Psychonegroendocrinology.VoL 17, No. 5, pp. 475-484, 1992
0306-,4530/92 $5.00+0.00 @1992 Pergamon Press Ltd.
Printed in Great Briton
CORTISOL LEVELS A N D VIGILANCE IN EATING DISORDER PATIENTS R. G. LAESSLE,1 M. FISCHER,1 M. M. FICHTER,2 K. M. PIRKE,1 J.-C. KRIEG1 Planck Institute of Psychiatry, Clinical Institute, Munich, and 2psychosomatic Hospital Roseneck, Prien/Chiemsee, Germany
1Max
(Received 11 June 1991; in finalform2 December 1991)
SUMMARY Elevated plasma levels of eortisol, as well as deficits in cognitive processes such as attention, have been observed in patients with eating disorders. The association between plasma cortisol and performance in a vigilance task was studied in 17 patients with bulimia nervosa or anorexia nervosa during the acute phase of their eating disorder. In comparison to normal young women, the patients had a significantly lower hit rate in a discrimination task and showed an impaired perceptual sensitivity index. They also displayed significantly longer reaction times to hits, but not to false alarms. Cortisol levels of the patients were significantly higher than those of the normal controls. When patients were divided according to their median cortisol level, the patients with higher levels performed significantly more poorly than did the patients with lower levels. In the total patient sample, cortisol levels showed a significant negative correlation with hit-rate (r=-.54) and a significant positive correlation with reaction time to hits (r=.70). Other clinical characteristics were not related to cognitive performance. These results suggest a possible role of cortisol in the development of attentional deficits in eating disorder patients.
INTRODUCTION PATIENTS with the eating disorders anorexia nervosa and bulimia nervosa are characterized by specific psychopathological features such as an obsessive drive for thinness and body dissatisfaction, which promote an extremely disturbed eating behavior. The chronic restriction of food intake of anorectic patients leads to severe underweight, while bulimic patients, due to alternating periods of fasting and overeating, usually stay within the normal weight range. A variety of metabolic and endocrine abnormalities, such as increased plasma levels of cortisol and ketone bodies and low plasma concentrations of triiodothyronine (T3), have been described in anorexia nervosa (Pirke & Ploog, 1986) and to a somewhat lesser degree in bulimia nervosa (Pirke et aL, 1985). Most of these abnormalities can best be explained as adaptation to starvation, caused either by chronic (in the case of anorexia nervosa) or by intermittent (in the case of bulimia nervosa) restriction of caloric intake (Fichter & Pirke, 1990; Kaye et al., 1990). Therefore, most of the metabolic and endocrine abnormalities seen in anorectic and bulimic patients do occur on a continuum, dependent not only on the degree of weight loss, but also on nutritional status, related to the structure of eating behavior (i.e., caloric intake, macronutrient composiAddress correspondence and reprint requests to: Dr. J.-C. Krieg, Max Planck Institute of Psychiatry, Clinical Institute, Kraepelinstral~ 10, D-8000 Miinchen 40, GERMANY. 475
476
R.G. LAESSLEet al.
