Journal ofAf&-tire Disorders. 26 ( 19921 17-24 0 1992 Elsevier Science Publishers B.V. All rights reserved 016%0327/92/$05.00
17
JAD 00921
linical correlates of response to examethasone uppression Test in epression: A Organisation Collaborative Study Markus Gastpar ’ Klinik
a, Ulrich Gilsdorf
a, M.T. Abou-Saleh
‘J and Trunc Ngo-Wlac
C
fib AHgerneine Psychiatrhe, Essex, FRG, ’ Urkersity of Licerpoo!, Liverpool, lJK and ’ World Heafh Qrganizatiaon, Gneua, Switzerland (Received 24 September 1990) (Revision received 20 May 1992) (Accepted 26 May 1992)
From 9 centres 293 patients took part in the WHO-collaborative study on Dexamethasone-Suppression-Test (DST) in depression to examine the relationship of psychopathological and psychiatric history information to cortisol-levels and suppression/ non-suppression status. Differences between the centres were large and significant on nearly all of the measures. The predictor analyses generally suffered from numerically weak correlations with many variables correlating to sex and age. Therefore analyses of the data were adjusted for centre-, sex-, and age-influences. The best predicting features of cortisol were ‘fitful, restless sleep’, ‘change of bodyweight’ and ‘affective disorders in blood relatives’. The last 2 items together with ‘hypersomnia’ and ‘ideas of insufficiency’ were the best predictors of suppression/nonsuppression status. However, statistical evidence did not seem to be strong enough to describe a typical symptom profile of a depressive cortisol suppressor or nonsuppressor.
Key words: DST; Major depression; Biological psychiatry
Preliminary results of ahe WHO-collaborative study on Dexamethasone-Suppression-Test (DST)
Carrespondence to: M. Gastpar, Director Winik fiir Allgemeine Psychiatric, Virchowstrasse 174, D-4300 Essen 1, FRG. ’ Present address: Department of Psychiatry, United Arab Emirates University, Al Ain, United Arab Emirates.
have been published (Coppen and Metcalfe, 1987). The present report contributes to the research on the relationship between response to the DST and clinical characteristics of depressed patients, such as psychopathological and psychiatric history information, based on the Standardized Assessment of Depressive Disorders (SADD), (Sartorius et al., 1983). Findings so far, relating psychopathological symptoms to hypercortisolism, are contradictory
and conscqucntly unsatisfying (cf. e.g. Miller ct 31.. 19x7). ~orc and rnorc conditions arc discussed which may intcrvcnc in thu regulatory mechanisms of the hypothalamic-pituitary-adrL‘nal axis. However, the more factors possibly influence cortisd secretion and/or bias the rcsponsc to DST, the more concomitant circumstances must be controlled to reach an acceptable cornparabidity of the samples studied and the resulting findings. Lack of comparability of the patient groups studied (e.g.. in sex and age), different cutpoints to define the abnormal response to DST. or differing statistical strategies (using the coraisol values transformed or untransformed. partialling out or not possibly intervening variables like sex and age) must lead to diverging results. This is cspccially true if the basical statistical associations. as indicated e.g. by the simple correlations between symptoms and cortisol, are not very strong. The present work will take into account the considerations just outlined as far as possible. Patients
and Methods
All patients had to fulfil the Rcscarch Diagnostic Criteria for major depressive disorder (Spitzer et al., 1978). From originally I3 research centres (set: Appendix 21 with 543 patients, rescarchcrs of 9 ccntrcs took part in a 2 day training session with the SADD (see Appendix 1) in BasIe and then documented the clinical picture and the psychiatric history of 304 patients on the SADD. In addition, the severity of the patient’s depressive symptoms was assessed on the giamilton-Rating-&ale for Depression (HRSD). (Hamiltcn? 1967). 11 patients were excluded from the analysis because their HRSD-score was less than 10, leaving the data of 293 patients for statistical analysis.
