The FDA's Decision to Consider Data From a Simplified Pragmatic Trial Design

Host Dr. John Sweetenham and guests Dr. Karen Reckamp and Dr. Harpreet Singh discuss the S2302 Pragmatica-Lung trial, a streamlined, real-world clinical trial that is poised to simplify and transform the entire clinical trials model as we know it.

TRANSCRIPT

Dr. John Sweetenham: Hello, I'm Dr. John Sweetenham, the associate director for cancer network clinical affairs at UT Southwestern's Harold C. Simmons Comprehensive Cancer Center and host of the ASCO Daily News Podcast.

Today, we are going to be discussing a streamlined, real-world clinical trial from the Southwest Oncology Group (SWOG), which is S2302, also known as the Pragmatica-Lung trial. This study is poised to simplify and transform the entire clinical trials model as we know it.

Joining me for this discussion is the trial's lead investigator, Dr. Karen Reckamp. Dr. Reckamp is a clinical professor, director of the Division of Medical Oncology, and associate director for clinical research at Cedars-Sinai Samuel Oschin Cancer Center. I'm also delighted to welcome Dr. Harpreet Singh, the director of 1 of 3 divisions of oncology at the U.S. Food and Drug Administration. She will discuss the FDA's views on streamlining clinical trials to reach more representative groups of patients and will also more broadly address some of the key questions that regulators consider when deciding on whether real-world data can substitute for randomized controlled trials.

Our full disclosures are available in the transcript of this episode, and disclosures relating to all episodes of the podcast are available on our transcripts at asco.org/DNpod.

Dr. Reckamp and Dr. Singh, it's a great pleasure to have you on the podcast today.

Dr. Karen Reckamp: Thank you for having us.

Dr. John Sweetenham: Dr. Reckamp, I'm going to start with you if I may and ask if you could give us some background on S2302, the Pragmatica-Lung trial for non-small cell lung cancer.

Dr. Karen Reckamp: Sure. The Pragmatica-Lung trial really started with the sub-study from Lung-MAP, which was called S1800A, and it was a randomized phase 2 trial that evaluated pembrolizumab and ramucirumab versus standard of care for patients who had previously received chemotherapy and immunotherapy with advanced non-small cell lung cancer and had had disease progression. And in this phase 2 trial, we found that there was an improvement in overall survival with a hazard ratio of 0.69% and survival of 14.5 months for pembrolizumab and ramucirumab, and 11.6 months for standard of care. And with that, we had again a randomized phase 2 trial, but the study was small. And so, trying to think about how to move this to the next level to get phase 3 data, we started thinking about how to do this in a way that would reduce the timelines and potentially move this treatment to patients more quickly than standard registrational, randomized phase 3 trials. And that's kind of where S2302 and Pragmatica-Lung started.

Dr. John Sweetenham: So, can you tell us in a little more detail how the dramatically streamlined pragmatic design of this trial is going to hopefully simplify trial design and trial conduct in the future beyond non-small cell lung cancer?

Dr. Karen Reckamp: I think the important piece of this—and we have Dr. Singh here to speak to the FDA part—but this has been a partnership with the FDA and CTEP [NCI's Cancer Therapy Evaluation Program]. Really, our goal was to try and find a way to run trials in a more streamlined way. One of our colleagues at CTEP during this process said, "If this is not making you uncomfortable, then you're not doing it right." So, the first thing we did was kind of lean into the discomfort because for those of us who have been writing trials and putting trials together for the last 20-plus years, this is dramatically different. And we're really looking at one question, and that is overall survival. We're trying to validate the overall survival we saw in S1800A. And with that, we stripped away all the unnecessary data collection that comes along with other types of registrational issues that come with randomized phase 3 trials. And then, we also looked at patient burden and really opened it up. So, again, pragmatically, making this practical, allowing investigators to be empowered to treat patients how they normally would in their own practice. And so, moving forward, again, for types of trials where we have drugs whose toxicity profiles are well known, they're used in practice but using novel combinations for a subset of patients who have limited treatment options available, this could really change the paradigm moving forward for these types of trials in multiple diseases.

Dr. John Sweetenham: Yeah, thanks. And in addition to simplifying the trial design, obviously, one of the goals here is to have a study population which is more representative of the patients who are seen typically in community practice. Hopefully, we'll overcome some of the known disparities that we see in clinical trial accrual. Could you speak just a little bit to how the study design and the organization of the study helps to achieve that?

