• Adrian Muwonge, Sydney Malama, Barend M de C Bronsvoort, Demelash Biffa, Willy Ssengooba, and Eystein Skjerve.
• Department of Food Safety and Infection Biology, Centre for Epidemiology and Biostatistics, Norwegian University of Life Sciences, Oslo, Norway; The Roslin Institute, College of Medicine and Veterinary Medicine, University of Edinburgh, Easter Bush, Midlothian, Edinburgh, United Kingdom.
• Plos One. 2014 Jan 1; 9 (6): e100720.

BackgroundThis study compared TB diagnostic tools and estimated levels of misdiagnosis in a resource-limited setting. Furthermore, we estimated the diagnostic utility of three-TB-associated predictors in an algorithm with and without Direct Ziehl-Neelsen (DZM).Materials And MethodsData was obtained from a cross-sectional study in 2011 conducted at Mubende regional referral hospital in Uganda. An individual was included if they presented with a two weeks persistent cough and or lymphadenitis/abscess. 344 samples were analyzed on DZM in Mubende and compared to duplicates analyzed on direct fluorescent microscopy (DFM), growth on solid and liquid media at Makerere University. Clinical variables from a questionnaire and DZM were used to predict TB status in multivariable logistic and Cox proportional hazard models, while optimization and visualization was done with receiver operating characteristics curve and algorithm-charts in Stata, R and Lucid-Charts respectively.ResultsDZM had a sensitivity and specificity of 36.4% (95% CI = 24.9-49.1) and 97.1%(95% CI = 94.4-98.7) compared to DFM which had a sensitivity and specificity of 80.3%(95% CI = 68.7-89.1) and 97.1%(95% CI = 94.4-98.7) respectively. DZM false negative results were associated with patient's HIV status, tobacco smoking and extra-pulmonary tuberculosis. One of the false negative cases was infected with multi drug resistant TB (MDR). The three-predictor screening algorithm with and without DZM classified 50% and 33% of the true cases respectively, while the adjusted algorithm with DZM classified 78% of the true cases.ConclusionThe study supports the concern that using DZM alone risks missing majority of TB cases, in this case we found nearly 60%, of who one was an MDR case. Although adopting DFM would reduce this proportion to 19%, the use of a three-predictor screening algorithm together with DZM was almost as good as DFM alone. It's utility is whoever subject to HIV screening all TB suspects.

23 keywords...

hide…