INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Human immunodeficiency virus (HIV) has had a profound impact on tuberculosis (TB), particularly in Africa (1, 2). HIV is the most common risk factor in Africa for reactivation of latent TB infection, and is also strongly associated with rapid progression from infection to disease (3, 4). As a result of the rise in incidence of TB in countries where both TB and HIV infection are common, TB control programs are being stretched to their limits (5).

The World Health Organization (WHO) has advocated better supervision of therapy for TB to ensure that patients are cured (6). A target cure rate of 85% has been set, although it is acknowledged that the high mortality rates among patients with HIV-related TB may make this unattainable.

Several studies of cohorts of African patients with TB with and without HIV infection have been reported (7). However, these studies either had difficulty in following their full cohorts of subjects, had incomplete data on causes of mortality and treatment failure, or used less effective drug regimens.

In order to define the impact of HIV on the outcome of treatment for TB, and to better understand the causes and risk factors for treatment failure or death, we prospectively followed a cohort of Southern African gold miners presenting with pulmonary TB during 1995.

Although the community of gold miners is not typical of most of Southern Africa, the study has considerable strengths. Medical resources are more readily available to these miners than in other program settings, and include sputum culture and drug susceptibility testing. Patients with drug-susceptible isolates of Mycobacterium tuberculosis are uniformly treated with rifampicin. Directly observed therapy (DOT) is practiced, and there are adequate facilities to admit patients with complications. Follow-up can be achieved both within the mine community and, for those who return to their homes, through the agency network that recruits miners and through which pensions and compensation are paid. Pulmonary TB is a compensable disease in miners, and autopsy examination of the cardiorespiratory organs is provided by law. The present study therefore provides the most comprehensive examination yet made of the impact of HIV on the outcome of TB.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population and Setting

The study population comprised 28,522 black men working in four gold mines in Westonaria, a district situated 50 km from Johannesburg, South Africa. The men were mainly migrant workers, living in hostels in the mine areas, who had come from rural areas of South Africa as well as from the neighboring countries of Botswana, Lesotho, Mozambique, and Swaziland, to which they returned on annual leave. Medical services are provided by the mining company without charge. Health care is provided at a 240-bed hospital and its satellite primary health care facilities situated in each of the hostels.

TB Control Program

Mine workers may self-report for medical care with any symptoms at any time, and most TB patients are found in this way. Active case finding comprises chest radiographic screening as well as contact tracing (one sputum smear and a chest radiograph) of men who sleep in the same room as persons with TB. Men clinically suspected of having TB have at least three sputum smear examinations, and a minimum of one specimen is cultured and routinely tested for drug-resistant organisms. M. tuberculosis is bacteriologically confirmed in 94% of patients who receive treatment for pulmonary TB.

Smear-positive patients are initially hospitalized, after which they report every weekday to the primary health-care facilities, where DOT is administered and recorded.

Treatment consists of 2 mo of isoniazid (INH), rifampicin, pyrazinamide, and ethambutol, followed by 4 mo of INH and rifampicin. All patients, regardless of their history of previous TB treatment, receive the same regimen except for those with multidrug resistance, who are treated according to their drug sensitivity pattern, which is available within 6 wk after testing for TB.

Patients are followed at a TB clinic at the hospital, with an average of three visits during the 6-mo treatment period. At each visit the patient is assessed clinically by a physician, a sputum smear is examined, and a chest radiograph is taken. At the end of treatment, sputum is cultured and, if positive, the organism is tested for drug susceptibility. Patients are kept in employment, returning to underground work when fit. Leave is discouraged while patients are taking treatment, but those who do go on annual leave (56 d) are given drugs for self administration and their TB record cards. Patients whose condition is terminal may choose to return home for their last months of life. They transfer out of the program and are provided with anti-TB drugs, a letter of referral to their nearest health center, and a pension.

Definitions

New TB was defined as TB in a patient who had never before received treatment for TB, and recurrent TB was defined as TB in a patient who had been previously treated and assessed as cured. This included patients who had documented sputum negativity at the end of treatment, and patients completing treatment whose sputa at the end of treatment were not examined and who were assessed as being cured on the basis of clinical and radiographic findings.

Study Cohort

All patients with new or recurrent pulmonary TB confirmed by at least one positive sputum culture between January 1, 1995 and December 31, 1995 were prospectively enrolled in the cohort. Patients with extrapulmonary TB were excluded. Prevalent cases, for which treatment had been started before 1995 were excluded, even if their sputum cultures were still positive in 1995. At enrollment, each patient was interviewed, and the patient's medical records were reviewed according to a standard form. For the few patients who were unavailable for interview, the mining company's human resources data bases were accessed for information about home area, educational level, occupation, and years of employment. All previous episodes of TB were recorded. CD4⁺ lymphocyte percentages and HIV status were determined for consenting patients.

