OUP user menu

Beat-to-beat T-wave amplitude variability in the long QT syndrome

Fabrice Extramiana, Charif Tatar, Pierre Maison-Blanche, Isabelle Denjoy, Anne Messali, Patrick Dejode, Frank Iserin, Antoine Leenhardt
DOI: http://dx.doi.org/10.1093/europace/euq137 1302-1307 First published online: 14 May 2010

Abstract

Aims Long QT syndrome (LQTS) is a primary electrical disease characterized by QT prolongation and increased repolarization dispersion leading to T-wave amplitude beat-to-beat changes. We aimed to quantify beat-to-beat T-wave amplitude variability from ambulatory Holter recordings in genotyped LQTS patients.

Methods and results Seventy genotyped LQTS patients (mean age 23 ± 15 years, 42 males, 50% LQT1, 39% LQT2, and 11% LQT3) and 70 normal matched control subjects underwent a 24-h digital Holter recording. Using the Tvar software (Ela Medical, Sorin group), the beat-to-beat variance of the T-wave amplitude (TAV in µV2) was assessed on 50-ms consecutive clusters during three 1-h periods: one with around average diurnal heart rate (Day Fast), one nocturnal period (Night), and one diurnal period with around average nocturnal heart rate (Day Slow). TAV was increased in LQTS patients during the two diurnal periods but not at night (during the Day Fast period, mean TAV was 34 ± 20 µV2 in LQTS patients vs. 27 ± 10 µV2 in controls, P < 0.05). This effect depended on the genotype. In LQT1, TAV was larger when compared with controls for both Day Fast and Slow periods, but in LQT2 only Day Fast shows higher TAV. Oppositely, in LQT3 the TAV was higher than in the control group during the Day slow period (mean TAV = 34 ± 20 vs. 25 ± 8 µV2 in controls, P < 0.05).

Conclusion In genotyped LQTS patients beat-to-beat T-wave amplitude variability was increased when compared with control subjects. That pattern was modulated by circadian influences in a gene-dependent manner.

  • Long QT syndrome
  • Holter ECG
  • T-wave variability

Introduction

The congenital long QT syndrome (LQTS) is characterized by prolongation of ventricular repolarization duration, susceptibility to Torsades de Pointes (TdP) and a risk of sudden death.1,2 Although genetic testing and expression studies have improved the diagnosis and the understanding, the clinical diagnosis of the syndrome is not always easy since identification of a prolonged QT interval is difficult for non-expert cardiologists but also because some patients with positive genetic test may have a normal QTc interval duration.1,3,4 The stratification for the risk of sudden death in affected patients is even more difficult.

The degree of QTc prolongation has been recognized as a powerful predictor of risk. Current risk stratification strategies include genetic evaluation, gender, and QTc duration dichotomized above or below 500 ms.5 Yet, syncope or even cardiac arrest can occur in ∼5% of LQTS-family members who have a normal QTc interval,1 thus underlining the need for other repolarization phenotypes for the identification of patients at risk of ventricular tachyarrhythmias.

LQTS is a primary repolarization disease and experimental models show that ventricular repolarization prolongation is arrhythmogenic only when associated with an increased dispersion.6 Experimental models of LQTS also show that the increase in spatial ventricular repolarization heterogeneity serves as the mechanism for T-wave alternans.7 Macroscopic T-wave alternans has been long observed in LQTS patients, in particular before the onset of TdP.810

T-wave alternans represents a particular form of T-wave variability displaying a 2/1 pattern. In the study by Zareba et al.8 this pattern was not predictive of arrhythmic events during follow-up. Other patterns of T-wave variability have been described. Variability in QT duration is increased in dilated cardiomyopathy and in the non-type 1 LQTS.11,12

The aim of our study was to quantify beat-to-beat T-wave amplitude variability from ambulatory Holter recordings in genotyped LQTS patients.

Methods

Patients

Data were extracted from Lariboisière database which includes Holter recordings (n = 1032), genetic testing when available (including mutation), demographic data, LQTS related symptoms (i.e. syncope, documented TdP, sudden cardiac death), and treatment (specifically beta-blocker treatment). For the present study only digital recordings of genotyped LQT1, LQT2, and LQT3 patients were used.

