How to Speedrun a New Drug Application

"it took 48 hours to write the application, instead of four months"

Feb 22, 2024

Alvea launched in 2022, and announced on its website that it was “building an engine for warp speed drug development.” In the middle of COVID-19, the company planned to show it was possible to dramatically reduce the time from molecule to trials for new medicines, starting with a new vaccine candidate tailored to BA.2, a particular COVID variant.

A year and a half later, Alvea shut down (for reasons we explore here). But before that, it became the first company to vaccinate an animal against Omicron, and proved that pharma companies can move far faster through the FDA approval process than they currently do.

Today we talk to Grigory Khimulya, the former CEO of Alvea, about we can get effective new drugs more quickly.

How’d this thing start, Grigory?

Initially, Kyle Fish and I started an AI antibody design company to do early-stage design of drug candidates against new pandemics in weeks, instead of the usual years. Alvea grew out of that, in part. It was a four person co-founding team: myself, Kyle, Ethan Alley, and Cate Hall. This is really how we got our first experience of the pharma industry and began to see bottlenecks.

In our experience, these bottlenecks are very common downstream of the initial discovery and design stages, and through the highly regulated steps, all the way to clinical trials.

Alvea started in response to the public health emergency from the Omicron variant. It’s been what, two years now? At the time, it was the most transmissible and immunoevasive COVID variant to date. There was lots of uncertainty about how dangerous it was going to be.

There was a huge public health need, especially in low- and middle-income countries that were very unlikely, in our view, to receive updated COVID vaccines in time. This was also an opportunity for me to see what the later stages of the process looked like and whether there was a way to get to clinical trials and to do drug development much faster than it's normally done.

This is where we've had some success, even though we ultimately made the decision to wind down. That's how it came together. Ethan and Cate came in that first week, as we were getting concerned about Omicron, it all started from there. If memory serves, we hired our first 20 or 30 teammates in the next two weeks.

As you were putting the team together, what did you imagine the roadmap would be? What was the initial plan for getting the thing off the ground?

What I've been excited about for a long time, the problem we’ve been solving, is taking a new medicine from design to human trials to results as quickly as humanly possible. This is of course particularly critical in pandemic emergencies.

But it's really important in peacetime drug development as well. Every wasted month and year of development time off a treatment for a disease that doesn't have one is costing lives. What we wanted to do is design a very simple and scalable COVID vaccine focused on Omicron. Initially we focused on the BA1 variant, and then we pivoted to the BA2 variant that emerged a bit later.

It’s a transition from this very minimally regulated world of basic and applied science, where drug designs come from, to a suddenly very regulated world of human clinical trials. You go through all the steps required, including animal experiments that test the safety and efficacy of the drug manufacturing at sufficient scale to provide for a clinical trial. Then of course you go through the clinical trial itself. We wanted to try and streamline every part of this process to get this new medicine to patients in the pandemic and see how well it worked.

Obviously, as a small, lean startup, you were able to hire quicker. You were able to do some things much faster. I think that's clear.

What was your original theory about being able to streamline the process? Why wouldn't the pharma industry be able to move at the pace that you were moving?

This is a really good question. I think my initial motivation was something like, “this is a really important problem.” I think this holds true for most people on our team. We didn't necessarily see all the ways in which it would be easy or difficult. We just had a sense that this is something that could be done a lot faster by a dedicated team in a situation where the emergency is extremely clear and salient.

We did an informal poll and many of our advisors were setting numbers like five years to go from an idea to submitting the clinical trial application to the regulator. Not even to running the clinical trial. It just didn't seem right.

Of course, we were not the only ones who were thinking about that. Some pharmaceutical companies achieved tremendous results during COVID, including Moderna and Pfizer. This was an inspiration for us, but we wanted to see how this could be done as a small team from design onward.

The way the industry is normally structured, there are drug design people, and there are early-stage biotech people who don't really think about the regulated stuff. They think about the science, they generate the candidates, and then there is a very inefficient, high transaction cost process where some of these candidates get picked up by big pharma and then developed further.

Just to provide more light on what we discovered, I'm going to tell a bit of a stylized story.

I love stylized stories.

Obviously, it works differently in every company. But every step: animal testing, clinical trial design and preparations, interaction with regulators, and manufacturing, among others, is usually done by a siloed department, and many of them are done in series.

One step gets done in one department and then it passes the gate and proceeds to the next department, but only after the first one has done its job. This can lead to the steps not being under a tight schedule, and a really slow and annoying process of actually assembling the regulatory package.

Is that package a new drug application?

