Two findings today failed for the same reason, and neither failure was about the intervention being ineffective.
In the tuberculosis trial, Xpert Ultra and urine LAM work — their sensitivity is not in question. The strategy failed because 39% of eligible control patients never produced a sputum sample, and offering the test to more people does nothing about that. In the stewardship trial, the guidelines were identical in both arms; only the delivery changed, and delivery was not what stopped three-quarters of prescriptions following them.
The habit is to ask, of any implementation study, what fraction of the target population actually received the intervention as intended, before reading the effect estimate. A trial of a strategy answers a different question from a trial of a test, and a strategy can fail while everything inside it works. When it does, the fix is in the denominator, not the technology.
- Ask what proportion of eligible patients actually received the intervention
- A test's sensitivity and a strategy's yield are different quantities
- Where uptake is the constraint, widening eligibility changes nothing
- Look for the step in the pathway where people are lost, not the tool
- Applies to screening programmes as much as to diagnostics
The statistics, in plain English
An intention-to-treat analysis counts everyone as randomised, which is the right way to estimate what a policy would achieve in practice. It is not the right way to judge whether a test works, because incomplete uptake dilutes the estimate toward no effect. Both readings are legitimate; confusing them is how a working test gets abandoned.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for infectious diseases, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free