Hi all,
Thank you, as always, for your support.
In Clinical Trials, data is usually locked after cleaning and stored with the analysis results. Is this also common for RWD? Should we extract and freeze the analysis dataset for each study to prevent updates?
I assumed this was necessary, but my colleague thinks it’s unlikely due to the large volume of data.
If not, I worry the data might change after analysis.
Currently, I store each study’s analysis population separately.
I appreciate your thoughts.
Good question. And yes, RWD and clinical trial data are dealt with differently. Locking RWD is almost never done: Their purpose is rarely research, but supporting healthcare of individual patients and paying for that. It’s a rolling corpus of information.
The purpose of locking a database in clinical trials is to prevent tinkering with the data if the result turns out not so good. Therefore, no analysis is allowed before database lock (with some exceptions for predefined interim analyses), and no data must be changed after database lock.
In the real-world, the analysis will change over time. And that’s ok, since the population changes and so do treatment paradigms. In fact, characterizing this change is one of purposes of RWE. But of course it does it so slowly that a day or week or month or even year or two more or less in most cases just changes some insignificant digit behind the decimal point.
Finally, tinkering with RWD is non trivial, if you want to make it look realistic and contain the subtle signals you are after. Ask the people who try to generate synthetic data.
2 Likes
Thank you very much for your detailed explanation. I now have a clear understanding of the differences in handling between clinical trial data and RWD.
The perspective you shared appears to be particularly relevant to large-scale studies. Except for special situations, I intend to avoid data locks in RWD studies. I work at an oncology hospital and am very interested in understanding how real-world changes occur when new treatments are introduced. Conducting OHDSI network studies to capture major trends is extremely valuable.
Thank you once again for your insightful comments!