0021-972X/91/7303-0541$03.00/0 Journal of Clinical Endocrinology and Metabolism Copyright © 1991 by The Endocrine Society
Vol. 73, No. 3 Printed in U.S.A.
Hormonal Responses during Insulin-Induced Hypoglycemia in Manic-Depressed, Unipolar Depressed, and Healthy Control Subjects* JAY D. AMSTERDAM AND GREG MAISLIN Depression Research Unit, Department of Psychiatry, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania 19104
ABSTRACT. Early studies using the insulin tolerance test (ITT) in affective disorders reported a blunted GH response in some depressed patients. However, results from subsequent studies were less consistent. These discrepancies resulted in part from limited sample sizes causing diagnostic heterogeneity. In the present study we examined hormonal response patterns during the ITT in 25 bipolar patients (19 depressed and 6 hypomanic), 91 unipolar depressives, and 51 healthy volunteers to determine whether distinct neuroendocrine response patterns characterized specific diagnostic subgroups. We measured the glucose, GH, PRL, and cortisol responses during a 75-min ITT and observed a diminished cumulative GH response in bipolar hypomanic patients compared to bipolar depressives (P = 0.004) and healthy volunteers (P < 0.001), and
a larger cumulative GH response in bipolar depressives compared to unipolar patients (P = 0.02). No differences were observed in either PRL or cortisol responses among subject groups. The present findings suggest that GH responses during the ITT may be more complex than initially described and partially dependent upon affective disorder subtype. Thus, bipolar patients may have distinctive GH response patterns compared to those of unipolar depressives and healthy controls. Moreover, some of the difference observed in the GH response to the ITT may result from the phase of the manic-depressive illness. These data indicate differences in neuroendocrine substrates between affective disorder subtypes. {J Clin Endocrinol Metab 73: 541548,1991)
T
HE INSULIN tolerance test (ITT) was one of the earliest strategies for assessing neuroendocrine disturbances in affective illness. Early studies demonstrated a blunted GH response during the ITT (1-5), and several investigators hypothesized that this disturbance might result from alterations in brain noradrenergic neurotransmission (3, 6-8). These early studies were confounded by a variety of methodological problems, including limited sample sizes with diagnostic heterogeneity, and subsequent studies demonstrated less consistent results (9-12). Thus, in the present study we have performed the ITT in diagnostic subgroups of affectively ill patients and healthy controls to determine whether specific neuroendocrine disturbances might characterize these subgroups. Limited attention has been given to the ITT in diagnostic subgroups of affectively ill patients, in particular
t h o s e with manic-depressive (bipolar) illness. In t h i s
regard, Sachar et al. (2) initially reported a blunted GH response in 5 of 13 patients with psychotic, neurotic, and bipolar depression; however, a subsequent study by this group in 8 unipolar, 5 bipolar, and 9 healthy controls failed to substantiate these findings (9). Casper et al. (4) administered the ITT to 13 depressed patients and an unspecified number of healthy controls and observed a diminished GH response in men, although no information was provided regarding the diagnostic subtype. More recently, Koslow et al. (10) observed no difference in the peak GH response during the ITT in 22 unipolar and 11 bipolar depressives compared to that in 21 healthy controls; these data confirmed earlier observations by our group comparing 22 bipolar patients to 22 healthy controls (13). Because most neuroendocrine studies of affective illness have been limited to small patient samples with substantial diagnostic heterogeneity, a closer examination of hormone responses during the ITT in specific diagnostic subgroups of affectively ill patients appeared warranted to determine whether distinct neuroendocrine response patterns characterized individual patient groups. We, therefore, assessed glucose, GH, PRL, and cortisol responses to insulin-induced hypoglycemia in
Received September 4,1990. Address all correspondence and requests for reprints to: Jay D. Amsterdam, M.D., Depression Research Unit, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104. * This work was supported in part by NIH Clinical Research Center Grant RR-00040, the Jack Warsaw Fund for Research in Biological Psychiatry, and NIMH Training Program in Neuropsychopharmacology Grant MH-14655.
541
AMSTERDAM AND MAISLIN
542
bipolar hypomanic and depressed patients and compared these findings to those obtained from unipolar depressives and healthy control subjects studied under similar experimental conditions.
