- Email: [email protected]
Tuberculosis: associations with HIV and socioeconomic status in rural Malawi
Tuberculosis:
associations
with
HIV and socioeconomic
status in rural Malawi
J. R. Glynn’, D. K. WarndorfFJ, S. S. Malema 2, V. Mwinukaz, J. M. P6nnighsus2, A. C. Crampin’ and P. E. M. Fine’ ‘London School of Hygiene and Tropical Medicine, Keppel Street, London WClE 7HT, UK; 2Karonga Prevention Study, PO Box 46, ChiZumba, Malawi Abstract Tuberculosis (TB) is associated with human immunodeficien~ virus (HIV) infection, increasing age and male sex, but less is known about other risk factors in developing countries. As part of the Karonga Prevention Study in northern Malawi, we conducted a retrospective cohort study in the general population to assess risk factors for the development of TB. Individuals were identified in 1986-89 and TB cases diagnosed up to 1996 were included. TB was confirmed in 62/l 1059 (0.56%) HIV negative individuals and 7/182 (3.9%) HIV positive individuals (relative risk 7.1, 95% confidence interval 3.2-15.7). This association was little altered by adjustment for age, sex or socioeconomic factors. The risk of TB was higher in those aged over 30 years than in younger individuals, in men than in women, in those engaged in occupations other than farming than in subsistence farmers, in those living in households with burnt brick dwellings than in those with less well built dwellings, and in those with some schooling than in those with none. These associations persisted after adjusting for age, sex, HIV status and population density. The absolute risks of TB were low in this study due to the passive follow-up and strict diagnostic criteria. The relative risk with HIV was of a similar magnitude to that measured elsewhere. Increased risks of TB with age and in men are expected. Associations with measures of higher socioecomomic status were unexpected. They may reflect a greater likelihood of diagnosis in this group. Keywords: tuberculosis, human iramunodeficiency virus, socioeconomic status, Malawi Introduction Human immunodeficiencv virus (HIV) is known to increase the risk of tuberculo&s (TB); but the magnitude of this increased risk is less certain and is likely to depend on degree of immunosuppression. Relative risks measured in different studies have varied from less than 2 to more than 20 (GLYNN, 1998). Cohort studies to date have been conducted in selected ~ouulations: antenatal clinic attenders, intravenous drug&&s, and commercial sex workers (SELWYN e6 al., 1989, 1992; BRAUN er ad., 199 I; ALLEN et aZ., 1992; LEROY et at., 1995; GIUCS et al., 1997; KJZIZER et al., 1997). The risks and relative risks of TB in these arouus mav be different from those in the general populaiion.* Many case-control studies of HIV and TB have also been conducted, but some have been marred by the use of unsatisfactory control groups or lack of adjustment for confounding (GLYNN et al., 1997). TB risks vary with age and sex, but less is known about other risk factors in developing countries. In industrialized countries, TB is strongly associated with poverty (DARBYSHIRE, 1995), but the association is less clear in Africa (SCHO~AN hi aZ., 1991) and is likely to change with HIV infection. We have assessed the factors influencing the development of TB in a general population cohort with known HIV status in a rural area of northern Malawi. Methods This investigation was conducted as part of the Karonga Prevention Study, a continuing study of mycobacterial disease and HIV in Karonga District. Malawi. The methods of the fuii study hav: been de&bed in detail elsewhere (P~~~HA~s et aZ., 1993; KARONGA I%EV?ZNTION TRIAL GROUP, 1996). House-to-house surveys of the total population (now estimated at 180 000) were conducted in 1981-84 and again in 1986-89 to assess the mevalence and incidence of leprosy and tuberculosis: The second survey formed the recruitment chase of a vaccine trial of repeat BCG (bacillus Calmetie-Guerin) and BCG plus killed Mycobacterium Zeprae. In the first survey, information was collected on education, occupation and housing conditions. In the second survey the schooling level of those Address for correspondence: Dr Judith Glynn, Infectious Disease EpidemioIo~ Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WClE 7HT, UK; e-mail [email protected]
currently in school was recorded, but no other socioeconomic information. Since the end of the second survey, Karonga Prevention Study staff have been stationed in health centres and the hospital to assess attenders for TB and leprosy, and additional house-tohouse surveys were conducted in selected areas. Patients with a cough of more than 3 weeks’ duration were encouraged to provide 3 sputum specimens, and patients with suspected extrapulmonary TB were referred for biopsv or further diagnostic procedures as appropriate. $putum smears were examined and cultures set up in the oroiect laboratorv in Chilumba, Malawi. Cultures whiih were macroscopically consistent with M. tuberculosis were sent to the UK for species identification and drug sensitivity testing. All diagnosed TB patients started treatment in the hospital, following Malawi government guidelines. As nart of an investigation of serological tests for leprosy, filter paper blood specimens were collected between 1987 and 1989 from all individuals in 2 rural areas at opposite ends of the district. The leprosy tests did not proveuseful and the specimens have continued to be stored at -20°C. With permission from the Health Sciences Research Committee of the Malawi Ministry of Health. these soecimens were tested for HIV. A particle agglutinati& assay for anti-HIV-l was used as an initial screening test, followed byMconfirmation of positive samples using an Organon enzyme-linked immunosorbent assay and a SerodiaTM particle agglutination assay. Individuals with known HIV status, on the basis of the filter paper tests from the late 198Os, formed the cohort for this study. Those aged less than 15 years at the time of testing and those who had had previous TB were excluded. r~o~a~on on housing, sihooling and occupation from the earlier survey was used where available. Housing conditions were defined only for those whose households owned the dwellings. Past schooling was recorded in the first survey, and current schooling in both surveys. For those who were still in education at the time of questioning, the occupation of the head of the household was used as a measure of the socioeconomic status of the household. Population density at the time of the second survey was calculated for each square kilometre of the district. -All individuals included in the study had been vaccinated at least once with BCG, and we have found no evidence that BCG is associated with TB in this population (KARONGA PREVENTION TRIAL GROUP, 1996).
