By DAVID SHAYWITZ, MD (2)
I’m deeply skeptical that I have much knowledge to impart to Harvard Business School (HBS) students. After all, they’re the ones clever enough to pursue a two year advanced degree (“six months of education crammed into two years,” they joke), while across town, my classmates and I ran gels, plated cells, memorized structures, and took call for a decade or more (in some cases) — and all for the privilege of eventually working for our fleece-vested colleagues (see also this 2011 Scott Gottlieb piece, and my 2012 Forbes post).
Even so, I was recently invited to appear as a guest on a new podcast out of HBS called “Under The Datascope,” where I answered questions about my experiences and perspective as a physician, scientist, technologist, drug developer, and investor. The episode (here),released today, is part of a series hosted by Gabriel Eichler and sponsored by the Kraft Precision Medicine Accelerator (Go Pats!) at HBS, featuring interviews with people working on and thinking about data, analytics, and precision medicine.
There’s a lot of content packed into the nineteen minute episode, and I thought it might make sense to capture some of the highlights – though I suspect the entire episode, and the series more generally, is likely to be of interest to readers.
Biomedical entrepreneurs drive science into durable application. After struggling during my clinical and research training with the persistent gap between promising science and clinical application, I came to appreciate that biomedical entrepreneurship represents the distilled essence of the translational impulse. (See this 2005 Nature Biotechnologycommentary, for example, this related version that was published in the San Francisco Chronicle, and this and thisfrom Forbes.)
Biomedical entrepreneurship requires humility and humanity, not tech fetishization and solutionism. Driving science into application requires not only the best (more precisely, the most suitable) technologies that are available, but also a deep sense of, and respect for, the complexities of biology and what I described as the “humanistic center of medicine and patient care.” (Regular readers will recognize this as a recurrent theme of this column — e.g. this 2011 post, “What Silicon Valley Doesn’t Understand About Medicine”).
Good doctors have always customized care. The mantra of precision medicine – “right drug for the right patient at the right time” – is not a radical new idea, peculiar to the molecular age. Admirable doctors have long tried to individualize treatments based not only on the biology of disease, as best it could be understood, but also based on the physician’s knowledge of the patient’s circumstances and preferences. It’s also critically important not to be excessively reductionist, and to recognize a person isn’t just the sum of their molecular mutations; everyone exists in a much broader context. See hereand here as well.
Beware of turning “check engine light” into memento mori. Ideally, technology can help capture and more deeply dimensionalize a patient’s experience with health and disease – including relevant environmental factors. At the same time, it’s important to appreciate most patients don’t want to be defined by their disease, nor do they want to be constantly reminded (by technology, for example) that they are sick – or could be. It’s a tricky balancing act. (See this characteristically on-point post by Lisa Suennen.)
Beware the first-mover disadvantage in digitalization.Management consultants have persuaded pharmas that they are (all!) seriously behind on their “digital transformation journeys,” which in practice creates tremendous pressure internally, as I’ve often discussed in this column (eg here). In short, there’s an edict coming down from the C-suite saying “we must digitally transform,” and then there’s the lived experience in the trenches, where people are trying to get their regular jobs done, and are suddenly confronted with the prospect of needing to allocate a lot of additional effort to achieve organization digital transformation objectives, often without much confidence that there will be palpable and personally meaningful benefits coming out the other end. I suspect this is what John Browne (former CEO of BP and author of a book I recently reviewed in the Wall Street Journal) had in mind when he cautioned against reflexive early adoption of new digital tech, pointing out that the tech often gets better and more reliable over time.
Horses for courses. The challenge with new technologies – like AI – is figuring out how and where to selectively apply them, and also to have appropriate expectations. The idea of sticking all your data in a pot, adding AI magic, and having it extrude a magical cure, for example, is an alluring fantasy. On the other hand, I am excited about opportunity for technology to robustify a lot of experimentation, and help generate more rigorous and replicable data sets – which, as I understand it, is a key part of what companies like Recursion and Insitro are doing. (You can hear Recursion’s CEO on Tech Tonics here, and Insitro’s here.)
Miracle Max: R&D head. It’s helpful to keep mind that even today, the fundamental problem faced by pharma companies is that they have no idea where their next major medicine is going to come from. Not only do most things fail, but when something actually succeeds, it’s viewed by senior management as a “miracle” – that’s precisely how both Merck’s Roger Perlmutter and Novartis’s Vas Narasimhan have described coming up with a successful new medicine. It’s kind of remarkable, when you think about it, to consider that we are working in what remains, at its core, a miracle-based business. This certainly highlights just how limited our existing understanding is (as Taleb and I discussed in the Financial Times here).