Yeah, in the US and some other countries it’s called the clinical trial application (CTA). It's all the documents that go to the regulators for them to eventually make a decision. [Astute readers may remember Statecraft interviewee Dr. Gilbert Honigfeld posing with all 327 volumes of the clinical trial application for clozapine.] It’s a summary, including hundreds of pages of supporting data for all different aspects.

The FDA benchmark, the average time it takes to write this package, is four months. You've got to get all the departments together and they’ve got to talk about it. They’ve got to plan it out and then actually execute the writing. For what its worth, this timeline is based on my recollection of what a few FDA experts anecdotally said was the expectation, but I so far haven't found solid stats which are this fine-grained, and I would love to see them if they exist.

From the start, we decided to do things a bit differently. Instead of doing this in sequence, in departments, we did it with a small, integrated team. And we did it in parallel as much as scientifically possible from a risk management perspective. Because it's a small team and everything was designed from the start to proceed in parallel to the final result, it took 48 hours for us to write the final application, instead of four months.

Is the reason that a big pharma company doesn’t operate on a more integrated scale like this just math? Is it that J&J is a massive operation, so it’s not feasible to do all the work in parallel?

I think in normal time, scale has a lot to do with it. In particular, it's the difference between working on many drug candidates in parallel, vs focusing on a single drug and running all the stages of the process in parallel. I think that’s the best way to think about it. The latter thing is what in our experience gives you a really big speed boost.

The approach that we've taken, and what’s really been enabled by this integration, parallelization, and a small team, is that we would always be looking across all the stages and all the activities, and asking, “What is the current bottleneck? What is the thing, right now, that is pushing back our delivery date of the clinical trial application and ultimately the clinical trial?”

Of course, a lot of these things just can't physically be done faster. Sometimes cells just need time to grow and there isn't a whole lot you can do about it. But in our experience, a lot of the slow steps are slow because of people, not physics or biology. We would basically focus the attention of everybody we could on the bottlenecking step and invest a somewhat unusual amount of effort into really going after it. So that's the second piece of it.

If Alvea was still around today, how much of that initial pace do you think you could maintain over time? How much of it was the consequence of an emergency environment and a tightly scoped project?

If you were still doing this a year from now, could you keep that pace in perpetuity?

My guess is yes, absolutely we would have. I think one thing that didn’t lend itself to this approach was the early stages of drug discovery themselves.

Our initial candidate was chosen and optimized to be quite simple and well-validated from a scientific standpoint. When, after our first clinical trial, we expanded into more speculative and challenging scientific areas and applied a similar kind of thinking to speeding it up, it turned out to be much, much harder.

Early-stage, risky science is very difficult to do on schedule, even with all the best conceptual speed-up approaches applied: parallelization, contingency and bottleneck optimization, etc.

The focus of our pivots from that point on was coming in on cases where the science is reasonably clear and there was a strong proof of concept: for example, in animals, which is still comparatively unregulated.

That’s one place where I think our initial assumption that we could speed everything up did not pan out. But as far as running through a new clinical trial goes, I think the approach generalizes. We saw that, in preparation for future clinical trials as we continued to operate.

What did you learn about the regulatory environment from that first clinical trial?

I think that was the most important learning. There’s a meme in Silicon Valley circles about people who are really into moving fast and breaking things. The caricature is that the FDA is the enemy of progress, medical regulators are the enemy of progress, and they're slowing everything down. On reflection, I don’t agree with that take, and our experience doesn’t really support it.

When you say on reflection, do you mean that you would have agreed with that take before this process?

No, I just would see that conversation going on in a lot of private and public places, and I was agnostic about it because I'd never interacted with regulators myself. I could see reasonable people pointing to lots of inefficiencies in how the FDA and any medical regulator operates. No doubt things can be improved, but there are a few things that I've come to appreciate.

What regulators are trying to do is actually really hard. People look at the drug development processes, and they notice that like 90% of the effort is focused on testing, verifying, and quality control, all these boring things after the creative, exciting part is done. There is a big difference between the hypothetical drug that is a formula or a design in the computer, and the actual drug that ends up in your vial.

In drug development especially, making a thing that is plausibly good is much, much easier than making something that is actually, reliably, very good. Deploying drugs to scale requires that reliability.

It’s a very hard socio-technical problem. All the different kinds of regulatory requirements, quality management, quality control, etc., that could be naively identified as red tape or boring paperwork that slow down the innovators are actually there to achieve that reliability.

Of course, when you get into the details, there are tons of ways this could be done more efficiently. But the fact that validation, testing, and ensuring that things are as they seem is 90% of the process is just the way the world works, not any fault of the regulators.

My second conclusion is that regulators are fundamentally reasonable, scientifically minded people who want to do good things for patients and get good drugs to them quickly. We’ve interacted the most with regulators in South Africa who have been, in a word, fantastic. In every case where we had a scientifically based opinion that some guideline was inappropriate in light of new evidence, they have been very receptive.

