- Email: [email protected]
Serial dexamethasone suppression tests in initial suppressors and nonsuppressors treated with electroconvulsive therapy
BIOL PSYCHLURY ~987;22:463-472
463
Serial Dexamethasone Suppression Tests in Initial Suppressors and Nonsuppressors Treated with ~le~tro~onvulsiv~ Therapy D. P. Devanand, Paolo Decina, Harold A. Sackeim, Nancy Hopkins, Hana Novacenko, and Sidney Malitz
Four dijqrerentmethods of quantiJ)ing the I-mg Dexamethasone Suppression Test (DST) were contrasted with serial testing in endogenous depressives receiving electroconvulsive therapy (ECTj. Of three contin~us measures in 38 patients with pre~eatment DSTs, only the log*tra~formed value for plasma cortisol was normally distributed, i~icating that it possessed superior psychometric properties. Pretreatment Hamilton Depression Rating Scores (HAM-D) correlated positively with pretreatment DST status, with a similar association noted between posttreatment DST status and HAM-D scores. There was no uniform eflect of ECT on the DST. Although pretreatment nonsuppressors showed a trend toward decreased postdexamet~sone co&sol values, initial s~pressors (cuto~: 5~g~dl) evidenced a significant increase in these values, and 35.3% of initial suppressors were nonsuppressors atjnul DST assessment. These trends were noted in the DST assessment done following the third ECT treatment, suggesting an efSect of regression to the mean. The findings highlight the importance of following initial DST suppressors in studies of this type.
Introduction A series of claims have been offered about the clinical significance of the Dexamethasone Suppression Test (DST) in depressed patients. These include: (1) high specificity and sensitivity in melancholia (Brown and Shuey 1980; Carroll et al. 1981); (2) high postdexamethasone plasma cortisol concentration at pretreatment predicts strong therapeutic improvement with antidepressants (Coppen et al. 1985); (3) severity of depressive symptomatology and maximum postdexamethasone plasma cortisol levels are associated in nonsuppressors (Greden et al. 1983); (4) DST nonsuppressors normalize progressively with clinical improvement, with “normalization” coinciding with or preceding by one or more weeks m~ifes~tion of euthymia (Greden et al. 1983, 1985); and (5) a ~~istently
From the Departmentof Biological Psychiatry, New York State Psychic Institute(D.P.D.); and Departmentof Psychiatry, College of Physicians aod Surgeons. Columbia University; and the Departmentof Psychology, New York University (H.A.S.). New York. NY. Address rep&t requestak Dr. D. P. Devaoand, Departmentof SioIogkaI Psycbiatty, New York State Psychiatric Institote, 722 West 168th Street, New York, NY 10032. Received March4, 1986; revised September 11, 1986.
8 1987 Society of Biological Psychiatry
GQO6-3223/871.$03.50
464
BIOL PSYCHIATRY 1987;22:463-472
D.P. Devanand et al.
abnormal DST is associated either with poor immediate outcome (Greden et al. 1985) or may predict early relapse (Papakostas et al. 1981; Insel and Goodwin 1983). Research on the clinical value of the DST has shifted from its diagnostic utility to its use as a prognostic tool, with several studies supporting this approach in patients treated with antidepressants (Holsboer et al. 1982; Greden et al. 1983, 1985), although a few reports have questioned the predictive value of serial DSTs (Amsterdam et al. 1981; Peselow et al. 1983). In contrast to the work with antidepressants, there are few studies that have reported on serial DSTs in depressed patients receiving electroconvulsive therapy (ECT). Initial reports suggested that DST nonsuppression changes to DST suppression prior to clinical response (Albala et al. 1981) and that the greater the decrease in postdexamethasone cortisol levels from pretreatment to posttreatment, the better the clinical outcome (Papakostas et al. 198 1). Zelnik et al. (1985) reported that failure to normalize was associated with relapse, although 27% of patients showed oscillating rhythms of postdexamethasone cortisol levels during ECT. On the other hand, Coryell (1982) found that whereas pretreatment nonsuppression predicted good clinical outcome rated globally, it was not related to HAM-D scores at discharge. In addition, 6-month follow-up did not reveal differences in clinical course between pretreatment suppressors and nonsuppressors in a group of melancholic patients treated with ECT. Qualls and Brown (1982) found that 100% of initial suppressors responded to ECT, whereas only 50% of their initial nonsuppressors were responders. Preliminary data from our group found that postdexamethasone cortisol levels increased during ECT in both pretreatment suppressors and nonsuppressors and remained elevated at posttreatment (Decina et al. 1983). A recent study by Coppen et al. (1985) suggested that pretreatment DST status may have a different predictive value for patients treated with antidepressants or with ECT. They found that the pretreatment DST was significantly related to outcome in a group treated with antidepressants, with only a nonsignificant trend in this direction in the ECT group. In this study, we prospectively followed a group of primary major depressives, endogenous subtype, with serial DSTs over the course of ECT. We examined whether or not pretreatment DST values predicted short-term clinical outcome, whether pre- and/or posttreatment DST values were related to severity of depressive symptomatology, and whether or not changes in the DST during treatment were associated with clinical outcome. In addition, two central methodological issues were addressed. First, there is a lack of consensus in the field concerning the appropriate measure to be derived from postdexamethasone cortisol values. Although a cutoff value of 5 pg/dl is often used for dichotomous classification, it is acknowledged that depending on assay technique, a different value may be required (Carroll 1982). Furthermore, there is the possibility that depending on the question being addressed, different cutoff values may be optimal. Regardless of the cutoff score used, dichotomous classification of patients runs the risk of loss of statistical power relative to using the full range of postdexamethasone cortisol values (Cohen 1969). In this light, alternative techniques have involved the use of the mean of postdexamethasone cortisol values (Decina et al. 1983; Coppen et al. 1985) and the peak postdexamethasone cortisol value (Greden et al. 1985). However, as cortisol values are usually skewed in their distribution (Carroll et al. 1981), transformation of the cortisol values to achieve normality may be necessary. Here, we examine the substantive issues in reference to all four methods: cutoff score, mean of raw values. peak value, and mean of log-transformed values.
Effect of ECT on DST
BIOL PSYCHlATRY 1981:22:463-412
465
Secondly, to our knowledge, serial DST studies have followed only patients who were pretreatment nonsuppressors. As a subgroup is being selected that falls at one extreme of a distribution, some proportion of subsequent “normalization” to nonsuppression in this subgroup should reflect regression to the mean. Furthermore, treatments such as ECT can be highly effective. Consequently, the predominant patterns may be both DST normalization and clinical recovery, without there being a meaningful association between them. For these reasons, we followed both initial suppressors and nonsuppressors.
