Background: The WHO 4-symptom screening tool, comprising any cough, weight loss, fever or night sweats, is a simple evidence-based tool designed for use in resource limited settings to rule out TB in people with HIV, so IPT can be safely initiated. We assessed WHO tool performance using data from the XPHACTOR study, which evaluates a novel algorithm to prioritise TB investigation amongst HIV clinic attendees in South Africa.
Methods: A systematic sample of adult clinic attendees were screened for TB using WHO tool. Sputum was tested with Xpert MTB/RIF if high priority for TB according to XPHACTOR algorithm (any of: cough, BMI<18.5, CD4<100, weight loss ≥10%). All were followed to 3 months, with reinvestigation if indicated. All had sputum and blood cultured for TB at 3 months. TB cases were classified as definite if Xpert+ or culture+ for M. tuberculosis at any point; probable if TB treatment commenced based on compatible radiology; possible if treated without bacteriological or radiological evidence. We calculated negative predictive value, the proportion who did not have TB out of those WHO-tool negative, to evaluate tool performance for ruling out TB.
Results: Among 3460 participants, at enrolment 898 (26%) were pre-ART and 2562 (74%) on ART; 70.5% were female. Amongst pre- vs. on ART, median age was 35 vs. 41 yrs, median CD4 399 vs. 466, 38.1% (342/898) vs. 29.9% (765/2562) were WHO tool +, and 8.1% (72/893) vs. 39.7% (1015/2558) had previous TB treatment (P<0.001).
Prevalence of TB in pre- and on-ART groups respectively was 56/898 (6.2%, 95%CI 4.7%-8.0%) and 67/2562 (2.6%, 95%CI 2%-3.3%). Amongst pre- vs. on ART TB diagnoses, 76.8% (43/56) vs. 89.6% (60/67) were definite, 16.1% (9/56) vs. 9% (6/67) probable, and 7.1% (4/56) vs. 1.5% (1/67) possible (P=0.13).
Negative predictive value of WHO tool (Table) was greater than 98% for all TB, when restricted to definite TB, and when stratified by pre vs. on ART group. Positive predictive values were low (see Table).
Conclusions: In this setting the WHO 4-symptom tool performs extremely well for the purpose for which it was designed, to rule out TB, both amongst ART experienced and ART naïve individuals. However the low positive predictive value necessitates clear guidance on further evaluation of those who screen positive, in order to rationalise further investigation and avoid burdening health care systems in resource-limited settings.