Business development in biopharmaceuticals, and due diligence of those deals, is as scientifically detailed as one can get. Even very early stage start-up companies come to meetings with a lot of experimental results to present.

That’s a big contrast to what I often see today from many aspiring life science startup companies, where there is minimal data and lots of aspirational descriptions. It seems that the current startup environment is being influenced more by the tech playbook of developing a Minimum Viable Product (MVP).

I suspect some of this is because those founders are getting advice that is not suitable for the life sciences industry. The other reason is that they are simply too early and have more work to do.

However, we can be sure that any startup company that did a deal with a large pharmaceutical company, or that got funded by a top tier life sciences venture fund, had plenty of data that received scrutiny by the partner/investor.

I will offer an example from my early experience in business development. I can’t give any more recent examples, because I need to respect Confidential Disclosure Agreements (CDAs) and respect professional confidentiality. All examples on this blog are always of situations where the CDA has long expired and the original entity is no longer around (such as because of a change of control).

Be wary of data from the wrong models

This post was inspired by yesterday’s piece on STAT about the problem of animal models of Alzheimer’s Disease (AD). As the author states:

We are overlooking another key driver for numerous translational failures: the overreliance on behavioral readouts from contrived transgenic rodent models to guide drug development for Alzheimer’s, Parkinson’s, and other neurologic diseases.

 In many areas of drug development, the most important preclinical metric is whether the compound showed signs of efficacy in an animal model, most often mice. But this breaks down when it comes to neurologic disorders that manifest themselves in distinctly human cognitive and functional behaviors.

Biopharma blogger Derek Lowe further adds:

I don’t know of too many who would stand up and defend the rodent passive avoidance response assay or the Morris swim maze or what have you as great front ends for Alzheimer’s research, but at the same time, they would be rather nervous about abandoning them completely.

Deciding what drug candidates to advance is really tough. Even the top companies with lots of experienced experts get it wrong. In the last year alone, Eli Lilly, Axovant, Roche, and Biogen all had failures in clinical stage Alzheimer’s drug candidates.

The startup company in 2004

In 2004, I was heading early stage business development at a biopharma company. We were approached by a startup company founded by two university professors. They claimed to have a small molecule compound for treating AD.

The prevailing hypothesis at the time was that AD pathogenesis was driven by the production of β-amyloid peptide that aggregated and deposited in brain tissue.

This company had performed controlled in vitro experiments where their compound inhibited the aggregation of β-amyloid peptide in solution. This was their proposed mechanism of action.

They also had in vivo data using a mouse model of AD. These were transgenic mice that rapidly accumulate amyloid deposits in their brain. However, when they were fed the compound to these mice, autopsies showed decreased accumulation of these deposits in the mice brain tissue. What’s more, this was dose-dependent. Feeding these mice more of their compound resulted in less amyloid deposit found during autopsy. This showed that the proposed mechanism of their compound was active physiologically.

Also, their compound was “apparently non-toxic.” In drug development, nothing is for sure, but this was a compound that had pre-existing data of safe consumption, and many of its analogs also had pre-existing data of safe consumption.

The only curiosity was the potency. A patient would have to take a lot of this compound, maybe up to 4 grams a day. That was unusual, but nothing to warrant a rejection upfront.

These results were intriguing and we had them in for a meeting to present their in vivo efficacy data (experiments to show the compound works in an animal model).

Their CEO arrived with two professors. One professor led the mouse behavioral model experiments and was an expert in using these models for the study of AD. The other professor was a world-renowned scientist in AD research with a huge list of prestigious scientific awards.

From our side, we had our Chief Scientific Officer, our VP of Discovery Research, our top pharmacology scientist, our head of molecular biology, a medicinal chemist, our senior director of behavioral pharmacology, and myself.

Our senior director of behavioral pharmacology was recently hired from a global pharmaceutical company. He had worked in two major global pharma companies and this was his area of expertise. He even authored a peer-reviewed paper on this topic.

The presentation from the company was flawless in showing the preclinical efficacy of their compound.

Morris water maze.

They used the Morris water maze test. It’s not really a maze. In this test, the mouse is placed in a tank of water at one exact location. There is a platform hidden somewhere in the tank, just under the surface of the water. Since the mouse doesn’t like water, it will swim around until it finds the platform. If this exact experiment is done every day, the mouse will remember where the platform is. A camera records the movement of the mouse as it swims around looking for the platform. Once it learns and remembers, the distance and time the mouse takes to find this platform becomes less.

Transgenic AD mice will forget the location of the platform as they grow older. The distance and time for them to find the platform will increase as they get older.

When they fed their compound to their transgenic AD mice, the distance and time for the mice to find the platform did not decrease. They showed many different experiments to make this point.

We could not find any fault with their studies.

Our senior director of behavioral pharmacology appeared troubled. He asked how they physically executed the studies. He wanted to make sure no one was in the room at the time of every experiment, because that would distract the mouse. He wanted to make sure it was done at the same time every day. In every element, the professor did the studies perfectly.

He asked if they did a “probe test.” The probe test looks at the distribution of time the mouse spends in each of the four quadrants of the tank. This corrects for issues such as if the mouse remembers the platform location and goes to that quadrant, but some non-cognitive impairment causes them to spend more time searching for the platform. The professor did the probe test, and the results were as expected.

Next, we looked at their mouse model. As I recall, there were about twelve murine models of AD at the time, of which three were used most often. Each strain of transgenic mouse had different molecular mechanisms that led to the generation of β-amyloid deposits and each had different physiologies and life cycle features.

In human AD patients, in addition to seeing a build-up of β-amyloid plaque, tau proteins in neurons are also observed to form neurofibrillary tangles over time. None of these mouse models expressed any tau neurofibrillary tangles.