Meth ods Patients with major depressive disorder, endogenous subtype (Research Diagnostic Criteria-RDC) participated in a clinical trial on the efficacy and side effects of ECT. RDC diagnoses were based on a structured interview using the Schedule for Affective Disorders and Schizophrenia (SADS). To be included in the ECT trial, patients had to have a pretreatment score of 18 or more on the Hamilton Rating Scale for Depression (24-item), have negative histories for organic brain syndrome and substance abuse, not have received ECT within the past year, not present with a serious concomitant medical condition, and provide informed consent. Furthermore, to be included in the DST substudy, patients had to meet the inclusion/exclusion criteria described by Carroll et al. (1981). With the exception of lorazepam (1 mg p.r.n. every 12 hr), patients were free of psychotropic medication for a minimum period of 5 days preceding ECT and did not receive psychotropic medication until following the last posttreatment DST. ECT was administered three times a week using a low-dosage titration procedure (Malitz et al. 1985). Constant current, bidirectional, square wave, brief pulse stimulation (MECTA) was used with the deliberate administration of subconvulsive electrical intensities to determine seizure threshold. Electrical dosage at seizure was kept minimally above seizure threshold throughout the course. A minimum seizure duration of 25 set in motor manifestations was required at each treatment session. Patients were randomized to either right unilateral (d’Elia) ECT or bilateral (frontotemporal) ECT. A minimum of 10 treatments was required prior to designation as a nonresponder to the particular modality. Otherwise, treatment length was determined naturalistically by a blind evaluation team on the basis of clinical response. The clinical evaluation team completed HAM-D ratings at pretreatment and the afternoon following each ECT and during the week following the end of the treatment course. The clinical raters were blind to all DST results. Criteria for classification as ECT responder were a minimum 60% decrease from pm- to posttreatment in HAM-D scores, a maximum posttreatment score of 16, and maintenance of the minimum 60% reduction for at least 1 week posttreatment, while remaining free of psychotropic medication. DST assessment involved an oral dose of 1 mg dexamethasone given at 1l:OO PM, with blood drawn for plasma cortisol the following day at 8:OOAM, 4:00 PM, and 11:OO PM. Plasma cortisol was assayed by adapting the competitive protein binding method described by Murphy (1967) with the modifications used by Carroll and Curtis (1976) and others (Novacenko et al. 1980). Intraassay coefficients of variation were 5.7% at cortisol value 3.7 pg/dl, 3.1% at 12.5 &dl, and 3.3% at 21.3 pg/dl; whereas interassay coefficients of variation were 6.9% at 3.7 Pg/dl, 4.6% at 12.5 pg/dl, and 5.3% at 21.3 l&d1 for this laboratory during the period of study. All assays for a particular patient were run in the same batch. Pretreatment DSTs were obtained in 38 patients. Table 1 presents demographic and
466
BlOL PSYCHIATRY 1987:22:463-472
Table 1. Demographic
D.P. Devanand et al.
and Psychiatric
Characteristics
Total sample Mean (SD) Initial n = 38
4x-e Sex (%female) Number of hospitalizations Number of depressive
of the Sample Pretreatment nonsuppressors Mean (SD)
Initial n = 20
Pretreatment suppressors Mean (SD) Initial n = 18
61.34 (13.08) 57.90 2.61 (2.96)
62.08 (12.49) 54.20 2.92 (3.46)
60.07 (13.95) 64.30 2.07 (1.67)
4.34 (6.48)
3.96 (2.99)
5.00 (9.89)
43.63 (16.89)
43.38 (16.48)
44.07 (17.54)
37.05 (29.21)
34.58 (28.60)
41.29 (29.76)
78.95% 13.27 (10.71)
75.00% 15.00 (14.52)
85.71% 1 I .44 (10.48)
9.29 (2.42)
9.79 (2.00)
8.43 (2.80)
31.29 (8.11)
33.20 (8.48)
29.44 (7.60)
episodes
Age at first episode Length of current episode (weeks) Unipolar Days from admission to first DST” Number of ECT treatments Pretreatment
HAM-D scores "A maximum 30.day interval was permitted.
psychiatric characteristics of the sample. As shown in Table 1, mean number of days from admission to the first DST assessment was 13.27 days, with no significant difference between pretreatment suppressors and nonsuppressors, suggesting that stress of hospitalization was not an important factor. In addition, mean number of days of psychotropic medication washout, other than sedative/hypnotics, was 14.3 (SD 10.72) days for the total sample and was comparable in the suppressor (mean 13.7 days, SD 9.80) and nonsuppressor groups (mean 14.6 days, SD 30.8), using an upper limit of 30 days duration for each patient. Subsequent DSTs were administered weekly (after every third ECT), and during the week following the last treatment while still medication free. Patients were studied 1 week post-ECT to rule out extremely transient clinical response and to avoid purely acute effects of ECT on the DST. In some patients, either lack of continuing patient cooperation or logistical problems resulted in loss of DST data during the ECT course and at posttreatment. At least one DST assessment during the course of ECT was available for 32 patients, and posttreatment data were available for 24 patients obtained drug-free during the week following ECT. The bulk of loss at posttreatment was a function of the ECT course ending within a few days following the last repeat DST and the rapid institution of alternative somatic treatment. No additional DSTs were conducted in three patients who had pretreatment DSTs. An a priori decision was taken prior to data analysis to include cortisol values as data if results from at least 2 of 3 samples (8:OOAM, 4:00 PM, and 1l:OO PM) drawn on that day were available. The clinical ECT trial was unusual in its method of seizure induction by titrating electrical intensities to just above seizure threshold. A marked difference in the efficacy
BIOL PSYCHIATRY 1987;22:463-472
Effect of ECT on DST
467
of bilateral and right unilateral ECT was obtained (Malitz and Sackeim 1986) in favor of bilateral ECT. Likewise, in the subsample reported on here, 10 of 16 bilateral ECT patients (62.5%) were classified as responders, whereas only 6 of 22 unilateral ECT patients (27.3%) were so classified, x*(l) = 4.72, p = 0.03. Consequently, all effects reported on here were examined separately for patients within each treatment condition. All significance levels were based on two-tailed tests.
Results At pretreatment, 20 of the 38 patients (52.6%) were classified as nonsuppressors using the traditional cutoff of 5.0 cLg/dl.pretreatment nonsuppressors (33.20 + 8.48, mean + SD) and suppressors (29.44 + 7.60) did not differ in HAM-D scores. At posttreatment, 11 of 24 patients (45.8%) were classified as nonsuppressors. At this time point, patients who were nonsuppressors (19.36 + 12.94) tended to have higher Hamilton scores than posttreatment suppressors (10.85 + 10.09), t = 1.73, df = 22, p = 0.098. Table 2 presents descriptive statistics, tests against normality, and correlations with Hamilton scores for the three continuous measures of cortisol secretion at pre- and posttreatment. At pretreatment, the distribution for both the mean of the three raw cortisol values and the peak value significantly departed from normality. The mean of the log-transformed cortisol values was normally distributed both at pre- and posttreatment. At pretreatment, all three measures showed significant associations with HAM-D scores. The same level of correlation was observed at posttreatment, although it did not reach significance in the case of the log-transformed measure, which was most likely due to the reduced sample size. These analyses suggest that the dichotomous classification was less sensitive than use of continuous measures to relations between postdexamethasone cortisol secretion and severity of depressive illness. Furthermore, the log-transformed measure was superior in not violating assumptions of normality.
