April 2, 2019 | Increasingly, the acquisition and digital analysis of health-related measurements is gaining merit in the clinical setting with the promise that artificial intelligence will increase the strength of diagnosis, prediction, and therapeutic efficacy.

Daniel Karlin, a board-certified psychiatrist, addictionologist, and clinical informatician with an academic appointment in psychiatry at Tufts University and prior experience at Pfizer, is pushing this envelope. His current work is with a company called Healthmode, where he is developing additional measures for clinical trials in partnership with pharmaceutical companies.

On behalf of Clinical Research News, Marina Filshtinsky spoke with Karlin about the impacts of digitization on clinical research.

Editor’s note: Marina Filshtinsky, a Conference Producer at Cambridge Healthtech Institute, is planning a plenary session on data-driven drug discovery and development at the upcoming World Pharma Week, in Boston, June 17-20. Karlin will serve as a panelist during the plenary session, sharing his knowledge of how health measure digitization can improve clinical research and development. Their conversation has been edited for length and clarity.

Clinical Research News: How has digitization of drug discovery and development impacted pharma R&D in general and specifically pre-clinical research and early clinical development?

Daniel Karlin: There has obviously been a ton of digitization of research in general; the sorts of basic science research that leads to drug discovery have been highly computerized, including computational models of receptors and receptor engagement, which can direct basic scientists to build compounds tailored to do the things we want with increasingly digitized ways of monitoring and measuring the likelihood of function with computational models. But in the early clinical space we've seen much less success with this.

There's been a chronic effort to try to incorporate digital into clinical research, but it's tricky. If you go back to the beginnings of these efforts, you see people including connected pedometers, like the earliest version of Fitbit, in clinical trials just because Fitbit was there and it seemed like a neat thing to add. They could add them, and they did them. None of this was badly intentioned, but there wasn't a whole lot of strategic robustness to it; there wasn't a much consistency in application, or clinical thinking about what we actually wanted to measure and the best way to measure it using these digital technologies. However, over the past few years, we've seen a real move toward thinking about this differently, thinking about how we can use digital in rational and logical ways to best measure the things we're interested in in the clinical development space. Though there are some early promising signs, with the exception of digital ePRO being administered on mobile phones, we haven't seen a whole lot of progress in the objective measurement of illness for regulatory or market access purposes.

How do technology and digital health fit in the overall strategy in the biopharma industry in day-to-day operations?

I think the level of digital integration varies a lot from one company to the next. There are some companies where digital is definitely more of a sideline and sort of sits apart from routine, day-to-day efforts. Integrating digital can be more difficult in those companies. There are other companies where digital is much more central to the overarching strategy, and is thought of as a standard part of clinical development, something that all teams should be striving to do. The way digital is organized in a company differs very much between companies, and the importance placed on it varies significantly. There really is no standard here; everything is quite different depending on where you look.

It seems that CNS diseases and neurology and psychiatry are the leading therapeutic areas with regard to digital efforts. How can we cover more therapeutic areas?

One of the reasons that you could think that psychiatry and neurology seem like they’re leading here is because those are areas where it's been very difficult to develop new drugs, partly because it can be exquisitely difficult to measure the things we’re interested in. If it's difficult to measure things, then people are looking for new ways to measure them. It's not that we've seen any sort of overwhelming success in drug development in psychiatry and neurology. We're not exactly seeing tons of new mechanisms and new drug approvals, though over the past few weeks there have been some new approvals worth paying attention. We’re actually seeing that it is quite difficult to develop drugs in these areas. That difficulty may be driving some of that interest that you are describing in identifying and using new measures. If it's hard to do, then maybe we need to find different way to do it. On the other hand, the things that we think these measures are most capable of measuring (often what they actually measure) are behavioral outflow. So for diseases that directly affect behaviors, it makes sense that you want to measure the behaviors. I think the pressure is in both directions as you think about where you might use the measures. There are pressures both from where they can best perform and from where do we need the most help measuring new things. We may see that both are true in psych and neuro; we need the help, and it's the right place to be doing this.

What is the right balance of partnering externally versus development of an internal digital AI team?

I'll tell you when I figure it out. If I knew the answer, then it would save everybody a lot of trouble. I think though, at least what I'm experiencing while trying to do this from the outside, while not being in a pharmaceutical company, is that some of the very fast-moving technological development stuff is easier to do when you're structured as a small startup and harder to do when you're in a big pharma company. I'm finding that the sort of thing that would have been a challenge in my previous world, are definitely more doable for me in the startup environment. That's one data point, one bit of evidence that maybe there is something to be said for the environment I am working in now, but there’s got to be a balance between these players, and like all things where the incumbents are changing rapidly, the balance will take time to figure out. There are some roles that are best played by people who are experienced in their pharma companies, and there are some things that we as a startup are capable of doing much more quickly and much more efficiently. The real art of this is finding the right person, at  the  right entity for each task, and finding ways to build partnerships that play off the complementary strengths of the different participants. Rather than seeing it as all or nothing—partner or go it alone—the ideal seems to be finding ways to structure it so that everyone is doing the things that they are best suited to do within a partnership.

That's the goal of our partnerships. We are always thinking, “how do we align tasks and incentives as best we can so that each group and individual involved in the partnership accomplishes the things that it is best situated to do well and most efficiently.”