Public-private mix for control of tuberculosis and TB-HIV in Nairobi, Kenya: outcomes, opportunities and obstacles.

SETTING: Nairobi, the capital of Kenya. OBJECTIVE: To promote standardised tuberculosis (TB) care by private health providers and links with the public sector. DESIGN AND METHODS: A description of the results of interventions aimed at engaging private health providers in TB care and control in Nairobi. Participating providers are supported to provide TB care that conforms to national guidelines. The standard surveillance tools are used for programme monitoring and evaluation. RESULTS: By the end of 2006, 26 of 46 (57%) private hospitals and nursing homes were engaged. TB cases reported by private providers increased from 469 in 2002 to 1740 in 2006. The treatment success rate for smear-positive pulmonary TB treated by private providers ranged from 76% to 85% between 2002 and 2005. Of the 1740 TB patients notified by the private sector in 2006, 732 (42%) were tested for human immunodeficiency virus (HIV), of whom 372 (51%) were positive. Of the 372 HIV-positive TB patients, 227 (61%) were provided with cotrimoxazole preventive treatment (CPT) and 136 (37%) with antiretroviral treatment (ART). CONCLUSION: Private providers can be engaged to provide TB-HIV care conforming to national norms. The challenges include providing diagnostics, CPT and ART and the capacity to train and supervise these providers.

Cost-effectiveness of polymerase chain reaction versus Ziehl-Neelsen smear microscopy for diagnosis of tuberculosis in Kenya.

BACKGROUND: Laboratory services, particularly in large sub-Saharan cities, are overstretched, and it is becoming difficult both for patients and health staff to adhere to the diagnostic procedures for tuberculosis. Alternative techniques would be welcome. The polymerase chain reaction (PCR) has the potential to be cost-effective. We compared the cost-effectiveness of two diagnostic strategies, Ziehl-Neelsen (ZN) on three specimens followed by chest X-ray (CXR), and AMPLICOR MTB PCR on the first specimen only. METHODS: Three sputum samples were collected from tuberculosis (TB) suspects attending the Rhodes Chest Clinic, Nairobi. All samples were subjected to ZN, PCR and Löwenstein-Jensen culture used as gold standard. CXR was used to diagnose smear-negative TB. Cost analysis included health service and patient costs. RESULTS: Costs per correctly diagnosed case were US dollar 41 and dollar 67 for ZN and PCR, respectively. When treatment costs were included, including treatment of culture-negative cases, PCR was more cost-effective: dollar 382 vs. dollar 412. CONCLUSION: PCR may be an alternative in settings with many patients. PCR is patient friendly, CXR is not necessary and, unlike ZN, its performance is hardly affected by the human immunodeficiency virus. PCR can handle large numbers of specimens, with results becoming available on the same day.

A comprehensive comparison of Ziehl-Neelsen and fluorescence microscopy for the diagnosis of tuberculosis in a resource-poor urban setting.

SETTING: Nairobi City Council Chest Clinic, Kenya. OBJECTIVES: To establish the efficiency, costs and cost-effectiveness of six diagnostic strategies using Ziehl-Neelsen (ZN) and fluorescence microscopy (FM). DESIGN: A cross-sectional study of 1398 TB suspects attending a specialised chest clinic in Nairobi subjected to three sputum examinations by ZN and FM. Lowenstein-Jensen culture was used as the gold standard. Cost analysis included health service and patient costs. RESULTS: Of 1398 suspects enrolled, 993 (71%) had a complete diagnostic work-up involving three sputum specimens for ZN and FM, culture and chest X-ray (CXR). Irrespective of whether ZN or FM was used on one, two or three smears, the overall diagnostic process detected 92% culture-positive cases. Different strategies affected the ratio of smear-positive to smear-negative TB; however, FM was more sensitive than ZN (P < 0.001). FM performance was not affected by the patient's HIV status. The cost per correctly diagnosed smear-positive case, including savings, was 40.30 US dollars for FM on two specimens compared to 57.70 US dollars for ZN on three specimens. CONCLUSION: The FM method used on one or two specimens is more cost-effective and shortens the diagnostic process. Consequently, more patients can be put on a regimen for smear-positive TB, contributing to improved treatment and reducing transmission.