Table 2. PostdexarnethasoneCortisol Secretion: Descriptive Statistics, Tests against Normality, and Correlationswith Hamilton Scores Mean cortisol Pretreatment (n = 38) MGUI SD
Skew Kurtosis r with PIE-HAM-D r with post-HAM-D r with change in HAM-D Posttreatment (n = 24) Mean SD
Skew Kurtosis r with post-HAM-D “Departsfrom normality. “p < 0.05.
‘p < 0.01. “p < 0.001.
5.28 4.92 1.36”~~ 1.61”.b 0.42’ 0.23 0.04 5.12 4.59 0.74 -1.12 0.43*
Peak cortisol
Mean log cortisol
7.44 6.40 0.87”” -0.14 0.41’ 0.18 0.07
0.64 0.34 0.05 -0.75 0.34b 0.01 0.07
7.50 6.12 0.56 - 1.10 0.436
0.63 0.33 0.30 -1.13 0.38
468
BIOL PSYCHIATRY 1987:22:463-472
Predictors
D.P. Devanand et al.
of Outcome
To examine whether or not pretreatment DST values predicted clinical outcome, for each of the three continuous DST measures, correlations were computed with both the posttreatment Hamilton scores and the percentage reduction in Hamilton scores (see Table 2). Dichotomously classified pretreatment suppressors and nonsuppressors were also compared on these two outcome measures. Analyses conducted within and across ECT treatment modalities (unilateral and bilateral) revealed no significant effects. Likewise, no significant effects emerged when clinical outcome was based on dichotomous classification as responder or nonresponder. For example, 8 of 20 (40%) initial nonsuppressors were responders, whereas 7 of 18 (39%) initial suppressors were responders. These analyses indicated that pretreatment DST status was not predictive of response to ECT, regardless of DST measure or ECT modality.
Effects of ECT To examine effects of ECT on the DST, for each of the three continuous measures, a mean was taken of all DST assessments during the ECT course. Repeated measures Analyses of Variance (ANOVAs) were conducted that contrasted initial suppressors and nonsuppressors in DST values at pretreatment, during treatment, and posttreatment (initial classification, 2 levels, X time, 3 levels). An ANOVA was conducted on each of the three continuous DST measures. All three ANOVAs revealed a main effect of initial classification, with all ps cO.005. As expected, cortisol values were higher for initial nonsuppressors. There were no main effects of time, indicating that receiving ECT, by itself, had no uniform effect on cortisol values. However, in all three ANOVAs, there was a significant interaction between initial classification (suppressor or nonsuppressor) and time (pre- versus post-ECT), with all ps cO.05. Figure 1 illustrates the interaction in the case of the log-transformed values. During and following the ECT course, initial nonsuppressors tended to have lower values than at pretreatment. In contrast, initial suppressors tended to change toward the direction of nonsuppression. For each measure, paired t-tests revealed that in initial suppressors, the increase from pretreatment to during ECT was significant (all ps ~0.05). The increase in this group from pre- to posttreatment was marginally significant for the measure of the mean of raw cortisol values and the mean of log-transformed values (both ps CO. 1) and was significant for the peak value (p < 0.05). The smaller decreases seen in the initial nonsuppressor group did not reflect statistically significant change. Nonparametric analyses supported these findings. Of 15 initial suppressors, 13 (86.7%) increased in values (all measures) from pretreatment to the mean of assessments during ECT. This pattern characterized only 8 of 17 (47.1%) initial nonsuppressors, x’(l) = 5.54, p = 0.02. Additional analyses examined possible effects of modality. At pretreatment, equal rates of nonsuppression were obtained in the bilateral group (8 of 15, 53.3%) and in the unilateral group (12 of 23, 52.2%). Introducing modality as a factor in the ANOVA did not result in any new significant main effects or interactions. Both groups showed the pattern of increased cortisol values in initial suppressors and decreased values in initial nonsuppressors. However, nonparametric analyses were suggestive that the pattern was somewhat more pronounced in the unilateral group. At posttreatment, 4 of 7 initial suppressors abnormalized in the unilateral group, whereas 1 of 6 initial suppressors did so in the bilateral group.
BIOL PSYCHIATRY 1987:22:46?-472
Effect of ECT on DST
469
0.80.72 $
0.6-
%
Figure 1. Mean cortisol values (log-
“, 0.50” 2 0” 4
0.4-
z 2”
0.3-
transformed) at pretreatment, during the ECT course, and at posttreatment as a function of pretreatment status as DST suppressoror nonsuppressor.
0
0.2-
0. I 0
Initial Non- suppressors
0 Initial
1
fre ECT
Suppressors
During ECT
Pas; ECT
The basic pattern of initial suppressors increasing in postdexamethasone cortisol values, while initial nonsuppressors decreased, could reflect an artifact of regression to the mean, an acute effect of ECT, or a pattern of change related to fluctuation in clinical state. Changes in cortisol levels were examined from pretreatment to following the third ECT (n = 32), the first repeat DST following pretreatment evaluation. By this time point, the pattern of opposite directions of change had developed in initial suppressors and nonsuppressors. However, the interaction term in the ANOVAs only reached the trend level in ANOVAs on mean of raw values and peak values (both ps CO. 1) and was significant for the log-transformed values @ < 0.05). These analyses provided some support for the view that the pattern reflected an effect of regression to the mean. DST Change and Clinical Outcome It has been claimed that percentage reduction in Hamilton scores is associated with the magnitude of change from pretreatment to posttreatment in quantitative DST values (Papakostas et al. 1981; Coppen et al. 1985). In this sample, a trend in this direction was obtained, with the correlation ranging from 0.35 to 0.39 (0.09 > all ps > 0.05) for the three measures. It was noted, however, that within the right unilateral group, the correlation ranged from 0.54 to 0.66 @s < 0.05, n = 13) and was opposite in direction for the bilateral group (- 0.22 to - 0.30, NS, n = 11). The differences between these
470
BIOL PSYCHIATRY 1987;22:463-472
D.P. Devanand et al.
two groups in these correlations were all significant (p < 0.05).
To examine whether or not the changes in values were associated with clinical outcome using the 5.0 pg/dl cutoff, patients were classified into four groups on the basis of their initial and final DST assessments. Of 18 initial nonsuppressors, only 6 showed normalization. Of 17 initial suppressors, 6 showed abnormalization. The small numbers in these subgroups precluded statistical analysis. Nonetheless, it might be noted that four out of six patients who normalized were classified as responders, whereas only two of six patients who abnormalized were so classified.
