24-Hour QT variability in heart failure☆

Background 

Heart failure (HF) is associated with increased instability of repolarization,1 and this increased lability may contribute significantly to the increased propensity for catastrophic ventricular arrhythmias. The QT variability index (QTVI) and normalized QT variance (QTVN) are 2 validated methods for quantifying repolarization variability that have been found predictive of ventricular tachycardia or fibrillation recorded on implanted defibrillators in subjects with severe left ventricular systolic dysfunction.2 These have hitherto been applied to short-term Holter recordings (≤10 minutes) only, and it seems reasonable that longer ambulatory recordings may be more predictive of subsequent arrhythmias. Furthermore, if significant diurnal variation in QT variability is present, future studies of this phenomenon will need to take into account the time of day data were recorded. Finally, an exploration of diurnal changes in repolarization variability represents the first step to understanding whether beat-to-beat changes in the QT contribute to the described morning increase in risk for lethal arrhythmias. In this report, we apply the QT variability approach to high-resolution ambulatory recordings from 372 individuals enrolled in the Gruppo Italiano per lo Studio della Sopravvivenza nell'Insufficienza Cardiaca (GISSI HF) study. We present the first report, to our knowledge, of QT variability measurements during a 24-hour Holter recordings in an HF population.

Methods 

In a previously described patient population3 from the GISSI HF project, 372 twenty-four hour digital Holter monitor recordings were performed at the time of enrollment. Forty-two centers throughout Italy performed the Holter monitoring. Patient inclusion criteria included New York Heart Association (NYHA) class II-IV HF of any etiology. Patients were excluded if they had a revascularization procedure or an acute coronary syndrome within 1 month of study enrollment or if cardiac surgery was planned within the follow-up period. In addition, any noncardiac comorbidity that would limit their follow-up period in the study led to exclusion. All patients were recommended standard medical management for HF during the course of the study to include angiotensin-converting enzyme inhibitors, β-blockers, diuretics, digitalis, and spironolactone. Amiodarone, aspirin, and coumadin administration was allowed at the discretion of each attending physician. Patients were not excluded based on any estimates of left ventricular function. There were no sex or age limitations to the study enrollment.

Holter recordings were performed with a high-resolution (1000 Hz) digital 12-lead portable Holter monitoring system (model H12+; Mortara Instruments, Milwaukee, WI). The recordings were transferred to an investigator group blinded to patient characteristics and outcome data. Each recording was translated from its proprietary format into binary code using the program SuperECG (Mortara Industries).

A novel analysis program was created for use with 24-hour Holter data using MATLab software (The MathWorks, Natick, MA; programming courtesy of B. Fetics and R. Berger, Johns Hopkins University and Robin Medical, Baltimore, MD). In a custom graphical interface, the operator reviews the 8 nonderived leads from the Holter and chooses the “best” lead using the parameters (in order of importance) of the least artifact, most prominent T wave, and most recognizable P wave. The program then generates signal-averaged QT templates at regular intervals from the chosen lead. The template QT interval is automatically compared with each individual QT (from that lead) and stretched or compressed as necessary until a high degree of “fit” is achieved. A new value for the QT interval is thereby generated, and because it uses the entire T and U waves, the QT value is not dependent on defining the terminus of the T or U wave. This greatly reduces the susceptibility of the measurement to inaccuracies due to noise or interindividual disagreement. A normalized QTVI is then derived according to the following equation:

QT variability index and QTVN were generated for every 5-minute epoch. Epochs with greater than 10% rejected beats were excluded. Reasons for rejection of a beat are nonsinus origin, inclusion of the P wave in the QT template time region, excessive signal noise, and the beat that follows a beat of nonsinus origin.

Statistics 

Data were summarized using means (range) or frequency (percent). Pearson product-moment correlation coefficients were calculated to describe the associations among continuous variables. Univariate comparisons between patients in the highest quartile for QTVI or QTVN and patients in the lower quartiles were conducted using Student 2-sample t test for continuous variables and χ2 tests for dichotomous variables. Diurnal variation in QTVI and QTVN was examined using repeated-measures analysis of variance (ANOVA) with time as a within-subjects factor. Post hoc comparisons among time periods were adjusted for multiple comparisons using the Tukey-Kramer method. P values less than .05 were considered statistically significant. SAS 9.0 and SPSS 14.0 statistical software for Windows were used for data analysis.