Focus on participation and outcomes. Important opportunities for digital and data include (a) improved clinical trial participation in the fullest sense of the word – efforts that bring more trials to patients, and efforts that bring more of the patient to the trial; (b) leveraging real world data not as a substitute for randomized controlled trials (important as that possibility is in certain contexts), but rather as a mechanism of routinely assessing what actually matters most – real world performance and real world outcomes; I discussed this topic in this Clinical Pharmacology and Therapeutics commentary (open access). I’m confident the opportunities (and challenges) around real world data haven’t escaped the notice of the FDA’s Deputy Director in particular.
Most biopharmas aspire to create game-changing medicines, and desperately want to better define the patients most likely to benefit from them. I’ve been incredibly struck by the commitment of biotech and pharma researchers to develop novel, transformative therapies. From what I’ve seen, virtually all companies would love to pre-define exactly which patients would respond to a particular medicine, as this would result in large effect sizes necessitating comparatively small clinical trials, and would minimize the exposure of patients to experimental therapies unlikely to offer benefit.
Ars longa, vita brevis, and your tech solution is unlikely to “solve” drug R&D. I remain deeply skeptical about most tech assertions I’ve heard around solving, or at least meaningfully impacting drug development, as there are a huge number of discrete and often lengthy and expensive steps involved in making a drug, and even if you believe you’ve fixed one (generally early) problem there remain a slew of additional hurdles. (See this and everything else Derek Lowe has patiently written on the subject.) Pharma may be onto something by adopting the view that, basically, rather than evaluate and license or acquire your particular magical black box, why don’t you use it to generate compelling medicines, validated in the customary ways, and if the early drug data look good, we’ll pay you properly for the very real value you’ve generated, the same way other startups who develop promising early medicines are paid. And if you think you can generate great new drugs ten times as fast as conventional approaches – do it, we’ll pay you appropriately for each and every one. I’m sure if one or another of these approaches truly are able to generate real drugs much faster, then the platform will be acquired – unless of course by that point they’re the ones who acquire pharma’s manufacturing and distribution engine (and make no mistake: it would be inordinately satisfying to see that happen)!
This is what progress looks like; the answer isn’t to apply the rule of holes and stop digging. In his unnecessarily kind introduction of my i2b2 keynote in Boston earlier this year, Zak Kohane described me as a “lucid optimist,” explaining that I recognize the risks and challenges of technology, but in the long run, I find the opportunities compelling. He’s right. While I’m deeply cognizant of the complexities of biology and medicine, and the naive hype around a lot of tech, I’m incredibly optimistic, long-term, about the impact technology can have – indeed, has already had – on the development of therapeutics. Consider – as I’ve recently discussed – what has now become routine in discussions of cell therapy, techniques that would have been considered fanciful aspirations just a few years ago. I’m equally convinced that ultimately data and digital will profoundly impact medicine.
Tech: transformative, eventually. In my view, the two key lessons from the history of technology are (a) new technologies are critically important for advancing the science, and radically altering the questions we can ask, and the approaches we can contemplate; (b) it takes a very long time (as Brynjolfsson and McAfee point out in Second Machine Age, and as Byrnjolfsson reinforces in the Wall Street Journal, here) to figure out how to actually use a new technology, and to develop the complementary inventions and business process improvements required to capture effectively the benefits the new technology enables. (I’ve also frequently discussed this point in the context of implementation – see here, here, and references therein. This topic is discussed in depth in Learning by Doing, by James Bessen, and in The Nature of Technology, by W. Brian Arthur; a classic paper in the field is Paul David’s The Dynamo and The Computer, available here.)
Bonus: in honor of the Krafts, here is a bonus recommendation for health entrepreneurs: a captivating documentary (even if, for some reason, you’re not a fan of either the Patriots or of Tom Brady) called “The Brady 6,” highlighting Brady’s unexpected rise to superstardom by focusing on the career trajectories of the six quarterbacks drafted ahead of Brady in the 2000 draft (he went in the sixth round, and was the 199th pick overall). It’s an incredible story of persistence, determination, a bit of luck to be sure — and the importance of never counting yourself out, whether you’re an overlooked college quarterback or an aspiring entrepreneur. Perhaps we shouldn’t be surprised to see these qualities in Brady — after all, he grew up right here in Silicon Valley.
David Shaywitz, MD is a physician, investor and a partner at Takeda Ventures.