Cortisol Elevations Comparable to Those That Occur During Hypoglycemia Do Not Cause Hypoglycemia-Associated Autonomic Failure

A total of 12 healthy young adults (5 women and 7 men) aged 26 ± 7 years (mean ± SD) and with a BMI of 23.0 ± 2.6 kg/m² gave their written consent to participate in this study, which was approved by the Washington University Medical Center Human Studies Committee and was conducted at the Washington University General Clinical Research Center. They were in good health as assessed by medical history and physical examination and had normal fasting plasma glucose concentrations, hematocrits, and electrocardiograms. Aside from estrogen-progestin oral contraceptives, used by two of the women, they took no medications.

Each subject participated in two studies, separated by at least 2 weeks, that involved 2 consecutive study days. On day 1, in random sequence, they received either saline or cortisol (SoluCortef; Pharmacia and Upjohn, Kalamazoo, MI) infusions from 0930 to 1200 and from 1330 to 1600. The cortisol infusion rate (means ± SD) was 1.3 ± 0.2 μg · kg⁻¹ · min⁻¹. A dose of 1.4 μg · kg⁻¹ · min⁻¹ was used in seven subjects, 1.2 μg · kg⁻¹ · min⁻¹ in three subjects, and 1.0 μg · kg⁻¹ · min⁻¹ in two subjects. The initial dose used (1.4 μg · kg⁻¹ · min⁻¹) was calculated from the dose of 2.0 μg · kg⁻¹ · min⁻¹ used by Davis et al. (5) in an attempt to produce peak plasma cortisol concentrations of ∼26 μg/dl rather than those of ∼32 μg/dl produced by Davis et al. (5). Because our target was exceeded in the first seven subjects, lower doses were used in the remaining five subjects.

On day 2 on both occasions, hyperinsulinemic-stepped hypoglycemic clamps (18) were performed after an overnight fast. Insulin was infused in a dose of 2.0 mU · kg⁻¹ · min⁻¹ for 5 h, and variable infusions of 20% dextrose, based on plasma glucose measurements with a glucose oxidase method (Yellow Springs Analyzer 2; Yellow Springs Instruments, Yellow Springs, OH) every 5 min, were used to clamp plasma glucose concentrations at ∼85, 75, 65, 55, and 45 mg/dl in hourly steps. Insulin and glucose were infused through a catheter in an antecubital vein. Arterialized venous samples (for the analytes listed below) were drawn through an indwelling line in a dorsal hand vein, with that hand kept in a ∼55°C plexiglass box. Samples were drawn at −30, −15, 0, 30, 60, 90, 120, 150, 180, 210, 240, 270, and 300 min. At the same time points, the subjects scored (from 0 [none] to 6 [severe]) each of 12 symptoms: 6 neurogenic (autonomic) 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 (19). Heart rates and blood pressures (Propaq Encore; Protocol Systems, Beaverton, OR) were recorded serially, and the electrocardiogram was monitored throughout.

Plasma insulin (20), C-peptide (20), pancreatic polypeptide (21), glucagon (22), growth hormone (23), and cortisol (24) concentrations were measured with radioimmunoassays, plasma epinephrine, and norepinephrine concentrations with a single isotope-derivative (radioenzymatic) method (25). Serum nonesterified fatty acid (26), blood lactate (27), β-hydroxybutyrate (28), and alanine (29) concentrations were measured with enzymatic methods.

Data are reported as the means ± SE, except where the SD is specified. The data were analyzed by repeated-measures ANOVA. P values <0.05 were considered to indicate statistically significant differences.

On day 1, 2.5-h cortisol infusions raised plasma cortisol concentrations to 33.8 ± 3.0 μg/dl at 1200 and 37.0 ± 3.6 μg/dl at 1600 compared with corresponding levels of 9.6 ± 1.6 and 9.1 ± 1.1 μg/dl, respectively, during saline infusions (Fig. 1). When the data from the two subjects taking oral contraceptives were excluded, the peak plasma cortisol concentrations were 30.9 ± 2.2 μg/dl at 1200 and 33.7 ± 3.2 μg/dl at 1600 on the cortisol-infusion days.

