[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
Purchase Options:
[Skip to Content Landing]
Views 8,291
Citations 0
October 9, 2017

Knowledge for Precision MedicineMechanistic Reasoning and Methodological Pluralism

Author Affiliations
  • 1Departments of Medicine and Bioethics and Humanities, University of Washington, Seattle
  • 2Department of Laboratory Medicine, University of Washington, Seattle
JAMA. Published online October 9, 2017. doi:10.1001/jama.2017.11914

Precision medicine (PM) describes prevention, diagnosis, and treatment strategies that take individual variability into account.1 While PM aims to incorporate individual variability in genes, environment, and lifestyle, the emphasis in current practice is on personalized genetic profiling for diagnosis and risk assessment.

As genetic testing and interpretation advance, PM stands to move medicine away from the population-based knowledge that grounds evidence-based medicine (EBM) to the treatment of patients “based on a deep understanding of health and disease attributes unique to each individual.”2(p1842) Such understanding requires a different and broader concept of medical knowledge, the development of new methods for generating such knowledge, and approaches for incorporation into clinical practice. As PM advances, for some decisions it will replace the population-based “best evidence” of EBM with specific and detailed understanding of what makes an individual patient different from others. To practice PM, clinicians should reconsider current notions regarding the relative value of evidence, as case-based reasoning and understanding of mechanisms will figure more prominently.

The Importance of Variants of Uncertain Significance

Population-based data will remain important for informing current understanding of health and disease, but the nature of genetic variation means that it can no longer be seen as sufficient. Most actionable genetic variation in individuals derives from extremely rare or even unique variants. The average individual will have about 50 genomic mutations not present in either of his or her parents, most of these variants will not change protein function, and about 200 protein-coding family-specific variants inherited from relatively recent ancestors that are not present in variant databases.3 Such variants are generally designated as a “variant of uncertain significance” (VUS). In a medical model that prioritizes and relies on knowledge derived from population-based studies, these variants might be considered variants of unknowable significance.

The impossibility of population-based data, however, does not mean the significance of extremely rare variants is unknowable. Alternative strategies can provide understanding of their significance. For any patient-centered medicine, an ethical imperative exists to classify the medical significance of a VUS in a medically important gene, because an individual carrying a VUS will benefit from decreasing the uncertainty associated with the variation.3

Given the impossibility of conducting standard, population-based analyses when the number of individuals with a specific variant is very small, the methods required to reclassify rare VUS will necessarily differ from those given priority by EBM. Attempts to determine the significance of such a variant will emphasize mechanism rather than epidemiology. The focus on mechanism, with the development of new methodologies to aid in understanding the clinical significance of genetic variation, will ultimately drive the practice of PM.

Criteria proposed for use in classifying genetic sequence variants rely primarily on mechanistic reasoning and methodologies.4 Simple mechanistic reasoning alone may be enough to reclassify a VUS as either benign or pathogenic using a variety of allelic considerations, for instance, a variant that clearly causes a known effect, such as loss of gene function, that is in the same gene domain as all other pathogenic variants.

High-throughput and computational in silico analyses can provide predictions regarding the likely significance of a novel variant. Some computational analyses use inferences from evolutionary conservation because protein domains that are conserved over time are less likely to tolerate variation. Other computational analyses evaluate structural features that are predicted to change RNA splicing or protein folding comparing new variants to data on variants with known outcomes. Some analyses combine multiple approaches.

Functional studies of protein variants can provide some understanding of pathogenicity. Normal protein function strongly suggests a benign variation, whereas substantially altered protein function suggests pathogenicity. Family-specific variants affecting multiple, related individuals can be the subjects of family cosegregation studies, an epidemiologic approach to calculate the likelihood of a specific mutation being causative given family relationships. Even correlation with individual clinical presentation, phenotype, or both, may aid in variant classification, particularly in oncology. However, in almost all situations, correlation of information from multiple sources is considered necessary for classification that will alter medical care. While the importance of mechanistic and Bayesian reasoning are emphasized in guidelines for interpretation of sequence variants, the underlying framework for integrating knowledge derived from various quantitative and qualitative methodologies remains controversial and underexamined.

Therapeutic Implications

Because the possibility of performing randomized clinical trials disappears when variant frequency reaches the family-specific range, any advances in treatment for disorders caused by rare variants must rely on other kinds of evidence. For example, in May 2017, the US Food and Drug Administration (FDA) expanded the indications for ivacaftor to persons with cystic fibrosis caused by 23 new residual function mutations based on the results of functional assays alone.5 The FDA also granted accelerated approval to pembrolizumab for use in any progressive metastatic solid tumor with microsatellite instability or mismatch repair deficiency regardless of tumor tissue origin or location. While some small, uncontrolled human trials are cited in support of approval, the broad tissue- or site-agnostic approval rests solely on mechanistic arguments.6 This shift in regulatory epistemology, acknowledged as an “alternative approach based on precision medicine” heralds a shift away from the primacy of population-derived knowledge in drug development and approval.5 Acceptance of mechanistic information as actionable knowledge seems likely to lead to the expansion of such techniques and arguments going forward.

