Biotech Startups And The Hard Truth Of Innovation

Posted May 20th, 2019 in Bioentrepreneurship, Biotech startup advice, Corporate Culture, Leadership | Leave a comment

Gary Pisano’s recent Harvard Business Review piece, The Hard Truth About Innovative Cultures, beautifully frames up how innovative corporate environments are frequently misunderstood. Innovative startups aren’t just about being cool and nimble, having beer taps in the kitchen, or an endless bounty of swag. Pisano sums up the harder, harsher reality of truly innovative environments: “These cultures are not all fun and games.”

Pisano’s piece struck a huge chord with my own experience working with both large pharma and small startups over the past 20 years.  And many of the observations are in line with my post from July 2017 on Distinctive Biotech Corporate Cultures.

While reading Pisano’s piece, my brain kept saying “Yes!” and “So true!” every few sentences.  I tweeted out my endorsement of the piece back in January but have finally gotten around to blogging on it; in this post, I’d like to add some “practitioner” color-commentary to Pisano’s observations.

Here are his five major attributes about innovative cultures:

  1. Tolerance for Failure but No Tolerance for Incompetence.
  2. Willingness to Experiment but Highly Disciplined.
  3. Psychologically Safe but Brutally Candid.
  4. Collaboration but with Individual Accountability.
  5. Flat but Strong Leadership.

I’m going to share some reflections on each from my vantage point as a biotech investor.

Tolerance for failure requires an intolerance for incompetence. 

We work in a high risk field: less than 5% of drug discovery projects ever make it to market. We have to embrace failure. When projects or investments are killed, we should openly acknowledge it in the light of day, not bury the dead in the night, and celebrate the learnings that can be gleaned: as we did with both the Quartet Medicine and OnQity write-offs.  Innovative cultures have to convey that its “ok to fail” and foster quality learning loops.

But it’s not ok to fail due to sloppy science or poor decision-making. Not asking the tough killer questions, or using the right controls and comparators, can lead to chasing false positives for too long – which is not acceptable. It’s just too easy to keep drug programs going, overhype the data to get the next dollop of funding, or play around but never truly test the science – simply because we can rather than we should. It’s incompetent not to do the killer experiments, and do them as well as possible. If you are waving your hands too much, maybe the science is too early. Try to invalidate the central biological hypothesis early, before mountains of capital put too much weight on the scales of judgement – leading to low stringency and poor decisions biased by the sunk cost fallacy and the practicalities of being “too big to fail”.

All of this requires not only competence, but a shared view that the collective time of truly competent people is the scarcest and most valuable resource in an innovative culture: don’t waste that resource on ideas that are passed their sell-by-date (or never should have had a date!). Chalk up a failure and spend your time on better things.  Move onto the next project, or the next startup, or back to big pharma. But move on nonetheless. Competence, by definition, can’t tolerate complacency around the decision to move on.

There’s another corollary of this observation. We all fail at hiring sometimes, as it’s impossible to hire only truly competent people when you are scaling an organization beyond more than a couple dozen employees. Mistakes in hiring happen, and that’s ok. But there can be no tolerance for keeping underperforming hires around just because we’ve got a “nice” culture. Robust performance management systems need to be in place to weed out those folks who aren’t delivering with high performance. I’m not suggesting it has to be the Jack Welch “fire the bottom 10%” rule – but if you aren’t transitioning some people out of a startup over time than you are almost certainly tolerating too much incompetence. It’s just the practical reality that we all make mistakes in hiring over time. Sadly, I’ve been involved with companies that have tolerated mediocrity for too long. It enables weaker researchers to keep hand-waving around their data, when all the competent folks around them cringe. It feels benign and “nice” not to fire people, but honestly it is enormously corrosive over time.

Nothing reeks worse in a company than tolerance for incompetence: everyone smells it, knows it’s there.  And knows the leadership are allowing it. That stinks for everyone, and negatively affects the overall corporate culture.

Willingness to experiment requires rigorous discipline.

Comfort with ambiguity and a lack of structure, and often process, is frequently key to finding innovative ways of doing things. But in biotech startups this loose structure doesn’t condone doing willy-nilly science – or crazy “Friday afternoon experiments” every day of the week.

