Hypertrophic cardiomyopathy clinical phenotype is independent of gene mutation and mutation dosage

Over 1,500 gene mutations are known to cause hypertrophic cardiomyopathy (HCM). Previous studies suggest that cardiac β-myosin heavy chain (MYH7) gene mutations are commonly associated with a more severe phenotype, compared to cardiac myosin binding protein-C (MYBPC3) gene mutations with milder phenotype, incomplete penetrance and later age of onset. Compound mutations can worsen the phenotype. This study aimed to validate these comparative differences in a large cohort of individuals and families with HCM. We performed genome-phenome correlation among 80 symptomatic HCM patients, 35 asymptomatic carriers and 35 non-carriers, using an 18-gene clinical diagnostic HCM panel. A total of 125 mutations were identified in 14 genes. MYBPC3 and MYH7 mutations contributed to 50.0% and 24.4% of the HCM patients, respectively, suggesting that MYBPC3 mutations were the most frequent cause of HCM in our cohort. Double mutations were found in only nine HCM patients (7.8%) who were phenotypically indistinguishable from single-mutation carriers. Comparisons of clinical parameters of MYBPC3 and MYH7 mutants were not statistically significant, but asymptomatic carriers had high left ventricular ejection fraction and diastolic dysfunction when compared to non-carriers. The presence of double mutations increases the risk for symptomatic HCM with no change in severity, as determined in this study subset. The pathologic effects of MYBPC3 and MYH7 were found to be independent of gene mutation location. Furthermore, HCM pathology is independent of protein domain disruption in both MYBPC3 and MYH7. These data provide evidence that MYBPC3 mutations constitute the preeminent cause of HCM and that they are phenotypically indistinguishable from HCM caused by MYH7 mutations.


Introduction
Hypertrophic cardiomyopathy (HCM) is a commonly inherited myocardial disease with an estimated prevalence of 1 in 200 [1]. HCM is characterized by abnormal thickening of the left ventricular wall that cannot be explained by abnormal loading conditions or metabolic disorder [2]. In patients with HCM, cardiomyocytes enlarge, and interstitial fibrosis and myocardial disarray are apparent [3]. Clinical outcomes of HCM are highly variable such that many patients are asymptomatic, or mildly symptomatic, presenting with symptoms such as dyspnea, angina, syncope, lightheadedness, and heart palpitations [2]. In severe cases, HCM is potentially life-threatening and associated with high risk of heart failure, atrial fibrillation, stroke, and sudden death [4]. Importantly, HCM is the leading cause of sudden cardiac death in young athletes [5]. However, even among HCM, it is difficult to predict disease severity owing to the variable penetrance of the disease and insufficient genotype-phenotype correlation [6].
Genetically linked HCM has been defined mostly as a disease of the sarcomere [7]. Mutations in genes encoding sarcomeric proteins account for over 95% of all HCM cases. Over 1,500 autosomal dominant mutations in at least 14 genes encoding cardiac sarcomere proteins are associated with HCM [7][8][9][10]. Mutations in cardiac β-myosin heavy chain (MYH7) and cardiac myosin binding protein-C (MYBPC3) occur in most cases [11,12], while cardiac troponins -I (TNNI3),-T (TNNT2), and-C (TNNC1), α-tropomyosin (TPM1) and the myosin light chains (MYL2 and MYL3) are less frequently associated with HCM. These mutations display a wide range of phenotypes, from normal appearing heart, or mild hypertrophy, to severe hypertrophy and increased risk for life-threatening ventricular arrhythmias [1,13,14]. In patients with HCM, new mutations in sarcomeric proteins are increasingly recognized, but genotypephenotype correlation and the underlying molecular mechanisms of HCM pathogenesis have not been fully defined. Two lines of reasoning have been advanced to explain the pathogenic effects of MYH7 and MYBPC3 mutations. In some studies, MYBPC3 mutants have been associated with mild to moderate phenotype and a less severe form of HCM [15,16] with late clinical onset [17,18], while other studies have described MYBPC3 mutations to be associated with more severe phenotype [19][20][21][22]. Furthermore, some studies have demonstrated that HCM patients with multiple causative gene mutations have poorer prognosis in terms of earlier disease onset, increased LV hypertrophy, and increased frequency of heart failure and sudden cardiac death, when compared to those carrying a single mutation [11,[23][24][25][26]. Studies have also reported that 70% of the mutations in MYBPC3 are truncations and cause a more severe HCM phenotype than that associated with missense and in-frame deletions [27,28].
