Influence of Autonomic Neuropathy on QTc Interval Lengthening During Hypoglycemia in Type 1 Diabetes

A total of 109 adults who have type 1 diabetes and attend our outpatient clinics initially underwent screening for autonomic neuropathy, excluding those with hypertension, ischemic heart disease, peripheral vascular disease, nephropathy (urine albumin-to-creatinine ratio <2.5 mg/mmol in all subjects), and active proliferative retinopathy. Sixty subsequently attended for BRS measurement, 24 of whom agreed to have their QTc interval measured during experimental hypoglycemia. We approached an additional nine patients who had known CAN; four agreed to participate. We also studied eight individuals without diabetes (fasting blood glucose <6 mmol/l) matched with the diabetic groups for age, sex, and BMI. No participant was taking medication that is known to influence the autonomic nervous system or QT interval, and all had a normal resting electrocardiogram (ECG). Any therapy other than insulin was withdrawn 48 h before study. The North Sheffield Research and Ethics committee approved the protocol, and all participants gave written informed consent.

Participants were asked to avoid stimulants and exercise for 12 h beforehand, attended at least 2 h after eating, and rested for 10 min before recordings were undertaken. Capillary glucose was measured on arrival. We rescheduled one participant, who experienced biochemical hypoglycemia the day before his study.

Traditional cardiac autonomic function tests were performed according to a protocol reported by O’Brien et al. (17): 1) resting heart rate, 2) R-R interval response to the Valsalva maneuver, 3) R-R interval response to deep breathing (expiration [E]-to-inspiration [I] ratio; E:I difference), 4) R-R interval response to standing (30:15 ratio), and 5) postural blood pressure response. Diabetic participants were designated CAN+ when two or more tests were abnormal (17).

BRS can be considered as the change in heart rate per unit change in blood pressure. For BRS measurement, we used the “sequence” method, based on the automated identification of spontaneous sequences of three or more consecutive beats characterized by either a progressive rise or decline in systolic blood pressure (SBP) and associated changes in R-R interval (18). The sequence technique correlates strongly with BRS measurements obtained using traditional invasive methods (19). Blood pressure was tracked using a Portapres (TNO-TMP Biomedical Instrumentation, Amsterdam, Netherlands), a device measuring beat-to-beat changes in finger arterial pressure using a volume-clamp method. There is a strong correlation between blood pressure measurements obtained using this technique and intra-arterial recordings (20). Before and after BRS measurement, blood pressure was taken manually with a conventional sphygmomanometer to verify Portapres accuracy. For each participant, a minimum of three supine datasets were collected, each consisting of a 3-min recording period.

Analog blood pressure and ECG data were digitized using a PCM-DAS08 card (Amplicon) attached to an on-line personal computer (Toshiba 4090CDS) sampling at 200 Hz. A QRS detection algorithm was used to measure R-R intervals obtained from the three-lead surface ECG. Computer software was developed using Matlab (Mathworks, Nantucket, MA) to identify SBP changes with corresponding increases and decreases in R-R interval length. An “up sequence” (+RR/+SBP) was identified when three or more consecutive SBP readings increased by at least 0.5 mmHg, with a corresponding increase in R-R interval. A “down sequence” (−RR/−SBP) was highlighted when three or more consecutive SBP readings fell, with a reduction in R-R interval. The regression coefficient of these sequences (the slope of the regression line between SBP and pulse interval) was taken as an overall measure of BRS (ms/mmHg) (21). Total BRS was obtained by averaging the +RR/+SBP and −RR/−SBP regression coefficients. SDs of the recordings were computed for up, down, and total BRS values.

BRS was measured in 69 diabetic participants (51 CAN−, 18 CAN+). We classified participants as having normal or impaired BRS according to published reference values (22). Those with a mean supine BRS value below the 10th centile relative to age-matched healthy control subjects were designated BRS−; BRS+ participants had a mean BRS above the 25th centile. CAN− participants whose BRS fell between the 10th and 25th centiles were excluded from further study.

Diabetic participants were assigned to one of three groups: 1) “normal” autonomic function (CAN− with normal BRS [BRS+]), 2) “mild” autonomic neuropathy (CAN− with impaired BRS [BRS−]), or 3) advanced autonomic neuropathy (CAN+).

