In many sub-Saharan African countries infection with HIV is a common cause of morbidity and mortality. In Zambia, figures show that in 1994, 25–32% of the child-bearing women in urban areas were infected with HIV . In this region, latent tuberculosis (TB) infection is common. Sixty-two per cent of young HIV-negative adults attending a sexually transmitted diseases (STD) clinic in Lusaka in 1990 were tuberculin skin test (TST) positive . HIV is the greatest risk factor for the progression from latent to active TB with an annual risk of 5–8% and the consequent association between the epidemic of HIV and TB is now widely recognized .
TB is an HIV-associated infection that is both curable and preventable . There is also evidence that TB may lead to a more rapid progression of HIV disease as a result of immune activation [5,6]. Preventive therapy for TB, the treatment of a latent infection to prevent reactivation, might therefore be appropriate in countries such as Zambia, which have high rates of both HIV and TB infection.
In the pre-HIV era, preventive therapy has been shown to reduce the incidence of active TB by more than 80% . In the HIV era in industrialized countries, preventive therapy has also been shown to prevent the reactivation of latent infection in HIV-infected individuals [8,9]. The results of three randomized controlled trials of preventive therapy from developing countries have been published [10–12]. The study from Haiti  showed that 12 months of daily isoniazid preventive therapy reduced the incidence of TB and also reported a delay in progression of HIV infection to AIDS. The study from Kenya  reported a low efficacy of daily isoniazid for 6 months, and the most recent study  demonstrated shortterm protection for daily isoniazid and multidrug regimens with isoniazid and rifampicin in Ugandan adults.
We have conducted a study to assess the efficacy of two, self-administered, intermittent regimens in the prevention of TB in persons infected with HIV. The study was conducted in Lusaka, Zambia and compared the regimens of 6 months of isoniazid (H), 3 months of rifampicin/pyrazinamide (RZ) and placebo. This study raises important questions about the feasibility of using TB preventive therapy in high TB and HIV prevalence countries.
Study sites, screening, and eligibility criteria
The study was conducted at the University Teaching Hospital (UTH) in Lusaka, Zambia. Patients were referred to the clinic for possible inclusion in the study from several sites within the city: the Primary Care Clinic; the STD outpatient clinic; other private clinics within Lusaka; and a primary health clinic in one of the high-density residential areas. Individuals seen from these sites had not been tested for HIV and were referred to the clinic for HIV pre- and post-test counselling and testing. Additional sites included an HIV voluntary counselling and testing centre and the blood bank clinic within UTH that provides medical care for blood donors. Individuals referred from these two sites were already known to be HIV positive and were aware of their HIV status.
Baseline screening of potential study subjects consisted of: HIV serology using two enzyme-linked immunosorbent assays (ELISA) based on different antigens (Organon Technika, Brussels, Belgium; and Behring Enzygnost Anti-HIV-1/HIV-2, Marburg, Germany), complete blood count and differential leucocyte count, liver enzyme assay (serum creatinine and aspartate transaminase measurements), chest radiograph and a TST performed with 0.1 ml intracutaneous tuberculin purified protein derivative RT 23 SSI, 2 tuberculin units (TU)/0.1 ml (Staten Seruminstitut, Copenhagen, Denmark) and read after 48–72 h.
Patients were eligible for inclusion in the study if they were HIV positive (two positive ELISA tests), were over 15 years of age, and gave written consent. Patients were excluded if they had a previous history of treatment for TB, had abnormal liver function tests, had evidence of TB (pulmonary or extra-pulmonary), were pregnant, or were unable to attend the study clinic. All subjects with abnormal chest radiographs or respiratory symptoms or cough had three consecutive sputum samples examined for Mycobacterium tuberculosis. Any patient found to have TB by smear or cultures was referred to the TB clinic for treatment.
Enrolment and follow-up
Subjects enrolled into the study were randomly assigned to receive 3 months of rifampicin (R) (600 mg) twice a week and pyrazinamide (Z) (3500 mg), or 6 months of isoniazid (H) (900mg) twice a week, or 3 months of a twice weekly placebo that matched RZ, or 6 months of a twice weekly placebo that matched H. A computer program generated the randomization code in blocks of 30 (H 10, RZ 10, H placebo 5, RZ placebo 5) to ensure that subjects were equally distributed between treatment groups throughout recruitment. Each treatment allocation was kept in a sealed envelope and labelled with its appropriate treatment number. Subjects were assigned treatment numbers in numerical order. The randomization code was held by an independent party until the end of the study.
