|Home | About | Journals | Submit | Contact Us | Français|
A prolonged QT interval corrected for heart rate (QTc) is a major risk factor in patients with the long-QT syndrome (LQTS). However, heart rate-related risk in this genetic disorder differs among genotypes.
We hypothesized that risk-assessment in LQTS patients should incorporate genotype-specific QT correction for heart rate.
The independent contribution of four repolarization measures (the absolute QT interval, Bazett’s, Fridericia’s, and Framingham’s correction formulae) to the risk of aborted cardiac arrest or sudden cardiac death during adolescence, before and after further adjustment for the RR interval, was assessed in 727 LQT1 and 582 LQT2 patients. Improved QT/RR correction was calculated using a Cox model, dividing the coefficient on log(RR) by that on log(QT).
Multivariate analysis demonstrated that in LQT1 patients 100 msec increments in the absolute QT interval were associated with a 3.3-fold increase in the risk of life-threatening cardiac events (p=0.020), and 100 msec decrements in the RR interval were associated with a further 1.9-fold increase in the risk (p=0.007), whereas in LQT2 patients, resting heart rate was not a significant risk factor (HR=1.11; p=0.51; p-value for heart rate × genotype interaction = 0.036). Accordingly, analysis of an improved QT correction formula showed that patients with the LQT1 genotype required a greater degree of QT correction for heart rate (improved QTc=QT/RR0.8) than LQT2 patients (improved QTc=QT/RR0.2).
Our findings suggest that risk stratification for life-threatening cardiac events in LQTS patients can be improved by incorporating genotype-specific QT correction for heart rate.
The identification of genes associated with the congenital long-QT syndrome (LQTS) has had a major impact on understanding the molecular basis for ventricular arrhythmias and sudden cardiac death (SCD) in young patients without structural heart disease.1 Numerous advances have been made in the identification of risk factors for cardiac events in LQTS patients.2–6 However, there remains a substantial challenge to explain the widely observed genotype-phenotype variability of this genetic disorder.
A prolonged QT interval corrected for heart rate (QTc) is a major risk factor for cardiac events in LQTS patients,4–5, 7–8 and is usually assessed using the exponential Bazett 9 and Fridericia 10 correction formulae. However, heart rate-related risk of cardiac events in LQTS patients has been shown to occur in a gene-specific manner.3, 11–12 Specifically, 82% of the lethal arrhythmic episodes in LQT1 patients, who harbor mutations that impair the IKs current, are associated with exercise and faster heart rates, whereas in LQT2 patients, in whom IKs current is normal, symptoms occur mostly during night or with arousal triggers.11 This may be due to the fact that the repolarizing current IKs activates during increased heart rate, and is essential for QT interval adaptation during tachycardia.13 Thus, it has been shown that faster heart rates are associated with increased arrhythmic risk in LQT1 patients 12, 14 and it is possible that RR interval provides incremental prognostic information to mere assessment of QTc as currently calculated by QT correction formulae particularly in LQT1 patients with fast resting heart rates.
The present study was designed to evaluate whether risk stratification for life-threatening cardiac events in LQTS patients who carry the common LQT1 and LQT2 genotypes should incorporate a genotype-specific QT correction for heart rate.
The study population was drawn from subjects enrolled in the International Long-QT Syndrome Registry,2 for whom follow-up data were available from age 10 through 20 years. Genetic testing was performed on 3,374 members of 443 families enrolled in the registry, and identified 1309 patients with the LQT1 (727 patients from 165 proband-identified families with the KCNQ1 mutations) and LQT2 (582 patients from 176 proband-identified families with the KCNH2 mutations) genotypes. Patients with a mutation in both KCNQ1 and KCNH2 genes were not included in the current analysis. Mutations were identified by standard genetic tests.15–16 Informed consent for genetic and clinical studies was obtained from all subjects.
