A Shot in the Arm - Jason Burke on Health and Life Sciences

Back in November, I blogged briefly about protocols in summarizing a presentation I delivered at Medidata's User Conference.  This week, Medidata's Glen de Vries returned the favor by kicking off SAS' annual SAS Drug Development User Connection meeting.  For those customers that attended, thank you (and Glen) for making it such a wonderful event.  The SAS team was extremely pleased with all of the great conversations, and I never cease to me amazed at the wonderfully creative things our customers do with software.  As before, Glen's talk reaffirmed our shared interest in seeing the progression of machine-readable representations of research protocols. 

If you don’t already know, there are at least 2 distinct meanings of a protocol in the health and life sciences ecosystem. In the world of drug development, a protocol is a specification for a research study design – it basically defines an experiment. In the world of clinical care, a clinical protocol is a recommended and prescribed course of treatment for a given disease state or condition. If you have a patient enter an oncology ward with skin cancer, a clinical protocol lays out how the regimen of care should be administered.

Though at first glance these two definitions appear quite dissimilar, when you look a little deeper, you’ll notice something interesting: the concepts behind protocols are actually quite similar. In fact, both markets are trying to improve their protocols with very similar objectives. The following table highlights some of the different concepts behind protocols in drug development and patient care, and how they map across the two market segments.

If you follow the evolution of each market’s protocol to its logical extreme, they could actually mean the same thing. Consider for a moment an adaptive drug trial. Let’s assume that the research protocol is designed in such a way as to incorporate incoming clinical observations to iteratively improve the efficacy of the study – a practice that offers both ethical and research benefits. If we kept that study running for a long time – say, 5 years – wouldn’t the current view of how to make the treatment outcome successful be an ideal basis for a clinical protocol?

If this were a new drug, this research protocol would offer the best insight available into how to drive the best patient outcomes. An argument can be made that ideally there should not be a distinction between a research protocol and a clinical protocol. An investigation into the safety and efficacy of a drug treatment would be an ongoing lifecycle-long process of aggregating data from around the world related to patient exposure and experiences with the treatment; continuously offering the best guidance available on outcomes while gathering additional information in real time to iteratively improve future treatments. It could even be said that the best research protocol design is actually a clinical protocol specification that accounts for the continuous improvement of clinical endpoints.

Some might argue that this is in effect what happens today, with researchers collaborating with medical thought leaders regarding research design, and medical practitioners sharing outcomes and experiences in standardizing clinical protocols. But I would counter that the way it happens today is not like what I am describing, and here is why:

1. Because of their current focus on drug safety and efficacy, research protocols today do not usually give equal time to investigating the factors influencing outcomes other than the drug itself. We have seen a gradual change in phase 3-4 trials in this direction, but more priority and attention is still needed.
    
2. Today, there are multiple levels of abstraction between what is uncovered in clinical trials, and what is actually put into medical practice.  I'm describing a process that removes that abstraction, providing more direct access to real-world experiences.  This distinction leads to...
   
   
3. The process I'm describing does two things not commonly done today: it places complete transparency between researchers and real-world medicine, and it creates a shared ownership of long-term health outcomes with both researchers and practitioners.
   
   
4. Clinical protocols are not modified based on strong analytical models per se – they are modified based on human experience. And whereas I am a strong advocate for the role of human experience in the practice of medicine, I am also a strong advocate of using real data to overcome limitations in the human understanding of biological processes.
   
   
5. There is a considerable time lag – years or more – between the generation of research outcomes and the incorporation of that information into clinical protocols. If analytics were an “in stream” activity – providing real-time insight into the performance of treatments across diverse population factors – patients could benefit more quickly from what is actually known about a treatment.  Note that I am not talking about clinical guideline rules engines increasingly found around EMRs -- I'm talking about predictive modeling and multivariate analysis.

Of course, I'm glossing over all of the things that actually make research and clinical protocols unique.  But my question is do those differences mandate that, in the world of electronic patient information and real-time communication, those two protocols must be separate?  Or could we combine them into something more useful for researchers, physicians, and patients?  What do you think?