LQTS cases were matched 1/1 for gender and age with control subjects. For the adult control population Holters recorded in healthy volunteers during intensive QT studies were used. To match the paediatric LQTS population, we used Holter recordings of healthy children with atypical symptoms, without underlying cardiac disease and with a normal resting ECG.

Holter ECG recordings and analysis

Digital 24-h ECG recordings (Ela Medical, Sorin group, Le Plessis Robinson, France) were sampled at 200 Hz with a resolution of 2.5 µV. ECG recordings were edited to ensure that all cardiac beats of sinus origin were accurately identified and that non-sinus beats as well as artefacts had been excluded for quantitative analysis.

QT duration was measured from Holter tracings at RR = 1000 ms (n = 42) or at the RR interval closest to 1000 ms and then corrected using the Bazett's formula (n = 28).

In order to evaluate potential circadian influences, three 1-h time periods were subsequently selected based on the hourly heart rate table: (i) period representative of diurnal heart rate (Day, Fast), (ii) period representative of nocturnal heart rate (Night), and (iii) diurnal period with slow heart rate (Day, Slow).

ECG and corresponding beat annotations were then processed by the T-wave analysis software developed by Ela Medical, Sorin group.13

Briefly, for each of the 1-h period, the ECG stream was split into clusters made of 60 QRS-T consecutive complexes. Three different filters were applied and clusters were excluded from analysis in case of (i) presence of ventricular or atrial premature contraction, (ii) high RR interval variability defined as RR changes larger than 20% above or below the mean cluster RR interval, and (iii) noise level exceeding 10 µV.

The variance of the T-wave amplitude (TAV in µV2), defined as the average of squared deviations from the mean, was assessed on the 60 consecutive sinus beats within a given cluster. TAV was computed on eight consecutive 50 ms T-wave segments (TAV1–TAV8) following QRS offset (defined as QRS onset +120 ms; Figure 1).

Figure 1

The variance of the T-wave amplitude (TAV in µV2) was assessed on the 60 consecutive sinus beats within a given cluster. Variance of the T-wave amplitude was computed on eight consecutive 50 ms T-wave segments (TAV1–TAV 8) following QRS offset (defined as QRS onset +120 ms). Then mean and maximum TAV was calculated.

All 60-beat clusters within a single 1-h period were tabulated and all suitable (i.e. not filtered as defined above) 60-beat based TAV data were averaged.

Mean TAV was defined as the average TAV from T-wave segments 1 to 8 and Max TAV as the maximum TAV from time windows 1 to 8.

Statistical analysis

Data are presented as mean ± SD. Differences between groups were evaluated using ANOVA with a Fisher's post test when applicable. Circadian influences were assessed using ANOVA for repeated measures. In order to look for technical (noise level) and clinical (age, heart rate, QTc duration) confounding factors for TAV, single and multiple regression analyses were performed on the global population (i.e. LQTS patients and control subjects, n = 140). In all cases a P-value <0.05 was considered as significant. Statistical analysis was performed using Statview 5.0 (SAS Institute, Inc, NC, USA).

Results

Study population

Among the 1032 Holter recordings included in the LQTS database 480 patients had a positive genotype for LQTS 1, 2, or 3. Among these 480 recordings, 111 were in digital format, obtained from 70 different genotyped LQTS patients. These 70 LQTS patients were matched for age and gender with 70 control subjects. Table 1 summarizes demographic and clinical data in both groups.

View this table:
Table 1

Demographic and clinical data in LQTS patients and control subjects

Control subjects (n = 70)LQTS patients (n = 70)P-value
Age (years)21.5 ± 1123.0 ± 150.47
Sex ratio M/F42/2842/28
LQTS type
 LQT135 (50%)
 LQT227 (39%)
 LQT38 (11%)
Symptoms025 (36%)
Beta-blockers045 (64%)
Heart rate (Day Fast)78 ± 1671 ± 12<0.01
Heart rate (Day Slow)70 ± 1563 ± 13<0.01
Heart rate (Night)65 ± 1261 ± 11<0.05
QTc day399 ± 18470 ± 46<0.0001
QTc night409 ± 23491 ± 52<0.0001

Among the 70 LQTS patients, 25 were symptomatic patients (22 with syncope and 3 with a history of aborted sudden cardiac death). Mean diurnal QTc duration was 470 ± 46 ms in LQTS patients and 399 ± 18 ms in control subjects (P < 0.0001). Most of LQTS patients (64%) were on beta-blockers at the time of Holter recording (Table 1). LQTS patients off beta-blockers were those referred to our clinic before any treatment initiation. Controls were all off-drug.