Treatment outcome was assessed 6 mo after treatment was begun. To assess compliance, patients were reinterviewed, and the clinic records and DOT register were reviewed. The mortality status of patients who had transferred out of the program was determined by home visits, organized through the mines' employment agency network, both in rural areas of South Africa and in neighboring states.

The study was approved by the Committee for Research on Human Subjects of the University of the Witwatersrand, and informed written consent was obtained from the subjects.

Chest Radiography

Patients had posteroanterior (PA) chest radiographs at diagnosis. Radiographs were jointly assessed by two readers (R.G.T. and P.S.), who were blinded to the patients' HIV status and clinical data. Radiographs were considered to be typical of TB if there were predominantly upper-lung-field infiltrates and/or fibrocavitation, or miliary disease. A radiograph was considered atypical if changes, including cavitation, were present predominantly in the lower lung fields; if there was isolated hilar and/or mediastinal lymphadenopathy; if there was lobar consolidation; or if there were no changes.

For the purpose of assessing radiographic changes, each lung was classified into upper, middle, and lower zones. Severity was classified as minimal if there was slight to moderate involvement in one or both lungs but the total extent of lung affected did not exceed one zone; moderate if multiple zones were affected but with intervening areas of normal lung; and extensive if multiple zones were affected with minimal normal lung present.

Laboratory Methods

All sputum specimens were inoculated into Bactec 12B vials (Becton Dickinson, Sparks, MD) after appropriate decontamination through standard N-acetylcysteine-sodium hydroxide digestion and concentration (13). Once a positive growth index was recorded, aliquots were transferred to both a Bactec 7H11 vial for susceptibility testing and to a Lowenstein-Jensen slope for morphologic assessment and culture confirmation. Isolates of acid-fast bacilli were identified by DNA- RNA hybridization (Gen-Probe AccuProbe, San Diego, CA) and standard biochemical testing (13). Susceptibility testing was done with the BACTEC system (14). INH was tested at 0.1 mg/L, rifampicin at 2 µg/ml, ethambutol at 7.5 mg/L, and streptomycin at 6.0 mg/L. Any difficulties encountered with this method were resolved by retesting the isolate by the conventional proportional-count method, using Lowenstein-Jensen medium (15). Susceptibility testing was undertaken at Lancet Laboratories in Johannesburg.

T-cell analysis was done by flow cytometry on a FACSort instrument (Becton Dickinson, San Jose, CA) with SimulSet monoclonal antibodies. We used relative levels of CD4⁺ and CD8⁺ lymphocytes, expressed as percentages of total lymphocytes, rather than using absolute counts, because the former method is associated with less variation (16). A CD4⁺ percentage of > 28 corresponds to an absolute CD4⁺ cell count of  500/µl, a percentage of 14 to 28 corresponds to a count of 200 to 499 /µl, and a percentage of less than 14 corresponds to a count of less than 200/µl (17). T cells were analyzed at the National Institute for Virology in Johannesburg.

Treatment Outcomes

The following definitions were used in determining the outcome of patients in the study:

Treatment interrupted: departure from the mine at any time after starting treatment, without notifying the TB control program staff. These patients could therefore not be referred to another medical facility, and were lost to follow-up.

Transfer out: transfer to another medical facility, with mortality status undetermined at 6 mo after treatment began.

Death: death within 6 mo of entering the study, regardless of cause.

Failure: smear and/or culture positivity at 6 mo after treatment began.

Cure: completion of treatment with at least two negative sputum smears during the course of treatment, and either a negative culture or a negative smear or inability to produce sputum at the end of treatment, but in the presence of clinical wellness.

Single drug resistance: resistance to one first line anti-TB drug (INH, rifampicin, streptomycin or ethambutol).

Multidrug resistance: resistance to at least INH and rifampicin.

Compliance was assessed in terms of attendance at scheduled TB clinic visits. Attendance was defined as irregular if one or more appointments were missed. Patients who went on leave could not be supervised owing to the paucity of health-care services in rural areas.

Statistical Analysis

We used the SAS statistical package (SAS Institute, Cary, NC) to analyze the study data (18). We used the chi-square test in two-way tables to test the hypothesis of no association between health outcome and individual risk factors. We used the Mantel-Haenszel chi-square statistic to test the alternative hypothesis of a linear association between health outcome and increasing category of a risk factor. We calculated relative risk (RR) as a measure of association between health outcome and individual risk factors; RR = p1/p2, where pi = nii/ni and I = 1,2.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Three hundred seventy-six sputum culture-positive patients were enrolled in the cohort, giving an incidence of pulmonary TB of 1,318 per 100,000. Mixed M. tuberculosis and mycobacteria other than M. tuberculosis were present in seven (1.9%) patients, four of whom were HIV infected and three of whom were not.