Plasma glucose targets were approximated and glucose concentrations were similar on the day after saline infusion and the day after cortisol infusion (Fig. 2). Plasma insulin and C-peptide concentrations (Fig. 3) and the glucose infusion rates required to maintain the glucose steps (Fig. 4) were similar on both occasions.

Plasma epinephrine (Fig. 5) and norepinephrine (Fig. 6) responses to hypoglycemia were not reduced on the day after cortisol infusion. Neurogenic (Fig. 7) and neuroglycopenic (Fig. 8) symptom scores were also not reduced on the day after cortisol infusion; indeed, they were increased (P = 0.0015 and 0.0104, respectively). Plasma pancreatic polypeptide, glucagon, growth hormone, and cortisol responses to hypoglycemia were unaltered on the day after cortisol infusion (Table 1). Serum nonesterified fatty acid, blood lactate, β-hydroxybutyrate, and alanine concentrations were similar on both occasions (Table 2). Similarly, heart rates, systolic blood pressures, and diastolic blood pressures did not differ (Table 3).

These data do not support the suggestion that the cortisol response to antecedent iatrogenic hypoglycemia mediates HAAF in diabetes (5–10). Compared with intravenous infusions of saline, intravenous infusions of cortisol that raised plasma cortisol concentrations during euglycemia to (and indeed above) levels that occur during hypoglycemia did not reduce the sympathoadrenal (plasma epinephrine and norepinephrine) or neurogenic (autonomic) symptom responses to hypoglycemia the following day. Thus, these physiological cortisol elevations did not reproduce the key features of HAAF in diabetes (1–4).

Using a generically similar experimental design—interventions on day 1 and quantitation of the responses during hyperinsulinemic-hypoglycemic clamps on day 2—antecedent hypoglycemia (rather than cortisol elevations in the absence of hypoglycemia) has consistently been found to reduce autonomic neuroendocrine and neurogenic symptom responses to a given level of subsequent hypoglycemia in healthy subjects (17) and patients with type 1 diabetes (2) as well as those with advanced type 2 diabetes (3). In patients with type 1 diabetes, recent antecedent hypoglycemia has been shown to shift glycemic thresholds for autonomic, neurogenic symptom, and cognitive dysfunction responses to subsequent hypoglycemia to lower plasma glucose concentrations (2,30), impair glycemic defense against hyperinsulinemia (2), and reduce detection of hypoglycemia in the clinical setting (31). Given the present findings, one cannot attribute these effects of antecedent hypoglycemia to the cortisol response to that hypoglycemia.

We infused lower doses of cortisol on day 1 than Davis et al. (5) because we intended to produce mean cortisol elevations comparable to those that occur during hypoglycemia (∼26 μg/dl) (2,6,8,10). They infused cortisol intravenously in a dose of 2.0 μg · kg⁻¹ · min⁻¹ and produced mean peak plasma cortisol concentrations of ∼32 μg/dl (5). Based on those data, we started with a cortisol infusion dose of 1.4 μg · kg⁻¹ · min⁻¹ (n = 7) and then doses of 1.2 μg · kg⁻¹ · min⁻¹ (n = 3) and 1.0 μg · kg⁻¹ · min⁻¹ (n = 2); the cortisol dose (mean ± SD) was 1.3 ± 0.2 μg · kg⁻¹ · min⁻¹. Despite these lower cortisol doses, peak cortisol concentrations (mean ± SE) were 33.8 ± 3.0 μg/dl at the end of the first infusion and 37.0 ± 3.6 μg/dl at the end of the second infusion. If the data from two women using estrogen-containing oral contraceptives (which raise corticosteroid binding globulin and thus total plasma cortisol concentrations) were excluded, the corresponding mean peak cortisol concentrations were 30.9 ± 2.2 and 33.7 ± 3.2 μg/dl, respectively. Despite these somewhat supraphysiological (with respect to hypoglycemia) cortisol elevations on day 1, sympathoadrenal and neurogenic symptom responses to hypoglycemia were not reduced on day 2.