Therefore, the primary purpose of this study was to further define the spectrum of mutations in a cohort of patients examined in a specialized center for HCM and define the effect of compound mutations on clinical pathogenesis. We also asked if individual gene mutations and the protein domains they affect could be correlated with the severity of HCM. Our data demonstrate that MYBPC3 gene mutations are the predominant cause of HCM with high penetrance. Additionally, the severity of HCM was shown to be dependent on the predicted pathogenicity of the mutation, irrespective of the gene (MYBPC3 and MYH7), protein domain affected, or compound effects. 42.9±17.3 and 45.9±18.3 years, respectively. The overall age distribution between the groups was comparable with no significant bias in any group ( Fig 1B and Table 1). The study population was uniformly distributed in terms of age and left ventricular (LV) function with the exception of six patients who presented with HCM that had progressed to the decompensated stage (dHCM) (Fig 1C, black marks). The clinical characteristics of the study population, as evaluated on the basis of American Heart Association (AHA) clinical guidelines, showed that 24.0% (n = 36) had HCM without obstruction, while 25.3% (n = 38) had hypertrophic obstructive cardiomyopathy (HOCM) (Fig 1D). Interestingly, concentric hypertrophy (hypertension with IVS>10mm and IVS/PWT 1.1) was identified in 4.7% (n = 7), but Study subject characteristics across three cohorts and clinical evaluation profile. A, Distribution of males and females in the study population represented as percentages. B, Age and gender distribution among the three cohorts, probands with definitive HCM/HOCM or dHCM clinical phenotype (symptomatic carriers, SC), family members with no HCM phenotype (asymptomatic carriers, AC), and non-carriers (NC) unrelated community-based population confirmed by echocardiography to have normal cardiac structure and function. Data are represented as means with standard error of mean (SEM) C, Age-based distribution of left ventricular ejection fraction across the three cohorts; red markers indicate HCM, blue markers indicate AC, and green markers indicate NC. Black markers indicate dHCM subjects with ejection fraction 45%. D, Diagnostic subclassification of subjects with gene mutation represented as a percentage of total. E-J panels show various echocardiographic parameters that reflect HCM severity presented as mean with standard error of mean. IVS, interventricular septum thickness (E); LVPW, left ventricular posterior wall (free wall) thickness (F); E/e' ratio, ratio of early transmitral flow (E) to left ventricular early diastolic velocity (e') (G); E/A ratio, ratio of early (E) to late (A) ventricular filling velocities (H); LVEF, left ventricular ejection fraction (I); LVOT-Gdt, LV outflow tract peak gradient (J). * indicates p value of 0.01 or lower by one-way ANOVA analysis.
only among asymptomatic carriers and not within non-carriers (data not shown). Hypertension and diabetes were observed in 38.0% (n = 57) and 7.3% (n = 11) of subjects, respectively (Table 1). Electrocardiographic assessment of patients and first-degree family members showed no significant ECG abnormalities in 55 (36.7%), paced rhythm in 14 (9.3%), and bundle branch block in 10 (6.7%) patients (6 with right and 4 with left bundle branch block) ( Table 2).
Electrocardiographic analysis showed that the majority of subjects had normal sinus rhythm, while subjects with and without HCM phenotype showed some of the features commonly seen among HCM patients. These included ventricular hypertrophy (n = 20), bundle branch blocks (n = 10) and left atrial enlargement 16 (13.9%). Five subjects had previously undergone septal reduction, and 14 had either an implanted pacemaker or implantable cardioverter defibrillator (ICD) ( Table 2). Phenotypically, septal and left ventricular posterior wall thickness and E/e' ratio was significantly higher in HCM patients compared to both asymptomatic carriers and non-carriers ( Fig 1D-1F, p<0.001). Interestingly, left ventricular ejection fraction in symptomatic carriers (68.4 ± 0.7%) was significantly higher than the non-carriers (61.2 ± 1.0%, p<0.014), and asymptomatic carriers (66.1 ± 0.9%, p<0.04) showed a significant increase in LVEF compared to the non-carriers ( Fig 1I). The resting left ventricular outflow tract (LVOT) gradient was !30mmHg in 15 subjects (12.8%) and increased by ! 50mmHg with provocation in 14 (12.0%) patients (data not shown). Of the 55 familial samples screened, 38 (69.1%, probands and first-degree family members) were clinically found to have HCM/ HOCM/dHCM, while 17 (30.9%) were asymptomatic carriers at the time of evaluation (S1 Table). The overall disease penetrance was 69.6% (80 of 115) of the subjects who were symptomatic carriers with HCM phenotype (S2 Table).