LYtuti.hxd nrtnlysis of For statistical analysis the dichotomization some catcgorial variables into O/ 1 -data was neccssary. Sex was coded as 0 for male and 1 for female patients. The psychopathological part of the SADD consists of a list of 40 symptoms and signs and 17 psychiatric history items (see Appendix 1). Symptoms of category ‘last month’ were used for anaIysis and dichotomized by assigning 0 (absent) to the original ratings of absent or unknown and 1 (present) to miid or severe. Missing data were treated as symptom absent. The psychiatric history items ‘History of psychiatric disorder (affective) in blood relatives (Itcrn 411’ to ‘Precipitating stress (Item SO) were coded as 0 (absent) and 1 (prcscnt). The number of prcvisus depressive episodes and the number of previous manic episodes (items 5 1 and 53) were summed to “Number of previous episodes’ and the greater of the number of years since first depressive episode and number of years since first manic episode was used as ‘Number of years since first episode’. Concerning DST-Status, code 0 was assigned to suppression or normal response, if the cortisol value was less than SO [ng/ml]. and code 1 to nonsuppression. As dependent variable for regrcscion analysis, the log Wtransformation of cortisol was chosen, because it had the better distribution characteristics of skewness (syrnmetry) and kurtosis (heavytailedness). The cortisol values were analyzed in two stages: I1 1 the log lo-transformed cortisol values by means of stepwise regression analysis, and (2) the categorized DST response (suppression/ nonsuppression status) by stepwise logistic regression analysis (a statistical technique to predict binary events and with far fewer assumptions than discriminant analysis). In either case the selection process was stopped, if a variable with a probabillity greater than 0.05 was entered into ahe qrec%ictar set. tioa 0P ahe Sample
of dexamethasone was given to the patients’at 23 h and k$ood was sampled between 15 and 16 h the following day. Technical details are described in the first report on the present project Coppen and Metcalfe, 1987). One
mg
and
Evduation
sf
Centre Differences
Out of 293 patients, considered for statistical analysis, 105 were male and 188 were female. According to KID-9 diagnostic classification,
13.7% had non-endogenous
and 86.3% had endogenous depression, whilst on the Newcastle-diagnostic index OKIS), 28.$%1 and 72.0% had non-endogenous and endogenous depression iC
in type of onset (xz = 15.28, df = 8, P < 0.053) and duration of present depressive condition (F = 3.36, df = 8/283, P < O.OOl), whereas no significance was reached for number of previous episodes and number of years since 1 episode.
spectivdy.
between centres 1s comiderable, which was confirmed by ANOVAs with the 9 centres as independent variabie: for cortisol, Iog IQ-cortisol, and age the F-values were 7.50, 15.36, and 6.18, respectively (with df = S/284 and P < .OOOl for each F-test). The centrewise presence of psychopathological symptoms and psychiatric history items is shown in Fig. I. !$epwise discriminant analysis selected as the 5 symptoms with the greatest differences between t!re centres: hypochondriasis (Symptom No. 3l), change of perception of time (No. 271, feelings of pressure (No. 231, other somatic signs and symptoms (No. 241, and slowness/ retardation of thought (No. 10). Among the psychiatric history items the greatest differences were found for: continual psychic stress (No. 49), precipitating stress (No. %I), stressful events in childhood (No. 44), other affective disturbances in adult life (No. 47), and traits of premorbid personality (No. 48). All the items mentioned contributed highly significant to the variation between centres. Table 1 presents the centrewise descriptive statistics for the remaining psychiatric history variables. The centres were significantly different The
TABLE
variatim
Results
As shown there were large differences between the centres on nearly all variables. Furthermore initial exploratory analyses showed mtobest correlations between many symptoms and the variables of sex and age (sex correlated highest with ideas of reference, - 0.16, decrease of libido, - 0.16, and hypochondriasis, - 0.16; age correlated with decrease of libido, -0.28, hypochon driasis, - 0.26, and constipation, - 0.21). To avoid confounding the results by centre-inffuences and these two demographic variables, it was decided to do partial correlations adjusting for centres, sex and age. This is formally done by ‘forcing in’ centres, sex and age into the predictor set as the first variables and then to start the stepwisc selection process. Swade et al. (1987) found the cortisol levels ir! the 4 winter months (Nov. to Feb.) to be Iower than the cortiscr values in the remaining months of the year. This relation could not continously be observed in our sample, and so we renounced to
1
Centre-information Centre
for psychiatric histow variables N
Number of prev .tius episodes Mean
St. Dev
Nundxr of years since 1. episode
Meal1
St. Dev
Duration of present condition (months)
Type of onset of present condition Clearly sudden
Slow, insidious
Mean
St. Dev.