Patients We examined 116 out-patients from the Depression Research Unit of the Hospital of the University of Pennsylvania: 25 had bipolar disorder (19 depressed and 6 hypomanic) and 91 had unipolar depression. Descriptive features of the subject groups are displayed in Table 1. Diagnostic assessments for all subjects were performed before neuroendocrine testing by two research psychiatrists using a semistructured diagnostic interview based upon the Schedule for Affective Disorders and Schizophrenia (SADS) format (14). All patients fulfilled DSM III-R criteria of the American Psychiatric Association (15) for major depression, single or recurrent episode, or for bipolar disorder, depressed or hypomanic subtype. The symptom severity of depression was more than 18 on the 17-item Hamilton Depression Rating Scale (16) at the time of neuroendocrine testing. Patients with characterological and dysthymic disorder and patients with psychotic mania were excluded. All subjects were drug free for a minimum of 10 days before testing, and most were drug free for longer periods. No patient had received a neuroleptic drug, lithium carbonate, or electroconvulsive therapy within 3 months of testing. All patients were without physical illness and had no clinically meaningful laboratory abnormalities. None had a past history of endocrinopathy or chronic medical illness, and all subjects were within 20% of their ideal body weight. None of the women was taking oral contraceptive or replacement hormone therapy. Premenopausal women were studied without regard to the phase of their menstrual cycle. Controls Fifty-one drug-free healthy volunteers were also tested under similar conditions (Table 1). All were given a semistructured psychiatric interview and were free of psychiatric illness. All subjects were in good physical health, and none had a history 1. Mean (±SD) age and age range of subject groups
25 11 14
Mean ± SD age 30 ± 11" 32 ±12 29 ±9
Age range 18-56 18-56 20-53
19 6 91 37 54 51 24 27
31 ±14 30 ±10 39 ± 12 39 ±13 38 ± 11 34 ±12 31 ±10 38 ±12
20-56 18-53 18-68 21-68 18-64 21-60 21-56 23-60
Subject group
n
Bipolar Men Women Depressed Hypomanic Unipolar
Men Women Controls
Men Women
of endocrinopathy or recent use of corticosteroids or other hormone preparations. Finally, none had a family history of major psychiatric illness in either a primary or secondary relative. Procedures
Subjects and Methods
TABLE
JCE & M • 1991 Vol 73 • No 3
' Significant difference from unipolar patients (P < 0.05).
All subjects freely participated after providing informed consent. The ITT was performed at 0900 h after an overnight fast. After inserting a 19-gauge indwelling venous catheter in an antecubital vein, the subject remained at bed rest for 15 min, and then baseline blood samples were obtained for glucose, GH, PRL, and cortisol levels at -15 and 0 min. Crystalline synthetic (glucagon-free) regular insulin (0.1 U/kg BW) was injected as an iv bolus, and blood samples were obtained at 10, 20, 30, 40, 50,60, and 75 min for glucose, GH, PRL, and cortisol responses. An hypoglycemic response 50% or more of the fasting glucose level was necessary for inclusion in the hormone data analyses, as previously described (11, 13). Assays Glucose concentrations were measured using glucose oxidase-derived techniques (17). GH concentrations were determined by means of a double antibody RIA technique with a kit from Cambridge Nuclear (Billerica, MA) (13, 18). The lower limit of sensitivity was 0.8 ng/mL, and the intra- and interassay coefficients of variation were 8.4% and 14.8%, respectively. PRL concentrations were measured by a double antibody saturation-type RIA (13, 19). This assay was developed using antibody obtained from the NIH, National Pituitary Agency (prepared by Dr Parlow, Director, Pituitary Hormone and Antisera Center, Harbor-UCLA Medical Center, Torrance, CA), and iodinated hormone and PRL standards were purchased from Cambridge Medical Diagnostics (Billerica, MA). The primary antibody (AFP-3) was highly specific and was used at a dilution of 1:125,000 for a final tube dilution of 1:500,000. The lower limit of detection was 2.0 ng/mL, and the intra- and interassay coefficients of variation were 6.24% and 11.62%, respectively. Assays were simultaneously performed on samples from patients and control subjects. All samples from a single subject were assayed together in duplicate. Cortisol concentrations were measured by means of a single antibody RIA technique (13, 19). The intraassay coefficient of variation was 7.7%, while the interassay coefficients of variation for the low (mean, 9.8 Mg/dL) and high (mean, 32.5 ng/ dL) cortisol range were 18.1% and 14.5%, respectively. The lower limit of sensitivity was 1.0 /xg/dh. Statistical methods After obtaining basal hormone concentrations, responses were computed over time as the centered cumulative response (CCR) value. The CCR is similar to the area under the curve, except that the response curve is terminated before its return to the baseline value. It is computed using a trapezoidal approximation to the total integrated area under the response curve after subtracting the basal value. Time units were defined
