To the Editor: We thank Dr Benn for his letter titled “Prenatal Counseling and the Detection of Copy-Number Variants”1 and agree that prenatal testing for copy-number variations (CNVs) differs from karyotyping, in which, more frequently, the patient can be given a clearer phenotypic expectation. Clarifying uncertainty is why we have attempted to estimate penetrance by including a wide range of possible associated disorders.2 We also agree that prenatal microarray testing should be performed in the context of careful thought and counseling. Unclear results can be found regardless of careful use of the test, and our estimates provide one tool to aid counseling in such situations.

To obtain valid estimates for penetrance from CNV frequencies, it is important to know the fraction of the population with an abnormal phenotype that results in being referred for microarray testing. Our assumption that this fraction is approximately equal to the frequency of pediatric conditions with a genetic component—5% in a Canadian epidemiological study3—has some limitations, as pointed out by Dr Benn. Certain conditions may be diagnosed without microarray testing, although the rate of single-gene disorders in the Canadian study was only 0.36%.3 Moreover, a subset of individuals with those conditions may still be referred for microarray testing, including individuals with conditions that can be caused by microdeletions (e.g., neurofibromatosis and cystic kidneys) or individuals who have atypical presentations of their diagnosed condition, for which clinicians wish to rule out other genetic factors altering the phenotype. Certain multifactorial conditions, if isolated, may not be a sufficient cause for microarray testing but are frequently part of syndromic presentations in individuals who are referred for microarray testing. Microarray testing may also be performed in a subset of individuals who have a condition, such as fetal alcohol syndrome, that does not have a genetic component, because it is important to rule out genetic causes before attributing their phenotypes to teratogens. Overall, although this raises the possibility that the 5% frequency may be an overestimate, further analysis of our data does not support Dr Benn’s suggested reduction to 1%. For example, if we compare the population frequencies of known genetic syndromes such as Williams syndrome and Smith–Magenis syndrome due to microdeletions (1/7,5004 and 0.9/15,000,5 respectively) with the frequencies of these deletions in our patient population (110/48,637 and 46/48,637, respectively), it suggests that our testing population comes from a 6% subset of the population with abnormal phenotypes. This may be considered an upper limit, given that we may be underascertaining these syndromes if some cases are diagnosed through other methods such as fluorescence in situ hybridization. Finally, as Dr Benn points out, factors contributing to disease have likely changed since the Canadian study. Some conditions, like autism, are on the rise and may help to counterbalance the subset of the 5% that are not being tested by microarray.

As we state in our original report,2 the controls used are not known to be disease free, and this can cause underestimation of penetrance. If we recalculate penetrance assuming controls are completely unscreened (having a probability of disease (P(D)) of 0.0512), as described in the supplemental methods by Vassos et al.,6 the three CNVs with the highest penetrances have new estimates that are outside of their original confidence intervals: distal 16p11.2 deletions, 100%; proximal 16p11.2 deletions, 84.1%; and distal 1q21.1 deletions, 56.7%. However, these are likely overestimates, given that the controls were adults, and pediatric disease is likely to be underrepresented in that population.

Dr Benn raises concerns about falsely attributing disease causation to CNVs. Our calculations are based on the assumption that the CNV is contributory in all cases in which it is identified. As models for disease causation are shifting toward interaction of multiple genetic changes, including CNVs,7 we believe this to be an acceptable assumption. Furthermore, by examining only CNVs with enrichment in cases, we ensure that we are not falsely attributing causation. Finally, we have excluded prenatal cases from our data to ensure that our testing population is made up exclusively of individuals with known abnormal phenotypes.

We thank Dr Benn for discussing some limitations of our estimates. There is some degree of uncertainty in our estimates, and it is important to keep that in mind when counseling. However, we believe that our 5% estimate for disease frequency is a more reasonable approximation than 1%. Furthermore, it is common to quote a background risk to expectant parents of 3–5% for a child with congenital anomalies, developmental delay, or intellectual disabilities.8 If the counseling session includes framing the problem in terms of the high end of that estimate, then these penetrance estimates could be useful. For example, upon the identification of a 15q11.2 deletion, a couple could be counseled that this may double the chance of the child having congenital anomalies, developmental delay, or intellectual disabilities, changing the risk from the 5% background risk to closer to 10%.

Disclosure

J.A.R. is an employee of Signature Genomic Laboratories, a subsidiary of PerkinElmer. E.E.E. is on the scientific advisory boards for Pacific Biosciences, SynapDx, and DNAnexus. The other authors declare no conflict of interest.