For each patient, data on personal and family history, cardiac events, and therapy were systematically recorded at each visit or medical contact. Clinical data were recorded on prospectively designed forms and included patient and family history and demographic, electrocardiogram (ECG), therapeutic, and cardiac event information. Upon enrollment in the Long QT Registry, a 12-lead ECG was obtained from each patient. From the first recorded ECG, the duration of the QT and RR intervals were assessed from lead II. The observer who made the QT and RR measurements was blinded to clinical status of patients. Follow-up data regarding β-blocker therapy included the starting date, type of β-blocker, and discontinuation date in case it occurred. Among patients who died, the usage of a β-blocker before death was determined retrospectively. The reported analyses used the LQTS analytic database version 18.
The primary end point of the study was time to aborted cardiac arrest ([ACA] requiring external defibrillation as part of the resuscitation) or LQTS-related SCD (death abrupt in onset without evident cause, if witnessed, or death that was not explained by any other cause if it occurred in a nonwitnessed setting), whichever occurred first, from age 10 through 20 years. We focused on life threatening events during adolescence, which is a time period with a high event risk among the LQTS population as demonstrated previously.7
We analyzed the risk associated with the following four repolarization measures: the absolute QT interval (Model 1); the QT interval corrected for heart rate using Bazett’s formula (Model 2: QTc[b] = QT/RR1/2); the QT interval corrected for heart rate using Fridericia’s formula (Model 3: QTc[f] = QT/RR1/3); and the QT interval corrected for heart rate using the linear Framingham formula 17 (Model 4: QTc[fram] = 0.154(1 - RR) + QT). To evaluate the incremental prognostic contribution of heart rate to the analyzed repolarization parameters among carriers of each genotype, all models were analyzed before- (Models A) and after- (Models B) further adjustment for the RR interval.
Bazett 9 and Fridericia 10 derived their QT correction formulae by regressing log(QT) on log(RR) in a population of normal subjects, and observing that the slope was −½ and −1/3, respectively. These analyses result in the following log-transformed formulae: log(QTc(b)) = log(QT) – ½log(RR); and log(QT(f)) = log(QT) – 1/3log(RR). We fit a Cox model that included both log(QT) and log(RR). If Bazett’s formula were optimal for risk stratification among LQT1 and LQT2 genotype carriers, the estimated coefficient on log(RR) should equal −1/2 times that on log(QT); if Fridericia’s formula were to provide optimal risk stratification, then the estimated coefficient on log(RR) would be −1/3 times that on log(QT). But neither Bazett’s nor Fridericia’s correction formula was specifically designed for risk stratification based on a specific endpoint, and it is possible that improved correction for heart rate might differ, depending on the intended use of the correction formula.
Using this methodology we also estimated an “improved” QT correction for heart rate among carriers of each genotype, by using the derived coefficients from the log(QT) and log(RR) model, using the following formula: QTc [new] = QT/RRα (α = −βlog(RR)/βlog(QT)). By “improved,” we refer here to risk stratification for SCD or ACA.
The clinical characteristics of study patients by genotype were compared using the chi-square test and the Fisher’s exact test for categorical variables, and the t-test or the Mann-Whitney-U test for continuous variables.
Cox proportional hazards regression models were used to assess the independent contribution of each of the four repolarization measures, with and without further adjustment for the RR interval, to the development of ACA or SCD during follow-up in LQT1 and LQT2 patients. Prespecified covariates included in each of the models were: syncope (defined as transient loss of consciousness that was abrupt in onset and offset), treatment with β-blockers, and gender (male age 10–15 years). The development of syncope and the administration of β-blocker therapy from age 10 through 20 years in an affected individual were evaluated in a time-dependent manner. We checked the proportional hazards assumption and found insufficient evidence of any violations of proportional hazards (p > 0.05 for each predictor interacted with time) after allowing gender to interact with age, since this particular violation of the proportional hazards assumption has already been well documented in our prior work.6–7 In a secondary analysis we standardized all QTc measures to have the same unit, analyzed hazard ratios (HR) per standard deviation (SD) of QTc and reported these data in the supplementary appendix.