People who have this critical take about regulators have often seen these guidelines that seem very harsh and uncompromising in how they state some process needs to work. But it's actually always a conversation with highly scientifically competent people who are capable of analyzing the evidence, hearing arguments and reconsidering what might be required in any given case. That’s been our experience.

Something that folks who are coming into it fresh may not think to do is don’t take anything as gospel. This is advice I only really started hearing after getting to people who are some of the most experienced in the world at dealing with this kind of regulatory process. It's always a conversation. It's always a conversation about science and safety for the patient, which is very much in everybody's interest: patients and regulators and companies.

It feels to me like regulators don't set their own risk tolerances. Regulators get blamed when something goes wrong, but don't get a whole lot of credit when they make things safe and good. Of course there’s legislation and Congress to set standards, but it’s really society, through the court of public opinion, that sets a risk tolerance threshold.

I’m really sympathetic to society potentially shifting the accepted risk tolerance toward getting more innovative medicines to patients faster. But that’s just not a decision that the FDA makes. It is a decision that happens at a much higher and more diffuse level in society. It feels like an aspect that I failed to appreciate in my naïve mindset before I went through the process, and that can be ignored by folks who are looking for a policy solution.

There’s an incentive set up by how the news works. There’s much more reporting on very rare issues compared to achievements in safety in the vast majority of the drug supply.

Can you give me some more context for the conversation between the drug sponsor and the regulators?

You have, on paper, some set of requirements. But those requirements may actually be negotiable? What's the process by which you negotiate that?

The high level frame on this process is it is up to the drug developer to propose what kinds of evidence would be sufficient. Given the scientific literature and everything that is known about this kind of drug, what would be sufficient to establish safety and efficacy? What would be sufficient regarding a significant unmet medical need to approve the first human clinical trial?

These guidelines are compilations of high level best practices and suggestions for a thousand specific decisions that you might make in the course of developing a drug. For example, you might need to run your safety testing in a certain number of different animal species, which of course is expensive and takes a long time.

Those are specifically guidelines and best practices rather than regulatory requirements, right?

Correct. There are regulatory requirements, which are good manufacturing practice, good laboratory practice, good clinical practice, so called GXP, and those are much more serious. They're also much higher level and have more to do with ensuring that every bit of quality control the regulators have agreed on for these kinds of cases is there.

But, in the case of guidelines, they speak much more precisely to the specific kind of drug that you might develop. For example, for us, this was a DNA plasmid-based vaccine, which is very different from many other kinds of drugs that FDA considers and has much more specific requirements for. In some ways, it does not need to be tested as extensively as a completely new, unprecedented drug because of its very long history of use in humans and many clinical trials that have come before.

This process happens concretely, in your conversations with regulators before your clinical trial application is submitted. There are different kinds of preliminary meetings with the FDA, and the drug developer is in a position to propose what they think is sufficient to establish safety, consistent with what’s normally required for a phase one clinical trial. Again, in our experience regulators are super reasonable and responsive to hard scientific arguments about why this or that thing might optimally be done differently.

Looking back, is there something you would have done differently in your Interactions with regulators?

Honestly it feels like the answer is no. I'm sure there were small things we could have changed in hindsight about this or that part of the application to preempt a question or something like that. But not in a big way. The problem in our experience was really not on the regulatory side.

The problem was with operational inefficiencies that run deep and fractally for every part of this process, and I think that suggests a focus for other efforts to speed it up. Getting operations parallelized, and speeding up the process of generating the data and arguments that would convince regulators, give you a much bigger bang for your buck than over-optimizing the regulatory process.

There are common-sense things that we did that I would recommend everybody do, which are to work with experienced consultants who have seen a lot of applications and to proactively address ways in which you might be doing things differently from how they would be normally done.

I think there are ways to implement the same regulatory requirements that lead to a very slow process or a very fast process. It's really in the implementation and prioritization, making all the pieces fit together very tightly on schedule, where you get the speed up.

Tell me a little bit more about the “fractally distributed” organizational problems that you encountered.

For sure. Again, the larger context of this is a normally serial process through siloed departments. The big changes are integrating the process, running it in parallel.

On top of that, there is a huge family of bottlenecks, some of them funny, some of them just frustrating, that we ran into and resolved. In terms of solutions, there are maybe three different clusters of things.

One is just using relatively straightforward software and automation tools. This is usually the easiest solution.

For example, when you work with any vendor for a pharmaceutical company, almost everybody requires an NDA to be signed. This by itself can eat up to two weeks of time on both ends of this transaction. We had automated this NDA signing process so that it would usually happen in hours. Many of our vendors would follow up and tell us how insanely fast this was and how it was the smoothest and fastest contracting experience that they had ever had.

