OHDSI Home | Forums | Wiki | Github

Forming external controls for comparative effectiveness estimation

(Seamus) #1

Hi All,

An important potential use of observational data in oncology for both regulatory and HTA decision making is in forming external control groups. With IPD from a trial (usually single arm) and external controls matching methods can be used to estimate comparative effectiveness.

In many ways, the OHDSI/EHDEN data networks offer a great opportunity to support this by helping with data identification and access.

However, a federated data network also acts against this because data sets (trial and observational) cannot be combined (unless both are in possession of those undertaking analysis).

I would be interested in hearing people’s perspective on the value of the OHDSI/EHDEN networks for estimating comparative effectiveness using external controls. Is this something anyone has given any thought to?


(Christian Reich) #2


This keeps popping up all the time. And the OHDSI approach can make it easy to define the comparator arm in a systematic way, so it can be executed at different sites.

But the underlying issues remains: Can you really match patients:

  • Defined using inclusion criteria in different settings (clinical trial criteria vs criteria in observational data)
  • With missing information
  • Across different sites
  • Without randomization

The answer probably is: In cancer the effect sizes are so big (particularly for toxicity) that you may not care. But subtle relative risks you almost certainly would not be able to tease out.

(Seamus) #3

Thanks for the response @Christian_Reich!