Patients with idiopathic dilated cardiomyopathy (DCM) are at high risk of sudden death, which could be prevented by implantable cardioverter-defibrillators (ICDs). However, the arrhythmic risk stratification of DCM patients is still an open issue (1- 2). Similar to what has been found in patients with previous myocardial infarction, depressed left ventricular ejection fraction (LVEF) is independently associated with increased risk (2). The use of left ventricular dysfunction as the only parameter to define DCM patients eligible for ICD translates into a large number of patients who do not benefit. There is, therefore, a need for other parameters to improve patient selection. The analysis of ventricular repolarization is an intriguing way to implement risk stratification. QT intervals and their dispersion at electrocardiogram (ECG) failed to demonstrate any role in predicting arrhythmic events (2).

The possibility to evaluate ventricular depolarization dynamicity during the 24-h period should provide better information. In comparison with ECG evaluation, the analysis of QT dynamicity and/or variability could offer a more complete assessment of ventricular depolarization, representing the expression of the complex interaction between arrhythmic substrate, heart rate, and autonomic nervous system activity (3). The potential usefulness of this kind of analysis has been demonstrated in patients with DCM (4), ischemic cardiomyopathy, and chronic heart failure (5- 7).

The aim of the present study was to evaluate the role of QT dynamicity in predicting major arrhythmic events in a group of patients with DCM.

The patients participating in this prospective study were selected from a series of consecutive patients referred to our institution between September 1998 and June 2005. Dilated cardiomyopathy was diagnosed on the basis of patients’ clinical history, physical examination, 12-lead ECG, chest radiography, echocardiography, left ventriculography, and coronary angiography according to the World Health Organization criteria (8). The patients with history of sustained ventricular tachycardia (VT) and/or ventricular fibrillation (VF) were excluded. The other exclusion criteria were previously described (9). At the time of enrollment, all of the patients were in stable clinical condition and had been taking conventional therapy for at least 3 months. At baseline, all of the participants underwent a physical examination, an ECG, an echocardiographic examination, and a 24-h ECG recording.

The local ethics committee approved the study, and all patients gave their written informed consent.

Twelve-lead ECGs were analyzed as previously described (10). Briefly, they were scanned by means of a flat scanner (HP ScanJet 5300C, HP, Palo Alto, California), with a resolution of 600 dots/inch (equivalent to <1 ms per dot). The QRS duration, QT interval, and the preceding RR interval were measured using specific software written in Visual Basic 6.0 language for PC-compatible computers (F.M.) that works with all Windows operating systems. The software provided the use of semiautomatic calculation. The ECGs were analyzed by a single operator (S.C.) and validated by another (M.I.). QRS duration was calculated in standard leads from the start to the end of the QRS complex, and the longest QRS was considered (11). The QT interval and the QT interval corrected for heart rate (Bazett’s formula) were calculated in all leads from the onset of the QRS complex to the end of the T-wave, at the point in which it returned to the isoelectric line (12). Leads with a small T-wave (<50 μV) were excluded (13). A detectable deflection after the T-wave was considered a U-wave when the interval between the end of the T-wave and the apex of the doubtful deflection was ≥100 ms (14). QT dispersion (QTd) was calculated as the difference between the longest and the shortest QT interval among the 12 leads (12).

The echocardiography recordings were made using a phased-array echo-Doppler system (Sonos 5500, Philips, Eindhoven, the Netherlands) equipped with a 3-Mhz transducer. After resting for 10 min, the patients were examined in the left lateral recumbent position using standard parasternal, short- and long-axis, and apical views. Baseline left ventricular end-diastolic diameter (LVEDD) and LVEF (Simpson’s rule) were calculated by a single operator (R.R.).

