936

CONCISE COMMUNICATION

Association of Pulmonary Tuberculosis with Increased Dietary Iron

Innocent T. Gangaidzo,1 Victor M. Moyo,4 Departments of 1Medicine and 2Chemical Pathology, University
Elisha Mvundura,1 George Aggrey,6 of Zimbabwe School of Medicine, Harare, and 3Nyadire Mission
Hospital, Mutoko, Zimbabwe; 4Johns Hopkins Oncology Center,
Nyasha L. Murphree,3 Hlosukwazi Khumalo,1
Johns Hopkins University, Baltimore, and 5Cell Biology
Thokozile Saungweme,1 Ishmael Kasvosve,2 and Metabolism Branch, National Institute of Child Health
Zvenyika A. R. Gomo,2 Tracey Rouault,5 and Human Development, National Institutes of Health, Bethesda,
Johan R. Boelaert,8 and Victor R. Gordeuk7 Maryland; 6School of Public Health, George Washington University
Medical Center, and 7Center for Sickle Cell Disease, Howard
University, Washington, DC; 8Unit of Renal and Infectious Diseases,
Algemeen Ziekenhuis Sint-Jan, Brugge, Belgium

Downloaded from http://jid.oxfordjournals.org/ at University of Georgia on June 20, 2015

To determine whether increased dietary iron could be a risk factor for active tuberculosis,
dietary iron history and human immunodeficiency virus (HIV) status were studied in 98
patients with pulmonary tuberculosis and in 98 control subjects from rural Zimbabwe. Ex-
posure to high levels of dietary iron in the form of traditional beer is associated with increased
iron stores in rural Africans. HIV seropositivity was associated with a 17.3-fold increase in
the estimated odds of developing active tuberculosis (95% confidence interval [95% CI], 7.4–
40.6; P ! .001), and increased dietary iron was associated with a 3.5-fold increase (95% CI,
1.4–8.9; P p .009). Among patients treated for tuberculosis, HIV seropositivity was associated
with a 3.8-fold increase in the estimated hazard ratio of death (95% CI, 1.0–13.8; P p .046),
and increased dietary iron was associated with a 1.3-fold increase (95% CI, 0.4–6.4; P p .2).
These findings are consistent with the hypothesis that elevated dietary iron may increase the
risk of active pulmonary tuberculosis.

Mycobacterium tuberculosis, which grows within phagosomes among these populations has been attributed solely to ingestion
of macrophages, has an elaborate system of mycobactin sidero- of high levels of dietary iron in a traditional fermented beverage,
phores for acquiring and storing iron [1, 2]. The growth of M. recent studies indicate that a genetic predisposition may also
tuberculosis in serum [3] and in mice [4] is enhanced by the be a factor [8]. Iron loading of macrophages could, conceivably,
addition of iron. On the other hand, the addition of iron inhibits enhance the growth of M. tuberculosis and impair the ability
experimental infection with M. tuberculosis in monocytes under of macrophages to suppress invading microorganisms. A recent
certain conditions [5]. Furthermore, the addition of monocytes statistical analysis of an autopsy series collected in South Africa
to cultures of M. tuberculosis in serum is inhibitory, with or in the 1920s found an association between high macrophage
without iron supplementation [6]. iron stores and death from tuberculosis [9]. Also, in a retro-
Dietary iron overload, which affects ⭓10% of some rural spective study of bone marrow iron stores, a history of tuber-
African populations, causes heavy iron deposits within both culosis was more common in human immunodeficiency virus
macrophages and parenchymal cells [7]. Although the condition (HIV)–infected patients with elevated iron levels [10]. In the
present study, we explored a potential relationship between a
history of dietary iron and tuberculosis in a setting where HIV
Received 25 September 2000; revised 24 May 2001; electronically pub-
lished 17 August 2001. infection is common.
Presented in part: 2d International Conference on HIV and Iron, Brugge,
Belgium, 31 March to 2 April 2000.
Written informed consent was obtained from all study participants. Human
experimentation guidelines of the Medical Research Council of Zimbabwe and Subjects and Methods
the US Department of Health and Human Services were followed.
Financial support: Office of Minority Health to the Cell Biology and Study subjects. We studied 98 patients with pulmonary tuber-
Metabolism Branch of the National Institute of Child Health and Human culosis from Nyadire Mission Hospital, Mutoko District, Zim-
Development (NICHD); Cell Biology and Metabolism Branch of NICHD babwe, and 98 community control subjects matched by age, sex,
(HD 3-3196); J. F. Kapnek Charitable Trust (Harare, Zimbabwe).
and area of residence. A preliminary analysis of haptoglobin poly-
Reprints or correspondence: Dr. I. T. Gangaidzo, Dept. of Medicine,
University of Zimbabwe School of Medicine, Box A178, Avondale, Harare, morphisms in these subjects has been published elsewhere [11].
Zimbabwe (gangaidz@ecoweb.co.zw). History of increased dietary iron. We estimated the amount of
traditional beer, which is prepared at home from local grains in
The Journal of Infectious Diseases 2001; 184:936–9
䉷 2001 by the Infectious Diseases Society of America. All rights reserved. nongalvanized steel containers, consumed over each subject’s life-
0022-1899/2001/18407-0020$02.00 time as an indicator of exposure to increased dietary iron. The
JID 2001;184 (1 October) Tuberculosis and Dietary Iron 937