tion, timing of meals) (Schweiger et al., 1988). It is very unlikely that differences in these biological alterations between anorexia and bulimia are of a qualitative nature due to the diagnostic category p e r se. There also is empirical evidence of impaired cognitive performance in both anorexia nervosa (Fox, 1981; Hamsher et al., 1981; Palazidou et al., 1990) and bulimia nervosa (McKay et al., 1986; Laessle et al., 1990). These deficits have been hypothesized to indicate a primary central nervous system dysfunction, which also might play a role in the pathogenesis and maintenance of these eating disorders (Hamsher et al., 1981). Another assumption is that the metabolic and endocrine changes which occur as a consequence of the disturbed eating behavior are involved in the impairment of specific cognitive functions. In a previous study of bulimic patients (Laessle et al., 1990), impaired vigilance was associated with higher levels of ~-hydroxybutyric acid, which as a ketone body serves as a source of brain energy supply during periods of fasting. Since ketone bodies, however, are incapable of sufficiently providing the energy requirement of all brain regions (Hawkins & Biebuyck, 1979), cognitive impairment could be directly related to an inadequate fuel supply. On the other hand, cognitive dysfunction also could be mediated by starvation-induced neuroendocrine changes. In particular, cortisol levels may play a crucial role, since there is evidence that glucocorticoids not only influence metabolic processes in the periphery, but also modulate central nervous system functions involved in stimulus perception and information processing (Henkin, 1975). Support for this comes from clinical studies which have revealed decreased concentration and memory in patients with Cushing's syndrome (Starkman & Schteingart, 1981). The severity of the patients' cognitive dysfunction was significantly correlated with their plasma levels of cortisol and ACTH, and it lessened with decreasing cortisol concentrations during treatment (Starkman et al., 1986). In depressed patients, higher mean urinary free cortisol levels were associated with a higher degree of impairment in cognitive abstracting ability (Rubinow et al., 1984), and in healthy subjects the administration of glucocorticoids influenced both attentional processes (Born et al., 1987) and performance on memory tasks (Wolkowitz et al., 1990). Born et al. (1987) concluded from their literature review and their own results on the effects of hydrocortisone on auditory event-related potentials that increasing levels of glucocorticoids may lead to a reduced sensitivity of the nonspecific sensory system. In the present study, the relationship between cortisol levels and cognitive performance in terms of sustained attention was explored in patients with an eating disorder, who were assumed would display substantial variability in their cortisol levels. It was hypothesized that patients with higher cortisol levels would perform worse on the vigilance task. METHODS Subjects
Seventeen female patients fulfillingDSM-III-R criteria (AmericanPsychiatric Association, 1987) for either anorexia nervosa (n = 9) or bulimia nervosa (n = 8) were assessed during the first week after hospital admission. The clinical characteristicsof the sample are shown in Table I. Twenty-twowomen served as controls for vigilance performance. The absence of any physical or psychiatric disorder was assured by a semistructuredinterview. The control subjects had a mean age of 25.0 yr (SD = 4,8 yr) and were of normal weight (body m a s s index [BMI]: 20,3+1.9 kg/m2). All patients and controls reported that they had not taken any psychotropic medication for at least 4 wk prior to study. The severity of the psychopathology of the eating disorder was assessed with the Eating Disorder Inventory (EDI) (Garner et al., 1983). (The 95th percentile of the total score in a normal female comparison sample is 17.) Depression was measured with the Depression Scale (DS) (von
477
CORTISOLAND VIGILANCE
Zerssen, 1976). (The 95th percentile of a normal female sample is 16.) As indicated in Table I, the patients displayed significant psychopathology. Cognitive task
Cognitive performance in terms of attention and information processing was measured with a continuous performance test (CFr). The CPT was originally developed by Rosvold et al. (1956) to investigate vigilance in brain-damaged subjects. Since then it has been widely used to assess sustained attention in a discrimination task. In the present study, a degraded-stimulus CPT in a computerized version was used, as described by Nuechterlein (1986). Stimuli were blurred single digits (0-9) which were presented for 75 msec every 1 sec on a Taxan Supervision HI monitor. Patients were told to respond by pressing a button with their dominant hand each time they detected the digit 0. No response was required for nontargets. Targets were presented irregularly, with a probability of 25%. Target and nontarget digits