TBANDHIVINhlALAWl
For the purposes of this study, patients were classified as having developed TB ifthey fulfilled 1 of the following criteria: sputum smear positive (excluding those with single smears with fewer than 10 acid-fast bacilli per 100 fields, and specimens from which the culture grew nontuberculous mycobacteria); M. tuberczdosis cultured from the sputum or other body fluid; histology suggestive of TB; clinical diagnosis of tuberculous lymphadenonathv. All other uatients diaanosed as having TB were excluded from these analyses, TB cases diagnosed up to 1996 were included. There was no active follow-up of the copulation so results are presented as risks and-odds ratios (not rates). Absolute risks are likelv to be low. but this should not bias the relative risks (although it-increases the sampling error). HIV status was available for the whole cohort only at the start of the neriod. Subseauent acauisition of HIV would lead to underestimation of the effect of HIV, as well as a failure to correct fully for the confounding effects of HIV when considering other factors. Analyses were therefore repeated considering only the first 3 years after HIV testing. HIV status was tested on the majority ofTB cases at the time of diagnosis of TB. The risks of TB in HIV negative individuals were also reanalysed after excluding patients found to be positive (or with unknown HIV status) at the time of diagnosis. Results Results were available for 11 24 1 individuals, of whom 182 were HIV positive at the time of testing in 1987-89. Nearly 80% of individuals were seen in the first survey and so had socioeconomic data available. Sixty-nine new cases of TB fulfilling the diagnostic criteria were identified, including 2 extrapulmonary cases (one with biopsyconfirmed lymph node TB, and the other with cultureconfirmed mberculous pleural effusion). TB was diagnosed in 3.9% (7/182) HIV uositive subjects and 056% (62/l 1059) of those who-were HIV’ negative. HIV positivity increased the risk of TB about 7-fold, and this was little affected by adjusting for age, sex or the socioeconomic variables (Table). Risks ofTB were the same in the 2 areas, and between heads and members of households after adjusting for age, sex and initial HIV status. There was no associationbetween population density and TB, and no difference in TB risk amone those who were seen in the first survey and those who w&e not seen. The risk ofTB was lower under the age of 30 years, but little different between those aged 30-44 and those aged 45 years and over. Males hada higher risk of TB than females, and this was predominantly seen in the oldest age group (x2 interaction between age and sex = 4.3, P = 0.1). Odds ratios [ORs] for TB in males compared to females, after adjusting for initial HIV status, were 0.85 (95% confidence interval [95% CI] O-29-25) in those aged under 30, 1.6 (95% CI 0.74-3.4) in those aged 30-44, and 3.6 (95% CI 1.5-8.9) in the oldest group. TB was more common in those with houses of burnt brick than in those with less well built dwellings, but there was little difference between the other types,fdwelling. The risk of TB was hither in neoule in occunations other than subsistence fan&g. ‘Y&isLwas a m&ed group of salaried workers, those employed in traditional trades, and casual workers; risks were high in all these subgroups. After adjusting for age, sex and HIV status the odds-ratios for the different occupation groups compared to the farmers were 3.7 (95% CI 1.5-9.5) for salaried workers, 2.1 (95% CI 0.50-9.2) for traditional traders and 3.1 (95% CI 1.0-9.1) for casual workers. These odds ratios were onlv sli&lv reduced bv adjusting for housing and schooling. ‘Those with no schooling had the lowest risks of TB, and narticularlv high risks were seen in the few with secondary or te&u$ education. The associations with socioeconomic variables persisted after adjustment for initial HIV status, age, sex and population density. The association with schooling was no longer
statistically significant after adjusting for housing and occupation (Table). Thirty people developed TB within 3 years, including 3 HIV positive individuals. Analysis of risk factors for TB within 3 years gave similar results to those obtained ti-om the full data set except that housing conditions no longer influenced development of TB. There was still a strong association with occupations other than farming (OR 4.0, 95% CI 1*5-10.8, after adjusting for age, sex and HIV) and with schooling. Of the 62 TB cases who were HIV negative at the time of the initial test, 8 were HIV nositive bv the time of diagnosis of TB, and 7 had no test available.-These 15 individuals were excluded, and the risks of develonine TB were assessed in the 11044 initially HIV n’egayive individuals. In this group, TB was still associated with increasing age and male sex. After adjusting for age and sex, housing conditions were still associated with the development of TB (OR O-21, 95% CI 0.05-0.94 and OR 0.50, 95% CI 0.23-1.1 for the unburnt brick or pounded mud and the wood or mud groups respectively, compared to those with burnt brick ho&s), but the effect of occupation was less strong (OR 2*4.95% CI O-95-5.9, comnarina those in occunations other than farming to subsisience Tanners) and the effect of schooling was lost. Discussion The observed risks of TB measured in this study were low. There are several reasons for this. Cases were detected passively, and so many may have been missed. Even the World Health Organization (WHO) target for case finding is only 70% and it is estimated that many places fail to achieve this. We used strict criteria for diagnosing TB for this study in order to improve specificity and the validitv of the relative risk estimates. During this period a further 48 individuals from the cohort were diaanosed as having TB bv the health services. The der&ninator for the-risk estimates was all those seen and tested during the second survey. Because there was no active follow-up we do not know what proportion of the cohort was alive and still living in the district at different times and so actually at -risk of developing and being seen with TB. The risks of TB also need to be considered in relation to the risk of infection with M. tuberculosis in the district. Several studies have measured the incidence of TB among HIV positive tuberculin positive individuals as between 5 and 10% uer annum (GLYNN. 1998). In this the annual risk of i&ection*has been esti_population _ mated as less than 1% (FINE et al., 1999). The median age of the HIV nositive individuals was 30 vears. so we can estimate that only approximately 26% of them are likely to have been infected with M. tuberculosis. Relative risks may be biased if the case detection is different in the different groups, if there is differential survival or migration, or if the risk factors are measured inaccurately. Migration may be more common in those who are HIV positive than those who are HIV negative, and high mortality rates are expected among the HIV positive persons. Both these effects would tend to reduce ihe estimate of the relative risk, but the 7-fold increased risk seen for HIV is similar to that measured in casecontrol studies in the whole district at 2 different times (PhNNlGHAUS et al., 1991; GLYNN et al., 1997), and to estimates from other populations (GLYNN, 1998). The increased risks with age and in men are also expected (RIEDER et al., 1989). Socioeconomic status was defined on the basis of data from the first survey and so will fail to take into account anv changes over time, leadina to misclassification of some ind:viduals. This~miscla&ication is likely to be similar for TB cases and other individuals. and so lead to underestimation of effects. Migration out of the district may be more common for those with higher compared to lower socioeconomic status and for those who are not farmers compared to farmers. This would also tend to
T. R. GLYi’W ETAL.
502
Table.