Results 

QT variance by patient characteristics 

A total of 372 Holter recordings were available for the analysis (Table 1). The univariate comparisons of important clinical variables between the highest quartile of QTVI and QTVN versus the remaining quartiles are summarized in Table 2. Frequency of female sex, history of nonsustained ventricular tachycardia, and left ventricular ejection fraction (LVEF) were not statistically significantly different in the higher quartile, although β-blocker use was marginally lower in the top quartile (P = .029). Mean NYHA category was significantly higher in the upper quartile for QTVI (P = .0004) and QTVN (P = .019). When NYHA classes were dichotomized into 2 versus 3 to 4, patients with a higher NYHA class had increased QTVI for the entire the 24-hour period, −0.855 versus −1.053 (P < .0001). Normalized QT variance was slightly higher during the 24-hour period, 0.219 versus 0.234 (P = .011) in the patients in NYHA 3-4. Comparing continuous values, QTVI correlated modestly but significantly with NYHA class (r = 0.22, P = .0003) and heart rate (r = 0.39, P < .0001), whereas QTVN did not correlate with either (r = 0.12, P = .053 for NYHA, r = 0.11, P = .07 for heart rate). Normalized QT variance correlated inversely with serum potassium (r = −0.22, P = .0002).

Table 1. Patient characteristics

		Total N		%
	Age (y)	Mean	64.8
		Range	22.3-89.6
	Sex	Male	303	81
		Female	69	19
	NYHA	2	296	79.6
		3	70	18.8
		4	6	1.6
	LVEF	<35%	257	69.1
		≥35%	115	30.9
	β-Blocker	No	112	31.0
		Yes	260	69.9
	H/o NSVT	No	242	65
		Yes	130	35.0

H/o NSVT indicates history of non-sustained ventricular tachycardia.

Table 2. Comparison of the high risk quartile of QTVN and QTVI to lower three quaRTILES

		QTVN comparison by highest quartile			QTVI comparison by highest quartile
		QTVN lower 3 quartiles	QTVN upper quartile	P	QTVI lower 3 quartiles	QTVI upper quartile	P
	β-Blocker	72.0%	64.2%	.209	74.0%	59.7%	.029
	H/o NSVT	28.0%	26.9%	.892	28.0%	26.0%	.892
	Heart rate	68.4	78.9	.147	67.9	71.0	<.0001
	Serum K+	4.5	4.2	.009	4.44	4.41	.738
	LVEF	32.8	32.8	.551	32.6	32.7	.962
	NYHA	2.2	2.3	.019	2.2	2.4	.0004

H/o NSVT indicates history of non-sustained ventricular tachycardia.

Diurnal fluctuation in QT variability 

A pronounced diurnal variation was demonstrated in both QTVI and QTVN. Both metrics were lowest during the midnight to 6 am time frame and increased throughout the day, peaking at noon to 6 pm, then decreasing 6 pm to midnight (see Fig. 1A). For QTVI, all 4 time points were significantly different by repeated-measures ANOVA with P < .001. For QTVN, the midnight to 6 am period was significantly different from all other periods (P < .0001), and there were significant differences between the noon to 6 pm and 6 pm to midnight periods (P < .0001) and between the noon to 6 pm and the 6 pm to midnight periods (P = .039). Given that QTVI represents a ratio, it is useful to consider whether the diurnal variations in QTVI are driven by shifts in the numerator or denominator. The denominator function represents normalized HRv, and when analyzed during the same periods, it seems that the normalized HRv changes little between the first 2 time points (P = .979), and then significantly drops after the noon hour (P = .001). These data suggest that the significant increase in QTVI between midnight to 6 am and 6 am to noon is driven by an increase in QT variance and not by a fall in normalized HRv, whereas the second increase after noon is driven principally by a fall in normalized HRv.

View Large Image Download to PowerPoint

Fig. 1. Comparison of mean QT variability for 24 hours in GISSI HF subjects. A, Mean ± SEM QTVI for 24 hours; all time points are significantly different by post hoc testing, P < .0001. B, Mean ± SEM QTVN for 24 hours; all time points are significantly different from 0 to 6 am by post hoc testing, P < .0001. C, Mean ± SEM in normalized HRv for 24 hours; there was significant diurnal variation (P < .001). After adjusting for multiple comparisons using Tukey procedure, no significant difference was found between the midnight to 6 am and the 6 am to noon time periods (P = .979), or between the noon to 6 pm and the 6 pm to midnight time periods (P = .683). All other differences were statistically significant (P < .001).

