Sleep-Related Hypoglycemia-Associated Autonomic Failure in Type 1 Diabetes: Reduced Awakening From Sleep During Hypoglycemia

We studied eight patients with type 1 diabetes and eight nondiabetic subjects matched for sex, age, and BMI; each subject gave their written consent to participate. The study protocol was approved by the Washington University Medical Center Human Studies Committee and conducted at the Washington University General Clinical Research Center. The characteristics of the participants are listed in Table 1. All had normal hematocrits, serum creatinine concentrations, and electrocardiograms, and none had a history of central nervous disease or cardiac arrhythmias. The diabetic patients had no evidence of classical diabetic autonomic neuropathy (as evidenced by a negative medical history and physical examination including normal electrocardiographic RR variation during deep breathing and the absence of orthostatic hypotension) and no active diabetic retinopathy. They had not experienced an episode of severe iatrogenic hypoglycemia (requiring the assistance of another individual) over the 3 months before the study and had no self-monitored blood glucose levels (measured at least three times a day) <72 mg/dl (<4.0 mmol/l) during the week before the study. (If the latter occurred the study was postponed until that criterion was met.)

Studies were performed after a fast of at least 10 h. Patients took their last prestudy dose of NPH insulin at least 12 h before each study or their last dose of glargine insulin at least 24 h before each study. The patients were admitted to the research center 12 h before each study. Their diabetes was managed with variable dosages of intravenous regular insulin to hold plasma glucose concentrations in the 80–120 mg/dl range over the 10 h before each study. Nondiabetic subjects were admitted 1 h before each study. Studies in any one given subject were separated by at least 2 weeks.

Intravenous lines were inserted into an antecubital vein (for insulin and glucose infusions) and into a hand vein, with that hand kept in a 55°C Plexiglas box (for arterialized venous blood sampling) at −60 to −30 min. The subjects remained supine from −30 through 300 min. Hyperinsulinemic (2.0 mU · kg⁻¹ · min⁻¹) stepped hypoglycemic clamps (hourly steps at 85, 75, 65, 55, and 45 mg/dl) (11) were performed on three occasions in random, computer-generated sequence in both groups. These were performed in the morning (0730–1230) and twice during the night (2100–0200), once with the subject awake and once with the subject allowed to sleep starting at 1130 h. Hyperinsulinemic (2.0 mU · kg⁻¹ · min⁻¹), euglycemic (85 mg/dl) clamps (11) were performed in the morning on a fourth occasion in the nondiabetic subjects. These clamps were accomplished by variable intravenous infusions of 20% dextrose based on arterialized venous plasma glucose measurements every 5 min at bedside (Yellow Springs Analyzer 2; Yellow Springs Instruments, Yellow Springs, OH). Arterialized venous blood samples for the analytes listed below were drawn at −15 and 0 min and then every 30 min through 300 min. Blood pressures and heart rates (Propaq Encore; Protocol Systems, Beverton, OR) were recorded at −15 and 0 min and every 30 min through 300 min. The electrocardiogram was monitored throughout. Symptoms of hypoglycemia were quantified, also at 30-min intervals, by asking the subjects to score on a scale of zero (none) to six (severe) each of 12 symptoms: 6 neurogenic symptoms (adrenergic: heart pounding, shaky/tremulous, and nervous/anxious; cholinergic: sweaty, hungry, and tingling) and 6 neuroglycopenic symptoms (difficulty thinking/confused, tired/drowsy, weak, warm, faint, and dizzy) based on our published data (12). Symptoms were not, of course, assessed during sleep.

Polysomnographic recordings for determination of sleep stages were made using 14 electrodes (including left and right electrooculograms, electromyograms, and four scalp electroencephalograms). Staging was hand scored by a single technician using the Rechtschaffen and Kales (13) standardized scoring for each 30-s epoch. Recordings were performed from 0000 through 0200 (i.e., at nominal plasma glucose steps of 55 and 45 mg/dl) during all 16 sleep studies. However, data were available for analysis only for the eight nondiabetic subjects and four of the eight diabetic patients.

Plasma glucose was measured with a glucose oxidase method (Yellow Springs Analyzer 2). Plasma insulin (14), C-peptide (14), glucagon (15), pancreatic polypeptide (16), growth hormone (17), and cortisol (18) were measured with radioimmunoassays. Plasma epinephrine and norepinephrine were measured with a single isotope derivative (radioenzymatic) method (19). Serum nonesterified fatty acid (20) and blood β-hydroxybutyrate (21), lactate (22), and alanine (23) were measured with enzymatic methods.

Data in this manuscript are expressed as means ± SE, except where the standard deviation is specified. Time- and condition-related data were analyzed by general linear model repeated measures ANOVA after adjustment for any differences at baseline. Sleep data were analyzed by t test. P < 0.05 was considered to indicate statistically significant differences.

