A toolkit has been developed jointly by the WHO Global Tuberculosis Programme and the Special Programme for Research and Training in Tropical Diseases for conducting a CAD calibration study in a new setting. The toolkit consists of three parts.
- Part A introduces CAD calibration studies, including their designs, outcomes of interest and guidance for interpreting study findings and their application in a TB programme.
- Part B is a generic study protocol, with the proposed research method, data collection and analysis. The protocol can be adapted by users to seek ethics approval. Detailed information assists users in conducting operational research, including study procedures and sample size estimates for various study designs and sampling options, a generic study protocol, data collection tools and an online tool for data analysis.
- Part C is a guide to use of an online tool, “CAD for TB detection”, to support data analysis, to determine appropriate thresholds by demonstrating the practical implications of various CAD thresholds, including for true- and false-positive CAD readings and the costs incurred for follow up confirmatory testing. Part C also includes case studies and guidance to help users interpret the results generated by the online tool and their practical application to local contexts.
The full toolkit can be found online at https://tdr.who.int/activities/calibrating-computer-aideddetection-for-tb.
In general, a CAD calibration study requires identification of the population in which CAD is to be used, sampling of the population to obtain CAD scores and the TB status of a subset of the population, and use of that data to calculate the sensitivity and specificity of CAD at different potential thresholds. The implications of the thresholds in terms of yield of true- and false-positive cases detected and the cost of diagnostic evaluation can then be used to guide the choice of a CAD threshold for a given implementation according to the goals of the screening or triage programme.