Influence of Cardiac Autonomic Neuropathy on Heart Rate Dependence of Ventricular Repolarization in Diabetic Patients

The population enrolled in this retrospective study comprised 27 subjects (12 women, 15 men; mean age 51.8 years, range 39–66 years), divided into three groups of subjects (age and sex-matched): nine control subjects, nine type 2 diabetic patients with DAN (DAN+), and nine type 2 diabetic patients without DAN (DAN−).

The diabetic subjects were patients who were referred to Jean Verdier University Hospital as being beyond the scope of a protocol designed to assess silent myocardial ischemia (13,14). They were selected according to strict inclusion criteria: absence of coronary artery disease (no history of angina or myocardial infarction, a negative exercise ECG stress test, a negative thallium-201 myocardial scintigraphy with intravenous infusion of dipyridamole, and no significant alteration of ST segment on a 48-h ambulatory ECG recording) and normal left ventricular function (as assessed by echocardiographic indexes). None of the patients were taking drugs known to prolong the QT interval (in particular, β-blockers were withdrawn at least 48 h before entry into the study).

Control subjects were asymptomatic volunteers recruited at Lariboisière University Hospital who showed a complete negative cardiovascular evaluation, including blood pressure, resting 12-lead ECG, echocardiogram, exercise stress test, and Holter monitoring. All were free of any cardiovascular medication.

DAN was defined as an alteration in at least one of the three bedside clinical tests evaluating parasympathetic innervation to the heart, as proposed by Ewing et al. (10) and as previously described (14). Briefly, the RR interval variations were measured during the tests, using a computerized device (Q-med). The Valsalva test, conducted with the patient seated, consisted of forced exhalation and maintaining a pressure of 40 mmHg for 15 s. The result was expressed as the ratio RR maximum/RR minimum. The test was done three times consecutively, and the mean value for this ratio was taken. The deep breathing test was performed in a previously trained subject and consisted of the subject taking six deep breaths in 1 min in a supine position. The results were expressed as the mean value for the ratio of maximal heart rate/minimal heart rate. During the lying-to-standing test, the maximal heart rate (about the 15th beat)/minimal heart rate (about the 30th beat) ratio was calculated. The results of the three tests were compared with those from a control series, with age taken into account (15). The reproducibility of these methods has been demonstrated in diabetic patients: the coefficients of variations are 9.2, 12.6, and 6.4% for the Valsalva, deep-breathing, and lying-to-standing tests, respectively (16).

The QT intervals were manually measured by two cardiologists blind to the protocol on a 12-lead resting ECG printed at a paper speed of 25 mm/s, according to the method of Lepeschkin and Surawicz (17). In each lead, the QT interval was the mean of three successive individual intervals. The RR interval used for QT rate correction was the mean of all the sinus available RR intervals, and the rate correction was made by the Bazett formula (square root of the mean RR interval). QT dispersion was calculated as the difference between the longest and shortest QT interval from all available leads.

A 24-h ambulatory Holter was recorded and further digitized at 128 samples per second. Tapes were then manually validated and transferred onto a personal system, dedicated to heart rate variability (HRV) and QT analysis. In addition, a diurnal period corresponding to the 8 consecutive awake hours with the highest heart rate, and a nocturnal period corresponding to the 4 consecutive sleeping hours with the lowest heart rate were separately analyzed.

The mean sinus RR interval was calculated together with three time-domain HRV parameters: SDNN (standard deviation of consecutive RR intervals), a global index of sinus variability, and two vagal activity indexes—PNN-50 (percentage of normal consecutive RR intervals differing by >50 ms) and RMSSD (root mean of squared successive differences) (18). The difference between the mean nocturnal and diurnal RR intervals was also computed. HRV was also studied in the frequency domain. The power spectra obtained were integrated over low-frequency (LF) and high-frequency (HF) bands, as recommended by the International Task Force on HRV (18). The LF/HF ratio was calculated and log transformed to normalize the distribution.