Discussion TO summarize the findings, four different methods of quantifying the DST were examined with serial testing in depressed patients receiving ECT. Three continuous measures of postdexamethasone cortisol secretion-mean of raw cortisol values, peak cortisol value, and mean of log-transformed cortisol values-all showed positive correlations with depression severity scores, both at pre- and posttreatment. The traditional dichotomous classification of patients as suppressors and nonsuppressors was not significantly associated with depression scores at pre- or posttreatment, presumably due to the loss of power resulting from discrete categorization. Of the three continuous measures, only the logtransformed score was normally distributed both at pre- and posttreatment assessments, suggesting that this measure possesses superior psychometric properties. No evidence was obtained that the pretreatment DST was useful in predicting immediate clinical outcome of ECT. This negative finding concurs with the reports of Coppen et al. (1985) and Lipman et al. (1985) and is in partial agreement with the results of Coryell (1982). There was no uniform effect of ECT on postdexamethasone cortisol values. However, during and following the ECT course, initial suppressors evidenced a significant increase in these values. Pretreatment nonsuppressors showed a nonsignificant trend toward decreased values. The change toward higher values in pretreatment suppressors was highly consistent, and 35.3% of initial suppressors were nonsuppressors at their final DST assessment. This pattern could be an artifact of regression to the mean, may be due to an acute effect of ECT, and/or may be related to changes in clinical state. As the pattern was seen as early as following the third ECT (the first reassessment) and the groups chosen for being at the extremes of a distribution moved closer together, this suggests that regression to the mean may have played a role. This phenomenon should be studied further by repeating DSTs prior to the start of somatic treatment. That regression to that mean could not account completely for the changes observed in DST values was evidenced by the associations at posttreatment between depression ratings and cortisol values and by the association between changes in cortisol values from pre- to posttreatment and change in clinical state. These associations indicated that degree of reduction in cortisol values was related to degree of clinical improvement. Interestingly, the relation with degree of clinical improvement was observed only in the right unilateral group and not in the bilateral ECT group. This raises the possibility that ECT electrode placements may differ in their impact on the DST. It has been speculated that bilateral ECT may have more pronounced effects on hypothalamic functioning than right unilateral ECT (e.g., Abrams and Swartz 1985). Previous research on the serial DST in patients treated with ECT has not examined effects of electrode placement. This study was conducted with an unusual method of ECT administration. The extent
Effect of ECT on DST
BIOL PSYCHIATRY t987:22:463-472
471
which the DST findings were a function of utilizing a titrated, low-dosage ECT technique is unknown. The serial DST findings reported here should be reexamined with traditional, higher dosage ECT techniques. In this sample, there was a major difference in the efficacy of unilateral and bilateral ECT. Furthermore, there was an indication that the two modalities differed in effects on the DST, with the right unilateral group displaying an association between serial postdexamethasone cortisol values and changes in symptomatology. The fact that unilateral ECT was particularly weak in antidepressant effects may partly account for the association within this group between cortisol values and symptom change. Overall, the findings highlight the need to follow pretreatment suppressors in future serial DST studies. To our knowledge, this is the first report that a substantial proportion of pretreatment suppressors abnormalize with somatic treatment. There were indications here that such an abnormalization may be associated with poor clinical outcome. Regardless of its possible clinical significance, in this type of research, following initial suppressors is necessary to control for the effect of regression to the mean when using cutoff cortisol values or avoiding restriction of range in examining continuous measures of postdexamethasone cortisol secretion.
to
Supported in part by NlMH Grant MH35636.
References Abrams R, Swartz CM (1985): ECT and prolactin release. 1. Effects of stimulus parameters. Convul Ther 1:115-119. Albala AA, Greden JF, Tarika J, Carroll BJ (1981): Changes in serial Dexamethasone Suppression Tests among unipolar depressives receiving electroconvulsive treatment. Biol Psychiatry 16:X51-560. Amsterdam JD, Winokur A, Caroff S (1981): Dexamethasone Suppression Test as a prognostic tool: 2 case reports. Am J Psychiatry 138:979-980. Brown WA, Shuey I (1980): Response to dexamethasone and subtype of depression. Arch Gen Psychiatry 37:747-75 1. Carroll BJ (1982): The Dexamethasone Suppression Test for melancholia. Br J Psychiatry 140:292-304. Carroll BJ, Curtis GC (1976): Neuroendocrine identification of depressed patients. Aust NZ J Psychiatry 10: 13-20. Carroll BJ, Feinberg M, Greden JF, et al (1981): A specific laboratory test for the diagnosis of melancholia. Arch Gen Psychiatry 38: 15-22. Cohen J (1969): Statistical Power Analysis for the Behavioral Sciences. Orlando, FL: Academic. Coppen A, Milln P, Harwood J, Wood K (1985): Does the Dexamethasone Suppression Test predict antidepressant treatment success? Br J Psychiatry 146:294-296. Coryell W (1982): Hypothalamic+uitary-adrenal axis and ECT response. Psychiatry Res 6:283-292. Coryell W, Zimmerman M (1983): The Dexamethasone Suppression Test and ECT outcome: A six-month follow-up. Biol Psychiatry 18:21-27. Decina P, Sackeim HA, Malitz S (1983): Prognostic value of serial Dexamethasone Suppression Tests during and following ECf. Psychopharmacol Bull 19:85-87. Greden JF, Gardner R, King D, Grunhaus L, Carroll BJ, Kronfol Z (1983): Dexamethasone Suppression Tests in antidepressant treatment of melancholia: The process of normalization and test-retest reproducibility. Arch Gen Psychiatry 401493-500.
472
BIOL PSYCHIATRY 1987:22:463-412
D.P. Devanand et al.
Greden JF, Haskett R, Zelnik T, Grunhaus L, Tiongco D (1985): Serial hypothalamicpituitary-adrenal monitoring in depressed patients. Presented at the Society of Biological Psychiatry Meeting, Dallas, TX, May 15-19. Holsboer F, Steiger A, Hofschuster E (1982): Repeated Dexamethasone depressive illness. J Affect Dis 4:93-101.
Suppression
Test during
lnsel TR, Goodwin FK (1983): The Dexamethasone Suppression Test: Promises and problems of diagnostic laboratory tests in psychiatry. Hasp Commun Psychiatry 34: 1134-1138. Lipman RS, Backup C, Bobrin Y, et al (1985): The Dexamethasone Suppression Test and clinical outcome in hospitalized depressives treated with ECT. Presented at the Annual Meeting of the National Association of Private Psychiatric Hospitals, Marco Island, FL, January 30. Malitz S, Sackeim HA (1986): Electroconvulsive NY Acad Sci 46256-65.
therapy: Clinical and basic research issues. Ann
Malitz SM, Sackeim HA, Decina P, Kanzler M, Kerr B (1985): The efficacy of ECT: Dose-response interactions with modality. Murphy BEP (1967): Some studies of protein binding of steroids and their application to the routine micro and ultramicro measurement of various steroids in body fluids by competitive protein binding radioassay. J Clin Endocrinol Metab 271973-990. Novacenko H, Halpem FS, Sachar EJ (1980): An improved cortisol. Clin Chem 26:312-3 13. Papakostas Y, Fink M, Lee J, et al (198 1): Neuroendocrine and outcome with ECT. Psychiatry Res 455-64.
protein binding assay for plasma
measures in psychiatric patients: Course
Peselow Y, Goldring N, Fieve R, Wright R (1983): The Dexamethasone Suppression Test during depression and following recovery from a depressive episode. Psychopharmacol Bull 19:87-91. Qualls CB, Brown WA (1982): Dexamethasone Presented at the New Research Proceedings, Abst no. NR 36.
Suppression Test subgroups differ in ECT response. American Psychiatric Association Annual Meeting,
Zelnik T, Albala AA, Haskett R, Grunhaus L, Greden JF (1985): Serial neuroendocrine patterns in depressives treated only with electroconvulsive therapy. Presented at the Society of Biological Psychiatry Meeting, Dallas, TX, May 15-19.