HORMONAL RESPONSES DURING HYPOGLYCEMIA in 0.5-h increments for the purpose of computing the CCR value. Hence, the units for the GH CCR value were nanograms per mL/0.5 h, while the CCR values for glucose, cortisol, and PRL were milligrams per dL/0.5 h, micrograms per dL/0.5 h, and nanograms per mL/0.5 h, respectively. Statistical comparisons of the mean (±SD) glucose and hormone responses were then made between the bipolar hypomanic and depressed patients, and these, in turn, were compared to the values obtained from unipolar patients and healthy controls. The comparison between bipolar hypomanic and bipolar depressed patients was examined using the pooled t test of the mean CCR value (unless the hypothesis of equal variances was rejected at a = 0.05, in which case the t test for unequal variances was employed). The comparison was repeated using the Wilcoxon rank sum test to determine whether the result was caused by the presence of extreme values (20). Repeated measures analysis of variance (ANOVA) was used to test for significant differences in the shapes of the glucose and hormone response curves during the ITT. To maximize the statistical power for these tests over the widest range of potential differences in shape, it was decided that the null hypothesis would be rejected if 1) the univariate approach Greenhouse-Giesser adjusted time by group interaction F statistic was significant at a = 0.025, or 2) the multivariate approach Wilkes X F statistic was significant at a = 0.025 (18). This approach maintained an overall a level no greater than 0.05, yet increased the statistical power in cases when only one of the two approaches yielded significant findings. Pearson and Spearman correlations between the basal and CCR values were also examined to determine whether an analysis of covariance (ANCOVA), holding the basal value constant, was necessary. Additional analyses consisted of comparing hormone responses among the four subject groups (bipolar hypomanic, bipolar depressed, unipolar depressed, and healthy controls) using parametric and nonparametric procedures. Pairwise comparisons between subject groups were then performed, as described for the bipolar patients (see above), and t tests were used to describe any statistical differences. Because there were five separate comparisons for this analysis, results for each hormone response (holding the a at
Vol. 73, No. 3 Printed in U.S.A.
Hormonal Responses during Insulin-Induced Hypoglycemia in Manic-Depressed, Unipolar Depressed, and Healthy Control Subjects* JAY D. AMSTERDAM AND GREG MAISLIN Depression Research Unit, Department of Psychiatry, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania 19104
ABSTRACT. Early studies using the insulin tolerance test (ITT) in affective disorders reported a blunted GH response in some depressed patients. However, results from subsequent studies were less consistent. These discrepancies resulted in part from limited sample sizes causing diagnostic heterogeneity. In the present study we examined hormonal response patterns during the ITT in 25 bipolar patients (19 depressed and 6 hypomanic), 91 unipolar depressives, and 51 healthy volunteers to determine whether distinct neuroendocrine response patterns characterized specific diagnostic subgroups. We measured the glucose, GH, PRL, and cortisol responses during a 75-min ITT and observed a diminished cumulative GH response in bipolar hypomanic patients compared to bipolar depressives (P = 0.004) and healthy volunteers (P < 0.001), and
a larger cumulative GH response in bipolar depressives compared to unipolar patients (P = 0.02). No differences were observed in either PRL or cortisol responses among subject groups. The present findings suggest that GH responses during the ITT may be more complex than initially described and partially dependent upon affective disorder subtype. Thus, bipolar patients may have distinctive GH response patterns compared to those of unipolar depressives and healthy controls. Moreover, some of the difference observed in the GH response to the ITT may result from the phase of the manic-depressive illness. These data indicate differences in neuroendocrine substrates between affective disorder subtypes. {J Clin Endocrinol Metab 73: 541548,1991)
T