beverage has a high ferrous iron level and a low alcohol content, patients with pulmonary tuberculosis, with adjustment for age and
and consumption is associated with increased iron stores, as as- serum AST level.
sessed directly by examination of liver tissue and indirectly by mea-
suring serum ferritin concentration and transferrin saturation [7].
The estimate of traditional beer consumption was based on inter- Results
views conducted before blood samples were obtained by researchers
who were fluent in the local languages and knowledgeable of the Patients with tuberculosis and control subjects. Demographic
local culture. The number of liters of traditional beer consumed in and clinical characteristics for patients and control subjects are
a typical beer-drinking day was multiplied by the number of days summarized and compared in table 1. For patients with pul-
the individual usually drank per month, and this monthly total monary tuberculosis, the laboratory measurements of hemato-
was then multiplied by 12 times the number of years during which logic and iron status presented in table 1 were done during weeks
the individual had drunk traditional beer. This estimate provides 1–3 of treatment. HIV seropositivity was more common in the
a broad approximation of lifetime traditional beer consumption,
patients with tuberculosis than in control subjects, whereas high
because consumption may not be uniform over time and because
levels of dietary iron in the form of traditional beer was com-
information was obtained by recollection. Increased dietary iron

Downloaded from http://jid.oxfordjournals.org/ at University of Georgia on June 20, 2015

