Polygenic Risk Scores Move Closer to IPF Clinical Use: David Zhang, MD

Higher polygenic risk scores are associated with an increased risk in idiopathic pulmonary fibrosis (IPF), according to a recent study using real-world electronic health records (EHR) and biobank data.

The study was reviewed by David Zhang, MD, a pulmonary disease medicine specialist at Columbia University Irving Medical Center, who further explained the practical implications of these findings in an interview with The American Journal of Managed Care^®.

In this Q&A, Zhang highlights how polygenic risk scores could serve to improve diagnostic accuracy and aid clinicians in identifying IPF disease risk.

This transcript was lightly edited for clarity.

AJMC: What new insight does this study add beyond earlier discoveries, and how has your own research examining rare genetic variants, polygenic risk, and telomere biology interactions that may influence pulmonary fibrosis risk complemented those findings?

Zhang: I think it's very exciting to see some of the new findings from this study, which focused more on polygenic risk and common variation but applied it in a way that is more practical and takes a step forward toward what implementation of these risk tools might look like in a real-life setting.

A lot of what we looked at in previous studies was establishing the importance and contributions of various genetic risk factors. The next step after finding those discoveries is figuring out how they would apply in real life and what challenges might come up when applying them.

Like anything, it's one thing to find something; it's another thing to put that thing to use. This paper has done an excellent job taking that next step and seeing how it applies in a real-world setting.

AJMC: One of the most interesting findings was that higher polygenic risk scores were associated not only with an IPF diagnosis but also with transplant-free survival. How should clinicians interpret that result, and do you think the score is identifying patients with more aggressive disease or primarily identifying patients who truly have IPF rather than another interstitial lung disease?

Zhang: The score was associated with more specific IPF diagnoses and, at the same time, with outcomes, especially transplant-free survival. But in the real-world data, having a more specific IPF diagnosis itself is also associated with worse transplant-free survival.

It's important to be careful when interpreting that result because one plausible explanation is that the polygenic score is doing a better job picking out IPF cases that are more similar to the curated clinical cases from which the score was derived. These clinically curated cases are those that expert clinicians have carefully examined and deemed to truly be IPF.

I think it's an equally plausible explanation that the score is better at picking up classic IPF cases or even cases that are actually diagnosed with IPF.

There were previous studies looking at a similar score, although not the exact same polygenic score, within a carefully curated IPF cohort and found the opposite effect. That's why this question is still somewhat up for debate.

In the other paper, there was a dichotomy between IPF patients who had rare disease-associated mutations having worse outcomes and, on the other side, patients with a greater polygenic risk score having a comparatively more favorable outcome. It's hard to reconcile those 2 results without considering the possibility that the score is picking up more classic IPF cases.

AJMC: In your editorial, you write that this study offers a “sneak peek” at the promise of polygenic risk assessment in IPF. What aspects of the findings make you optimistic that we’re moving closer to clinical implementation?

Zhang: The study is powerful because it looked at multiple different real-world EHR databases. What we've seen previously with earlier studies in the UK Biobank and All of Us datasets, we're now seeing time and time again that there is value across multiple large EHR databases.

This replication alone is something that is commendable, and it was a tremendous effort that the authors were able to undertake. I think it means that this tool has value in real-world data.

I think what the study is highlighting now are some of the more nuanced challenges that remain to take that risk prediction and refine it in a way that is clinically meaningful. The bigger question now is how this will matter to a clinician considering an IPF diagnosis, or even a clinician who has an IPF patient and is trying to understand how the score informs prognosis.

I think now we're getting into the nitty-gritty of it. We've gone past the premise that this works in carefully curated clinical databases and that this has a signal in real-world databases. Now we're really talking about the finer details, which is exciting to me because this is another step closer before it can actually make its way into clinical care.

I think we touched on the main value of the study, and again, it should not be understated how much effort it was to do this work across multiple independent large EHR data sets. I think that's the most impressive part for me—the ability to show this across so many different datasets and find largely consistent signals.

At the same time, all the limitations that I mentioned, I view those less as limitations and more as opportunities to refine them. The paper has really highlighted the remaining work.

AJMC: Although the study reported good discrimination, the positive predictive value remained relatively low because IPF is a rare disease. What does that tell us about where polygenic risk scores may—or may not—fit into routine care?

Zhang: That's the big challenge. The context of how the score is used, like many other genetic variables, is going to be very important.

In real-world clinical practice, we are not envisioning that the score would be applied to a general unselected population. Like any other test, there are positive predictive value, negative predictive value, and other performance characteristics that have to be taken into consideration when deciding when and how to use that test and for what purpose.

A lot of these tests are used for rare diseases, and deciding what context to use them in is going to be quite important.

To me, it seems like this is a test that would supplement and definitely not replace—and that's not something anyone is suggesting—but rather supplement clinical assessment. For somebody who has a suspected IPF diagnosis, I think the score could have a lot of extra value, maybe for determining how classic the patient is compared with a classic IPF case, or even for considering prognosis when IPF is in the differential.

Again, this study took it to the next step, which is to say that across multiple real-world datasets, this tool has value. Now, getting into the nitty-gritty of which specific settings this should be deployed in becomes the next question.