HE INSULIN tolerance test (ITT) was one of the earliest strategies for assessing neuroendocrine disturbances in affective illness. Early studies demonstrated a blunted GH response during the ITT (1-5), and several investigators hypothesized that this disturbance might result from alterations in brain noradrenergic neurotransmission (3, 6-8). These early studies were confounded by a variety of methodological problems, including limited sample sizes with diagnostic heterogeneity, and subsequent studies demonstrated less consistent results (9-12). Thus, in the present study we have performed the ITT in diagnostic subgroups of affectively ill patients and healthy controls to determine whether specific neuroendocrine disturbances might characterize these subgroups. Limited attention has been given to the ITT in diagnostic subgroups of affectively ill patients, in particular
t h o s e with manic-depressive (bipolar) illness. In t h i s
regard, Sachar et al. (2) initially reported a blunted GH response in 5 of 13 patients with psychotic, neurotic, and bipolar depression; however, a subsequent study by this group in 8 unipolar, 5 bipolar, and 9 healthy controls failed to substantiate these findings (9). Casper et al. (4) administered the ITT to 13 depressed patients and an unspecified number of healthy controls and observed a diminished GH response in men, although no information was provided regarding the diagnostic subtype. More recently, Koslow et al. (10) observed no difference in the peak GH response during the ITT in 22 unipolar and 11 bipolar depressives compared to that in 21 healthy controls; these data confirmed earlier observations by our group comparing 22 bipolar patients to 22 healthy controls (13). Because most neuroendocrine studies of affective illness have been limited to small patient samples with substantial diagnostic heterogeneity, a closer examination of hormone responses during the ITT in specific diagnostic subgroups of affectively ill patients appeared warranted to determine whether distinct neuroendocrine response patterns characterized individual patient groups. We, therefore, assessed glucose, GH, PRL, and cortisol responses to insulin-induced hypoglycemia in
Received September 4,1990. Address all correspondence and requests for reprints to: Jay D. Amsterdam, M.D., Depression Research Unit, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, Pennsylvania 19104. * This work was supported in part by NIH Clinical Research Center Grant RR-00040, the Jack Warsaw Fund for Research in Biological Psychiatry, and NIMH Training Program in Neuropsychopharmacology Grant MH-14655.
541
AMSTERDAM AND MAISLIN
542
bipolar hypomanic and depressed patients and compared these findings to those obtained from unipolar depressives and healthy control subjects studied under similar experimental conditions.
Patients We examined 116 out-patients from the Depression Research Unit of the Hospital of the University of Pennsylvania: 25 had bipolar disorder (19 depressed and 6 hypomanic) and 91 had unipolar depression. Descriptive features of the subject groups are displayed in Table 1. Diagnostic assessments for all subjects were performed before neuroendocrine testing by two research psychiatrists using a semistructured diagnostic interview based upon the Schedule for Affective Disorders and Schizophrenia (SADS) format (14). All patients fulfilled DSM III-R criteria of the American Psychiatric Association (15) for major depression, single or recurrent episode, or for bipolar disorder, depressed or hypomanic subtype. The symptom severity of depression was more than 18 on the 17-item Hamilton Depression Rating Scale (16) at the time of neuroendocrine testing. Patients with characterological and dysthymic disorder and patients with psychotic mania were excluded. All subjects were drug free for a minimum of 10 days before testing, and most were drug free for longer periods. No patient had received a neuroleptic drug, lithium carbonate, or electroconvulsive therapy within 3 months of testing. All patients were without physical illness and had no clinically meaningful laboratory abnormalities. None had a past history of endocrinopathy or chronic medical illness, and all subjects were within 20% of their ideal body weight. None of the women was taking oral contraceptive or replacement hormone therapy. Premenopausal women were studied without regard to the phase of their menstrual cycle. Controls Fifty-one drug-free healthy volunteers were also tested under similar conditions (Table 1). All were given a semistructured psychiatric interview and were free of psychiatric illness. All subjects were in good physical health, and none had a history 1. Mean (±SD) age and age range of subject groups
25 11 14
Mean ± SD age 30 ± 11" 32 ±12 29 ±9
Age range 18-56 18-56 20-53
19 6 91 37 54 51 24 27
31 ±14 30 ±10 39 ± 12 39 ±13 38 ± 11 34 ±12 31 ±10 38 ±12
20-56 18-53 18-68 21-68 18-64 21-60 21-56 23-60
Subject group
n
Bipolar Men Women Depressed Hypomanic Unipolar
Men Women Controls
Men Women
of endocrinopathy or recent use of corticosteroids or other hormone preparations. Finally, none had a family history of major psychiatric illness in either a primary or secondary relative. Procedures
Subjects and Methods
TABLE
JCE & M • 1991 Vol 73 • No 3
' Significant difference from unipolar patients (P < 0.05).