Decentralisation of tuberculosis treatment from the main hospitals to the peripheral health units and in the community within Machakos district, Kenya.

SETTING: A rural district, Machakos, in Kenya, facing decreasing national resources for health and an increasing tuberculosis (TB) caseload fuelled by the human immunodeficiency virus (HIV). OBJECTIVE: To evaluate the impact on district TB programme performance of decentralising TB treatment by providing ambulatory care in the hospital and peripheral health units and in the community. METHODS: A comparative study of district TB programme performance before and after the decentralisation of TB services at the end of 1997. To facilitate ambulatory care, ethambutol replaced streptomycin in the new treatment regimen. FINDINGS: The number of patients registered in the control period (1996) was 1141, of whom almost 100% were admitted during the intensive phase of TB treatment, and in the intervention period (1998 and 1999), it was 3244, of whom only 153 (4.7%) required admission in the intensive phase. Of 3244 TB patients (all forms) registered in the intervention period, the number (%) choosing the different options for directly observed treatment (DOT) supervision were: hospital clinic 1618 (49.9%), peripheral health unit 904 (27.9%), community volunteer 569 (17.5%) and hospitalisation 153 (4.7%). The options were found to be acceptable to patients, their families and health staff. The treatment outcomes among new sputum smear-positive pulmonary TB patients were similar in the intervention and control cohorts, with treatment success rates of 88% vs. 85% and death rates of 4% vs. 6%, respectively. Treatment completion was significantly higher among new sputum smear-negative and extra-pulmonary TB patients in the intervention than in the control cohort (79% vs. 48%, respectively). CONCLUSION: The decentralisation of the intensive phase of TB treatment resulted in maintenance of good TB programme performance, while Machakos hospital closed its TB wards. A separate paper describes the cost-effectiveness of this approach. The National Tuberculosis Control Programme plans to adopt this approach as national policy.

Drug Resistance to AIDS; Tuberculosis and Malaria in the African Region

With just 10of the world population; sub-Saharan Africa has the highest burden of HIV/AIDS; tuberculosis and malaria in the world. Both access to and adequate utilization of eff ective treatment with quality-assured medicines are crucial for reducing the disease burden. However; eff orts to improve access to treatment are hampered by the development of HIV; TB and malaria drug resistance. This is a result of genetic mutations and is a major threat to control of HIV/AIDS; TB and malaria. HIV drug resistance can be minimized by good antiretroviral treatment (ART) programmes; removal of barriers to continuous access to ART and reduction of HIVtransmission. Recent surveys conducted at antenatal clinics in several countries in the African Region estimated that HIV resistance to all drug classes is less than 5. A global HIV drug resistance network established in 2001 supports countries in capacity building and guidance on standard procedures for monitoring HIV drug resistance. Multidrug-resistant TB (MDR-TB) and extensively drug-resistant TB (XDR-TB) are principally a result of inadequate or poorly administered treatment regimens. The new WHO Stop TB Strategy launched in 2006 identifies management of MDR-TB as a core component of TB control. The magnitude of MDR-TB in the African Region is still unknown. In 2007; 27 countries notifi ed MDR-TB cases; and six reported at least one case of XDR-TB. Following widespread resistance to chloroquine and sulphadoxine-pyrimethamine all malaria-endemic countries except two in the Region have changed the treatment policy to artemisinin-based combination therapy (ACT). The main method of monitoring antimalarial drug resistance is through therapeutic efficacy testing. Todate there has been no confi rmed resistance to ACTs in the African Region. Given the emergence and spread of resistance to HIV; TB and malaria drugs; the purpose of this paper is to describe the issues and challenges and propose a way forward with regard to the prevention and control of such resistance

Tuberculosis re-treatment outcomes within the public service in Nairobi, Kenya.

