There was no endpoint to measure, and almost nobody to measure it in.

A client wanted European approval for an ultra-rare neurodegenerative disease off a single pivotal study. No validated endpoints, no natural-history data, a handful of patients scattered across the continent.

In a pool that small the risk isn't finding patients. It's whether every one you find reaches the end producing clean, evaluable data. So the step most teams tick off as routine, clinician and patient training, was what the approval rested on. With the sponsor, we ran rehearsed data fire drills across a dozen specialized centers in several countries: clinicians, patients and families practicing exactly how each observation was captured and recorded. Treated as part of the team, they stayed, and measurement held. A scattered cohort came out as full evaluable and intent-to-treat populations.

None of that is clever. It's training, taken seriously.