Screening for active tuberculosis among patients accessing antiretroviral therapy in sub-Saharan Africa

We read with interest the article by Were et al. concerning a simple screening tool for tuberculosis (TB) among HIV-infected patients attending antiretroviral treatment (ART) services in Uganda.¹ There is a great need for effective intensified case finding in such clinical settings since TB is a major cause of morbidity, mortality and nosocomial disease transmission. Early detection of active TB would enable rapid initiation of TB treatment while reliable exclusion of disease in the remainder would permit use of isoniazid preventive therapy.

This screening tool detected any one of four symptoms or signs of TB (cough ≥3 weeks, fever ≥4 weeks, lymphadenopathy or body mass index ≤18 kg/m²) and was assessed as having a sensitivity of 99% and a negative predictive value of 100% using their case definition and gold standard for TB. However, the apparent sensitivity of a screening tool is of course dependent on the gold standard used. In this study this was a TB diagnosis made according to the Uganda Ministry of Health diagnostic guidelines. Diagnoses of pulmonary TB were made in symptomatic patients with either two positive sputum smears or chest radiographs compatible with TB and no improvement of symptoms after two weeks of broad-spectrum antibiotics. Symptoms were therefore an integral component of the gold standard diagnosis. Patients with minimally symptomatic or asymptomatic disease will largely have been excluded.

The true gold standard in TB screening studies in HIV-infected individuals is culture of sputum from all participants (regardless of symptoms) together with appropriate investigations for extrapulmonary disease. Using a culture-based gold standard, many studies report that the sensitivity of symptom screening in such patients is limited, indicating that disease is asymptomatic in a significant proportion of patients.² In our own study of patients just prior to ART initiation in South Africa, a combination of four possible symptoms had a sensitivity of just 78%³ and this did not improve appreciably with inclusion of lymphadenopathy and body mass index ≤18 kg/m². Thus, consistent with other studies, approximately one fifth of these TB patients with advanced immunodeficiency had sub-clinical culture-positive disease.²

The negative predictive value of symptom screening in our study was just 73%.³ In contrast, Were et al. reported a negative predictive value of 100%, suggesting their screening tool had exceptional reliability for excluding TB.¹ Again, this may simply reflect failure of the study to investigate patients who did not have overt symptoms of TB. This is problematic as it suggests that their screening tool alone could be used to reliably exclude TB in HIV-infected patients prior to initiation of isoniazid preventive therapy. However, significant rates of clinically unrecognised culture-positive TB in numbers of studies indicate that this is typically not the case.²

One of the great challenges of ART programmes in sub-Saharan Africa is high early mortality of which TB is a leading cause.⁴ In the pre-ART era, extremely high rates of clinically unrecognised disseminated TB were found in post-mortem studies of patients dying of AIDS.⁵ We suspect that a major burden of undiagnosed TB may similarly exist among patients with advanced immunodeficiency dying just prior to ART and during early treatment. Development of potential strategies to address this high mortality must explore means of detecting not only symptomatic TB but also disease which is not clinically apparent.