The study subjects and clinicians knew whether a subject was on a 3 or 6 month regimen, but not whether the subject was on preventive therapy or a placebo. The study was therefore double-blind.
Subjects were provided with drug calendars and instructed to mark on the calendar each day that a dose of the drug was taken as a reminder of when to take the drug. Follow-up visits to the clinic, which included completion of a questionnaire and a physical examination, were conducted monthly for the first 6 months and thereafter at 3 monthly intervals for 30 months. Repeat HIV testing was performed on all subjects at the first month's review, liver function at monthly intervals during the treatment period and complete blood count at 6 monthly intervals throughout the study period. CD4 counts were measured on a sample of study subjects and will be reported separately (provisional data analysis indicates a strong correlation between lymphocyte count and CD4 count). Subjects who became pregnant or who developed adverse side-effects were taken off preventive therapy.
Attempts were made to follow all patients until the end of the study in April 1996. Several visits were made to patients not responding to the initial reminders. Subjects not attending the review clinic were contacted after a week by either delivering or posting a letter of reminder to their residential or business address.
Compliance with preventive therapy
Compliance was assessed by pill collection and subject reports. Subjects were defined as probably compliant if they collected all their tablets, reported taking more than 80%, and returned within one month of taking the last dose. Subjects receiving all the tablets, who reported taking more than 80% of them, but who returned more than one month after taking the last dose, were defined as possibly compliant. Those who received, or reported taking, less than 80% of the tablets were defined as non-compliant.
The study outcomes were TB, death and adverse drug reactions. During the follow-up period, any subject presenting with respiratory symptoms had three sputum specimens examined for M. tuberculosis by direct microscopy and culture. Sputum samples were examined using fluorescence microscopy and culture on a Lowenstein-Jensen slope incubated at 37°C for 8 weeks. Positive cultures were examined for drug sensitivity using the resistance ratio method .
Any patient developing respiratory symptoms suggestive of TB had a chest X-ray, was asked to submit three sputum specimens and was treated with a broad spectrum antibiotic for 7 days. If there was no improvement, they were given a different antibiotic. If there was still no improvement they were started on TB treatment and the response monitored by clinical and chest X-ray findings. Some patients were too sick to complete this protocol and were started on TB treatment on the basis of radiological and clinical grounds. Cases diagnosed as TB during the course of the study were therefore classified as follows:
- Tuberculosis-subdivided as:
- Confirmed tuberculosis: if there was a positive smear or culture for M. tuberculosis or positive histopathology.
- Presumed tuberculosis: a) Pulmonary infiltrates and clinical symptoms, completing the clinical protocol described above and responding to TB treatment within 2 months; or b) Pleural or pericardial effusion with a documented response to TB treatment within 2 months.
- Probable tuberculosis: Patients with respiratory symptoms and radiological features suggestive of TB, with no positive sputum results but who were commenced on TB treatment before antibiotics were given because of the severity of the patient's clinical condition.
The cause and date of death was determined by review of hospital notes or interview of relatives. Information on deaths occurring outside the hospital was ascertained from relatives, when the subject's residential address was visited.
During the treatment phase, subjects were screened for adverse events. Subjects were removed from the study if they developed biochemical hepatitis, defined as a serum aspartate transaminase (AST) greater than twice normal (normal AST 5–45 IU).
Sample size and data analysis
It was assumed that: 15–25% of subjects in the placebo group would develop TB within the 3 year follow-up period; the efficacy of H or RZ would be approximately 50%; 70% of subjects would have a study outcome or complete follow-up; and 30% of subjects would have no follow-up. For the study to have at least an 80% power of obtaining a significant rate ratio (RR) (P < 0.05), at least 350 subjects needed to be recruited in each group. Three hundred and sixty subjects were recruited into each treatment group. The placebo group was divided into two groups for the drug administration (180 subjects received RZ placebo and 180 received H placebo), but was combined for the analysis, after checking that the two placebo groups did not differ with respect to the study outcomes.