Demographic characteristics of the cohort at presentation are shown according to HIV status in Table 1. Overall, 55.5% of the cohort were migrants from neighboring countries: 56.9% from Lesotho, 22.5% from Botswana, 11.5% from Mozambique, and 8.6% from Swaziland; one patient was from Malawi.

Clinical characteristics related to TB infection according to HIV status are shown in Table 2. There was no difference in the distribution of recurrent TB in HIV-positive and HIV-negative patients. Of the subjects with recurrent TB, 79.3% had had one, 17.1% had had two, and 3.6% had had three previous episodes.

Single-drug and multidrug resistance rates were not related to HIV status (Table 2). However, drug resistance was significantly associated with a history of recurrent TB (data not shown): of previously untreated patients, 90.8% were fully sensitive, 8.9% had single-drug resistance (INH 8.1%, streptomycin 0.8%), and 0.3% were multidrug resistant, as compared with 75.3% of previously treated patients who were fully sensitive, 18.2% who had single-drug resistance (INH 16.9%, rifampicin 1.3%), and 6.5% who were multidrug resistant (single drug resistance, p = 0.011; multidrug resistance, p = 0.001). All patients with recurrent TB had previously been treated at this group of mines with INH, rifampicin, pyrazinamide, and ethambutol. DOT was practiced, but compliance had not been recorded.

There were no significant differences between the new and recurrent TB groups in terms of delay, which was defined as more than 24 h between the laboratory diagnosis of TB and the start of treatment.

Table 3 shows the clinical features of the 182 HIV-infected patients for whom CD4⁺ cell counts were available, according to CD4⁺ cell percentage. A CD4⁺ cell percentage of less than 14%, corresponding to an absolute CD4⁺ cell count of less than 200/ul, was found in 48.9% of the patients. The level of immune suppression was not associated with recurrent TB or drug resistance.

Table 4 shows the outcome of the cohort after 6 mo of follow-up. The outcome is shown separately according to HIV status and history of recurrent TB. For the entire cohort, the treatment interrupted rate was very low, and loss to follow-up was not associated with HIV status or recurrent TB (two patients were discharged from employment, two never returned from leave, and four left the mine of their own accord).

Six patients who transferred out of the program were alive at 6 mo thereafter, but it was not known whether they were cured. Three patients had end-stage acquired immune deficiency syndrome (AIDS), two HIV-negative patients with combined TB and silicosis were compensated for occupational lung disease and returned home, and one patient requested transfer.

There were 28 deaths, 26 of which occurred in HIV-positive patients; 13 patients died in the mine hospital and 13 died in rural areas (two were on leave and 11 patients had returned home because of end-stage AIDS). One HIV-negative patient died at home from burns before commencing treatment. The HIV status of the other patient was unknown; he died at home before starting treatment. Death was not significantly associated with a history of recurrent TB.

There were 25 patients in the treatment failure category. Treatment failure was strongly associated with recurrent TB (p < 0.001, odds ratio [OR] = 4.89; 95% confidence interval [CI] 2.12 to 11.30) but not with HIV infection.

A total of 309 patients were cured. Of these, 300 had a negative culture, five had a negative smear but culture was not done, and four were unable to produce sputum but were clinically well. There was a strong association between HIV status and cure (p  0.009); however, if those patients who died are excluded (and thus no longer contribute to the infectious pool), the total cure rate increases from 82.2% to 88.8%, and there is no longer an association between cure and HIV infection. Cure rates were significantly lower in those patients with recurrent TB than in those with new TB (68.3% versus 86.1%), even when patients who died were excluded (76.7% versus 92.0%). Data for cure of all patients still in the program after 6 mo of follow-up are also shown. Among patients who were smear positive on entry into the study, the overall smear positive cure rate was 81.8% (225 of 275 patients).

Risk factors for treatment failure after 6 mo of TB therapy for patients with new and recurrent TB are shown in Table 5.

For the new TB group, the factors associated with significantly increased risk were having taken leave and irregular clinic attendance. There was no association with drug resistance, all of the treatment failures having been fully sensitive.