OBJECTIVES: This study was undertaken to describe treatment outcomes in patients started on a re-treatment drug regimen, assess the quality of follow up procedures and the adequacy of the currently advocated re-treatment drug regimen in Nairobi, Kenya. DESIGN: A retrospective study. SETTING: Mbagathi District Hospital (MDH), Nairobi, a public hospital that serves as the Tuberculosis (Tb) referral centre for Nairobi. MATERIALS AND METHODS: The Tb register at the MDH was used to identify patients who were on the re-treatment regimen for Tb. Case records for these patients were then retrieved. From these sources, information on age, sex, HIV status, previous and current tuberculosis disease and drug regimens, adherence to treatment and treatment outcomes, was obtained. Descriptive statistics was used to analyse the data. RESULTS: Of the total of 4702 patients registered at the MDH between 1996 and 1997, 593 (12.6%) were patients with either recurrent Tb, returning to treatment after default or had failed initial treatment. Of the 593 patients, case records were unavailable for 168 and 17 were children below the age of ten in whom the diagnosis of Tb was uncertain making a total of 185 patients who were excluded from the study. Of the remaining 408 patients, 77 (18.9%) were cured, 61 (15.0%) completed treatment without confirmation of cure, two (0.5%) defaulted, six (1.5%) died and 262 (64.2%) had no outcome information. There were no treatment failures. Treatment success defined as cure or treatment completion was achieved in 94.5% of the 146 patients in whom outcome data were available. HIV positive patients had a statistically significant poorer success rate (34/40, 85%) when compared with HIV negative patients (104/106, 94%), p=0.004. Mycobacterium tuberculosis culture and drug susceptibility testing, was not done. CONCLUSION: The high number of patients with no treatment outcome information at the MDH is worrying, as these patients may harbour drug resistant bacilli and reflects an inadequate follow up service for Tb re-treatment in Nairobi. However, where treatment outcomes could be assessed, the currently advocated re-treatment regimen achieved a high success rate. These observations point to an urgent need to improve Tb documentation and follow up procedures within the public service in Nairobi in order to forestall the emergence and spread of drug resistant Tb.

Multi-sectoral approach to DOTS--the only hope.

This article reviews both the global and Kenyan tuberculosis (TB) situations and gives a manager's perspective of the management of a national TB control program with limited resources. It demonstrates how complex it can be to manage a control program with a 500% increase in case load and a 60% decline in real per capita health expenditure during the last 20 y. The picture is complicated even further by a high prevalence of TB among the country's nomadic population inhabiting the arid and semi-arid parts of the country. Acceptable treatment results are achieved through a combination of appropriate strategies. However, the future picture is grim unless more personnel are used to implement the DOTS strategy.

Surveillance of drug-resistant tuberculosis and molecular evaluation of transmission of resistant strains in refugee and non-refugee populations in North-Eastern Kenya.

SETTING: Three refugee camp complex clinics and an adjacent non-refugee treatment centre in North-Eastern Kenya. OBJECTIVES: To use conventional and molecular epidemiology tools to determine: 1) the prevalence of drug resistance in newly diagnosed patients with smear-positive pulmonary tuberculosis in refugee and non-refugee populations; 2) risk factors for resistance in the two populations; and 3) whether IS6110 restriction fragment length polymorphism (RFLP) and spoligotyping showed similarities in DNA fingerprinting patterns of drug-resistant isolates that could infer transmission within and between the two populations. RESULTS: Of 241 isolates from the camps, 44 (18.3%) were resistant to one or more drugs, seven of which (2.9%) were multidrug-resistant TB (MDR-TB). Of 88 isolates from the non-refugees, five (5.7%) were resistant to one or more drugs without MDR-TB. Drug resistance was higher in the camps than in the non-refugee population (OR = 3.7; 95%CI 1.42-9.68; P < 0.007). Resistance was significantly higher in one camp compared with the other two, despite a comparable ethnic distribution. Unusually, females were more associated with drug resistance than their male counterparts in both populations (OR = 2.3; 95%CI 1.2-4.8; P = 0.008). There was evidence of transmission of streptomycin-resistant strains in the refugee population. DNA fingerprints of resistant strains from the non-refugee population were unique and different from those in the refugee camps. CONCLUSION: The observed high levels of drug resistance and MDR-TB, combined with evidence of transmission of strains resistant to streptomycin in the refugee population, suggest a need for strengthened TB control programmes in settings with a high risk of developing drug-resistant strains.