Data were double-entered in dBase III Plus version 1.1 (Ashton-Tate 1986) and analysed in EpiInfo version 6 (Centers for Disease Control, Atlanta, GA 30333, USA) and Stata version 5 (Stata Corporation, College Station, TX 77840, USA). Before breaking the code, it was agreed that: definite/probable TB and death would be the main outcomes; the main analysis would be based on intention to treat; and RRs would only be adjusted for any variables that resulted in at least a modest change in RR. Follow-up times were calculated separately for the analysis of TB and death. Most analysis was conducted separately for RZ versus placebo, H versus placebo, and H or RZ versus placebo. RRs and 95% confidence intervals (CI) were estimated using Cox proportional hazards models, and significance was assessed using the likelihood ratio test. RRs were also compared in different strata of variables that might modify the effect of preventive therapy: adherence to treatment, TST, crowding, total lymphocyte count, and haemoglobin level. These variables were chosen a priori on the basis of the results of other studies. Progression to TB or death in the three treatment groups was compared using Kaplan-Meier curves. All analysis of TB is for Ai) confirmed and Aii) presumed unless stated otherwise.
The protocol received ethical approval from the Ethics Committee of the University of Zambia and the Ministry of Health, and the London School of Hygiene and Tropical Medicine. A data safety and monitoring board was established to conduct interim analyses and to advise on the continuation of the study.
Screening and enrolment
Between August 1992 and July 1994, 3278 patients were seen in the study clinic and offered HIV testing, of whom 2063 tested positive. Of these, 1080 (52%) agreed to participate in the study. The main reason for non-enrolment among the remaining 983 patients was that they did not return after the initial interview (81%), although a small number had possible TB (5%) or a past history of TB (3%). After enrolment, 27 individuals were excluded from the study because they did not satisfy the inclusion criteria: 22 individuals (placebo 8, H 7, RZ 7) were found to be HIV negative on subsequent testing; three individuals (placebo 1, H 1, RZ 1) were duplicates (they returned to the clinic a long time after initial recruitment, did not reveal that they had already attended and were recruited for a second time by mistake; these individuals were censored at second recruitment); 1 subject (placebo) revealed a previous history of treatment for TB after recruitment; and 1 subject (RZ) was discovered shortly after recruitment to have a TB-positive culture on a sample given before recruitment. Therefore, 1053 subjects remained in the study of whom 352 received placebo, 350 received H, and 351 received RZ.
Characteristics of subjects at enrolment
The demographic, clinical and laboratory characteristics of the three treatment groups are shown in Table 1. The three groups were similar with respect to all characteristics, except that subjects in the RZ group had a slightly lower lymphocyte count and a lower haemoglobin level, although neither of these differences were statistically significant.
Compliance with treatment, adverse drug reactions
A total of 777 subjects (74%) were defined as compliant, of whom 718 (92%) were defined as probably compliant and 59 (8%) as possibly compliant (see Table 2).
Twenty-nine subjects (3%) were withdrawn from the study because of adverse drug reactions. The types of reactions reported were: four with raised liver enzymes (placebo 0, H 3, RZ 1), seven with rash (placebo 0, H 1, RZ 6); 11 with gastrointestinal symptoms (placebo 1, H 5, RZ 5); and seven with other complaints (placebo 2, H 3, RZ 2). Eleven subjects became pregnant while taking treatment and their medication was discontinued although follow-up continued (placebo 2, H 7, RZ 2).
A total of 1053 subjects were followed for a total of 1631 person-years (py) (median = 1.8 years). Of the 1053 people in the study, 248 (24%) had a study outcome (63 TB only, 152 death only, and 33 TB and death), 473 (45%) were seen within the last 6 months of the study, and 332 (32%) were not seen within the last 6 months of the study (115 did not return to the clinic after enrolment and 217 returned to the clinic but did not complete follow-up. The 332 ‘losses’ were equally distributed between the three treatment groups (placebo 106, H 114, RZ 112).
Effect of preventive therapy on tuberculosis
During the course of the study, 96 cases of TB/probable TB were diagnosed (59 TB and 37 probable). Thirty-six of the 59 TB cases were confirmed and 23 were presumed (six with infiltrates and 17 with effusions). Thirty-seven cases of probable TB were diagnosed: 12 of these patients had a documented response to TB treatment, three patients died early in treatment, and 17 were lost to follow-up or had no documented response. The remaining five patients were started on treatment by clinicians other than those on the study team.
