Why did qt dispersion die?

Cardiac Electrophysiology Review 2002;6:295–301
C 2002 Kluwer Academic Publishers. Manufactured in The Netherlands.
Why Did QT Dispersion Die?
Pentti M. Rautaharju
EPICARE Center, Wake Forest University School of Medicine,
Winston-Salem, North Carolina, USA
Abstract.
Background: Considerable controversy ex-
ings. The interest in QTD is still very much alive, ists about the meaning of QT dispersion (QTD). The
and in the decade since its introduction it became working hypothesis of the present paper was that the
the most fashionable topic in the realm of QT in- necessary although not sufficient condition for the va-
vestigations since the long period of domination of lidity of QTD concept is the association of QTD with
QT and QT rate adjustment studies after Bazett nondipolar voltage (NDPV) in T waves of the 12-lead
published his formula in 1920. Therefore, the title Methods and Results: ECGs of 4890 subjects, 966
that was suggested for the present paper is rather with coronary heart disease (CHD) and 3844 considered
provocative and challenging. It is provocative with CHD-free were processed using computer programs for
its connotation that QTD is dead. It is challenging measurement of the ratio of the first two eigenvalues
in demanding proof that the concept is indeed in- (E2/E1), nondipolar voltage (NDPV), terminal T wave
valid. The present paper will first summarize the direction and ECG estimate of left ventricular mass
arguments presented for and against the QTD con- (LVM). The mean NDPV in T wave was 11 µV (SD 3.9),
cept, followed by presentation of some new data with 6 µV (SD 1.3) in terminal 40 ms. NDPV alone ex-
relevant to this intriguing controversy.
plained only 6% and NDPV, E2/E1 and LVM combined
13% of QTD variance. There was a modest increase in
the fraction of subjects with QTD >60 ms among subjects
with NDPV in terminal T > 7 µV compared to those with
Arguments for and against the validity
NDPV ≤ 7 µV (15% vs. 10%). A more profound increase
of QT dispersion concept
was associated with terminal T wave direction deviat-
Most of the QTD publications have enthusiasti- ing from normal (37% vs. 12% among those with nor-
cally supported the concept. There is a definite mal direction), reflecting dipolar rather than nondipo-
publication bias—reviewers and editors of profes- lar components.
sional journals as well as investigators in general Conclusions: The association between QTD and
tend to have the attitude that negative results do NDPV is weak, and QTD is unlikely to represent any
not warrant particularly serious consideration. In meaningful myocardial repolarization event in the in-
the majority of the reports, the support presented terval domain. It seems more logical to use direct mea-
surement of NDPV as a potential marker of localized
comes from circumstantial, indirect evidence as- dispersion and heterogeneity of ventricular repolariza-
sociating QTD with excess risk of adverse events tion for evaluation of the risk of adverse cardiac events.
in a large variety of conditions summarized in anextensive monograph by Malik and Batchvarov [3] Key Words.
QT, QT dispersion, cardiovascular, electro-
and in other review articles [4–6].
cardiography
against the validity of the QTD concept are Background and the Present State
summarized in Table 1. The main argumentpresented against the concept is that morphologic QT dispersion (QTD) was introduced in 1990 by T waveform variations associated with dipolar a British group of scientists [1]. One year follow- components of repolarization can produce large ing the introduction of QTD, the group reportedresults from a clinical trial demonstrating reduc-tion in QTD by sotalol [2]. This finding aroused theinterest in the QTD concept among clinical inves-tigators and electrocardiographers and the pub- The author thanks Mr. James Warren, M.Sc. for his con- lication activity increased steadily, showing char- tributions to the development of the ECG MorphologyProgram, and Mr. Charles Campbell, B.Sc. and Mrs.
acteristics of an epidemic with a relatively long Zhu-Ming Zhang, M.D., for their contributions to var- incubation period. A similar proliferation of com- ious ECG processing and data analysis tasks.
munications was seen in scientific meetings ofprofessional societies such as the annual sci-entific sessions of the American Heart Associa- Address correspondence to: Pentti M. Rautaharju, M.D., Ph.D., tion and the American College of Cardiology. The Suite 505, Piedmont Plaza Two, 2000 West First Street, most recent MEDLARS literature search lists 488 communications with QTD in the subject head- 295
296
Table 1. Arguments against the QT dispersion concept and against equating conceptually QT dispersion with
dispersion of ventricular repolarization
1. Measured QTD determined primarily by dipolar The range of QTD in leads generated from strictly components and do not represent dispersion of dipolar components is of the same order of 2. Interlead differences in measured QT largely Long QT values obtained when terminal T vector is in determined by T wave “loop” morphology and by direction along the axis of the lead vector in a given the projection of terminal T wave vector on lead and short when near 90◦ angle.
