Despite what you may have heard, incidental findings in whole genome sequencing may not be so different from incidental findings in other areas of medicine.
In response to the precipitous drop in the costs of whole genome and exome sequencing, groups are now implementing this technology in a variety of research and clinical settings. Whole genome and exome sequencing (WGS) has remarkable power to detect causal genetic disease variants in populations of patients who had eluded diagnosis despite extensive diagnostic odysseys, particularly when one considers the challenges of interpreting the vast amounts of data coming off of the next generation sequencing machines.
A consequence of doing WGS is that one ascertains all of the variants from the portion of the genome that is being analyzed. Some of these variants will occur in genes. The vast majority of these variants will have no effect, but a small subset will be deleterious. In some cases the mutation will affect one allele in a gene associated with an autosomal recessive condition (CFTR and cystic fibrosis) in which case the individual would be a carrier, but would not have any manifestations of the disease. More infrequently a deleterious mutation in a gene could predispose the individual to disease. Examples could include autosomal dominant cancer predisposition genes (e.g. BRCA1/2); autosomal recessive genes where an individual is found to have a mutation in each allele (HFE and hemochromatosis); and X-linked genes where males or female hemizygous carriers could be identified (PRPS1 and Charcot-Marie-Tooth X-linked recessive 5 or GLA and Fabry cardiomyopathy). In some cases these variants will be identified as being causally related to the indication for performing WGS, but there is also the possibility that these could be found in situations where WGS was done for other reasons. In the latter case the findings would be considered incidental or secondary. The recognition that clinically significant variants would be discovered in the course of WGS has generated much discussion from a variety of groups involved in the performance and interpretation of WGS. For this article I will focus on one aspect of the debate: The contention that incidental information in the context of WGS represents a novel problem in medicine and, as such, consent to testing should allow the patient and/or provider the choice to opt out of receiving clinically important variant information that is unrelated to the indication for testing in order to protect patients from unneeded or unwanted stress.
Is this new? Is this different?
Based on 30 years of medical practice I can say unequivocally no-incidental findings are an inherent part of medical practice. Most agree that whether uncovered as part of a history and physical examination or as a consequence of a diagnostic test this is a part of medicine. Perhaps the more relevant question is: Are incidental findings from WGS different from those found in traditional medical practice? Here there is much more disagreement. A non-exhaustive list of potential differences include: impact on other family members; identification of risk for adult-onset conditions in children; lack of information about the risk of specific variants (or findings for non-genetic situations) in particular information about the penetrance and variability of expression of the identified variant; and cost to the system of confirming incidental findings. While these all merit consideration, I've experienced each of these issues in my general pediatric practice.
In a well-child examination, I identified a heart murmur that I thought was not an innocent childhood murmur. Echocardiography identified a hypertrophic cardiomyopathy that ultimately required heart transplantation for the infant. Most of the cardiomyopathies are genetic with the majority being autosomal dominant (impact on other family members). Recommendation was made for echoes to be done on the parents and siblings of the patient. While these turned out to be normal, the variability of expression and penetrance for these conditions in families presented uncertainty in counseling. Should serial echoes be done over time and, if so, in both parents and children? At what age are we certain that a cardiomyopathy would be present on an echo? Clearly these issues affect not only the clinical issues but also impact the cost of care to the system related to the incidental finding. This case occurred long before genetic testing, but application of genetic testing in the form of a cardiomyopathy gene panel could have helped with some of the questions. If a causal variant was identified in the child, is it present in one of the parents? If not, this is likely a new mutation and other family members would not be at increased risk. If so, the questions of variability of expression and penetrance would still need to be addressed, but at least the number of family members that would be offered screening is restricted to those who carry the variant lowering the impact on the cost of care.
I performed a well-child examination on a 3 year old asymptomatic child of missionaries who were temporarily back in the US. They informed me that their mission was in an area where tuberculosis is endemic, so their child had received a BCG vaccine. Because the vaccine rendered the recommended screening test (intradermal PPD) uninterpretable, I performed a screening chest X-ray to check for occult disease, as the child would be attending day care while in the US. The X-ray revealed a 10 cm paraspinous mass which, at surgery, turned out to be a cystic neuroblastoma. In reviewing the X-ray finding, none of the consulting physicians could definitively define the risk of the specific finding and thus the need for surgery. Cystic neuroblastoma is a very rare tumor type, so even with a pathologic diagnosis, it was unclear if additional treatment with chemotherapy and radiation was necessary as some of these tumors do not progress, and occasionally resolve spontaneously. Ultimately no additional treatment was given and the child did well on interval visits. Several questions occur: If the X-ray had not been done, would the mass have ever led to problems? Would it have spontaneously regressed? Should additional treatment have been given beyond surgery? What are the risks and benefits of such treatment?