The 2-channel 24-h ECG recordings (model 445A, Del Mar Avionics, Irvine, California) were always begun between 8:00 am and 9:00 am. During the day, the subjects were allowed to undertake all of their usual activities. The ECG recordings <16 h in duration or with <90% of the recording suitable for analysis were excluded (15). The recordings were analyzed by using the ELATEC system (ELA Medical, Mountrouge, France). A manual overread was also performed by an investigator (A.S.). The presence of nonsustained ventricular tachycardia (NSVT) was defined as more than 3 consecutive ventricular premature beats at >120 beats/min (2). Heart rate variability was assessed by measuring mean normal RR cycle length and the standard deviation of normal RR intervals (SDNN). In order to evaluate QT dynamicity, all the tapes first underwent a 200 Hz A/D conversion with an 8 bit and 10 mV amplitude resolution, as previously described (16). The digitized signals were processed for QT dynamicity by the software ELATEC, which has proven to be a clinically feasible method of assessing ventricular repolarization dynamicity (4- 5,7,16). The recordings were divided into templates of 30 s each (2,880 templates for each 24-h recording), and for each template the algorithm automatically measured the QT apex (QTa), QT end (QTe), and the RR interval. The T-wave apex was determined by fitting a parabola through the peak of the T-wave (17), whereas the T-wave end by the intersection of the tangent of the downslope of the T-wave with the isoelectric baseline. The software also computed the slopes of the linear regressions of QTe and QTa values plotted against the corresponding RR interval (QTe/RR and QTa/RR) (5,7). The QT dynamicity indexes considered were: the QT-slope (the slope of the regression line between QT and RR during the 24-h period) and the mean QT corrected for heart rate (Bazett’s formula). QT dynamicity was analyzed by a single operator (A.S.) and validated by another (M.I.). The calculation was made on the basis of only 1 lead (CM5 or, if not analyzable, CM2) and of templates whose amplitude was ≥0.15 mV. The correct measurement of QT for each template was verified, and the templates with incorrect QT evaluation were deleted.

Patients were followed up as outpatients in our heart failure clinic for at least 6 months. The clinical end points were major arrhythmic events defined as spontaneous VT, VF, or sudden death (i.e., death within 1 h after the onset of symptoms in a previously medically stable patient, death during sleep, or unwitnessed death) (2). The end points were reviewed and classified by 2 senior investigators (C.F., M.P.) who were blinded to the baseline evaluation.

The continuous variables are expressed as mean values ± SD. The continuous variables were compared by using Student t test. The correlations between the variables were analyzed by means of Pearson’s linear correlation. In the case of patients experiencing multiple events, the analysis was restricted to the first event. An individual was considered censored when she/he underwent cardiac transplantation or died from nonarrhythmic events. Cox proportional hazards model was used to assess the association of the study variables with the events (hazard ratio [HR] and 95% confidence interval [CI] for risk factors are given). A Cox proportional hazards model was fitted for each of QT dynamicity parameter that was significantly associated with arrhythmic events at univariate analysis, including other significantly associated clinical variables. Beta-blockers were included in the model as they may influence outcome (9). The multivariate analysis was also adjusted for the presence of QRS duration ≥120 ms. The HR for a continuous variable refers to the risk ratio per unit of the analyzed variable unless specified otherwise. The event-free curves were based on Kaplan-Meier analysis stratified by median slope of linear regression analysis of QTe and RR intervals (QTe-slope) and LVEF, and NSVT, and compared using log-rank test. Statistical measures of sensitivity, specificity, positive and negative predictive value were computed from the survival values at 36 months. The analyses were made using Statistica 6.1 software (StatSoft Inc., Tulsa, Oklahoma). The p values of <0.05 were considered statistically significant.

From the 241 consecutive patients who agreed to participate, 179 were considered for the study, and their clinical characteristics are shown in (Table 1). Twenty-three patients with computable QT dynamicity were excluded from the analysis because 13 of them had a positive history of VT and/or VF, and 10 had ventricular paced rhythm. In 39 patients, QT dynamicity analysis was not computable, due to atrial fibrillation in 16, paced atrial rhythm in 6, frequent ventricular or supraventricular ectopic beats in 5, poor quality of the 24-h ECG recordings, and/or T-wave not analyzable in 12. At the time of enrollment, 66 patients had a QRS ≥120 ms due to left bundle branch block in 64 (97%) and to right bundle branch block in 2 (3%).

The QTe-slope significantly correlated with LVEF (r = −0.38, p < 0.001), LVEDD (r = 0.16; p = 0.034), SDNN (r = −0.43, p < 0.001), and mean QTe corrected for heart rate (r = 0.44, p < 0.001). No significant correlations were found when age, QRS duration, and QTd were considered. Female patients had higher values of QTe-slope (0.22 ± 0.07 vs. 0.19 ± 0.07, respectively, p = 0.010), as did the patients in New York Heart Association functional class III versus those in class I/II (0.23 ± 0.08 vs. 0.19 ± 0.07, p = 0.009), patients with NSVT (0.22 ± 0.08 vs. 0.19 ± 0.07, p = 0.003) and QRS duration >120 ms (0.22 ± 0.07 vs. 0.19 ± 0.07, p = 0.011). The QTe- and QTa-slope significantly correlated with each other (r = 0.69; p < 0.001), but the mean QTe-slope was significantly greater than QTa-slope (0.20 ± 0.07 vs. 0.18 ± 0.07, p < 0.001).