Determinants of TAV variability

The mean TAV displayed circadian influences with higher TAV values during the fast diurnal period (31 ± 16 µV2) followed by the slow diurnal period (27 ± 11 µV2), the nocturnal TAV being the smallest (25 ± 14 µV2), paired ANOVA P < 0.001.

Using single regression, mean TAV was weakly but significantly correlated with age and ECG noise but not with the average heart rate of the corresponding cluster. No correlation between TAV and QTc was evidenced (Table 2).

View this table:
Table 2

Determinants of T-wave amplitude beat-to-beat variability

Single regressionMultiple linear regression
αR2P
Age1.4*10−40.02<0.01–1.9*10−4
Mean heart rate7.*10−50.0030.15–1.7*10−4
QTc−8.8*10−60.0020.53
Noise level0.0220.05<0.012*10−3

Using multiple linear regressions, TAV was negatively correlated to age and heart rate and positively to ECG noise, multivariate R2 = 0.08, P < 0.0001 (Table 2).

LQTS patients vs. control subjects

Mean and maximal TAV were increased in LQTS compared with control subjects during the day but not at night (Table 3). In control subjects, the highest TAV values were observed in the central part of the T-wave (TAV 3–5, 150–250 ms), whereas in LQTS patients highest TAV values occurred later (TAV 5–7, 250–350 ms; Figure 2).

View this table:
Table 3

T-wave amplitude variability in LQTS patients and control subjects

TAV µV2Control subjects (n = 70)LQTS patients (n = 70)P-value
Day FastNoise4.97 ± 1.934.61 ± 1.870.32
Mean TAV27 ± 1034 ± 20<0.05
Max TAV37 ± 1846 ± 25<0.05
Day SlowNoise4.53 ± 1.842.93 ± 1.70<0.0001
Mean TAV25 ± 830 ± 180.06
Max TAV35 ± 1444 ± 25<0.05
NightNoise3.32 ± 1.352.49 ± 1.58<0.01
Mean TAV26 ± 924 ± 120.47
Max TAV35 ± 1435 ± 180.81
Figure 2

Variance of the T-wave amplitude value for each of the eight T-wave segments (TAV1–8) in control subjects (blue bars) and LQTS patients (red bars) during the fast day period (upper panel), the slow day period (mid panel), and at night (lower panel).

Symptomatic status was not associated with significantly different TAV when compared with asymptomatic LQTS patients.

In the LQTS group, the TAV values did not show statistically significant differences between those taking and those not taking beta-blockers.

The average noise was similar in control subjects and LQTS patients during the fast diurnal period but was significantly lower in LQTS patients during the slow diurnal and nocturnal periods (Table 3).

Genotype influences

The diurnal TAV values were dependent on LQTS genotypes (Table 4). In LQT1, TAV was larger when compared with controls for both Day Fast and Slow periods, but in LQT2 only Day Fast shows higher TAV. Oppositely, in LQT3 the TAV was higher than in the control group during the Day slow period (mean TAV = 34 ± 20 vs. 25 ± 8 µV2 in controls, P < 0.05).

