7 January 2016
The aim of so-called precision medicine is to improve the safety and effectiveness of healthcare by targeting interventions to individuals in whom they will be most effective, thereby avoiding unnecessary and potentially harmful interventions and possible adverse reactions from inappropriate dosages. In theory this targeting can be achieved by characterising what makes each of us ‘individual’ by measuring a range of physiological parameters including (but not limited to) our genome sequence, and then using this knowledge in combination with what we know about how the effectiveness of different interventions varies with these parameters to decide on the best course of treatment (or prevention).
The success of precision medicine depends strongly, therefore, on being able to accurately relate variations in our genomes to variations in disease risk or treatment response, and as a paper published this week in JAMA shows, we are still some way from achieving this goal. The findings illustrate clearly two of the major hurdles that are yet to be overcome in transforming the promise of genomics into the reality of precision medicine. The authors show, using an innovative electronic medical record data mining approach, that in an unselected population of patients there are a significant number carrying potentially pathogenic variants in genes known to be associated with potentially fatal cardiac arrhythmias, but who have no reported clinical signs of arrhythmia.
This disconnect between the presence of potentially pathogenic mutations in known disease-related genes and the presence of clinical signs of the corresponding disease is not new, but this study nevertheless gives those of us who remain optimistic about the future of precision medicine two particularly important reasons to pause and reflect:
As the PHG Foundation and many others have noted, the quality and completeness of the information from which predictions of the pathogenicity of genomic variants are made leaves a lot to be desired. Furthermore, variation in laboratory protocols for annotating variants, even where the same underlying knowledge base is available, also contributes to a significant degree of variation in interpretation between laboratories. Deciding whether a genetic variant is or is not likely to be harmful is rarely easy, even for experts, and until enough information is amassed on other individuals with the same variant and their clinical status a definitive answer may be impossible.
The bottom line: Precision medicine does not have much of a future until the knowledge on which it is founded becomes both more accurate and complete. Achieving this will require a huge research effort to characterise the functional consequences of the many rare genetic variants of unknown significance that may contribute to disease (gene editing techniques should help here), and an equally huge effort to curate and ‘clean up’ existing databases on which the currently shaky foundations of genomic medicine are being built.
Clinical geneticists can be relatively (although not absolutely) confident about interpreting the potential pathogenicity of genomic variants when they are being analysed in the context of known family history and patients presenting with phenotypes that ‘fit’ the known disease-association of the genes being tested. However, one of the goals of precision medicine is to go beyond genomic diagnosis in the people with detectable disease and to enable personalised disease prevention, by using genomic and other biological information to predict and mitigate future disease risk in the absence of any overt clinical phenotype in the person being examined. In the case of the cardiac arrhythmia genes described in this paper, this could mean offering an implanted cardiac defibrillator to people carrying ‘pathogenic’ mutations to mitigate their risk of sudden cardiac death. This is an invasive and costly measure.
If, as this paper clearly suggests, it is not currently possible to predict accurately which variants in these cardiac arrhythmia genes are genuinely pathogenic, it is likely that attempts to do so will result either in patients being falsely reassured that they are not at risk, or falsely advised that they (and potentially their family members) are at risk of sudden death and require a major medical intervention to mitigate this risk.
The bottom line: Clinicians and researchers involved in the return of so-called incidental/secondary/additional findings from whole genome analysis, where disease risk is being predicted in the absence of clinical phenotypes, must proceed with extreme caution, particularly where invasive and potentially harmful interventions may be recommended as a consequence.
While the sceptics will seize on these findings as evidence that precision medicine as a concept is doomed to fail, this interpretation is overly pessimistic in my view. There is nothing in the paper / research that shows that genomics cannot in principle be used (in combination with other clinical information) to better characterise individuals and improve management of their existing diseases and their future disease risks.
What the paper does remind us is that there is still a lot of work to be done to make essential improvements to the knowledge base on which the ‘precision’ part of precision medicine is founded, and even more work to be done to improve the systems for sharing and analysing that knowledge across health systems, to ensure consistent delivery of safe and high quality information and interventions to patients. As ever, the trick is to balance the optimism necessary to inspire and drive forward this work, with the realism needed to manage the expectations of those people, be they patients or professionals, who are so eagerly anticipating the many promised benefits for health of the precision medicine era.