|Home | About | Journals | Submit | Contact Us | Français|
The patients with the long QT syndrome type-1 (LQT-1) have an impaired adaptation of the QT interval to heart rate changes. Yet, the description of the dynamic QT/RR coupling in genotyped LQT-1 has never been thoroughly investigated.
We propose a method to model the dynamic QT/RR coupling by defining a transfer function characterizing the relationship between a QT interval and its previous RR intervals measured from ambulatory Holter recordings. Three parameters are used to characterize the QT/RR coupling: a fast gain (GainF), a slow gain (GainL), and a time constant (τ). We investigated the values of these parameters across genders, and in genotyped LQT-1 patients with normal QTc interval duration (QTc<470 ms).
The QT/RR dynamic profiles are significantly different between LQT-1 patients (97) and controls (154): LQT-1 have longer QTc interval (453±35 vs. 384±26 ms, p<0.0001), and an increased dependency of the QT interval to previous RR changes revealed by a larger GainL (0.22±0.06 vs. 0.18±0.07, p<0.0001) and GainF (0.05±0.02 vs. 0.03±0.01, p<0.0001). Importantly, LQT-1 patients have a faster QT dynamic response to previous RR changes described by τ: 122±44 vs. 172±92 beats (p<0.0001). This faster QT dynamic response of the QT-RR dynamic coupling remained in LQT-1 patients with QTc in a normal range (<430 ms).
The measurement of QT/RR dynamic coupling could be used in patients suspected to carry a concealed form of the LQT-1 syndrome, or to provide insights into the types of arrhythmogenic triggers a patient may be prone to.
The combination of clinical research, genetics and molecular biology in the study of the long QT syndrome (LQTS) has played a crucial role in our understanding of the disease. So far, the syndrome has been associated with mutations from 13 different genes. Yet, the mutations in the KCNH2 (LQT-2) and KvLQT-1 (LQT-1) genes represent more than 90 % of the clinical cases. The KCNH2 and KCNQ1 mutations are associated with different ion channel dysfunctions,1,2 and gene specific triggers of arrhythmias.3 In LQT-1, it is recognized that exercise (catecholamine) is the primary trigger because of the malfunctioning IKs channels that leads to less effective shortening of the QT intervals during tachycardia than in normal individuals. While in the LQT-2 patients, emotion/auditory stimulation have been widely reported as primary triggers.3 These observations support the importance of the dynamic response of ventricular repolarization to changes in heart rate 4 and the interest of studying the dynamic features of the QT/RR coupling from the surface ECG in patients with the LQTS.
Because close to 40% of patients with the long QT syndrome can have a non-diagnostic QTc at rest, several investigators have designed tests to reveal the presence of concealed LQTS by provoking QT prolongation through heart rate increase.5,6,7 A few studies have strived to measure the QT/RR coupling using static linear or nonlinear model.8,9,10 But, there is a lack of information pertaining to the characterization of the dynamic relationship between QT and immediately preceding RR intervals in LQTS patients. Consequently, we evaluate a novel technique to quantify the QT interval response to changes in RR values from Holter ECG recorded in LQTS patients. In this study, we hypothesize that this technique could identify different QT/RR dynamic coupling between LQT-1 patients and controls, and these profiles could help identifying the genotyped LQT-1 patients who present near normal QTc interval duration.
The study relies on a set of ECG files and clinical information shared through the Telemetric and Holter ECG Warehouse (THEW: www-thew-project.org) 11 hosted by the University of Rochester Medical Center (NY, USA). We analyze the information from a group of healthy individuals (N=203) and a cohort of LQTS patients (N=480). The first group is used as controls. Clinical information files in ASCII format, and ECG files in ISHNE format, were downloaded from the THEW repository. This database includes genetic testing (including mutation), demographic data, and treatment (specifically the presence of beta-blocker treatment at the time of the recording).12 The two databases used in this analysis were the Healthy Individuals (E-HOL-03-203-003) and the congenital LQTS (E-HOL-03-0480-013) database. This latter one was donated by the Hospital Lariboisière (Paris, France) to the THEW initiative.8,9
These recordings were recorded at a sampling frequency of 200 Hz. Two ECG epochs of 30 min. in length were extracted manually from the diurnal part of the ECG files in the Holter recordings from healthy and LQTS patients. The criteria used to extract epochs from the Holter ECGs ensure that the ECG epochs include sufficient variation in heart rate while being from interval in which QT and RR interval can be measured accurately: 1) slow change of RR amplitude >150 ms; and 2) good ECG signal quality that was visually assessed by rejecting the epochs with low-amplitude T-wave and/or signal loss and/or signal saturation. Any epoch with more than 20% of the rejected beats were excluded from the analysis.
