Precision medicine (PM) is of growing importance in healthcare and was the focus of a major push by the Obama Administration through the Precision Medicine Initiative (PMI) launched in 2015. Chilmark uses Ferryman and Pitcain’s definition of precision medicine as the “effort to collect, integrate, and analyze multiple sources of genetic and non-genetic data and applying data analytics and machine learning/AI to develop insights about health and disease that are tailored to an individual.” With the growth in both biological knowledge and the ability to collect behavioral, social determinants and environmental data, the ability to customize therapies and care pathways in growing dramatically.
Metabolomics, proteomics, microbiomics and beyond are offering substantially more biological data and insights on unique trajectories of diseases. Furthermore, we are now seeing market penetration of wearables grow significantly which raises the possibility for more precise behavioral data and medical-grade vital signs monitoring offers even more analytical opportunities to assess risk and monitor the progression of chronic diseases. Add the growing number of social determinants data analytics offerings and our ability to match patients with not only appropriate biopharmaceutical interventions, but even broader care pathways and social supports is beginning to improve dramatically. More granular data means more finely tuned ways of performing risk stratification for assignment to different care regimes.
Translating the tremendous growth in data into clinical insights falls into the hands of AI/ML platforms. The rapid growth in investment in AI and cloud computing are beginning to create the foundations for the PM market of the future, but we still have quite a few years ahead of us until it becomes a mainstream clinical practice.
However, the rapid growth of data sources and types of data does not automatically translate into improvements for the clinician and patient at the point of care. The burden of work in managing this huge influx of data is growing. There is little doubt that AI and the cloud will bring a great deal of computational power to health systems, but we are still early in the journey towards PM becoming a mainstream reality. There are a number of limitations in the current health IT and EHR infrastructure that make the clinical realities of PM challenging.
First, integrating genomic test data into EHRs is a bit tricky. Historically there has been a lack of structured data and a consensus on the best way to integrate genetic test results into the EHR environment. The Sync 4 Science (S4S) program that is part of the PMI is utilizing SMART on FHIR to integrate data from EHRs and the All of Us Research Program . There is also a lack of storage space, but perhaps even more importantly, integrating genetic test reports that can be 20 pages long in the context of clinical decision support has not been an easy task. Making genetic data clinically actionable can be difficult when the science has significant degrees of uncertainty. The eMERGE Network across over a dozen academic research centers has been making a great deal of progress in recent years in addressing the challenge of integrating genomic data into EHRs. Some of the remaining issues include how to best handle genetic knowledge that may also impact a patient’s family who may also need to be tested for similar genetic conditions. As new insights in basic science emerge, the meaning of test results in the past can change over time. This can result in very different conclusions from the original context when the test was taken and create uneasy clinical encounters for patients and clinicians.
The rapid growth of data sources and types of data does not automatically translate into improvements for the clinician and patient at the point of care.
In our upcoming Precision Medicine Insight Report, we take a look at a number of companies from Syapse to Flatiron Health to see how -omic data is being integrated and the new tools that are becoming available to clinicians, researchers and patients to find the best therapeutic option or clinical trial. We look at platforms for aggregating data and translating genetic testing into the clinical context as well as platforms that match patients to the best therapeutic options and trials.
There has also been some tension in the medical and public health spheres over the tendency of PM to focus on the individual at the expense of public health. Advocates for precision medicine have tended in recent years to construct the emerging field as a continuum from the individual to precision public health. This explains why the All of Us Research Program has had a strong emphasis on minorities and women in their work. If we take a close look on what actually drives outcomes at the population level we see that interventions for cancer, at least, that focus on public health approaches can account for at least 40-50% of the prevented cancers. PM has the potential to identify population cohorts sharing similar characteristics that may need more customized public health approaches in the future. Moreover, we will likely need to see a balance of some rare diseases where the emphasis may be on focused interventions on the individual. For diseases where social determinants and behavior weigh more heavily, we will see a mix of precision medicine and public health approaches that may draw upon insights from microbiomics, for example, but scale via public health interventions.
This brings us to the perennial question of who is going to pay for precision medicine in order for it to scale? This question is going to pose some significant challenges given the business model of biopharma which has historically focused on one-size-fits-all approaches and now must demonstrate robust outcome data for drugs for much smaller market segments based on -omic data. This issue becomes even more complex when we factor in the need to build the evidence base for smaller cohorts. Already we have seen some value-based contracts for drugs based on the associated outcomes and how different cohorts of patients respond. Tiered pricing models exist corresponding to relative outcomes. Others in the re-insurance industry are looking at how more cooperation around data sharing will be needed as well as the creation of “cure funds” that the insurance industry creates to cover the cost for prohibitively expensive treatments.
It is clear that both pharma and payers will need to come together to chart a clear path to lower costs and risks across the spectrum. This means better leveraging AI to bring down development costs, clear signals on what will be paid for, and better engagement with patients to collect better data in an environment where trust is lacking. These challenges along with the need for clinically relevant data and ways to integrate the insights from a rapidly expanding data ecosystem means we still have a long way to go until PM is mainstream across clinical practice. Trust is also a central element that is going to take time until both patients and clinicians see the fruits of PM tools come to reality.
 Formerly known as the PMI Cohort Program, the program was developed as part of the PMI to support long-term research with data from a cohort of one million participants. It was renamed as the All of Us Research Program to reflect the diversity of the country but is still frequently referred to as the “cohort program.”