Fifteen patients died during the follow-up (39 ± 22 months), 13 of them for cardiac causes, including 4 who died after acute decompensated heart failure and 9 who died suddenly. Ten patients showed sustained VT and 5 VF. Two patients who showed VT and VF were among those who died for progression of heart failure. The clinical characteristics of patients with and without arrhythmic events are shown in (Table 2). Univariate analysis showed that LVEF, NSVT, SDNN, QTe-slope were significantly associated with major arrhythmic events (Table 2). The QTe-slope was also significantly associated with arrhythmic events also when patients with and without QRS duration ≥120 ms were evaluated separately (p = 0.046 and p = 0.001, respectively). At multivariate analysis, only the QTe-slope, LVEF, and NSVT were significant predictors of events, regardless of SDNN, a QRS duration ≥120 ms, or beta-blocker therapy (Table 2). (Figure 1) shows the Kaplan-Meier curves for arrhythmic events of the patients dichotomized for median QTe-slope values (0.19). (Figure 2) shows the contribution to arrhythmic risk stratification offered by combining LVEF (<35% vs. >35%), NSVT, and QTe-slope (>0.19 vs. ≤0.19). Arrhythmic events were more frequent among patients with NSVT and a low LVEF (panel A) and those with a low LVEF and steeper QTe-slope (panel B). No significantly higher risk was observed among patients with a higher LVEF and NSVT or steeper QTe-slope. When all 3 variables were considered together (panel C), the patients with a low LVEF and NSVT or a steeper QTe-slope were found to be at highest arrhythmic risk.

When the major arrhythmic events after 36 months of follow-up were considered (18 events), the presence of an LVEF of >35% (42% of the population) was associated with a low probability of arrhythmic events (3%). In the subgroup with LVEF ≤35% (58% of population), the probability of events was of 20%, with 93% sensitivity and a 46% specificity. (Figure 3) shows the probability of the occurrence of the arrhythmic events and the proportion of population on the basis of the presence or absence of NSVT and a QTe-slope above or below the median value in patients with an LVEF of ≤35%. The presence of NSVT and QTe-slope >0.19 defined a small population with the highest probability of events with 62% sensitivity and 88% specificity. The presence of NSVT and/or a QTe-slope of >0.19 showed 90% sensitivity and 60% specificity in identifying patients with arrhythmic events (Figure 3).

The QTe-slope was also significantly associated with total mortality of patients at univariate (HR for 0.05 increase 1.92; 95% CI 1.43 to 2.59; p < 0.001) as well as at multivariate analysis (HR for 0.05 increase 1.86; 95% CI 1.21 to 2.86; p = 0.005) after correction for beta-blocker therapy, QRS >120 ms, and all predictors at univariate analysis (age, New York Heart Association functional class, LVEDD, LVEF, NSVT, and SDNN).

The main finding of this study is that the evaluation of ventricular repolarization dynamicity by means of slope of QTe/RR interval regression analysis independently identifies DCM patients prone to experience major arrhythmic events.

Ventricular repolarization is a critical time in the cardiac cycle playing a considerable role in the pathophysiology of malignant arrhythmias. Its clinical evaluation should provide parameters that reflect cardiac electrical instability and, therefore, the increased risk of ventricular arrhythmias (18- 19). Twelve-lead ECG measurements of the QT interval (i.e., QT interval and QTd) are considered a global index of the duration and dispersion of repolarization in the ventricular myocardium (3) and have been largely investigated in order to better identify patients with various clinical conditions prone to experience major arrhythmic events. However, the results of these studies are conflicting (20- 23). This could be due to the fact that ECG parameters poorly reflect the complexity of ventricular repolarization process depending on different dynamic components as transmembrane ion currents, heart rate, and autonomic nervous system activity (3). The analysis of QT interval dynamicity and/or variability from 24-h ECG monitoring should offer a better tool in order to evaluate the complex interaction between arrhythmic substrate (i.e., dispersion of refractoriness) and its dynamic determinants. In this study, we used the computation of slope of RR/QT intervals regression analysis and we found that the steeper the QTe-slope the greater the arrhythmic risk. In pathophysiological terms, a steeper QT slope may reflect an excessive shortening of QT with fast rate and/or an excessive lengthening of QT interval with slower heart rates (3,24- 25). Both these conditions have been suggested to reflect a greater risk of arrhythmic events. It is interesting that also other parameters analyzing repolarization changes under dynamic conditions have shown a predictive role in risk stratification. Although not evident from the results of MACAS (Marburg Cardiomyopathy Study) (2), it has recently been shown that T-wave alternans can predict death or sustained ventricular arrhythmias in patients with either ischemic or nonischemic left ventricular dysfunction (26).