were presented in a pseudorandorn sequence, with the restrictions that identical digits never follow one another and that targets be preceded by each nontarget digit an equal number of times. A total of 480 stimulus trials (120 targets and 360 nontargets) were presented over a test period lasting just over 8 min. Although the stimuli were given continuously, for the analysis of changes in performance over time, the trials were divided into three 2.7-rain blocks of 160 trials each. Subjects were given 160 practice trials before performing the task. During this practice period, targets and nontargets were presented at the same rate as in the subsequent main task. A personal computer controlled the presentation of stimuli and recorded the responses in all phases of the experiment. Measures of vigilance were signals detected (hit rate) and responses to nontargets (false-alarm rate). In addition, the signal detection theory indices of "perceptual sensitivity" and "response criterion" were calculated according to Grief (1971). Signal detection theory provides a means of distinguishing among various components of vigilance performance. Perceptual sensitivity reflects the degree to which the frequency distribution of the internal perceptual evidence generated by signal trials is distinguished from that generated by the noise trials. Low perceptual sensitivity indicates greater difficulty in discriminatingbetween targets and nontargets. The response criterion level assesses the amount of perceptual evidence needed prior to making a decision to respond to a stimulus as a signal. A low hit rate may reflect an impaired discrimination ability but might also be due to a high response criterion, meaning that a person is generally more cautious in making a decision. In addition to these indices of vigilance, the response times for hits and false alarms were monitored. The CI'T testing of the patients and the controls was carried out in the afternoon between 1500h and 1700h. The last food intake for all subjects was at lunch time between 1200h and 1300h. Biochemical measures
Morning concentrations of plasma cortisol and T3 were measured on the day of CPT assessment. At 0800h,
TABLE I. CLINICALAND END(X:RINECHARACTERISTICSOF THE PATIENTSAMPLE Total sample
Anorexia nervosa
Bulimia n e r v o s a
( n = 17)
( n = 9)
( n = 8)
Age (years) Body mass index ( k g / m 2)
24.4 + 4.0
23.8 + 3.8
25.0 + 4.4
16.4+ 1.6
13.8+ 1.9
18.9:t: 1.3
Duration of illness (months)
83.4 + 43.1
79.1 + 43.5
88.1 + 45.1
14.1 + 17.4
2.6+4.9
27.1 + 17.3
Depression Scale
34.8+ 14.0 19.6 ± 9.7
37.9 + 12.3 22.3 + 8.8
31.8+ 15.8 16.9 + 10.4
Cortisol ( n g / m l )
132.2 ± 34.1
142.1 ± 38.1
121.0 ± 27.2
0.68± 0.21
0.68± 0.21
0.67± 0.20
Frequency of bulimic episodes (per week) Eating Disorder Inventory
T3 ( n g / m l ) Values are mean+ SO.
478
R.G. LAESSLEet al.
after overnight fasting, a butterfly needle was inserted into a forearm vein. Blood samples were collected at 10min intervals from 0830h-0930h. The patients remained supine throughout the entire procedure. Immediately after blood collection was completed, the plasma was separated by centrifugation and kept frozen at -80 ° C until analysis. Cortisol was measured in a sample pooled from the repeated blood samples obtained from each individual subject by radioimmunoassay, with kits supplied by Serono (Freiburg, Germany). The intra-assay variability was 4.6% at an average concentration of 98.8 ng/ml, and the inter-assay variability was 7.1% at an average concentration of 109 ng/ml. Cortisol values obtained by this sampling procedure have been demonstrated to be a valid index of 24-hr cortisol secretory activity in eating disorder patients and in normal controls (Doerr et al., 1980; Heufelder et al., 1985). T3 also was analyzed in the pooled sample from each patient with radioimmunoassay kits (Serono, Freiburg, Germany). The intra-assay variability was 3.9% at an average concentration of 1.29 ng/ml, and the inter-assay variability was 5.4% at an average concentration of 1.95 ng/ml. Ten healthy young women, with a mean (+SD) age of 24.7+2.2 yr and a mean BMI of 21.3+1.6 kg/m2 served as controls to provide normal values for cortisol and T 3, which were obtained under identical sampling conditions as described for the patients. These subjects did not take part in the CPT testing. Statistical analyses
Comparisons between two groups were done by t test. CPT results between groups were compared by a twofactorial multivariate analysis of variance (MANOVA) for repeated measures (O'Brian & Kaiser, 1985). As a within-subjects factor, the three trial blocks during the CPT were used. RESULTS Cortisol and T3 levels of the patients are depicted in Table I. Anorectic and bulimic patients did not differ in their mean values. The mean cortisol value of the total sample of eating disorder patients was significantly higher than that of the normal controls (rnean+SD: 132.2+ 34.1 ng/ml vs. 95.0+39.8 ng/ml; p
0306-,4530/92 $5.00+0.00 @1992 Pergamon Press Ltd.