( -=I I week) BliSk
Brussels Copenhagen Irvine Lucknow Moscow Munich Nagasaki Utrecht
39 36 13 26 25 54 57 19 24
3.14 3.w 2.06 2.40
All centres
293
3.43 3.57 2.09 3.82
3.7k 2.44 2.66 3.94 2.3h 4.21 4.70 1.75 5.58
7.92 6.70 S.60 5.89 7.S6 7.85 9.48 3.55 7.89
9.16 6.89 6.58 6.72 9.79 7.66 10.48 4.32 7.Y3
2 8 2 3 5 IO 20 c ;
37 28 11 23 20 44 37 14 IS
3.40 6.14 8.81 9.56 5.46 4.75 6.23 6.71 9.90
4.81 5.61 h*lB 9.64 6.51 5.18 7.06 5.92 8.05
3.03
3.89
7.46
8.38
61
232
6.24
6.74
I .8I)
:nK: BM3
Baa
BbcI
BbcI
I
-
26 PbLilsido
27 PacTi-
kiiU-
Ehcl
Ebci
3bcE
%bct
BbcI
28 Suicide
BbCI
29 Guilt
30 lnsamcg
-_
3l HypSbn
%bcI
Bhcl
%lbCL
%Mx
BMII 32 hg-!Svwi
BXI
Fig. 1. Centrewise percentages of symptom presence. Vertical axis is scaled in steps rtf IOCi. Centre abbreviations: B/Fxle; b/Brussels: C/Copenhagen: I/Irvine: L/Lucknow: M/Moscow; m/Munich; N/Nagasaki r;nd U/Utrecht. Full description of items in the appendix.
introduce an additional variable to be partialled ;)ut controlling for season;iI effects.
The simple correlation coefficients between log-transformed cortisol and the 40 symptoms were generally low ( iphest coeffcients found: 0.17 with symptom a kly, 0.15 with change of body weight, - 0.13 with decrease of libido). The same weak relationshi was true for the psychiatric history items set (highest correlation: 0.10 with stressful1 events in childhood). Stepwise analyses of regression were carried out for the symptom- and psychiatric history parts of the SAWD separately. afk:t having nartialled out the influences of centres, sex and age. The results are presented in Table 2 (for descriptive purposes, the simple and partial correlations are added). For predicting cortisol by symptoms and signs the following 2 items were selected: fitfull/ restless sleep and change of body weight, the first item associated with lower and the second with higher cortisol values. Within the set of psychiatric history items, only one item was seIected: patients with affective disorders in blood relatives
hypersomnia, change of body weight, and ideas of insufficiency Gable 3). Nonsuppressors had a higher prevalence of hypersomnia and change of body weight, whereas suppressors were higher in ideas of insufficiency. The classificatory part of analysis gave the folIowing results: centres, sex, and age as the only predictors classified 66.5% of suppressors and 69.6% of nonsuppressors correctly. Adding the 3 selected symptoms increased the percentage of the correctly classified patients to 72.9% for suppressors and 72.5% for nonsuppressers, with an average gain of about 5%. From the pool of psychiatric history variables one item was selected: affective disorders in blood relatives. It was related to the status of suppression and increased the number of correctly classified patients for not more than 1%. The simple positive correlations between symptoms like anxiety/ tension and aggression (see above) would have supported a stress-activation model of DST. Gut this initial impression was corrected after having partialled out the cenTABLE
3
Analysis of Logistic Regression (Cutpoint 50)
for Prediction
of DST-Status
had lower cortisol values.
With a cortisol cut-off point of 50 ng/ml, 155 from 293 patients were suppressors (52.9%) and 138 patients were non-suppressive (47. I%). Stepwise logistic regression analysis selected (after having partialled out centre-, sex- and ageinfluences)
3 symptoms with a probability
< 0.05:
Partial R
Proba5iiity
Predictors (adjusted for centres, sex. and age)
0.0996 0.0835 - 0.0742
0.0196 0.0335 0.0480
S~mnjxoms and Signs Hypersomnia Change of bady weight Ideas of insufficiency
-0.1408
0.0029
Psychiarrrc Hirtory Affective disorders in blood relatives
trc influences: ‘anxiety/ tension* and ‘aggression’ lost ail their predictive power, after having included Moscow into the predictor set. The reasons for this can be seen in Fig. 1: Moscaw has no positive ratings in the symptom of aggression (Item 5) and is lower than the other centres in anxiety/ tension (Item 41. In a similar way the symptom of ‘decrease of libido’ was no more important after having parGalled out the influences of sex and age. It was stronger associated with sex/age than with cortisol.