To assess and compare the discriminatory ability of the models among carriers of each genotype, we calculated Harrell-Kremers generalized c-index of concordance 18 (http://mayoresearch.mayo.edu/mayo/research/biostat/upload/80.pdf) for each of the time-dependent adjusted Cox models (without including RR in the models), and compared this to the c-index based on replacing QTc by QTc(new).
Grouped jackknife estimates of standard errors were compared with the standard large-sample estimates assuming independence in order to determine whether adjusting inferences for the potential dependencies due to family membership appeared necessary.19 A 2-sided p<0.05 significance level was used for hypothesis testing. The statistical software used for the analyses was SAS version 9.2 (SAS Institute Inc, Cary, NC).
Baseline clinical and ECG characteristics and cardiac events during follow-up in carriers of the LQT1 and LQT2 genotypes are shown in Table 1. Gender distribution and the frequency of a family history of a LQTS-related SCD were similar in the 2 genotype groups. Baseline heart rates and the corrected QT intervals were similar among LQT1 and LQT2 patients, whereas the absolute QT interval was significantly longer among LQT2 patients as compared with LQT1 patients. The baseline RR and QT intervals were normally distributed in both LQT1 and LQT2 patients (Figure 1).
Multivariate analysis showed that the exponential Bazett formula was the only correction method that independently predicted the risk of life-threatening events among LQT1 patients, whereas among LQT2 patients all analyzed QT correction formulae exhibited a similar association with outcome (Table 2A). Thus, among LQT1 patients 100 msec increments in QTc(b) were associated with a significant 2.8-fold increase in the risk of life threatening cardiac events, whereas among LQT2 patients 100 msec increments in QTc(b), QTc(f) and QTc(fram) were associated with a similar (2.1–2.2 fold) increase in the risk of ACA or SCD (Table 2A).
When the RR interval was included as an additional covariate in each of the 4 models, the effects of heart rate were shown to be significantly different between LQT1 and LQT2 patients (Table 2B).
In LQT1 patients, each of the 4 repolarization measures was demonstrated to be a more powerful (>3-fold risk increase) and significant predictor of outcome in the models that included the RR interval compared with the respective models that did not include heart rate as a covariate. Multivariate analysis in LQT1 patients showed that 100 msec increments in the absolute QT interval were associated with a 3.3-fold increase in the risk of ACA or SCD (p=0.020), and 100 msec decrements in the RR interval were associated with a further 1.9-fold increase in the risk (p=0.007; Table 2B). Further adjustment for heart rate improves risk stratification for LQT1 patients, compared with currently used QT correction formulae.
In LQT2 patients, the risk associated with each repolarization measure was similar before and after adjustment for heart rate, and the RR interval was not a significant predictor of outcome in any of the models (Table 2B). Further adjustment for heart rate does not improve risk stratification for LQT2 patients, compared with currently used heart-rate corrected QT formulae.
Similar results were found when we standardized all QTc measures to have the same unit (standard deviation of each QT corrected formula) before running the Cox models as reported in the supplementary appendix (Tables 1S and 2S).
A significant heart rate x genotype interaction was shown in the adjusted model that included the absolute QT and RR intervals (Table 2B). This finding indicates that the heart rate associated risk is significantly greater in LQT1 patients (HR = 1.88) as compared with LQT2 patients (HR = 1.11; p-value for heart rate x LQT1/LQT2 interaction = 0.036).