For the recurrent TB group, significant relative risk factors were having taken leave and single-drug resistance. In order to determine whether leave and single-drug resistance were independent risk factors for treatment failure, we first stratified the analysis by single-drug resistance. Leave was an important risk factor in patients without drug resistance (RR = 6.17, 95% CI: 0.89 to 42.56), but not in those with single-drug resistance (RR = 1.00, 95% CI: 0.08 to 12.56). When as a second measure we stratified the analysis by leave,we found that single-drug resistance was a more important risk factor in patients who did not go on leave (RR = 18.50, 95% CI: 2.17 to 157.46) than in those who did (RR = 3.00, 95% CI: 0.29 to 31.63).

In the recurrent TB group who were HIV positive, treatment failure was associated with decreasing CD4⁺ cell percentage, but this did not attain statistical significance. This group also showed a nonsignificant increase in the risk of extensive radiologic change.

In neither the new nor the recurrent TB group was there any association of infection with age, education, length of employment, or admission weight (data not shown).

Of the six patients with multidrug resistant TB, three were still in the program at 6 mo: one had treatment failure and two had negative sputum cultures at the end of the 6 mo period, and were therefore classified as cured. However, because of these latter two patients' multidrug resistant disease, treatment was continued, and both patients subsequently had sputum cultures positive for drug-resistant organisms. Two patients with multidrug resistant TB who also had AIDS died, and a further patient had terminal AIDS and chose to return home.

Table 6 shows the factors associated with death in HIV-positive patients. Of the 190 HIV-positive patients enrolled in the cohort, four were lost to follow-up and 26 died. The risk of death increased with CD4⁺ lymphocyte depletion (chi-squared result for trend: p < 0.001). However, because no deaths occurred among patients with a normal CD4⁺ cell percentage (> 28%), the CIs associated with the RR for death are very wide.

In patients with a CD4⁺ lymphocyte percentage of less than 14%, the case fatality rate was 22.1%, as compared with 7.7% in those with a CD4⁺ lymphocyte count of 14% to 28%.

Delay in starting treatment was not an issue, since all patients started treatment within 4 d of the diagnosis of TB, except for one patient who died before treatment began.

Table 7 shows characteristics of the HIV-positive patients who died. Six patients died within the first 30 d after diagnosis. Autopsy examination was performed on the cardiorespiratory organs of 12 of the 13 men who died while in mining service. Only five patients died of autopsy-confirmed TB; all died within 30 d of diagnosis, and all had extensive radiographic changes, with three having CD4⁺ lymphocyte counts of more than 14%. There were four documented deaths from cryptococcal pneumonia, which occurred from 29 to 61 d after the diagnosis of TB. These four patients were more immune compromised than the others, and all had atypical chest radiographs.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This prospective cohort study of Southern African gold miners provides the most reliable data to date on the impact of HIV on the outcome of treatment of patients with pulmonary TB. The data include both CD4⁺ lymphocyte percentages and anti-TB drug susceptibility. Vital status was available for 98% and TB status for 96% of the cohort by the end of the follow-up period. Autopsy examination of the lungs was performed on nearly 50% of those who died.

However, there are differences between the mining population and the general population. The incidence of pulmonary TB is much higher in the former, at 1,318 versus 400 per 100,000 (19). This is partly attributable to silica dust exposure, which is a known strong risk factor for development of TB (20, 21), but may also reflect comprehensive case detection in the mining population, which includes radiologic surveillance and contact tracing. However, silica dust exposure is not associated with a worse treatment outcome (22). The dormitory accommodation of miners, with up to 18 men sharing a room, may be distinct from most African households. Nevertheless, urban deprivation in many cities in the region may also lead to overcrowding.

Of the study cohort, 49% were HIV positive. Apart from younger age, there were no significant differences in these patients' demographic characteristics versus those of HIV-negative patients (Table 1). Overall, the proportion of recurrences of TB was high (22%), which may have been due to the high prevalence of infection in the mining environment. Program factors may also have played a part in the high recurrence rate, particularly the lack of supervision when miners are on leave, and the use of the same regimen for retreatment of TB for patients who are not multidrug resistant (as recommended by the South African national TB guidelines at the time of the study). The high recurrence rate also increased the overall treatment failure rate. It is known that all programs for treating TB find higher failure rates in cases of recurrent disease.

We found no association between HIV infection and a history of previous treatment for TB, as has been previously described (23). When rifampicin-containing regimens are used, recurrence rates are similar in HIV-infected and HIV-noninfected patients (7, 24). Further studies are required to determine whether recurrences are due to relapse or reinfection with a different strain of M. tuberculosis and the associated risk factors.

In this study, dual infection with atypical mycobacteria was present in seven patients, and was not associated with HIV status. This probably reflects local conditions, since infection with atypical mycobacteria is well recognized in South African gold miners (27). Dual infection has not been previously described in Africa.