Risk factors for the development
of tuberculosis
in Karonga
District
Malawi,
1987-1996
Odds ratios”
Risk factor HIV status Negative Positive Sex Female Male Age group (years) 15-29 30-44 245 House construction Burnt brick Unburnt brick/pounded YU;&$m;d/temporary
mud
Schooling None l-5 years primary 6-8 years primary Second~/te~a~ Unknown Occupation Farming Other Unknown Position in household Head Member Other
Adjusted for age, sex and HIV
Adjusted for age, sex, HIV, housing, schooling, and occunation
No. of subiects
No. with tuberculosis
11059 182
62 (0.56%) 7 (3.9%)
:.4 (3.3-16-6)
i-3 (2.7- 14.5)
6037 5204
27 (0.45%) 42 (0.81%)
1 l-9 (l-1-3.0)
i-5 (O-90-2.6)
4932 3059 3250
14 (0.28%) 27 (0.88%) 28 (0.86%)
:*I (l-6-6.0) 3.1 (1-6-6.0)
:.2 (1.6-6.5) 3.4 (1.6-6.8)
1506 1423 5021 3291
17 (1.1%) 4 (0728%) 27 (054%) 21 (0.64%)
2374 3616 2563 208 2480
9 (0.38%) 26 (0.72%) 10 (0.39%) 4 (1.9%) 20 (0.81%)
7757 679 2805
:; yyy I9 (0.68%)
:.2 (1.6-6-4) 1-6 (0.88-2.8)
i-9 (1.4-6~0) 1.9 (0.75-4-6)
3425 7597 219
36 (1.1%) 3; ?)43%)
0187 (0*41-1.8)
of75 (O-39-1.5)
. 0*\5 (0*085--0.76) 0.44 (0~24-0.8 1) 0.53 (0.28- 1.0)
:.o l-3 4.0 1.9
(0.93-4.4) (0-49-3-4) (1.1-138) (O-81-4.2)
0.;7 (0.09 l-0.83) 0.50 (0.27-0.96) 0.40 (0.13-1.2)
i-0 1.0 1.7 1.6
(O-89-4.3) (O-38-2*8) (0.42-6.6j (0-49-52)
Area North south Population density (per km2) Cl00 100-199 200-299 2300
6151 5090
37 (0.60%) 32 (0.63%)
2678 2693 3017 2853
15 19 16 19
(056%) (0.71%) (0.53%) (0.67%)
i-2 (0*60-2.4) 0.92 (0.45-1.9) 1.2 (0.60-2.4)
l-2 (0.58-2.3) 0.92 (0.45- 1.9) 1.0 (0.52-2-l)
Seen in first survey Individual Household/family No one
8918 938 1385
53 (0.59%) 7 (0.75%) 9 (0.65%)
:.6 (0.70-3.5) 0.97 (O-46-2.0)
Note b -
member
i-0 (O-64- 1.7)
0194 (O-57-1.6)
“95% confidence interval in parentheses. bSocioeconomic factors known only for those seen in first survey.
reduce the relative risks, though this may be offset by any excess mortality among those with lower socioeconomic status compared to those with higher status. Population density was defined per square kilometre and may not reflect the density in the immediate locality of the individual. The HIV status was defined for all individuals only at the start of the study, HIV has increased rapidly in this district over the period of the study, so some individuals classified as negative will have become positive @LYNN et al., 1997). This leads to underestimation of the relative risk of TB by HIV status and to a failure to control completely for the confounding effect of HIV when considering other risk factors. Two approaches were used to investigate this further: first, considering only the first 3 years of follow-up (so that the role of incident HIV
cases leading to TB was minimized); and second, considering HIV negative individuals only and excluding HIV positive (or HIV status unknown) TB cases. In these analyses better housing and occupations other than farming were still associated with TB, after adjusting for other factors. The effect of schooling was lost in the analysis restricted to HIV negative individuals, and in all analyses after adjusting for occupation and housing. These results suggest that, although the associations between TB and measures of higher socioeconomic status could be partly due to residual confounding due to HIV incidence, this was not the full explanation. Relations between poverty and TB caot be assnmed to be the same in At&a as they are in ‘western’ countries. Indeed, even within 1 country they may be different in different ethnic groups (HAWKER et af., 1999). If urban
503
overcrowding is one of the key elements of poverty which leads to increased TB risks in ‘western’ nations, it would not be surprising if associations were different in a rural African community. ‘Better’ housing might increase the risk of transmission of M. tuberculosis due to poor ventilation, and the outdoor life of farmers could be protective. In England and Wales the standardized mortality ratio for deaths from respiratory tuberculosis among agricultural labourers is consistently low (data from decennial supplements from the UK Office of Population Censuses and Surveys), in line with lower rates of tuberculosis in rural than in urban areas (HOR%‘I%?2et al., 1969). This contrasts with the USA, where very high rates of tuberculosis have been found among migrant farm labourers, attributable to substandard living conditions and poor health care (CIESIELSIU et al.. 1991. 1994: MCKBNNA etai.. 1996). TB could be associated with higher” socioeconomic status if this increased the risk of acquiring infection (for example, by increased travel) or due to an increased likelihood of diagnosis of the disease. It would be interesting to investigate the socioeconomic risk factors associated with TB. after adiustine for HIV. in different settings. To the extent to which these differences reflect different chances of being diagnosed with TB, they may give clues to the relative adequacy of case finding in different sections of the community. Acknowledgements Umil 1996 the Karonga Prevention Study was funded primarily by LEPRA (the British Leprosy Relief Association) and ILEI? (the International Federation of Anti-Leprosy Organizations) with contributions from the WHOAJNDPWorld Bank Special Programme for Research and Training in Tropical Diseases. Since 1996 the WellcomeTtust has been the principal funder. J. R. G. is supported by the British Department for International Development. We thank the Govemment of the Republic of Malawi for their interest in and support of the Project and the Malawi Ministry of Health for permission to publish the paper. We thank Dr Laura Rod&es for helpful discussions and Dr Dave Leon for comments on an earlier draft. References Allen, S., Batungwanayo, J., Kerlikowske, K., Lifson, A. R., Wolf, W., Granich, R., Taelman, H., van de Perre, P., Serufilira, A., Bogaerts, J., Slutkin, G. & Hopewell, P. C. (1992). Two-year incidence of tuberculosis in cohorts of HIV-infected and uninfected urban Rwandan women. Anrerican Review ofRespimtory Disease, l&1439-1444. Braun, M. M., Badi, N., Ryder, R. W., Baende, E., Mukadi, Y., Nsuami, M., Matela, B., Willame, J.-C., Kaboto, M. & Heyward, W. (199 1). A retrospective cohort study of the risk of TB among women of childbearing age with HIV infection in Zaire. American Review of Respiratoy LXrease, 143, 501-504. Ciesielski, S. D., Seed, J. R., Esposito, D. H. & Hunter, N. (1991). The epidemiology of tuberculosis among North Carolina migrant farm workers. Journal of the American Medical Association, 265, 1715- 1719. Ciesielski, S., Esposito, D., Protiva, J. & Piehl, M. (1994). The incidence of tuberculosis among North Carolina migrant farmworkers, 199 1. American 3otmd of Public Health, 84, 1836-1838. Darbyshire, J. H. (1995). Tuberculosis: old reasons for a new increase? British M~~3ou~, 310,954-955. Fine, P. E. M., Bruce, J., Pijnnighaus, J. M., Nkbosa, P., Harawa: A. & Vynnycky, E. (1999). Tuberculin sensitivity: conversions and reversions in a rural African population. International Journal of Tuberculosis and Lung Disease, 3, 962-975.