The results in this sample of 293 deprcssivc patients show large: variations between centres on nearly al! mcasurcs and confirm the influence of age anil sex on the DST. Results also show differences in the psychopathollogical profile of dcpressivc patients from different centres with greatest differences for hypochondriasis, change of perception of time, and feelings of pressure. Within the set of psychiatric history items, centre differCiUXS wer2 greatest for vari;lbtcs relating to stress: continuai psychic stress, precipitating stress, and stressful events in childhood. Multiple regression analysis showed an association between higher cortisol vaIues and the presence of change of body weight and bctwcen lower cortisol values and the prcscnce of fitful/ restless sleep. From thy ps!‘ckiatry history set, presence of affective disorders in blood relatives was connected with lower cortisol levels. Logistic regression analysis showed non-suppressors to have higher prcv;lPence of symptoms of hypersomnia and change of bodvweight, 1 whereas suppressors had higher prevalence of ideas of insufficiency, Concerning the psychiatric history, suppressors had a higher prevalence of affective disorders in blood relatives. all the variables mentioned contributed significantly to the variation of cortisol values or LXXst;rtus respectively. Their predictive power, howCVX, evaluated on the basis of simple and partial correlations, the proportion of variance explained or the number of correctly classified patients, is
too weak, and the resulting predictor pattern not clear enough, to describe the clinical picture of the typical suppressor or nonsuppressor within our sample of depressive patients. Our modest findings are not in contradiction with the results of related works in this field. Miller and Nelson (1987) present a table which contains ‘Some Reported Dexamethasone Suppression Test-Symptom Associations (p. 772)‘. This overview of 12 studies impressively makes clear that a continuous symptom profile for depressive nonsuppressors obviously does not exist. The rather moderate correlations and partly contradictory associations between psychopathslogical symptoms and suppression/ nonsuppressionstatus do not promise much success for further investigations. The approach of a stress-activation model for DST on a biological level (e.g. AbouSaleh, 1985; Mellsop et al., 1985) seems to be more fruitful for future research.
Abou-Saleh. M.T. (1985) Dexamethasone suppression tests in psychiatry: Is there 3 place for an integrated hypothesis’? Psych&r. Dev. 3. 275-306. Coppen. A. and Metcalfe, M. (1987) WHO collaborative study: The dexamcthasone suppression test in depression. Br. J. Psychintty 150. 45c)-462. Hamilton, M. (19671 Development of a rating scale for primary depressive iilness. Br. .I. Clin. Sot. Psychnl. 6, 27% 2%. Mellsop, G.W., Huttun, J.D. and Delahunt. J.W. (1985) Dexamethasone suppression test as a simple measure of stress’? Br. Med. J., 2961, 1804-1086. Miller, K.B. and Nelson. J.C. (19871 Does the Dexamethasone Suppressir;n Test relate to subtypes, factors, symptoms, or severity’? Arch. Gen. Psychiatry, 44, 769-774. Sartorius, N., Davidian. H., Err&erg, G.. Fenton. F.R.. Fujii, I.. Gastprlr, M.. Gulbinat. W.. Jahtensky, A., Kielhotz. P.. Lehmann, HI.. Naraghi, M.. Shimizu, M., Shinfuku, N. and Takahashi, R. (1983) Depressive disorders in different cultures. yeport on the WHO Collaborative Study on Standardized Assessment of Dupreshlve Disorder. World Health Orgsnizatiun. Geneva. Spirzer. R.. Endicott. J. and Rnhbins. E. (1978) Research diagnostic criteria: rationale and reliubility, Arch. Gen. Psychiatry, 35, 773-782. Swade. C., Metcalfe, M., Cuppen. A.. Mrndlewicz, J. and Linkowski, P. (1987) Seasonal variations in the Dexamethasone Suppression Test, J. Affect. Disord.. 13, Y-11.