3. Abnormal T wave morphology has a strong Abnormal strictly dipolar morphology patterns can cause large interlead differences in QT. Theserepresent variability in amplitude/time domainrather than any physiologically meaningfulintervals related to repolarization events 4. Overall technical variability of QTD measurement General current consensus is that QTD values in is so large that no feasible threshold can be excess of 100 ms can be considered abnormal. Such established to separate normal from abnormal large variations are commonly occurring with QTD. Short- and long-range variability also 5. Variations in lead vector strength, T wave amplitude Most Toffset detection algorithms use fixed thresholds and noise level cause additional QTD variation so that T wave amplitude variations from projectionof strictly dipolar components may cause largevariations in QTD 6. Presence of non-dipolar components in body surface Their presence is a necessary condition for detection of the end of localized ventricular repolarization and localized dispersion from QT measurements variations in QTD (Argument 1 in Table 1).
lead variations of the measured QT in limb leads Lee et al. [7] reported that QTD in 12-lead ECG are due to projection differences, variations in lead generated by transformation from a strictly strength when a constant threshold value is cho- dipolar source was 53 (SD 49) ms and it was not sen for defining the end of the T wave and a variety significantly different from QTD in the original 12-lead ECG (49 (23) ms)). Argument 2 relates to the fact that T wave loop morphology (“flat” T prompted publication of an editorial in 2000 in vector loop) is an important determinant of QTD European Heart Journal titled: “QT dispersion: [8]. A similar T wave morphology descriptor is time for an obituary?” by Malik [13] who con- T wave “complexity” expressed as the ratio of cluded: “Despite the serious difficulties with the the second and the first eigenvalues (E2/E1) [9].
concept and despite the fact that a clear publica- Abnormal T wave morphology certainly increases tion bias exists towards positive findings, the huge QTD. However, dipolar components alone leave a number of studies showing some meaningful re- large part of QTD variability unexplained as will sults with QT dispersion measurements cannot be Argument 3 in Table 1 relates to methodologi- In summary, while the arguments against the cal problems in determination of QTD. The overall validity of QTD have considerably weakened the technical variability, procedural differences and practical utility of QTD, they have not proven that other factors induce so large indeterminacy in es- the concept is invalid, and the voluminous publi- timation of QTD that it has not been possible to es- cations supporting the concept QTD have failed tablish a definition for abnormal QTD as pointed to produce evidence fulfilling the necessary and out by Malik et al. [3]. In contrast, Macfarlane even less the sufficient condition for the validity et al. have concluded that the upper normal limit of the concept. The crucial necessary although not of 50 ms is “highly specific” [10]. It is conceivable sufficient condition for the validity of QTD con- that with improved methodology QTD measure- cept is that nondipolar components in body surface ment may become more meaningful if the concept ECG exist during ventricular repolarization and otherwise can be proven valid. The fact that only that these in turn are of sufficient magnitude to two of the six limb leads have independent signal have a significant association with QTD. Without components implies that only one pair of QT differ- such evidence, the whole QTD concept has been ences can be a valid measure of QTD [11]. Lead-to- denounced by some as “fallacy” [14].
