Episode Transcript
Hello everyone. Welcome to the next episode of TransPerfect LifeSciTalks. I'm Mark Wade. Today I'm joined by Bill Byrom, who is VP of Product Intelligence and Positioning at Signant Health. Welcome Bill.
Yeah, great to be here. Mark. Good to see you again.
Bill and I know each other a long time. Bill has his Ph.D. From Strathclyde and was in Nottingham undergrad, so we know each other a long time. So I'm really glad that you came on for our LifeSci Talks and specifically today I wanted to talk about oncology and capturing data in that patient cohort.
So, what is the historic experience of capturing data from those type of patients?
Yeah, I think if we look at how we've done things traditionally and probably how we're still doing things mainly now, there's a few things. I think one is, we use various quality-of-life instruments to measure things like physical function and other aspects of quality of life. And we usually do these when patients come in for a visit just before they get the next cycle of treatment. So, it's usually site-based administration. They'll sit down, complete their questionnaires, they'll either do it on paper or they'll do it electronically on a tablet. And, there's a question mark really over whether that's really the right time to be doing this. You know, this is when patients are feeling well enough to get the next cycle of treatment.
So you're asking them stuff about the effects of treatment and the toxic effects that they've been suffering, but they're actually, at this point in time, they're feeling better than they have done for the last few weeks. So, it's not necessarily the optimal time to do it. So I think that's one interesting thing about how we do things now. And if we think about why that's a problem, when we think about recall, Mark, we think about when's the right time to be asking patients a question about how they've been feeling over the last few days or week or whatever. And, there's two elements of bias that can creep in when we think about recall. One is can I actually physically remember something? You asked me what I ate this time last week, I can't remember that.
I'm never gonna remember that. Ask me what I ate yesterday. Yes. I can tell you that. So that's kind of one aspect, just not being able to remember, that's probably less relevant here. But the one that's possibly more relevant is what we call response shift. And that's a patient- Because their condition has changed because they're actually feeling better, does that color the way that they perceive that they felt three weeks ago? And, we call that response shift, and that's a bias that we're kind of quite concerned about when we're just measuring at these points, when actually it's the time when the patient's feeling at their best.
Do you remember years ago we used to talk about the "parking lot syndrome?" And in many ways, it is like that, with this revisionist history. Here, the disease state is actually coloring their recall. But, it's the exact same thing. I think. Can I unpack that slightly? Because when you look about the instruments, when you talk about the instruments that we use, how specialized are they? I mean, are they designed for purpose?
Well, I think there's been some good experience with some of the instruments, and I think this perhaps leads us into understanding how they're used and what they're used for in regulating decision making actually. And, if we think about- Ari Gnanasakthy has published a couple of papers, two or three papers where there's been a review of oncology drug approvals, and he looked at the labeling claims that those drugs have put out on their labels and seen how many of those have got ones based on COA data. And, it's not a great picture. And particularly for the US approvals, it's not a great picture. So I think 2010 to 2014, there were 3/40 drugs that got approved by FDA for oncology that actually had PRO data in the label claims.
And then he looked again, 2012 to 2016, and found no drugs at all out of 45 approvals. Whereas, on the European side, it's a little bit of a better picture. About half, 21/45 approvals, in that same timeframe had label claims based on PRO data. And I think that starts to illustrate some of the concerns that FDA have around how good these instruments are and how able they are to measure the things that FDA are specifically interested in when they want to think about the related labeling.
That's a good point. And the FDA draft guidance, I believe you published a paper on this, but the FDA draft guidance, does that go a long way to speak to this?
Yeah, absolutely. I mean, this has been bubbling on for quite a long time. The publication of this guidance, or at least the draft guidance is really quite significant. I published a summary of what it means for us as a little white paper that Signant health published recently. But, I think there are three things to that guidance. The first thing is around the specificity of the measures and making sure that they are able to measure specifically a number of different domains that FDA are interested in. So we talk about those in a minute, but the other aspects of the guidance are around the frequency of assessments and what we talked about before, this idea of just before a cycle starts, when the patient's feeling at their best, that isn't the right time to be asking questions about the toxic effects of the drugs, etc.
And, they've defined a more frequent approach to measuring certain domains like physical function or a symptomatic side effects. And they want to see those measured more frequently, not just at the clinic visits. And then the third part is the importance of collecting a reason for missing data. And, they're very worried about the bias that can be introduced by patients not completing assessments. And maybe they're not completing them because they're just physically not well enough to. And, then how can you deal with that in an analysis? You know, you've got to make sure, actually, if I just treat it as missing at random, I'm gonna, I'm gonna bias that estimate, because actually they were feeling really sick at that point. Right. So, those are three things they really focused on. And that specificity thing gets to the point that you were talking about, which is, are the instruments fit for purpose at this point? And, I think some of them are, but some of them aren't.
I agree. There's a couple of things in there. I mean, the whole concept of the patients being too sick to complete these instruments, I think that's terribly important. And, if they're doing it at home and, maybe the caregiver or whoever is minding them isn't able to help them, are they too sick to complete it? If they're not, well, what do we do with the lack of data? The lack of data is just as useful as the data itself, isn't it? In many ways.
