Special Report: Evaluating Evidence Supporting a Role for Genetic Markers in Diagnosis, Determining Disease Predisposition, Prognosis, or Predicting Therapeutic Response
Background: While genetic tests are ubiquitous in medicine, evidence that results inform decisions leading to benefit is often lacking.
Objective: To outline a conceptual framework for appraising evidence supporting a role for genetic markers in assessing disease predisposition, diagnosis, and prognosis, or predicting response to therapy.
Method: Important characteristics of evidence are outlined for analytic validity (assay technical performance) and clinical validity (assessing disease predisposition, diagnosis, prognosis, or predict therapeutic response). Aspects of clinical validity are explored including discrimination (e.g., distinguishing individuals who experience some outcome or disease progression from those who do not); calibration (closeness of risk estimates to those observed); relevance of relative effect measures; classification into risk strata by conventional methods followed by reclassification based on additional information provided by a marker; and validation. The interface between evidence and decision frameworks to establish clinical utility is detailed.
Main results: Identifying a genetic marker does not affect morbidity and mortality; however, decisions taken as a result of such testing may. A decision framework allows linking clinical validity to decisions and, subsequently, to benefits and harms or clinical utility. Appraising clinical validity can present challenges. Relative effect measures are of limited use determining the ability of a marker to inform disease predisposition, prognosis, or predicting response to therapy. Risk classification that can inform decision-making can facilitate appraising likely clinical benefit.
A marker must accurately and reliably discriminate—for predisposition, those who will develop disease from those who will not; for diagnosis, those with and without disease; and for prognosis, diseased patients experiencing outcomes of interest from those who will not. Tools or models that include a single or multiple genetic markers should estimate outcome probabilities close to those observed (i.e., well-calibrated). Classification with established tools followed by reclassification using genetic marker(s) with defined risk categories may provide the most informative strategy to convey discrimination. Validation or evidence that the value of a marker is generalizable to other samples and settings is required.
Genetic markers can also be used to predict treatment response. Methods to assess prediction are similar those used for prognosis but conditional on treatment. Evidence supporting clinical utility for predictive markers is most definitively established in randomized, controlled trial(s) comparing therapy directed by the marker to other strategies.
Benefit (clinical utility) will be most convincing when a genetic test is accurate and improves clinical decision-making. Determining that a net clinical benefit results requires a decision framework informed by evidence from multiple sources.
Conclusions: Evaluating evidence supporting a role for genetic markers in assessing disease predisposition, diagnosis, and prognosis, or predicting response to therapy can be complex. Defining benefit requires demonstrating not only test accuracy (reproducible, reliable, and identifies the outcomes or development of disease of interest) but that test results inform and improve decision-making and relevant health outcomes.