Penetrance of Deleterious Clinical Variants—Reply

In Reply Regarding our recent article, Ms Bastarache and Dr Peterson suggest that identification of cases using EHR data may lead to lower estimates of penetrance. While they raise relevant points, we would like to provide clarification. First, Bastarache and Peterson state that a substantial fraction of individuals with disease lack the relevant ICD diagnosis codes used to estimate penetrance in our study; however, they provide limited evidence for this. The study mentioned is for AATD, a disease that was not included in our study. Extrapolating this to all diagnosis codes and genetic diseases is problematic. Individuals with pathogenic AATD variants may lack ICD codes due to incomplete penetrance, as cautioned by the authors of the AATD study and supported by another study showing normal lung function in most individuals with pathogenic AATD genotypes, likely due to ascertainment bias inflating penetrance estimates. To minimize this problem, our study design used 2 large EHR-linked biobanks with unselected participants. Other studies have also found high rates of inflated pathogenicity and penetrance of disease predisposition variants, consistent with the findings of our study.