Dr. Karen Reckamp: So, I think it's on several levels, but we are looking to allow this to be more generalizable and allow a more diverse population into this study. First, by again stripping down the eligibility criteria to only the absolute essential criteria for understanding our scientific question. And so, we don't require imaging studies to be uploaded or presented. If the patient has progression, it's based on the investigator's opinion. And so, we don't need to be searching for outside scans or things like that. We don't have tissue requirements, and we don't even actually have lab requirements. If this is a patient, you would treat your standard of care practice with the standard of care regimens; those are the labs that you do. So, it's all based on standard of care. So, by doing this based on standard of care, it really allows almost any patient to enroll. And then we have outreach. We have our DEI group and our community practices very well engaged to make sure that we have broad reach. Having this open through NCTN [NCI's National Clinical Trials Network] will make sure that we get this to multiple practices in far-reaching parts across the United States.

Dr. John Sweetenham: Yeah, I think that's excellent. And you've already alluded to the fact that the data collection requirements for the study are going to be kind of pared down to the absolute minimum and that's going to include, I believe, toxicity reporting as well. So, can you comment a little bit on that and, specifically, what your plans are for reporting toxicity in this trial?

Dr. Karen Reckamp: Yes, you're correct. This is significantly pared down from what we're used to doing. And so, most clinical trial offices are struggling with staffing and making sure that their patients have enough staff and that they have enough staff to get patients onto trials efficiently, and then getting the data in is always a challenge for sites. So, we have really, again, working with our partners, working with the FDA, and with CTEP, we have minimized what we are going to collect on patients. So, we're collecting survival and vital status on patients. We are collecting the background standard information that we collect on kind of prior therapies, and we are collecting only unexpected grade 3 and higher adverse events.

And so, thinking about these drugs—ramucirumab and pembrolizumab—we know how these drugs work, we know the toxicity profile, we're using them in combinations and single agents in multiple tumor types. And so, thinking about most of the immune-related adverse events wouldn't even be reportable because they're expected. And so, a large number of data that is normally collected would not be collected here. And, as noted, we don't collect scans, we don't collect labs, we're not collecting con-meds, start and stop dates. A lot of that burden of data collection, but also data auditing and queries, goes away. It should be a significantly easier trial to perform by sites.

Dr. John Sweetenham: And can you just update us on the status of the trial right now?

Dr. Karen Reckamp: We're in the process of pre-activation, and so, if you're an NCTN site, you can actually go in and do some pre-training and take a look at the draft protocol. And we are anticipating approval sometime in early March.

Dr. John Sweetenham: Great. Congratulations on getting this trial launched and underway because I know that the word "groundbreaking" is used a lot, but I think that, obviously, if this trial proves to be the success that it looks like it will be, then it's going to have, I think, major implications for study design in the future.

And that's going to lead me to ask a couple of questions to Dr. Singh. And the first one of those is the FDA's decision to consider data from a simplified pragmatic trial design like this, which uses more limited clinical information, is really kind of almost revolutionary. And could you comment a little on this from the FDA perspective and how you think it's going to influence the future of clinical trials and the future of cross-trials?

Dr. Harpreet Singh: Well, thanks so much for the question, and thanks for having me. I want to push back on that just a little bit because I think, for what it's worth, the FDA has been advocating for trial efficiencies in oncology for many years. And, of course, as you know, our current commissioner, Dr. Robert Califf, is very vested in this concept. And certainly, the idea of pragmatic trial has been there in the field of cardiology for some time.

In terms of this trial, in particular, in coming to oncology, I do think actually putting pen to paper and drafting the protocol, which we did really in cohesion with SWOG and many calls with Karen and others who were a part of this, that collaborative piece, I think, is groundbreaking because what we saw here was a great deal of discomfort, actually around everything that we were stripping down. We sensed a lot of discomfort in terms of including various, like you mentioned, safety issues, safety reporting not being perhaps as rigorous as we're accustomed to seeing at the FDA. And certainly, investigators are accustomed to collecting other endpoints besides overall survival, like time to progression, but the real-world version of that, or time to next therapy.

And so, one very difficult lesson that I've had to learn, and we've had to learn, is that you have to really learn to say no to some very interesting trial design elements that are not essential to the big question here, which is, does this combination regimen offer an improvement in survival over the control? So, while I do think the actual organizational and structural piece of this, now that it's actually stood up, is groundbreaking, I think that the idea of pragmatic trials and incorporating clinical care into the idea of answering a clinical question as opposed to the traditional randomized clinical trial is a concept that's been around. I'm just thrilled to see it actually occur in this very, I think, ideal setting for patients with lung cancer.

Dr. John Sweetenham: Yeah, absolutely. I want to broaden the scope of what we're discussing here just a little bit, perhaps to talk a little more about the "real world" and "real-world data." More real-world data is being considered in regulatory decision-making. And one of the questions I have, again, from an FDA perspective, is that everyone still, I think, regards randomized controlled trials as the gold standard for evaluating efficacy if not effectiveness, of various interventions. What are the key questions that you consider when deciding whether any kind of real-world data analysis is a good substitute for a randomized controlled trial?

