Abstract 16172: Predicting the Presence of a Mutation Resulting in Familial Hypercholesterolemia - Development of a Prediction Model in a Cohort of 64,000 Subjects
Introduction: Familial hypercholesterolemia (FH) is a hereditary disease that warrants early diagnosis to prevent premature cardiovascular disease (CVD) by initiating therapy. A definitive diagnosis is made by demonstrating a causal mutation. This might not only increase therapy adherence, but is also an important requirement for cascade screening. However, DNA analysis is costly and careful selection of subjects is important. Unfortunately, the accuracy of current selection criteria is poor.
Hypothesis: We set out to develop a model to predict the presence of an FH causing mutation in persons referred by general practitioners.
Methods: All participants in the Dutch FH screening program from January 1994 till January 2014 were included. Cross-sectional data was available on medical history, lipid profile and DNA analysis. The primary outcome was the presence of a deleterious FH mutation. We developed a prediction model using multivariable logistic regression analysis.
Results: Our study population consisted of 25,809 FH patients and 38,297 unaffected relatives. Our final model included age, gender, levels of LDL-cholesterol, HDL-cholesterol and triglycerides, history of CVD, use of statins and other lipid lowering therapy, smoking and alcohol. The performance was good: the area under the receiver operating characteristic curve (AUC) was 85.2% (95% CI: 84.9 - 85.5). The model was well calibrated with a slope of 1.02 (1 is optimal). Internal validation was excellent: the AUC was unaltered in 100 bootstrap samples. The performance of the model can be illustrated by selecting subjects with a predicted probability of 30% or lower. This would identify 87.0% of all unaffected subjects, and avoid testing in 46.1% of the population.
Conclusions: We developed a model to predict the presence of a deleterious FH mutation in subjects referred by general practitioners. Our model showed good discrimination and calibration, with no signs of overfitting during internal validation. After external validation, we will develop an interactive web-based calculator or smart phone application that can be used to calculate the probability of the presence of an FH mutation in individual subjects and will facilitate the use of our prediction model in daily clinical practice.
Author Disclosures: J. Besseling: None. J.B. Reitsma: None. G.K. Hovingh: None. B.A. Hutten: None.
- © 2014 by American Heart Association, Inc.