Of the 69 subjects with type 1 diabetes, 28 (10 BRS+, 9 BRS−, 9 CAN+) underwent sequential euglycemic-hypoglycemic clamps, along with 8 nondiabetic participants. All refrained from stimulants and strenuous exercise for 12 h beforehand. To reduce the risk of antecedent hypoglycemia (diabetic participants), evening insulin dose was reduced by 20–25% the day before study and participants were instructed to check capillary glucose at bedtime, at 0300, and before breakfast. Symptomatic hypoglycemia during the preceding 3 days or a blood glucose <4 mmol/l precluded individuals from participating on that occasion.

Diabetic participants were admitted at 0800 on the study day, having consumed a light breakfast but omitted their morning insulin dose. On arrival, soluble insulin (Human Actrapid; Novo Nordisk Laboratories, Copenhagen, Denmark) was infused via an 18-g cannula inserted into an antecubital vein to maintain capillary blood glucose between 5 and 10 mmol/l.

Between 1100 and 1130, a retrograde cannula for sampling was inserted into a dorsal hand vein and kept patent with a slow 0.9% saline infusion; this hand rested in a heated chamber at 55°C to arterialize venous blood. Baseline data were collected at least 30 min later. When blood glucose was stable at ∼5 mmol/l (around 1230), insulin infusion rate was increased to 60 mU · min⁻¹ · m⁻² and a 20% glucose solution was administered. Glucose infusion rate was adjusted according to blood glucose measurements obtained every 3–5 min. After a 60-min euglycemic period (5 mmol/l), glucose was allowed to fall over 30 min to a target of 2.5 mmol/l and held at 2.5 mmol/l for an additional 60 min.

Physiological and ECG data were collected at baseline and every 30 min during each glycemic plateau (expressed as 5 mmol/l: E0, E30, and E60; 2.5 mmol/l: H0, H30, and H60). Plasma catecholamines were measured at baseline, E60, H30, and H60, and free insulin was measured at E60 and H60.

Blood glucose was measured by a glucose oxidase method (YSI 2300 Stat Plus analyzer; Yellow Springs Instruments, Yellow Springs, OH). HbA_1c was measured using Primus 385 and 330 instruments (Primus, Kansas City, MO), using a boronate affinity and high-performance liquid chromatography method (23). Plasma potassium, magnesium, calcium, and albumin were measured using a dry slide automated analyser (Vitros 250; Ortho Clinical Diagnostics, Rochester, NY). Samples for catecholamine analysis were taken into chilled tubes that contained 75 μl of EGTA-glutathione, centrifuged promptly, and stored at −80°C until assayed by high-performance liquid chromatography and electrochemical detection (24). Free-insulin samples were centrifuged promptly, and 0.5 ml of the supernatant was added to chilled tubes that contained 0.5 ml of polyethylene glycol buffer. After centrifugation, the resulting supernatant was stored at −80°C until analyzed by radioimmunoassay (Pharmacia & Upjohn Diagnostics AB, Uppsala, Sweden). Inter- and intra-assay coefficients of variation were 9.9 and 8.5%, respectively (epinephrine), 3.9 and 1.2% (norepinephrine), and 5.8 and <5% (insulin).

A custom-built, high-resolution ECG analysis system, acquiring three bipolar orthogonal X, Y, and Z leads, was used for QTc measurement (25). For each lead, signals from sinus beats were amplified, digitized, filtered, and averaged online over a 1-min recording period.

A semiautomated gradient method (25,26) was used, with a tangent automatically applied to the downward slope of the T wave at the point of maximum gradient, and the end of the T wave was taken as the intersection of this tangent with the baseline. The tangent position was verified and adjusted if necessary by a single observer who was blinded to all other data. The values for QT interval and mean associated heart rate were then displayed on the computer screen, and QTc was derived from QT using the modified Bazett formula (27). Finally, the root mean square QTc across the three leads was calculated.

Heart rate, blood pressure, sweating, and tremor responses were determined using previously described methods (28). Briefly, heart rate and blood pressure recordings were made using an automatic oscillometric sphygmomanometer (Accutorr 4; Datascope, Huntingdon, U.K.). Finger tremor was measured using a ring accelerometer placed on the index finger of the outstretched hand over a 1-min period, and sweating was determined using a ventilated chamber evaporimeter placed on the sternum, which records the difference in humidity between inflowing and outflowing air.

Comparisons between variables measured at several time points were made using repeated-measures ANOVA to evaluate between-group differences and within-subject changes over time. In the event of a significant group by time interaction, summary measures were compared by one-way ANOVA followed by Scheffe post hoc test for multiple comparisons. Group comparisons not involving time were also made by one-way ANOVA with Scheffe subanalysis. The relationship between plasma epinephrine and QTc length was explored using Spearman rank correlation. Data are presented as mean ± SE unless otherwise stated, and P < 0.05 was judged significant. Analyses were performed using the statistical package SPSS version 10.0 (SPSS, Chicago, IL).