Printed in Great Briton
CORTISOL LEVELS A N D VIGILANCE IN EATING DISORDER PATIENTS R. G. LAESSLE,1 M. FISCHER,1 M. M. FICHTER,2 K. M. PIRKE,1 J.-C. KRIEG1 Planck Institute of Psychiatry, Clinical Institute, Munich, and 2psychosomatic Hospital Roseneck, Prien/Chiemsee, Germany
1Max
(Received 11 June 1991; in finalform2 December 1991)
SUMMARY Elevated plasma levels of eortisol, as well as deficits in cognitive processes such as attention, have been observed in patients with eating disorders. The association between plasma cortisol and performance in a vigilance task was studied in 17 patients with bulimia nervosa or anorexia nervosa during the acute phase of their eating disorder. In comparison to normal young women, the patients had a significantly lower hit rate in a discrimination task and showed an impaired perceptual sensitivity index. They also displayed significantly longer reaction times to hits, but not to false alarms. Cortisol levels of the patients were significantly higher than those of the normal controls. When patients were divided according to their median cortisol level, the patients with higher levels performed significantly more poorly than did the patients with lower levels. In the total patient sample, cortisol levels showed a significant negative correlation with hit-rate (r=-.54) and a significant positive correlation with reaction time to hits (r=.70). Other clinical characteristics were not related to cognitive performance. These results suggest a possible role of cortisol in the development of attentional deficits in eating disorder patients.
INTRODUCTION PATIENTS with the eating disorders anorexia nervosa and bulimia nervosa are characterized by specific psychopathological features such as an obsessive drive for thinness and body dissatisfaction, which promote an extremely disturbed eating behavior. The chronic restriction of food intake of anorectic patients leads to severe underweight, while bulimic patients, due to alternating periods of fasting and overeating, usually stay within the normal weight range. A variety of metabolic and endocrine abnormalities, such as increased plasma levels of cortisol and ketone bodies and low plasma concentrations of triiodothyronine (T3), have been described in anorexia nervosa (Pirke & Ploog, 1986) and to a somewhat lesser degree in bulimia nervosa (Pirke et aL, 1985). Most of these abnormalities can best be explained as adaptation to starvation, caused either by chronic (in the case of anorexia nervosa) or by intermittent (in the case of bulimia nervosa) restriction of caloric intake (Fichter & Pirke, 1990; Kaye et al., 1990). Therefore, most of the metabolic and endocrine abnormalities seen in anorectic and bulimic patients do occur on a continuum, dependent not only on the degree of weight loss, but also on nutritional status, related to the structure of eating behavior (i.e., caloric intake, macronutrient composiAddress correspondence and reprint requests to: Dr. J.-C. Krieg, Max Planck Institute of Psychiatry, Clinical Institute, Kraepelinstral~ 10, D-8000 Miinchen 40, GERMANY. 475
476
R.G. LAESSLEet al.