I really don't have an answer for why this is not how it normally works, but we've encountered a lot of places where a simple – for folks in Silicon Valley simple – software and automation infrastructure was extremely helpful.

I need more of an answer than that. You can't just tell me you don't know why you guys were so effective.

Okay, I think part of it is the siloing between departments. The value of this kind of solution is much clearer if you are looking at the whole process and wanting to optimize the bottleneck at every step for a single drug you really care about getting to the patients as quickly as humanly possible.

If you are running tons of these drugs in parallel through this sequential gated process and the whole thing takes five years, and that's the industry standard, then saving a week on the NDA negotiation side might not feel like such a big deal.

I think that’s part of it, but I will be honest with you. I don't have a satisfying answer to this for myself, and I'm very frustrated by it. I expected there would be a better way but I'm still at a bit of a loss. This is not to say that we are the only folks who are moving in this direction. Vial, for example, is one company that is working in this direction, not quite as fast as us yet. But this is not widespread by any means.

Another big pattern is that, for some reason, for a lot of these key processes that really move the needle on speed, the standard operating procedure for the industry is to talk to maybe three to five different vendors, compare them across a bunch of categories, and then pick one and go forward with them. That never seemed to work for us. We would approach it by finding every single vendor in the world who does the thing that we need done, finding the best people, and then going in and very closely redesigning and managing their process for maximum speed. Practically, this involves parallelization and then bottleneck hunting in the vendor’s process to identify ways to make it faster.

A good example of that was the manufacturing of the drug itself, of the DNA plasmid that was our vaccine’s main active component. Our initial quotes from the first few vendors were like two years. “It takes two years. There is no way around that. This is just how long it takes.” Then we found some folks who said, “It’s going to be hard, but we can do it in a year.” Then, once we have come in and looked at it deeply and redesigned it in collaboration with these folks, we ended up doing it in just over two months if memory serves.

This is the kind of speed up that's possible. Part of that is really taking advantage of platform and modular processes. A lot of vendors, folks who ostensibly do the same thing over and over again for many different pharma companies, won’t do exactly the same thing: they develop a new, custom process for every customer. It's a little bit complicated, but to simplify it a lot, there are often intellectual property reasons for it. If you have your own patented process, which was developed specifically for you to manufacture your drug, you have some increased protection in future intellectual property disputes.

But it just ends up eating a lot of time. Often, these vendors have a platform modular process with reasonable defaults on lots of different parameters that they can use to produce a very high-quality product.

Taking advantage of that wherever possible is a really big thing. You see it across other fields as well. In the formal academic work on mega-projects, modularity has emerged as a really big factor for projects that don't go over budget and over time. This was a big thing for us.

One last thing that I think is unavoidable if you want to move this quickly is strategic in-housing. If everything else fails, if you can't find a vendor who can do things quickly enough, you can do it yourselves. This usually happens for relatively small steps of the process that are nevertheless critical and end up bottlenecking. This is where the culture and the people are crucial, because, of course, it’s expensive and really difficult.

For the small number of really important bottlenecks, you need to have an emergency mindset. You end up having to do things like building out an entire tissue culture facility in-house in a matter of weeks, faster than any vendor quote.

To summarize, automate as much as possible, look at every single vendor and redesign and manage their processes for speed, and then strategically in-house the most important bottlenecking step for which everything else fails.

I guess it's more of a family of responses to a whole zoo of specific small inefficiencies, rather than just one thing that makes a huge difference. Perhaps that's the other reason why this is relatively rare and why we've been able to achieve these results. It’s a big leap to a better place, at least in terms of speed, that’s difficult to do piecemeal; we developed it in response to a pandemic emergency. You need to change a hundred things and you need to apply this approach a lot to really see the kind of timelines that we've been able to see.

Is there anything else I should have asked you?

Perhaps one thing to add is what to take away from the ultimate decision to wind down. There is never a single explanation for a decision like that. Of course, changes in the funding environment have been factors, but there are also factors connected to the specific risk profile of the science that we decided to pursue.

I really think there is a huge path forward to substantially speeding up drug development and clinical trials and that there has been awfully little innovation on that front, all things considered.

I'm excited for folks to keep working on this problem. I'm optimistic about large language models automating a lot of this work in ways that we got started on. This feels like a profound opportunity right now, and the wind down of any one company really doesn't invalidate this concept for me.

I think our results – getting from idea to first in-human trials in less than six months – really speak for themselves in terms of what's possible.

Thanks to Chloe Holland for her judicious edits to this transcript.