was defined as an estimated lifetime consumption of 11000 L of parable. Compatible with the presence of an active and chronic
traditional beer [12]. infection, patients with pulmonary tuberculosis had lower hemo-
Treatment for pulmonary tuberculosis. Patients were hospital- globin concentrations and higher ESRs and serum ferritin con-
ized for 2 months for treatment with isoniazid (usual dose, 300 mg/ centrations, compared with those for control subjects. Patients
day), rifampicin (10 mg/kg/day; usual dose, 450–600 mg/day), and with tuberculosis also had higher transferrin saturations.
ethambutol (25 mg/kg/day; usual dose, 800–1200 mg/day) by Serial measurements of iron and hematologic status in tuber-
mouth, administered under direct observation, and streptomycin culosis patients. Serial measurements during the 6 months of
(15 mg/kg/day) by intramuscular injection. Thereafter, isoniazid antituberculosis treatment and 3 additional months are sum-
and rifampicin were continued at the same doses for another 4 marized in table 2, according to the subject’s HIV status. Serum
months on an outpatient basis, with the patients being followed
ferritin concentrations did not differ significantly according to
up monthly at the hospital. Patients were followed up from 1 week
HIV status at any time point, and transferrin saturations dif-
to 9 months after the start of treatment.
fered on only one occasion. In contrast, the hemoglobin con-
Blood collection and analysis. Among patients with pulmonary
tuberculosis, an effort was made to obtain fasting, morning blood centrations were significantly lower, and the ESRs were signifi-
samples on a weekly basis during the 2-month hospitalization and cantly higher in HIV-seropositive subjects at most time points.
then monthly, until 9 months from the start of therapy. Among Taken as a whole, serum ferritin levels decreased significantly
control subjects, blood was obtained for 2 successive days. Vitamin
C (2 g) was given orally to each subject 24 h before blood samples Table 1. Clinical and demographic characteristics for patients with
were obtained. Erythrocyte sedimentation rates (ESRs), complete pulmonary tuberculosis and for control subjects.
blood cell counts, serum aspartate aminotransferase (AST) levels, Patients with
serum ferritin concentrations (Ramco Spectroferritin Enzyme Im- pulmonary Control
munoassay), serum iron concentrations, and total iron-binding ca- tuberculosis subjects
Characteristic (n p 98) (n p 98) P
pacities (methods modified from the International Committee for
Standardization in Hematology) were measured, and transferrin Age, mean years Ⳳ SD 36 Ⳳ 13 37 Ⳳ 14
saturations were calculated. The HIV 1&2 Rapid Test (Makro Male:female 54:44 54:44
No. (%) with high levels of
Medical Supplies) was used to test for antibodies to HIV-1. dietary iron, as deter-
Statistical analysis. After preliminary inspection, blood test mined by consumption of
results for patients with tuberculosis were grouped as follows: 1–3, traditional beer 23 (23.5) 20 (20.4) .7
4–6, and 7–9 weeks and 3, 4–6, and 7–9 months after the start of No. (%) HIV positive 68 (69) 15 (15) !.001
a b
AST level, mean U/L Ⳳ SD 38 Ⳳ 17 29 Ⳳ 11 !.001
treatment for tuberculosis. Continuous variables were compared
Hemoglobin level,
by use of Student’s t test, and proportions were compared by use mean g/dL Ⳳ SD 9.4 Ⳳ 2.1
c
14.3 Ⳳ 1.8 !.001
of Fisher’s exact test. Serum ferritin concentrations followed a Erythrocyte sedimentation
c
skewed distribution and were log-transformed. Analysis of variance rate, mean mm/h Ⳳ SD 108 Ⳳ 43 26 Ⳳ 24 !.001
Ferritin level, geometric
with adjustment for age and sex was used to compare measures of c
mean mg/L (SD range) 498 (152–1517) 70 (21–236) !.001
iron status in control subjects, according to dietary iron exposure. Ferritin:AST level, geomet-
c b
In patients with tuberculosis, linear regression was used to deter- ric mean mg/U (SD range) 13.5 (4.4–41.4) 2.5 (0.7–8.6) !.001
mine the relationship between measurements at weeks 1–3 and at Transferrin saturation,
c
mean % Ⳳ SD 43 Ⳳ 21 33 Ⳳ 16 .001
succeeding time points. Logistic regression modeling was used to
examine the relationships of dietary iron exposure and HIV sero- NOTE. AST, aspartate aminotransferase; HIV, human immunodeficiency
positivity with the development of tuberculosis among all 196 sub- virus infection.
a
n p 93; Measured at study entry (usually weeks 1–3 of antituberculosis
jects, with adjustment for serum AST activity. Cox proportional treatment).
hazards modeling was used to examine the relationships of dietary b
n p 97.
c
iron exposure and HIV seropositivity with mortality among the 98 n p 74–77; Measured during weeks 1–3 of antituberculosis treatment.
938 Gangaidzo et al. JID 2001;184 (1 October)