View this table:
Table 4

T-wave amplitude variability in control subjects and in the different LQTS subtypes

TAV µV2Control subjects (n = 70)LQT1 (n = 35)LQT2 (n = 27)LQT3 (n = 8)ANOVA P
Day FastHeart Rate78 ± 1669 ± 12*72 ± 1375 ± 9<0.05
Noise4.97 ± 1.934.70 ± 1.844.74 ± 2.053.53 ± 1.210.43
TAV332 ± 1432 ± 1337 ± 2923 ± 70.32
TAV433 ± 1439 ± 1640 ± 3025 ± 80.11
TAV527 ± 1043 ± 21*43 ± 31*29 ± 6<0.001
TAV625 ± 1141 ± 19*40 ± 32*32 ± 12<0.001
TAV726 ± 1334 ± 15*38 ± 32*29 ± 10<0.05
TAV827 ± 1032 ± 1438 ± 3225 ± 80.08
Mean TAV27 ± 1034 ± 1337 ± 30*25 ± 8<0.05
Max TAV37 ± 1848 ± 2146 ± 3234 ± 130.08
Day slowHeart Rate70 ± 1563 ± 14*63 ± 12*64 ± 13<0.05
Noise4.53 ± 1.843.04 ± 1.862.80 ± 1.602.73 ± 1.05<0.0001
TAV329 ± 1230 ± 2420 ± 828 ± 200.11
TAV432 ± 1335 ± 2324 ± 1030 ± 210.08
TAV528 ± 1140 ± 26*28 ± 1337 ± 26<0.01
TAV623 ± 842 ± 24*31 ± 1639 ± 16*<0.001
TAV723 ± 834 ± 22*28 ± 1443 ± 22*<0.001
TAV824 ± 1031 ± 22*27 ± 1938 ± 18*<0.05
Mean TAV25 ± 833 ± 21*24 ± 1134 ± 20<0.05
Max TAV35 ± 1450 ± 28*34 ± 1948 ± 21<0.01
NightHeart Rate65 ± 1260 ± 1161 ± 1165 ± 70.12
Noise3.32 ± 1.652.66 ± 1.792.17 ± 1.312.69 ± 1.16<0.05
TAV328 ± 1124 ± 1318 ± 9*17 ± 6*<0.01
TAV431 ± 1328 ± 1421 ± 11*18 ± 6*<0.01
TAV530 ± 1232 ± 1524 ± 11*20 ± 6<0.05
TAV626 ± 1135 ± 19*26 ± 1124 ± 4<0.01
TAV725 ± 1131 ± 1824 ± 1128 ± 50.10
TAV825 ± 1129 ± 1723 ± 1530 ± 80.28
Mean TAV26 ± 927 ± 1421 ± 921 ± 40.09
Max TAV35 ± 1440 ± 2030 ± 1632 ± 60.10
  • Post test P < 0.05 * vs. contrôle, vs. LQT1, vs. LQT2.

  • TAV1 and 2 have been omitted because they were not different between groups whatever the period considered.

At night, the TAV values in LQT1 patients were similar to those of control subjects whereas LQT2 and three patients showed a significant decrease in TAV when compared with control subjects (Table 4).

Discussion

Using digital Holter ECG recordings, we found that beat-to-beat T-wave amplitude variability is increased in LQTS patients when compared with age and gender-matched control subjects. Furthermore, we show that this T-wave pattern shows both circadian and genotype influences.

Determinants of beat-to-beat T-wave amplitude variability

ECG technology has profound clinical implications. In two previous studies, TAV data based on short (10 or 20 min) recordings with a XYZ lead configuration and 1000 Hz sampling rate have been reported.13,14 Using the same software, we analysed 24-h recordings with a two or three non-orthogonal lead configuration and a 200 Hz sampling rate. Despite these technical differences, our TAV values in control subjects (around 25 µV2) are similar.14 Ambulatory recordings provide large RR interval ranges as well as different autonomic environments but are theoretically associated with higher noise than resting ECGs. However, we could analyse Holter data with noise levels similar to those reported during resting conditions.14

Even after selection of low-noise ECG segments, we could still evidence a significant correlation between T-amplitude variability and noise. It is unlikely that the observed mean differences in T-wave amplitude variability between our study groups can be explained by differences in mean noise level, since the later was higher in the control group. Nevertheless, the level of signal to noise ratio in studies aiming to evaluate T-wave amplitude variability from Holter recordings should be systematically reported.

