|Home | About | Journals | Submit | Contact Us | Français|
Electrocardiographic QT interval prolongation is a risk factor for sudden cardiac death (SCD) and drug-induced arrhythmia. The clinical correlates and heritability of QT interval duration in African Americans have not been well studied despite their higher risk for SCD compared to non-Hispanic whites. We sought to investigate potential correlates of the QT interval and estimate its heritability in the Jackson Heart Study (JHS).
The JHS comprises a sample of African Americans residing in Jackson, Mississippi, of whom 5302 individuals with data at the baseline examination were available for study. JHS participants on QT-altering medications, with bundle branch block, paced rhythm, atrial fibrillation/flutter or other arrhythmias were excluded resulting in a sample of 4660 individuals eligible for analyses. The relation between QT and potential covariates was tested using multivariable stepwise linear regression. Heritability was estimated using SOLAR in a subset of 1297 JHS participants in 292 families; the remaining sample included unrelated individuals. In stepwise multivariable linear regression analysis, covariates significantly associated with QT interval duration included RR interval, female sex, QRS duration, age, lower potassium, hypertension, body mass index, coronary heart disease, diuretic use, and Sokolow-Lyon voltage (p ≤ 0.01 for all). The heritability of QT interval duration in age-, sex- and RR-interval-adjusted and fully-adjusted models was 0.41 (SE 0.07) and 0.40 (SE 0.07, p<10−11 for both), respectively.
There is substantial heritability of adjusted QT interval in African Americans supporting the need for further investigation to identify its genetic determinants.
Sudden cardiac death (SCD) claims 300,000 lives annually in the United States.1 African Americans have higher rates of SCD than European Americans but the causes of the difference in event rates are unknown.2–5 Risk factors for SCD in the general population include coronary artery disease, hypertension, tobacco use, electrocardiographic QT interval prolongation, left ventricular hypertrophy, diabetes, body mass index and family history.6–9 The increased risk of SCD in African Americans could result from environmental differences, including diet or smoking, from disparities in access to health care or from aggregation of multiple risk factors with strong biologic underpinnings, such as increased rates of hypertension and left ventricular hypertrophy.10
Family history is a consistent risk factor for SCD.6,11–13 Whether heritable factors contribute to the higher risk of SCD in African Americans compared to European Americans is unknown. A common missense variant in the SCN5A sodium channel (S1102Y) found exclusively in individuals of African ancestry has been found to increase risk of ventricular arrhythmias and sudden cardiac death.14,15 A common missense variant in the beta-2 adrenergic receptor (Gln27Glu) that is more frequent in African Americans has been found to increase risk of SCD in European Americans, and had borderline significance in a small African American sample.16 While an aggregation of many such genetic variants could contribute incrementally to differential SCD risk, existing studies of SCD in African and European Americans lack statistical power to define the full spectrum of common and rare variants that contribute to SCD risk.
The electrocardiographic QT interval is a reflection of myocardial depolarization and repolarization time. QT interval prolongation17,18 or shortening19 is associated with increased risk of SCD in community-based samples as well as in familial congenital syndromes of long and short QT duration and ventricular arrhythmias.20–22 Moreover, QT interval prolongation and ventricular arrhythmia upon exposure to cardiac and non-cardiac medications is a life-threatening side effect and a major barrier to drug development.23 The widespread availability of QT interval measurements in multiple cohorts makes it an attractive intermediate SCD trait and enables well-powered studies of the determinants of myocardial repolarization in the general population.
It has been shown that RR-interval-, sex- and age- adjusted QT interval duration in community-based samples of European ancestry is heritable.24 Heritability, quantifying the tendency of related individuals to have more similar trait values than unrelated individuals, estimates the proportion of variation in a trait that is explainable by additive genetic factors. Heritability in the Framingham Heart Study, a sample of European ancestry, has been estimated at 0.35, indicating that 35% of variation in adjusted QT interval duration was attributable to additive genetic factors.24 While a large-scale study by Dekker et al. has established several clinical correlates of QT interval duration on univariable analysis in African American community-based samples,18 limited data on the independent correlates of QT interval duration on multivariable analysis or its heritability in African Americans are currently available.