Table 2. Chronological summary of measurements of iron status in Iron status according to dietary iron content. To confirm re-
patients with tuberculosis at Nyadire Mission Hospital, Zimbabwe, sults from other studies demonstrating higher iron status with
according to human immunodeficiency virus (HIV) infection status.
increased dietary iron, we compared indirect measures of iron
HIV-negative patients HIV-positive patients
status in control subjects, according to the presence or absence
Measurement, time Value n Value n P of increased dietary iron. After adjustment for age and sex, the
Serum ferritin geometric mean serum ferritin concentration was 115 mg/L (SE
concentration, range, 90–148 mg/L) in 20 control subjects with a history of in-
a
ng/mL
Weeks creased dietary iron, compared with 62 mg/L (SE range, 56–70
1–3 411 (145–1169) 24 514 (155–1706) 52 .5 mg/L) in 78 subjects without increased dietary iron (P p .038).
4–6 317 (92–1094) 25 321 (100–1030) 40 1.0 Similarly, the ratio of serum ferritin concentration to AST level
7–9 232 (63–861) 21 236 (56–989) 35 1.0
Months was 4.9 mg/U (SE range, 3.8–6.5 mg/U) in control subjects with
3 161 (48–542) 16 138 (33–577) 37 .7 increased dietary iron, compared with 2.1 mg/U (SE range, 1.9–2.4
4–6 126 (39–338) 21 91 (26–318) 41 .3 mg/U) in those without increased dietary iron (P p .009). The
7–9 87 (19–401) 10 47 (18–119) 11 .3
adjusted mean Ⳳ SE transferrin saturation was 39% Ⳳ 4% in

Downloaded from http://jid.oxfordjournals.org/ at University of Georgia on June 20, 2015

Transferrin
b
saturation, % control subjects with increased dietary iron, compared with
Weeks 32% Ⳳ 2% in those without increased dietary iron (P p .11).
1–3 43 Ⳳ 19 24 43 Ⳳ 22 53 .9
4–6 38 Ⳳ 14 25 39 Ⳳ 18 40 .9 Relationship between dietary iron and development of pul-
7–9 31 Ⳳ 14 20 41 Ⳳ 18 35 .035 monary tuberculosis in patients and control subjects. Among
Months 84 subjects who were both HIV seronegative and negative for
3 32 Ⳳ 15 17 37 Ⳳ 16 37 .2
4–6 42 Ⳳ 22 22 40 Ⳳ 18 41 .7 increased dietary iron, 17 (20.2%) had tuberculosis. Among 29
7–9 51 Ⳳ 26 9 38 Ⳳ 12 11 .2 subjects who were HIV seronegative but were positive for in-
Hemoglobin creased dietary iron, 13 (44.8%) had tuberculosis. Among 69
b
level, g/dL
Weeks subjects who were HIV seropositive but were negative for in-
1–3 10.8 Ⳳ 2.1 23 8.8 Ⳳ 1.9 53 !.005 creased dietary iron, 58 (84.1%) had tuberculosis. Among 14
4–6 11.1 Ⳳ 2.1 27 10.0 Ⳳ 1.4 59 .02 subjects who were both HIV seropositive and positive for in-
7–9 11.8 Ⳳ 1.8 28 10.7 Ⳳ 1.6 58 .008
Months creased dietary iron, 10 (71.4%) had tuberculosis. Logistic re-
3 12.8 Ⳳ 1.9 19 11.3 Ⳳ 2.0 51 .008 gression modeling indicated that, after adjustment for the de-
4–6 13.5 Ⳳ 2.3 26 12.1 Ⳳ 1.8 47 .01 gree of dietary iron and liver function, as assessed by serum
7–9 13.9 Ⳳ 2.1 11 12.7 Ⳳ 1.9 14 .2
Erythrocyte sedi- AST level, HIV seropositivity was associated with a 17.3-fold
mentation increase in the estimated odds of developing tuberculosis (95%
b
rate, mm/h confidence interval [95% CI], 7.4–40.6; P ! .001). Modeling also
Weeks
1–3 96 Ⳳ 35 23 113 Ⳳ 45 53 .1 indicated that, after adjustment for HIV status and liver func-
4–6 89 Ⳳ 42 27 118 Ⳳ 32 60 .002 tion, increased dietary iron was associated with a 3.5-fold in-
7–9 81 Ⳳ 36 28 110 Ⳳ 34 58 .001 crease (95% CI, 1.4–8.9; P p .009).
Months
3 62 Ⳳ 47 19 95 Ⳳ 40 52 .01 Mortality in patients with pulmonary tuberculosis. Eighteen
4–6 47 Ⳳ 44 26 86 Ⳳ 41 47 !.005 of the 98 patients died during the period up to 9 months after
7–9 31 Ⳳ 38 10 68 Ⳳ 43 15 .033 the start of antituberculosis treatment. Compared with the pa-
NOTE. Antituberculosis treatment was given for 6 months. P values indicate tients who survived, those who died were older, tended more
significance level for comparison between HIV-negative and HIV-positive subjects
often to have a history of high levels of dietary iron, and had
at each time point.
a
Data are mean (SD range). higher initial serum ferritin concentrations and transferrin satu-
b
Data are mean Ⳳ SD. rations. Cox proportional hazards modeling indicated that, after
adjustment for age, dietary iron exposure, and liver function, as
between all time periods from 1–3 weeks to 7–9 months (P ⭐
assessed by serum AST level, HIV seropositivity was associated
.023), hemoglobin concentrations increased significantly be-
with an estimated 3.8-fold increase in the hazard ratio of death
tween all time periods up to 4–6 months (P ! .005 ), and ESRs
(95% CI, 1.0–13.8; P p .046). The modeling also indicated that,
decreased significantly for each time period between 7–9 weeks
after adjustment for age, HIV status, and serum AST level, pa-
and 7–9 months (P ⭐ .033). Serum ferritin concentrations and
tients with increased dietary iron had a 1.3-fold increase in the
hemoglobin levels determined during weeks 1–3 of antituber-
hazard ratio of death (95% CI, 0.4–6.4; P p .2).
culosis treatment correlated strongly with measurements during
each successive time period (r ⭓ .7 and P ⭐ .002 for serum
ferritin concentration and r ⭓ .5 and P ⭐ .021 for hemoglobin Discussion
concentration). Transferrin saturations determined during
weeks 1–3 also correlated significantly with later measurements Because inflammation associated with active tuberculosis
at most time periods. tends to increase serum ferritin concentration and decrease
JID 2001;184 (1 October) Tuberculosis and Dietary Iron 939