The ECG records were up-sampled to 400 Hz to increase the resolution of RR and QT detection. The singular value decomposition (SVD) was used to reduce the number of leads to two leads (first two eigenleads). Then, the two eigenleads were combined in one signal as described in
where ev1 and ev2 are the eigenleads. 13 The S(t) signal was analyzed with our custom-designed software ScopeWin (Institute of Scientific Instruments, Brno, Czech Republic) to obtain a continuous series of RR and QT intervals. The QT interval duration was determined from the onset of the QRS complex to the end of the T wave, defined as the crossing between the isoelectric line and the tangent to the descending T wave. A semiautomatic method of QT detection was used. Subsequently, the results of beat detection (RR intervals) were compared with the annotations records from the THEW database. Beats that were not normal (i.e. non sinus beats), or some doubt exist about proper detection of T-wave end, were flagged for manual review. The QT intervals from these beats were extracted from the analysis. All disagreements were visually controlled and corrected.
The parameters characterizing the QT/RR dynamic were obtained from the transfer function model described in prior reports.14,15 This type of model of QT/RR coupling predicts the QT interval based on the history of RR intervals. The TRF model step response, i.e. QT reaction following sudden change of heart rate, fully agrees with known QT step response measured in patients with pacemaker 16 and in experiments based on isolated heart 17.
The TRF method can briefly be described as follow: three QT dynamic parameters (GainL, GainF and τ) are computed. The TRF model defines the next recursive relation between RR and QT intervals such as: RRx=RR-mean(RR), and QTx=QT-mean(QT),
where rrx(i) and qtxm(i) are i-th values of variables (RRx, QTxm) and QTxm is model QT without mean level, i.e. QTm=QTxm+mean(QT). The coefficients a1, b2 and b3 are fitted by minimizing the residuum between QT and QTm. The QT step response is computed from the fitted parameters a1, b2 and b3 in order to provide a more meaningful description of the characteristics of the TRF. The GainL, the GainF, and the τ values are computed from the step response which describes the behavior of the dynamic system (QT/RR dynamic coupling) when its input changes from zero to one in a very short time. The steps response provides information about the time and gain of the system to reach a QT stationary state after either a deceleration or an acceleration of the heart rate. Thus, the proposed model does not separate heart-rate deceleration from heart-rate acceleration; rather it provides a model for the characterization of the overall “QT/RR dynamic coupling”.
We describe the three parameters of the model as follow:
The figure 1 provides an example of the QT, RR and the QT/RR coupling step response curve of the TFR model. We report the heart rate corrected QT (QTc) computed as QTc=mean(QT)+(1000-mean(RR))× GainL The subject specific, dynamic correction is used.
The computation of the coefficients of the transfer function and the coefficient of its step response were done using MATLAB, (Mathworks, Matick, MA). The method is described in more details in 15.
We report the average values of QTc, RR, GainL, GainF, and τ across the study groups. A comparison of ECG variables was performed using t test for independent sample. In case the data was not normally distributed the non-parametric test from Kruskall-Wallis was used instead. The data are given as mean± standard deviation, and differences were accepted as significant for p<0.05.
We scanned the Holter ECGs from the THEW database to identify the recordings with 30-min epochs that were meeting the selection criteria described above. At least one epoch could be found in 97 LQT-1 recordings and in 154 recordings from healthy individuals. An additional 28 records were excluded from healthy individuals and 52 records in LQT-1 subjects because: 1) they did not include enough valid QT intervals due to the low signal quality or/and, 2) the correlation between detected QT and RR values was lower than 0.3.
In the LQT-1 cohort (N=97, age 23±16 years) were 40 men and 57 women, 49 of them were on beta blocker therapy (20 men and 29 women), 48 were off beta blocker (20 men and 28 women). In the healthy cohort, we identified 154 healthy subjects, age: 36±14 yrs, including 73 women (47%).
