How to prove your drug wins when you can’t run a trial for every competitor?

Idorsia, a global biopharmaceutical company, set out to show that competing insomnia treatments underperformed against its own on the outcomes patients report directly. Proving that is normally the job of a double-blind randomised controlled trial. Running one against every rival treatment was financially unviable and left no room to adapt. Idorsia needed credible, comparative evidence, and the standard playbook did not offer a way to get it. 

Challenge: proving superiority with no viable path to the evidence

Patient-reported outcomes are the measures that matter in insomnia. They are also the measures that conventional real-world evidence cannot see. PROMs are rarely captured in structured data, so the usual RWE methods have nothing to compare.

That left Idorsia with two routes, both closed. An RCT for each competing treatment was too expensive and too rigid to justify. Established RWE methodology could not measure the patient-reported effects the comparison depended on. The data existed. The signal Idorsia needed sat in unstructured physician notes, where no standard method could reach it.

Solution: measuring patient outcomes from physicians' free text with AI

Working with Idorsia, we built a new way to define patient-reported outcomes from data that was never structured for the purpose: the free text physicians write in clinical notes.

The system is called DiSMOL. It uses natural language processing to read clinical notes and detect signs of daytime impairment in insomnia patients, the kind of signal that diagnosis codes alone miss. At its core is an abstract ontology that quantifies PROM-like signals from doctors' free text, a step no one had taken before this study. That let us measure treatment effects across different therapies using real-world data, and it opened a class of RWE studies that had been considered impossible.

The methodology and results were later peer-reviewed and published in Nature Communications Medicine, which puts the approach on record as a validated scientific method rather than an internal claim.

Impact: turning unused notes into evidence Idorsia could act on

The work turned unstructured data that had been sitting unused into real-world evidence Idorsia could act on. Three things followed.

It set a precedent. The project introduced a new way of defining PROMs in RWE studies, one the wider field can now build on.

It is reusable. The same method transfers to other therapeutic areas, so the investment reaches beyond a single insomnia programme.

It supported the commercial case. Robust real-world evidence strengthened Idorsia's market position and gave its commercial strategy firmer ground to stand on.

More broadly, the study is a working demonstration that AI-driven RWE can complement conventional clinical trials, and in some cases stand in for them.

‍