Of 109 type 1 diabetic subjects initially screened, 9 were classified as CAN+. The diagnosis of CAN was confirmed in all of the nine additional individuals who had known CAN and were also studied. BRS was unequivocally low (below the 10th centile) in 15 of 51 CAN− participants (29%) and fell between the 10th and 25th centiles in 5 CAN− participants, leaving 31 (61%) with “normal” BRS. In contrast, 14 of 18 CAN+ participants (78%) exhibited low BRS, with 3 showing intermediate (10th to 25th centiles) results.

Cardiac autonomic function data relating to the hypoglycemic clamp participants are summarized in Table 1. BRS was above the 25th centile relative to age-matched control subjects in only one CAN+ participant.

CAN+ participants tended to be older than the nondiabetic and CAN− participants. The groups were otherwise well matched (Table 2).

Blood glucose was similar among the groups (Fig. 1). Within each group, plasma free-insulin concentrations were similar at steady-state euglycemia (E60) and hypoglycemia (H60; BRS+, 82.1 ± 5.7 vs. 76.8 ± 5.3 mU/l, P = 0.26; BRS−, 76.7 ± 2.8 vs. 70.7 ± 5.1 mU/l, P = 0.17; CAN+, 71.7 ± 5.6 vs. 77.8 ± 6.2 mU/l, P = 0.48; N, 82.2 ± 5.5 vs. 80.4 ± 7.3 mU/l, P = 0.76). There were no group differences in insulin concentrations at euglycemia (F = 0.96, P = 0.42) or hypoglycemia (F = 0.466, P = 0.71).

Plasma epinephrine concentrations (Fig. 2) were similar at baseline among the groups (F = 0.88, P = 0.46). There was a significant difference between the groups with respect to hypoglycemic epinephrine responses (F = 6.21, P < 0.001). Nondiabetic participants exhibited greater epinephrine responses (peak increase from baseline 5.50 ± 0.82 nmol/l) than the diabetic groups (BRS+, 2.93 ± 0.81 nmol/l, P = 0.065 vs. control subjects; BRS−, 2.37 ± 0.43 nmol/l, P = 0.02 vs. control subjects; CAN+, 1.31 ± 0.32 nmol/l, P = 0.001 vs. control subjects). Epinephrine responses tended to be blunted in CAN+ participants compared with the other diabetic groups, but this was not statistically significant (vs. BRS+, P = 0.35; vs. BRS−, P = 0.71).

Basal plasma norepinephrine concentrations were similar between the groups (F = 0.32, P = 0.81). The groups differed with respect to hypoglycemic norepinephrine responses (F = 2.48, P = 0.038), with nondiabetic participants tending to show greater norepinephrine increases (peak rise above baseline 1.42 ± 0.36 nmol/l) than the diabetic groups. However, there were no significant differences between individual groups (BRS+, 0.67 ± 0.19, P = 0.3 vs. control subjects; BRS−, 0.68 ± 0.23, P = 0.32 vs. control subjects; CAN+, 0.80 ± 0.30, P = 0.49 vs. control subjects).

Plasma potassium (Fig. 3) was similar at baseline between the groups (F = 1.612, P = 0.21). By the end of euglycemia, potassium was significantly lower in all groups (P < 0.001); the euglycemic decrement in potassium was similar between the groups (F = 0.806, P = 0.59).

During hypoglycemia, the decrement in potassium differed significantly between the groups (F = 2.543, P = 0.029). This was largely accounted for by a twofold greater potassium fall in the nondiabetic participants (0.59 ± 0.06 mmol/l) when compared with the diabetic groups (BRS+, 0.29 ± 0.07 mmol/l, P = 0.047 vs. control subjects; BRS−, 0.30 ± 0.09 mmol/l, P = 0.08 vs. control subjects; CAN+, 0.33 ± 0.06 mmol/l, P = 0.12 vs. control subjects). The diabetic groups exhibited similar potassium falls (BRS+ vs. BRS−, P = 0.998; BRS+ vs. CAN+, P = 0.98; BRS− vs. CAN+, P = 0.998). Plasma calcium (corrected for albumin) and magnesium concentrations were similar among the groups and remained unchanged during the studies (data not shown).

At baseline, heart rate differed between the groups (F = 7.42, P = 0.001), being greater in CAN+ participants than in nondiabetic (P = 0.015) and BRS+ (P = 0.001) participants; there were otherwise no significant group differences (Table 3).