Mutation profile of study cohort
Eighteen genes currently known to cause HCM were analyzed using next-generation sequencing (NGS) in these samples (S3 Table). Within the study cohort (n = 150), 39.3% had genetic variations in MYBPC3, and 20.7% had variations in MYH7, followed by cardiac troponin T (5.3%), α-galactosidase (2.7%), and tropomyosin and myosin light chain 3 (2% for both). One hundred twenty-five genetic mutations were found among the 18 genes, 82 of them being distinct changes. Of these distinct mutations, MYBPC3 contributed 50.4%, while MYH7 contributed 24.8% (Fig 2A). Of all potentially pathogenic mutations, including likely pathogenic, pathogenic and variations of conflicting pathogenicity, identified among the subjects (n = 115), it is significant to note that unique MYBPC3 and MYH7 contributed 50% and 24.4%, respectively, together accounting for about 74.4% of the total pathogenic mutational load ( Fig 2D). These results suggest a very clean population of HCM subjects, asymptomatic carriers and unrelated non-carriers. Upon closer scrutiny, we identified all classes of genetic variations in the dataset with most (73.6%, n = 92) being missense, followed by nonsense (7.2%, n = 9) and frameshift mutations (6.4%, n = 8). A total of 12.8% (n = 16) had variations in intronic or the upstream regulatory region of the gene (Fig 3A and 3B). Seven of the 63 mutations in MYBPC3 were novel, while 2 novel mutations were identified among the 31 MYH7 mutations (Table 3).

MYBPC3 mutations account for most genetically linked HCM
Disease penetrance with respect to individual genes was calculated to be 77.8% (n = 21) for MYBPC3 and 71.4% (n = 10) for MYH7. A broader classification provided a penetrance of 78.4% (40 of 51) for sarcomeric genes and 66.7% (2 of 3) for non-sarcomeric genes (Table 4). Clinically, MYBPC3 was associated with all major phenotypes, particularly HCM and HOCM, but also dHCM (S2 Table). These data show that MYBPC3 mutations are the predominant cause of HCM in our patient cohort (n = 59), significantly higher than MYH7 (n = 31), in contrast to previous reports [29,30]. The remaining genetic contributions were similar to those already published, which showed that mutations from sarcomeric genes were far greater (~90%) than mutations from non-sarcomeric genes in less than 10% of the samples.

HCM phenotype is independent of underlying gene mutation
Since MYH7 gene variations were previously reported as having a more severe phenotype compared to MYBPC3 mutations, we compared phenotypic outcomes that resulted from variations in these two highly mutable genes. We found that MYBPC3 (n = 59) and MYH7 (n = 31) were still the genes most frequently associated with HCM. However, no significant phenotypic differences in symptomatic carriers were found between those caused by MYBPC3 or MYH7, either by echocardiogram or by comparing other sarcomeric (n = 18) with non-sarcomeric gene mutations (n = 12) (Fig 4). We also compared symptomatic and asymptomatic carriers, and, again, no significant difference was found between the two genetic variants (data not shown). Taken together, these data show that the resultant HCM phenotype was independent of the offending mutation, even though MYH7 and MYBPC3 were the most dominant cause of HCM.
Mutation dose increases the risk, but not severity, of HCM Double mutations, defined as the presence of one mutation, which is classified as pathogenic, likely pathogenic, or conflicting interpretation of pathogenicity, and a second non-benign mutation, were identified in 10 (6.7%) subjects within the cohort. Nine of the ten subjects exhibited HCM phenotype, but further investigation revealed that the one asymptomatic   subject had an MYH7 mutation that could be characterized as conflicting pathogenicity, while the second MYBPC3 mutation was defined as benign, resulting in exclusion of the subject from this subgroup. Distinct compound gene mutations, together with their effect on the gene product and likelihood of pathogenicity, are detailed in S4 and S5 Tables. While both single and double mutants exhibited highly significant phenotypic difference compared to control population (Table 5), no phenotypic differences were observed between these two mutant groups (Fig 5A-5F). From these data, it is apparent that compound mutations may substantially increase the risk of developing HCM, without increasing the severity of the HCM phenotype. However, a larger population of patients presenting with compound mutations is required to confirm this.
