Incidental results can come up in many areas of medicine. In genomic testing, they are practically guaranteed.
Much has been written about the promise of genome-scale sequencing in medicine, both for diagnosis of rare monogenic disorders and also perhaps as part of routine patient care. Yet no matter the indication, the vast majority of genetic variants in a given individual will be unrelated to the reason the test was performed, and thus could be considered "incidental" or "secondary" findings. Of these, nearly all will have minimal or no clinical implications, but each individual will have a heterogeneous assortment of clinically significant but unpredictable genetic variants:
- A small fraction of individuals will be found to have a mutation that indicates the existence of a previously unsuspected genetic disorder with clear clinical consequences (ranging from very rare disorders such as Lynch syndrome to more common conditions such as hemochromatosis).
- Most individuals will be found to carry a handful of heterozygous mutations for conditions with recessive inheritance patterns.
- Everyone will have a complement of common genetic variants that have been implicated in pharmacogenomic variation or associated with modestly shifting their risk for common multifactorial disorders.
The task of a clinically-oriented genomic analysis will therefore be to effectively communicate to patients the range of potential incidental findings and to outline a structured process for identifying and reporting the relevant variants, balancing principles of medical ethics such as the physician's "duty to warn" with the patient's autonomy to determine what genomic information they desire to know about themselves.
How, then, do we communicate the wide range of potential incidental findings in a genome sequence? Making the task more difficult is that if we are to engage in any kind of meaningful informed consent with patients, the possible range of results waiting to be discovered in an individual's genome must be communicated to them prior to sequencing so that they can make informed decisions. One cannot wait until one happens to find, say, a presenilin mutation that confers a virtual certainty of Alzheimer's disease and then ask the patient if they want to know about the potentially disturbing mutation that was found.
Different metaphors have been used to describe genomic incidental findings with respect to incidental findings in other areas of medicine. One familiar comparison is with the radiographically-detected "incidentaloma," for which the differential diagnosis might include both benign and malignant entities. In any given study, a radiologist must routinely assess numerous minute details that vary slightly from "normal," and decide whether and how to comment on them. Physicians do not routinely make their patients aware of the possibility of such findings when ordering imaging studies, yet they must cope with the consequences of such findings, including the potential need for follow-up imaging studies or more invasive biopsies. The challenge of applying this model to genomics is that while not infrequent, incidental findings in radiology are not the rule. On the contrary, they will be universally discovered when one's genome is sequenced and thus clear guidelines must be established for how to deal with them in a way that balances efficiency, provider duty and patient autonomy.
Another comparison can be made to laboratory biochemical findings, in which different flags are used to draw attention to values that are outside of the normal range. Laboratory "critical" values are those that must be attended to immediately due to the potential for imminent danger to the patient. In the setting of an acutely ill patient, such critical values can be of great diagnostic importance. In an otherwise healthy patient undergoing routine screening, the detection of a critical laboratory abnormality is more likely to represent a spurious analytic error than a true finding, and is made more likely when multiple assays are run simultaneously.
Both the radiology and the clinical laboratory metaphors articulate the essential idea that certain incidental findings are of sufficient importance to warrant reporting to physicians and patients as a matter of course. However, neither accommodates either the frequency of their discovery in genomics (a certainty with each genome sequenced) nor the broad range of potential findings seen in genomic medicine—from highly deterministic to modestly predictive, from generally benign to potentially devastating, and from completely preventable to utterly futile with regard to possible medical interventions. Instead, these metaphors relate primarily to clinical situations in which the abnormal finding indicates a need for specific actions that would reasonably be expected to benefit the patient. In contrast, only a small minority of genomic findings rise to a level of clinical "actionability." And critically, while some findings may be scientifically and clinically valid (for example, the discovery of the presenilin mutation that confers near certainty of early-onset Alzheimer disease), such results will be ardently desired by some patients and just as adamantly not desired by others.
So while we can learn from the dilemma of the radiologist and the traditional laboratorian, such metaphors ultimately fail to fully capture the magnitude of the problem when it comes to genome sequencing, given that each individual will have millions of positions that differ from the "reference" genome sequence, the overwhelming majority of which are of no clinical importance. The hazards of interpreting incidental genomic findings have been capably outlined elsewhere. Further complicating the genomic situation is that our knowledge base is evolving very rapidly. The sheer number of genetic variants and the extremely low prior probability that such findings truly indicate the presence of a genetic disorder result in an imperative to avoid overwhelming the system with incidental findings of dubious clinical validity or utility, and instead to focus only on those variants that are highly likely to represent disease-causing mutations.
Clinically-oriented analysis of incidental findings in genome sequence data must therefore address these points of concern with a methodical, reproducible approach that allows both the clinician to efficiently discharge their responsibility while affording maximal autonomy to the patient. It seems to us that a case-by-case assessment of incidental findings could lead to unsystematic results that are not consistent between analyses. We therefore prefer the establishment of an a priori framework for the analysis of genomic incidental findings. In our proposed model, genes are organized into categories ("bins") based on clinical actionability (Bin 1), clinical validity without direct actionability (Bin 2), and no clinical implications (Bin 3). This model can then be used to facilitate pre-test informed consent so that it can be made clear that some types of incidental findings (Bin 1, expected to be very rare) would be obligatorily returned as a matter of course because the information would directly affect that individual's or their family members' clinical management, while other findings (Bin 2) reside in the realm of individual preference and are not routinely divulged without appropriately-informed decisions by the individual. Such an approach is enabled only by the appropriate use of computational algorithms that select variants for further review and reporting based on a high threshold of likely pathogenicity, thus streamlining the analysis and ensuring that each sample is subjected to precisely the same analytic criteria. Finally, the application of a structured informatics approach to analyzing the incidentalome provides "versioning" and allows one to revisit the analysis periodically since the underlying choices about which genes and variants are clinically "actionable" will certainly change as genomic medicine advances.
The genome is a big place, full of information that varies from worthless to life-saving, from innocuous to terrifying. Approaching it in a planned and methodical manner will help patients, the public and providers alike navigate its immensity.
Jonathan S. Berg, MD, PhD, is Assistant Professor in the Department of Genetics at the University of North Carolina School of Medicine and Principal Investigator for the Carolina-Georgia Center of the Cancer Genetics Network.
James P. Evans, MD, PhD, is Bryson Distinguished Professor of Genetics and Medicine at the University of North Carolina School of Medicine.
1. Green ED, Guyer MS; National Human Genome Research Institute. "Charting a course for genomic medicine from base pairs to bedside." Nature. 2011 Feb 10;470(7333):204-13.
2. Kohane IS, Hsing M, Kong SW. "Taxonomizing, sizing, and overcoming the incidentalome." Genet Med. 2012 Apr;14(4):399-404.
3. Berg JS, Khoury MJ, Evans JP. "Deploying whole genome sequencing in clinical practice and public health: meeting the challenge one bin at a time." Genet Med. 2011 Jun;13(6):499-504.
4. Berg JS, Adams M, Nassar N, Bizon C, Lee K, Schmitt CP, Wilhelmsen KC, and Evans JP. "An informatics approach to analyzing the incidentalome." Genet Med. In press.