During hypoglycemia, heart rate increased in the control (peak increase from baseline 12 ± 4.3 bpm; P = 0.02), BRS+ (12 ± 2.3 bpm; P < 0.001), and BRS− (10 ± 1.8 bpm, P < 0.001) groups; by contrast, CAN+ participants did not show a significant rise in heart rate (peak increase 5 ± 3.3 bpm; P = 0.36). However, hypoglycemic heart rate changes did not differ significantly between the groups (F = 0.85, P = 0.56).

SBP was similar at baseline among the groups (F = 1.16, P = 0.34). SBP tended to increase during hypoglycemia in all groups, although this was significant only for the nondiabetic (peak increase 13 ± 2.2 mmHg; P = 0.001) and BRS+ (10 ± 1.9 mmHg; P = 0.02) groups. Overall, there was no significant difference between the groups with respect to change in SBP during hypoglycemia (F = 0.637, P = 0.82).

Baseline diastolic blood pressure (DBP) was similar among the groups (F = 0.79, P = 0.51). Although DBP tended to fall during hypoglycemia in the BRS+ and control groups, the change in DBP was not significantly different between the groups (F = 1.069, P = 0.39).

At baseline, finger tremor was similar among all groups (F = 1.80, P = 0.17). There was a significant difference in hypoglycemic tremor responses between the groups (F = 2.016, P = 0.047). Although tremor increased during hypoglycemia in all groups, this tended to be blunted in CAN+ participants (peak incremental tremor 0.049 ± 0.010 g · m⁻² · min⁻¹), particularly in comparison with the control (0.163 ± 0.044, P = 0.06 vs. CAN+) and BRS+ groups (0.142 ± 0.023, P = 0.12 vs. CAN+).

Sweating increased across all groups during hypoglycemia. The magnitude of the sweating response did not differ between the groups (F = 0.59, P = 0.70).

Baseline QTc was similar between the groups (control, 378 ± 4 ms; BRS+, 377 ± 9 ms; BRS−, 380 ± 8 ms; CAN+, 383 ± 6 ms; F = 0.124, P = 0.95; Fig. 4). During euglycemia, there was a small (10–20 ms) but significant increase in QTc across all groups; these increases were similar between the groups (F = 0.94, P = 0.49).

During hypoglycemia, QTc lengthened significantly among all four groups (P < 0.001). However, there was considerable interindividual variability. BRS+ participants exhibited peak QTc increases above baseline between 13 and 90 ms (mean, 61 ms). QTc lengthened by 4 to 84 ms (mean, 41 ms) in BRS− participants, by −9 to 49 ms (mean, 25 ms) in CAN+ participants, and by 28 to 106 ms (mean, 61 ms) in the nondiabetic participants.

The groups differed with respect to the magnitude of hypoglycemic QTc lengthening (F = 3.442, P = 0.007). When expressed as peak increase above baseline, CAN+ participants showed blunted QTc increases in comparison with both the control (P = 0.038) and BRS+ (P = 0.025) groups. There were no significant differences among the control, BRS+, and BRS− groups.

Maximum QTc interval during hypoglycemia did not differ significantly between the groups (F = 1.731, P = 0.18), although there was a trend toward lower peak QTc in CAN (408 ± 10 ms) and BRS− (421 ± 11 ms) participants when compared with BRS+ (439 ± 13 ms) and control (439 ± 10 ms) groups.

There was a significant positive correlation (Fig. 5) between peak incremental plasma epinephrine response and the magnitude of QTc lengthening (r_s = 0.43, P = 0.009).

Our data show that type 1 diabetic patients with CAN (as evidenced by either abnormal conventional cardiovascular tests or impaired BRS) do not exhibit greater hypoglycemic QTc lengthening than those without CAN. Indeed, CAN+ participants tended to show the smallest QTc increases. Thus, our findings refute the hypothesis that autonomic neuropathy is an important factor increasing the risk of QTc lengthening during hypoglycemia in young adults with type 1 diabetes. This suggests that autonomic neuropathy is not an important contributory factor to sudden death from hypoglycemia in young people with diabetes.

Although the mechanisms underlying these deaths are not fully understood, there is strong circumstantial evidence implicating hypoglycemia and associated cardiac arrhythmia (1,6). In Tattersall and Gill’s series, most deaths occurred at night, and 14 patients had experienced significant nocturnal hypoglycemia during the preceding months. Generally, the bedclothes were not disturbed, and no anatomical lesion (e.g., neuropathological evidence of hypoglycemia) was found at autopsy. Therefore, it seems unlikely that the deaths were related to hypoglycemic brain damage. The evidence points to a sudden cardiac or respiratory arrest as the terminal event. It is noteworthy that, although apparently uncommon, both supraventricular (29,30) and ventricular (31–33) arrhythmias have been reported during hypoglycemia.