463
Serial Dexamethasone Suppression Tests in Initial Suppressors and Nonsuppressors Treated with ~le~tro~onvulsiv~ Therapy D. P. Devanand, Paolo Decina, Harold A. Sackeim, Nancy Hopkins, Hana Novacenko, and Sidney Malitz
Four dijqrerentmethods of quantiJ)ing the I-mg Dexamethasone Suppression Test (DST) were contrasted with serial testing in endogenous depressives receiving electroconvulsive therapy (ECTj. Of three contin~us measures in 38 patients with pre~eatment DSTs, only the log*tra~formed value for plasma cortisol was normally distributed, i~icating that it possessed superior psychometric properties. Pretreatment Hamilton Depression Rating Scores (HAM-D) correlated positively with pretreatment DST status, with a similar association noted between posttreatment DST status and HAM-D scores. There was no uniform eflect of ECT on the DST. Although pretreatment nonsuppressors showed a trend toward decreased postdexamet~sone co&sol values, initial s~pressors (cuto~: 5~g~dl) evidenced a significant increase in these values, and 35.3% of initial suppressors were nonsuppressors atjnul DST assessment. These trends were noted in the DST assessment done following the third ECT treatment, suggesting an efSect of regression to the mean. The findings highlight the importance of following initial DST suppressors in studies of this type.
Introduction A series of claims have been offered about the clinical significance of the Dexamethasone Suppression Test (DST) in depressed patients. These include: (1) high specificity and sensitivity in melancholia (Brown and Shuey 1980; Carroll et al. 1981); (2) high postdexamethasone plasma cortisol concentration at pretreatment predicts strong therapeutic improvement with antidepressants (Coppen et al. 1985); (3) severity of depressive symptomatology and maximum postdexamethasone plasma cortisol levels are associated in nonsuppressors (Greden et al. 1983); (4) DST nonsuppressors normalize progressively with clinical improvement, with “normalization” coinciding with or preceding by one or more weeks m~ifes~tion of euthymia (Greden et al. 1983, 1985); and (5) a ~~istently
From the Departmentof Biological Psychiatry, New York State Psychic Institute(D.P.D.); and Departmentof Psychiatry, College of Physicians aod Surgeons. Columbia University; and the Departmentof Psychology, New York University (H.A.S.). New York. NY. Address rep&t requestak Dr. D. P. Devaoand, Departmentof SioIogkaI Psycbiatty, New York State Psychiatric Institote, 722 West 168th Street, New York, NY 10032. Received March4, 1986; revised September 11, 1986.
8 1987 Society of Biological Psychiatry
GQO6-3223/871.$03.50
464
BIOL PSYCHIATRY 1987;22:463-472
D.P. Devanand et al.
abnormal DST is associated either with poor immediate outcome (Greden et al. 1985) or may predict early relapse (Papakostas et al. 1981; Insel and Goodwin 1983). Research on the clinical value of the DST has shifted from its diagnostic utility to its use as a prognostic tool, with several studies supporting this approach in patients treated with antidepressants (Holsboer et al. 1982; Greden et al. 1983, 1985), although a few reports have questioned the predictive value of serial DSTs (Amsterdam et al. 1981; Peselow et al. 1983). In contrast to the work with antidepressants, there are few studies that have reported on serial DSTs in depressed patients receiving electroconvulsive therapy (ECT). Initial reports suggested that DST nonsuppression changes to DST suppression prior to clinical response (Albala et al. 1981) and that the greater the decrease in postdexamethasone cortisol levels from pretreatment to posttreatment, the better the clinical outcome (Papakostas et al. 198 1). Zelnik et al. (1985) reported that failure to normalize was associated with relapse, although 27% of patients showed oscillating rhythms of postdexamethasone cortisol levels during ECT. On the other hand, Coryell (1982) found that whereas pretreatment nonsuppression predicted good clinical outcome rated globally, it was not related to HAM-D scores at discharge. In addition, 6-month follow-up did not reveal differences in clinical course between pretreatment suppressors and nonsuppressors in a group of melancholic patients treated with ECT. Qualls and Brown (1982) found that 100% of initial suppressors responded to ECT, whereas only 50% of their initial nonsuppressors were responders. Preliminary data from our group found that postdexamethasone cortisol levels increased during ECT in both pretreatment suppressors and nonsuppressors and remained elevated at posttreatment (Decina et al. 1983). A recent study by Coppen et al. (1985) suggested that pretreatment DST status may have a different predictive value for patients treated with antidepressants or with ECT. They found that the pretreatment DST was significantly related to outcome in a group treated with antidepressants, with only a nonsignificant trend in this direction in the ECT group. In this study, we prospectively followed a group of primary major depressives, endogenous subtype, with serial DSTs over the course of ECT. We examined whether or not pretreatment DST values predicted short-term clinical outcome, whether pre- and/or posttreatment DST values were related to severity of depressive symptomatology, and whether or not changes in the DST during treatment were associated with clinical outcome. In addition, two central methodological issues were addressed. First, there is a lack of consensus in the field concerning the appropriate measure to be derived from postdexamethasone cortisol values. Although a cutoff value of 5 pg/dl is often used for dichotomous classification, it is acknowledged that depending on assay technique, a different value may be required (Carroll 1982). Furthermore, there is the possibility that depending on the question being addressed, different cutoff values may be optimal. Regardless of the cutoff score used, dichotomous classification of patients runs the risk of loss of statistical power relative to using the full range of postdexamethasone cortisol values (Cohen 1969). In this light, alternative techniques have involved the use of the mean of postdexamethasone cortisol values (Decina et al. 1983; Coppen et al. 1985) and the peak postdexamethasone cortisol value (Greden et al. 1985). However, as cortisol values are usually skewed in their distribution (Carroll et al. 1981), transformation of the cortisol values to achieve normality may be necessary. Here, we examine the substantive issues in reference to all four methods: cutoff score, mean of raw values. peak value, and mean of log-transformed values.
Effect of ECT on DST
BIOL PSYCHlATRY 1981:22:463-412
465
Secondly, to our knowledge, serial DST studies have followed only patients who were pretreatment nonsuppressors. As a subgroup is being selected that falls at one extreme of a distribution, some proportion of subsequent “normalization” to nonsuppression in this subgroup should reflect regression to the mean. Furthermore, treatments such as ECT can be highly effective. Consequently, the predominant patterns may be both DST normalization and clinical recovery, without there being a meaningful association between them. For these reasons, we followed both initial suppressors and nonsuppressors.