297
The publication by Malik in 2000 was the first were rejected from ECG data files. The 12SL paper documenting that significant amounts of ECG Program (GE Medical Systems Information, nondipolar components indeed exist in the T wave of the standard 12-lead ECGs of normal subjects and of patients with hypertrophic and dilated car- (Novaheart Inc., Winston-Salem, NC) was used diomyopathy and survivors of acute myocardial for quantitative vector analysis of repolarization infarction [15]. However, QTD did not correlate waveform patterns. The program has modules for significantly with nondipolar components except extraction of dipolar components and distributions for borderline correlation in patients with hyper- in 12 preferential spatial directions of the ST-T trophic cardiomyopathy (p = 0.03), and the au- subinterval vectors in a rhombododecahedron ref- thor concluded that QTD is unrelated to nondipo- erence frame [17,18]. In addition, the QT Guard program of the GE Medical Systems Informationwas used for QTD measurement. ECG estimates The Objective of the Present Study
of left ventricular mass were determined usinggender- and race-specific algorithms of the NOVA- The present communication will address the question of the determinants of QTD includingnondipolar components in normal subjects and Data analysis
subjects with coronary heard disease (CHD). The Differences in QTD distributions were first primary objective was to determine the magni- examined in subgroups stratified according to tude of nondipolar components (square root of to- the CHD status, the ratio of the eigenvalues of tal dipolar energy or nondipolar voltage, (NDPV)) the first two principal components (E2/E1) and the in the T wave of the standard 12-lead ECG and to magnitude of the nondipolar components in evaluate if the association of QTD with nondipolar the T waves of the standard 12-lead ECG. The sig- components is sufficiently strong to fulfill the nec- nificance of the differences between group means essary condition for the validity of QTD concept.
was performed using the t-test. Multiple regres-sion models were used to evaluate the association Study groups
of various variables of interest with QTD. In subse- ECG data of 4,890 subjects (60% females, 40% quent analyses, stratification was performed into males) from community-based populations were dichotomized categories according to the magni- selected from the files of the EPICARE Center, a tude of the nondipolar components and spatial dis- central ECG laboratory for population studies and tribution of the terminal T wave in a window from clinical trials. An older adult group was selected (Toffset − 40 ms) to Toffset. A ratio test for performed for the study (64 years and older) in order to ob- to evaluate the significance of the differences in tain an adequately large subgroup of subjects with the proportion of QTD exceeding 60 ms in various clinical evidence of coronary heart disease (CHD).
stratified subgroups. All analyses were performed ECGs with QRS duration ≥ 120 ms and those with using Microsoft Excel Version 5.0 (Microsoft Cor- an electronic pacemaker, atrial fibrillation or flut- poration, Seattle, Wa) and SAS/STAT Version 8.0 ter were excluded from the study. The CHD group of 966 subjects was selected on the basis of con-firmed history of myocardial infarction or anginapectoris or invasive cardiac procedures related to Results
coronary artery disease. The remaining 3844 sub-jects of the study group were considered CHD-free.
Mean values with standard deviations for vari- Silent MI by ECG alone was not used as an exclu- ables of key interest concerning QTD are listed sion criterion. From the total group of 4890 sub- in Table 2. The range of QTD values was wide in jects, a subgroup of 4810 was used for some more both study groups, with the mean value in total detailed analyses with a more complete set of ECG study group 34 ms (23.5 ms). QTD exceeded 75 ms data related to quantitative morphologic T wave in 5% of the subjects and 95 ms in 2%. QTD was analysis, including evaluation of the terminal T 7 ms wider in the CHD group than in the CHD- free group (p ≤ 0.001) and also the eigenvalue ra-tio E2/E1 differed significantly between the two ECG methodology
groups, 14% (13.9) in the CHD-free and 19% (16.5) ECGs were recorded in resting supine state follow- in the CHD group (p < 0.01). The question of ing strictly standardized procedures for ECG ac- key interest in the present context is the mag- quisition, including electrode placement [16]. All nitude of nondipolar components in the standard ECGs received at the EPICARE Center were in- 12-lead ECG in relation to QTD. The NDPV was spected visually to detect technical errors, miss- 11 µV (14.8), exceeding 18 µV in 5% of the sub- ing leads and inadequate quality, and such records jects. Thus, there are nondipolar components of 298
Table 2. Mean values (SD) of total and precordial QT dispersion, nondipolar voltage and eigenvalue ratio in
subjects with coronary heart disease (CHD) and in CHD-free subjects
Fraction (%) with NDPV > 15 µV *p < 0.05, **p < 0.01, ***p < 0001 for one-tailed test for difference between group means or ratios.
†QT adjusted to ventricular rate by linear function of RR.
‡QT dispersion.
&NDPV = nondipolar voltage.
Table 3. Contingency table for the longest and the shortest QT in chest leads*
*Omitted were 80 ECGs with measured QTD = 0 for technical reasons.