The ratio of the estimated coefficients on log(QT) and log(RR) in the Cox models were different for LQT1 vs. LQT2 patients (Table 3). In LQT1 patients, the ratio of the estimated coefficients was about 0.8, significantly different from the ratio implied by the correction of either Bazett (0.5) or Fridericia (0.33), neither of which sufficiently corrects for RR. In LQT2 patients, the ratio of the estimated coefficients on log(RR) and log(QT) was about 0.2 (Table 3). This was not statistically significantly different from the ratio implied by either Bazett’s (0.5) or Fridericia’s formula (0.33), though it is obviously closer to Fridericia than to Bazett. Thus, although we have insufficient evidence to reject any of the currently used QTc formulae for LQT2 patients, Fridericia’s (and the Framingham) correction appears slightly superior to Bazett’s, and the data hints that all might slightly over-correct for heart rate among LQT2 patients.
The differences in the c-indices between models based on various QTc measures are relatively small in absolute terms, but QTc(new) is the best measure overall; wstatistically significant (Supplementary data, Table 3s). The mean (SD) values of QT and the various versions of QTc, stratified by whether or not ACA/SCD was observed, are presented in Table 4. QTc(new) appears better separated as compared with other correction formulae. Among potential dichotomizations of QTc(new), a threshold of 550ms resulted in the best model fit (the highest c-index) for predicting life threatening events.
Several important clinical implications emerge from the current study of LQTS patients: 1) resting heart rate is an independent predictor of life-threatening cardiac events in LQT1, but not LQT2 patients, with a statistically significant difference in heart rate-related risk between genotypes; 2) current formulae that correct the QT interval duration for heart rate have important limitations for risk assessment in LQT1 patients; and 3) risk stratification for life-threatening cardiac events in LQTS patients may be improved by incorporating a genotype-specific correction of the QT interval for heart rate.
The identification that genes responsible for LQTS encode different ion channels involved in the control of repolarization has led to important advancements in current knowledge regarding genotype-phenotype correlation in hereditary arrhythmogenic disorders. Data from the International LQTS Registry have shown that life-threatening arrhythmias in LQTS patients occur under specific circumstances and in a gene-specific manner.11 It has been shown that, despite the fact that LQT1 and LQT2 patients harbor mutations that affect the potassium channels, triggers for life-threatening cardiac events among carriers of these 2 genotypes are distinctly different. Patients with the LQT1 genotype have been consistently reported to have a high frequency of arrhythmic events during activities that are associated with increased sympathetic activity and faster heart rates, such as vigorous exercise and swimming, whereas patients with the LQT2 genotype have been shown to be at a relatively low risk during exercise.11, 20–22 The difference in heart rate-related risk between genotypes may reflect that different potassium currents are involved in these two LQTS genotypes. IKs shortens ventricular repolarization with fast heart rates and catecholamines; in LQT1, there is a reduction in this compensatory response. Fast heart rates have been shown to lead to accumulation of IKs, even in the absence adrenergic stimulation.13 Thus, impaired IKs accumulation in LQT1 patients, as a result of malfunction of channels containing a mutant subunit, may contribute to reduced adaptation of action potential duration to increasing heart rates. This concept is supported by an experimental model for LQT1, in which IKs blockade greatly enhances the probability of torsade de pointes in the presence of catecholamines, 23 and by the fact that LQT1 patients have an exaggerated QTc interval prolongation during physical activity.24 Our findings are consistent with previous observations regarding the effect of impaired IKs function on the relationship between heart rate and the duration of ventricular repolarization in LQT1 patients. However, in contrast to previous studies that evaluated the QT – RR relationship during exercise testing, 24–25 we have shown that increasing resting heart rate is an independent risk factor for life-threatening cardiac events in LQT1 patients. In addition, Schwartz et al.12 and Brink et al.14 have previously reported that resting heart rate plays a significant modulating role on the risk for cardiac events among carriers of a specific KCNQ1 mutation (A341V); the present study extends these findings to a wide range of KCNQ1 mutation carriers.
