Diagnostic Test Accuracy in Childhood Pulmonary Tuberculosis: A Bayesian Latent Class Analysis

Area of Science:

• Infectious Diseases
• Pediatric Pulmonology
• Diagnostic Accuracy Studies
• Biostatistics

Background:

• Diagnosing childhood pulmonary tuberculosis (CPTB) is complex due to the lack of a definitive reference standard.
• Accurate diagnostic tools are crucial for timely and appropriate treatment of CPTB in hospitalized children.
• Previous evaluations of CPTB diagnostic tests have been limited by methodological constraints.

Purpose of the Study:

• To evaluate and compare the diagnostic accuracy of five common tests for childhood pulmonary tuberculosis (CPTB).
• To estimate the prevalence of CPTB in a South African pediatric cohort.
• To assess the impact of test performance on treatment decisions in suspected CPTB cases.

Main Methods:

• Bayesian latent class analysis was employed to assess the accuracy of five diagnostic tests for CPTB.
• Data were collected from 749 hospitalized children suspected of having CPTB in South Africa (2009-2014).
• Tests evaluated included mycobacterial culture, smear microscopy, Xpert MTB/RIF, tuberculin skin test (TST), and chest radiography.

Main Results:

• Estimated CPTB prevalence was 27%.
• Sensitivities varied: TST (75%), chest radiography (64%), culture (60%), Xpert MTB/RIF (49%), and smear microscopy (22%). TST sensitivity decreased in malnourished and HIV-infected children.
• High rates of antituberculosis treatment were observed in both CPTB-negative (46%) and CPTB-positive (93%) cases, indicating overtreatment and undertreatment.

Conclusions:

• No single diagnostic test demonstrated perfect accuracy for CPTB, highlighting the need for multi-test approaches.
• The Tuberculin Skin Test (TST) showed the highest sensitivity but its performance is affected by nutritional status and HIV co-infection.
• Substantial overtreatment of CPTB-negative children and undertreatment of CPTB-positive children underscore the diagnostic challenges and treatment implications.