Bold-faced figures indicate lead pairs with the longest and shortest QT in two adjacent leads in 1303 of the subjects (27.1%).
sufficient magnitude that could conceivably have ity. The question is whether QT distribution is a significant influence on QTD. The mean NDPV truly dispersed or is there some measure of reg- values differed by 1 µV between the CHD-free and ularity in the distribution among adjacent ECG the CHD groups (p < 0.01). Although the mean leads. The rationale behind this question is the group difference is negligible, the possibility that consideration that irregularity or dispersion of the the group difference will become more pronounced functional refractory period or repolarization time at the higher range of QTD and NDP values has at localized level over relatively short distances to be considered. Table 2 lists also the fraction of enhances propensity for triggered activity and subjects with QTD exceeding 60 ms and NDPV ex- re-entry. Such localized myocardial dispersion ceeding 10 µV and their combination. Ratio tests should produce QTD in adjacent ECG leads pro- indicated that both fractions were significantly vided that nondipolar components are present in larger in the CHD group than in the CHD-free sufficient magnitude in body surface ECGs.
group, as was the fraction with subjects exceeding Joint distribution of the chest leads with the both thresholds (p < 0.001 for all).
longest and the shortest QT are shown in Table 3.
Limb leads were omitted from this table becausethe lead vectors of the limb and chest leads are Are QT intervals really dispersed?
spatially in different planes in image space so that The origin of the word dispersion is dispersere, dipolar projection differences can be expected to to scatter, implying separation and moving apart be larger between these two sets of leads. A close in different directions without order or regular- examination of Table 3 revealed that the longest 299
Table 4. Correlations between variables of key interest
Table 5. Fraction (%) of subjects with QTD >60 ms in
in relation to determinants of QTD subgroups stratified by terminal T wave orientation,coronary heart disease status and nondipolar voltage *NDPV = nondipolar voltage in T wave.
†E2/E1 = ratio of the first two principal components.
and the shortest QT often resided between two ad- jacent chest leads, for instance between V1 and V2 in 675 (14%) and overall between two adjacentleads in 1303 subjects (27%).
*ILA = inferior-left-anterior, AL = anterior-left, L = left in a12-directional duododecahedron reference frame.
†CHD = Coronary heart disease.
Determinants of QTD
‡p < 0.001 for ratio test for group differences.
NDPV as a possible indicator of the dispersion ofventricular repolarization and E2/E1 ratio as anindex of morphologic T wave abnormalities in time 6 µV (1.5) in the CHD group. Thus, the amplitude domain not directly related to any repolarization of the NDP components in the terminal T wave intervals are of key interest in this context as pos- were approximately one half of that in the total sible determinants of QTD. In addition, ECG es- timate of left ventricular mass (LVM) and QRS The fraction of QTD values exceeding 60 ms was duration were also considered because any then evaluated in subgroups stratified by the di- changes in ventricular excitation may cause sec- rection of the terminal T wave (normal vs. abnor- ondary repolarization abnormalities. Correlation mal), the CHD status, and the fraction of NDP matrix in Table 4 shows relatively modest and voltage in this terminal T window exceeding 7 µV equal level of correlation between QT and NDPV (r = 0.25) and E2/E1 ratio (r = 0.27) and a lowercorrelation with LVM (r = 0.12). Multiple regres-sion models (not shown) with NDPV and E2/E1 ratio entered as simultaneous covariates into re- gression on QTD produced R-square value 0.075.