Gilks, C. F., Godfrey-Faussett, P., Batchelor, B. I. F., Ojoo, J. C., Ojoo, S. J., Brindle, R. J., Paul, J., Kimari, J., Bruce, M. C., Bwayo, J., Plummer, F. A. & Warrell, D. A. (1997). Recent transmission of tuberculosis in a cohort of HIV-l-infected female sex workers in Nairobi, Kenva. AIDS. 11.91 l-918. Glynn, J. R. (1998). Resurgenchof tuberculosis and the impact of HIV infection. British Medical Bulletin, 54,579-593. Glynn, J. R:, Warndofi, D. K., Fine, P. E. M., Msiska, G. K., Munthah, M. M. & Pdtmighaus, J. M. (1997). The impact of HIV on morbidity and mortality from tuberculosis in subSaharan Africa: a study in rural Malawi and review of the literature. Health Transirion Review, 7, supplement 2,75-87. Hawker, J. I., Bakhshi, S. S., Ah, S. & Farrington, C. P. (1999). Ecological analysis of ethnic diITerences in relation between tuberculosis and poverty. &it&k Medical 30~~4 319, 1031-1034. Horwitz, O., Wilbeck, E. & Erickson? P. A. (1969). Epidemiological basis of tuberculosis eradication. 10. Longitudinal studies on the risk of tuberculosis in the general population of a low-prevalence area. Bulletin of the World Heakk Organization, 41,95-113. Karonga Prevention Trial Group (1996). Randomised controlled trial of single BCG, repeated BCG, or combined BCG and killed Mycobactetium Zeprae vaccine for prevention of leprosy and tuberculosis in Malawi. Lancet, 348, 17-24. Keixer, S. T., Langendam, M. W., van Deutekom, H., Coutinho, R. A. & van Ameijden, E. J. C. (1997). Tuberculosis and HIV-infection in a cohort of drug users, Amsterdam 1986-1995. Zntenaatinaf Journal of Tuberculosis and Lung Disease, 1, supplement, SI57-S158. Leroy,V., Msellati, P., Lepage, P., Batungwanayo, J., Hitimana, D. G., Taelman, H., Bogaerts, J., Boineau, F., van de Perre, P., Sunonon, A., Safamon, R. & Dabis, F. (1995). Four years of natural history of HIV-l infection in African women: a prospective cohort study in Kigali (Rwanda), 1988-1993. 3oumal of Acquired Immune Delicinzcv Syndromes and Human ~bw&qy,9 415-421. - McKenna, M. T., Hutton, M., Cauthen. G. & Onorato. I. M. (1996):The association‘between occupation and tuberculosis. A population-based survey. American Journal of Respiratoy and Critical Care Medicine, 154,587-593. Ponnighaus, J. M., Mwanjasi, L. J., Fine, P. E. M., Shaw, M. A., Turner, A. C., Oxborrow, S. M., Lucas, S. B., Jenkins, P. A., Steme, J. A. C. & Bliss, L. (1991). Is HIV infection a risk factor for leprosy? International Journal of L.eprosy, 59, 221-228. Pijtmighaus, J. M., Fine, P. E. M., Bliss, L., Gruer, P. J. K., Kapira Mwamondwe, B., Msosa, E., Rees, R. J. W., Clayton, D., Pike, M. C., Steme, J. A. C. & Oxborrow, S. M. (1993). The Karonga Prevention Trial: a leprosy and tuberculosis vaccine trial in Northern Malawi-L Methods of the vaccination phase. Leprosy ReGw, G&338-356. Rieder, H. L., Cauthen, G. M., Comstock, G. W. & Snider, D. E., jr (1989). Epidemiology of tuberculosis in the United States. EnidemiologicReviews. 11.79-98. Schoeman,]. H., Westaway, MI S. & Neethling, A. (199 1). The relationship between socioeconomic factors and pulmonary tuberculosis. International Journal of Epidemiology, 20, 435-440. Selwyn, P. A., Hartel, D., Lewis, V. A., Schoenbaum, E. E., Vermund, S. H., Klein, R. S., Walker, A. T. & Friedland, G. H. (1989). A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeficiency virus infection. New EnglandJournal of Medicine, 320, 545-550. Selwyn, P. A., Sckell, B., Alcabes, P., Friedland, G. H., Klein, R. S. & Schoenbaum, E. E. (1992). High risk of active tuberculosis in HIV-infected drug users with cutaneous ~~~930um~ of the American Medical Association, 268, Received 29 March 2000
2000; accepted for publication
I9 April
associations
with
HIV and socioeconomic
status in rural Malawi
J. R. Glynn’, D. K. WarndorfFJ, S. S. Malema 2, V. Mwinukaz, J. M. P6nnighsus2, A. C. Crampin’ and P. E. M. Fine’ ‘London School of Hygiene and Tropical Medicine, Keppel Street, London WClE 7HT, UK; 2Karonga Prevention Study, PO Box 46, ChiZumba, Malawi Abstract Tuberculosis (TB) is associated with human immunodeficien~ virus (HIV) infection, increasing age and male sex, but less is known about other risk factors in developing countries. As part of the Karonga Prevention Study in northern Malawi, we conducted a retrospective cohort study in the general population to assess risk factors for the development of TB. Individuals were identified in 1986-89 and TB cases diagnosed up to 1996 were included. TB was confirmed in 62/l 1059 (0.56%) HIV negative individuals and 7/182 (3.9%) HIV positive individuals (relative risk 7.1, 95% confidence interval 3.2-15.7). This association was little altered by adjustment for age, sex or socioeconomic factors. The risk of TB was higher in those aged over 30 years than in younger individuals, in men than in women, in those engaged in occupations other than farming than in subsistence farmers, in those living in households with burnt brick dwellings than in those with less well built dwellings, and in those with some schooling than in those with none. These associations persisted after adjusting for age, sex, HIV status and population density. The absolute risks of TB were low in this