Meth ods Patients with major depressive disorder, endogenous subtype (Research Diagnostic Criteria-RDC) participated in a clinical trial on the efficacy and side effects of ECT. RDC diagnoses were based on a structured interview using the Schedule for Affective Disorders and Schizophrenia (SADS). To be included in the ECT trial, patients had to have a pretreatment score of 18 or more on the Hamilton Rating Scale for Depression (24-item), have negative histories for organic brain syndrome and substance abuse, not have received ECT within the past year, not present with a serious concomitant medical condition, and provide informed consent. Furthermore, to be included in the DST substudy, patients had to meet the inclusion/exclusion criteria described by Carroll et al. (1981). With the exception of lorazepam (1 mg p.r.n. every 12 hr), patients were free of psychotropic medication for a minimum period of 5 days preceding ECT and did not receive psychotropic medication until following the last posttreatment DST. ECT was administered three times a week using a low-dosage titration procedure (Malitz et al. 1985). Constant current, bidirectional, square wave, brief pulse stimulation (MECTA) was used with the deliberate administration of subconvulsive electrical intensities to determine seizure threshold. Electrical dosage at seizure was kept minimally above seizure threshold throughout the course. A minimum seizure duration of 25 set in motor manifestations was required at each treatment session. Patients were randomized to either right unilateral (d’Elia) ECT or bilateral (frontotemporal) ECT. A minimum of 10 treatments was required prior to designation as a nonresponder to the particular modality. Otherwise, treatment length was determined naturalistically by a blind evaluation team on the basis of clinical response. The clinical evaluation team completed HAM-D ratings at pretreatment and the afternoon following each ECT and during the week following the end of the treatment course. The clinical raters were blind to all DST results. Criteria for classification as ECT responder were a minimum 60% decrease from pm- to posttreatment in HAM-D scores, a maximum posttreatment score of 16, and maintenance of the minimum 60% reduction for at least 1 week posttreatment, while remaining free of psychotropic medication. DST assessment involved an oral dose of 1 mg dexamethasone given at 1l:OO PM, with blood drawn for plasma cortisol the following day at 8:OOAM, 4:00 PM, and 11:OO PM. Plasma cortisol was assayed by adapting the competitive protein binding method described by Murphy (1967) with the modifications used by Carroll and Curtis (1976) and others (Novacenko et al. 1980). Intraassay coefficients of variation were 5.7% at cortisol value 3.7 pg/dl, 3.1% at 12.5 &dl, and 3.3% at 21.3 pg/dl; whereas interassay coefficients of variation were 6.9% at 3.7 Pg/dl, 4.6% at 12.5 pg/dl, and 5.3% at 21.3 l&d1 for this laboratory during the period of study. All assays for a particular patient were run in the same batch. Pretreatment DSTs were obtained in 38 patients. Table 1 presents demographic and
466
BlOL PSYCHIATRY 1987:22:463-472
Table 1. Demographic
D.P. Devanand et al.
and Psychiatric
Characteristics
Total sample Mean (SD) Initial n = 38
4x-e Sex (%female) Number of hospitalizations Number of depressive
of the Sample Pretreatment nonsuppressors Mean (SD)
Initial n = 20
Pretreatment suppressors Mean (SD) Initial n = 18
61.34 (13.08) 57.90 2.61 (2.96)
62.08 (12.49) 54.20 2.92 (3.46)
60.07 (13.95) 64.30 2.07 (1.67)
4.34 (6.48)
3.96 (2.99)
5.00 (9.89)
43.63 (16.89)
43.38 (16.48)
44.07 (17.54)
37.05 (29.21)
34.58 (28.60)
41.29 (29.76)
78.95% 13.27 (10.71)
75.00% 15.00 (14.52)
85.71% 1 I .44 (10.48)
9.29 (2.42)
9.79 (2.00)
8.43 (2.80)
31.29 (8.11)
33.20 (8.48)
29.44 (7.60)
episodes
Age at first episode Length of current episode (weeks) Unipolar Days from admission to first DST” Number of ECT treatments Pretreatment
HAM-D scores "A maximum 30.day interval was permitted.
psychiatric characteristics of the sample. As shown in Table 1, mean number of days from admission to the first DST assessment was 13.27 days, with no significant difference between pretreatment suppressors and nonsuppressors, suggesting that stress of hospitalization was not an important factor. In addition, mean number of days of psychotropic medication washout, other than sedative/hypnotics, was 14.3 (SD 10.72) days for the total sample and was comparable in the suppressor (mean 13.7 days, SD 9.80) and nonsuppressor groups (mean 14.6 days, SD 30.8), using an upper limit of 30 days duration for each patient. Subsequent DSTs were administered weekly (after every third ECT), and during the week following the last treatment while still medication free. Patients were studied 1 week post-ECT to rule out extremely transient clinical response and to avoid purely acute effects of ECT on the DST. In some patients, either lack of continuing patient cooperation or logistical problems resulted in loss of DST data during the ECT course and at posttreatment. At least one DST assessment during the course of ECT was available for 32 patients, and posttreatment data were available for 24 patients obtained drug-free during the week following ECT. The bulk of loss at posttreatment was a function of the ECT course ending within a few days following the last repeat DST and the rapid institution of alternative somatic treatment. No additional DSTs were conducted in three patients who had pretreatment DSTs. An a priori decision was taken prior to data analysis to include cortisol values as data if results from at least 2 of 3 samples (8:OOAM, 4:00 PM, and 1l:OO PM) drawn on that day were available. The clinical ECT trial was unusual in its method of seizure induction by titrating electrical intensities to just above seizure threshold. A marked difference in the efficacy
BIOL PSYCHIATRY 1987;22:463-472
Effect of ECT on DST
467
of bilateral and right unilateral ECT was obtained (Malitz and Sackeim 1986) in favor of bilateral ECT. Likewise, in the subsample reported on here, 10 of 16 bilateral ECT patients (62.5%) were classified as responders, whereas only 6 of 22 unilateral ECT patients (27.3%) were so classified, x*(l) = 4.72, p = 0.03. Consequently, all effects reported on here were examined separately for patients within each treatment condition. All significance levels were based on two-tailed tests.
Results At pretreatment, 20 of the 38 patients (52.6%) were classified as nonsuppressors using the traditional cutoff of 5.0 cLg/dl.pretreatment nonsuppressors (33.20 + 8.48, mean + SD) and suppressors (29.44 + 7.60) did not differ in HAM-D scores. At posttreatment, 11 of 24 patients (45.8%) were classified as nonsuppressors. At this time point, patients who were nonsuppressors (19.36 + 12.94) tended to have higher Hamilton scores than posttreatment suppressors (10.85 + 10.09), t = 1.73, df = 22, p = 0.098. Table 2 presents descriptive statistics, tests against normality, and correlations with Hamilton scores for the three continuous measures of cortisol secretion at pre- and posttreatment. At pretreatment, the distribution for both the mean of the three raw cortisol values and the peak value significantly departed from normality. The mean of the log-transformed cortisol values was normally distributed both at pre- and posttreatment. At pretreatment, all three measures showed significant associations with HAM-D scores. The same level of correlation was observed at posttreatment, although it did not reach significance in the case of the log-transformed measure, which was most likely due to the reduced sample size. These analyses suggest that the dichotomous classification was less sensitive than use of continuous measures to relations between postdexamethasone cortisol secretion and severity of depressive illness. Furthermore, the log-transformed measure was superior in not violating assumptions of normality.
Table 2. PostdexarnethasoneCortisol Secretion: Descriptive Statistics, Tests against Normality, and Correlationswith Hamilton Scores Mean cortisol Pretreatment (n = 38) MGUI SD
Skew Kurtosis r with PIE-HAM-D r with post-HAM-D r with change in HAM-D Posttreatment (n = 24) Mean SD
Skew Kurtosis r with post-HAM-D “Departsfrom normality. “p < 0.05.