73
Appendix I
Standardized Assc~sment of Depressive Disorder s (SADD) (A) Symptoms and Signs item Description
Sadness, depressed mood Joyless, inability to enjoy Hopelessness Anxiety and/or
tension
Aggression Irritahilty lack of energy Disruption of social functioning Desire to be alone Subjective experience of slowness and retardation of thought Indecisiveness Lack of self-confidence Loss of interest Loss of ability to concentrate Subjective experience of loss of memory Early awakening inability to fall asleep Fitful. restless sleep Hypersomnia Lack of appetite (B) Psychiatric History 11~m Description
4x 49
Affective disorders in blood relatives Other disorders in blond relatives Early separation Stressful event: in childhood Psychopathological symptoms in childhood Psychopathological symptoms in adolescence Psychopathological symptoms. syndromes and disorders in adult life other than past or present affective disturbances Pronounced traits of premorbid personality Continual psychic stress
St1
Precipitating stress
Item
Description
21 22
Change of body weight Constipation Feelings of pressure Other somatic signs and symptoms Other psychological symptoms Decrease of libido Change of perception of time
23 24 25 26 27 28 29 30
Suicidal ideas
36 37 38 39 4U
Feelings and/or ideas of guilt and self-reproach Ideas of insufficiency, inadequacy and worthlessness Hypochondriasis Ideas of impoverishment Ideas of ref’erence and/or per:;ecution Other delusions Disorders of perception: illusions/hallucinations Psychomotor retardation Psychomotor restlessness and agitation Diurnal fluctuation: Worse in the morning Diurnal fluctuation: Worse in rhe evening Physical disease or infirmity
Item
Description
51
No. of previous depressive episodes No. of years since 1. depressive episode No. of previous manic episodes No. of years since 1. manic episode No. of intervals free of symptoms Type of onset of current episode Duration of present condition
31
3’ 33
34 35
52 53
54 5s 56 57
Items of par1 (A) can be rated for the time interval of last month. and if no positive rating can be done for “any time in this episodr’. The irltensity of part (Al-items is ratehle as (a) absent, (b) present mild, (c) present crrntinously or in severe form, and (d) unknown. if the interviewer is not sure or information is not sufficient. Items 31 to !&Ican be rated as (a) absent or (b) present.
Appendix 2
This report was prepared on behalf of the collaborating investigators from the WI-IQ designated centrcs in this study at: MRC Neumps~~l~jatry Research Laburatory, Epsum (Lead Centre):
A. Coppen, M.T. Abou-Sal&, M. Metcalfe, 9. Harwood, C. Swade, P. Milln, J. Baily, Fsychiatrisclae UniL-ersithtsklinik Basd, Basel: P. JSielholz, M. Gastpar, G. Gastpar, Chigrres Unirersitaires tie Bruxelles, Brztsselss J. Mendlewicz, P. Linkowski, National Imtittite for- Nerrous and
Menrnl LXsuases, Brldupest: M. Arato, Z. Rhimer, Serr*ice Hopitn?ro-elrril’~l.sitL7irede Psychiatric, C’nsablcr~m: D. Moussaoui, K. Chihabeddirle, M. Touhani. Psyclmchcrrtist~~ Jmfitube, Coperthoget~: 0.9. Rafaelsen, I-I. Dam, A. Gjerris, L’pzir*ersi@of Califamiu, Inine: W.E. Bunney Jr., D. Garber, ,a/. Zetin, Ming G~cwge)LcMedical College, Lucknow: f3.B. Sethi, J.K. Trivedi, A. Bhiman, Abl C’PticnResenrch Centrc far Menfai Health, A.M.S., MOSCQW; ME. Vartanian, B.S. Beliayev, E.I.
Psychiatri.yche Kh%ik und Poliklinik der Unir’c sitiit Miinchen, Munich: I-I. Hippius, H. KIcin, Nagasaki UMzrsity School of Medicine, Nagasaki: R. Takahashi, Y. Nakane, S. Yoshimoto, Hokkaido University School of Medicine, Sap~~oro: L Yamashita, I. Koyama, T. Qhmori, Uvliw&y Hospi~ni, Utrecht: H.M. van Praag, D. Ceulemans, W.M.A. Verhoeven, Dicision of Mental Health, World Health Organization, Geneva, Switzerlmd: P.V. jMorozov, T. Ngo-Khac.
Minsker,
Sv.K. Eukanina,
17
JAD 00921
linical correlates of response to examethasone uppression Test in epression: A Organisation Collaborative Study Markus Gastpar ’ Klinik
a, Ulrich Gilsdorf
a, M.T. Abou-Saleh
‘J and Trunc Ngo-Wlac
C
fib AHgerneine Psychiatrhe, Essex, FRG, ’ Urkersity of Licerpoo!, Liverpool, lJK and ’ World Heafh Qrganizatiaon, Gneua, Switzerland (Received 24 September 1990) (Revision received 20 May 1992) (Accepted 26 May 1992)
From 9 centres 293 patients took part in the WHO-collaborative study on Dexamethasone-Suppression-Test (DST) in depression to examine the relationship of psychopathological and psychiatric history information to cortisol-levels and suppression/ non-suppression status. Differences between the centres were large and significant on nearly all of the measures. The predictor analyses generally suffered from numerically weak correlations with many variables correlating to sex and age. Therefore analyses of the data were adjusted for centre-, sex-, and age-influences. The best predicting features of cortisol were ‘fitful, restless sleep’, ‘change of bodyweight’ and ‘affective disorders in blood relatives’. The last 2 items together with ‘hypersomnia’ and ‘ideas of insufficiency’ were the best predictors of suppression/nonsuppression status. However, statistical evidence did not seem to be strong enough to describe a typical symptom profile of a depressive cortisol suppressor or nonsuppressor.
