At some point, it would make sense to “mine” the data from the EMR. Again, I am not contesting the claim to “short” the clinical trial process or commenting on its ethics. Lay press is clearly inadequate to layout nuances or highlights the decision making tree without the risk of oversimplification.
EMR’s, however, require a substantial reboot. It is a fantastic idea but only if debate shifts to maximising the user interface for better capture of data and automating repetitive workflows. In my conceptual presentation, I had argued for automating the process by extracting contextual information that would aim to reduce friction and eliminate the possibility of physician burnout.
Still, the moot point: Can data mining eliminate the clinical trials? The surrogates for “survival” are being loosely defined and their boundaries shifted to include “last-time-patient-went-in-for-a-refill”. If it isn’t outrageous, then your senses have numbed by this time.
We need better research in palliative, improve access to healthcare, rather than the quixotic debates around “AI and machine learning” for clinical trials. When something works, it will work in the future as well.
Amgen used real-world data from leukemia patients to serve as a comparison for a small clinical trial in which all patients in the trial received its leukemia drug Blincyto. The FDA last year used the analysis in its decision to approve a new use for Blincyto treating patients who are in remission but have some cancerous cells that put them at risk for relapse.
Amgen would have had to enroll at least 50% more patients in the clinical trial to have a standard control arm, said Elliott Levy, Amgen’s senior vice president of global development.
Such trials “typically involve the potential to be randomized to existing standard of care or even no therapy, when what patients want is the opportunity to be treated with a promising experimental” drug, he said.