Genetic screening for hypertrophic cardiomyopathy in large, asymptomatic military cohorts

Sudden cardiac death (SCD) is one of the leading causes of mortality in the U.S. military and competitive athletes. In this study, we simulate how genetic screening may be implemented in the military to prevent an SCD endpoint resulting from hypertrophic cardiomyopathy (HCM). We created a logistic regression model to predict variant pathogenicity in the most common HCM associated genes MYH7 and MYBPC3. Model predictions were used in conjunction with the gnomAD database to identify frequencies of pathogenic variants. Extrapolating these variants to a military population, lives saved and cost benefit analyses were conducted for screening for HCM related to pathogenic variants in MYH7 and MYBPC3. Genetic screening for HCM followed by echocardiography in individuals with pathogenic variants is predicted to save an average of 2.9 lives per accession cohort, based on historical cohort sizes, and result in a break‐even cost of ~$7 per test. The false positives, defined as disqualified individuals for military service who do not have HCM, are predicted to be 0 individuals per accession cohort. This study suggests that the main barriers for the implementation of genetic screening for the U.S. military are the low detection rate and variant interpretation.


| INTRODUCTION
Sudden cardiac death (SCD) is the most common cause of nontraumatic death in the military (Eckart et al., 2004), and hypertrophic cardiomyopathy (HCM) is the leading cause of SCD in competitive young athletes (Maron, Doerer, Haas, Tierney, & Mueller, 2009). HCM is a condition where the left ventricular wall of the heart is abnormally enlarged which compromises the heart's ability to pump blood, especially during exercise. It is formally defined as left ventricular wall thickness ≥ 15 mm on transthoracic echocardiography when secondary causes such as hypertension, aortic stenosis, or infiltrative cardiomyopathies are absent (American College of Cardiology Foundation/ American Heart Association Task Force on et al., 2011). HCM has variable expressivity and incomplete penetrance, and the estimated prevalence ranges from 1 in 500 to 1 in 200 individuals (Ommen, 2011;Semsarian, Ingles, Maron, & Maron, 2015). Those with HCM have an estimated 0.39% annual incidence of SCD, which increases to 0.84% if the equivalent events such as appropriate implantable cardioverter defibrillator (ICD) shock or successful cardiopulmonary resuscitation are taken into account (Weissler-Snir et al., 2019;O'Mahony et al., 2018). Factors that increase the risk of sudden cardiac death include maximal wall thickness, family history of SCD, left atrial diameter, and non-sustained ventricular tachycardia (O'Mahony et al., 2014;Weissler-Snir et al., 2019). Individuals who are identified as having the disease can be treated with lifestyle changes, which include limiting physical exertion, pharmacotherapy, or invasive procedures such as ICD placement, surgical septal myectomy, or alcohol septal ablation (Weissler-Snir et al., 2019).
An estimated 83% of HCM cases with a positive genetic test are attributed to pathogenic or likely pathogenic variants in the genes MYH7 and MYBPC3, while other genes explain only ≤ 2% per gene (Alfares et al., 2015;Cirino & Ho, 1993;Mademont-Soler et al., 2017;Walsh et al., 2017). Genetic testing in those studies employed Sanger sequencing or next-generation sequencing for a limited number of genes. Considering the decreasing prices and an increased availability of genomic sequencing, population-level testing for HCM and inclusion of new genes as a panel are becoming plausible. Currently, there is little data to evaluate effectiveness of genetic screening for HCM in military or athletic populations. Young athletes are similar to military recruits in age and physical demands, and several studies have looked at the effectiveness of screening young athletes with electrocardiogram (ECG) (Corrado et al., 2005) (De Castro et al., 2016). In this report, we explore the feasibility and cost-effectiveness of screening military recruits for HCM.

| METHODS
To analyze the benefits of screening military personnel for HCM, we used the following sequential stages: genome simulation, disease simulation, HCM screening, sudden death simulation, and cost analysis for the military. Figure S1 diagrams the workflow for these sequential stages. The population size was based on historical accession data, and we consider an "accession cohort" to be the officers and enlisted members who joined the military in a given year. Below is a brief description of simulation, more detail is found in the Supplemental Methods.