We used hyperinsulinemic-stepped hypoglycemic clamps, with hourly steps of ∼85, 75, 65, 55, and 45 μg/dl, to assess autonomic neuroendocrine and neurogenic symptom responses to hypoglycemia on day 2. Davis et al. (5) used hyperinsulinemic-single-step (to ∼50 μg/dl for 2 h) hypoglycemic clamps and generally analyzed responses during the final 30 min. Thus, the duration of plasma glucose levels below the glycemic thresholds for glucose counterregulatory (including autonomic neuroendocrine) responses (rev. in 30) was nearly 3 h in the present study and nearly 2 h in the study of Davis et al. (5). Therefore, the differences in the findings cannot be attributed to a shorter duration, or lesser magnitude, of hypoglycemia in the present study. Furthermore, the differences cannot be explained by the fact that we analyzed absolute plasma epinephrine and norepinephrine and neurogenic symptom score values rather than increments from baseline, because there were no differences in the baseline values for these parameters under the two study conditions. Cortisol was infused during ambient euglycemia, rather than during hyperinsulinemic clamped euglycemia (5), on day 1 in the present study. Thus, there was greater antecedent hyperinsulinemia in the earlier study, but that was the same on both study occasions. Despite the apparent similarity of the peak cortisol levels produced on day 1 in both studies, we suspect that the critical issue is the antecedent plasma cortisol dose. We found that marked α_1–24ACTH-induced cortisol elevations (to ∼45 μg/dl) reduced sympathoadrenal and neurogenic symptom responses to hypoglycemia the following day (16). Here, using the same experimental design and methods on day 2, we find that less-marked antecedent cortisol elevations did not do so.

Additional data are seemingly inconsistent with the suggestion that cortisol mediates hypoglycemia-associated autonomic failure in diabetes. Cortisol (among other) responses to nocturnal hypoglycemia are virtually absent during polysomnography-documented sleep in patients with type 1 diabetes (32,33). Nocturnal hypoglycemia, with the patients ostensibly asleep, has been shown to reduce sympathoadrenal and neurogenic symptom responses to hypoglycemia the following morning (30). To the extent that the latter patients were asleep during the nocturnal hypoglycemia, their reduced sympathoadrenal and neurogenic symptom responses to hypoglycemia—the key features of hypoglycemia-associated autonomic failure (1–4)—the following morning cannot, therefore, be attributed to a cortisol response to the antecedent nocturnal hypoglycemia.

To this point we have focused on the sympathoadrenal and neurogenic symptom responses to hypoglycemia because those are directly relevant to hypoglycemia-associated autonomic failure in diabetes (1–4). Among the other end points measured, plasma glucagon, growth hormone, and cortisol responses to hypoglycemia on day 2 were also not reduced after cortisol infusion on day 1. Growth hormone, but not glucagon or cortisol, responses to hypoglycemia were reduced after pharmacological (ACTH-induced) elevations of cortisol in our earlier study (16). In that study the plasma pancreatic polypeptide response to hypoglycemia, a marker of the parasympathetic response, was reduced significantly. It appeared to be reduced slightly in the present study. However, parasympathetic (in contrast to sympathetic cholinergic) responses are not known to be involved in the glucose counterregulatory or symptomatic responses to hypoglycemia (34). Finally, there were no differences in the glucose infusion rates required to maintain the hypoglycemic clamps on day 2 after saline or cortisol infusions on day 1 (biological support for the absence of differences in the measured glucagon and epinephrine responses) and no differences in the metabolic (lactate, nonesterified fatty acid, β-hydroxybutyrate, and alanine) or hemodynamic (heart rate and systolic and diastolic blood pressure) responses.

In summary, plasma cortisol elevations during euglycemia that are comparable to (and indeed above) levels that occur during hypoglycemia did not reduce sympathoadrenal or neurogenic symptom responses to subsequent hypoglycemia. Therefore, these data do not support the suggestion that the cortisol response to antecedent iatrogenic hypoglycemia mediates hypoglycemia-associated autonomic failure in diabetes.

This work was supported, in part, by U.S. Public Health Service/National Institutes of Health Grants R37 DK27085, M01 RR00036, P60 DK20579, and T32 DK07120 and a fellowship award from the American Diabetes Association.

The authors gratefully acknowledge the analytical assistance of Krishan Jethi, Cornell Blake, Joy Brothers, Zina Lubovich, and Michael Morris; the assistance of the nursing staff of the Washington University General Clinical Research Center in the performance of this study; and the assistance of Karen Muehlhauser and Janet Dedeke in the preparation of this manuscript.