Derivation and Validation of a Screening Model for Hypertrophic Cardiomyopathy Based on Electrocardiogram Features.

Background: Hypertrophic cardiomyopathy (HCM) is a widely distributed, but clinically heterogeneous genetic heart disease, affects approximately 20 million people worldwide. Nowadays, HCM is treatable with the advancement of medical interventions. However, due to occult clinical presentations and a lack of easy, inexpensive, and widely popularized screening approaches in the general population, 80-90% HCM patients are not clinically identifiable, which brings certain safety hazards could have been prevented. The majority HCM patients showed abnormal and diverse electrocardiogram (ECG) presentations, it is unclear which ECG parameters are the most efficient for HCM screening. Objective: We aimed to develop a pragmatic prediction model based on the most common ECG features to screen for HCM. Methods: Between April 1st and September 30th, 2020, 423 consecutive subjects from the International Cooperation Center for Hypertrophic Cardiomyopathy of Xijing Hospital [172 HCM patients, 251 participants without left ventricular hypertrophy (non-HCM)] were prospectively included in the training cohort. Between January 4th and February 30th, 2021, 163 participants from the same center were included in the temporal internal validation cohort (62 HCM patients, 101 non-HCM participants). External validation was performed using retrospectively collected ECG data from Xijing Hospital (3,232 HCM ECG samples from January 1st, 2000, to March 31st, 2020; 95,184 non-HCM ECG samples from January 1st to December 31st, 2020). The C-statistic was used to measure the discriminative ability of the model. Results: Among 30 ECG features examined, all except abnormal Q wave significantly differed between the HCM patients and non-HCM comparators. After several independent feature selection approaches and model evaluation, we included only two ECG features, T wave inversion (TWI) and the amplitude of S wave in lead V1 (SV1), in the HCM prediction model. The model showed a clearly useful discriminative performance (C-statistic > 0.75) in the training [C-statistic 0.857 (0.818-0.896)], and temporal validation cohorts [C-statistic 0.871 (0.812-0.930)]. In the external validation cohort, the C-statistic of the model was 0.833 [0.825-0.841]. A browser-based calculator was generated accordingly. Conclusion: The pragmatic model established using only TWI and SV1 may be helpful for predicting the probability of HCM and shows promise for use in population-based HCM screening.