We sought to define the clinical correlates of QT interval duration in 4660 Jackson Heart Study (JHS) participants and to determine the heritability of multivariable-adjusted QT interval duration in the family-based subset of 1500 JHS participants. Demonstration that a significant proportion of multivariable-adjusted QT interval variation in African Americans is attributable to genetic factors would support efforts to identify those genetic factors which could then be examined for association with SCD risk or drug-induced arrhythmias in the general population.
The Jackson Heart Study (JHS) cohort includes individuals recruited from the greater Jackson, Mississippi area25 some of whom were originally participants in the Atherosclerosis Risk in Communities Study (ARIC).26 Men and women aged 35–84 years were included as well as family members aged 21 or older who agreed to participate in the family study component27 with a final study cohort of 5302 participants aged 21–95. DNA samples were extracted from all JHS participants who consented to genetic studies. The pedigree data stored in the software package Progeny 2000 (www.progenygenetics.com) were checked for inconsistencies including same-sex married couples, pedigree loops, incompatible age, and ID problems. A total of 1500 individuals in 292 families had microsatellites genotyped by the Mammalian Genotyping Service (research.marshfieldclinic.org/genetics/home/index.asp). Using microsatellite markers, pedigree relationships underwent modification to resolve Mendelian and single-marker inconsistencies.27
Of 5302 total JHS participants in the unrelated (n = 3802) and family (n = 1500) components, those without informed consent (0.2%) or electrocardiographic data (0.3%), those on digoxin (1.5%) or a QT-prolonging medication (4.8%, listed in Supplemental Table 1), and those with left or right bundle branch block (2.2%), atrial fibrillation or atrial flutter (0.4%), other arrhythmias (0.3%), or a pacemaker (0.2%) were excluded, resulting in the final sample of 4660 individuals who were eligible for analysis (n = 3363 in the unrelated sample and n = 1297 in the family sample). The total sample of 4660 individuals contributed to association analysis of clinical covariates and the related subsample of 1297 individuals in 292 families contributed to the heritability analyses.
The baseline examination consisted of a home interview, self-administered questionnaires, and a clinic visit. Medications taken in the prior 2 weeks were brought to clinic and transcribed verbatim with subsequent coding by a pharmacist using the Medispan dictionary (www.medispan.com/index.aspx) with classification according to the Anatomical Therapeutic Chemical Classification System (www.who.int/classifications/atcddd/en/). QT-prolonging medications were annotated using the lists of medications maintained by qtdrugs.org (www.arizonacert.org/medical-pros/drug-lists/drug-lists.htm, accessed November, 2006). Digoxin use, which shortens QT interval duration, was also annotated. After an overnight fast, anthropometric and seated blood pressure measurements were obtained and venipuncture was performed. Blood pressure was measured after a 5 minute rest by trained staff in the right arm of seated participants, whose back and arm were supported, using an appropriately sized cuff and a Hawksley random-zero sphygmomanometer. The average of 2 measures taken one minute apart was recorded. Hypertension was defined as a measured blood pressure ≥140/90 mmHg or use of antihypertensive medication. Presence of coronary heart disease (CHD) was determined by self-reported history, physician-diagnosed- or ECG-determined myocardial infarction, or self-reported angioplasty. Presence of diabetes was established by self-report or by a measured fasting glucose of ≥126 mg/dl, or use of insulin and/or oral hypoglycemic agents . Blood samples from the fasting venipuncture were analyzed for several analytes including potassium, glucose, total cholesterol, high-density lipoprotein cholesterol, and triglycerides standardized according to the Centers for Disease Control and Prevention as previously described.25 Low-density lipoprotein cholesterol was calculated according to the Friedewald equation. Higher alcohol use was defined as consumption of >14 drinks per week for men and >7 drinks per week for women. Four physical activity index scores, for Active Living, Work, Sport, and Home and Family Life, were calculated based on the physical activity cohort survey administered during the home induction interview. For each index, the values ranged from one (low) to five (high). The four individual index values were summed to obtain the total physical activity index (ranging from 4 to 20).28
The clinical examination included supine 12-lead digital electrocardiography using a Marquette MAC/PC digital electrocardiograph to record, store, and transmit ECG data via phone modem to the Electrocardiographic Reading Center (ECGRC) at the University of Minnesota.25 The ECGRC used the Minnesota Code Modular ECG Analysis System (MC-MEANS) computer program for all baseline examination ECGs, which has performed well when compared to other computer algorithms and manual reads by cardiologists.29,30 MC-MEANS generates a representative averaged beat and determines onsets and offsets of P, QRS, and T waves simultaneously over all leads. The PR interval is the time from onset of the P wave to onset of the QRS complex. The QRS duration is the time from the onset of the Q wave to the offset of the S wave. The QT interval is the time from the onset of the Q wave to the offset of the T wave. These data are the basis for all continuous measurements and are coded using a morphology decision logic algorithm. These preliminary codes were then subjected to hierarchical exclusion rules from the Minnesota Code. Sokolow-Lyon voltage was calculated as the sum of the S-wave amplitude in lead V1 and the greater of the R-wave amplitudes in V5 and V6.