I mean that can be quite informative, can't it? So I think that it starts to introduce a challenge and one of the things that FDA have have asked for is this more frequent measurement. But not for everything. So, thinking about the practicalities of a patient completing a QLQC30 and then a disease-specific questionnaire on top of that, and then maybe an EQ5D and doing all of that at home in one sitting when they're not feeling well is certainly not very practical. But, I don't think that's what they're asking for. If you kind of look at what they're saying, they split down the areas that they're interested into five different domains.
So they've got disease related symptoms, symptomatic AEs, and overall side effect impact measure, and then a measure of physical function and a measure of role function. And it's only a couple of those domains that they actually think you should be measuring every week, at least for the first few weeks of treatment. You know, maybe the first two or three months. And that would be the adverse events and the physical function. The other aspects which are more sort of generic role function, disease-related symptoms, health related quality of life, they can be less frequent. And so, I think what we're kind of getting towards is, is there a happy medium?
That's exactly what I was gonna say!
Yeah. Ask a few questions to cover these particular domains and be able to ask them more frequently, because we're not overburdening. So, hopefully we get to a point where we are collecting enough data, and we're able to address the things that the FDA are asking for.
That was exactly my point. I'm always concerned that the burden is too great on the patient, and especially in this cohort. Burden is a different animal, isn't it? So that that's the fine line. That's a balancing act, isn't it?
Yeah.
One of the things that I did want to ask you about is item banks, right?
Okay.
To the ignorant of us, walk us through this idea that, that you can pick and choose and put an instrument together, because there's so much work involved. Can you speak to that? I'm just genuinely curious.
I think maybe just backing up a minute. So there's probably two places where I think item banks are really interesting. So one is in the rapid assembly of a new instrument. So maybe we want to develop an instrument for a cancer that we haven't got a standard instrument for. And so it enables us, perhaps, to do that a little more rapidly because all the questions, all the items in the item back are already validated. We know that patients cancer patients understand and interpret those questions in the way intended. So some of that work has already been done. I think the other area where an item bank can be valuable is in addressing this specificity question. So, where we've got instruments that perhaps aren't as specific, aren't measuring those specific domains that the FDA are asking for, it might be that supplementing with a number of items from an item bank might get us to that place much quicker.
And so, we think about the EORTC, and as we think about cancer, we've got the promised item, but we also got the EORTC one and, Ithink EORTC has maybe a thousand items now. And these are all items that have been taken from their existing validated questionnaires or have been generated subsequently to fill gaps that they think they need. And I think the way that they can be used is exactly as we said just to rapidly assemble new questionnaires or new subscales or to supplement existing instruments.
Because the overlap is always an issue for me.
That's right. And, think when we look at something like, if we're using the QRQC30 and then we're using a BR23, is it the breast cancer or wanting to go on top of that, is there some overlap between those two or probably not too much because they're designed to go together. But if we're using that with a Fact G or with something else, there is significant overlap in these scales, and that's a real problem when we want to think about patient burden and wanna make sure that we're just asking them enough questions to measure the things that we need to measure. Because these are people who are sick, they're sick patients, and they don't want to be overburdened with this.
Can the item bank speak to that entirely? In other words, can you, for the specificity, can you choose those items, do the work to make sure that it doesn't increase a bias, all that good stuff that you would normally do? Can you use that on its own? Can that stand on its own?
I think so. What I've seen done, and there's been some really good work done, I think Jill Bell published with her colleagues a really nice example of how to use the EORTC item bank. So, they were looking at a number of fairly rare cancers for which there weren't any specific questionnaires developed. And, they went off and did some additional concept elicitation work in patients to find out what were the most meaningful symptoms, etc. That they were experiencing. And, then they did a mapping exercise. They went and looked at the QLQC30. Well, actually a lot of these mapped to that instrument already, so we have those covered, but there's a few here, there's a dozen or so of symptoms which aren't covered by the QLQC30, but that are covered by the wider EORTC item bank. And, they could pull those in and then suddenly they've got an instrument that's now specific and tailored to that particular set of set of cancers.
To be honest, that's the real power of this, isn't it? I didn't I actually didn't appreciate the power of that.
Yeah. I think it gives us a headstart. I mean, we've still got to do work. We've still got to do the concept elicitation. We've got to prove the content validity of what we've put together. And there's probably some work we've got to do at the backend around scoring and some psychometrics and things, but the actual item development piece has already been done if we're able to just pull from the library. And that's quite a big chunk of work that's already done and ready for us to use. So, I think it's a nice illustration I think of how to use this and also how to develop something that's specific to back to the FDA guidance that's specific enough to meet their requirements, and do it in quite a nice, simple manner like that.
Well, we'll have to sit and wait for the final guidance from the FDA. Yeah. That's great. This is Bill, thank you so much for today. I appreciate that. Bill Byrom is VP of Product Intelligence and Positioning at Signant Health. I'm Mark Wade. Thank you for joining us for LifeSci Talks. Thanks.
Thank you, Mark.