tion, timing of meals) (Schweiger et al., 1988). It is very unlikely that differences in these biological alterations between anorexia and bulimia are of a qualitative nature due to the diagnostic category p e r se. There also is empirical evidence of impaired cognitive performance in both anorexia nervosa (Fox, 1981; Hamsher et al., 1981; Palazidou et al., 1990) and bulimia nervosa (McKay et al., 1986; Laessle et al., 1990). These deficits have been hypothesized to indicate a primary central nervous system dysfunction, which also might play a role in the pathogenesis and maintenance of these eating disorders (Hamsher et al., 1981). Another assumption is that the metabolic and endocrine changes which occur as a consequence of the disturbed eating behavior are involved in the impairment of specific cognitive functions. In a previous study of bulimic patients (Laessle et al., 1990), impaired vigilance was associated with higher levels of ~-hydroxybutyric acid, which as a ketone body serves as a source of brain energy supply during periods of fasting. Since ketone bodies, however, are incapable of sufficiently providing the energy requirement of all brain regions (Hawkins & Biebuyck, 1979), cognitive impairment could be directly related to an inadequate fuel supply. On the other hand, cognitive dysfunction also could be mediated by starvation-induced neuroendocrine changes. In particular, cortisol levels may play a crucial role, since there is evidence that glucocorticoids not only influence metabolic processes in the periphery, but also modulate central nervous system functions involved in stimulus perception and information processing (Henkin, 1975). Support for this comes from clinical studies which have revealed decreased concentration and memory in patients with Cushing's syndrome (Starkman & Schteingart, 1981). The severity of the patients' cognitive dysfunction was significantly correlated with their plasma levels of cortisol and ACTH, and it lessened with decreasing cortisol concentrations during treatment (Starkman et al., 1986). In depressed patients, higher mean urinary free cortisol levels were associated with a higher degree of impairment in cognitive abstracting ability (Rubinow et al., 1984), and in healthy subjects the administration of glucocorticoids influenced both attentional processes (Born et al., 1987) and performance on memory tasks (Wolkowitz et al., 1990). Born et al. (1987) concluded from their literature review and their own results on the effects of hydrocortisone on auditory event-related potentials that increasing levels of glucocorticoids may lead to a reduced sensitivity of the nonspecific sensory system. In the present study, the relationship between cortisol levels and cognitive performance in terms of sustained attention was explored in patients with an eating disorder, who were assumed would display substantial variability in their cortisol levels. It was hypothesized that patients with higher cortisol levels would perform worse on the vigilance task. METHODS Subjects
Seventeen female patients fulfillingDSM-III-R criteria (AmericanPsychiatric Association, 1987) for either anorexia nervosa (n = 9) or bulimia nervosa (n = 8) were assessed during the first week after hospital admission. The clinical characteristicsof the sample are shown in Table I. Twenty-twowomen served as controls for vigilance performance. The absence of any physical or psychiatric disorder was assured by a semistructuredinterview. The control subjects had a mean age of 25.0 yr (SD = 4,8 yr) and were of normal weight (body m a s s index [BMI]: 20,3+1.9 kg/m2). All patients and controls reported that they had not taken any psychotropic medication for at least 4 wk prior to study. The severity of the psychopathology of the eating disorder was assessed with the Eating Disorder Inventory (EDI) (Garner et al., 1983). (The 95th percentile of the total score in a normal female comparison sample is 17.) Depression was measured with the Depression Scale (DS) (von
477
CORTISOLAND VIGILANCE
Zerssen, 1976). (The 95th percentile of a normal female sample is 16.) As indicated in Table I, the patients displayed significant psychopathology. Cognitive task
Cognitive performance in terms of attention and information processing was measured with a continuous performance test (CFr). The CPT was originally developed by Rosvold et al. (1956) to investigate vigilance in brain-damaged subjects. Since then it has been widely used to assess sustained attention in a discrimination task. In the present study, a degraded-stimulus CPT in a computerized version was used, as described by Nuechterlein (1986). Stimuli were blurred single digits (0-9) which were presented for 75 msec every 1 sec on a Taxan Supervision HI monitor. Patients were told to respond by pressing a button with their dominant hand each time they detected the digit 0. No response was required for nontargets. Targets were presented irregularly, with a probability of 