Pathogenic missense mutations are location-agnostic MYBPC3 and MYH7 encode the sarcomeric proteins cardiac myosin binding-C (cMyBP-C) and β-myosin heavy chain (β-MHC), respectively. As major contractile proteins, both have multiple domain-dependent binding partners. We divided cMyBP-C domains into three functional clusters: C0-C2 that directly interacts with and regulates myosin, C3-C6 that acts as a tether, and C7-C10 that anchors cMyBP-C to the thick filament. Here we compared mutations in these regions of cMyBP-C and the three domains of β-MHC, including motor, actin-binding, and coiled-coil/rod (Fig 6), to echocardiographic outcome. We found no significant association between the site of the mutation and the functional effects of HCM in either cMyBP-C  (Fig 7) or in β-MHC (Fig 8). Unfortunately, the small sample size made it difficult to determine domain association in β-MHC.

Discussion
This study analyzed the interaction of multiple cardiovascular mutations in key causative genes in a cohort of 150 HCM patients, family members who are asymptomatic carriers and non-carriers. The effects of domain-specific mutations and double-dosing mutations in MHY7 and MYBPC3 were also investigated. HCM is a unique genetic cardiac disease, which, despite being very common, has a high degree of variability in phenotypic expression and disease penetrance [31]. Even among familial subjects, the variation in phenotypic expressions of the same genotype can be significant [7,31]. While most studies have identified specific variations that track with the severity of the phenotype [32,33], the present study looked at all variations seen in the top 18 genes known to cause HCM, and we identified those associated with severe phenotype, irrespective of gender, age and other confounding variables. Among our study subjects, we found that disease penetrance was very high at 69.6% (80 of 115 genotype-positive subjects) and that a high proportion of the familial population (family of probands) presented with an HCM phenotype (38 of 55, 69.0%). This is higher than that previously reported (41% and 15%) [17]. According to Marian and Braunwald [34], over 1,500 mutations have been identified in 26 genes reported to cause HCM. These include nine sarcomeric genes (ACTC1, MYBPC3, MYH7, MYL2, MYL3, TNNC1, TNNI3, TNNT2, and TPM1) and nine non-sarcomere-encoding genes (CAV3, GLA, LAMP2, MTTG, MTTI, MTTK, MTTQ, PRKAG2, and TTR). It was found that more than 94% of HCM subjects expressed mutations in at least one of these genes (34). Interestingly, more than 51.2% of subjects analyzed carried a mutation in MYBPC3, while only about 24.8% of patients with HCM carried mutations in MYH7. These results stand in contrast to a previous study [35], but they are consistent with similarly sized patient studies, one from a Chinese population [36] and one from a larger Canadian population [37]. While the former study only evaluated MYBPC3 and MYH7 [36], the latter study was a meta-analysis that evaluated over 1,400 patients with multiple sarcomeric genes [37]. Mutation analysis of the 18 genes studied in the present work showed that some genetic variations could be identified in both symptomatic carriers and asymptomatic first-degree relatives of mutant carriers. Thus, determining the contribution of these mutations to the phenotype was difficult, as they were likely noncontributing private mutations [9,38]. Accordingly, such mutations were removed from analysis, and we focused only on gene mutations in that were likely to contribute to the pathogenesis of HCM. Contrary to prevailing evidence that MYH7 mutations result in a more severe phenotype than MYBPC3 mutations [17,35,39,40], we found both MYBPC3 and MYH7 to be strongly pathogenic with no significant differences in echocardiographic or clinical parameters of patients with mutations in either gene. The relative size of the MYBPC3 and MYH7 genes to other HCM-associated sarcomeric genes may increase the probability for disease causing mutations, which could be one of the reasons for the increased frequency of HCM-causing mutations in these two genes [41]. The large gene size (>3000bp) of MYBPC3 and MYH7 could be an important factor in the frequency of mutations identified in both these genes relative to other smaller sarcomeric genes (~<1500bp). Owing to increased baseline frequency of mutations, it is also possible that the number of sublethal (pathogenic, but not lethal in pediatric age groups) mutations that slip through the evolutionary filter is also higher compared to diseases like cancer. Among the genes analyzed, MYBPC3 mutations contributed to more than half of all HCM cases, consistent with previous reports [36,37]. Interestingly, immediate family members who also carried the variation, but did not have clinically apparent HCM at the time of evaluation, still had increased left ventricular ejection fraction and abnormal diastolic function compared to non-carriers, outcomes suggestive of subclinical HCM [42] Nine patients with HCM also had double mutations compared to 105 subjects with single mutation. In the present study, a subject was considered to carry a double mutation if 1) one mutation was determined to be pathogenic, likely pathogenic, or have conflicting interpretation of pathogenicity, and 2) a second mutation was determined as non-benign. Importantly, all subjects who presented with double mutation also had a strong cardiomyopathy phenotype [43]. Notwithstanding, comparison of HCM patients with either single or double mutation did not display any significant difference in echocardiographic characteristics, indicating that the presence of a secondary mutation does not significantly alter the pathogenesis of HCM or clinical phenotype when compared to a single mutation. This is to be expected since HCM is a dominant disease such that any failure (single mutation) in the critical path would lead to a phenotype, and acute genetic lesions would be embryonically lethal.
