Myriad Genetics issued an interest 8-I (impactful news update) that it was discontinuing a planned analysis of GeneSight pharmacogenetic clinical data. The data design involved using an administratively constructed comparator arm (aka "control" arm) with claims data.
They write, the Company directed the third-party vendor to review its data warehouse to identify a group of patients that best matched the characteristics of the GeneSight patient group.... the Company concluded that any data comparing the GeneSight patient group against this non-GeneSight control group were not reliable. Therefore, the Company has determined to discontinue the analysis for this part of the study.
SEC notice here.
See also a 2024 Am J Psychiat publication by Baum et al., The addition of these new data do not alter the recommendations of the 2018 report, or the advice of the U.S. Food and Drug Administration, that the evidence does not support the use of currently available combinatorial PGx tools for treatment selection in major depressive disorder. It updates the "American Psychiatric Association Council of Research Workgroup on Biomarkers" position.
MYGN has a share price dip + recovery:
###
Don't Over-hype "Administrative Virtual Control Arms"
The issue is the limits of using administratively constructed databases as a comparator arm as opposed to using actual tracked patients (randomized) as a comparator arm. I have heard senior CMS staff in the past year tout the potential expansion of "Coverage with Evidence Development" in non randomized trials due to use of administrative control arms, and how that science had developed so much.
Well....
My impression it is still very difficult to rely on administratively implied or virtual control groups, especially when outcome differences are relatively small. The CMS YMCA pre-diabetes intervention study (published in Health Affairs, not the most rigorous scientific/economic journal) used an administratively constructed comparator group and got outlandishly favorable outcomes, even if a prediabetes patient had a single brief YMCA session. I would have concluded that if that was the case, the YMCA treatment group (people showing up at gyms!!!) was probably simply different than the unobserved, imputed control group (people the same age but sitting in their cardiologist's office).
##
See also a May 2024 article in JAMA on inferring causality from observational trials; Dahabreh et al.
