Dr Karl Kilgore Outlines Challenges of Collecting Real-World Data on CAR T-Cell Therapy

March 18, 2020

Several challenges arise when using real-world data derived from claims to study the impacts of CAR T-cell therapy on Medicare patients, said Karl Kilgore, PhD, senior research analyst at Avalere Health.

Several challenges arise when using real-world data derived from claims to study the impacts of CAR T-cell therapy on Medicare patients, said Karl Kilgore, PhD, senior research analyst at Avalere Health.

Transcript:

What challenges arise when collecting data in the real world on CAR T-cell therapy outcomes, use, toxicities, and other information?

Our data are based on claims period. That's what we use. There's a lot of good things about claims data. In our case we have many databases, in the claims area. We have data for commercial population, for managed Medicaid, for the Medicare Advantage population. We didn't use that for this study. We focused on the Medicare patients specifically. That was one of the initial goals of the study. Future plans, we'll see. But anyway, what we have, this data resource we use, is the Medicare 100% fee for service data. All Medicare patients in our study are Part A and B. We also have Part D data. So drugs in the retail channel, okay. CMS had not released those data at the time that we needed to go to press, but it'll be released very shortly. We will supplement the data then. So the challenges are what you can get out of claims data. It's not medical records. So we may know for example, one of the conditions that excluded patients from the clinical trials was severe cytopenia. So if the patients were severely anemic or severely thrombocytopenic, for example, they would exclude them or delay them from participation in the trial. In our patients, we know over a third of these patients in the 2 weeks prior to CAR T, we know that they had severe enough cytopenia, for someone to have put that diagnosis in the chart. All patients are going to be cytopenic because of the treatments that they get, because of the chemotherapy, because of the lymphodepletion that they get in preparation for getting CAR T. So that's a limitation. Some of our hypotheses are descriptive and tentative. So, in that particular example, in the talk itself, we share the prevalence rates of cytopenia in our sample. We can suggest that some of these patients might have been cytopenic enough to be excluded from the trial, but we can't prove it. We can't demonstrate it. So it's limitations of claims data. There are gaps in the claims data. If a patient, for example, is getting their drugs paid for using a supplemental plan, so we have costs, we have the claims. But if neither the patient nor CMS is paying for this, we may not see the claim for it in the Medicare fee for service database. It's the nature of the claims data. As experts in this area, we have ways of cleaning the data, knowing what we can do valid, and what we can do that is less so, ways of proxying what we don't have in the records. I'll give you an example of that, if you like. We, in our study, we know who expired in the 6 months, we know which patients expired. Another indication that CAR T might have failed or that a treatment might have failed, in our case CAR T, is that they initiate a subsequent line of treatment. So we don't know. Clinicians would love for us to be able to report on relapse, on refraction. Did the treatment not work? We don't know, but we can proxy that and estimate that by looking to see if the patients received a subsequent line of treatment after the particular treatment were most interested in.