Key words: DST; Major depression; Biological psychiatry
Preliminary results of ahe WHO-collaborative study on Dexamethasone-Suppression-Test (DST)
Carrespondence to: M. Gastpar, Director Winik fiir Allgemeine Psychiatric, Virchowstrasse 174, D-4300 Essen 1, FRG. ’ Present address: Department of Psychiatry, United Arab Emirates University, Al Ain, United Arab Emirates.
have been published (Coppen and Metcalfe, 1987). The present report contributes to the research on the relationship between response to the DST and clinical characteristics of depressed patients, such as psychopathological and psychiatric history information, based on the Standardized Assessment of Depressive Disorders (SADD), (Sartorius et al., 1983). Findings so far, relating psychopathological symptoms to hypercortisolism, are contradictory
and conscqucntly unsatisfying (cf. e.g. Miller ct 31.. 19x7). ~orc and rnorc conditions arc discussed which may intcrvcnc in thu regulatory mechanisms of the hypothalamic-pituitary-adrL‘nal axis. However, the more factors possibly influence cortisd secretion and/or bias the rcsponsc to DST, the more concomitant circumstances must be controlled to reach an acceptable cornparabidity of the samples studied and the resulting findings. Lack of comparability of the patient groups studied (e.g.. in sex and age), different cutpoints to define the abnormal response to DST. or differing statistical strategies (using the coraisol values transformed or untransformed. partialling out or not possibly intervening variables like sex and age) must lead to diverging results. This is cspccially true if the basical statistical associations. as indicated e.g. by the simple correlations between symptoms and cortisol, are not very strong. The present work will take into account the considerations just outlined as far as possible. Patients
and Methods
All patients had to fulfil the Rcscarch Diagnostic Criteria for major depressive disorder (Spitzer et al., 1978). From originally I3 research centres (set: Appendix 21 with 543 patients, rescarchcrs of 9 ccntrcs took part in a 2 day training session with the SADD (see Appendix 1) in BasIe and then documented the clinical picture and the psychiatric history of 304 patients on the SADD. In addition, the severity of the patient’s depressive symptoms was assessed on the giamilton-Rating-&ale for Depression (HRSD). (Hamiltcn? 1967). 11 patients were excluded from the analysis because their HRSD-score was less than 10, leaving the data of 293 patients for statistical analysis.
LYtuti.hxd nrtnlysis of For statistical analysis the dichotomization some catcgorial variables into O/ 1 -data was neccssary. Sex was coded as 0 for male and 1 for female patients. The psychopathological part of the SADD consists of a list of 40 symptoms and signs and 17 psychiatric history items (see Appendix 1). Symptoms of category ‘last month’ were used for anaIysis and dichotomized by assigning 0 (absent) to the original ratings of absent or unknown and 1 (present) to miid or severe. Missing data were treated as symptom absent. The psychiatric history items ‘History of psychiatric disorder (affective) in blood relatives (Itcrn 411’ to ‘Precipitating stress (Item SO) were coded as 0 (absent) and 1 (prcscnt). The number of prcvisus depressive episodes and the number of previous manic episodes (items 5 1 and 53) were summed to “Number of previous episodes’ and the greater of the number of years since first depressive episode and number of years since first manic episode was used as ‘Number of years since first episode’. Concerning DST-Status, code 0 was assigned to suppression or normal response, if the cortisol value was less than SO [ng/ml]. and code 1 to nonsuppression. As dependent variable for regrcscion analysis, the log Wtransformation of cortisol was chosen, because it had the better distribution characteristics of skewness (syrnmetry) and kurtosis (heavytailedness). The cortisol values were analyzed in two stages: I1 1 the log lo-transformed cortisol values by means of stepwise regression analysis, and (2) the categorized DST response (suppression/ nonsuppression status) by stepwise logistic regression analysis (a statistical technique to predict binary events and with far fewer assumptions than discriminant analysis). In either case the selection process was stopped, if a variable with a probabillity greater than 0.05 was entered into ahe qrec%ictar set. tioa 0P ahe Sample
of dexamethasone was given to the patients’at 23 h and k$ood was sampled between 15 and 16 h the following day. Technical details are described in the first report on the present project Coppen and Metcalfe, 1987). One
mg
and
Evduation
sf
Centre Differences
Out of 293 patients, considered for statistical analysis, 105 were male and 188 were female. According to KID-9 diagnostic classification,
13.7% had non-endogenous
and 86.3% had endogenous depression, whilst on the Newcastle-diagnostic index OKIS), 28.$%1 and 72.0% had non-endogenous and endogenous depression iC
in type of onset (xz = 15.28, df = 8, P < 0.053) and duration of present depressive condition (F = 3.36, df = 8/283, P < O.OOl), whereas no significance was reached for number of previous episodes and number of years since 1 episode.
