Author: Bret Stetka, MD
Results from a new study suggest that a simple blood test can identify patients who are unlikely to respond to TNF inhibitor medications, commonly prescribed treatments for rheumatoid arthritis (RA). Researchers from Scipher Medicine, a biotechnology company based in Waltham, Massachusetts, in collaboration with a Dutch academic research group at the Sint Maartenskliniek, have shown that Scipher’s PrismRA® molecular test can accurately predict which patients with RA are unlikely to respond to the TNF inhibitor drugs adalimumab or etanercept, based on the European League Against Rheumatism (EULAR) response criteria.
Both medications are among the most expensive approved for any condition, and guidance on who should and who should not receive them would go a long way to help curtail healthcare costs and see that patients with RA are prescribed effective, personalized therapies. Study co-author and Principal Scientist at Scipher, Johanna Withers, PhD, talks about the newly released research.
This was work done in partnership with the Sint Maartenskliniek clinic in the Netherlands. They’d been following a cohort of patients with RA who had initiated either adalimumab or etanercept as their first biologic therapy. They were looking at EULAR response outcomes specifically, which are criteria we hadn’t explored yet.
We partnered with them to obtain blood samples and molecular information from these patients in order to understand whether or not PrismRA test results determined before the start of treatment would be predictive of EULAR response outcomes.
Yes. Most of our previous work focused on outcomes commonly used in the US, like ACR50, CDAI or DAS28. This new study focused on EULAR response outcomes, another RA treatment response measure more often used in European studies.
PrismRA is a 23-biomarker test that assesses RNA levels derived from blood. It also integrates information from 4 clinical features to enhance the predictive ability. These are anti-CCP, one of the autoantibodies that is common in RA patients, sex, body mass index and patient global assessment.
The study included 68 patients, who were on average 57 years old and 63% were female. 84% initiated etanercept, and the rest, adalimumab. Predictions of treatment response using PrismRA were made before treatment was started, while their responses – again, based on the EULAR criteria – were assessed at 6 months.
Among the 68 patients in the study, those predicted to be non-responders to TNF inhibitor therapy based on their molecular signature were 4 times less likely to achieve a EULAR good response after 6 months of TNF inhibitor therapy.
These results are on par with the results we see with other clinical outcome measures in RA, like ACR50 and DAS28. Regardless of which measure happens to be a provider’s preferred assessment for their patients with RA, they can confidently use the PrismRA test to inform their treatment decisions.
The treatment response outcomes each incorporate different assessments of a patient’s RA biology and integrate this information in different ways, using different algorithms. Joint counts are a common measure, but some will also use the acute phase reactants, like CRP and ESR, or the patient global assessment. The different response definitions don’t necessarily correlate well with each other, given that they each use a different selection and weighting of clinical assessments.
This means that you can’t assume that just because someone doesn’t achieve an ACR50, that they’re also not going to achieve a EULAR good response. There is some variability as to whether or not a patient achieves one or another of these response outcomes. This study gives us a sense that PrismRA is testing for true non-responders to these two TNF inhibitor therapies, and that the predictions of inadequate response are not biased toward one particular outcome measure.
Yes, it’s called the Monte Carlo simulation. It’s a way of modeling variability in the data. We know that the different clinical assessments that are used in RA like joint counts and patient global assessments all have inherent variability, whether it’s from observer variability, measurement error, or between-subject variability.
The Monte Carlo simulation is a way to model that variability. Using the data we got from each patient, we can model repeated assessments using known inter- and intra-individual variability and probabilities. For each of 2000 such simulations, we calculate response outcomes. This gives us a confidence score as to how true it is that a particular patient did or did not respond to treatment. It accounts for the variability you get from measurement error, which is really important when you’re talking about precision medicine since you want to be confident that a person is a true responder or a true non-responder – and that their outcome is not biased by their measurement being taken on a particular day, or at a particular time. For example, if you measured a patient a day or week later, they could have a slightly different perceived response to their treatment. This simulation addresses this potential confounder. This Monte Carlo simulation approach was used in this study for EULAR response outcomes, but it could be applied to any treatment response outcome in any disease area.
We’re excited about gathering more real-world evidence to show the impact on patients and providers of incorporating PrismRA into guiding prescription choices and care. We’re also investing in our RA program in general, looking at response prediction to other RA drug classes, like the JAK inhibitors. These are two areas we’re very interested in.