Dr. Harpreet Singh: Well, thank you for the question about how FDA considers real-world data when we consider this to be appropriate. There are many nuances to this. So, first of all, what is real-world data? And there's actually a distinction between real-world data, which is just simply a source used in observational studies traditionally. But real-world data is not specifically a trial design; it's just data. Whereas real-world evidence, which is evaluating the benefits and risks which are derived from real-world data, may come from things like electronic health records. It's not either-or.

So, for example, in a pragmatic trial, you could use a blended approach where you have some components of real-world data or real-world practice, which we may consider kind of part of real-world data, but while retaining some elements of randomized control trials. So, I think when FDA considers real-world evidence, so I'll say that instead of data, it usually would be a source like a very high-quality registry or data obtained through a very well-designed observational study. And this would be in settings of perhaps super rare diseases in which randomization is either not feasible or, in some cases, where you may have preliminary data which suggests that randomization is not ethical. But we agree with the general idea that the gold standard is randomization.

And that's what I love about this pragmatic trial, is that you are retaining the benefit of randomization while bringing pragmatic elements in, bringing the trial to patients and really incorporating clinical practice into the trial, as opposed to the reverse, where you're having patients enrolled on a traditional trial where the visits are outside of routine.

Dr. John Sweetenham: Thanks for drawing that distinction between real-world evidence and real-world data because I think the two expressions are sometimes used a little carelessly, as maybe I just did. But certainly, one of the things that I've observed over the last several years since we started to incorporate real-world data or real-world evidence into our kind of oncology lexicon is that real-world data has been used in a fairly relaxed, let's say, way and certainly any relatively small series which has been registry based or retrospective, there's been a tendency to use this term called real-world data, which personally, I've certainly seen applied to patients who are undergoing very intensive therapies such as CAR T.

And certainly, when I look at the patient characteristics in those elements of so-called real-world data, it's a long way from the real world that I'm familiar with in my own practice. And so, I do think that the term has been used very loosely. And your point about real-world evidence is an important one, I think. People are still questioning whether real-world evidence in oncology is truly valid. And I think to some extent, you've already answered that question.

Do you think that there are mistakes and pitfalls that investigators can avoid when they're looking at real-world evidence?

Dr. Karen Reckamp: Sure. I mean, I think the first point of clarification is, are we looking at this evidence to support use of an oncology drug in clinical practice, or are you an investigator working to bring real-world evidence to the FDA for drug approval? But either way, no matter what scenario you're in, I think the first question you must ask yourself is, is this data fit for purpose? And what does that actually mean, 'fit for purpose'? And I think it goes to things like, are the patients well-matched? So, there's this very complex process, but the concept is not complex of propensity score matching, which our statisticians do for us beautifully. But this idea of are the patients in this data set that you're looking at, this is just a collection of data, right? How relevant is it to the patient in front of you? Is there some sort of matching that's going on in terms of patient characteristic?

After that, you have to ask yourself about this kind of array of epidemiologic biases that are inherent in non-randomized comparisons. Like, is this contemporaneous data? So, if this data set came from a group of patients who started their therapy—this goes to the idea of the index date, okay, start of therapy—has the standard of care changed? Has supportive care become increasingly better? Obviously, the answer to that is yes. And so, if you have these contemporaneous mismatches, then can you actually really rely on this real-world data or evidence, either one, as you're applying it to your patient? So, I think if it's for regulatory purposes, certainly you could avoid many mistakes by coming to the FDA early and often, which we always recommend. And if it's you as a clinician, as a health care provider, looking at this collection of data, I think you do have to walk yourself through in a really basic kind of logical process of, "how well does this data apply to the scenario in which I want to use this therapy?" So, index date, selection, timing, patient characteristics, things like that.

Dr. John Sweetenham: Yeah, great, thanks. And I'd like to maybe ask both of you for your comments on one final question, and this is circling back to the S2302 study. Intrinsic to the study design and the concept is that the population in this study will be truly representative of the "real world." My two questions to both of you will be, first of all: What is the gold standard for representative? In other words, what does that really mean to have a representative population of patients with advanced non-small cell lung cancer? And secondly, do you have any safeguards in place in the course of the study designed to make sure that that study population doesn't get skewed in some way so that it becomes unrepresentative of the real world? So, Dr. Reckamp, maybe I can start with you and ask you for your comments about that.

Dr. Karen Reckamp: Thank you. I think that's a very good question and something that we grappled with as we designed this study and really did keep coming back to that. So, I think when we talk about representation, most randomized trials don't have broad representation. They are very specific populations that we curate in order to take as much variability out of the trial as possible so that we can investigate just the experimental arm versus standard of care or whatever we're evaluating. And here, we've consciously made an effort to say we want to know how this works in a real practice and make this as generalizable as possible while still being safe. So, we have the premise of keeping patients safe as the number one goal of this trial. And then, we want to look at the survival data.