The protein products of MYBPC3 and MYH7, cMyBP-C and β-MHC, are key components of the sarcomere, and both exert their functions through extensive protein-protein interactions. In particular, extensive interactions take place between cMyBP-C and β-MHC. With such close ties, mutations in either of the two genes may significantly alter the structure and function of cardiac muscle, but together, they contribute to >80% of all familial HCM [43]. β-MHC is a globular protein with a motor (M) domain, which includes two actin-binding domains (AB) that directly interact with the actin fiber tracks (thin filaments) and a coiled-coil (CC) domain that intertwines with the CC domain of other myosins. Thus, the M domain allows force transmission on the Z-lines and, in turn, the whole myocyte for contraction ( Fig  6). On the other hand, cMyBP-C consists of 11 domains labeled C0 to C10, consisting of immunoglobulin-like or fibronectin-like domains. It crosslinks the thick and thin filaments of the sarcomere and regulates actomyosin interaction, thereby controlling the rate and force of contraction. To achieve this, cMyBP-C can be broadly divided into three functional clusters: C0-C2 that can differentially interact with the myosin motor heads and actin filament, C3-C6 that acts as a neck/tether, and C7-C10 that intertwines with β-myosin coiled-coil tail to anchor in the thick filament (Fig 6). In the present study, we analyzed whether mutations in any of the interacting domains between cMyBP-C and β-MHC could affect the pathogenicity of the HCM-causing mutations. Surprisingly, we did not find any change in phenotypic effects, as reflected by echocardiographic findings, among the various protein domain mutations of cMyBP-C or β-MHC in our study population. Despite multiple protein-protein interactions between cMyBP-C and β-MHC, it is clear from the data that a pathogenic mutation in any location of either protein would result in the failure of the protein as a whole, in turn causing HCM phenotype independent of mutation change in the DNA at any particular locus. In conclusion, we reported that MYBPC3 is the most commonly mutated gene among our study population with hypertrophic cardiomyopathy, almost twice that of MYH7, which was previously suggested as the most common genetic cause of familial HCM. While it is possible that the higher frequency of MYBPC3 is specific to our study population, these results are consistent with previous studies from different ethnic groups [36,37,44] indicating it is a generalized phenomenon and not based on, or biased by, race or ethnicity. Based on NGS, we evaluated the prevalence, pathogenicity, frequency and location of mutations relative to the clinical phenotype, and we provided strong evidence that MYBPC3 and MYH7 can be considered coequal pathogenic causes of familial HCM whose pathogenicity is independent from the location of the mutation on the protein. We also provided clear evidence that compound mutations increase the risk of HCM pathogenesis, and that asymptomatic carriers display subclinical hypertrophic features in terms of wall thickness and left ventricular ejection fraction. Study participants ranged in age from 8 months to 92 years and, moreover, showed strong and consistent HCM phenotype at baseline. Consequently, this study may provide an invaluable platform for the longitudinal evaluation and monitoring of these study subjects who were asymptomatic carriers during the study period.