| Stage 1: Genome simulation
First, we simulated the presence/absence of pathogenic variants. To determine which variants would be considered pathogenic, we identified 90 variants in ClinVar database for MYH7 and MYBPC3 that were rated pathogenic or likely pathogenic with no conflicting interpretations and 346 that were rated benign or likely benign with no conflicting interpretations (www.ncbi.nlm.nih.gov/clinvar/). Variant pathogenicity for other variants was predicted using a logistic regression model that had been trained on the high-certainty ClinVar variants. The selected logistic regression model used allele frequency, combined annotation dependent depletion (CADD), dbscSNV, and indicated amino acid change as predictors (see Supplementary Methods for modeling details and predictor descriptions). We then constructed a ranked list of pathogenic variants for simulation, composed of the 90 ClinVar variants with high certainties of pathogenicity and no conflicting interpretations, followed by additional variants ranked by their model-based likelihoods of pathogenicity.
To simulate variant frequencies as they occur in the general population, we used the gnomAD database to estimate how frequently pathogenic variants should occur in our simulated population. The gnomAD database was used as a surrogate of the military population due to the asymptomatic nature of both populations. Of the six high-certainty pathogenic variants available in the 1,000 Genomes database (Siva, 2008), all had no co-segregation, and thus we worked under the assumption of independent and separate inheritance for each pathogenic allele. Equation (2)  In the case of genetic screening followed by transthoracic echocardiogram for positive genetic test, "screening positive for HCM" is defined as a positive echocardiogram. The simulations were run for 1,000 cohorts.  (Sharra, 2015), and a RAND corporation study (Dahlman, 2007

| RESULTS
Results are summarized in Table 2. In this report, we display averages

| Screening sensitivity and FDR
Case 2 (echocardiogram only) has ideal sensitivity and FDR, 1 and 0 respectively. Case 4 (genetic screening; positive genetic test followed by transthoracic echocardiogram) is the next most favorable mix, with a sensitivity of 30% and FDR of 0 (Table 2).

| Lives saved
As expected, the most sensitive screening strategy is also predicted to save the most lives. Case 2 (echocardiogram only) is expected to save an average of 9.5 lives per cohort (95% CI,9.3-9.7) (assuming historical cohort size and officer/enlisted composition); however, the monetary cost of screening every individual is too high to make this a practical approach for screening large populations. Case 4 (genetic screening; positive genetic test followed by echocardiogram), as noted above with the next most favorable sensitivity and FDR values, saves an average of 2.9 (95% CI, 2.8-3.1) lives per recruit cohort (Table 2).
T A B L E 1 Screening cases being compared in cost/benefit analysis False positive (discharged without HCM) Case 1: No screening n/a n/a 0 n/a n/a Case 2: Echocardiogram only 1 (1-1) 0 (0-0) 9.52 (9.33-9.71) n/a 0 (0-0) 3.3 | False positives (discharged from U.S. military in the absence of HCM) The only case resulting in false positives is Case 3 (genetic screening only). It is predicted to erroneously discharge 240 service members per cohort (95% CI, 239-241), which is unacceptably high.

| Break-even costs
Considering costs of training, death gratuities, and screening procedures, as designated in Supplementary Table 1, we compared the overall cost of no screening versus each screening strategy. The "break-even genetic test cost" shown in Table 2 is the maximum cost of a genetic test where the military would start to see a cost benefit to screening.

| DISCUSSION
The leading cause of nontraumatic death in the military is SCD (Eckart et al., 2004;Eckart and others, 2011). HCM is the most common cause of SCD in young athletes (Maron et al., 2009), and although screening athletes for cardiac risks is controversial, a mandatory screening program using ECG reduced the incidence of SCD in young athletes (Corrado and others, 2005 (Mehlman & Li, 2014). Given the incomplete penetrance of most genetic conditions, special care must be taken before labeling an asymptomatic individual with a disease or syndrome. In this study, we confirmed the positive genetic screening findings with an echocardiogram.
Most cases of HCM, the most common cause of SCD, can be attributed to pathogenic variants in MYH7 or MYBPC3, but the detection rate using the two variants is only around 30%. The detection rate is expected to improve over the time with the addition of new genes and reclassification of variants of unknown significance. A limitation in our simulation study is the assumption that the pathogenic variant list constructed for MYH7 or MYBPC3 is accurate, that is, that the variants are truly pathogenic. We acknowledge that variant classification is a complex process that requires a trained geneticist and should follow the ACMG guidelines (Richards et al., 2015). However, by prioritizing Clin-Var's high certainty pathogenic variants and using predictions from a disease-specific classification model, we have constructed a variant list that best reflects current scientific knowledge. The averages reported in Table 2 would be consistent with any list of pathogenic variants whose total frequency is P(VAR), but, clearly, practical implementation would require validation of which particular variants comprise this list.