Design: Pragmatic community trial with paired screen-positive design
Population: 20,023 adult household members in Lesotho and South Africa
Primary outcome: TB case yield (noninferiority, margin −10%) and cost per case
Effect: 94.5% vs 82.2% of cases found (difference −12.3%, 95% CI −23.0 to −1.6); $5,454 vs $7,486 per case

A pragmatic community trial in Lesotho and South Africa screened 20,023 adult household members, 23% living with HIV, using two approaches in the same people: computer-aided detection (CAD4TBv7) on digital chest X-ray alone, or CAD followed by a point-of-care CRP test for intermediate scores. Xpert MTB/RIF Ultra confirmed cases.

Of 73 people with TB found by either approach, CAD alone identified 69 (94.5%) and CAD plus CRP 60 (82.2%), a difference of −12.3 percentage points (95% CI −23.0 to −1.6). That failed the −10% noninferiority margin. Cost per case found was US$5,454 with CAD alone and $7,486 with CRP added, 37% more.

Adding a CRP step to save confirmatory tests seemed sensible, but it lost cases and saved nothing. The study was published in July 2026, but it bears directly on active case-finding programmes, including India's, which already use CAD on chest X-rays. Settings differ — this was a high-HIV population — but there is no case here for an added CRP filter.

In community TB case-finding, use CAD on chest X-ray as the triage step without a CRP filter
Confirm CAD-positive results with a molecular test such as Xpert Ultra
Adding CRP lost about one in eight cases and raised cost per case found
The trial population had high HIV prevalence, so yields may differ elsewhere

Why it matters

An extra triage test that looks economical can quietly cost cases and money.

The statistics, in plain English

Noninferiority required the CRP approach to miss no more than 10 percentage points more cases. The confidence interval for the difference reached −23.0, beyond that margin, so it failed. A paired design compares both strategies in the same people, which removes differences between groups.