Case studies

Problems our clients brought to us, what we did, and what it meant for their submission. Client names are left out.

A questionnaire score that could not detect change

Your questionnaire may not detect change in the treatment group. Often the cause is how its items are weighted.

1
The problem
Our client's questionnaire did not detect change. Every item carried the same weight, so the score could not separate patients well or show when they improved.
2
What we did
We used item response theory to find which items best separated levels of severity.
From this, we derived a new scoring system.
3
The result
With the new score, our client showed regulators that patients in the treatment group improved in the aspects of health that matter to them, compared with the control group.

A treatment can help patients on every measure except the one a wearable records. Further analysis can find the change.

1
The problem
The treatment worked on every measure but one. The raw daily step count for the treatment group showed no overall improvement.
2
What we did
We wrote and ran a revised analysis plan.
It measured each patient's change against their own weekly pattern of activity.
3
The result
Our client showed regulators that the treatment group kept up and improved their activity levels.
This evidence was key to an expanded label claim.

In a rare disease, it is hard to show that a measure works. Some methods still give clear evidence with very few patients.

1
The problem
Our client needed psychometric evidence for a measure in a rare disease.
This type of study would ideally have more than 100 patients. Only 15 were available.
2
What we did
We assessed the measure's psychometric properties at the level of the individual patient.
We then used clear graphs to show the supporting trends and descriptive results.
3
The result
Our client showed regulators that the measure worked as expected.
The graphs told a clear story of the measure and how it tracked change in the disease.