Plasma glucose concentrations were clamped at target levels in both the nondiabetic and diabetic subjects (Fig. 1). Plasma insulin concentrations were raised comparably under all conditions in both groups (Fig. 2). Plasma C-peptide concentrations declined during hyperinsulinemic euglycemia and to a greater extent during hypoglycemia in the nondiabetic subjects (Fig. 2); plasma C-peptide was undetectable in the diabetic subjects (Fig. 2).

During the night asleep studies, sleep efficiency (the percent of time asleep) was significantly higher in the diabetic than in the nondiabetic subjects during the 55- and 45-mg/dl hypoglycemic steps (1200–0200) (P = 0.0227), particularly during the 45-mg/dl hypoglycemic step (0100–0200) (P = 0.0190) (Table 2, Fig. 3). Similarly, the time in sleep stages I through IV (i.e., non-REM sleep) was greater in the diabetic subjects (P = 0.0272 and 0.0191, respectively) (Table 2). During the final 30 min of the 45-mg/dl hypoglycemic step, sleep efficiency was 77 ± 18% in the diabetic subjects patients and 26 ± 8% in the control subjects (P = 0.0109) (Fig. 3).

In both the nondiabetic and diabetic subjects, increments in plasma epinephrine (Fig. 4) and norepinephrine (Fig. 5) concentrations during hypoglycemia were similar in the morning awake and night awake states. In the nondiabetic subjects, the plasma epinephrine response to hypoglycemia at night was not reduced significantly during sleep, although the epinephrine levels during sleep appeared to be lower than those when the subjects were awake at the 65- and 55-mg/dl glucose steps (Fig. 4). In the diabetic subjects, the plasma epinephrine response to hypoglycemia at night was reduced (P = 0.0010) during sleep compared with when the subjects were awake (Fig. 4). In the nondiabetic subjects, the plasma norepinephrine response to hypoglycemia at night was not reduced significantly during sleep, although the norepinephrine levels during sleep appeared to be lower than those when the subjects were awake at the 65- and 55-mg/dl glucose steps (Fig. 5). In the diabetic subjects, the plasma norepinephrine response to hypoglycemia at night was not significantly reduced (P = 0.0838) during sleep compared with when the subjects were awake, although there was no apparent response during sleep (Fig. 5).

In both the nondiabetic and diabetic subjects, increments in neurogenic (autonomic) (Table 3) and neuroglycopenic symptom scores (Table 4) during hypoglycemia were similar during the morning awake and night awake states. Symptoms were not assessed during sleep.

In the nondiabetic subjects, increments in plasma glucagon concentrations during hypoglycemia were similar in the morning awake and night awake states (Fig. 6). The plasma glucagon response to hypoglycemia at night was not reduced during sleep (Fig. 6). In diabetic subjects, there were no glucagon responses to hypoglycemia under any of the study conditions (Fig. 6).

In the nondiabetic subjects, increments in plasma pancreatic polypeptide concentrations during hypoglycemia were reduced (P = 0.0003) during the night awake compared with the morning awake state (Fig. 7). There was no effect of sleep on the pancreatic polypeptide response (Fig. 7). In diabetic subjects, the small pancreatic polypeptide responses to hypoglycemia were similar in the morning and night awake states (Fig. 7), but reduced (P = 0.0034) in the night asleep state (Fig. 7).

Plasma cortisol levels at baseline were lower at night than in the morning in both groups (Fig. 8). In the nondiabetic subjects, increments in plasma cortisol concentrations during hypoglycemia, after being adjusted for baseline differences, were greater (P = 0.0005) in the night awake than in the morning awake state (Fig. 8). That was also the case (P = 0.0384) in diabetic subjects (Fig. 8). There was no significant effect of sleep on the cortisol response to hypoglycemia in the nondiabetic subjects (Fig. 8). In diabetic subjects, the plasma cortisol response to hypoglycemia at night was reduced (P = 0.0069) during sleep compared with when the patients were awake (Fig. 8).

In the nondiabetic subjects, increments in plasma growth hormone concentrations during hypoglycemia were similar in the morning and night awake states (Fig. 9). There was no significant effect of sleep on the plasma growth hormone response to hypoglycemia (Fig. 9). In diabetic subjects, the increments in plasma growth hormone during hypoglycemia were greater (P = 0.0197) in the night than in the morning awake state (Fig. 9). The growth hormone response to hypoglycemia was not reduced during sleep (Fig. 9).

The glucose infusion rates required to maintain the hypoglycemic steps were higher in the morning awake than in the night awake state in both nondiabetic and diabetic subjects (P < 0.0001 and P = 0.0156, respectively). In the nondiabetic subjects, the glucose infusion rate required to maintain the lower glucose steps at night was higher (P = 0.0022) when they were asleep than when they were awake (Fig. 10). In diabetic subjects, the required glucose infusion rates were similar when the subjects were awake and asleep (Fig. 10).