Participant baseline characteristics such as age, body mass index (BMI), systolic and diastolic blood pressure, use of antihypertensive medication, use of diuretics, hypertension, fasting glucose, fasting lipid profile (total, LDL, and HDL cholesterol and triglyceride level), serum potassium level, type 2 diabetes status, CHD status, physical activity index and electrocardiographic QT interval, RR interval, PR interval (lead II) and QRS duration were studied. The characteristics were summarized descriptively, in sex-pooled and sex-specific analyses, using means and standard deviations for continuous parameters and using percentages for categorical parameters. Sex-specific means, standard deviations (SD), interquartile ranges (IQR) and ranges for QT interval adjusted for heart rate using Bazett's correction (QTc) were computed.31
Linear regression analyses were performed to identify baseline demographic and clinical characteristics associated with QT interval duration. Initially, linear regression models that included QT interval duration as the dependent variable and a clinical characteristic along with three concomitant variables (age, sex, and RR interval) as independent factors were utilized to explore the relationship of each clinical characteristic individually with QT interval duration. Factors that were statistically significant (p<0.05) in these models were selected for stepwise model building. A separate analysis of systolic and diastolic blood pressure, hypertension status, and use of anti-hypertensive medications, all of which individually were significant in the age-, sex-, and RR-adjusted analysis, was conducted using a linear regression model with stepwise selection, after forcing in age, sex, and RR interval. Based on the model R-square statistic and Mallows' Cp statistic, only hypertension status was chosen for inclusion in the final model as the predictor of QT interval with the greatest explanatory power among blood pressure-related factors. All factors identified for the final model selection were then tested using a multivariable linear regression model with stepwise backward elimination and retention of variables with p<0.05 in the final model.
Finally, two linear regression models were examined to obtain heritability estimates for the sample. The first model included only age, sex, and RR interval (model 1), while age, sex, RR interval and all additional factors with p<0.05 in the multivariable model were included in the second model (model 2). Heritability analyses in either model were restricted to the 1297 participants in 292 families with complete information for all characteristics in model 2 to allow comparison. The values of the standardized adjusted residuals, Winsorized at 3 standard deviations, obtained from the two models, were analyzed using variance components methods implemented in SOLAR, which parses genetic and nongenetic components of variation in a trait, to estimate heritability.32
Baseline characteristics after exclusion of individuals on QT-altering medications, or with left or right bundle branch block, arrhythmia or paced rhythm are presented in Table 1. The sample population was obese on average with mean BMI 31.7 kg/m2 and had a high prevalence of hypertension (60%), diabetes (17%), and CHD (5.6%). The mean QT interval was longer in women (416.5 msec, SD 30.6) compared to men (407.4 msec, SD 30.0, p for difference = ) as expected. The mean RR interval was 981 msec (SD 159) in men and 942 (148) in women. We compared the heart rate-adjusted QT interval in the 4660 men and women not on any QT-altering medications to that in the 223 individuals taking QT-prolonging medications and in the 45 individuals taking digoxin, known to shorten QT interval duration. Men who were not taking any QT-altering medication had a QTc of 413.1 msec (SD 25.3), compared to 422.7 msec (SD 25.3, p for difference = 0.008) for men on QT-prolonging medication and 412.3 msec for men on digoxin (SD 31.5, p for comparison to group on no QT-altering medication = 0.90). Women not taking any QT-altering medication had a QTc of 430.8 msec (SD 24.0), compared to 440.9 msec (SD 25.4, p < 0.0001) for women on QT-prolonging medication and 406.9 msec (SD 33.8, p < 0.0001) for women on digoxin.