Epinephrine infusion lengthens QTc interval in normal participants (34), predominantly through a β-adrenoceptor-mediated process, whereby the cardiac action potential is prolonged by delayed inactivation of inward calcium currents (35). In this study, we observed a relationship, albeit weak, between the magnitude of the hypoglycemic epinephrine response and the degree of QTc lengthening. These observations suggest that hypoglycemic QTc lengthening is predominantly mediated through the sympatho-adrenal response, although there seems to be wide interindividual variability in myocardial epinephrine sensitivity (7,36). Type 1 diabetic patients with autonomic neuropathy tend to exhibit impaired plasma epinephrine responses to hypoglycemia in comparison with their non-neuropathic counterparts (37,38), and this seems the most likely explanation for our findings.

We did not restrict recruitment according to HbA_1c or match it over the three diabetic groups because of the limited number of participants. However, we made some attempt to ensure that HbA_1c in the three diabetic groups was similar because glycemic control modulates the sympatho-adrenal response and hence could influence the degree of QT lengthening. Glycemic control tended to be poor in the diabetic participants, perhaps not surprising in those with diabetic neuropathy. It is possible that both the hypoglycemic sympatho-adrenal response and the magnitude of QT lengthening would have been smaller if we had studied participants with better glycemic control. However, because HbA_1c was similar across the groups (with a trend to higher values in CAN+ participants), we do not believe that the level of glycemic control influenced our findings.

Our laboratory-based study inevitably has limitations when considering its relevance to the clinical situation. Because the depth of experimental hypoglycemia must be limited for ethical reasons, we have used QT interval as a surrogate marker of abnormal cardiac repolarization. Thus, the influence of autonomic dysfunction on cardiac electrical pathophysiology during clinical hypoglycemic episodes must remain somewhat uncertain, as is the influence of QT duration on arrhythmia risk. However, there is good evidence in families with congenital LQTS that the risk of ventricular arrhythmia is related to QT interval (39), and this is also the case in acquired QT lengthening caused by medication (40) or after myocardial infarction (41).

Experimental hypoglycemia is induced using greater circulating insulin concentrations than those generally observed clinically, and this tends to produce greater potassium falls than those that occur during clinical episodes. This might be expected to produce greater QTc lengthening. However, we recently showed that, in comparison with the rise in plasma epinephrine, an associated fall in potassium has a relatively minor effect on hypoglycemic QTc prolongation (8). Similarly, preventing a fall in plasma potassium does not prevent QTc lengthening induced by epinephrine infusion (36).

The clinical relevance of our data in relation to the risks of intensive insulin therapy is also uncertain. Nocturnal hypoglycemia is common in both adults and children with type 1 diabetes whether they are undertaking intensive insulin therapy or not, although it is likely to occur more frequently in those who aim for tight glucose targets. Thordarson and Sovik (4) noted a rising incidence of hypoglycemic death during the 1980s and attributed this in part to the increasing use of intensive insulin therapy. Conversely, tight glycemic control is associated with diminished sympatho-adrenal responses, which, although increasing the risk of severe hypoglycemia, might actually reduce the degree of abnormal cardiac repolarization and any associated arrhythmic risk. Nevertheless, we advocate the use of approaches to reduce nocturnal hypoglycemia, such as insulin analogs or insulin pump therapy, in those who are at particular risk during intensive insulin therapy.

In conclusion, we have demonstrated that the presence of autonomic neuropathy does not result in greater QTc lengthening during moderate experimental hypoglycemia in type 1 diabetic participants. Rather, our data support the view that patients with the strongest hypoglycemic sympatho-adrenal responses, typically those with relatively short disease duration and poor glycemic control, may be most at risk for significant QTc lengthening and perhaps sudden death. Other factors, including mutations or polymorphisms in candidate genes, also need to be explored in future work because this might identify those who are at particular risk of this rare but devastating condition.

We acknowledge the support of the Juvenile Diabetes Foundation International and the Novo Nordisk U.K. Research Foundation. We are grateful to the diabetes research nurses Mishell Cunningham and Sue Hudson for assistance and to David Forster and Sujata Dissanayake (University of Nottingham) and Fiona Harding (University of Newcastle) for performing catecholamine and free-insulin assays, respectively.