Their other professor was there to answer our questions about why they think the particular transgenic mouse strain that they used in their studies was appropriate. This was not an objective assessment, because it rested on our view of the expertise and plausibility of the arguments of this esteemed scientist.

The deal decision

In post-assessment, almost all of us thought this startup company did a good job advancing their compound to this point. The data looked excellent. We were speechless when it came to objective concerns about the data. It was the most promising deal we had seen of dozens over the last year.

The only hold-out was from our new senior director of behavioral pharmacology. He did not feel comfortable moving ahead. We pressed him on why, and I in particular drilled him a lot. He still did not offer any answer.

Our executive team rejected the deal. Most of us were disappointed, I in particular, because my job was about closing a good deal.

Everyone else asked me why we turned this down. It seemed as if we didn’t have the courage to do any deal. My explanation was that if we were to hire a top behavioral pharmacologist and to relocate him to work for us, we need to trust his judgement, or else what are we saying to our top scientists?

Also, if we went ahead, this was no trivial financial decision.

  • There would be an Investigation New Drug (IND) application that would cost about $1 million to put together at the time.
  • There would be a Phase I study using a dozen volunteers to establish safety and to determine dosing.
  • A Phase II would require up to 400 patients for up to 2 years.
  • Assuming all goes well to that point, a Phase III study would run into thousands of patients for years.

The “back-of-the-envelope” costing of clinical studies is by a per patient per year basis. It was probably $15,000 per patient back then. These numbers add up to huge costs. If we were to advance this, we would have to go to the stock market to raise financing. If we were going to do that, we would have needed a much more compelling drug candidate, like a monoclonal antibody. This small molecule did not even target a receptor nor enzyme. It would not have generated much investor excitement.

To take such a big bet and risk the livelihood of our company, we would have needed a strong conviction that this compound would be successful for treating AD in real human patients before doing any licensing deal with this startup company.

Business development of science is different

In business development practiced in most industries, there is a tangible product that can be sold, or there is a prototype or Minimum Viable Product that can be used as a demonstration.

In life sciences companies with only potential products in development (such as this startup company founded by these two professors), investors assess deals for the reward and the risk. The reward comes from a large market size for the potential product.  The risk is the probability of failure during the R&D process, which is very high in the life sciences industry.

For investors to assess risk of the startup company failing (and the probability of success), having as much experimental results as possible is crucial. The scientific data – and the experts in the company who can explain the data – are the assets of the company.

The type of data shown by the professors in this case study is a good example of the level of detail required for an investor or alliance partner to make a decision.

Now let’s look at business development from our side.

In all other industries, business development is about selling: the more the better.

In deal closures for life sciences companies, it is about deal support done by lawyers, accountants and consultants: the more the better.

However, when it comes to sourcing deals and finding the best one(s) to take our company to the next level, success is about not doing deals that will cost a lot of money and then fail. It is also about executing that one strategic deal that is transformational, and if the company is big enough, then a series of these transformational or bolt-on deals.

When I left this job a few years later, a recruiter asked me about how many deals I did. Her response to my answer was “Is that all? You looked at hundreds of companies and only did a few deals?” She spoke to me as if I was a loser.

I was caught off guard by the naivety of that perspective. She viewed business development as if I was in sales or as if I was professional service provider. I should have explained it to her.

The outcome of this case study is telling.

Lessons from the aftermath

Less than a year later, the two professors’ startup company closed a deal to license their AD drug candidate to a small Toronto-based pharmaceutical company.

This Toronto-based pharma company raised financing and took the drug candidate through IND and Phase I.

Then in Phase II, nine deaths occurred. These fatalities were never linked to the drug candidate. There were also a large number of serious adverse effects, particularly increased rates of infection in patients taking the drug candidate. These required the clinical study to reduce the dose level dramatically.

In the end, all clinical development stopped.

As for the Toronto-based pharmaceutical company, it is no longer around.

I follow as best as possible what happened to every deal and project in which I worked. I think this is helpful to learn and improve practice, just as pro athletes and chess players study their post-game performance to get better.

Many years later, I asked our former senior director of behavioral pharmacology: “your instincts saved our company from a disaster. In retrospect, what made you feel this was not going to work out?”

His answer was that he still did not know.  He was an expert in these animal models. He reaffirmed that it was just his gut feel after doing this line of work for so long.

In this conversation, I finally realized that he did not know because he was struggling with that dawning realization that these models were not predictive of real Alzheimer’s Disease, a fact more known today, but only emerging as a question to a rare few experts at the time this deal took place, including to himself.

That meant that this startup company had no meaningful data about the efficacy of their drug candidate!

He also added that these patients would be taking a lot of this drug: grams a day. Where is it all going?

Yes, it then hit me too: there was no drug metabolism and pharmacokinetic (DMPK) data. If the drug candidate’s mechanism of action was postulated to prevent physicochemical aggregation, DMPK data was needed to support that.

They would have generated that data eventually for the IND. FDA requires it. However, they missed something crucial in those results and went into Phase I and then Phase II.

The DMPK had something to do with the adverse effects. None of us saw that coming, because our review team never worked through this line of analysis. This exemplifies the unpredictability of clinical studies, especially if you don’t consider all angles.

That recruiter may have thought I was a loser for not doing enough deals, but she had no clue about this job.

The Toronto-based small pharmaceutical company that closed the deal instead of us ended up costing their investors a lot of sunk costs, and they bore the negative outcomes.

My final learning point is that it is important to trust your best scientists and to work hard to understand their line of thinking. I am proud that our team made the right choice.

The importance of scientific data in biopharma deals
Share this...