In the initial linear regression analyses (adjusting only for age, sex, and RR interval), the following factors were significant predictors of QT interval duration (p<0.05) when added individually: age, sex, RR interval, QRS duration, PR interval, hypertension, systolic blood pressure, diastolic blood pressure, antihypertensive medication, diuretic use, Sokolow-Lyon voltage, body mass index, current smoking, physical activity index, prevalent coronary heart disease, and serum potassium (Table 2). In a stepwise linear regression analysis of hypertension status, systolic and diastolic blood pressure and anti-hypertensive use, hypertension status entered into the model first and explained the greatest proportion of variation in QT interval duration compared to the other three variables and was thus used in subsequent analyses for ease of interpretation (data not shown). In the backward stepwise multivariable linear regression analysis, the following factors were independently positively associated (p<0.05) with increase in QT interval duration: RR interval, female sex, QRS duration, age, hypertension, BMI, CHD status, diuretic use and Sokolow-Lyon voltage (Table 3). Serum potassium was negatively associated with QT interval duration (p<0.0001). The majority of variation in QT interval duration was explained by RR interval (47%), female sex (5%), QRS duration (3%) and age (2%). All other factors explained less than 1% of variation in QT interval duration. The correlation of residuals from the age-, sex- and RR-interval-adjusted model (54% variation explained) and those from the fully-adjusted model (58% variation explained) was 0.99.
The estimated heritability of RR interval-, sex- and age-adjusted QT interval residuals was 0.40 (SE 0.07, p<10−11) and for residuals from the fully-adjusted model shown in Table 3 was 0.41 (SE 0.07, p<10−11, Table 4).
In this study, several independent clinical correlates of QT interval duration in a large, African American community-based sample were identified and were consistent with several correlates previously reported in the literature. It was demonstrated for the first time that adjusted QT interval duration was a heritable trait with a substantial proportion of variation explained by additive genetic factors in African Americans.
QT interval duration was found to be positively associated with RR interval, female sex, QRS duration, age, hypertension, body mass index, coronary heart disease, use of diuretics, and Sokolow-Lyon voltage and negatively associated with serum potassium. These observations were consistent with findings reported by Dekker et al. in whites and blacks from the Atherosclerosis Risk in Communities study that increasing QTc quintiles showed significant trends for increasing age, hypertension prevalence, and body mass index and decreasing serum potassium on adjustment for age, sex and race when these factors were considered individually.18 No significant associations of triglycerides, total cholesterol, HDL cholesterol, or diabetes with QT interval was found in JHS regression models adjusted for age, sex and RR-interval but were by Dekker et al. This could reflect chance differences between JHS and ARIC cohorts or reduced power in the JHS sample because of the smaller total sample size. The two studies included only 1601 individuals in common who were examined between 1987 and 1989 in ARIC and between 2000 and 2004 in the JHS. Alternatively, some of the associations in the analyses by Dekker et al. could have been driven by effects in the white component of the ARIC cohort in the race-adjusted analyses in the combined sample.
In the Jackson Heart Study, there was a strong association of increasing QT interval with diuretic use and with decreasing serum potassium. It would be expected that diuretic use might be associated with QT interval duration in a community-based sample because of its relationship with total body potassium loss. The observation that serum potassium and diuretic use were independently significant in the same model could reflect the known impact of diuretic-induced hypomagnesemia on QT interval duration33 the incomplete correlation between one-time serum potassium measures and total body potassium stores.
