Prediction of Medium-Term Outcome by Cortisol Response to the Combined Dexamethasone-CRH Test in Patients With Remitted Depression

A large body of clinical and preclinical evidence regarding corticotropin-releasing hormone (CRH) supports the notion that hyperactive CRH-containing neurons in conjunction with coexpressed and cosecreted vasopressin account not only for enhanced secretion of ACTH and corticosteroids but also for a number of behavioral and autonomic changes that are regular concomitants of depression (1). This evidence is based on data from patients suffering from depression, in whom levels of CRH were found to be high in cerebrospinal fluid and hypothalamic neurons. According to neuroanatomical studies in rats, these hypothalamic CRH nerve terminals innervate, among other sites, the locus ceruleus and the central nucleus of the amygdala, two structures implicated in behavioral and autonomic response to stress and anxiety. These associations have led to the hypothesis that peripherally measurable hypothalamic-pituitary-adrenocortical (HPA) abnormalities reflect a central pathology that is causally related to depression. Compatible with this hypothesis are the results of studies showing that persistently abnormal HPA regulation during antidepressant treatment indicates an unfavorable response and, in turn, normalization of HPA activity precedes resolution of depressive psychopathology (2). In the present study we extended this approach by using the combined dexamethasone-CRH test with inpatients after initiation of antidepressant treatment and before discharge after the patients recovered. This test has been proven to pick up HPA abnormalities with greater sensitivity than do other neuroendocrine tests (2). We hypothesized that patients in whom the cortisol response after the dexamethasone-CRH test either was high on both occasions or changed from a low to a high level would have a worse medium-term outcome than those in whom the response pattern was low on both occasions or changed from high to low.

METHOD

Originally 42 inpatients were admitted to the study, but two who did not respond to various treatments within 6 months were excluded from analysis. The remaining 40 patients (14 men and 26 women) were aged 25 to 79 years (mean=50.10, SD=14.52) and met the DSM-IV criteria for a major depressive disorder according to results of a semistructured interview. Written informed consent was obtained after the procedures had been fully explained. Patients with major medical disorders were excluded. If the subjects scored at least 18 points on the Hamilton Depression Rating Scale, a dexamethasone-CRH test was administered 5–8 days after initiation of antidepressant treatment of the physician’s choice. This time delay was to rule out confounds resulting from hospitalization and the cessation of any drug treatment received before admission. An oral dose of 1.5 mg of dexamethasone was given at 11:00 p.m., and the next day the patient rested supine on a bed for at least 30 minutes before 100 µg of CRH was infused. Blood was collected every 15 minutes between 3:00 and 4:15 p.m. Just before discharge and after the patients had recovered from the depressive episode, i.e., showed a drop in the Hamilton depression score of at least 50% from baseline or had a score below 10, a second dexamethasone-CRH test was administered. After discharge, all patients continued to take medication and were examined three times (at 3 weeks, 3 months, and 6 months) during the observation period. Clinical relapse was defined as an increase in the Hamilton depression score to more than 16.

The probability of relapse during the 6 months after discharge was estimated by using the cortisol response to the two dexamethasone-CRH tests, selected demographic and illness-related characteristics (age, sex, and duration of inpatient treatment), and logistic regression analysis. The likelihood ratio served as the criterion for determining variables (predictors) to be removed from the model, and the Wald statistic was used to test the coefficients of the predictors for significance. Qualitative or categorical variables were coded according to the dummy coding scheme. To extract an easily accessible predictor from the cortisol responses, we arbitrarily defined a dichotomous variable, called “high cortisol response”; a value of 1 was assigned for this variable if the maximum level of the cortisol responses on both test occasions was above 150 ng/ml or the level on the second test was more than 25 ng/ml higher than the level on the first, and a value of 0 was assigned in all other cases.

RESULTS

The mean score on the Hamilton depression scale after admission and initiation of antidepressant treatment was 27.0 (SD=6.3). After a mean of 95 (SD=14) days of treatment, the mean score had dropped to 3.4 (SD=3.5). The maximum cortisol responses on the first test were between 7.5 and 297.5 ng/ml (mean=89.40, SD=71.28), and on the second test they were between 6.1 and 236.4 ng/ml (mean=66.51, SD=66.54). As illustrated in Figure 1, the patients who relapsed and those who did not relapse had different time course curves. Logistic regression analysis revealed that our model with outcome (relapse versus nonrelapse) as the dependent variable, high cortisol response, and the aforementioned demographic and illness-related variables as predictors classified the observed data well. The predictions were correct for 27 (90%) of the 30 nonrelapsed patients and eight (80%) of the 10 relapsed patients. Furthermore, of the variables considered, only high cortisol response significantly influenced the probability of a relapse (β=–1.79, Wald χ²=12.90, df=1, p=0.0003). The relapsed and nonrelapsed patients were indistinguishable in their initial Hamilton depression scores and comorbidity rates.

All patients had received adequate doses of antidepressants throughout the study. For at least 7 days before both neuroendocrine tests they had received constant doses. Seventeen patients were also being treated with benzodiazepines when they had the first test. Of the 25 patients who had high cortisol responses, 11 (44%) were taking benzodiazepines in combination with antidepressants, and six (40%) of the 15 patients with normal responses were receiving benzodiazepines. These rates rule out the possibility that drug interference obfuscated the neuroendocrine test results. At test 2, none of the patients was receiving benzodiazepines.

DISCUSSION

The results of the present study suggest that the cortisol responses to the combined dexamethasone-CRH test predict medium-term outcome in patients with remitted depression. The study used a slightly modified version of the dexamethasone-CRH test, with only five cortisol samples per test (2). In contrast to previous studies, the current protocol explored the predictive potential of the test, independent of the type of drug treatment. Under these naturalistic conditions we continued to observe that antidepressants tend to normalize HPA abnormality in people with major depression (2). In an extension of previous findings, we found that although the patients were indistinguishable in terms of psychopathology ratings and comorbidity according to initial diagnostic assessment, only those who had persistently severe HPA dysregulation were prone to relapse. The relapse rate of 25% within 6 months agrees with previous data (3) but is higher than that found when antidepressant treatment is combined with psychotherapy on a more systematic schedule (4). Aside from the practical nature of this test as feedback for the clinician, the finding reported here corroborates earlier evidence of a causal relationship between HPA abnormality and depression (5). It has been proposed that abnormal dexamethasone-CRH test results indicate low corticosteroid receptor function and, further, that antidepressants may act by counteracting this deficit (6). The present findings are consistent with this concept and suggest the need for a broader-based evaluation of neuroendocrine function tests that are easily applied and may be useful as predictors of clinical course.

Received Oct. 23, 1997; revisions received April 24, July 30, and Oct. 9, 1998; accepted Nov. 19, 1998. From the Max Planck Institute of Psychiatry. Address reprint requests to Dr. Zobel, Max Planck Institute of Psychiatry, Kraepelinstrasse 10, D-80804 Munich, Germany; [email protected] (e-mail)