The incidence rates of TB were 4.94, 2.74, and 3.16 per 100 py, in the placebo, H and RZ groups, respectively (Table 2). The crude RRs were 0.56 (95% CI: 0.30–1.05, P = 0.065), 0.65 (95% CI: 0.35–1.19, P = 0.157), and 0.60 (95% CI: 0.36–1.01, P = 0.057) for H versus placebo, RZ versus placebo, and H or RZ versus placebo, respectively. The incidence rates of TB/probable TB were 8.06, 4.94, and 4.65 per 100 py, in the placebo, H and RZ groups, respectively. The crude RRs were 0.62 (95% CI: 0.38–0.99, P = 0.045), 0.58 (95% CI: 0.35–0.95, P = 0.026), and 0.60 (95% CI: 0.40–0.89, P = 0.013) for H versus placebo, RZ versus placebo, and H or RZ versus placebo, respectively. When adjusted for variables that might be associated with risk of TB (i.e. age, crowding, BCG scar, TST result, lymphocyte count, haemoglobin level), these RRs did not change substantially. Therefore, crude RRs only have been presented for all analysis of TB.
Crude RRs were also compared in different strata of variables that might modify the effect of preventive therapy: time since enrolment (Table 3); and adherence to treatment, TST result, crowding, lymphocyte count, and haemoglobin level (Table 4). The rate of TB since enrolment shows an increase over follow-up time. The efficacy of preventive therapy falls with increasing time, so that after 18 months (about one year after completion of treatment) there is less of an effect, although these results are based on a small number of events during each time interval, and the test for interaction was not statistically significant (P = 0.25). Similar results were observed when we analysed TB/probable TB: RRs were 0.37, 0.50, and 0.81 for 0–5.9, 6–17.9, and over 18 months after enrolment, respectively, and P = 0.38 for interaction.
Table 4 shows crude RRs for TB, stratified for adherence to treatment, TST result, crowding, lymphocyte count and haemoglobin level. The efficacy of preventive therapy varied with the TST results (P = 0.05 for interaction) and the lymphocyte count (P = 0.031 for interaction). The effect of preventive therapy on the incidence of TB is most marked in those with a lymphocyte count of 2 × 109/l or higher, or a TST of 5 mm or greater. In those with a lymphocyte count of 2 or higher, the incidence rate of TB is 20/259 = 7.7 per 100 py in the placebo group and 13/510 = 2.5 per 100 py in the H and RZ groups combined (RR = 0.33, 95% CI: 0.16–0.68, P = 0.002).
Data were available on 694 tuberculin tests placed and read at enrolment, of which 524 (76%) were negative. In those with a TST of 5 or greater, the incidence rate of TB is 9/98 = 9.2 per 100 py in the placebo group and 4/162 = 2.5 per 100 py in the H and RZ groups combined (RR = 0.27, 95% CI: 0.08–0.87, P = 0.021). These results are further illustrated by the Kaplan-Meier survival curves in Figs. 1 and 2, which show progression to TB stratified by treatment group and TST result, and by treatment group and lymphocyte count, respectively. Our data shows that the efficacy of preventive therapy is greatest in those with a haemoglobin level of 10 mg/dl or greater. Non-compliance marginally reduces the efficacy of preventive therapy, although this result is based on a small number of events and person-years in the non-compliant individuals.
RRs for TB stratified by TST and lymphocyte count is most marked in those with a lymphocyte count of 2 × 109/l or over and TST of 5 mm or greater. The rate of TB is 6/56 = 10.71 per 100 py in the placebo group and 2/95 = 2.10 × 100 py in the H and RZ groups combined (RR = 0.19, 95%CI: 0.04–0.94, P = 0.026) (see Table 4).
Culture results were available for 26 subjects and sensitivity results were available for 12. One specimen showed resistance to isoniazid and this subject was in the placebo group.
Effect of preventive therapy on mortality
There were 185 deaths in the study. The mortality rates were 9.62, 10.02, and 11.76 per 100 py, in the placebo, H and RZ groups, respectively (Table 2). The crude RRs were 1.05 (95% CI: 0.73–1.50, P = 0.80), 1.24 (95% CI: 0.87–1.76, P = 0.23), and 1.14 (95% CI: 0.83–1.55, P = 0.41) for H versus placebo, RZ versus placebo, and H or RZ versus placebo, respectively. These RRs did not change when adjusted for potential confounders.
As expected, the mortality rate rose with increasing follow-up, but there was no obvious increase or decrease in efficacy over time (Table 3). A higher mortality rate was associated with non-compliance, a TST of 5 or less, a lymphocyte count of 2 or less, and a haemoglobin level of 10 or less (Table 4). The higher RRs associated with RZ in those with TSTs of 5 or greater or haemoglobin level of 10 or less should be interpreted with caution: both are based on a small number of events. None of the RRs in Table 4 changed substantially after adjustment for confounders.