transferrin saturation [13], in this study, we chose to gauge iron and tuberculosis are common, the possible association of high
status by exposure to increased dietary iron in the form of levels of dietary iron with tuberculosis has potential public
traditional beer. Estimated traditional beer consumption has health implications. Of note, iron overload may be found in
been shown to correlate with iron status in both community urban areas where consumption of traditional beer is less com-
and hospitalized rural African subjects [7, 8], and we observed mon [15]. Further studies to examine the relationship between
serum ferritin concentrations to be significantly higher in con- iron status and tuberculosis seem to be warranted.
trol subjects with increased dietary iron, even after adjustment
for liver dysfunction, as reflected in serum AST level. We found References
a significant association between exposure to high levels of di-
1. De Voss JJ, Rutter K, Schroeder BG, Barry CE. Iron acquisition and meta-
etary iron and the presence of pulmonary tuberculosis. We also
bolism by mycobacteria. J Bacteriol 1999; 181:4443–51.
found a trend toward higher mortality in the patients with 2. De Voss JJ, Rutter K, Schroeder BJ, Su H, Zhu Y, Barry CE. The salicylate-
tuberculosis who had exposure to high levels of dietary iron, derived mycobactin siderophores of Mycobacterium tuberculosis are essential
but the trend was not statistically significant. for growth in macrophages. Proc Natl Acad Sci USA 2000;97:1252–7.