study due to the passive follow-up and strict diagnostic criteria. The relative risk with HIV was of a similar magnitude to that measured elsewhere. Increased risks of TB with age and in men are expected. Associations with measures of higher socioecomomic status were unexpected. They may reflect a greater likelihood of diagnosis in this group. Keywords: tuberculosis, human iramunodeficiency virus, socioeconomic status, Malawi Introduction Human immunodeficiencv virus (HIV) is known to increase the risk of tuberculo&s (TB); but the magnitude of this increased risk is less certain and is likely to depend on degree of immunosuppression. Relative risks measured in different studies have varied from less than 2 to more than 20 (GLYNN, 1998). Cohort studies to date have been conducted in selected ~ouulations: antenatal clinic attenders, intravenous drug&&s, and commercial sex workers (SELWYN e6 al., 1989, 1992; BRAUN er ad., 199 I; ALLEN et aZ., 1992; LEROY et at., 1995; GIUCS et al., 1997; KJZIZER et al., 1997). The risks and relative risks of TB in these arouus mav be different from those in the general populaiion.* Many case-control studies of HIV and TB have also been conducted, but some have been marred by the use of unsatisfactory control groups or lack of adjustment for confounding (GLYNN et al., 1997). TB risks vary with age and sex, but less is known about other risk factors in developing countries. In industrialized countries, TB is strongly associated with poverty (DARBYSHIRE, 1995), but the association is less clear in Africa (SCHO~AN hi aZ., 1991) and is likely to change with HIV infection. We have assessed the factors influencing the development of TB in a general population cohort with known HIV status in a rural area of northern Malawi. Methods This investigation was conducted as part of the Karonga Prevention Study, a continuing study of mycobacterial disease and HIV in Karonga District. Malawi. The methods of the fuii study hav: been de&bed in detail elsewhere (P~~~HA~s et aZ., 1993; KARONGA I%EV?ZNTION TRIAL GROUP, 1996). House-to-house surveys of the total population (now estimated at 180 000) were conducted in 1981-84 and again in 1986-89 to assess the mevalence and incidence of leprosy and tuberculosis: The second survey formed the recruitment chase of a vaccine trial of repeat BCG (bacillus Calmetie-Guerin) and BCG plus killed Mycobacterium Zeprae. In the first survey, information was collected on education, occupation and housing conditions. In the second survey the schooling level of those Address for correspondence: Dr Judith Glynn, Infectious Disease EpidemioIo~ Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WClE 7HT, UK; e-mail [email protected]
currently in school was recorded, but no other socioeconomic information. Since the end of the second survey, Karonga Prevention Study staff have been stationed in health centres and the hospital to assess attenders for TB and leprosy, and additional house-tohouse surveys were conducted in selected areas. Patients with a cough of more than 3 weeks’ duration were encouraged to provide 3 sputum specimens, and patients with suspected extrapulmonary TB were referred for biopsv or further diagnostic procedures as appropriate. $putum smears were examined and cultures set up in the oroiect laboratorv in Chilumba, Malawi. Cultures whiih were macroscopically consistent with M. tuberculosis were sent to the UK for species identification and drug sensitivity testing. All diagnosed TB patients started treatment in the hospital, following Malawi government guidelines. As nart of an investigation of serological tests for leprosy, filter paper blood specimens were collected between 1987 and 1989 from all individuals in 2 rural areas at opposite ends of the district. The leprosy tests did not proveuseful and the specimens have continued to be stored at -20°C. With permission from the Health Sciences Research Committee of the Malawi Ministry of Health. these soecimens were tested for HIV. A particle agglutinati& assay for anti-HIV-l was used as an initial screening test, followed byMconfirmation of positive samples using an Organon enzyme-linked immunosorbent assay and a SerodiaTM particle agglutination assay. Individuals with known HIV status, on the basis of the filter paper tests from the late 198Os, formed the cohort for this study. Those aged less than 15 years at the time of testing and those who had had previous TB were excluded. r~o~a~on on housing, sihooling and occupation from the earlier survey was used where available. Housing conditions were defined only for those whose households owned the dwellings. Past schooling was recorded in the first survey, and current schooling in both surveys. For those who were still in education at the time of questioning, the occupation of the head of the household was used as a measure of the socioeconomic status of the household. Population density at the time of the second survey was calculated for each square kilometre of the district. -All individuals included in the study had been vaccinated at least once with BCG, and we have found no evidence that BCG is associated with TB in this population (KARONGA PREVENTION TRIAL GROUP, 1996).