The place of death was ‘in hospital’ for 74 (40%) of subjects, ‘at home’ for 66 (36%), ‘other’ such as their village for 19 (10%) and ‘not known’ for 26 (14%). Broad cause of death categories showed no difference between the treatment groups.
This study has demonstrated that 6 months of twice weekly isoniazid or 3 months of a twice weekly combination of rifampicin and pyrazinamide was effective in the prevention of TB in HIV-infected individuals in Lusaka, Zambia. The protective effect appeared to be greatest in subjects who were tuberculin reactive and in those with a higher lymphocyte count or haemoglobin level. No significant effect on mortality was detected with either regimen.
Currently, three studies have been published reporting on randomized clinical trials of preventive therapy in high TB and HIV prevalence areas. A study from Haiti  demonstrated a reduction in the incidence of TB among HIV-infected subjects receiving 12 months of daily isoniazid, which was more pronounced in those subjects who were TST positive. The second study from Kenya  showed no statistically significant effect for daily isoniazid for 6 months overall, but did indicate some preventive effect in the TST-positive subjects. The third study from Uganda  showed that 6 months of self-administered isoniazid taken daily, reduced the risk of TB by 67% in subjects with positive tuberculin skin tests (TST). In addition, daily regimens of 3 months of isoniazid and rifampicin, and isoniazid, rifampicin and pyrazinamide substantially reduced the risk of TB, but the reduction did not reach statistical significance. The only other information on studies on randomized clinical trials of TB preventive therapy in HIV-infected adults is a conference report from a study in Zambia , which showed a reduction in the incidence of TB in persons taking 12 months of daily isoniazid therapy.
This study is an important contribution to the development of recommendations for TB preventive therapy in high TB and HIV prevalence countries, and raises important questions about the feasibility of using preventive therapy in these areas. The study was presented in February 1998 at a meeting co-sponsored by the WHO and UNAIDS to develop a ‘policy statement on preventive therapy against tuberculosis in people living with HIV’. First, it shows that an intermittent self-administered twice weekly regimen of 6 months of isoniazid is effective. Intermittent therapy has the potential advantage of lowering toxicity, adverse reactions and the cost. Second, it demonstrates that persons with more advanced immunosuppression are less likely to benefit from preventive therapy, and suggests a total lymphocyte count as a potential screening opportunity for identifying HIV-infected persons who would benefit most from TB preventive therapy [15,16]. Third, it supports the findings of the Ugandan study that multidrug regimens containing rifampicin are also effective, and finally, it raises the possibility that lifelong preventive therapy or re-prophylaxis will be needed in high prevalence areas where there is frequent exposure to TB.
In high TB prevalence areas HIV-related TB arises by both reactivation and reinfection. In HIV infection, preventive therapy for TB works either by reducing the latent bacillary load to remove the possibility of reactivation of infection as immunity decreases in a person infected with M. tuberculosis, or it may prevent the establishment of latent infection. An important question is the durability of preventive therapy in areas where continued exposure to infection occurs. Before the HIV epidemic, in low prevalence countries where re-exposure to infection is rare, the protective effect of isoniazid preventive therapy lasts for up to 19 years or may be lifelong . In this study, the analysis of the occurrence of TB cases with respect to time since enrolment suggests a decline in effect after the first year of the study so that by 18 months the rate of TB in the treated groups was similar to that in the placebo group (Fig. 1). These data need to be interpreted with caution because the numbers are small and the results are not significant. For both groups it would appear that preventive therapy is effective during treatment and shortly thereafter, suggesting continued exposure to infection. The limited duration of the protective effect of preventive therapy in high TB prevalence areas such as Lusaka or Nairobi  raises the question of the possibility for lifelong preventive therapy or the need for re-prophylaxis.
The protective effect of the preventive therapy was greatest among subjects with a positive TST. This was the same finding as in the studies from Haiti, Kenya and Uganda [10–12], suggesting that preventive therapy is preventing the reactivation of latent infection. In that study preventive therapy was offered to all HIV-positive subjects regardless of their tuberculin reaction, in recognition that in the HIV-positive subject anergy to tuberculin may not necessarily indicate lack of infection. An earlier limited tuberculin survey had shown that there was a higher rate of anergy among the HIV-positive than HIV-negative individuals, indicating that tuberculin testing in this population would play a limited role in assessing individuals with a latent infection [2,16]. However, from our findings it would appear that the tuberculin reaction may be useful in identifying those subjects who would benefit the most from preventive therapy because they are both infected with M. tuberculosis and not anergic. The Ugandan preventive therapy study  indicates that persons who are anergic are less likely to benefit from preventive therapy.