Importantly, although the Bazett’s correction formula was identified in the present study as the best predictor of life-threatening events among known formulae for LQT1 patients, our data suggest that currently available QT correction for heart rate do not provide sufficient risk-assessment among carriers of the LQT1 genotype. We have shown that risk-assessment among carriers of this genotype can be improved by incorporating a genotype-specific QT correction formula that assumes a more linear relationship between the duration of ventricular repolarization and heart rate than that provided by QTc(b) or QTc(f). Our data demonstrate a significant heart rate x genotype interaction effect, indicating that the risk associated with resting heart rate is significantly different between LQT1 and LQT2 patients. Thus, in LQT2 patients, in whom IKs function is not affected, a QTc exponential formula of QT/RR0.2 appears to describe accurately the relationship between heart rate and ventricular repolarization, whereas in LQT1 patients risk stratification is improved by using a much stronger QT/RR0.8 correction formula. All models in the present study were adjusted for baseline and time-dependent risk factors and showed that 100msec increase in QTc [new] was associated with 2.9-fold (p=0.004) and 2.1-fold (p=0.013) increased risk for ACA/SCD among LQT1 an LQT2 patients respectively, in addition to the effects of gender, time dependent syncope, and beta blockers.
When a baseline ECG is assessed in a non-genotyped LQTS patient, we suggest using Bazett’s correction formula; however, an important implication for a patient who has been diagnosed as a carrier of the LQT1 genotype would be that the RR interval provides incremental prognostic information to mere assessment of QTc by Bazett’s correction formula. Thus, we recommend using the improved QT correction formulae when the specific genotype is known. It should be noted that any given value of QTc means something different for each correction formula as the scales are not identical. When using the new correction formulae, a threshold of QTc (new) of 550 ms was associated with the best model fit for predicting life threatening events.
In the present study we evaluated risk factors for LQTS-related life-threatening cardiac events during adolescence, a time-period that has been shown to be associated with the highest event rate in patients with this genetic disorder 7 and when heart rates are usually faster than 60 bpm. However, the phenotypic expression of LQTS has been shown to be age-dependent.7–8 Thus, it is possible that the risk associated with heart rate may be different in older LQTS patients. Further studies are needed to validate the current findings in different age-groups of affected LQTS patients.
Due to sample size limitations, we did not exclude patients treated with β-blockers; nevertheless, all multivariate models included adjustment for time-dependent β-blocker therapy. In a secondary analysis excluding patients treated with β-blockers, we observed consistent results regarding the QT-RR relationship and the suggested improved QT correction formulae among LQT1 and LQT2 patients. Our study is underpowered to assess the yield of the new formulae and event rates within fixed time-periods by dichotomizing survival time and is also underpowered to calculate the sensitivity and specificity of the new formulae.
The current study population, despite being the largest genotyped population to assess QT/RR relation to arrhythmic risk, is not sufficient for validation of specific correction formulae without validations in independent populations. The present study is an overall risk assessment, whereas in clinical practice risk assessment should be individualized since some lethal events can occur in LQT1 patients in the absence of rapid heart rate. QTc alone should not be viewed as a complete diagnostic test, as there are several other important and known risk factors.
Investigations of clinical aspects and basic causal mechanisms of the LQTS have provided novel and important insight into the fundamental nature of the electrical activity of the human heart and to the relationship between disturbances in ion flow and cardiac disease. Our findings suggest that an understanding of the genotype-phenotype relationship in this genetic disorder can lead to improved criteria for risk stratification for life-threatening arrhythmic events in affected patients.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest disclosure: This work was supported in part by research grants HL-33843 and HL-51618 to the University of Rochester Medical Center from the National Institutes of Health, Bethesda, Maryland. Dr. Moss reports receiving a research grant from GeneDx; Dr. Kaufman research grant from CardioDx and St. Jude Medical. Dr. Ackerman has a consulting relationship and license agreement/royalty arrangement with PGxHealth and received consultant fees from Medtronic, Biotronik, Boston Scientific, and St Jude Medical.
This research was carried out while Dr. Alon Barsheshet was a Mirowski-Moss Career Development Awardee at the University of Rochester Medical Center, Rochester, NY.