spectivdy.
between centres 1s comiderable, which was confirmed by ANOVAs with the 9 centres as independent variabie: for cortisol, Iog IQ-cortisol, and age the F-values were 7.50, 15.36, and 6.18, respectively (with df = S/284 and P < .OOOl for each F-test). The centrewise presence of psychopathological symptoms and psychiatric history items is shown in Fig. I. !$epwise discriminant analysis selected as the 5 symptoms with the greatest differences between t!re centres: hypochondriasis (Symptom No. 3l), change of perception of time (No. 271, feelings of pressure (No. 231, other somatic signs and symptoms (No. 241, and slowness/ retardation of thought (No. 10). Among the psychiatric history items the greatest differences were found for: continual psychic stress (No. 49), precipitating stress (No. %I), stressful events in childhood (No. 44), other affective disturbances in adult life (No. 47), and traits of premorbid personality (No. 48). All the items mentioned contributed highly significant to the variation between centres. Table 1 presents the centrewise descriptive statistics for the remaining psychiatric history variables. The centres were significantly different The
TABLE
variatim
Results
As shown there were large differences between the centres on nearly all variables. Furthermore initial exploratory analyses showed mtobest correlations between many symptoms and the variables of sex and age (sex correlated highest with ideas of reference, - 0.16, decrease of libido, - 0.16, and hypochondriasis, - 0.16; age correlated with decrease of libido, -0.28, hypochon driasis, - 0.26, and constipation, - 0.21). To avoid confounding the results by centre-inffuences and these two demographic variables, it was decided to do partial correlations adjusting for centres, sex and age. This is formally done by ‘forcing in’ centres, sex and age into the predictor set as the first variables and then to start the stepwisc selection process. Swade et al. (1987) found the cortisol levels ir! the 4 winter months (Nov. to Feb.) to be Iower than the cortiscr values in the remaining months of the year. This relation could not continously be observed in our sample, and so we renounced to
1
Centre-information Centre
for psychiatric histow variables N
Number of prev .tius episodes Mean
St. Dev
Nundxr of years since 1. episode
Meal1
St. Dev
Duration of present condition (months)
Type of onset of present condition Clearly sudden
Slow, insidious
Mean
St. Dev.
( -=I I week) BliSk
Brussels Copenhagen Irvine Lucknow Moscow Munich Nagasaki Utrecht
39 36 13 26 25 54 57 19 24
3.14 3.w 2.06 2.40
All centres
293
3.43 3.57 2.09 3.82
3.7k 2.44 2.66 3.94 2.3h 4.21 4.70 1.75 5.58
7.92 6.70 S.60 5.89 7.S6 7.85 9.48 3.55 7.89
9.16 6.89 6.58 6.72 9.79 7.66 10.48 4.32 7.Y3
2 8 2 3 5 IO 20 c ;
37 28 11 23 20 44 37 14 IS
3.40 6.14 8.81 9.56 5.46 4.75 6.23 6.71 9.90
4.81 5.61 h*lB 9.64 6.51 5.18 7.06 5.92 8.05
3.03
3.89
7.46
8.38
61
232
6.24
6.74
I .8I)
:nK: BM3
Baa
BbcI
BbcI
I
-
26 PbLilsido
27 PacTi-
kiiU-
Ehcl
Ebci
3bcE
%bct
BbcI
28 Suicide
BbCI
29 Guilt
30 lnsamcg
-_
3l HypSbn
%bcI
Bhcl
%lbCL
%Mx
BMII 32 hg-!Svwi
BXI
Fig. 1. Centrewise percentages of symptom presence. Vertical axis is scaled in steps rtf IOCi. Centre abbreviations: B/Fxle; b/Brussels: C/Copenhagen: I/Irvine: L/Lucknow: M/Moscow; m/Munich; N/Nagasaki r;nd U/Utrecht. Full description of items in the appendix.
introduce an additional variable to be partialled ;)ut controlling for season;iI effects.