By including all persons with HIV infection, irrespective of their TST status, the study has shown that subjects with a total lymphocyte count of 2 × 109/l or greater (at enrolment) or a haemoglobin level of 10 g/dl or greater, i.e. persons with less advanced immunosuppression, were more likely to benefit from preventive therapy than individuals with low total lymphocyte counts or a low haemoglobin level. An alternative to using the TST to determine who would benefit from preventive therapy in this population would be the lymphocyte count or haemoglobin level as indicators of the degree of immunosuppression. Such a strategy would exclude fewer potential patients from the benefits of preventive therapy and facilities for measuring lymphocyte counts and haemogloblin levels are more widely available in Zambia than reagents for TSTs.
Two weaknesses of the study are the diagnosis of TB and the number of subjects lost to follow-up. However, for the diagnosis of TB, all results are consistent for all of our definitions and the losses of subjects are similar for each treatment group therefore making it unlikely to bias the results. The diagnosis of TB in HIV-positive individuals in Africa is not without difficulty , and we believe that most of the subjects classified as probable TB were likely to have TB, despite failing to meet our criteria for a definite diagnosis. When all TB cases are included in the analysis, the estimate of efficacy is very similar but more statistically significant. Preventive therapy significantly reduced the number of subjects put on TB treatment during the study period.
In 1997, the ethics of placebo-controlled trials in Africa were highlighted in the New England Journal of Medicine with the publication of the Ugandan preventive therapy study [12,19]. Our study was developed in the early 1990s before the publication of the Haiti data and went through the regular channels of ethics committees in the UK and Zambia. In addition, an independent monitoring committee was established to conduct interim analyses to determine if the trial should be stopped. Two interim analyses were conducted between 1992 and 1996. At the time of the interim analyses the committee discussed the new scientific information available on preventive therapy and whether this altered the ethics of the trial. After the results of this trial, the study subjects in the placebo group were treated with a 6 month course of twice weekly isoniazid.
The current international recommendation is that in low TB prevalence countries all HIV-positive persons with a positive tuberculin reaction should be given TB preventive therapy . However, as already stated, new updated recommendations are currently being prepared by WHO and UNAIDS. The study has shown that preventive therapy is effective in reducing the incidence of TB in dually infected individuals in a high TB prevalence area, albeit for a limited period of time, probably as a result of re-exposure to infection. Its possible role in preventing the acquisition of new infections suggests a role for preventive therapy in people regularly working in crowded environments where exposure to infection is likely. The cost-benefit ratio of preventive therapy will be higher in those who are likely to expose others to infection, such as teachers, because a larger number of secondary cases will also be prevented . The greater effect observed in those with early HIV disease suggests that facilities such as a voluntary testing and counselling centres may be the best recruitment site for such individuals, particularly as these individuals tend to be asymptomatic and so are unlikely to have active TB.
In conclusion, TB preventive therapy with 6 months of isoniazid and 3 months of rifampicin/pyrazinamide, self-administered twice weekly, prevents TB in persons with HIV infection. The effect is greatest in persons with early HIV disease, who have a positive TST or a lymphocyte count of 2 × 109 or greater, indicating that preventive therapy should be provided early in HIV disease. The effect of preventive therapy may, however, not be sustained because of continued exposure to infection after treatment, suggesting the possible benefit from either lifelong treatment or repeated courses of treatment.
We would like to thank all the staff on the Zambart project: Benita Halwiindi, Sister Kamshasha, Sister Muyunda, Sis J. Moomba, Violet Zulu, Mr G. Phiri, Mr K. Phiri, Nasilele Muchula and Sylvia Mulambo. Thanks also go to the other collaborating departments at the University Teaching Hospital such as the Department of Pathology and Microbiology with special reference to the Haematology, Virology and Biochemistry Departments and the Radiology Department and the Chest Disease Laboratory of the Ministry of Health. We would also like to thank Dr Rachel Baggaley, Fr Michael Kelly and the staff at the Kara Counselling and Training Trust. We are grateful for the support and encouragement we received from Dr Mario Raviglione, Dr David Cohen and Dr Paul Nunn in the Global Tuberculosis Programme, at the World Health Organisation and from Dr Jos Perriens in the Global Programme on AIDS.