‘p < 0.01. “p < 0.001.
5.28 4.92 1.36”~~ 1.61”.b 0.42’ 0.23 0.04 5.12 4.59 0.74 -1.12 0.43*
Peak cortisol
Mean log cortisol
7.44 6.40 0.87”” -0.14 0.41’ 0.18 0.07
0.64 0.34 0.05 -0.75 0.34b 0.01 0.07
7.50 6.12 0.56 - 1.10 0.436
0.63 0.33 0.30 -1.13 0.38
468
BIOL PSYCHIATRY 1987:22:463-472
Predictors
D.P. Devanand et al.
of Outcome
To examine whether or not pretreatment DST values predicted clinical outcome, for each of the three continuous DST measures, correlations were computed with both the posttreatment Hamilton scores and the percentage reduction in Hamilton scores (see Table 2). Dichotomously classified pretreatment suppressors and nonsuppressors were also compared on these two outcome measures. Analyses conducted within and across ECT treatment modalities (unilateral and bilateral) revealed no significant effects. Likewise, no significant effects emerged when clinical outcome was based on dichotomous classification as responder or nonresponder. For example, 8 of 20 (40%) initial nonsuppressors were responders, whereas 7 of 18 (39%) initial suppressors were responders. These analyses indicated that pretreatment DST status was not predictive of response to ECT, regardless of DST measure or ECT modality.
Effects of ECT To examine effects of ECT on the DST, for each of the three continuous measures, a mean was taken of all DST assessments during the ECT course. Repeated measures Analyses of Variance (ANOVAs) were conducted that contrasted initial suppressors and nonsuppressors in DST values at pretreatment, during treatment, and posttreatment (initial classification, 2 levels, X time, 3 levels). An ANOVA was conducted on each of the three continuous DST measures. All three ANOVAs revealed a main effect of initial classification, with all ps cO.005. As expected, cortisol values were higher for initial nonsuppressors. There were no main effects of time, indicating that receiving ECT, by itself, had no uniform effect on cortisol values. However, in all three ANOVAs, there was a significant interaction between initial classification (suppressor or nonsuppressor) and time (pre- versus post-ECT), with all ps cO.05. Figure 1 illustrates the interaction in the case of the log-transformed values. During and following the ECT course, initial nonsuppressors tended to have lower values than at pretreatment. In contrast, initial suppressors tended to change toward the direction of nonsuppression. For each measure, paired t-tests revealed that in initial suppressors, the increase from pretreatment to during ECT was significant (all ps ~0.05). The increase in this group from pre- to posttreatment was marginally significant for the measure of the mean of raw cortisol values and the mean of log-transformed values (both ps CO. 1) and was significant for the peak value (p < 0.05). The smaller decreases seen in the initial nonsuppressor group did not reflect statistically significant change. Nonparametric analyses supported these findings. Of 15 initial suppressors, 13 (86.7%) increased in values (all measures) from pretreatment to the mean of assessments during ECT. This pattern characterized only 8 of 17 (47.1%) initial nonsuppressors, x’(l) = 5.54, p = 0.02. Additional analyses examined possible effects of modality. At pretreatment, equal rates of nonsuppression were obtained in the bilateral group (8 of 15, 53.3%) and in the unilateral group (12 of 23, 52.2%). Introducing modality as a factor in the ANOVA did not result in any new significant main effects or interactions. Both groups showed the pattern of increased cortisol values in initial suppressors and decreased values in initial nonsuppressors. However, nonparametric analyses were suggestive that the pattern was somewhat more pronounced in the unilateral group. At posttreatment, 4 of 7 initial suppressors abnormalized in the unilateral group, whereas 1 of 6 initial suppressors did so in the bilateral group.
BIOL PSYCHIATRY 1987:22:46?-472
Effect of ECT on DST
469
0.80.72 $
0.6-
%
Figure 1. Mean cortisol values (log-
“, 0.50” 2 0” 4
0.4-
z 2”
0.3-
transformed) at pretreatment, during the ECT course, and at posttreatment as a function of pretreatment status as DST suppressoror nonsuppressor.
0
0.2-
0. I 0
Initial Non- suppressors
0 Initial
1
fre ECT
Suppressors
During ECT
Pas; ECT
The basic pattern of initial suppressors increasing in postdexamethasone cortisol values, while initial nonsuppressors decreased, could reflect an artifact of regression to the mean, an acute effect of ECT, or a pattern of change related to fluctuation in clinical state. Changes in cortisol levels were examined from pretreatment to following the third ECT (n = 32), the first repeat DST following pretreatment evaluation. By this time point, the pattern of opposite directions of change had developed in initial suppressors and nonsuppressors. However, the interaction term in the ANOVAs only reached the trend level in ANOVAs on mean of raw values and peak values (both ps CO. 1) and was significant for the log-transformed values @ < 0.05). These analyses provided some support for the view that the pattern reflected an effect of regression to the mean. DST Change and Clinical Outcome It has been claimed that percentage reduction in Hamilton scores is associated with the magnitude of change from pretreatment to posttreatment in quantitative DST values (Papakostas et al. 1981; Coppen et al. 1985). In this sample, a trend in this direction was obtained, with the correlation ranging from 0.35 to 0.39 (0.09 > all ps > 0.05) for the three measures. It was noted, however, that within the right unilateral group, the correlation ranged from 0.54 to 0.66 @s < 0.05, n = 13) and was opposite in direction for the bilateral group (- 0.22 to - 0.30, NS, n = 11). The differences between these
470
BIOL PSYCHIATRY 1987;22:463-472
D.P. Devanand et al.
two groups in these correlations were all significant (p < 0.05).
To examine whether or not the changes in values were associated with clinical outcome using the 5.0 pg/dl cutoff, patients were classified into four groups on the basis of their initial and final DST assessments. Of 18 initial nonsuppressors, only 6 showed normalization. Of 17 initial suppressors, 6 showed abnormalization. The small numbers in these subgroups precluded statistical analysis. Nonetheless, it might be noted that four out of six patients who normalized were classified as responders, whereas only two of six patients who abnormalized were so classified.