Effective tuberculosis control and health sector reforms in Kenya: challenges of an increasing tuberculosis burden and opportunities through reform.

During the period from 1980 to 1997, the annual number of new tuberculosis cases increased four-fold in Kenya, and had reached approximately 50,000 cases by 1998. During the same time period, the government per capita expenditure on health dropped from US$9.5 to US$3.5. Since 1983, Kenya has been decentralising financial responsibility and decision-making power to the districts. In addition, the late 1980s saw the introduction of cost-sharing schemes for most health services, excluding tuberculosis (TB) treatment. In the midst of these changes, a dual epidemic of TB and HIV/AIDS emerged, and is presently over-burdening the traditional public health system. In response, the National Leprosy and Tuberculosis Control Programme (NLTP) is seeking a wider network of service providers and new approaches to the prevention and treatment of TB in the country. The history of health sector reform in Kenya is summarised and the role of the NLTP in these reforms assessed. Recent approaches taken by the NLTP to sustain effective TB control, which draw on the environment of a changing and flexible health system, are expressed. Participation of the NLTP in components of health sector reform, particularly decentralisation, integration, financing through cost-sharing and public/private mix, are highlighted.

HIV sero-prevalence among tuberculosis patients in Kenya.

OBJECTIVE: To determine HIV seroprevalence among tuberculosis patients and the burden of HIV attributable tuberculosis among notified patients in Kenya. DESIGN: A cross-sectional anonymous unlinked HIV seroprevalence survey. SETTING: Tuberculosis diagnostic clinics of the National Leprosy Tuberculosis Programme in 19 districts. SUBJECTS: One thousand nine hundred and fifty-two newly notified tuberculosis patients. INTERVENTIONS: Selection and registration of eligible subjects followed by obtaining 5 ml of full blood for haemoglobin testing and separation of serum for HIV testing by ELISA. MAIN OUTCOME MEASURES: HIV seroprevalence per district and burden of HIV attributable tuberculosis among tuberculosis patients. RESULTS: A total of 1,952 eligible patients were enrolled. The weighted seroprevalence in the sample was 40.7% (range 11.8-79.6% per district). The seroprevalence was significantly higher among females and patients with sputum-smear negative tuberculosis. Chronic diarrhoea, female sex, oral thrush and a negative sputum were independent risk factors for HIV infection. The Odds ratio for HIV infection in female tuberculosis patients aged 15-44 years, was 5.6 (95% CI 4.5-6.9) compared with ante-natal clinic attenders. The population attributable risk was 0.22 in 1994. CONCLUSION: The HIV epidemic has had a profound impact on the tuberculosis epidemic in Kenya and explains about 41% of the 94.5% increase of registered patients in the period 1990-1994 and 20% of all registered patients in 1994. Repetition of the survey with inclusion of a more representative control group from the general population may provide a more accurate estimation of the burden of HIV attributable tuberculosis.