TBANDHIVINhlALAWl
For the purposes of this study, patients were classified as having developed TB ifthey fulfilled 1 of the following criteria: sputum smear positive (excluding those with single smears with fewer than 10 acid-fast bacilli per 100 fields, and specimens from which the culture grew nontuberculous mycobacteria); M. tuberczdosis cultured from the sputum or other body fluid; histology suggestive of TB; clinical diagnosis of tuberculous lymphadenonathv. All other uatients diaanosed as having TB were excluded from these analyses, TB cases diagnosed up to 1996 were included. There was no active follow-up of the copulation so results are presented as risks and-odds ratios (not rates). Absolute risks are likelv to be low. but this should not bias the relative risks (although it-increases the sampling error). HIV status was available for the whole cohort only at the start of the neriod. Subseauent acauisition of HIV would lead to underestimation of the effect of HIV, as well as a failure to correct fully for the confounding effects of HIV when considering other factors. Analyses were therefore repeated considering only the first 3 years after HIV testing. HIV status was tested on the majority ofTB cases at the time of diagnosis of TB. The risks of TB in HIV negative individuals were also reanalysed after excluding patients found to be positive (or with unknown HIV status) at the time of diagnosis. Results Results were available for 11 24 1 individuals, of whom 182 were HIV positive at the time of testing in 1987-89. Nearly 80% of individuals were seen in the first survey and so had socioeconomic data available. Sixty-nine new cases of TB fulfilling the diagnostic criteria were identified, including 2 extrapulmonary cases (one with biopsyconfirmed lymph node TB, and the other with cultureconfirmed mberculous pleural effusion). TB was diagnosed in 3.9% (7/182) HIV uositive subjects and 056% (62/l 1059) of those who-were HIV’ negative. HIV positivity increased the risk of TB about 7-fold, and this was little affected by adjusting for age, sex or the socioeconomic variables (Table). Risks ofTB were the same in the 2 areas, and between heads and members of households after adjusting for age, sex and initial HIV status. There was no associationbetween population density and TB, and no difference in TB risk amone those who were seen in the first survey and those who w&e not seen. The risk ofTB was lower under the age of 30 years, but little different between those aged 30-44 and those aged 45 years and over. Males hada higher risk of TB than females, and this was predominantly seen in the oldest age group (x2 interaction between age and sex = 4.3, P = 0.1). Odds ratios [ORs] for TB in males compared to females, after adjusting for initial HIV status, were 0.85 (95% confidence interval [95% CI] O-29-25) in those aged under 30, 1.6 (95% CI 0.74-3.4) in those aged 30-44, and 3.6 (95% CI 1.5-8.9) in the oldest group. TB was more common in those with houses of burnt brick than in those with less well built dwellings, but there was little difference between the other types,fdwelling. The risk of TB was hither in neoule in occunations other than subsistence fan&g. ‘Y&isLwas a m&ed group of salaried workers, those employed in traditional trades, and casual workers; risks were high in all these subgroups. After adjusting for age, sex and HIV status the odds-ratios for the different occupation groups compared to the farmers were 3.7 (95% CI 1.5-9.5) for salaried workers, 2.1 (95% CI 0.50-9.2) for traditional traders and 3.1 (95% CI 1.0-9.1) for casual workers. These odds ratios were onlv sli&lv reduced bv adjusting for housing and schooling. ‘Those with no schooling had the lowest risks of TB, and narticularlv high risks were seen in the few with secondary or te&u$ education. The associations with socioeconomic variables persisted after adjustment for initial HIV status, age, sex and population density. The association with schooling was no longer
statistically significant after adjusting for housing and occupation (Table). Thirty people developed TB within 3 years, including 3 HIV positive individuals. Analysis of risk factors for TB within 3 years gave similar results to those obtained ti-om the full data set except that housing conditions no longer influenced development of TB. There was still a strong association with occupations other than farming (OR 4.0, 95% CI 1*5-10.8, after adjusting for age, sex and HIV) and with schooling. Of the 62 TB cases who were HIV negative at the time of the initial test, 8 were HIV nositive bv the time of diagnosis of TB, and 7 had no test available.-These 15 individuals were excluded, and the risks of develonine TB were assessed in the 11044 initially HIV n’egayive individuals. In this group, TB was still associated with increasing age and male sex. After adjusting for age and sex, housing conditions were still associated with the development of TB (OR O-21, 95% CI 0.05-0.94 and OR 0.50, 95% CI 0.23-1.1 for the unburnt brick or pounded mud and the wood or mud groups respectively, compared to those with burnt brick ho&s), but the effect of occupation was less strong (OR 2*4.95% CI O-95-5.9, comnarina those in occunations other than farming to subsisience Tanners) and the effect of schooling was lost. Discussion The observed risks of TB measured in this study were low. There are several reasons for this. Cases were detected passively, and so many may have been missed. Even the World Health Organization (WHO) target for case finding is only 70% and it is estimated that many places fail to achieve this. We used strict criteria for diagnosing TB for this study in order to improve specificity and the validitv of the relative risk estimates. During this period a further 48 individuals from the cohort were diaanosed as having TB bv the health services. The der&ninator for the-risk estimates was all those seen and tested during the second survey. Because there was no active follow-up we do not know what proportion of the cohort was alive and still living in the district at different times and so actually at -risk of developing and being seen with TB. The risks of TB also need to be considered in relation to the risk of infection with M. tuberculosis in the district. Several studies have measured the incidence of TB among HIV positive tuberculin positive individuals as between 5 and 10% uer annum (GLYNN. 1998). In this the annual risk of i&ection*has been esti_population _ mated as less than 1% (FINE et al., 1999). The median age of the HIV nositive individuals was 30 vears. so we can estimate that only approximately 26% of them are likely to have been infected with M. tuberculosis. Relative risks may be biased if the case detection is different in the different groups, if there is differential survival or migration, or if the risk factors are measured inaccurately. Migration may be more common in those who are HIV positive than those who are HIV negative, and high mortality rates are expected among the HIV positive persons. Both these effects would tend to reduce ihe estimate of the relative risk, but the 7-fold increased risk seen for HIV is similar to that measured in casecontrol studies in the whole district at 2 different times (PhNNlGHAUS et al., 1991; GLYNN et al., 1997), and to estimates from other populations (GLYNN, 1998). The increased risks with age and in men are also expected (RIEDER et al., 1989). Socioeconomic status was defined on the basis of data from the first survey and so will fail to take into account anv changes over time, leadina to misclassification of some ind:viduals. This~miscla&ication is likely to be similar for TB cases and other individuals. and so lead to underestimation of effects. Migration out of the district may be more common for those with higher compared to lower socioeconomic status and for those who are not farmers compared to farmers. This would also tend to
T. R. GLYi’W ETAL.
502
Table.
