The Evolving Biopharma Regulatory Landscape: Q&A with Harpreet Singh, MD

In a conversation with Pharmaceutical Executive, Harpreet Singh, former FDA Oncology Division Director and current chief medical officer at Precision for Medicine, discussed the regulatory landscape for biopharma, emphasizing the importance of early and frequent engagement with the FDA.

Singh noted the shift from two pivotal trials to one for drug approvals, highlighting the nuanced evolution in evidence standards. She praised the Commissioner's National Priority Voucher Program for accelerating drug development but called for more transparency in the selection process. Singh also discussed the potential of AI, specifically the FDA's closed system Elsa, to streamline data review. Finally, she stressed the need for better communication between biopharma and the FDA to improve their relationship.

A transcript of Singh’s conversation with Pharmaceutical Executive can be found below.

Pharmaceutical Executive: You spent years inside the FDA shaping oncology drug policy, now you're advising the companies that navigate it. What does the regulatory landscape look like from the other side and has anything surprised you?
Harpreet Singh: I can say it's probably much more facile for somebody like me, who was inside for nearly a decade. I think that the principles of kind of early and often engagement and watching precedent apply, and I think understanding each individual division and the decision makers for each therapeutic area is really crucial. The FDA is a huge institution. They're data driven, but they're not a monolith. So, you know, having that inside track of information and understanding people's thought patterns and a divisions kind of behavior around a certain disease has been really critical. I think what has surprised me, in terms of our clients, you know, just, just maybe, what the perceptions are around decision making at FDA, I don't know how deeply appreciated it is, how data driven the decisions that are being made at FDA are, and I think what's also surprised me is actually the very high regard that people have for FDA oncology in particular, And for a lot of the initiatives that were brought forward during my tenure, and even though some of these may have caused some increases in drug development timeline or even some initial confusion, I think overall, people have really responded well and viewed FDA favorably over the years,

PE: Does FDA’s shift from two pivotal trials to one represent a genuine evolution in how the agency thinks about evidence?
Singh: I think this is very nuanced. I think it's more of an evolution versus a revolution, and that is because as long as I've regulated drugs in the oncology space, and I housed rare diseases, rare oncologic diseases, we've always used one pivotal trial, and we were always kind of a deviation from the norm, and there's very strong biologic and scientific rationale for that, and the rationale is that we were treating life threatening diseases.

So the idea that you could walk into a randomized, typically, but a registrational design, pivotal trial, and use that one adequate and well controlled trial, which is the technical language that demonstrates safety and effective efficacy, to approve this drug for patients with some risk on the back end right, fully characterizing safety, or perhaps long term outcomes for life threatening or rare diseases, has always been acceptable when you pour that logic into non imminently life threatening diseases, because diabetes, multiple sclerosis, cardiovascular disease, these are ultimately diseases that can take one's life, but they're more chronic comorbidities. So understanding that people would need to be on these therapies, likely for much longer, you think about it, you're on a statin almost for life, you're on a diabetes medication. These are chronic conditions, treating much larger swaths of the population.

So in doing that, the fundamental question is, can you answer all the scientific questions in one pivotal trial? And classically, the answer to that, from a regulatory standpoint, has been no, and therefore we need to do two adequate and well controlled trials to account for the statistical variation and the power and kind of the Alpha allocation and P value, and also just accounting for the large variability in the populations that you'll ultimately treat.

Do I think that there are opportunities to streamline drug development and push some of that risk into the post marketing setting, particularly where the therapies are transformational, an example of this for me would be Car T's. Lupus does if we're seeing transformational responses in a non-eminently life-threatening disease, but a disease with great comorbidity that can lead to long term organ damage, we would need two adequate and well controlled trials, depending on the magnitude of the initial trials results.

I would argue, in a case like that, given the modality, given the nature of the disease and what we're hoping will continue. Which are these really transformative outcomes, that's an example. Or I would say I'm okay with one pivotal trial, provided it's well designed, well controlled, and we have good measures to get additional data post market

PE: With multiple drug approvals now confirmed, Where do you think the CNPV program has delivered on its promise, and where has it fallen short?
Singh: I think the program's done a good job of highlighting promising therapies across different areas of public health. We saw a lot in oncology. We've seen some in antibiotics. We've seen some in pain management. Now with the psychedelics, this is new. So I like the idea that we're using these methods to accelerate, really just the back end of drug development. We need to be very clear about this. These vouchers are being awarded after the trial has been designed, after it's been executed, predicated on results that we're seeing, and so I don't know that they're fundamentally or dramatically altering the landscape.

My point is, approving drugs very rapidly is not a novel concept at FDA. In fact, it's something that there's a playbook for and so the voucher takes that, applies it with slightly different prerequisite criteria broadly across the agency. So, I think it's delivered on expanding the reach of these accelerated programs, where I think we need a little more from FDA is on the transparency around how these vouchers are given or designated. We know it's been widely reported that some of these companies have not requested vouchers, they are just receiving them and what is the process by which they're being selected? And then one question that has certainly come out is, you know, what are the implications post approval of receiving a voucher? So there's a little bit of opaqueness around this, and I think that's where we need more from the agency.

PE: With FDA’s recent implementation of AI, how do you think this can improve the agency’s internal processes and what effects could this have on drug approval processes?
Singh: Yeah, I really believe in harnessing the power of AI and I do think that FDA is a place that employs very highly trained scientists to review volumes of data. So I was not at FDA, when their AI system, Elsa was rolled out, but I am in very close touch and have a very active pipeline into FDA. The major difference that people may not realize is FDA’s Elsa system is a closed system, which means exactly as it sounds, it can only draw from its internal data.

So, if you were looking at a novel drug in breast cancer, and you wanted Elsa to compare that asset to FDA approved therapies, and how does this stack up, and how does this compare? Which would be very key in decision making. Elsa at this stage, to my knowledge would not be able to provide some of that pivotal data that you need for critical thinking.

Where I do think there is likely great potential is taking a bunch of data, aggregating it and giving it to the reviewer in a salient way, or summarizing the key points, or finding signals that perhaps are less obvious. I was trained in a very classic way at FDA, where this process even evolved in my tenure, where I was actually in data sets. I had some programming knowledge, if you will, so I could sort and find doing line item data review, which is how we were classically trained.

So I have that training. I do believe you can train an AI agent to do some of that work we were doing. It just it all goes into the training of the agent. I believe there's a lot of efficiency that can be derived, even with the closed system for kind of summarizing and pushing up key data. I believe the critical thinking skills and understanding what the landscape looks like and what the risk benefit looks like overall, still remains with the reviewer, and that's the critical thinking that we really want our FDA staff to be doing.