) Nutritional status and weight gain in patients with pulmon-
ary tuberculosis in Tanzania. Trans R Soc Trop Med Hyg 90,
162–166.
References Krarup HB, Drewes AM & Madsen PH (1998) A quantitative HCV-
Beisel WR (2001) Nutritionally acquired immune deficiency syn- PCR test for routine diagnostics. Scand J Clin Lab Invest 58,
dromes. In Micronutrients and HIV Infection, pp. 23 –42 [H 415–422.
Friis, editor]. Boca Raton, FL: CRC Press. McAlister FA, Straus SE, Sackett DL & Altman DG (2003) Analysis
Bhutta ZA, Bird SM, Black RE, et al. (2000) Therapeutic effects of and reporting of factorial trials: a systematic review. JAMA 289,
oral zinc in acute and persistent diarrhea in children in developing 2545 –2553.
countries: pooled analysis of randomized controlled trials. Am J Ministry of Health (1991) Manual of the National Tuberculosis and
Clin Nutr 72, 1516 – 1522. Leprosy Programme (NTLP). Geneva: Ministry of Health.
Bhutta ZA, Black RE, Brown KH, et al. (1999) Prevention of diarrhea Mugusi FM, Rusizoka O, Habib N & Fawzi W (2003) Vitamin A
and pneumonia by zinc supplementation in children in developing status of patients presenting with pulmonary tuberculosis and
countries: pooled analysis of randomized controlled trials. Zinc asymptomatic HIV-infected individuals. Dar es Salaam, Tanzania.
Investigators’ Collaborative Group. J Pediatr 135, 689 – 697. Int J Tuberc Lung Dis 7, 804–807.
Cegielski JP & McMurray DN (2004) The relationship between mal- Paton NI, Chua YK, Earnest A & Chee CBE (2004) Randomized con-
nutrition and tuberculosis: evidence from studies in humans and trolled trial of nutritional supplementation in patients with
experimental animals. Int J Tuberc Lung Dis 8, 286 –298. newly diagnosed tuberculosis and wasting. Am J Clin Nutr 80,
Fawzi WW, Mbise RL, Hertzmark E, Fataki MR, Herrera MG, 460–465.
Ndossi G & Spiegelman D (1999) A randomized trial of vitamin Ramakrishnan U & Huffman SL (2001) Multiple micronutrient mal-
A supplements in relation to mortality among human immunodefi- nutrition. In Nutrition and Health in Developing Countries, pp.
ciency virus-infected and uninfected children in Tanzania. Pediatr 365–391 [RD Semba and MW Bloem, editors]. Totowa, NJ:
Infect Dis J 18, 127–133. Humana Press.
Fawzi WW, Msamanga GI, Hunter D, Renjifo B, Antelman G, Bang Range N, Andersen AB, Magnussen P, Mugomela A & Friis H (2005)
H, Manji K, Kapiga S, Mwakagile D, Essex M & Spiegelman D The effect of micronutrient supplementation on treatment
(2002) Randomized trial of vitamin supplements in relation to outcome in patients with pulmonary tuberculosis: a randomized
 Downloaded from https://www.cambridge.org/core. IP address: 175.176.64.130, on 09 Jul 2021 at 08:28:54, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1079/BJN20051684
770 N. Range et al.

controlled trial in Mwanza. Tanzania. Trop Med Int Health 10, UNAIDS (2004) Report on the Global AIDS Epidemic: 4th Global
826–832. Report. Geneva: UNAIDS.
Shankar AH & Prasad AS (1998) Zinc and immune function: the bio- Wilkinson RJ, Llewelyn M, Toossi Z, Patel P, Pasvol G, Lalvani A,
logical basis of altered resistance to infection. Am J Clin Nutr 68, Wright D, Latif M & Davidson RN (2000) Influence of vitamin D
447S– 463S. deficiency and vitamin D receptor polymorphisms on tuberculosis
Tang AM, Graham NM, Kirby AJ, McCall LD, Willett WC & Saah among Gujarati Asians in west London: a case-control study.
AJ (1993) Dietary micronutrient intake and risk of progression to Lancet 355, 618–621.
acquired immunodeficiency syndrome (AIDS) in human immuno- World Health Organization (1991) Basic Laboratory Methods in
deficiency virus type 1 (HIV-1)-infected homosexual men. Am J Medical Parasitology. Geneva: WHO.
Epidemiol 138, 937– 951. World Health Organization (1995) Model Prescription Information:
Tang AM, Graham NM & Saah AJ (1996) Effects of micronutrient Drugs Used in Parasitic Diseases. Geneva: WHO.
intake on survival in human immunodeficiency virus type 1 infec- World Health Organization (2003) Treatment of Tuberculosis:
tion. Am J Epidemiol 143, 1244– 1256. Guidelines for National Programmes. Geneva: WHO.
Thurnham DI (2005) An overview of interactions between micronu- World Health Organization (2004) WHO Report 2003. Global
trients and of micronutrients with drugs, genes and immune Tuberculosis Control: Surveillance, Financing and Planning.
mechanisms. Nutr Res Rev 17, 211 –240. Geneva: WHO.