Capability example

A point-of-care test, evidenced where it is actually used.

Laboratory performance proved nothing about the ward. The study was designed around the three settings the device would meet.

All capability examples

Sector

Diagnostic devices

Device type

Point-of-care analyser, Class B

Services involved

Clinical research, quality and regulatory

Duration

14 months, protocol to submission

The challenge

Sensitivity and specificity, measured in the wrong room.

The existing evidence came from trained operators, controlled samples and a bench. The claim was for a ward, a clinic and a patient home. Those three settings differ in operator training, sample handling, ambient conditions and how long a result waits before it is read, and every one of those differences moves the number that matters.

What we did

Designed the study around the settings, not the device.

We built a protocol that enrolled in all three settings with the operators who would really use it, prespecified the agreement analysis against the reference method, and defined in advance what an unusable result was so that it could be counted rather than argued about. Ethics approvals ran in parallel across the participating sites, and the statistical analysis plan was fixed before the first sample.

What it took

The operator got better mid-study.

Agreement improved measurably over the first weeks at two sites. That is a training effect, and left unaddressed it inflates the performance claim. We had prespecified a learning-curve analysis, so we could show where performance stabilised and base the claim on the stable period, then feed the difference back into the instructions for use and the training material rather than into the marketing number.

The outcome

What came out of it.

480 participants enrolled across 4 sites and three use settings.
Agreement with the reference method inside the pre-specified limits in all three settings.
2 deficiency questions at review, both administrative, none on clinical evidence.
A learning-curve analysis that shaped the instructions for use rather than being discovered at review.