The average QTc, RR, GainL, GainF, and τ values are reported in Table 1 for the healthy vs. the LQT-1 patients off beta-blocker. Also, we report their values between genders for both the control and the LQT-1 groups excluding the patients on beta-blockers. Finally, we evaluated whether the QT-RR dynamic coupling is different between controls and LQT-1 patients with a concealed form of the syndrome or namely a range of normal QTc interval duration (370<QTc<430 ms). This group includes 98 recordings from healthy individuals (age 36±14 years, 44 men), and 24 ECGs from LQT-1 patients (age 21±16 years, 16 men).
The control group had a significantly weaker (p<0.000001) fast gain (0.03±0.01) than slow gain (0.18±0.07) which is consistent with a normal hysteresis i.e. the long term response to heart rate changes is larger than the immediate response.16 All investigated ECG parameters were statistically different between the LQT-1 and the control groups. The LQT-1 patients had lower heart (RR: 453±35 vs. 384±26 ms, p<0.00001), prolonged QTc interval, and a shorter time adaptation (122±44 vs. 172±92 beats, P<0.00001). This was associated with an increased fast (GainF: 0.03±0.01 vs. 0.05±0.02, p<0.00001) and slow gains (GainL: 0.18±0.07 vs. 0.22±0.06, p<0.00001). Interestingly, the gender differences in GainL (higher in women: 0.19±0.07 vs. 0.17±0.06, p<0.01) and τ values (longer in men: 186±109 vs. 156±65 beats, p<0.05) in healthy were not present in LQT-1 patients. Indeed, LQT-1 women and LQT-1 men presented a similar short time adaptation (119±41 vs. 125±48 beats, p=NS) combined with a strong slow gain (0.23±0.07 vs. 0.21±0.06, p=NS). Finally, the comparison of the characteristics of the QT-RR dynamic coupling for controls and LQT-1 patients with QTc interval in normal ranges, i.e. 370<QTc<430 ms, suggested that the intrinsic QT adaptation is abnormal event in patients with normal QTc interval. Indeed in this subgroup of individuals, the trends toward higher fast gain (0.037±0.01 vs. 0.030±0.01, p=0.02) and shorter time adaptation constant (116±45 vs. 167±67, p<0.001) was still strong.
The analysis of the set of recordings of LQT-1 patient on and off beta-blockers was implemented yet it did not show any difference. We explained this result by the lack of enough RR variation in patients on beta-blockers.
The curves describing the QT interval variation following an abrupt change in heart rate for the healthy, LQT1 and concealed LQT1 and healthy (QTc <370,430> ms) are in figure 2. The mean levels of QT parameters from Table 1 were used at the simulation. It shows that the QT duration during increased heart rate in controls is significantly shorter than n LQT-1 patients, not looking on nearly similar QTc in concealed groups.
We present a computerized method to quantify the QT/RR dynamic coupling by measuring the beat-to-beat QT and RR intervals in a manually-selected set of ECG epochs extracted from standard ambulatory Holter recordings. We evidenced statistically significant differences between the coefficients of the models characterizing the QT/RR dynamic coupling of LQT-1 patients and healthy individuals independently from their QTc interval duration. Our results suggest that the computerized QT and RR analysis and the modeling of the dynamic coupling between the QT and RR intervals could be used as a clinical tool to evidence an impairment of the adaptation of the ventricular repolarization to changes in heart rate. And therefore, it could provide insights into the presence of a concealed form in suspected LQT-1 patients.
Molecular biology combined with clinical observations led to uncover gene-specific abnormalities of ion currents, phenotypic ECG patterns, and multiple arrhythmogenic pathways. The phenotypic expression of the mutations on the surface electrocardiogram (ECG) has been widely reported using multiple quantitative approaches: T wave morphology19,20,21, T wave lability 22,23,24, and QT adaptation to heart rate changes.25 These methods were primarily designed to improve the diagnosis of LQTS, and to elucidate arrhythmogenic mechanisms leading to the frequent life-threatening cardiac events in the LQTS patients.