Risk factors for the development
of tuberculosis
in Karonga
District
Malawi,
1987-1996
Odds ratios”
Risk factor HIV status Negative Positive Sex Female Male Age group (years) 15-29 30-44 245 House construction Burnt brick Unburnt brick/pounded YU;&$m;d/temporary
mud
Schooling None l-5 years primary 6-8 years primary Second~/te~a~ Unknown Occupation Farming Other Unknown Position in household Head Member Other
Adjusted for age, sex and HIV
Adjusted for age, sex, HIV, housing, schooling, and occunation
No. of subiects
No. with tuberculosis
11059 182
62 (0.56%) 7 (3.9%)
:.4 (3.3-16-6)
i-3 (2.7- 14.5)
6037 5204
27 (0.45%) 42 (0.81%)
1 l-9 (l-1-3.0)
i-5 (O-90-2.6)
4932 3059 3250
14 (0.28%) 27 (0.88%) 28 (0.86%)
:*I (l-6-6.0) 3.1 (1-6-6.0)
:.2 (1.6-6.5) 3.4 (1.6-6.8)
1506 1423 5021 3291
17 (1.1%) 4 (0728%) 27 (054%) 21 (0.64%)
2374 3616 2563 208 2480
9 (0.38%) 26 (0.72%) 10 (0.39%) 4 (1.9%) 20 (0.81%)
7757 679 2805
:; yyy I9 (0.68%)
:.2 (1.6-6-4) 1-6 (0.88-2.8)
i-9 (1.4-6~0) 1.9 (0.75-4-6)
3425 7597 219
36 (1.1%) 3; ?)43%)
0187 (0*41-1.8)
of75 (O-39-1.5)
. 0*\5 (0*085--0.76) 0.44 (0~24-0.8 1) 0.53 (0.28- 1.0)
:.o l-3 4.0 1.9
(0.93-4.4) (0-49-3-4) (1.1-138) (O-81-4.2)
0.;7 (0.09 l-0.83) 0.50 (0.27-0.96) 0.40 (0.13-1.2)
i-0 1.0 1.7 1.6
(O-89-4.3) (O-38-2*8) (0.42-6.6j (0-49-52)
Area North south Population density (per km2) Cl00 100-199 200-299 2300
6151 5090
37 (0.60%) 32 (0.63%)
2678 2693 3017 2853
15 19 16 19
(056%) (0.71%) (0.53%) (0.67%)
i-2 (0*60-2.4) 0.92 (0.45-1.9) 1.2 (0.60-2.4)
l-2 (0.58-2.3) 0.92 (0.45- 1.9) 1.0 (0.52-2-l)
Seen in first survey Individual Household/family No one
8918 938 1385
53 (0.59%) 7 (0.75%) 9 (0.65%)
:.6 (0.70-3.5) 0.97 (O-46-2.0)
Note b -
member
i-0 (O-64- 1.7)
0194 (O-57-1.6)
“95% confidence interval in parentheses. bSocioeconomic factors known only for those seen in first survey.
reduce the relative risks, though this may be offset by any excess mortality among those with lower socioeconomic status compared to those with higher status. Population density was defined per square kilometre and may not reflect the density in the immediate locality of the individual. The HIV status was defined for all individuals only at the start of the study, HIV has increased rapidly in this district over the period of the study, so some individuals classified as negative will have become positive @LYNN et al., 1997). This leads to underestimation of the relative risk of TB by HIV status and to a failure to control completely for the confounding effect of HIV when considering other risk factors. Two approaches were used to investigate this further: first, considering only the first 3 years of follow-up (so that the role of incident HIV
cases leading to TB was minimized); and second, considering HIV negative individuals only and excluding HIV positive (or HIV status unknown) TB cases. In these analyses better housing and occupations other than farming were still associated with TB, after adjusting for other factors. The effect of schooling was lost in the analysis restricted to HIV negative individuals, and in all analyses after adjusting for occupation and housing. These results suggest that, although the associations between TB and measures of higher socioeconomic status could be partly due to residual confounding due to HIV incidence, this was not the full explanation. Relations between poverty and TB caot be assnmed to be the same in At&a as they are in ‘western’ countries. Indeed, even within 1 country they may be different in different ethnic groups (HAWKER et af., 1999). If urban
503
overcrowding is one of the key elements of poverty which leads to increased TB risks in ‘western’ nations, it would not be surprising if associations were different in a rural African community. ‘Better’ housing might increase the risk of transmission of M. tuberculosis due to poor ventilation, and the outdoor life of farmers could be protective. In England and Wales the standardized mortality ratio for deaths from respiratory tuberculosis among agricultural labourers is consistently low (data from decennial supplements from the UK Office of Population Censuses and Surveys), in line with lower rates of tuberculosis in rural than in urban areas (HOR%‘I%?2et al., 1969). This contrasts with the USA, where very high rates of tuberculosis have been found among migrant farm labourers, attributable to substandard living conditions and poor health care (CIESIELSIU et al.. 1991. 1994: MCKBNNA etai.. 1996). TB could be associated with higher” socioeconomic status if this increased the risk of acquiring infection (for example, by increased travel) or due to an increased likelihood of diagnosis of the disease. It would be interesting to investigate the socioeconomic risk factors associated with TB. after adiustine for HIV. in different settings. To the extent to which these differences reflect different chances of being diagnosed with TB, they may give clues to the relative adequacy of case finding in different sections of the community. Acknowledgements Umil 1996 the Karonga Prevention Study was funded primarily by LEPRA (the British Leprosy Relief Association) and ILEI? (the International Federation of Anti-Leprosy Organizations) with contributions from the WHOAJNDPWorld Bank Special Programme for Research and Training in Tropical Diseases. Since 1996 the WellcomeTtust has been the principal funder. J. R. G. is supported by the British Department for International Development. We thank the Govemment of the Republic of Malawi for their interest in and support of the Project and the Malawi Ministry of Health for permission to publish the paper. We thank Dr Laura Rod&es for helpful discussions and Dr Dave Leon for comments on an earlier draft. References Allen, S., Batungwanayo, J., Kerlikowske, K., Lifson, A. R., Wolf, W., Granich, R., Taelman, H., van de Perre, P., Serufilira, A., Bogaerts, J., Slutkin, G. & Hopewell, P. C. (1992). Two-year incidence of tuberculosis in cohorts of HIV-infected and uninfected urban Rwandan women. Anrerican Review ofRespimtory Disease, l&1439-1444. Braun, M. M., Badi, N., Ryder, R. W., Baende, E., Mukadi, Y., Nsuami, M., Matela, B., Willame, J.-C., Kaboto, M. & Heyward, W. (199 1). A retrospective cohort study of the risk of TB among women of childbearing age with HIV infection in Zaire. American Review of Respiratoy LXrease, 143, 501-504. Ciesielski, S. D., Seed, J. R., Esposito, D. H. & Hunter, N. (1991). The epidemiology of tuberculosis among North Carolina migrant farm workers. Journal of the American Medical Association, 265, 1715- 1719. Ciesielski, S., Esposito, D., Protiva, J. & Piehl, M. (1994). The incidence of tuberculosis among North Carolina migrant farmworkers, 199 1. American 3otmd of Public Health, 84, 1836-1838. Darbyshire, J. H. (1995). Tuberculosis: old reasons for a new increase? British M~~3ou~, 310,954-955. Fine, P. E. M., Bruce, J., Pijnnighaus, J. M., Nkbosa, P., Harawa: A. & Vynnycky, E. (1999). Tuberculin sensitivity: conversions and reversions in a rural African population. International Journal of Tuberculosis and Lung Disease, 3, 962-975.