In the nondiabetic subjects, blood lactate (Table 5), serum nonesterified fatty acid (Table 6), blood β-hydroxybutyrate, and blood alanine (data not shown) concentrations were similar under all hypoglycemic conditions. In the diabetic subjects, increments in blood lactate during hypoglycemia were reduced at night when the subjects were asleep (P = 0.013) compared with in the morning when they were awake and tended to be reduced at night when the subjects were awake (P = 0.0976) (Table 5). In addition, lactate levels were slightly higher during the asleep compared with the awake study (P = 0.0038) (Table 5).

Heart rates were similar during hypoglycemia under all study conditions in both groups (data not shown). Similarly, there were no differences in systolic or diastolic blood pressures during hypoglycemia (data not shown).

These data document markedly reduced awakening during hypoglycemia in patients with type 1 diabetes, a novel finding that could be attributed to diabetic patients’ reduced sympathoadrenal responses to hypoglycemia during sleep, which were also documented by the present data. Sleep efficiency (the percent of time asleep) was threefold greater in the diabetic subjects than in the matched nondiabetic control subjects late in the 45-mg/dl hypoglycemic steps; the diabetic subjects were asleep ∼75% of the time, whereas the control subjects were awake ∼75% of the time. Comparable data are not available from the study of Jones et al. (10), as they did not study nondiabetic control subjects at night while awake.

These data also document a reduced plasma epinephrine, and perhaps norepinephrine, response to hypoglycemia during sleep in adults with uncomplicated type 1 diabetes, findings similar to those of Jones et al. (10) in adolescents with type 1 diabetes. A reduced plasma pancreatic polypeptide response to hypoglycemia in type 1 diabetic patients is also documented. Thus, taken together, the data indicate that the autonomic nervous system response to a given level of hypoglycemia is reduced during sleep in patients with type 1 diabetes. Because arousal is a recognized adrenergic manifestation of the sympathoadrenal response to hypoglycemia (e.g., the symptom nervous/anxious can be blocked by administration of catecholamine antagonists) (12), it is reasonable to attribute the reduced awakening from sleep in the patients to their reduced sympathoadrenal responses during sleep. Our data and those of Jones et al. (10) also document a reduced adrenocortical cortisol response to hypoglycemia during sleep in the patients. However, in contrast to the findings of Jones et al. (10), the present data do not demonstrate significantly reduced plasma epinephrine, norepinephrine, or cortisol (or pancreatic polypeptide) responses to hypoglycemia during the sleep study in nondiabetic individuals.

Under all study conditions—morning or night, awake or asleep—the diabetic subjects compared with matched nondiabetic control subjects exhibited absent glucagon responses to hypoglycemia and reduced autonomic—adrenomedullary (plasma epinephrine and norepinephrine), sympathetic neural (neurogenic symptoms and plasma norepinephrine), and parasympathetic neural (plasma pancreatic polypeptide)—responses to a given level of hypoglycemia, as expected (1). Although the loss of glucagon response in the diabetic subjects was absolute, autonomic responses could be elicited, but the glycemic thresholds for those responses were shifted to lower plasma glucose concentrations, again as expected (1). The absent glucagon and reduced epinephrine responses were reflected biologically by the higher glucose infusion rates required to maintain the lower hypoglycemic steps. The reduced sympathoadrenal (sympathetic neural and adrenomedullary) responses were reflected by the reduced neurogenic (autonomic) symptom responses to a given level of hypoglycemia in the diabetic compared with the nondiabetic control subjects.

In general, subjects’ responses to hypoglycemia were remarkably similar in the morning and night awake states. Sympathoadrenal responses and the resultant neurogenic symptom responses to hypoglycemia were similar in the morning and night awake states in nondiabetic control subjects and, despite differences in the absolute values, in the diabetic subjects. However, the parasympathetic neural (plasma pancreatic polypeptide) response to hypoglycemia was reduced at night in the awake nondiabetic subjects; the small pancreatic polypeptide response in the awake diabetic subjects was not reduced significantly. The glucagon response to hypoglycemia, like the epinephrine and norepinephrine responses, was similar in the morning and at night in the awake nondiabetic subjects; there were no glucagon responses in the diabetic subjects, as expected. Baseline plasma cortisol concentrations were substantially lower at night, also as expected. However, after baseline adjustment, the cortisol responses to hypoglycemia were enhanced during the night in both the awake control subjects and the awake diabetic subjects. Notably, the glucose infusion rates required to maintain the hyperinsulinemic glucose clamps were lower throughout (i.e., during euglycemia as well as hypoglycemia) at night compared with in the morning in both groups. This finding implies relative insulin resistance at night, a finding at variance with an earlier report (24). Thus, aside from the parasympathetic neural response, these data indicate that there is no diurnal variation per se in the physiological responses to hypoglycemia. Because the parasympathetic response has no known direct role in defense against developing hypoglycemia (1), it follows that there is no diurnal variation, independent of sleep, in the physiology of glucose counterregulation.