Clinical observations reported by Tan and colleagues 28 described very different triggering mechanisms between LQT-1 and LQT-2 patients. Specifically, the proximate LQT-1 triggers were delayed after depolarization, whereas in LQT-2 patients pause-dependent torsades de pointes and early after depolarization are most commonly reported. The proposed ECG method could help identifying the susceptibility of patients to specific types of arrhythmic triggers such as adrenergically driven stimuli (exercise and emotion) during which a lack of QT adaptation to fast heart rate acceleration (normally triggering IKs activation) plays a crucial role. We report a GainL (slow gain) significantly larger in LQT-1 than healthy individuals. By definition this parameter is equivalent to the QT/RR slope, the parameter define the amplitude of QT steady state change given by RR change with amplitude 1. Our findings are consistent with previous results31 and with current reports describing a steeper diurnal slope (stronger gain in LQT-1 patients than in controls, [Couderc et al. 2011, Data submitted, Circulation: Arrhythmias and Electrophysiology]). Reconciling these results with the type of triggers in LQT-1 patients is challenging because of the large spectrum of risk factors present in LQT-1 patients and their modulation by mutation location [Costa et al., 2011, data submitted Heart Rhythm Journal]. Yet, it is expected that a prolong QT interval combined with a steep positive relationship between QT and RR intervals leads to large range of QT interval duration and a propensity to arrhythmias. The second interesting factor is described as a time constant (τ), which is the time needed for the QT interval to reach its stationary value. The LQT-1 patients present significantly smaller values than controls which one could translate into a reduced redundancy of repolarizing current within the ventricle, or namely, a limited repolarization reserve 27. A shorter time constant could reflect the impaired ability of the cardiac cells to timely restore their repolarizing capacity.
The role of gender in the LQTS has been thoroughly investigated because clinical phenotypes are known to be strongly modulated by gender and sexual hormones28. In our findings, we detected a significantly stronger slow gain and shorter time constant (τ) in healthy women than in men which is consistent with an increased propensity for arrhythmias in women exposed to drug-induced QT prolongation in comparison to men 29. However, this gender-specific profile of the QT-RR dynamic profile was not present in LQT-1 patients. The increased gain and reduced time constant detected in healthy women are exacerbated in LQT-1 patients for both genders. Interestingly, the men LQT-1 patients present an average reduction of τ equivalent to 61 beats, while in LQT-1 women, this reduction is almost half (37 beats in average). Consequently, one would speculate that LQT-1 men, despite a shorter QTc interval than LQT-1 women, present a more profound relative impairment of QT/RR dynamic coupling than LQT-1 women. As recently reported by Costa et al. [Costa et al., 2011, data submitted Heart Rhythm Journal] from 1,051 genotyped LQT-1 patients, LQT-1 men have higher probability of aborted cardiac arrest and sudden cardiac death than women. Consequently, our next step will be to investigate whether there is an association between a reduction of τ values and the occurrence of events in LQT-1 patients primarily in adult men.
Finally our investigation of the characteristics of the QT/RR dynamic coupling for patients and healthy individuals with QTc between 370 and 430 ms. suggested that the time constant τ is independent from the QTc interval prolongation. Indeed, the reduced value of τ values remains significantly smaller in this subgroup of LQT-1 patients with normal range of QTc interval durations than in the set of recordings from controls with the same range of QT values. Consequently, we believe that the proposed method carries the unique ability to identify patients with increased risk for arrhythmias independently from the presence or not of a prolongation of the QTc interval.
The LQT-1 patients have different diurnal responses of QT interval to rapid changes in heart rate characterized by stepper long term response and a faster adaptation of QT to RR change than healthy individuals. These characteristics where not modulated by gender despite the well described QTc prolongation in women. Such type of analysis could be applied to patients suspected to carry LQT-1 mutations to help the decision around genotyping, prophylactic therapy and susceptibility of the patients to recognized event triggers.
This work was partially supported by grant KONTAKT No. ME09050 of the Ministry of the Education of the Czech Republic, by European Regional Development Fund - Project FNUSA-ICRC No. CZ.1.05/1.1.00/02.0123, by the National Heart, Lung, and Blood Institute of the US Department of Healthy and Human Services grant U24HL096556 and by grant No. P102/12/2034 of the Grant Agency of the Czech Republic.
Conflict of interest: None.