PIP: This cross-sectional survey determined HIV seroprevalence among tuberculosis patients and the burden of HIV attributable tuberculosis among notified patients in Kenya. Data were collected from 1952 patients. The information gathered included demographic data, date of treatment initiation, type of patient, type of tuberculosis, sputum-smear results, and data concerning the signs and symptoms related to tuberculosis and HIV disease. Findings demonstrated that the weighted seroprevalence in the study sample was 40.7% (range, 11.8-79.6% per district), which is significantly higher in females and patients with sputum-smear negative tuberculosis. Chronic diarrhea, female sex, oral thrush, and negative sputum were independent risk factors for HIV infection. The odds ratio for HIV infection in female tuberculosis patients aged 15-44 years was 5.6 compared with antenatal clinic attenders.

Tuberculosis and the HIV epidemic: increasing annual risk of tuberculous infection in Kenya, 1986-1996.

OBJECTIVES: The purpose of this study was to assess the impact of the increased incidence of tuberculosis (TB) due to HIV infection on the risk of TB infection in schoolchildren. METHODS: Tuberculin surveys were carried out in randomly selected primary schools in 12 districts in Kenya during 1986 through 1990 and 1994 through 1996. Districts were grouped according to the year in which TB notification rates started to increase. HIV prevalence in TB patients and changes in TB infection prevalence were compared between districts. RESULTS: Tuberculous infection prevalence rates increased strongly in districts where TB notification rates had increased before 1994 (odds ratio = 3.1, 95% confidence interval = 2.3, 4.1) but did not increase in districts where notification rates had increased more recently or not at all. HIV prevalence rates in TB patients were 50% in districts with an early increase in notification rates and 28% in the other study districts. CONCLUSIONS: Countries with an increasing prevalence of HIV infection will need additional resources for TB control, not only for current patients but also for the patients in additional cases arising from the increased risk of TB infection.

Altitude: a determinant for tuberculosis in Kenya?

OBJECTIVE: To determine to what extent tuberculosis incidence is associated with altitude. METHODS: Notification rates were obtained from all 41 districts in Kenya in the period 1988-1990; the mean altitude of each district was estimated. Data on indicators of socio-economic status such as literacy rate and infant mortality rate were obtained from the 1989 census, as well as data on other potential confounders such as urbanisation and median household size. RESULTS: The notification rate of new smear-positive tuberculosis was 32/100000 overall, varying between districts from 5 to 222/100000. Notification rates steeply reduced with increasing altitude (r = -0.71; 95% confidence interval [CI] -0.51 to -0.83). At altitudes of 1000 m or more the notification rates were less than 30% of those in districts at altitudes below 500 m, also after adjustment for confounding. CONCLUSION: Tuberculosis incidence in Kenya decreases strongly with increasing altitude. If the association is not due to unknown confounding factors, a range of potential biological explanations needs to be explored.

Antituberculosis drug resistance surveillance in Kenya, 1995.

SETTING: Twenty-two of the 42 administrative districts in Kenya. OBJECTIVE: To determine the prevalence of drug resistance in newly diagnosed patients with pulmonary tuberculosis, to determine possible risk factors associated with resistance, and to establish standard routine surveillance of drug resistance. DESIGN: Cross-sectional study. METHODS: Sputum samples from newly diagnosed patients with smear-positive pulmonary tuberculosis were analysed using standard procedures. RESULTS: Of 638 patients, 85% were culture positive for Mycobacterium tuberculosis. Of 491 patients tested for susceptibility to isoniazid, streptomycin, rifampicin and ethambutol, 90.8% had fully sensitive strains and 9.2% had a strain resistant to one or more drugs. Of 445 patients with no history of previous chemotherapy, 6.3% had a resistant strain. Of 46 patients with a history of previous chemotherapy, 37% had a resistant strain. No resistance to either rifampicin or ethambutol was detected. There was a strong association between previous chemotherapy and resistance. Resistance was not associated with age or sex. High concordance between Kenya's results and those of the Mycobacterium Reference Unit in the UK on both drug-sensitive and drug-resistant strains indicates that clinically significant and comparable data can be obtained from laboratories employing unsophisticated and inexpensive standard procedures. CONCLUSION: Rates of initial drug resistance are still low in Kenya. The increase in acquired resistance to isoniazid requires monitoring.