In a world where a researcher can do only a few of many possible experiments, which ones should be chosen? The only way to prioritize the experiments for a program is to understand the questions you are asking – and the answers you are likely to get from the data. Once chosen, laying out the goals and milestones of the work is key. And then measure up the outcomes against those. Applying the age-old maxim – “Measure twice, cut once” – is a great rule for scientific exploration.

Experimentation works best in applied R&D when discipline is applied to the thought-process and the decision-tree before actually doing the work: defined a priori, specific go/no-go criteria need to be in place for where to take the project or initiative next.

Further, as Pisano notes, experiments should be selected for their learning value, and designed “rigorously to yield as much information as possible relative to the costs.” Essentially, it’s applying the logic of capital efficiency to innovative experimentation.

Don’t confuse free-ranging scientific exploration of a new modality, or new drug target, with a lack of rigor. Even experiments done with “early hit” molecules, designed in the spirit of learning, need to be done as rigorously as possible. I’ve seen programs get derailed by early negative signals caused by “Hail Mary pass” in vivo experiments done with suboptimal hit series; hope and prayer studies are occasionally ok to do over time, but full transparency around the decisions, criteria, and outcomes needs to be in place.

Lastly, just because a biotech doesn’t have formal committees to designate “stage gate transitions” doesn’t mean this lack of “process” or “structure” should lack discipline.  I’m very confident that we have Development Candidate criteria in our startups that look as robust as Big Pharma’s, maybe without Lipinsky Rule biases – we just deploy them without cumbersome governance committees, associated box-checking exercises, or a strict “cover-your-ass” mentality to every one of them.

Fundamentally, scientific discipline has to be intrinsic to a biotech startup’s culture or they will inevitably drink their own Koolaid and lose their way.

Psychological safety requires comfort with brutal candor.

Organizational “safety” is hugely important: a culture where individuals can discuss the truth without fear of vengeful consequences. Almost every biotech has values up on the wall that reflect transparency, openness, and challenge; but the big question is do companies and their executives “walk the talk” and live these values with both safety and candor.

What if the Emperor has no clothes? Can scientists challenge the CSO’s conclusions, or the CSO’s proposed experimental next steps? Can researchers critically (respectfully) evaluate other team members’ data? Can the CEO’s positioning of the status of a lead program be questioned by the team? These are the types of situations that reveal a lot about the culture in practice inside a biotech startup.

If the CSO or CEO is hand-waving around the scientific strengths of the story, and no one feels able to challenge them, what happens?  Real innovative thinking suffers. And Boards are often fed this pablum rather than something of substance.

Having multiple mediums for communicating can create the right cultural outlets for this type of safe “champion and challenge” culture, everything from town hall style company discussions, open dialogues within working teams, cross-functional employee focus groups, discussions with the HR lead, or even anonymous suggestion boxes. Yelling a lot in the big company meeting about a concern isn’t always the right forum, so folks should be encouraged to find the right one. But good companies don’t let intellectual dissension fester quietly without an outlet – it needs to be shared in some format or another. As Pisano writes, “Unvarnished candor is critical to innovation because it is the means by which ideas evolve and improve.”

Our own investment model at Atlas is aimed to fostering this “champion and challenge” culture. We’re a flat, equal set of investing partners. There is no emperor (with or without clothes). There is no one boss that squelches discussion. We openly and respectfully challenge each other, and base a lot of our decisions on mutual trust in the partnership.

Collaboration but with individual accountability.

This fourth observation from Pisano is exactly in line with what I described previously as: “There is no ‘I’ in team, but know who has the “D” (#6 in this post).

Team work, across different disciplines, is critical to success in biotech, probably more so than some other innovative sectors; integrating biology, chemistry, tox, pharmacology, clinical, manufacturing and many other functions over the course of many years. Collaboration is the lifeblood of the drug R&D endeavor.

But, as Pisano notes, “too often, collaboration gets confused with consensus.”  This is spot-on.  Consensus in a world of biological ambiguity is hard. Further, consensus can be a huge bottleneck on the timelines, as energy is spent trying to get all the stakeholders on the same page. Eventually, an individual just has to make the decision. I called this the “D” in my post on culture. Someone needs to integrate the inputs but make a decision. And then be held accountable to that decision.