Evaluating the influence of physiological factors on T-amplitude variability, we found that age and heart rate are both negatively and independently correlated with T-wave amplitude variability. The same trend has been previously described in LQT3 patients.15 Oppositely, experimental studies have shown the importance of fast pacing rate to unmask action potential duration (APD) variability at the cellular level.16,17 In our study, the mean heart rate was below 80 beats per minute, i.e. an heart rate threshold usually not associated with alternating APD.17 In resting conditions, it is conceivable that T-wave amplitude variability may be more influenced by preceding RR interval duration variability than by the consequence of restitution properties at faster heart rates. One can speculate that at cardiac frequency below the APD alternating threshold the T-wave variability is also dependent on RR interval changes. Might this be true, the reduced RR variability inherent to fast heart rates could lead to a decrease in T-wave amplitude variability. This hypothesis could also be applied to explain the decreased T-wave variability with increasing age. It is well documented that heart rate variability tends to decrease with aging.18

Our data also show that the level of T-wave variability is dependent on the circadian period considered. More specifically, periods with a supposedly higher sympathetic tone are characterized by higher T-amplitude variability. This circadian effect could be the consequence of direct adrenergic influences on ventricular restitution properties.19

In summary, our data underline that the determinants of beat-to-beat amplitude variability are complex and warrant further experimental and clinical validation.

Beat-to-beat T-wave amplitude variability in the long QT syndrome

LQTS is a primary repolarization disease in which arrhythmias are related to increased spatial ventricular repolarization heterogeneities,6 a condition leading to unique T-wave amplitude fluctuations such as T-wave alternans.7,20 Macroscopic T-wave alternans may be observed in LQTS patients before the onset of TdP.810 It was thus tempting to hypothesize that LQTS patients may present augmented beat-to-beat fluctuations in T-wave amplitude.

Our results show that the LQTS syndrome is associated with an increased beat-to-beat T-wave amplitude variability on the body surface ECG. This increased T-wave variability was not found to be different in symptomatic and asymptomatic LQTS patients. It is noteworthy that a similar finding has been also described for visible T-wave alternans on resting ECGs8 as well as in a recent study using microvolt T-wave alternans determined during a stress test.21 Thus, the prognostic yield of such parameters in risk stratification of LQTS patients remains questionable. A possible explanation could be that ventricular repolarization heterogeneities are not critically increased in periods without arrhythmic event.

The ECG differences between the three LQTS genotypes evaluated in the present study appear complex and related to the circadian period considered. The reason for such gene-dependent T-wave variability pattern with different autonomic environment is unclear. The relationship between LQTS type and autonomic conditions at time of symptom occurrence has been well documented.22 Whether the different autonomic effects on T-wave variability in the different LQTS types reflect real differences in arrhythmogenic substrate remain to be determined.

Study limitations

Given the sample size, in particular for LQT3 patients, our case–control study results should be interpreted with caution.

The majority of our LQTS patients were on beta-blocker therapy during the Holter recording. This might have biased the results but it would have been ethically difficult to withdraw, event transiently, the treatment to obtain off-drug data.

In the present study, we choose to excluded from analysis clusters including premature contractions or with high (>20%) RR interval variability. We thus intended to normalize the clusters to avoid comparing different conditions. However, by selecting only stable RR environment, we may have evaluated the segments of recordings with the lower T-wave amplitude variability. It has been shown that QT duration may fluctuate dramatically after fluctuations in RR intervals in LQTS patients.23 It might be interesting to evaluate T-wave amplitude variability after abrupt changes in RR interval duration.

To address important issues in clinical management of LQTS patients such as diagnosis value of T-wave variability in particular in patients with borderline QTc durations, larger Holter databases have to be built up. These databases would also allow investigating its prognostic yield in asymptomatic LQTS patients. This study is descriptive and mainly hypothesis-generating.

Conclusions

T-wave amplitude variability quantification can be assessed using Holter recordings during daily activities. In genotyped LQTS patients we could demonstrate that malfunctioning repolarizing channels are associated with an increased T-wave amplitude variability and that this pattern is modulated by circadian influences in a gene-dependent manner. Further studies are warranted to determine the potential diagnostic and prognostic values of this ventricular repolarization index.

Conflict of interest: none declared.

References

View Abstract