At the time of a sport's physical on a 13 year old girl, a family history was obtained that was clearly consistent with a breast/ovarian cancer syndrome in the mother's family. No genetic testing had been done in the family, but segregation analysis showed that the patient's mother was at a 50% risk, thus the patient was at a 25% chance of carrying a highly penetrant mutation in a cancer predisposition gene (most likely BRCA1/2). In the context of obtaining a medical history the incidental finding of a family history of breast/ovarian cancer identified high risk for an adult onset condition in an adolescent. The mother was referred for genetic counseling. If testing confirms a BRCA mutation in the patient's mother, should the child be tested? If so, when? If testing is deferred to adulthood to preserve autonomy, who is responsible for communicating the information to the child when she reaches adulthood? Has her ability to choose already been compromised? When should surveillance begin? When should chemoprevention and/or prophylactic surgery be offered? These questions reflect uncertainty not only in the risk of the specific finding, variable expression and penetrance but also introduce a host of questions about choice and autonomy.
The questions raised by these examples illustrate that while the application of WGS in the clinic may impact the number of incidental findings quantitatively compared to any other single test, the questions raised by the incidental findings are not qualitatively different.
An instructive analogy?
One "advantage" of having been in practice for several decades is that it is easier to understand historical perspectives one has personally experienced. In thinking about clinical WGS and incidental findings, the analogy of chemistry panels seems useful, if somewhat imperfect. The primary motivation for the introduction of chemistry panels into practice was economies of scale. If one wanted to order a test covering sodium, potassium, calcium and phosphorus, it was cheaper to run these tests on a panel of 20 (or 36 or more) than it was to run two or more of the tests individually (a similar argument to WGS over gene by gene testing). When chemistry panels moved into clinical practice, with very few exceptions the entire panel was ordered and reported, as this was more convenient. The challenge is that if you are doing 20 tests where the normal range is defined by 95% confidence intervals derived from populations, there is a near certainty that one of the 20 tests will be out of range (i.e. 'abnormal'). Clinician responses to these out of range results varied. An aspartate aminotransferase (AST) a couple of points out of range in a context where liver disease was not suspected (i.e. a low prior probability of disease for the specific test) could reasonably be ignored and was by many if not most clinicians. However, more compulsive (or less secure) clinicians would pursue more specific liver function tests, or even imaging studies despite the low prior probability of disease leading to higher likelihood of subsequent false positive out of range tests and added cost to the system without additional benefit and, in some cases, creation of harm. An extreme example of this is whole body CT scanning as a screening test, where 37% of scans detect an 'abnormal' finding that vast majority of which are inconsequential given that the testing is being done outside of a clinical indication, therefore the prior probability of disease is low. Indiscriminate use of incidental findings from WGS has the same potential to lead to cascades of evaluation, particularly if the finding is not contextualized through the use of other information like family history. In chemistry testing, the trend has been to require clinicians to order the specific tests they are interested in based on the clinical context. For economy's sake, the tests are still run as part of a larger panel; however, only the results of the requested tests are reported to the clinician. This eliminates the requirement to follow-up on other tests that are statistically out of range, but are unlikely to be of clinical significance. The exception is a value that is so out of range that it must be assumed to be clinically significant. Findings such as these are categorized as "panic values" and the clinician will be contacted even though they did not order the test. The challenge as I see it over the near future for WGS is to develop a list of genetic 'panic values' based on an accepted level of certainty of pathogenicity for the variant and known clinical interventions so that if a given variant is found that is not relevant to the indication for the test it will still be reported. An example could be a known deleterious mutation in the MLH1 gene that causes Lynch syndrome discovered through WGS in an individual being evaluated for adult onset deafness. Disclosure of this information could have important implications for the patient, in that earlier and more frequent colonoscopy has the ability to identify and remove pre-cancerous adenomatous polyps dramatically reducing the risk of developing colon cancer. At present several organizations, including the American College of Medical Genetics and Genomics, the National Human Genome Research Institute and the Evaluation of Genomic Applications in Prevention and Practice as well as many private groups are examining the content and ramifications of such a list. Implicit in this work is the requirement for a regularly updated centralized repository of well-annotated deleterious variants that can be accessed by laboratories and clinicians to aid in the interpretation of variants identified through WGS.
Finally, there is one additional issue relevant to this topic that should be addressed. Some have called for consent for WGS to allow patients and/or providers to 'opt out' from receiving incidental findings. If this were to be implemented it would be true genetic exceptionalism; in no other area of medical care do we allow patients to 'opt out' of receiving clinically significant incidental findings. Take the examples presented above and imagine the conversation with the parents: "I'm going to do a physical examination and I may find a heart murmur, but I want you to tell me now if you want me to disclose if I find something since it may not be information you want to deal with." While meant to elicit a smile, the reality is that this is not the standard of medical practice nor does it reduce the clinician's liability. What it does require is agreeing on an acceptable level of certainty about the clinical impact of the variant that can inform the decision to disclose as has been done with incidental findings in all other areas of medicine. As such the aforementioned proactive attempts to address incidental findings from WGS are most welcome.
Marc Williams, MD, is Director of Geisinger Health System's Genomic Medicine Institute.