Biopharma R&D timelines often make real accountability difficult, however. Take decision like selecting one path or another, such as nominating a specific Development Candidate vs waiting for another to emerge, or picking an initial clinical setting vs another. These decisions take years to manifest into outcomes, positive or negative, to be able to reflect on whether they were the “right” decision. Sometimes one never really knows whether it was the best decision or not as science doesn’t always reveal itself that way. With timelines this long, ensuring the proper information, discussion, and challenge were integrated into key decisions is a big part of the accountability equation.

Endless, often headless, committee meetings in big organizations can crush accountability by de facto forcing consensus decisions. The flip side of accountability in these larger organizations is the “CYA” culture (“cover your a—“): if I have to put my name on this decision, I’m going to make sure 99.9% of the ambiguity and uncertainty is wrung out before making it. Innovative cultures can’t be run with 99%+ confidence intervals around decisions. Risk is essential, and working under 60-75% confidence is much more in line with nimble, smart risk-taking. Decisions need to be informed better than coin toss, but often the risks and unknowns remain significant.  Understanding how to make innovation-enhancing decisions as individuals and teams in this environment is crucial.

Lastly, consensus frequently kills the outliers. Within venture firms, there are countless examples of firms passing on seed-stage deals because they were crazy ideas and consensus didn’t exist, only then to watch them grow into unicorns. Nearly a decade ago, Atlas changed its approach to seed investing in that any individual partner could back a “seed deal” (size delimited) without consensus support. This empowered partners to take some risk. As those investments mature, we generally move towards a team consensus as capital intensity increases. But at the earliest phases of the decision process, when uncertainties are high and getting to consensus could prevent leaning forward into a “risky” seed, we think this approach works. For Atlas, mutual trust and respect often gets you to implicit consensus in the end.

Flat but strong leadership.

Organizational charts are fine to capture the structured layers of a company, which is important.  Who is reporting to who, and who is responsible for what functions.  How big are certain functions.  And using them to forecast how the company will look in 1-3 years is a useful tool. All this is valuable.

But the org chart only tells you about structure.  It does not reveal “cultural flatness”, as Pisano calls it, that reflects the level of empowerment around ownership. Culturally flat organizations often push decision-making closer to those with first-hand knowledge of the critical information.

It also means that C-level executives can’t sit in their corner office and make decisions in a vacuum.  To quote Pisano, “For senior leaders, it requires the capacity to articulate compelling visions and strategies (big-picture stuff) while simultaneously being adept and competent with technical and operational issues.”  This is the essence of a great biotech science leader in my mind: able to frame the big vision of a 60K foot view while also being able to engage at the 6 inches view of a critical experiment or study design when appropriate.

In short, cultural flatness goes both ways: C-level executives being versatile with the details of the programs and platform, with researchers being able to engage their leadership with contributions beyond the hierarchy.


Innovative cultures don’t just emanate from cool Patagonia vests.

Building and leading this kind of culture in a startup biotech isn’t easy. It requires a mix of seemingly contradictory behaviors, like the tolerance for failure but not incompetence. And the pieces of culture outlined by Pisano aren’t really an a la carte menu – they all need to be integrated into the DNA of a startup in order to encode for a high performing culture.

Summed up, and repeating the phrase from the beginning of the blog, Pisano nails it: “These cultures are not all fun and games.”


Getting Clinical On New Drug Launches

Posted May 13th, 2019 in Pharma industry, R&D Productivity | Leave a comment

The IQVIA Institute released a data-rich report in April 2019 on the state of clinical R&D in the US pharmaceutical industry titled “The Changing Landscape of Research and Development: Innovation, Drivers of Change, and Evolution of Clinical Trial Productivity”.  The report is a fascinating deep-dive into lots of data cuts around what’s going on.

The increasing role of biotech in marketing their own drugs was the subject of much of the coverage on the report (here, here), which is indeed an interesting trend to see.  But rather than opine on that observation or the full scope of the report, I wanted to highlight a number of findings, largely glossed over elsewhere, relevant to the state of clinical trial activity and approvals that are relevant to biotech investors.

With clinical trial activity in the industry up 35% over the past few years, with over 4700 trials starting in 2018, it’s worth looking at some of the clinical trial metrics, especially since the features of clinical trials, such as the number of arms, numbers of patients, and years of trial follow-up are major drivers of costs.

