Grant: Project Grant

Tuberculosis (TB) is the leading cause of death from infectious disease as it killed 1.5 million people in 2014. This is mostly because each year, 3.7 million people with TB are not diagnosed or treated. An important reason why so many people with TB are never diagnosed, is that in poor countries where TB is common, tests to diagnose TB are too expensive to use on everyone that might have TB. A possible solution is to first test people with a low-cost triage test-to know which people need to have the more expensive tests. A candidate triage test, is a digital chest x-ray (DXR) automatically analyzed by software for detecting TB (CAD4TB). To assess this, we will enrol 2272 people with symptoms of TB who go to a public hospital in Pakistan-a country where TB is common, and HIV is rare. (This is relevant as almost 90% of people with TB don't have HIV.) All participants will undergo DXR read by CAD4TB. All will also have sputum tested with the best TB test, culture. Culture will tell us which patients truly have TB, and which ones don't. This will allow us to calculate the proportion of people with TB, whose x-rays CAD4TB correctly flagged as abnormal. It will also allow us to calculate the proportion of people without TB, that CAD4TB correctly did not flag. We will then know if CAD4TB meets minimum recommendations for accuracy. We will also calculate the cost-effectiveness of DXR analyzed by CAD4TB. To do so, we will estimate 3 things for an approach where DXR with CAD4TB is used as a triage test: 1) total costs of diagnosing someone with TB, 2) number of TB cases diagnosed, 3) number of people that die from TB. These same things will be estimated for status quo approaches, i.e. that don't use DXR and CAD4TB. We will then determine how much extra does DXR with CAD4TB cost and how much does it help in diagnosing more persons with TB, and even saving more lives. This study will advance our understanding about this promising new tool in the fight against TB.