QT analysis from long-term ECG was mainly based on the assessment of the heart rate dependence of the QT interval. This was evaluated by calculating the QT/RR relation on ECG templates obtained with a selective beat averaging technique, as previously reported (19). Briefly, the QT apex interval (from beginning of QRS to apex of the T wave; QTa interval) was measured over the preceding RR intervals, but in relative stable rate conditions (RR-2 = RR-1 ± 50 ms and RR min-1 = RR-1 ± 15 ms), and each individual QRS-T was incorporated into the closest RR template. In our experience, with the selective beat−averaging technique, the slopes of the QT/RR relation are steeper and the correlation coefficients are higher than with the beat-to-beat approach (all individual sinus beats), and the within-subject reproducibility is good, as has been previously shown (20).

For QT/RR relation, repolarization duration was analyzed by a fully automated technique. The beginning of the QRS was identified on the vector magnitude of the ECG signal using the first derivative of the initial deflection of the QRS. The apex of the T wave was defined as the peak of the parabola interpolated through the samples of the T wave.

In addition to the QT heart rate dependence, the amplitude of the nocturnal QT lengthening, defined as the mean difference between nocturnal and diurnal QT intervals over the range of common RR intervals (rate independent analysis), was analyzed (21).

All data are given as means ± SD. Between-group comparisons were computed by nonparametric tests (Wilcoxon and χ² for unpaired series). Correlations between continuous parameters were calculated according to a linear regression model.

There were no significant differences in the major clinical characteristics between the three groups (Table 1). However in the DAN+ and DAN− subjects, BMI was higher and hypertension more frequent than in control subjects. Peripheral neuropathy and use of ACE inhibitors were both more frequent in DAN+ than in DAN− patients. Based on the results of the cardiac autonomic function tests in the DAN+ group, five subjects had mild DAN (one abnormal test), one had confirmed DAN (two abnormal tests), and three had severe DAN (three abnormal tests).

The RR interval was significantly shorter (P < 0.05) in the DAN+ and DAN− than in control subjects (735 ± 94, 735 ± 77, and 899 ± 161 ms, respectively). However, no significant differences were observed for the QT or QTc interval (402 ± 56, 404 ± 23, and 390 ± 24 ms, respectively). The magnitude of QT dispersion was also similar in the three groups (38 ± 10, 39 ± 6, and 34 ± 9 ms, respectively).

Mean RR intervals were consistently shorter in the diabetic subjects when compared with the control subjects (Table 2). The SDNN was significantly reduced in DAN+ and DAN− subjects compared with in control subjects. The diurnal SDNN was significantly shorter in the DAN+ than in the DAN− subjects, and the nocturnal SDNN was significantly shorter in DAN+ than in control subjects. For parasympathetic indexes, only the RMSSD on the 24-h period tended to be lower in the DAN+ than in the DAN− group (P = 0.051). In DAN+ patients, the LF/HF in the 24-h period analysis and during the diurnal and nocturnal period was significantly lower than in control and DAN− subjects and was reversed during the night.

Correlation coefficients of the QTa/RR slopes based on selective beat averaging were much higher than for beat-to-beat analysis (0.80–0.95) (Table 2; Fig. 1). An autonomic modulation of the QTa/RR relation was evidenced in control subjects using this technique, and this trend (steeper diurnal than nocturnal slopes) was preserved in DAN− subjects. Most importantly, this long-term modulation was significantly different in DAN+ subjects, with a reversed day/night pattern: the nocturnal QT rate dependence was increased and the difference between diurnal and nocturnal QTa/RR slopes was negative.

There was a nonsignificant trend toward a decreased nocturnal QT lengthening in DAN+ subjects when compared with control and DAN− subjects (Table 2).

The goal of this study was to assess the long-term influences of the ANS on the QT heart rate dependence in a model of diabetic patients with parasympathetic denervation. The major findings can be summarized as follows: 1) QT heart rate dependence evaluation discriminated better between control subjects and DAN+ subjects than the time-domain HRV parameters; 2) the nocturnal QT heart rate dependence evaluation was increased in DAN+ subjects when compared with DAN− and control subjects; and 3) the physiological long-term autonomic modulation of the QT heart rate dependence was impaired in DAN+ subjects, with a reversed day/night pattern.