Study design and population
This study conforms to the principles of the Declaration of Helsinki, and it was approved by the respective Institutional Review Boards. Written informed consent was obtained from all study participants. All HCM patients and their relatives evaluated between July 2010 and August 2016 at the HCM specialized clinic at Aurora St. Luke's Medical Center, Milwaukee, WI, were included. Normal control subjects, who were non-carriers for gene defects in the 18-gene panel, were recruited throughout the upper Midwest of the United States and were clinically evaluated and imaged at Loyola University Chicago, Maywood, IL. Consenting participants were clinically evaluated, and genetic evaluation was offered for their first-degree relatives and non-carriers. Eighty-two consecutive subjects and 33 of their first-degree relatives were included in this study, together with 35 non-carriers. Subjects were selected based on their diagnosis of HCM, as determined by maximal left ventricular wall thickness of ! 15mm, following echocardiographic testing and clinical symptoms, such as chest pain, dyspnea, fatigue, fainting and palpitations. Subjects with secondary factors, such as aortic valve stenosis and myocardial infarction, were excluded from the study. Both Loyola University Chicago and St. Luke's Medical Center Institutional Review Boards approved the genome-phoneme analyses.

Electrocardiogram
All participants were tested with 12-lead ECG with rhythm strip to assess for electrical disturbance and potential arrhythmic disorders associated with HCM [45,46].

Echocardiography
Cardiac function and morphometry were measured, as described previously by the authors [47]. M-mode and two-dimensional echocardiographic studies were performed from standard transthoracic windows using Acuson Sequoia™ (Siemens Medical Solutions, Malvern, PA) and Vivid 7 (GE Healthcare, Waukesha, WI). The echocardiographic equipment was used in accordance with the recommendations of the American Society of Echocardiography. Examinations were performed in a blinded manner by experienced sonographers, and cardiologists interpreted the results [48]. Left ventricular (LV) posterior wall thickness and interventricular septum (IVS) were determined in millimeters using M-mode measurements. Simpson's method was used to measure left ventricular ejection fraction (LVEF), and diastolic dysfunction was evaluated using E-wave and A-wave diastolic transmitral velocities, E/A ratio, ratio of early transmitral flow (E) to left ventricular early diastolic velocity (e'), E/e' ratio and deceleration time. These measurements were obtained using spectral pulsed-wave Doppler recordings at the tip of mitral valve leaflets in the apical four-chamber view. Left atrial area and length were measured in apical four-chamber and two-chamber views. Left atrial volume was obtained in orthogonal views and calculated using the following algorithm: 0.85 x Left Atrial Area (Four-Chamber) x Left Atrial Area (Two-Chamber or Apical Long Axis) / (Average of the Two Lengths obtained from Orthogonal Views).
Genetic testing DNA sequence analysis of 18 genes associated with HCM was performed at GeneDx (Gaithersburg, MD 20877). Briefly, peripheral blood samples, or oral rinses, were collected from participants, and DNA was isolated from the samples. DNA quantity and quality were checked, and a specific amount of DNA was used to perform targeted gene sequencing of the 18 genes (S1 Table). This was achieved using at least a 100bp paired-end read using the Tru-Sight Cardio sequencing panel kit on Illumina HiSeq sequencers (Illumina, San Diego, CA, USA). Resultant raw DNA reads were checked for quality and trimmed to remove adapter sequences. The sequences were then aligned against the human reference genome build (GRCh37/UCSC hg19), the variations in the sequences collated, reports from GeneDx further analyzed using online resources, e.g., ExAC, dbSNP and HGV, and variant pathogenicity established using the Clinvar database. Among the eighteen cardiomyopathy genes, the presence of variations in multiple genes in an individual was of particular interest. For this analysis, double mutation was defined as the presence of one mutation classified as pathogenic, likely pathogenic, or conflicting interpretation of pathogenicity, and a second non-benign mutation.

Statistical analysis
Wilcoxon rank-sum tests were used to assess the differences between patient groups for continuous variables, and chi-square test, or Fisher's exact test, was used for categorical variables. Continuous variables were summarized as mean ± standard error of mean (SEM); discrete variables were described as number and percentage. In order to identify predictors for outcomes, logistic regression analysis was used to calculate odds ratio with 95% CI and to examine the association of each variable with adverse cardiovascular events. All tests were two-tailed, and a P-value <0.05 was considered statistically significant. All statistical analyses were performed with Prism, version 7.4 (GraphPad Inc., La Jolla, CA).
Supporting information S1