I’ve been working with a client recently who really lifted the lid on just how impossible the expectations can be for COA (clinical outcome assessment) teams. To set the scene: a COA strategy isn’t something you whip up overnight. It can take years to develop properly. Understanding a patient’s condition, mapping out a conceptual model, selecting or building a questionnaire that aligns with that model, or even developing a de novo measure. Testing it in the intended population. And that’s before we even get to the psychometric validation work that we at SeeingTheta focus on.
So when a client tells me they’ve been asked what measures to include in a clinical trial (with the end goal of generating evidence for a label claim, no less) and they’ve been given a timeline of just a few weeks, we all know we’re not just fighting an uphill battle. We’re trying to move the mountain on which the battle is taking place.
A Quick Rewind
At SeeingTheta, we specialise in psychometrics. We also advise on COA strategy for regulatory purposes. And we’re not afraid of a challenge. Some clients come to us when they need outside-the-box thinking. Personally, I’ve never really been in the box. Or near it. And when I do encounter one, I usually end up breaking it. Which, as it turns out, is exactly what this client needed.
When You Don’t Have Time (or Data)
So how do you come up with a COA strategy in a few weeks — with no existing literature base, no prior conceptual model, and no access to patients in that timeframe?
Short answer: you don’t. Not properly, anyway. Instead, you start asking the right questions and work through them one at a time until you arrive at something that’s at least directionally sound.
In this case, the disease area had:
- No existing conceptual model
- No published patient interviews outside of work adjacent to our health-related quality of life (HRQoL) field
- No disease-specific questionnaires
But we didn’t let that stop us. Instead, we asked:
- What does the disease look like? Are there other conditions with a similar aetiology or symptom profile?
- Do conceptual models exist in those other areas?
- Have any prior clinical trials in this or related conditions used COA endpoints?
- Are there generic or symptom-specific instruments that align with what we know about this disease, or that could reasonably be applied based on clinical overlap?
Building from Adjacency
Using these questions as a framework, we explored related disease areas and found a wealth of information. We leaned on clinical descriptions of symptom manifestation in our target condition, mapped overlaps with adjacent diseases, and from this we were able to draft a provisional conceptual model. We layered in considerations like disease trajectory, progression risk factors, and temporal symptom patterns.
With this foundation, we mined ClinicalTrials.gov and trial registries, pulled out previously used COAs, and crosswalked their item content against our model. This led us to select two instruments that, while not disease-specific, covered the core aspects of HRQoL relevant to our target condition.
Is It Enough?
Probably not. But it’s a start. It’s a hypothesis. It’s something to validate with clinical colleagues. It’s something to take to regulators and explain: “Here’s what we’ve done, and here’s why.”
It’s not perfect science. But it’s science. And it’s a lot more justifiable than throwing in the SF-36 and hoping for the best.