Characteristics of the QT interval are recognized as powerful discriminators between two populations, in the long QT syndrome and after myocardial infarction (3,4). In the present study, the QTc did not differ significantly among DAN+, DAN−, and control subjects, a finding in accordance with a recently published meta-analysis (11). The fundamental reason is that applying any formula supposes QT dynamicity to be a physiological constant. This study demonstrates that this is not so. Dynamicity is a characteristic of individuals, modulated by the ANS. Bellavere et al. (1) found a steeper heart rate dependence in DAN+ subjects than in DAN− and control subjects (calculated by measuring resting QT interval and QT interval during different autonomic tests). Ewing et al. (10) also reported a significantly steeper QT/RR relation in DAN+ diabetic subjects. Our study differs from the latter in that the data were taken separately in the daytime and nighttime.

In our population, the time-domain HRV parameters did not correctly discriminate DAN+ and DAN− subjects, except for diminished global HRV indexes. However, in DAN+ patients, the LF/HF ratio, an index of sympathovagal balance, was significantly reduced during the diurnal and nocturnal periods and was reversed during the night.

Differences in the circadian behavior of QT/RR slopes observed between DAN+ subjects versus DAN− and control subjects may be related to an altered autonomic modulation rather than to an intrinsic myocardial abnormality. Indeed, our diabetic patients were not affected by coronary heart disease, left ventricular dysfunction or hypertrophy, or significant thyroid disease, and were not taking antiarrhythmic drugs (including β-blockers), all factors known to affect QT interval duration, dispersion, or variability. Both DAN+ and DAN− groups included the same proportion of hypertensive patients and had similar left ventricle mass indexes (Table 1). To our knowledge, neither hypertension nor antihypertensive drugs (except for β-blockers) have been reported as playing a role in the cardiac autonomic function tests or in QT interval heart rate−dependence evaluation (22). Therefore, if one accepts that the ventricular substrate was not impaired in our patients, then the reversed day-night pattern appears to be associated with the cardiac neuropathy itself.

Long-term circadian influences on ventricular repolarization in healthy subjects have been described by Browne et al. (21), who showed that QT duration was increased by 25 ms at night in a heart rate−independent manner. This was confirmed in diabetic neuropathy and coronary artery disease (6,7). Our group was the first to demonstrate that ANS activity exerts a long-term modulation on QT heart rate dependence (5,23,24). Our DAN+ subjects were recruited on the basis of clinical parasympathetic denervation. One could assume that the observed differences in autonomic modulation of QT heart rate dependence are attributable to parasympathetic denervation. In keeping with this finding are results from in vitro and in vivo studies. Studies on canine myocardium have reported an increase of action potential duration and ventricular refractory periods with vagal stimulation (acetylcholine), suggesting the inverse effect of parasympathetic blockade (25). In humans, parasympathetic blockade with atropine (with or without atrial pacing) has been shown to result in a higher QT heart rate dependence (i.e., QT interval decreases more at high heart rates) at baseline and during exercise (26).

The duration of the QT interval and QT dispersion have emerged as prognostic factors in many clinical settings, such as in postmyocardial infarction, congenital long-QT syndrome, and chronic heart failure and in large subsets of the general population (2–4). In DAN+ patients, an abnormally high incidence of sudden death (9,10) and an abnormal QT interval duration and/or variability have been noted, suggesting a link between the two (1,10). Ewing et al. (10) also found a longer QT interval in diabetic patients who died during follow-up (mostly sudden deaths) as compared with patients who survived. Recent studies have assessed the prognostic significance of an increased QT dispersion in type 2 diabetes (12). A long-term follow-up on a large cohort of diabetic patients should help in assessing the prognostic significance of QT heart rate−dependence evaluation, in particular in patients with DAN.

In conclusion, in diabetic patients affected by mostly mild parasympathetic denervation, QT heart rate dependence assessed from Holter data was found to be clearly impaired; furthermore, a reversed day-night pattern was observed, associated with an impaired sympathovagal balance. The analysis of ventricular repolarization could represent a sensitive index of the progression of neuropathy; this possibility should be tested in a larger series of patients. The evolution and potential prognostic impact of a reversed day-night pattern with steep nocturnal QT/RR relation still remains to be defined.

This study was made possible by a grant from Alfediam and Lilly France. The authors acknowledge Fabio Badilini for his contribution in QT analysis.