Downloaded from http://jid.oxfordjournals.org/ at University of Georgia on June 20, 2015

Our study, which was performed at a rural African hospital, 3. Kochan I. The role of iron in bacterial infections, with special considerations
of host–tubercle bacillus interaction. Curr Top Microbiol Immunol 1973;
has several limitations. First, iron status was estimated indi-
60:1–30.
rectly through history of increased dietary iron rather than di- 4. Lounis N, Maslo C, Boelaert JR, et al. Impact of iron loading and iron
rectly by bone marrow aspirate or liver biopsy. Second, we chelation on murine tuberculosis. Clin Microbiol Infect 1999; 5:687–92.
could not measure CD4⫹ T cell counts and HIV-1 RNA levels 5. Byrd TF. Tumor necrosis factor a (TNF-a) promotes growth of virulent
and thus were unable to determine the stage of HIV infection. Mycobacterium tuberculosis in human monocytes: iron-mediated growth
suppression is correlated with decreased release of TNF-a from iron-
Third, increased dietary iron in rural Africa derives from a
treated monocytes. J Clin Invest 1997; 99:2518–29.
fermented beverage [7, 12], and alcoholism itself increases the 6. Rook GA, Steele J, Ainsworth M, Champion BR. Activation of macrophages
risk of developing tuberculosis [14]. However, African tradi- to inhibit proliferation of Mycobacterium tuberculosis: comparison of the
tional beer has low alcohol content, and specific histologic effects of recombinant gamma-interferon on human monocytes and mu-
changes of alcoholism are usually not found at liver biopsy in rine peritoneal macrophages. Immunology 1986; 59:333–8.
7. Gordeuk VR. Hereditary and nutritional iron overload. Balliere’s Clin Hae-
patients with African iron overload [7]. Fourth, iron overload
matol 1992; 5:169–86.
may be associated with cirrhosis [7], and we were unable to 8. Moyo VM, Mandishona E, Hasstedt SJ, et al. Evidence of genetic trans-
determine whether cirrhosis was present in the subjects with mission in African iron overload. Blood 1998; 91:1076–82.
increased dietary iron. 9. Gordeuk VR, McLaren CE, MacPhail AP, Deischel G, Bothwell TH. As-
Although our findings seem to be consistent with the hy- sociation of iron overload in Africa with hepatocellular carcinoma and
tuberculosis: Strachan’s 1929 thesis revisited. Blood 1996; 87:3470–6.
pothesis that macrophage iron-loading in African iron overload
10. de Monyé C, Karcher DS, Boelaert JR, Gordeuk VR. Bone marrow macro-
[7] may be associated with impaired macrophage function and phage iron grade and survival of HIV-seropositive patients. AIDS 1999;
increased risk for active tuberculosis, there are alternative in- 13:375–80.
terpretations. In view of in vitro work indicating that added 11. Kasvosve I, Gomo ZAR, Mvundura E, et al. Haptoglobin polymorphism
iron may inhibit the growth of M. tuberculosis in macrophages, and mortality in patients with tuberculosis. Int J Tuberc Lung Dis 2000;4:
771–6.
under certain circumstances [5], it is possible that the increased
12. Moyo VM, Gangaidzo IT, Gomo ZA, et al. Traditional beer consumption
transferrin saturation in patients with high levels of dietary iron and the iron status of spouse pairs from a rural community in Zimbabwe.
(table 1) might open up the extracellular space as an additional Blood 1997; 89:2159–66.
niche for M. tuberculosis replication. Also, if any of the subjects 13. Morris CD, Bird AR, Nell H. The hematological and biochemical changes
of severe pulmonary tuberculosis. Q J Med 1989; 73:1151–9.
with high levels of dietary iron had cirrhosis, then the general
14. Feingold AO. Association of tuberculosis with alcoholism. South Med J
immunocompromised status associated with cirrhosis could 1976; 69:1336–7.
represent a risk for tuberculosis independent of iron status. 15. Gangaidzo IT, Moyo VM, Saungweme T, et al. High prevalence of iron
In sub-Saharan Africa, where iron overload, HIV infection, overload in urban Africans in the 1990s. Gut 1999; 45:278–83.