Genomic screening and genomic diagnostic testing—two very different kettles of fish
Genome and exome sequencing
Genomic testing can be misunderstood as being determinative, when in reality it is the same as all other tests and context is essential for its correct interpretation. Two hypothetical cases of testing for Marfan syndrome demonstrate how clinicians should contextualize genomic test results and the implementation of Bayes theorem in clinical decision-making.
Genome and exome sequencing (GS/ES) are rapidly becoming more widely used and providing unprecedented ability to diagnose individuals with rare or unexpected genetic disorders quickly and accurately. The power of these sequencing techniques is in their breadth and hypothesis-generating nature: they test for nearly all Mendelian disorders . GS/ES is a powerful diagnostic tool, but like any other clinical test, it has true positives, true negatives, false positives, and false negatives. It is essential to understand these attributes both in the diagnostic setting and in the screening setting. The key to understanding variant pathogenicity and how to contextualize the clinical implications is based on Bayes theorem. Here, using two hypothetical GS/ES testing scenarios the practical utility of Bayes in genomic testing will be illustrated.
A young man presents to his internist for a routine checkup and the clinician notes that he has facial and skeletal features of Marfan syndrome that do not reach the threshold for major diagnostic specificity. He also has a history of high myopia but no known lens dislocation. There is no family history of Marfan syndrome, but several of his maternal relatives are tall with a vague history of unexplained sudden death in one. The internist sends the young man for an echocardiogram, which shows an aortic root diameter to body surface area ratio that is just over the 95th centile. On the basis of this evidence, she estimates that there is about a 75% chance that the patient may have Marfan syndrome: there are some signs of the disorder, but not enough for clinical diagnosis. Genome sequencing is ordered and returns a pathogenic variant in FBN1 (pathogenic is defined as ≥ 99% likely to be associated with the disease).
Importantly, the likelihood that the patient has the disorder (here 99.95%) is not numerically equivalent to the probability of pathogenicity of the variant (which is ≥ 99%).
(Note here that 0.00013 is the overall prevalence of Marfan syndrome, about 1/7500.) On the basis of what is known at this point, the odds are that this toddler does not have Marfan syndrome. The dramatic change here is due to the prior probability, which was 75% in the first scenario but about 1/7500 in the second scenario. Like all tests, GS/ES is challenged by the false-positive rate, which in these scenarios is the likelihood that a pathogenic variant might not actually be causative of disease. This is implicit in the description that it is ≥ 99% likely to be causative, not 100%. The critical lesson from scenario 2 is that the prior probability of disease (1/7500 vs 75%, screening vs diagnostic) is a critical determinant of the likelihood of the diagnosis.
While it is most likely that this toddler does not have Marfan syndrome, one should not dismiss the diagnosis. There are low risks of serious medical complications of Marfan syndrome in young children, so it is reasonable for the pediatrician to recheck some of the physical findings for Marfan syndrome and, if these features are absent, adopt a watch and wait approach. He could continue with regular pediatric well checks and when the girl is older, and clinically reassess and update the interpretation of the variant. Genetics knowledge is improving rapidly and a great deal will be learned in the coming years. If the variant is still considered to be pathogenic, a more thorough clinical evaluation for Marfan should be undertaken. This could include a referral to a clinician who is experienced and confident of their skills with Marfan syndrome, an ophthalmologic evaluation to assess ectopia lentis specifically, and an echocardiogram. This suite of findings can be evaluated by a clinician expert in Marfan syndrome to determine whether further workup is required, whether a diagnosis can be made and management instituted, or whether the family can be reassured that there is no sign of the disorder and a further watch and wait approach is appropriate.
These examples estimate the likelihood that the individual actually has the diagnosis, based on what was known clinically before the test and after a GS/ES test result. There are many more factors to consider in genomic diagnosis; for example, penetrance (the likelihood the patient has manifestations of the disease if they have the disease) has to be taken into account. Marfan syndrome has very high penetrance, although a number of the manifestations are age-dependent . Thus, the absence of obvious signs of the disorder in the toddler (scenario 2) should not permit the pediatrician to dismiss the possibility that signs could develop over the coming years (age-dependent penetrance). It should also be noted that calculations such as these are more complex when a disorder has low penetrance. Although there are nuances and complexities, the conclusion is clear: GS/ES results must be contextualized in a Bayesian framework to be valid clinically.
In the end, genomic testing is more similar to, than it is different from, a hematocrit or serum sodium test result. All three tests are extremely useful if interpreted correctly, given the clinical context in which they are used. The critical concepts to recognize are that the pathogenicity of the variant is not the likelihood that the patient has the disease, any more than the accuracy of a hemoglobin result is the likelihood that the patient has anemia. The clinical context in which the testing was done is a major determinant of the diagnosis of the patient. Much of the confusion surrounding genomic testing is based on misconceptions of genetic determinism: that one can determine ones’ status with certainty on the basis of a genomic or genetic test result. Genetic testing can be powerful and useful in both of the scenarios described above, but Bayes theorem must be taken into consideration.
Conclusions and future directions
Bayes theorem applies to everything that clinicians do, whether assessing the clinical significance of a fever or that of a GS/ES result. Bayes theorem is how clinical context can be incorporated into genomic testing to allow rational clinical decision-making. By contextualizing genomic test results, clinicians can better manage their patients in both diagnostic and screening contexts.
The author thanks Harry Dietz, Robert Nussbaum, and Heidi Rehm for reviewing a prior draft of this manuscript and providing comments and suggestions.
LGB wrote and edited the manuscript. The author read and approved the final manuscript.
This work was supported by National Institutes of Health intramural research grant HG200359.
The author is an uncompensated member of the Illumina Corp. Medical Ethics Board, and receives honoraria from Cold Spring Harbor Press.
- 4.Kalia SS, Adelman K, Bale SJ, Chung WK, Eng C, Evans JP, et al. Recommendations for reporting of secondary findings in clinical exome and genome sequencing, 2016 update (ACMG SF v2.0): a policy statement of the American College of Medical Genetics and Genomics. Genet Med. 2017;19:249–55.CrossRefGoogle Scholar
- 5.Dietz H. Marfan syndrome. In: Adam MP, Ardinger HH, Pagon RA, et al., editors. GeneReviews. Seattle: University of Washington; 1993–2019. https://www.ncbi.nlm.nih.gov/books/NBK1335/. Accessed 13 Nov 2019.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.