Cost-effectiveness of the polymerase chain reaction versus smear examination for the diagnosis of tuberculosis in Kenya: a theoretical model.

SETTING: A major out-patient tuberculosis clinic in Nairobi, Kenya. OBJECTIVE: To ascertain the cost-effectiveness of the polymerase chain reaction (PCR) for the diagnosis of tuberculosis in an urban setting in a developing country. DESIGN: A cost-effectiveness analysis of PCR and direct smear microscopy examination based on theoretical modelling. The cost-effectiveness was expressed in costs per correctly diagnosed tuberculosis patient for each of the two diagnostic techniques. Data were obtained from the literature, from the staff and the register at the health facility and from structured interviews with patients. Assumptions were made when no data were available. RESULTS: The PCR is expected to be more specific and sensitive than the routine procedure for diagnosis, but it is also more costly. The routine procedure based on direct smear microscopy turned out to be 1.8 times as cost-effective as PCR. CONCLUSION: It is concluded that the PCR method can potentially be a cost-effective screening procedure for tuberculosis, provided that the largest contributing cost component, the costs of the PCR-kit, can be reduced substantially.

Cost-effectiveness and total costs of three alternative strategies for the prevention and management of severe skin reactions attributable to thiacetazone in the treatment of Human Immunodeficiency Virus positive patients with tuberculosis in Kenya.

SETTING: Severe skin reactions due to thiacetazone (T) in Human Immunodeficiency Virus (HIV) positive tuberculosis patients have been reported in several publications, one of them from Kenya. However, the abandoning of T may not be feasible in Kenya as this may increase the cost of drugs by about three-fold per regimen. OBJECTIVE: To compare the cost-effectiveness and total cost of three strategies in which T is replaced with ethambutol (E). DESIGN: Three strategies are compared with a baseline strategy in which T is not replaced. The indicator for cost-effectiveness is the cost-per-averted-death attributable to T. RESULTS: Education of patients on the possibility of side-effects and replacement of T with E is the most cost-effective strategy at HIV prevalence rates of 1-90%. Abandonment of T and replacement with E is the most cost-effective at over 90% HIV prevalence. CONCLUSION: In Kenya, education of patients on the possibility of skin reactions should be preferred at low range HIV prevalence rates. Routine HIV testing would be the most attractive strategy in the middle range, and total replacement of T with E is to be preferred in the higher range of HIV prevalence.

PIP: In Kenya, the National Leprosy Tuberculosis Programme (NLTP) used previously reported data from Nairobi to compare the cost-effectiveness and total costs of a hypothetical strategy with three intervention strategies for the prevention and management of severe skin reactions caused by thiacetazone in treating HIV-positive patients with tuberculosis (TB). The hypothetical strategy was continued use of thiacetazone despite adverse skin reactions. The intervention strategies included patient education about possible side effects of anti-TB drugs (discontinue use if skin rash develops, report situation to clinic, replace thiacetazone with ethambutol when other skin diseases have been excluded), abandonment of thiacetazone and replacement with ethambutol, and HIV testing and pre- and post-test counseling. NLTP currently used the education strategy. It assumed a mortality rate of 5%. When the HIV prevalence rate is 1-90%, the education strategy is the most cost-effective strategy. In terms of total costs, the education strategy was also the most inexpensive strategy regardless of the HIV prevalence. At an HIV prevalence rate greater than 65%, the abandonment of thiacetazone strategy was the cheapest strategy. When the assumed mortality rate was 3%, the cost per averted death for the education strategy was reduced from about US$120 to about US$80 and the education strategy became the most cost-effective strategy over the entire range of HIV prevalence. In addition, the cost of HIV testing significantly increased the cost per averted death. Thus, the findings of this study are truly sensitive to different program conditions. Based on these findings, the authors recommended that the education strategy be applied with a range of HIV prevalence of 1-45%, that HIV testing be applied with a range of 46-72%, and that total abandonment be applied with an HIV prevalence greater than 72%.

