Designing for Approval from Day One: Q&A with Mwango Kashoki, MD, MPH, Parexel

The FDA's plausible mechanism framework is reshaping how sponsors approach early development for individualized therapies—raising the stakes on first-in-human trial design and pushing evidentiary planning to the front of the process.

To explore this further, Applied Clinical Trials spoke with Mwango Kashoki, MD, MPH, SVP and global head of regulatory strategy at Parexel, about how the framework redefines substantial evidence, where master protocols are most likely to break down, and what sponsors should be doing right now to stay ahead of finalization.

ACT: What does the FDA's plausible mechanism framework actually change for sponsors developing individualized therapies in ultra-rare conditions?

Kashoki: The biggest change is that under the framework, the first-in-human study has the potential to serve as the pivotal trial in support of product approval. Historically, FDA didn't consider or discuss licensure considerations as early as first-in-human—it was just too early. There were other aspects of product development and evidence maturation being considered first.

So the plausible mechanism framework acknowledges the need for these considerations in the context of individualized therapies for rare, severely debilitating, or life-threatening diseases. As described in the guidance, there are particular challenges that come with trying to apply a more traditional clinical trial approach—the randomized controlled trial, etc.—to these patient populations and conditions. You really need to think differently about what evidence base would be adequate and sufficient to support approval.

With the issuance of the guidance, the FDA is now more formally encouraging and assisting companies to design their initial trials as the pivotal trials. For sponsors of genome editing and other types of individualized therapies, this means that even earlier in development, they have greater clarity on the nature and scope of clinical, CMC, and non-clinical data necessary to support approval. That's the main change for companies working with these types of therapies.

ACT: How are regulators likely to interpret "substantial evidence" thresholds when patient populations are too small for traditional trial designs?

Kashoki: For any drug development program, the question of whether there's substantial evidence of effectiveness is always complex. It's a decision that takes into account multiple data inputs, and you can see from the series of guidances the agency has issued on this topic that the determination goes beyond number of trials or statistical significance of the primary outcome. Making a substantial evidence finding is really an ascertainment of the totality of the relevant data.

Under the plausible mechanism framework, sponsors need to ensure they have truly identified a well-characterized molecular or genetic abnormality that's the underlying cause of the disease, demonstrated that their product targets this underlying abnormality or its pathogenic pathway, and shown that there is an exposure response based on clinical outcomes or biomarkers. All of this allows the FDA to accept biologically plausible mechanistic data and evidence, as well as pharmacodynamic and biomarker data, as part of the confirmatory evidence package that supports the primary outcomes from the clinical trials.

So when you have small populations in trials of individualized therapies, what the agency is looking for is the effect on clinical outcomes or appropriate surrogate biomarker outcomes, as well as this confirmatory evidence that strongly reflects disease improvement or disease elimination due to the drug's effects on the underlying mechanisms or genetic abnormality.

As an example, if you have a population that has, as a matter of a genetic mutation, hyperammonemia, and you see a reduction in serum ammonia levels to physiologic levels in patients who didn't have that prior to the therapy—that reduction in serum ammonia is a really robust surrogate biomarker. Combined with all the other mechanistic data, you can say there is substantial evidence of effectiveness. I hope that explains how the agency would consider substantial evidence despite a relatively smaller number of patients in the clinical trials.

ACT: Where are master protocols most vulnerable to failure under this framework, and what should sponsors get right from the start?

Kashoki: Companies aim to accomplish several outcomes through master protocols, and the audience is likely well aware of the different designs that are part of the master protocol approach—basket designs and others—intended to understand potential drug effect across a variety of indications. If you have a master protocol with a basket design intended to enroll different populations with a variety of mutations, and you try to put that evaluation into one clinical study for an individualized therapy, you have to ensure that it is biologically plausible that the product would target each of the mutations for each of those diseases and translate that targeting into clinical benefit.

For example, if you have a gene replacement therapy and try to test it in a master protocol basket design, the effects may not translate as clinical benefit for persons who have dominant negative mutations, because gene replacement—increasing expression of the deficient gene—is not going to really help patients whose mutant protein causes interference with physiologic function. So you have to ensure that the types of conditions intended to be studied under the master protocol are indeed suitable and relevant for that individualized therapy.

There are also challenges around dosing. If there are conditions that could be targeted by the individualized therapy but one involves pediatric patients and another involves adults, you may have different dosing considerations. Different mutations and different age groups may also bring differences in baseline characteristics, disease severity, and progression over time. So whereas a master protocol in concept might be highly desirable, there are many considerations that have to be thought through in order for it to be truly useful as a means of evaluating an individualized therapy effect across a multitude of different conditions.

ACT: What are the biggest evidentiary risks around natural history data, and how should sponsors be thinking about external control comparability?

Kashoki: Where natural history data and other external control arm data sources present challenges is with respect to the extent that they differ from the clinical trial population. For rare diseases, the heterogeneity of both clinical course and potential patient response to treatment can confound the interpretability of results in trials with small sample sizes that also utilize an external control such as natural history data.

You have to try and address that confounding potential by ensuring you have as close as possible patient matching of the persons in the natural history dataset with respect to key baseline characteristics, disease characteristics, and so on—matching to those persons who will be included in the smaller prospective study of the disease being treated.

To achieve this close matching, what ends up happening is that you need a really large external or natural history database compared to the size of the population included in the interventional study, simply because it takes so much searching and confirmation to ensure that you have indeed achieved as close a match as possible with the study population. The heterogeneity in patient characteristics and disease course, and the potential differences in response to therapy even for individualized therapies, presents real challenges around the use of external control data, including natural history data.

ACT: How should sponsors be structuring their development plans right now to avoid delays when this guidance is finalized?

Kashoki: Even though the guidance is still in draft form, companies can still be planning accordingly. We know that guidance can stay in draft form for some time, and as more and more learning is acquired. So based on the principles outlined in the guidance, companies should be thinking very early on about a few things.

First, making sure they have plans to engage with the agency even before going into their first-in-human trial. Because as I said at the beginning, the first-in-human trial could potentially be your pivotal trial. Planning for a pre-IND meeting, potentially even an interact meeting, to discuss the company's plan to use the plausible mechanism framework is important. Getting early alignment with the agency on evidence expectations—what they would be looking for in terms of mechanistic, biomarker, and other data—is critical.

With that input from the agency, really invest in doing a robust non-clinical program. Make sure you have characterized safety as well as bioactivity, and plan for qualification of biomarkers if there are plans to use a surrogate endpoint—making sure it will meet the evidentiary standards for being reasonably predictive of the clinical outcome.

Where possible, invest in a trial design that includes adaptive and seamless designs so that you can quickly get to the right dose, the right patient population, and evaluation of relevant outcomes via biomarkers, clinical outcomes, and so on. And as we've been discussing, think about the use of external data, natural history data, and other kinds of data that can serve as the external control, because these are usually single-arm trials.

There's a lot to unpack from the guidance and a lot of planning that companies can start to do now. The key is making sure it's clear from the very start that the framework is intended to be leveraged, and being clear on not only the clinical trial design but that package of confirmatory evidence that would provide the robustness of data for the agency to make the determination that there is indeed substantial evidence of effectiveness.