The independent prognostic role of QT dynamicity is particularly interesting given the difficulties in stratifying the arrhythmic risk of DCM patients (1- 2) that limit, as a consequence, the potential use of ICDs in the primary prevention (27). Our results are consistent with those of the MACAS study demonstrating the relevant role of a low ejection fraction in predicting arrhythmic events in patients with DCM (2). We carefully selected DCM patients who were in sinus rhythm and had clinical characteristics similar to those of the patients enrolled in the MACAS study. Furthermore, most of our patients were taking conventional therapy including beta-blockers at the time of the enrollment. Multivariate analysis showed that not only LVEF but also QTe-slope and NSVT were significantly and independently associated with arrhythmic events. The most interesting information comes from the combination of these parameters whose integration allows the identification of a subgroup of patients with a low LVEF who were at higher risk of arrhythmic events. From the clinical point of view, these data suggest a possible different approach to a better selection of patients who will benefit from ICD implantation. Current American College of Cardiology/American Heart Association guidelines report a potential benefit from ICD implantation in patients with nonischemic cardiomyopathy (1) on the basis of the presence of a low LVEF and functional limitation, according to clinical trials that investigated the usefulness of ICD prophylactic implantation (27- 28). Also, in our study a very low rate of arrhythmic events in patients with LVEF >35% was found (i.e., high sensitivity and high negative predictive value, even if a low specificity was observed). The predictive value of LVEF was improved by integrating it with the other noninvasive arrhythmic risk parameters significantly associated with arrhythmic events (i.e., ventricular repolarization dynamicity and NSVT). Considering the patients with a low LVEF, no arrhythmic events were observed after 36 months in those without both NSVT and steep QTe-slope. The presence of NSVT or a steep QTe-slope increased both arrhythmic risk at 36 months (from 20% to 24%) and specificity (from 46% to 60%), with a high sensitivity (90%). When both NSVT and steep QTe-slope were present, a higher positive predictive value (43%) and specificity (88%) were obtained, even if with a lower sensitivity (62%).

The high percentage of patients taking beta-blockers at the time of enrollment (>70%) reinforces the prognostic role of QT dynamicity because it is known that beta-blockers modify heart rate, reduce the risk of major arrhythmic events, and, in patients with long-QT syndrome, prevent an abrupt increase in the QT interval at high heart rates (29).

The limitations of QT/RR slope analysis are strictly related to the feasibility of measuring it. We excluded patients with atrial fibrillation or paced rhythm, as well as those with abnormal ventricular repolarization, who accounted for 20% of the otherwise eligible patients. However, it was possible to analyze QT dynamicity in patients with a prolonged QRS duration mainly due to left bundle branch block, and, in these patients, the QTe-slope was also significantly associated with arrhythmic events, and its predictive value did not depend on the duration of QRS.

Analogously to the results of previous studies (4,6), we were not able to demonstrate an association of QT-a dynamicity with major arrhythmic events. Although the QTe and QTa slopes significantly correlated with each other, the mean QTa slope was less and did not remain significantly associated with arrhythmic events at multivariate analysis. These findings strengthen the hypothesis that the QTa and QTe slopes offer different information due to the fact that the QTa is highly affected by heart rate, whereas the end of the T-wave may reflect the interactions between ventricular repolarization, autonomic nervous system activity, and heart rate, as well as the activity of ventricular cells that may be greatly involved in arrhythmogenesis (6,30).

In conclusion, an increased QTe-slope in patients with DCM is associated with the occurrence of major arrhythmic events regardless of other clinical variables, thus suggesting its clinical usefulness in stratifying arrhythmic risk.

The authors would like to thank Cataldo Balducci, Angela Burdi, Mariella Vitone, and Anna Cavallo for their helpful cooperation.