25%. Target and nontarget digits were presented in a pseudorandorn sequence, with the restrictions that identical digits never follow one another and that targets be preceded by each nontarget digit an equal number of times. A total of 480 stimulus trials (120 targets and 360 nontargets) were presented over a test period lasting just over 8 min. Although the stimuli were given continuously, for the analysis of changes in performance over time, the trials were divided into three 2.7-rain blocks of 160 trials each. Subjects were given 160 practice trials before performing the task. During this practice period, targets and nontargets were presented at the same rate as in the subsequent main task. A personal computer controlled the presentation of stimuli and recorded the responses in all phases of the experiment. Measures of vigilance were signals detected (hit rate) and responses to nontargets (false-alarm rate). In addition, the signal detection theory indices of "perceptual sensitivity" and "response criterion" were calculated according to Grief (1971). Signal detection theory provides a means of distinguishing among various components of vigilance performance. Perceptual sensitivity reflects the degree to which the frequency distribution of the internal perceptual evidence generated by signal trials is distinguished from that generated by the noise trials. Low perceptual sensitivity indicates greater difficulty in discriminatingbetween targets and nontargets. The response criterion level assesses the amount of perceptual evidence needed prior to making a decision to respond to a stimulus as a signal. A low hit rate may reflect an impaired discrimination ability but might also be due to a high response criterion, meaning that a person is generally more cautious in making a decision. In addition to these indices of vigilance, the response times for hits and false alarms were monitored. The CI'T testing of the patients and the controls was carried out in the afternoon between 1500h and 1700h. The last food intake for all subjects was at lunch time between 1200h and 1300h. Biochemical measures
Morning concentrations of plasma cortisol and T3 were measured on the day of CPT assessment. At 0800h,
TABLE I. CLINICALAND END(X:RINECHARACTERISTICSOF THE PATIENTSAMPLE Total sample
Anorexia nervosa
Bulimia n e r v o s a
( n = 17)
( n = 9)
( n = 8)
Age (years) Body mass index ( k g / m 2)
24.4 + 4.0
23.8 + 3.8
25.0 + 4.4
16.4+ 1.6
13.8+ 1.9
18.9:t: 1.3
Duration of illness (months)
83.4 + 43.1
79.1 + 43.5
88.1 + 45.1
14.1 + 17.4
2.6+4.9
27.1 + 17.3
Depression Scale
34.8+ 14.0 19.6 ± 9.7
37.9 + 12.3 22.3 + 8.8
31.8+ 15.8 16.9 + 10.4
Cortisol ( n g / m l )
132.2 ± 34.1
142.1 ± 38.1
121.0 ± 27.2
0.68± 0.21
0.68± 0.21
0.67± 0.20
Frequency of bulimic episodes (per week) Eating Disorder Inventory
T3 ( n g / m l ) Values are mean+ SO.
478
R.G. LAESSLEet al.
after overnight fasting, a butterfly needle was inserted into a forearm vein. Blood samples were collected at 10min intervals from 0830h-0930h. The patients remained supine throughout the entire procedure. Immediately after blood collection was completed, the plasma was separated by centrifugation and kept frozen at -80 ° C until analysis. Cortisol was measured in a sample pooled from the repeated blood samples obtained from each individual subject by radioimmunoassay, with kits supplied by Serono (Freiburg, Germany). The intra-assay variability was 4.6% at an average concentration of 98.8 ng/ml, and the inter-assay variability was 7.1% at an average concentration of 109 ng/ml. Cortisol values obtained by this sampling procedure have been demonstrated to be a valid index of 24-hr cortisol secretory activity in eating disorder patients and in normal controls (Doerr et al., 1980; Heufelder et al., 1985). T3 also was analyzed in the pooled sample from each patient with radioimmunoassay kits (Serono, Freiburg, Germany). The intra-assay variability was 3.9% at an average concentration of 1.29 ng/ml, and the inter-assay variability was 5.4% at an average concentration of 1.95 ng/ml. Ten healthy young women, with a mean (+SD) age of 24.7+2.2 yr and a mean BMI of 21.3+1.6 kg/m2 served as controls to provide normal values for cortisol and T 3, which were obtained under identical sampling conditions as described for the patients. These subjects did not take part in the CPT testing. Statistical analyses
Comparisons between two groups were done by t test. CPT results between groups were compared by a twofactorial multivariate analysis of variance (MANOVA) for repeated measures (O'Brian & Kaiser, 1985). As a within-subjects factor, the three trial blocks during the CPT were used. RESULTS Cortisol and T3 levels of the patients are depicted in Table I. Anorectic and bulimic patients did not differ in their mean values. The mean cortisol value of the total sample of eating disorder patients was significantly higher than that of the normal controls (rnean+SD: 132.2+ 34.1 ng/ml vs. 95.0+39.8 ng/ml; p
