Dysmorphology-Based Prediction Model for Genetic Disorders in Infants With Congenital Heart Disease.

Abstract

Background: Genetic disorders are prevalent in patients with congenital heart disease (CHD), but genetic evaluations are underutilized and nonstandardized. We sought to quantify a dysmorphology score and develop phenotype-based prediction models for genetic diagnoses in CHD.

Methods: We used a test-negative case-control study of inpatient infants (<1 year) with CHD undergoing standardized genetic evaluations. We quantified a novel dysmorphology score and combined it with other clinical variables used in multivariable logistic regression models to predict genetic diagnoses identified by genetic testing.

Results: Of 1008 patients, 24.1% (243/1008) had genetic diagnoses identified. About half of the cohort were either nondysmorphic or mildly dysmorphic with dysmorphology scores ≤2. There were higher dysmorphology scores according to CHD class (P=0.0007), extracardiac anomaly-positive status (P<0.0001), female sex (P=0.05), and genetic diagnosis identified (P<0.0001). Multivariable logistic regression models quantified this effect further: each +1 increase in the dysmorphology score was associated with a 17% to 20% increased risk of genetic diagnoses (odds ratios, 1.17-1.20, P<0.0001). Extracardiac anomaly-positive status remained a stronger predictor of genetic diagnoses (odds ratios, 2.81-3.39). Nonetheless, about 10% of the cohort were minimally dysmorphic (dysmorphology scores ≤2), had isolated CHD, and were found to have genetic diagnoses, indicating that dysmorphology-based screening can be used to risk-stratify but not exclude genetic diagnoses.

Conclusions: The dysmorphology score is a novel screen for patients with CHD at high risk of having genetic diagnoses identified by genetic testing, including disorders not easily recognized by clinicians. We used these results to develop predicted probability plots for genetic diagnoses in patients with CHD.