Discussion TO summarize the findings, four different methods of quantifying the DST were examined with serial testing in depressed patients receiving ECT. Three continuous measures of postdexamethasone cortisol secretion-mean of raw cortisol values, peak cortisol value, and mean of log-transformed cortisol values-all showed positive correlations with depression severity scores, both at pre- and posttreatment. The traditional dichotomous classification of patients as suppressors and nonsuppressors was not significantly associated with depression scores at pre- or posttreatment, presumably due to the loss of power resulting from discrete categorization. Of the three continuous measures, only the logtransformed score was normally distributed both at pre- and posttreatment assessments, suggesting that this measure possesses superior psychometric properties. No evidence was obtained that the pretreatment DST was useful in predicting immediate clinical outcome of ECT. This negative finding concurs with the reports of Coppen et al. (1985) and Lipman et al. (1985) and is in partial agreement with the results of Coryell (1982). There was no uniform effect of ECT on postdexamethasone cortisol values. However, during and following the ECT course, initial suppressors evidenced a significant increase in these values. Pretreatment nonsuppressors showed a nonsignificant trend toward decreased values. The change toward higher values in pretreatment suppressors was highly consistent, and 35.3% of initial suppressors were nonsuppressors at their final DST assessment. This pattern could be an artifact of regression to the mean, may be due to an acute effect of ECT, and/or may be related to changes in clinical state. As the pattern was seen as early as following the third ECT (the first reassessment) and the groups chosen for being at the extremes of a distribution moved closer together, this suggests that regression to the mean may have played a role. This phenomenon should be studied further by repeating DSTs prior to the start of somatic treatment. That regression to that mean could not account completely for the changes observed in DST values was evidenced by the associations at posttreatment between depression ratings and cortisol values and by the association between changes in cortisol values from pre- to posttreatment and change in clinical state. These associations indicated that degree of reduction in cortisol values was related to degree of clinical improvement. Interestingly, the relation with degree of clinical improvement was observed only in the right unilateral group and not in the bilateral ECT group. This raises the possibility that ECT electrode placements may differ in their impact on the DST. It has been speculated that bilateral ECT may have more pronounced effects on hypothalamic functioning than right unilateral ECT (e.g., Abrams and Swartz 1985). Previous research on the serial DST in patients treated with ECT has not examined effects of electrode placement. This study was conducted with an unusual method of ECT administration. The extent
Effect of ECT on DST
BIOL PSYCHIATRY t987:22:463-472
471
which the DST findings were a function of utilizing a titrated, low-dosage ECT technique is unknown. The serial DST findings reported here should be reexamined with traditional, higher dosage ECT techniques. In this sample, there was a major difference in the efficacy of unilateral and bilateral ECT. Furthermore, there was an indication that the two modalities differed in effects on the DST, with the right unilateral group displaying an association between serial postdexamethasone cortisol values and changes in symptomatology. The fact that unilateral ECT was particularly weak in antidepressant effects may partly account for the association within this group between cortisol values and symptom change. Overall, the findings highlight the need to follow pretreatment suppressors in future serial DST studies. To our knowledge, this is the first report that a substantial proportion of pretreatment suppressors abnormalize with somatic treatment. There were indications here that such an abnormalization may be associated with poor clinical outcome. Regardless of its possible clinical significance, in this type of research, following initial suppressors is necessary to control for the effect of regression to the mean when using cutoff cortisol values or avoiding restriction of range in examining continuous measures of postdexamethasone cortisol secretion.
to
Supported in part by NlMH Grant MH35636.
References Abrams R, Swartz CM (1985): ECT and prolactin release. 1. Effects of stimulus parameters. Convul Ther 1:115-119. Albala AA, Greden JF, Tarika J, Carroll BJ (1981): Changes in serial Dexamethasone Suppression Tests among unipolar depressives receiving electroconvulsive treatment. Biol Psychiatry 16:X51-560. Amsterdam JD, Winokur A, Caroff S (1981): Dexamethasone Suppression Test as a prognostic tool: 2 case reports. Am J Psychiatry 138:979-980. Brown WA, Shuey I (1980): Response to dexamethasone and subtype of depression. Arch Gen Psychiatry 37:747-75 1. Carroll BJ (1982): The Dexamethasone Suppression Test for melancholia. Br J Psychiatry 140:292-304. Carroll BJ, Curtis GC (1976): Neuroendocrine identification of depressed patients. Aust NZ J Psychiatry 10: 13-20. Carroll BJ, Feinberg M, Greden JF, et al (1981): A specific laboratory test for the diagnosis of melancholia. Arch Gen Psychiatry 38: 15-22. Cohen J (1969): Statistical Power Analysis for the Behavioral Sciences. Orlando, FL: Academic. Coppen A, Milln P, Harwood J, Wood K (1985): Does the Dexamethasone Suppression Test predict antidepressant treatment success? Br J Psychiatry 146:294-296. Coryell W (1982): Hypothalamic+uitary-adrenal axis and ECT response. Psychiatry Res 6:283-292. Coryell W, Zimmerman M (1983): The Dexamethasone Suppression Test and ECT outcome: A six-month follow-up. Biol Psychiatry 18:21-27. Decina P, Sackeim HA, Malitz S (1983): Prognostic value of serial Dexamethasone Suppression Tests during and following ECf. Psychopharmacol Bull 19:85-87. Greden JF, Gardner R, King D, Grunhaus L, Carroll BJ, Kronfol Z (1983): Dexamethasone Suppression Tests in antidepressant treatment of melancholia: The process of normalization and test-retest reproducibility. Arch Gen Psychiatry 401493-500.
472
BIOL PSYCHIATRY 1987:22:463-412
D.P. Devanand et al.
Greden JF, Haskett R, Zelnik T, Grunhaus L, Tiongco D (1985): Serial hypothalamicpituitary-adrenal monitoring in depressed patients. Presented at the Society of Biological Psychiatry Meeting, Dallas, TX, May 15-19. Holsboer F, Steiger A, Hofschuster E (1982): Repeated Dexamethasone depressive illness. J Affect Dis 4:93-101.
Suppression
Test during
lnsel TR, Goodwin FK (1983): The Dexamethasone Suppression Test: Promises and problems of diagnostic laboratory tests in psychiatry. Hasp Commun Psychiatry 34: 1134-1138. Lipman RS, Backup C, Bobrin Y, et al (1985): The Dexamethasone Suppression Test and clinical outcome in hospitalized depressives treated with ECT. Presented at the Annual Meeting of the National Association of Private Psychiatric Hospitals, Marco Island, FL, January 30. Malitz S, Sackeim HA (1986): Electroconvulsive NY Acad Sci 46256-65.
therapy: Clinical and basic research issues. Ann
Malitz SM, Sackeim HA, Decina P, Kanzler M, Kerr B (1985): The efficacy of ECT: Dose-response interactions with modality. Murphy BEP (1967): Some studies of protein binding of steroids and their application to the routine micro and ultramicro measurement of various steroids in body fluids by competitive protein binding radioassay. J Clin Endocrinol Metab 271973-990. Novacenko H, Halpem FS, Sachar EJ (1980): An improved cortisol. Clin Chem 26:312-3 13. Papakostas Y, Fink M, Lee J, et al (198 1): Neuroendocrine and outcome with ECT. Psychiatry Res 455-64.
protein binding assay for plasma
measures in psychiatric patients: Course
Peselow Y, Goldring N, Fieve R, Wright R (1983): The Dexamethasone Suppression Test during depression and following recovery from a depressive episode. Psychopharmacol Bull 19:87-91. Qualls CB, Brown WA (1982): Dexamethasone Presented at the New Research Proceedings, Abst no. NR 36.
Suppression Test subgroups differ in ECT response. American Psychiatric Association Annual Meeting,
Zelnik T, Albala AA, Haskett R, Grunhaus L, Greden JF (1985): Serial neuroendocrine patterns in depressives treated only with electroconvulsive therapy. Presented at the Society of Biological Psychiatry Meeting, Dallas, TX, May 15-19.