By DAN HOUSMAN

We have been talking about Precision Medicine for a long time now but so far we are still in the infancy of using genetics to impact medical decision making. The human genome was sequenced in 2003,with the promise of rapid medical advances and genetically tailored treatments.  However, development and adoption of these treatments has been slow. Today with the advent of large cohorts, and in particular, the construction of the US Government’s Precision Medicine Cohort,[1] conditions are being set up for precision medicine to flourish.  In the PMI infographic,it states three reasons for ‘Why now?’ – sequencing of the human genome, improved technologies for biomedical analysis, and new tools for using large data sets. While I agree this is progress, I believe there area few fundamental other areas to be tackled in order to really get to the promise of precision medicine.

1. The concept of protocols is designed for mass production not mass personalization

Medicine is practiced using protocols and documented in EHRs. When written down they can look like a cook book or a choose-your-own-adventure story book. The intent is to codify medical knowledge into a guide that can be consistently used by all physicians to obtain the ideal outcome. But as a result of strict adherence to medical practice standards, they inherently choose paths that are well defined based on the assumption that most people are similar in their response to treatments. So over time the protocol is enhanced to suggest which decision is the right one to make with a given patient and a protocol will more often than not be ‘anti-precision,’ just in the same way that a factory is designed to make one size of jeans at a time rather than make a custom set of jeans for each customer visiting the store.

Protocols are analogous to laws. There are many parties, normally ones selling interventions or diagnostics, interested in wrestling for the control of a part of the protocol in order to optimize their own benefits. So like the laws, the number of places where a protocol can be diverged is dependent on the body that is building the protocol being able to deal with the divergent inputs from the many interested parties providing their evidence of why the protocol should include their suggested adjustment to improve patient benefit. There is nothing nefarious in this situation, as a better protocol will lead to a better set of patient options. But the model is just much more conducive to a small set of interventions and a small set of conditions being reviewed slowly to approve the changes in major paths than a precision model.

This paradigm of the protocol needs a major refresh to accomplish the goal of making a transition to a new world of precision medicine. For example,in HER-2 positive breast cancer patients we have learned to use Herceptin because it is a well understood variant and treatable because of the biology of the drug and cancer cells. But the Herceptin test is only a small part of cancer diagnostics and precision treatment options. There are 20 or possibly hundreds more key targets to test with increasing combinations looking more like signatures and combination therapies that are not as binary as a single test.As we start to tune-up the number of potential interventions and number of potential variations among patients, we are getting into the problem that having a simple decision tree quickly breaks down into complex graphs that may be impossible to document as a protocol. Furthermore, the body of evidence needed to support each decision becomes limited relative to published statistical ‘proof’ standards we have historically had in order to pick which path to take.

So – in short – if we want to have personalized medicine we need to figure out how to make a quantum jump that some industries have made such as media – from mass production – newspapers and books, to mass personalization –in the context of the current structure of regulatory science and ethics. We need to invest in more people with regulatory science and medical ethics to sort this out to figure out the right model since getting to bigger data only exacerbates this issue.

2. Economic realities of precision options

While there is a clear opportunity to provide a better service through mass personalization, there are some big financial issues to resolve for any first mover offering an intervention in the space. The current economic model for drug development still requires a drug to have a return on investment in the range of billions of dollars. This means that creating diagnostics that can significantly limit the potential addressable market are still economically disadvantageous. Now it is fair to say that it is unethical to not know the precise conditions when a drug will and won’t work. But for as long as it is impractical and difficult to answer to the question of precisely when a drug will not work, there are incentives to not invest heavily in seeking findings in this area without a regulatory group requiring it with clear guidelines and associated incentives or a ‘disruptive’ set of players in the market.

Currently,no one has found a way to disrupt the current pharma business model by identifying when drugs will not work and paying for that research with a diagnostic that is reimbursed to do so. So if I can sell tests that determine which 3 of 10 people will not benefit from a drug through a precision medicine diagnostic– in theory it is an ‘ounce of prevention’ for waste and should obtain the cost benefits of waste avoidance. But no break-out story about the AirBnB or Uber drug diagnostics company cutting down the cost of drugs in this way has been established. It hasn’t really happened yet. If such a business model were to become quickly dominant like these consumer disruptions, then drug development would need to assume that when there is variation in response, that it will be rapidly identified by such groups and capitalized on by the third party and not the drug maker. But that is not yet the current world we live in. Maybe it is going to change in a value-based care world where the total cost of care can create the opportunity to add this new ‘diagnostic’ player, but so far value-based care is not deeply penetrated enough into the overall model of US or global care. For now it takes a new reimbursement code to be able to introduce such tests and the value is hard to justify. My guess is that while the research could be done, the valuation of the tests are not sufficient to justify the kind of scale investment needed to identify the variants and the organizations with the capacity to do so are better suited to create new interventions.

3. Big Data technology companies involved in precision medicine aren’t medical device companies yet

Currently we have a number of technology companies that are ‘using Big Data’ to be able to support decisions. A big data cognitive engine is needed to break through the protocol driven system to formulate a model driven system where the decision path is more opaque but technology can still provide the right recommendations based on the body of available collected genetic and clinical evidence. As of today,systems like this haven’t been built and approved for broad medical use yet.One of the reasons these systems don’t exist yet is because such a system is a medical device and should comply with the same rules.We have seen the sorts of trouble that have occurred from the disposition of “medical device” being applied to Silicon Valley generated innovations.The regulatory world may need to catch-up to the capabilities and model of the newer technology, and the newer technologies will also need to admit that they need to modify their operating model to acquire FDA approvals to be trustable and involved in the liability of medical decision making.

Conclusion

While my view of the situation is sobering, I am optimistic about the transition from current medicine into a precision medicine world. It seems that most of the barriers have fallen from a technical point of view. We just need to work on the areas that are the key drivers. Some of these will get solved by market forces but others are still in the hands of regulators and leaders of the medical community managing the creation and diffusion of protocols. Big data businesses will get smart and involved in approving their systems as components of medical devices or they will not be players.New players will soonemerge that focus on reducing the waste in prescribing medicine that doesn’t work. Physicians will figure out within their societies how to transform their operating models including protocols and compensation structures to enable precision medicine. Let’s get this done.

[1] (https://www.nih.gov/precision-medicine-initiative-cohort-program)

Dan Housman is CTO at ConvergeHEALTH by Deloitte