Practicing Precision Medicine

Precision medicine explicitly prioritizes the individualization of care and focuses attention on unique characteristics of a particular patient. In this fashion, PM differs greatly from EBM, which seeks to determine the best course of action for a patient with an appeal to generalizable knowledge gained from population-based studies. The methods and tools of PM, being developed in large part to help classify rare variants, do not appear in the hierarchies of evidence promulgated within EBM, except occasionally in the bottom tier, categorized broadly as “in vitro research” or “mechanism-based reasoning.” To realize the goals of PM, the hierarchy of evidence pyramid must yield to a more horizontal conception of medical knowledge.

In particular, advancing the field of PM will require acknowledging the value and improving the methods of mechanistic reasoning, including ways to distinguish strong from weak mechanistic reasoning. The value of mechanistic reasoning directly correlates with the soundness of understanding of mechanisms. As genomics and proteomics advance, the scope and accuracy of mechanistic reasoning in diagnosis, risk assessment, and treatment will continue to improve.

As methodologies that rely on mechanistic inference become more reliable, clinical decision making will need to incorporate knowledge derived from a variety of sources to arrive at the best course of action for a particular patient. Clinicians will need to become more comfortable with the “untidy methodological pluralism”7 developing in medicine, in which relevant knowledge comes from a variety of sources and in which the value of that medical knowledge varies from patient to patient. Knowledge derived from populations remains informative for clinical decision-making but is no longer unconditionally preferred over knowledge derived from mechanistic understanding or methodologies. As long as EBM insists on a hierarchy of evidence emphasizing population-based knowledge, EBM and PM are not reconcilable.8 Precision medicine demands case-based reasoning, in which the relevant particulars of the individual patient must be elucidated and incorporated into clinical assessments and decisions.9

The major challenge of PM going forward will be to expand the individualized knowledge that can confidently be brought to bear, moving beyond genomics and proteomics to include aspects of lifestyle and environment as promised in its definition. Failure to do so would limit PM to genomic medicine, an important but insufficient description of optimal clinical medicine.

Back to top
Article Information

Corresponding Author: Brian H. Shirts, MD, PhD, Department of Laboratory Medicine, NW120, University of Washington, 1959 NE Pacific St, PO Box 357110, Seattle, WA 98195-7110 (shirtsb@uw.edu).

Published Online: October 9, 2017. doi:10.1001/jama.2017.11914

Conflict of Interest Disclosures: Both authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: Dr Shirts’s research is supported by grants from the National Institutes of Health (R21HG008513, 5P30 CA015704-39) and the Damon Runyon Cancer Research Foundation (DRR-33-15).

Role of the Funder/Sponsor: The funding sources had no role in the preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank Maya Goldenberg, PhD, and Benjamin Djulbegovic, MD, DSC, for feedback and suggestions.

Collins  FS, Varmus  H.  A new initiative on precision medicine.  N Engl J Med. 2015;372(9):793-795.PubMedGoogle ScholarCrossref
Feero  WG.  Introducing “Genomics and Precision Health”.  JAMA. 2017;317(18):1842-1843.PubMedGoogle ScholarCrossref
Shirts  BH, Pritchard  CC, Walsh  T.  Family-specific variants and the limits of human genetics.  Trends Mol Med. 2016;22(11):925-934.PubMedGoogle ScholarCrossref
Richards  S, Aziz  N, Bale  S,  et al; ACMG Laboratory Quality Assurance Committee.  Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.  Genet Med. 2015;17(5):405-424.PubMedGoogle ScholarCrossref
FDA expands approved use of Kalydeco to treat additional mutations of cystic fibrosis. 2017. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm559212.htm. Accessed September 14, 2017.
FDA approves first cancer treatment for any solid tumor with a specific genetic feature. 2017. https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm560167.htm. Accessed September 14, 2017.
Solomon  M.  Making Medical Knowledge. Oxford, England: Oxford University Press; 2015.Crossref
Nardini  C, Annoni  M, Schiavone  G.  Mechanistic understanding in clinical practice: complementing evidence-based medicine with personalized medicine.  J Eval Clin Pract. 2012;18(5):1000-1005.PubMedGoogle ScholarCrossref
Tonelli  MR.  Integrating evidence into clinical practice: an alternative to evidence-based approaches.  J Eval Clin Pract. 2006;12(3):248-256.PubMedGoogle ScholarCrossref