All subjects freely participated after providing informed consent. The ITT was performed at 0900 h after an overnight fast. After inserting a 19-gauge indwelling venous catheter in an antecubital vein, the subject remained at bed rest for 15 min, and then baseline blood samples were obtained for glucose, GH, PRL, and cortisol levels at -15 and 0 min. Crystalline synthetic (glucagon-free) regular insulin (0.1 U/kg BW) was injected as an iv bolus, and blood samples were obtained at 10, 20, 30, 40, 50,60, and 75 min for glucose, GH, PRL, and cortisol responses. An hypoglycemic response 50% or more of the fasting glucose level was necessary for inclusion in the hormone data analyses, as previously described (11, 13). Assays Glucose concentrations were measured using glucose oxidase-derived techniques (17). GH concentrations were determined by means of a double antibody RIA technique with a kit from Cambridge Nuclear (Billerica, MA) (13, 18). The lower limit of sensitivity was 0.8 ng/mL, and the intra- and interassay coefficients of variation were 8.4% and 14.8%, respectively. PRL concentrations were measured by a double antibody saturation-type RIA (13, 19). This assay was developed using antibody obtained from the NIH, National Pituitary Agency (prepared by Dr Parlow, Director, Pituitary Hormone and Antisera Center, Harbor-UCLA Medical Center, Torrance, CA), and iodinated hormone and PRL standards were purchased from Cambridge Medical Diagnostics (Billerica, MA). The primary antibody (AFP-3) was highly specific and was used at a dilution of 1:125,000 for a final tube dilution of 1:500,000. The lower limit of detection was 2.0 ng/mL, and the intra- and interassay coefficients of variation were 6.24% and 11.62%, respectively. Assays were simultaneously performed on samples from patients and control subjects. All samples from a single subject were assayed together in duplicate. Cortisol concentrations were measured by means of a single antibody RIA technique (13, 19). The intraassay coefficient of variation was 7.7%, while the interassay coefficients of variation for the low (mean, 9.8 Mg/dL) and high (mean, 32.5 ng/ dL) cortisol range were 18.1% and 14.5%, respectively. The lower limit of sensitivity was 1.0 /xg/dh. Statistical methods After obtaining basal hormone concentrations, responses were computed over time as the centered cumulative response (CCR) value. The CCR is similar to the area under the curve, except that the response curve is terminated before its return to the baseline value. It is computed using a trapezoidal approximation to the total integrated area under the response curve after subtracting the basal value. Time units were defined
HORMONAL RESPONSES DURING HYPOGLYCEMIA in 0.5-h increments for the purpose of computing the CCR value. Hence, the units for the GH CCR value were nanograms per mL/0.5 h, while the CCR values for glucose, cortisol, and PRL were milligrams per dL/0.5 h, micrograms per dL/0.5 h, and nanograms per mL/0.5 h, respectively. Statistical comparisons of the mean (±SD) glucose and hormone responses were then made between the bipolar hypomanic and depressed patients, and these, in turn, were compared to the values obtained from unipolar patients and healthy controls. The comparison between bipolar hypomanic and bipolar depressed patients was examined using the pooled t test of the mean CCR value (unless the hypothesis of equal variances was rejected at a = 0.05, in which case the t test for unequal variances was employed). The comparison was repeated using the Wilcoxon rank sum test to determine whether the result was caused by the presence of extreme values (20). Repeated measures analysis of variance (ANOVA) was used to test for significant differences in the shapes of the glucose and hormone response curves during the ITT. To maximize the statistical power for these tests over the widest range of potential differences in shape, it was decided that the null hypothesis would be rejected if 1) the univariate approach Greenhouse-Giesser adjusted time by group interaction F statistic was significant at a = 0.025, or 2) the multivariate approach Wilkes X F statistic was significant at a = 0.025 (18). This approach maintained an overall a level no greater than 0.05, yet increased the statistical power in cases when only one of the two approaches yielded significant findings. Pearson and Spearman correlations between the basal and CCR values were also examined to determine whether an analysis of covariance (ANCOVA), holding the basal value constant, was necessary. Additional analyses consisted of comparing hormone responses among the four subject groups (bipolar hypomanic, bipolar depressed, unipolar depressed, and healthy controls) using parametric and nonparametric procedures. Pairwise comparisons between subject groups were then performed, as described for the bipolar patients (see above), and t tests were used to describe any statistical differences. Because there were five separate comparisons for this analysis, results for each hormone response (holding the a at