Entering LVM as the third covariate increased R-square value to 0.13. Thus, there three pri- mary determinants of QTD combined explained QTD > 60 ms (%)
just 13% of the total QTD variance, leaving 87% of QTD in Relation to NDP Voltage
and the Direction of the Terminal T Wave
CHD Status No CHD;
T Direction ILA, AL or L
Spatial distribution of the mean terminal T vector CHD Status and Terminal T Direction
determined in a window from (Toffset − 40 ms) toT Fig. 1. The fraction of QTD exceeding 60 ms in
offset was determined in reference to 12 principal spatial directions in a duododecahedron reference subjects with NDPV > 7 µV (filled columns) compared frame [17,18]. In the CHD-free group, the direc- to those with NDPV ≤ 7 µV (open columns) insubgroups stratified by CHD status and direction of the tion of the terminal T wave vector was inferior- terminal T wave vector. Note the pronounced increase left-anterior, anterior-left or left in 95.5% of the in the fraction of abnormal QTD with abnormal subjects, corresponding closely to the spatial dis- terminal T wave direction both in CHD-free and CHD tribution of the normal mean T vector. The mean groups. In comparison, the fraction is significantly value of the NDP voltage in the terminal 40 ms different only in CHD-free subjects with normal window was 6 µV (1.3) in the CHD-free group and 300
(Table 5). Seven µV (approximately mean +1*SD) Overall, the correlations of QTD with NDPV was chosen in order to obtain an adequate num- and with other factor evaluated (E2/E1 and LVM) ber of subjects in various subgroups for statis- were low, and these three strongest determinants tical evaluation. The increase in the fraction of of QTD combined in multiple regression models QTD exceeding 60 ms with two-way stratifica- explained only a small fraction (13%) of the total tion by the NDP voltage, terminal T wave direc- QT variability. A prominent part of this residual tion and CHD status was significant (p < 0.001) QTD variability is likely to be due to morphologic in all of these major subgroups. However, the variations in T wave patterns (other than E2/E1 increments in the fraction of QTD > 60 were con- ratio) associated with dipolar rather than nondipo- siderably larger with stratification by T wave ori- lar sources and with methodological problems as- entation and also by CHD status than that by sociated with QTD measurement. Morphologic T stratification by NDP voltage, and the ratio test wave patterns are best evaluated in amplitude do- for NDPV subgroups was significant (p < 0.05) main and it is unlikely that lead-to-lead variations only in the CHD subgroup with normal terminal at the measured endpoint of the T wave are asso- ciated with any local variations in the endpoint ofmyocardial repolarization. Morphologic variationsin T wave patterns such as those manifested in in- Discussion
creased E2/E1 ratio and abnormal direction of theterminal T vectors are certainly associated with The results from the present study confirm the heterogeneity of ventricular repolarization. These findings by Malik et al. [15], demonstrating that aberrations are likely to reflect regional hetero- nondipolar components exist in the T waves of geneity (changes in transmural or apex-to-base ac- the 12-lead ECG. The mean value of NDPV was tion potential duration gradients). They may well 11 µV (14.8), exceeding 18 µV in 5% of the subjects.
take place with relatively small changes in tempo- Although these values are below visual resolu- ral profiles of action potential duration with small tion limits and smaller than amplitude thresholds or no changes in temporal dispersion of the end of used by many computer algorithms to define the end of the T wave, it is possible that the occurrence It also remains to be proven that the existence of nondipolar components at critical time points in of NDPV is significantly associated with definable certain leads can influence QTD measurements.
repolarization intervals or subintervals at myocar- The proportion of QTD exceeding 60 ms was dial level before the presence of NDPV can be ac- significantly higher (p < 0.001) among subjects cepted as a necessary condition to validate the with NDPV exceeding 7 µV than among subjects QT dispersion concept. Evidence for the sufficient below this threshold, although in subgroup analy- condition has to come from cardiac electrophysi- ses that the proportion was significantly different ological studies. Until such evidence emerges in only in the CHD-free subgroup with normal direc- future studies, low level correlation existing be- tion of the terminal T wave (p < 0.05).
tween NDPV and QTD leaves the validity of the An interesting observation was that the longest and the shortest QT often resided between two ad- It should be noted that correlation between jacent chest leads in 27% of the subjects, one half regional dispersion of ventricular repolarization of them between V1 and V2. The finding of such and QTD in surface leads [20] does not neces- localized dispersion of QT values is difficult to rec- sarily validate QTD concept. Dispersion of ven- oncile if QTD were solely due to projection differ- tricular repolarization may increase QTD due to ences from the dipolar components of the T wave.