Prospective studies have shown that QT interval duration is predictive of incident coronary heart disease, even after adjustment for known coronary heart disease risk factors in individuals of African or of European ancestry.18,34,35 In the JHS, the QT interval was shown to be positively associated with prevalent coronary heart disease, which could be due to an intrinsic predictive nature of adjusted QT interval on incident CHD or some reflection of acquired derangements of depolarization or repolarization that follow the development of CHD or aggregation of its risk factors.
In addition, a positive association was found between QT interval duration and increasing Sokolow-Lyon voltage, an electrocardiographic correlate of left ventricular mass. Left ventricular hypertrophy, measured by electrocardiography or echocardiography, has been associated with QRS lengthening (presumably from myocardial hypertrophy and fibrosis) and QT prolongation in hypertensive subjects.36,37 It is important to note that the large sample size of over 4000 individuals means that even clinically insignificant effects on QT interval duration can reach strong statistical significance. The partial R2, reflecting the fraction of variation explained by a clinical factor, was 0.47 for RR interval, 0.05 for female sex, 0.03 for QRS and 0.02 for age. All other covariates explain less than one percent of variation in QT interval and thus contribute little to the regression model; these could easily be dropped from analyses without altering results substantially. Because the QRS duration is a subcomponent of the QT interval, it may not be appropriate to adjust for it and most studies do not.
There is evidence that a substantial fraction of QT interval variability after adjustment for known covariates is explained by additive heritable factors in European-derived samples. The finding that 40–41% of variation in adjusted QT interval in African Americans is attributable to genetic factors is consistent with previous reports in individuals of European ancestry in the Framingham Heart Study (35%) and in the Family Heart Study (34%).24,38 It is important to establish that a heritable basis for a complex trait such as QT interval duration exists before embarking on genetic studies. Because of the known influences of diurnal variability, effects of adrenergic stimulation and measurement error on QT interval measures, it is certainly possible that an even greater proportion of the biologically meaningful component of QT interval variation is explained by genetic factors.
Strengths of the current report include the large sample size, the broad range of precisely measured covariates, the simultaneous adjustment for multiple risk factors and the availability of a family component to allow estimation of heritability of QT interval in African Americans. Limitations to the study must be acknowledged. The power to detect any true underlying association of a factor with QT interval is a function of both the strength of effect and the prevalence of the factor. Thus, the failure to find association of higher alcohol use or diabetes could be due to limited power. Lastly, QT interval measurements are limited by ambiguity of the T-wave offset and different computer algorithms can define this differently. However, the same well-validated computer algorithm was applied to the entire sample and any imprecision in T-wave offset would be expected to add noise to the QT interval trait and only bias association findings to the null.
Given the well-supported relationship of QT interval prolongation with sudden cardiac death in the general population, the high rates of SCD in African American populations and the substantial heritability of QT interval duration in the Jackson Heart Study, efforts to define the genetic basis of QT interval variation in populations of African ancestry are warranted The availability of DNA in the large, community-based Jackson Heart Study sampled without regard to phenotype27 will enable large-scale tests for association, including admixture mapping,39 and linkage to identify the genetic variants that underlie the heritable component of myocardial repolarization in people of African ancestry. Given the recent successes in identifying common variants that influence QT interval duration in European-derived samples,40 it is reasonable to expect that such variants will be operative in African-derived samples as supported by a recent report in 1,497 non-Hispanic blacks.41 Whether such variants contribute to risk for SCD or drug-induced arrhythmias will require further work in this and other cohorts.
We would like to thank the Jackson Heart Study participants and JHS staff for their important contributions to the study of determinants of health in African Americans.
Funding Sources: The Jackson Heart Study is supported by National Institutes of Health by contracts N01-HC-95170, N01-HC-95171, and N01-HC-95172 from the National Heart, Lung, and Blood
Institute and the National Center for Minority Health and Health Disparities. Dr. Newton-Cheh is supported by an NHLBI K23 award (HL080025), a Doris Duke Charitable Foundation Clinical Scientist Development Award, and a Burroughs Wellcome Fund Career Award for Medical Scientists.
Disclosures: Dr. Newton-Cheh is supported to study the genetic basis of QT interval duration by grants from the NIH, Doris Duke Charitable Foundation and the Burroughs Wellcome Fund. No other potential conflicts of interest were reported.