Precision (aka personalized) medicine (PM) describes prevention, diagnosis and treatment strategies that take into account the variability in genes, environment and lifestyle for each individual. While PM having already made an impressive impact on some fields of medicine, particularly oncology, the epistemology of precision medicine and its relationship to evidence-based medicine (EBM) has not received much attention.
To date, PM has focused almost exclusively on individual variability in genes, apparently leaving environment and lifestyle for later. The rapid progress in understanding basic mechanisms in genomics and proteomics, however, has already led to targeted therapeutic options for some individuals with diseases such as cancer or cystic fibrosis. The availability and low-cost of genetic testing, including full-gene sequencing, is providing individuals with a vast amount of personal information. Understanding the significance of such information requires a broad array of methodologies.
Several of the methodological tools of precision medicine fit well within the standard framework of EBM. For instance, the use of large databases allows for the exploration of associations between some genetic variants and disease. Similarly, genome wide association studies can be used to search for genetic variants that may be associated with a trait of interest. These epidemiologic approaches are helpful when the frequency of particular genetic variation in the population is relatively high.
Over half of the genetic variation in individuals, however, comes from single-nucleotide substitutions found in only a very few people, generally close family members. These variations are not amenable to population-based studies, given the extremely low allele frequency. When such changes are detected during genetic screening, laboratories report them as “variants of uncertain significance.” Such a designation provides no useful information to individuals who harbor these variants, often manifested in a gene where other variations are known to be associated with cancer risk. To aid persons with such variants, a variety of methodologies based upon mechanistic understanding and reasoning have been developed and deployed to help classify variants as either likely benign or likely pathogenic.
In some circumstances, an understanding of the effect of a variation can be made based upon mechanistic reasoning alone. For instance, a novel variation that produces a stop codon in the gene for CFTR (when the other allele also contains a CFTR CF-related variant) will certainly result in cystic fibrosis. (Even if a population-based study could be performed, it would be completely unnecessary.) In other cases, categorizing an extremely rare variant as pathologic or benign may be possible with the information gained from computational analysis and/or functional assays. Such methods rely upon the understanding of mechanisms, including principles of evolutionary conservation, the effects of similar mutations, and the impact of protein length, configuration and function changes. As the understanding of basic mechanisms improves, the predictive power of such tools improves as well.
Mechanistic reasoning also adds value to population-based evidence in PM. Associations found in large data bases or by genome-wide association studies are considered more likely to be of clinical significance when the occurring in a gene known or thought to be in the causal pathway for the trait or disease being examined. PM often involves an iterative process, with epidemiologic observations promoting specific mechanistic studies and the understanding of mechanisms leading to a search for difference-making evidence in particular genes.
The need for methodological pluralism on the part of PM is not surprising, as any strategy calling for the individualization of care cannot possibly rely upon population-based evidence alone. In the case-based reasoning required of precision medicine, mechanistic reasoning may be more compelling than population-based evidence, say for targeting specific tumor mutations rather than basing decisions on studies that classify tumors based upon tissue of origin. Regulators in the US have begun expanding the use of such targeted therapies (E.g. for cancer and cystic fibrosis) on the basis of mechanism, waiving the requirement of standard clinical trials.
PM requires evidence derived from mechanistic reasoning and methodologies. This requirement will only increase as factors related to environment and lifestyle are incorporated. Such evidence, however, continues to appear only on the lowest rungs of evidence hierarchies promulgated by EBM. While some have suggested that EBM and PM can be viewed as complementary to one another, such a call fails to acknowledge the inherent dissonance that arises when the two overlap. Rather, EBM and PM can co-exist in clinical practice only if EBM abandons its hierarchy of evidence and adopts a more horizontal approach to understanding the value of mechanistic knowledge alongside population-based studies. This call is certainly not new, it is a cornerstone of the EBM+ project, but the advent of precision medicine adds another argument for acknowledging the value of knowledge from a variety of sources and methods for clinical decision making.
University of Washington, USA
Clare Hall, Cambridge, UK