morphologic changes of dipolar origin rather than Frequent occurrence between V1 and V2 may be due to nondipolar components. It is imperative to in part due to the relatively large spatial angle demonstrate the association of localized timing of between the lead vectors of these two leads. QTD the end of myocardial repolarization with the end measurement problems, including overlapping U of the T wave in specific localized ECG leads, with and T waves may account, in part for the remain- nondipolar components as the direct link between these two factors as a sufficient condition for the The fraction of subjects with QTD exceeding validity of QTD concept. Until and unless such 60 ms was prominently associated with an abnor- evidence emerges, it is inappropriate to equate mal terminal T wave spatial direction, thus con- QTD with any meaningful myocardial repolariza- firming the findings of Kors et al. [8] about the importance of this T wave feature as a determi- In conclusion, the correlation between QTD nant of QTD. Abnormal terminal T wave direction and NDPV of the T wave is weak and there are and other morphologic T wave changes are in prin- profound methodological difficulties in QTD mea- ciple determined by T wave components of dipolar surement. Direct measurement of nondipolar com- ponents calculated easily by computer programs 301
rather than QTD should be a more logical choice measurement of QT dispersion. Circulation 1998; for evaluation as a potentially valuable although still unproven marker of localized heterogeneity of 11. Kors JA, van Herpen G. Measurement error as a source of QT dispersion: A computerized analysis.
Heart 1998;80:453–458.
12. Bacharov V, Yi G, Guo X, Savelieva I, Camm AJ, Malik M. QT interval and QT dispersion measure- References
ment with the threshold method depend on thresh-old level. Pacing Clin Electrophysiol 1998;21:2372– 1. Day CP, McComb JM, Campbell RW. QT dispersion: An indication of arrhythmia risk in patients with 13. Malik M. QT dispersion: Time for an obituary? long QT intervals. British Heart J 1990;63:342– Europ Heart J 2000;21:955–957 (editorial).
14. Rautaharju PM. QT and dispersion of ventricular 2. Day CP, McComb JM, Matthews J, Campbell RW.
repolarization: The greatest fallacy in electrocar- Reduction in QT dispersion by sotalol following diography in the 1990s. Circulation 1999;99:2477– myocardial infarction. Europ Heart J 1991;12:423– 15. Malik M, Acar B, Gang Y, Yap YG, Hnatkova K, 3. Malik M, Batchvarov V. QT Dispersion. In: Camm Camm AJ. QT dispersion does not represent elec- AJ (ed). Clinical Approaches to Tachyarrhythmias, trocardiographic interlead heterogeneity of ventric- Vol. 12, Armonk, NY: Futura Publishing Company, ular repolarization. J Cardiovasc Electrophysiol 4. Franz MR, Zabel M. Electrophysiological basis of 16. Rautaharju PM, Park L, Rautaharju FS, Crow R. A QT dispersion measurements. Progr Cardiovasc standardized procedure for locating and document- ing ECG chest electrode positions. Consideration of 5. Bachvarov V, Malik M. Measurement and interpre- the effect of breast tissue on ECG amplitudes in tation of QT dispersion. Progr Cardiovasc Diseases women. J Electrocardiol 1998;31:17–29.
17. Rautaharju PM, Punsar S, Blackburn H, Warren J, 6. Davey P. QT interval and mortality from coro- Menotti A. Waveform patterns in Frank-lead rest nary artery disease. Progr Cardiovasc Diseases and exercise electrocardiograms of healthy elderly men. Circulation 1973;48:541–548.
7. Lee KW, Kligfield P, Dower GE, Okin PM. QT dis- 18. Rautaharju PM, Warrren J, Wolf HK. Waveform persion, T-wave projection, and heterogeneity of re- vector analysis of orthogonal electrocardiograms.
polarization in patients with coronary artery dis- Quantification and data reduction. J Electrocardiol ease. Am J Cardiol 2001;87:148–151.
8. Kors JA, van Herpen G, van Bemmel JH. QT dis- 19. Rautaharju PM, Zhou SH, Park LP. Improved persion as an attribute of T-loop morphology. Circu- electrocardiographic models for left ventricular mass adjusted for body size, with specific algo- 9. Priori SG, Mortara DW, Napolitano C, Diehl L, rithms for normal conduction, bundle branch blocks Paganini V, Cantu F, Cantu G, Schwartz PJ. Evalu- and old myocardial infarction. J Electrocardiol ation of the spatial aspects of T-wave complexity in the long-QT syndrome. Circulation 1997;96:3006– 20. Zabel M, Portnoy PR, Franz MR. Electrocardio- graphic indexes of dispersion of ventricular repo- 10. Macfarlane PW, McLaughlin SC, Rodger JC. Influ- larization: An isolated heart validation study. J Am ence of lead selection and population on automated Coll Cardiol 1995;25:746–752.

Source: http://ave.dee.isep.ipp.pt/~lbf/AS4CARDIO/007.pdf