TABLE 1

Limitations observed in CAD4TB evidence-base

QUADAS-2 DomainLimitations observed in CAD4TB evidence basePublished and unpublished studies affected by limitation (number of studies with major concern/total)
Patient selectionExclusion of participants
Potential biased assessment of diagnostic accuracy
Limited generalisability to how software will be used in the field
Systematic screening (2/3)
Triage (1/6)
Index test (CAD4TB)CXR for training, and evaluation of accuracy originate from same dataset
Potential biased assessment of diagnostic accuracy
Limited generalisability to how software will be used in the field
Systematic screening (1/3)
Triage (2/6)
Threshold score not pre-specified
Potential biased assessment of diagnostic accuracy
Limited generalisability to how software will be used in the field
Systematic screening (2/3)
Triage (4/6)
Reference standard for TBComposite reference standard
Potential biased assessment of diagnostic accuracy
Systematic screening (0/3)#
Triage (1/6)
Use of NAAT as reference standard
Potential biased assessment of diagnostic accuracy due to limited sensitivity for smear-negative TB
Systematic screening (0/3)#
Triage (3/6)

TB: tuberculosis; CXR: chest radiography; NAAT: nucleic acid amplification test. #: microbiological tests were used to diagnose TB in all persons with symptoms or human-reader identified radiographic abnormalities; asymptomatic persons with radiographs interpreted as “normal” were assumed to not have TB, and did not undergo microbiological testing.