On March 5-6, 2014, I attended the 7th annual Future of Genomic Medicine (FoGM) conference [PDF 778.13 KB], hosted by Dr. Eric Topol at the Scripps Translational Science Institute. The audience included more than 500 participants from various fields including genomics, clinical medicine, laboratory medicine, industry, economics, social and communication sciences, patients and the press. The gathering featured a cast of outstanding presenters and panelists including Craig Venter, former Vice President Al Gore, the patient from the future [PDF 778.13 KB] and many others [PDF 778.13 KB]. This was an enthusiastic audience of early adopters who are working tirelessly to make the vision of genomic medicine a reality.
In my talk, I emphasized the current reality in this rapidly emerging field, including the need for both population and individual level perspectives to improving health, an evidence-based approach and the importance of collaboration between health care and public health. Nonetheless, the excitement at the conference for the bright future of genomic medicine was palpable with strong calls at the meeting for big data and implementation of whole genome sequencing, including the recently launched Craig Venter initiative, a private sector project to sequence thousands of people to discover causes of human diseases and interventions to address them.
A week after I came back from the meeting, we had a “reality check” in genomic medicine. In the March 12, 2014 issue of JAMA, Dewey et al. reported on their experience of implementing whole genome sequencing in 12 adult volunteers. Their genomes were sequenced, analyzed and reported to a panel of physicians. They found incomplete coverage of genes associated with Mendelian disorders, low reproducibility of detection of genetic variation, as well as uncertainty about clinically reportable findings. Greg Feero wrote an accompanying editorial reminding us that the field is still in its infancy and that its adoption in routine practice needs to be done with utmost care.
Dewey et al. presented data on the analytic performance of two commercially available WGS platforms. They show that many of the genes recently identified by the American College of Medical Genetics as having potentially high clinical impact were not covered sufficiently to achieve clinical variant detection without further confirmation. The authors also presented relatively high levels of disagreement in variant detection calls, especially for insertion/deletions in protein coding regions and those that could be related to disease. These platforms are known to have poor fidelity for these types of variants. Manual curation of about a hundred genetic variants for each participant required a median of 54 minutes per variant, resulting in moderate classification agreement between professionals. The process also reclassified 69% of genetic variants catalogued as disease causing in mutation databases to variants of lesser or unknown significance. There is widespread recognition that currently available databases have major limitations for clinical sequence interpretation. Therefore, the average practitioner right now should recognize that interpreting clinical implications consequences of rare and new genome variations remains difficult and costly even among experts.
Dewey et al. also explored the potential for downstream consequences of providing clinicians with genome data in healthy people. They shared standardized summary reports with three primary care providers and two medical geneticists not involved in the genome analysis, and examined their suggested action steps. The panelists suggested 1-3 additional tests or referrals per person costing several hundred dollars. There were fair to poor levels of agreement among physicians regarding what variations should be acted on, and what the next steps should be.
The selected small sample size of the study does not permit large scale generalization but the study points to current limitations in the use of WGS information in clinical care (especially in healthy people with low a priori risk of disease or disease precursors). Significant research work clearly remains to develop primary data and evidence synthesis to support the use of WGS in routine clinical care.
So how can we reconcile the excitement I witnessed at the Scripps meeting about the future of WGS in medicine and its current less than optimal reality in practice in 2014? I think we can have our cake and eat it too. At the Scripps meeting I reviewed a recently launched CDC 3-tier classification system of genomic applications, Using this schema, it is clear that the use of WGS in practice today is still a tier 3 application, especially among healthy individuals as the Dewey study showed. On the other hand, the potential for WGS in diagnosing rare, single-gene disorders seems to be more promising. Based on a recent technology assessment by Blue Cross Blue Shield Association [PDF 210.54 KB], we gave it a tier 2 status (potentially useful for informed decision making).
In case you have not checked the list lately, there are a growing number of tier 1 genomic applications that can improve health and prevent disease in millions of people. Some of these applications are poorly implemented in clinical practice, with significant disparities across segments of the population. The public health imperative calls for the need for widespread implementation of tier 1 applications now, across all segments of the population and the need to marshal public heath-health care collaboration to do so. At the same time, tiers 2 and 3 applications, including clinical WGS, require a robust translational research agenda before implementation in routine practice can be accomplished in a way that assures benefits will outweigh potential harms.
Although Francis Collins’s prediction in 1999 that the human genome will be used routinely in clinical practice by 2010 has not been quite fulfilled in 2014, we are on well on our way to make genomic medicine a reality in the foreseeable future.