The impact of HIV on resource utilization by patients with tuberculosis in a tertiary referral hospital, Nairobi, Kenya.

By using routinely collected data and results from research studies at the Infectious Diseases Hospital (IDH), Nairobi, we have begun to determine the scale of the increase in resource utilisation and treatment costs for tuberculosis control services caused by the HIV epidemic. New cases of tuberculosis registered annually at the IDH rose 61%, from 447 in 1985 to 720 in 1990. HIV seroprevalence among patients with tuberculosis rose from 7.5% in 1986 to 42% in 1990. The inpatient mortality rate rose from 8.4% in 1985 to 16.8% in 1989, but fell to 13.5% in 1990. HIV-positive patients were admitted to hospital on 2 or more occasions more often than HIV-negative patients (Relative risk (RR) = 2.46, 95% confidence intervals (CI), 1.1-5.7), but average duration of admission was similar for the 2 groups. Significantly more HIV-positive patients were prescribed antibiotics, antifungal agents, antidiarrhoeal agents, analgesics and corticosteroids than HIV-negative patients. Microbiological investigations, apart from those for tuberculosis, were performed more commonly among HIV-positive patients (RR = 2.0, 95% CI 1.0-4.2). Using this data, the average cost of ideal drug therapy, including antituberculosis drugs and treatment for intercurrent infections and other complications, was estimated using 1992 prices (ECHO, Coulsdon Surrey, UK). The costs were US$16.62 and US$32.94 for HIV-negative patients using 'standard' therapy (2STH/10TH) and short course therapy (2SHRZ/6TH) respectively, and US$41.18 for HIV-positive patients using a short-course regimen without thiacetazone (2EHRZ/6EH). The HIV epidemic is causing both an increase in the numbers of patients requiring treatment and an increase in the average cost of treatment per patients.(ABSTRACT TRUNCATED AT 250 WORDS)

Cross-sectional survey of HIV infection among patients with tuberculosis in Nairobi, Kenya.

Evidence from many countries suggests an association of human immunodeficiency virus (HIV) infection and tuberculosis of major public health significance. In order to begin assessing the impact of HIV on tuberculosis in Kenya, we have determined the HIV-1 seroprevalence among tuberculosis patients and compared the clinical characteristics of tuberculosis in HIV-positive and HIV-negative patients in two cross-sectional studies at the Infectious Disease Hospital (IDH) and the Ngaira Avenue Chest Clinic (NACC), Nairobi, Kenya. The diagnosis in 92% of all patients with pulmonary tuberculosis was confirmed by culture. The remainder were diagnosed on histological, clinical or radiological grounds. HIV seroprevalence among tuberculosis patients at IDH was 26.5% (52/196) compared to 9.2% (18/195) at NACC (P less than 0.001). There was no association between numbers of streptomycin injections in the previous 5 years and HIV infection. Positive sputum smear rates in HIV-positive patients were slightly lower than in HIV-negative patients at both study sites (71% vs 83% at IDH and 73% vs 82% at NACC) but the difference was not significant. Only Mycobacterium tuberculosis was isolated. Miliary disease was not associated with HIV infection. Persistent diarrhoea, oral candidiasis, generalized itchy rash, herpes zoster and generalized lymphadenopathy were all associated with HIV infection, but 46% (95% CI:38-54%) of all HIV-positive patients had none of the clinical features listed in the WHO Clinical Criteria for the Diagnosis of AIDS, apart from fever, cough and weight loss. Stevens-Johnson Syndrome was reported in 7/52 (13%) patients with HIV infection, and in 4/144 (3%) patients without (RR 4.85, 95% CI: 1.45-15.88).(ABSTRACT TRUNCATED AT 250 WORDS)