So, we actually did lower the bar a bit and changed our hazard ratio. Our hazard ratio was 0.69% for the phase 2 trial. We loosened that a little bit for the phase 3 trial, knowing that the patients that are coming on to trial are not going to be perfect patients, and there may be a little more coming together of those curves. That being said, randomization is what is supposed to wash away all sins, which has been said many times as we put this trial together. And so, the randomization is really the goal, to utilize the randomization process in order to make sure that there is balance and that we are getting representation on both sides that will help us understand how the investigational arm is really doing in this population. It's not going to be perfect, and we are allowing for performance status 2 patients. But I think we all believe that this is really important because there's a large proportion of patients who have performance status 2 who never go on to trials, but in the real world, we treat them generally with the standard of care options that we use. So, I think this is really important for moving things forward, and will be groundbreaking in that way, too.

Dr. John Sweetenham: Great, thank you. And Dr. Singh, just to add to that, will the FDA be looking at this from the perspective of making sure that the study population as a whole—accepting that randomization will hopefully cancel out some of the potential pitfalls there—but will the FDA be looking to make sure that the population as a whole is truly representative of what's out there in the real world?

Dr. Harpreet Singh: We always look at the population. We are always hopeful that, in general, the population is reflective of the disease for which—in this case, lung cancer. I think in this case, we were very hands-on with developing the protocol, and it is our hope and it's our expectation, and I think it very much will happen that you are going to see a very diverse and representative, more generalizable population here.

I just want to add a piece to this because remember that traditional randomized clinical trials typically do have a more homogeneous patient population because a lot of this is designed around a de-risking strategy when you're bringing new drugs to market. One of the reasons we felt so comfortable stripping away, as Karen mentioned, no lab criteria. If the clinician says, "I think you're fit for this regimen, go ahead and enroll them." We pushed for inclusion of PS 2 patients. We, the FDA, did. So, yes, we're going to be looking, but we do really hope that these really streamlined inclusion and exclusion criteria allow for that. And so there's other things too, like race, ethnicity, age. And so it starts with not excluding patients based on perhaps unfair or arbitrary cutoffs like labs. Not to say that performance status is arbitrary. But in this case, if the clinician deems you fit for this therapy, that is between the patient and the investigator and their judgment, which is really part of the element of real-world trials and this pragmatic element too.

I also wanted to add on this idea of diverse representation, we expect there to be a lot of extra, for lack of a better term, noise, in this trial, even though it's randomized. And so, part of the negotiation around designing this trial was the need for an increased sample size to try to account for some of what we expect to be perhaps unequal randomization, perhaps in terms of patient characteristics on either side, perhaps patients lost to follow-up, etc. And so, when we talk about pragmatic trials, one element is that you probably often may need an increased sample size to account for the increase in heterogeneity, not only in your patient population but perhaps in monitoring as well.

Dr. John Sweetenham: Well, thank you both, Dr. Reckamp and Dr. Singh, for a great discussion today and for sharing your insights on these developing trends in clinical trial design. Dr. Reckamp and Dr. Singh, we'll be watching closely to see how the trial performs in the coming months and advances the concepts of pragmatic trial design that Dr. Singh mentioned earlier within the FDA. We obviously are very excited to see whether this change in trial conduct will enable you to meet new groups of patients and ultimately improve outcomes for them. So, thanks once again for being with us today.

Dr. Harpreet Singh: Thanks so much.

Dr. John Sweetenham: And thank you to our listeners for your time today. If you value the insights that you hear on the ASCO Daily News Podcast, please take a moment to rate, review, and subscribe wherever you get your podcasts.

Disclaimer:

The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions. Guests on this podcast express their own opinions, experience, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.

Find out more about today's speakers:

Dr. John Sweetenham

Dr. Karen Reckamp

@ReckampK

Dr. Harpreet Singh

@harpreet_md

Follow ASCO on social media:

@ASCO on Twitter

ASCO on Facebook

ASCO on LinkedIn

Disclosures:

Dr. John Sweetenham:

Consulting or Advisory Role: EMA Wellness

Dr. Karen Reckamp:

Consulting/Advisory Role: Amgen, Takeda, AstraZeneca, Seattle Genetics, Genentech, Blueprint Medicines, Daiichi Sankyo/Lilly, EMD Serono, Janssen Oncology, Merck KGaA, GlaxoSmithKline, Mirati Therapeutics

Research Funding (Institution): Genentech/Roche, Janssen Oncology, Calithera Biosciences, Elevation Oncology, Daiichi Sankyo/AstraZeneca, Blueprint Medicines

Dr. Harpreet Singh:

None Disclosed