Taken at face value, the impact of sleep appeared to differ in the nondiabetic control and diabetic subjects. As noted earlier, the adrenomedullary (plasma epinephrine) and parasympathetic neural (plasma pancreatic polypeptide) responses to hypoglycemia of the diabetic subjects were reduced significantly, and the plasma norepinephrine response to hypoglycemia appeared to be reduced, during the night asleep compared with the night awake state. In contrast, during the nocturnal studies in the nondiabetic control subjects, the plasma epinephrine, norepinephrine, and pancreatic polypeptide responses to hypoglycemia were statistically similar when the subjects were awake and when they were ostensibly asleep. However, the epinephrine, norepinephrine, and pancreatic polypeptide responses appeared to be reduced at the 65- and 55-mg/dl hypoglycemic steps in the sleep compared with the awake nocturnal study. Notably, the significantly higher glucose infusion rates required to maintain the lower hypoglycemic steps during sleep suggest a biologically important difference. It is likely that the absence of an overall difference between the curves for the epinephrine, norepinephrine, and pancreatic polypeptide (and cortisol) responses was the result of the lower sleep efficiency in the nondiabetic control subjects. They were asleep only ∼25% of the time late in the lowest hypoglycemic step. Thus, the apparent lack of an impact of sleep in the nondiabetic control subjects may well have been the result of the fact that, for the most part, they were not asleep.

The significantly higher glucose infusion rates required to maintain the lower hypoglycemic steps during sleep in the nondiabetic subjects suggest a biological impact of our interpretation that their epinephrine responses were reduced earlier during the hypoglycemic clamps under that condition. The required glucose infusion rates were higher during hypoglycemia in the diabetic subjects, who had absent glucagon and reduced epinephrine responses to hypoglycemia, than in the nondiabetic control subjects, as expected. However, there was no difference in the glucose infusion rates required to maintain the lower hypoglycemic steps in the diabetic subjects during sleep, despite the subjects’ further reduced epinephrine responses. Thus it appears that this measure was not sufficiently sensitive to document the anticipated effect in the diabetic subjects, who had substantially impaired glucose counterregulation under all three study conditions.

The reduced cortisol response to hypoglycemia during sleep in diabetic subjects is relevant to the suggestion that it is the cortisol response to prior hypoglycemia that mediates reduced autonomic (including epinephrine) and symptomatic responses to subsequent hypoglycemia (25,26), that is, that cortisol is the mediator of hypoglycemia-associated autonomic failure in diabetes (1,6). Nocturnal hypoglycemia reduces the autonomic (including epinephrine) and symptomatic responses to hypoglycemia the next morning in patients with type 1 diabetes (27). To the extent that those patients slept during hypoglycemia, the present data suggest that the cortisol levels were not elevated during the nocturnal hypoglycemia. If so, the reduced responses to hypoglycemia the next morning cannot be attributed to cortisol elevations during the earlier nocturnal hypoglycemia.

The finding of a reduced cortisol response to hypoglycemia during sleep in diabetic subjects may also be relevant to the fact that Bendtson et al. (9) did not find reduced sympathoadrenal responses to hypoglycemia in their nocturnal study. Given their brisk cortisol responses, it is likely that the patients studied were to a large extent awake rather than asleep.

We conclude that autonomic responses to hypoglycemia are further reduced during sleep in patients with type 1 diabetes. Probably because of their reduced sympathoadrenal responses, these patients are substantially less likely to be awakened by hypoglycemia. Thus sleep should be added to recent antecedent hypoglycemia as a cause of hypoglycemia-associated autonomic failure in diabetes (1,6). This sleep-related hypoglycemia-associated autonomic failure, which impairs both physiological and ultimately (because of the lack of awakening) behavioral defenses against developing hypoglycemia, in the context of imperfect insulin replacement, likely explains the high frequency of nocturnal hypoglycemia in type 1 diabetes.

This work was supported, in part, by a Clinical Research Grant from the American Diabetes Association and by U.S. Public Health Service Grants R37-DK-27085, M01-RR-00036, and P60-DK-20579.

We acknowledge the technical assistance of Krishan Jethi, Cornell Blake, Joy Brothers, Michael Morris, and Sharon O’Neill; the assistance of the nursing staff at Washington University General Clinical Research Center; and the assistance of Karen Muehlhauser in the preparation of this manuscript.