BackgroundIntroduction of GeneXpert MTB/RIF (Xpert) assay has constituted a major breakthrough for tuberculosis (TB) diagnostics. Several patient factors may influence diagnostic performance of Xpert including sputum quality.ObjectiveWe carried out a prospective, observational, cross-sectional study to determine the effect of sputum quality on diagnostic performance of Xpert among presumed TB patients in Uganda.MethodsWe collected clinical and demographic information and two sputum samples from participants. Staff recorded sputum quality and performed LED fluorescence microscopy and mycobacterial culture on each sample. If both smear examinations were negative, Xpert testing was performed. We calculated diagnostic yield, sensitivity, specificity, and other indicators for Xpert for each stratum of sputum quality in reference to a standard of mycobacterial culture.ResultsPatients with salivary sputum showed a trend towards a substantially higher proportion of samples that were Xpert-positive (54/286, 19%, 95% CI 15–24) compared with those with all other sputum sample types (221/1496, 15%, 95% CI 13–17). Blood-stained sputum produced the lowest sensitivity (28%; 95% CI 12–49) and salivary sputum the highest (66%; 95% CI 53–77). Specificity didn’t vary meaningfully by sample types. Salivary sputum was significantly more sensitive than mucoid sputum (+13%, 95% CI +1 to +26), while blood-stained sputum was significantly less sensitive (-24%, 95% CI -42 to -5).ConclusionsOur findings demonstrate the need to exercise caution in collecting sputum for Xpert and in interpreting results because sputum quality may impact test yield and sensitivity. In particular, it may be wise to pursue additional testing should blood-stained sputum test negative while salivary sputum should be readily accepted for Xpert testing given its higher sensitivity and potentially higher yield than other sample types. These findings challenge conventional recommendations against collecting salivary sputum for TB diagnosis and could inform new standards for sputum quality.

No abstract

BackgroundPneumonia is a major cause of mortality among HIV-infected patients. Pneumonia severity scores are promising tools to assist clinicians in predicting patients’ 30-day mortality, but existing scores were developed in populations infected with neither HIV nor tuberculosis (TB) and include laboratory data that may not be available in resource-limited settings. The objective of this study was to develop a score to predict mortality in HIV-infected adults with pneumonia in TB-endemic, resource-limited settings.MethodsWe conducted a secondary analysis of data from a prospective study enrolling HIV-infected adults with cough ≥2 weeks and <6 months and clinically suspected pneumonia admitted to Mulago Hospital in Kampala, Uganda from September 2008 to March 2011. Patients provided two sputum specimens for mycobacteria, and those with Ziehl-Neelsen sputum smears that were negative for mycobacteria underwent bronchoscopy with inspection for Kaposi sarcoma and testing for mycobacteria and fungi, including Pneumocystis jirovecii. A multivariable best subsets regression model was developed, and one point was assigned to each variable in the model to develop a clinical predictor score for 30-day mortality.ResultsOverall, 835 patients were studied (mean age 34 years, 53.4% female, 30-day mortality 18.2%). A four-point clinical predictor score was identified and included heart rate >120 beats/minute, respiratory rate >30 breaths/minute, oxygen saturation <90%, and CD4 cell count <50 cells/mm3. Patients’ 30-day mortality, stratified by score, was: score 0 or 1, 12.6%, score 2 or 3, 23.4%, score 4, 53.9%. For each 1 point change in clinical predictor score, the odds of 30-day mortality increased by 65% (OR 1.65, 95% CI 1.39-1.96, p <0.001).ConclusionsA simple, four-point scoring system can stratify patients by levels of risk for mortality. Rapid identification of higher risk patients combined with provision of timely and appropriate treatment may improve clinical outcomes. This predictor score should be validated in other resource-limited settings.