The diurnal trend in QTVI was maintained when NYHA category was included in the analysis. NYHA was dichotomized to 2 versus 3 to 4 (see Fig. 1A). By repeated-measures ANOVA, QTVI for each period was significantly higher for those patients in the group with the higher NYHA classes (all P < .001). Subanalysis of QTVI during the hour of the Holter with the maximum heart rate demonstrated a significant QTVI increase for higher HF classes (P = .014). The hour with the lowest heart rate showed the same significant results (P = .005). Normalized QT variance was not significantly correlated with NYHA class (r = 0.11, P = .054).

Discussion 

The principle findings of this study are as follows: (1) repolarization lability measurements as represented by QTVI and QTVN can successfully be performed for full 24-hour Holter recordings in an HF population; (2) repolarization lability has a pronounced diurnal variation with the lowest during the period of midnight to 6am, increasing to a peak at noon to 6 pm, then decreasing 6 pm to midnight; and (3) increased QTVI and QTVN are associated with different phenomena, with QTVI correlating with NYHA class and heart rate and with QTVN correlating negatively with serum potassium.

Our finding of increasing QTVI with increasing NYHA HF classes is in line with previous studies of selected populations in short-term recordings. In the first description of the QTVI calculation technique, Berger et al1 reported increased QTVI in a population of dilated cardiomyopathy patients, as well as an association with NYHA class; this likely reflects the fact that both heart rate and heart rate variability are included in the denominator of this ratio function, and these correlate positively with NYHA class. As in the present study, Berger et al found no correlation between LVEF and QTVI. Raghunandan et al4 demonstrated that optimizing HF medical treatment lowered the QTVI. In Multicenter Automatic Defibrillator Implantation Trial II patients, the highest quartile of QTVI had an increased risk of implantable cardioverter defribrillator (ICD)-refractory mortality, consistent with more “pump-failure” death in the top quartile for QTVI. In this same study, top-quartile QTVN was significantly associated with increased incidence of ventricular fibrillation, suggesting that QTVI is more predictive of HF death, whereas QTVN may be a better predictor of arrhythmic death.2 The modest negative correlation between serum potassium and QTVN and lack of association with NYHA class supports the notion that this metric may be a better index of arrhythmic propensity. One possible explanation for the association between increased QTVN and reduced serum potassium would be loss of repolarization reserve due to decreased rapidly activating component of the delayed rectifier current in the setting of reduced extracellular potassium.5

The diurnal variation of QTVI and QTVN seen in the present study is consistent with previous reports using short-term recordings of QTVI. Previous studies6 using QTVI short-term (10 minutes) analysis in a postmyocardial infarction population found that the ratio between the standard deviation of QT intervals to the standard deviation of R-R intervals for 24 hours (termed QT/R-R variability ratio) was predictive of total mortality. Interestingly, although QT variability peaks in the morning, QTc values typically increase at night compared with day.7 Given that the time of greatest risk for sudden death is around the time of waking to midmorning,8, 9 QT variability is likely to prove a better predictor for sudden death than QTc is. Indeed in the Multicenter Automatic Defibrillator Implantation Trial II cohort, QTc was not predictive of events, whereas QT variability was independently predictive of ventricular tachycardia or ventricular fibrillation. Importantly, the increase in QTVI in the morning hours seems to represent a true increase in variability of repolarization and not simply a change in normalized HRv. Whether this increase in QT variability correlates with the timing and incidence of cardiovascular mortality and how it compares to pure measures of heart rate variability is not known.

The diurnal variation of QTVI and QTVN has important implications with regard to the interpretation of research studies on ventricular repolarization and their clinical applications. If there is a variation of these metrics from morning through afternoon, this may present a confounding variable to interpreting results from different times of day. With the introduction of 24-hour Holter QTVI and QTVN automated measurement presented in this report, it would be possible to deliver results for an entire day with minimal user input and time. We believe that this creates a substantially more robust as metric for possible clinical use.

In conclusion, to the best of our knowledge, this is the first report of QTVI and QTVN calculations for a continuous 24-hour Holter monitor tracing in an HF population.

More rigorous recording, analysis, and use of derived metrics of QT variability may be of help to better stratify patients at high risk of either arrhythmic or nonarrhythmic death such as those with HF. This hypothesis should be tested in proper clinical trials.