Gilks, C. F., Godfrey-Faussett, P., Batchelor, B. I. F., Ojoo, J. C., Ojoo, S. J., Brindle, R. J., Paul, J., Kimari, J., Bruce, M. C., Bwayo, J., Plummer, F. A. & Warrell, D. A. (1997). Recent transmission of tuberculosis in a cohort of HIV-l-infected female sex workers in Nairobi, Kenva. AIDS. 11.91 l-918. Glynn, J. R. (1998). Resurgenchof tuberculosis and the impact of HIV infection. British Medical Bulletin, 54,579-593. Glynn, J. R:, Warndofi, D. K., Fine, P. E. M., Msiska, G. K., Munthah, M. M. & Pdtmighaus, J. M. (1997). The impact of HIV on morbidity and mortality from tuberculosis in subSaharan Africa: a study in rural Malawi and review of the literature. Health Transirion Review, 7, supplement 2,75-87. Hawker, J. I., Bakhshi, S. S., Ah, S. & Farrington, C. P. (1999). Ecological analysis of ethnic diITerences in relation between tuberculosis and poverty. &it&k Medical 30~~4 319, 1031-1034. Horwitz, O., Wilbeck, E. & Erickson? P. A. (1969). Epidemiological basis of tuberculosis eradication. 10. Longitudinal studies on the risk of tuberculosis in the general population of a low-prevalence area. Bulletin of the World Heakk Organization, 41,95-113. Karonga Prevention Trial Group (1996). Randomised controlled trial of single BCG, repeated BCG, or combined BCG and killed Mycobactetium Zeprae vaccine for prevention of leprosy and tuberculosis in Malawi. Lancet, 348, 17-24. Keixer, S. T., Langendam, M. W., van Deutekom, H., Coutinho, R. A. & van Ameijden, E. J. C. (1997). Tuberculosis and HIV-infection in a cohort of drug users, Amsterdam 1986-1995. Zntenaatinaf Journal of Tuberculosis and Lung Disease, 1, supplement, SI57-S158. Leroy,V., Msellati, P., Lepage, P., Batungwanayo, J., Hitimana, D. G., Taelman, H., Bogaerts, J., Boineau, F., van de Perre, P., Sunonon, A., Safamon, R. & Dabis, F. (1995). Four years of natural history of HIV-l infection in African women: a prospective cohort study in Kigali (Rwanda), 1988-1993. 3oumal of Acquired Immune Delicinzcv Syndromes and Human ~bw&qy,9 415-421. - McKenna, M. T., Hutton, M., Cauthen. G. & Onorato. I. M. (1996):The association‘between occupation and tuberculosis. A population-based survey. American Journal of Respiratoy and Critical Care Medicine, 154,587-593. Ponnighaus, J. M., Mwanjasi, L. J., Fine, P. E. M., Shaw, M. A., Turner, A. C., Oxborrow, S. M., Lucas, S. B., Jenkins, P. A., Steme, J. A. C. & Bliss, L. (1991). Is HIV infection a risk factor for leprosy? International Journal of L.eprosy, 59, 221-228. Pijtmighaus, J. M., Fine, P. E. M., Bliss, L., Gruer, P. J. K., Kapira Mwamondwe, B., Msosa, E., Rees, R. J. W., Clayton, D., Pike, M. C., Steme, J. A. C. & Oxborrow, S. M. (1993). The Karonga Prevention Trial: a leprosy and tuberculosis vaccine trial in Northern Malawi-L Methods of the vaccination phase. Leprosy ReGw, G&338-356. Rieder, H. L., Cauthen, G. M., Comstock, G. W. & Snider, D. E., jr (1989). Epidemiology of tuberculosis in the United States. EnidemiologicReviews. 11.79-98. Schoeman,]. H., Westaway, MI S. & Neethling, A. (199 1). The relationship between socioeconomic factors and pulmonary tuberculosis. International Journal of Epidemiology, 20, 435-440. Selwyn, P. A., Hartel, D., Lewis, V. A., Schoenbaum, E. E., Vermund, S. H., Klein, R. S., Walker, A. T. & Friedland, G. H. (1989). A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeficiency virus infection. New EnglandJournal of Medicine, 320, 545-550. Selwyn, P. A., Sckell, B., Alcabes, P., Friedland, G. H., Klein, R. S. & Schoenbaum, E. E. (1992). High risk of active tuberculosis in HIV-infected drug users with cutaneous ~~~930um~ of the American Medical Association, 268, Received 29 March 2000
2000; accepted for publication
I9 April