Before diving in, some context on the data in the report: IQVIA tracks all the New Active Substances (NASs) that were launched in the US in 2018 (irrespective of what year they were approved). This includes 59 NASs for 2018, roughly half of which were orphan disease drugs. Here are a few of the key findings as they relate to clinical trial activity.

Features of the clinical trials in the regulatory approval packages:

  • Randomized controlled trials (RCTs) clearly remain the gold standard for approval: nearly 90% of drugs included RCTs in their regulatory packages.
  • Trials with active control arms are more common today, and were included in the regulatory filings of nearly half of these new drugs (46%), versus only 20% a few years ago. This highlights the increasing importance of comparative effectiveness in disease indications where existing standards of care are in place.
  • Only one registrational trial was required in over 40% (25/59) of these approvals; so much for the old “two well-controlled Phase 3 trials are required” regulatory guidance of the past.
  • Only Phase 1 or 2 studies did the trick in 12% (7) of the drugs’ approvals; these had no Phase 3 studies, so are obviously special situations.

Magnitude of patient exposures and time. 

  • There’s a huge range in the number of patients in these regulatory filings, as captured by the distribution below which was derived from the report. Roughly half of the approvals had fewer than 500 patients treated.

  • On the small end of the distribution, five drugs were approved with fewer than 100 patients of data. All were, as expected, orphan drugs. Four of these were approved on the basis on only a single trial. The 5th was elapegademase for ADA-SCID, which was approved on the basis of only 10 patients in two registrational trials!
  • The average number of patient-years of exposure supporting new drug launches between 2016-2018 has been 1800-3800 patient-years. Patient-years of exposure is often an important metric for understanding safety in particular, but also durability of efficacy
  • Orphan drugs had, on average, around 1000 patient-years of therapy over that period. While smaller than the non-orphan drugs, as expected, it’s not as small as I might have thought (e.g., could be 200 patients followed for 5-years of treatment or 5-years post-cell/gene therapy)
  • Collectively, achieving this magnitude of patient exposures has required a significant amount of development time: the cumulative years in the clinic and registration remains long and hasn’t changed as the average in recent years is approximately 12+ years. This rough average time holds for both orphan drugs and non-orphans, interestingly.
  • Although registrational trials for orphan drugs often include far fewer numbers of patients (about 5-fold fewer, ~430 vs ~2300 patients enrolled), these R&D programs still typically require long durations: in fact, the average trial duration for orphan drugs is actually 12% longer than for non-orphans, based on the 4-year averages (7.6 years vs 6.7 years). This was a surprising finding, and presumably reflects the broad range of orphan indications beyond just rare monogenic diseases.

These metrics highlight important aspects of the current clinical landscape.  Integrating these with our experience as early stage investors, and a few key takeaways emerge:

  • Active comparators with RCTs, even for orphan drugs in crowded classes, will likely be even more common going forward. Understanding comparability versus standard of care as early in development as practical will be increasingly important. Trying to run trials without them to save capital is not a recommended approach; plan the right study and convince investors to fund it, rather than cutting corners.
  • Single arm trials, poorly controlled, will continue to face challenges – as we all know from history. In cancer studies, this often leads to false positives, and thus late stage failures over time.  When they are used in rare orphan diseases, there are frequently no relevant long-term natural histories to rely on, creating further concerns. In these settings, companies should invest early (while in preclinical) to set up observational clinical studies aimed at creating adequate historic controls and benchmarks.
  • Going forward, we should expect that drug approvals will likely require ~1000 patient-years of exposure for “mainstream” orphan indications, and at least 2-3x more than that in non-orphan drugs. Gene and cell therapies with dramatic effects can be expected to have shorter aggregate exposures, but benefit from a constant accrual of patient-years even after a one-and-done intervention.
  • While special situations exist, unless there’s dramatic efficacy in early trials, very few drugs have or will be able to jump from Phase 1 or 2 straight to approval. It’s worth skeptically pressure-testing any pitches from teams that claim this expedited straight-to-approval approach is their base case plan.
  • Orphan drugs, while more efficient in some ways, are not dramatically faster in clinical development than non-orphan programs, due to the need to accumulate patient-years of safety and efficacy data from smaller overall patient numbers.

Thanks to the team at the IQVIA Institute for a stimulating and thought-provoking read.