Abstract
Background: Cough is common in pulmonary TB and other chronic respiratory infections. Identifying features that predict whether pulmonary TB is the cause would help target appropriate individuals for rapid and cost-effective screening, potentially limiting disease progression and preventing transmission to others. Methods: A systematic literature search for individual studies to answer eight key questions (KQs) was conducted according to established Chest Organization methods by using the following databases: MEDLINE via PubMed, Embase, Scopus, and the Cochrane Database of Systematic Reviews from January 1, 1984, to April 2014. Searches for KQ 1 and KQ 3 were updated in February 2016. An updated KQ 2 search was undertaken in March 2017. Results: Even where TB prevalence is greatest, most individuals with cough do not have pulmonary TB. There was no evidence that 1, 3, or 4 weeks’ duration were better predictors than cough lasting ≥ 2 weeks to screen for pulmonary TB. In people living with HIV (PLWHIV), screening for fever, night sweats, hemoptysis, and/or weight loss in addition to cough (any World Health Organization [WHO]-endorsed symptom) increases the diagnostic sensitivity for TB. Although the diagnostic accuracy of symptom-based screening remains low, the negative predictive value of the WHO-endorsed symptom screen in PLWHIV may help to risk-stratify individuals who are not close TB contacts and who do not require further testing for pulmonary TB in resource-limited settings. However, pregnant PLWHIV are more likely to be asymptomatic, and the WHO-endorsed symptom screen is not sensitive enough to be reliable. Combined with passive case finding (PCF), active case finding (ACF) identifies pulmonary TB cases earlier and possibly when less advanced. Whether outcomes are improved or transmission is reduced is unclear. Screening asymptomatic patients is cost-effective only in populations with a very high TB prevalence. The Xpert MTB/RIF assay on sputum is more cost-effective than clinical diagnosis. To our knowledge, no published comparative studies addressed whether the rate of cough resolution is a reliable determinant of the response to treatment or whether the rate of cough resolution was faster in the absence of cavitary lung disease. All studies on cough prevalence in Mycobacterium avium complex (MAC) lung disease, other nontuberculous mycobacterial infections, fungal lung disease, and paragonimiasis were of poor quality and were excluded from the evidence review. Conclusions: On the basis of relatively few studies of fair to good quality, we conclude that most individuals at high risk and household contacts with cough ≥ 2 weeks do not have pulmonary TB, but we suggest screening them regardless of cough duration. In PLWHIV, the addition of the other WHO-endorsed symptoms increases the diagnostic sensitivity of cough. Earlier screening of patients with cough will help diagnose pulmonary TB sooner but will increase the cost of screening. The addition of ACF to PCF will increase the number of pulmonary TB cases identified. Screening asymptomatic individuals is cost-effective only in groups with a very high TB prevalence. Data are insufficient to determine whether cough resolution is delayed in individuals with cavitary lung disease or in those for whom treatment fails because of drug resistance, poor adherence, and/or drug malabsorption compared with results in other individuals with pulmonary TB. Cough is common in patients with lung infections due to MAC, other nontuberculous mycobacteria, fungal diseases, and paragonimiasis.