HIV-positive patients presented across a wide spectrum of immune deficiency states (Table 3). As elsewhere in Africa (28, 29), immune suppression was associated with atypical radiographs and less cavitation. Nevertheless, many HIV infected patients were similar to noninfected patients: 61% had typical radiographs, 35% had cavitation, and 69% had a positive sputum smear.

The cure rate in patients with no previous history of treatment for TB was 86% (Table 4). If those patients who died are excluded, the cure rate was similar in HIV-positive and HIV-negative patients (89% versus 88%), which has also been observed in other studies (9, 26). Our study also shows that this was true for patients with advanced immunosuppression, although the numbers for analysis were small.

The WHO has emphasized the importance of directly supervised treatment for TB. Our study confirms that with adequate supervision, very few patients (2, 1%) with a first episode of TB remain infectious after treatment. Noncompliance with DOT as a result of patients' going on leave was strongly associated with treatment failure in our study: 17 of 25 (68%) failures occurred in patients who went on leave, even though they were given drug supplies for their time away from the program and were individually counseled. Southern Africa has a highly mobile population, and this finding highlights the need for TB control programs to be standardized and integrated across the region, and for supervision to be available at the most peripheral level.

It has been reported that among patients with new TB, the presence of resistance to INH and/or streptomycin should not affect the outcome of treatment, provided that four drugs (including INH, rifampicin, and pyrazinamide) are used in the treatment regimen (30). It was encouraging that in our patient cohort, none of the 19 patients with new TB and single-drug resistance (of whom 11 were HIV positive) experienced treatment failure.

We found that treatment failure was more common in patients with recurrent TB (Table 5), and that taking leave of the program and single-drug resistance were significant and independent risk factors. Previous studies have shown that cure rates of over 90% can be achieved in patients with recurrent disease who have bacilli resistant to INH and/or streptomycin, provided that extended, five-drug regimens are used (30). National treatment protocols in South Africa have recently been amended accordingly, and such a regimen has since been introduced into the TB control program in the gold mines.

In the patient cohort in our study, mortality was 13.7% among HIV-positive patients, as compared with 0.5% among HIV-negative patients. The high mortality rate in HIV-positive patients with TB is well known (7, 26, 31). The very low mortality that we found in HIV-negative patients may have been due to earlier recognition of disease through active case finding and access to sophisticated health care. It is unlikely that any program in Southern Africa will achieve much improvement on the mortality observed in the cohort described here.

In this study, early mortality (within 1 mo of diagnosis) among HIV-positive patients was due to TB. Despite a similar radiologic severity of disease at presentation, five of 49 HIV-positive patients with extensive disease died of autopsy-proven TB in the first month, as compared with none of 46 HIV-negative patients. The HIV-positive patients who died of TB were not the most immune compromised, and early detection of these patients may reduce TB-specific mortality. The role of immune modulators needs to be considered in the group of patients with extensive radiologically demonstrated disease in association with only moderate immune suppression.

The present study is the first cohort study with autopsy data to support the contention that much of the excess mortality of HIV-positive patients who die while receiving treatment for TB is due to nontuberculous, AIDS-related conditions (1, 7).

After TB, cryptococcal pneumonia was the most common cause of death among our patients. These patients died later and had more advanced immunosuppression, with atypical radiographs. Although cryptococcal meningitis is well recognized in sub-Saharan Africa, pulmonary involvement is rarely reported (4). Deterioration in a patient already receiving anti-TB treatment should prompt an aggressive search for additional infections or empirical therapy with antifungal and antibacterial agents. Studies of prophylaxis against bacterial and fungal pathogens in deeply immunosuppressed patients with TB are needed.

In our study, profound immune suppression (CD4⁺ lymphocyte percentage < 14) was a strong risk factor for death, with no deaths occurring in patients who had a normal CD4⁺ lymphocyte count. The other factors associated with death were low weight, recurrent TB, sputum smear positivity, drug resistance, and atypical radiographic changes.

Conclusion

In the regions of Africa that are most affected by HIV, up to 70% of cases of pulmonary TB are coinfected with HIV (1, 2), and approximately 10% of patients die. Our study had broadly similar findings, with 50% of cases coinfected with HIV and an overall fatality rate of 7.4%. The mortality rate among HIV-infected patients was 26 times that of HIV-noninfected patients; however, autopsy data showed that more than half of the patients who died did so from nontuberculous causes. The HIV-positive patients who survived responded as well to treatment for TB as the HIV-negative patients, even if they had advanced immune suppression. Despite the very high prevalence of TB in this population, this study clearly demonstrates the efficacy of a well resourced TB control program in effecting cure and saving lives.