The simple correlation coefficients between log-transformed cortisol and the 40 symptoms were generally low ( iphest coeffcients found: 0.17 with symptom a kly, 0.15 with change of body weight, - 0.13 with decrease of libido). The same weak relationshi was true for the psychiatric history items set (highest correlation: 0.10 with stressful1 events in childhood). Stepwise analyses of regression were carried out for the symptom- and psychiatric history parts of the SAWD separately. afk:t having nartialled out the influences of centres, sex and age. The results are presented in Table 2 (for descriptive purposes, the simple and partial correlations are added). For predicting cortisol by symptoms and signs the following 2 items were selected: fitfull/ restless sleep and change of body weight, the first item associated with lower and the second with higher cortisol values. Within the set of psychiatric history items, only one item was seIected: patients with affective disorders in blood relatives
hypersomnia, change of body weight, and ideas of insufficiency Gable 3). Nonsuppressors had a higher prevalence of hypersomnia and change of body weight, whereas suppressors were higher in ideas of insufficiency. The classificatory part of analysis gave the folIowing results: centres, sex, and age as the only predictors classified 66.5% of suppressors and 69.6% of nonsuppressors correctly. Adding the 3 selected symptoms increased the percentage of the correctly classified patients to 72.9% for suppressors and 72.5% for nonsuppressers, with an average gain of about 5%. From the pool of psychiatric history variables one item was selected: affective disorders in blood relatives. It was related to the status of suppression and increased the number of correctly classified patients for not more than 1%. The simple positive correlations between symptoms like anxiety/ tension and aggression (see above) would have supported a stress-activation model of DST. Gut this initial impression was corrected after having partialled out the cenTABLE
3
Analysis of Logistic Regression (Cutpoint 50)
for Prediction
of DST-Status
had lower cortisol values.
With a cortisol cut-off point of 50 ng/ml, 155 from 293 patients were suppressors (52.9%) and 138 patients were non-suppressive (47. I%). Stepwise logistic regression analysis selected (after having partialled out centre-, sex- and ageinfluences)
3 symptoms with a probability
< 0.05:
Partial R
Proba5iiity
Predictors (adjusted for centres, sex. and age)
0.0996 0.0835 - 0.0742
0.0196 0.0335 0.0480
S~mnjxoms and Signs Hypersomnia Change of bady weight Ideas of insufficiency
-0.1408
0.0029
Psychiarrrc Hirtory Affective disorders in blood relatives
trc influences: ‘anxiety/ tension* and ‘aggression’ lost ail their predictive power, after having included Moscow into the predictor set. The reasons for this can be seen in Fig. 1: Moscaw has no positive ratings in the symptom of aggression (Item 5) and is lower than the other centres in anxiety/ tension (Item 41. In a similar way the symptom of ‘decrease of libido’ was no more important after having parGalled out the influences of sex and age. It was stronger associated with sex/age than with cortisol.
The results in this sample of 293 deprcssivc patients show large: variations between centres on nearly al! mcasurcs and confirm the influence of age anil sex on the DST. Results also show differences in the psychopathollogical profile of dcpressivc patients from different centres with greatest differences for hypochondriasis, change of perception of time, and feelings of pressure. Within the set of psychiatric history items, centre differCiUXS wer2 greatest for vari;lbtcs relating to stress: continuai psychic stress, precipitating stress, and stressful events in childhood. Multiple regression analysis showed an association between higher cortisol vaIues and the presence of change of body weight and bctwcen lower cortisol values and the prcscnce of fitful/ restless sleep. From thy ps!‘ckiatry history set, presence of affective disorders in blood relatives was connected with lower cortisol levels. Logistic regression analysis showed non-suppressors to have higher prcv;lPence of symptoms of hypersomnia and change of bodvweight, 1 whereas suppressors had higher prevalence of ideas of insufficiency, Concerning the psychiatric history, suppressors had a higher prevalence of affective disorders in blood relatives. all the variables mentioned contributed significantly to the variation of cortisol values or LXXst;rtus respectively. Their predictive power, howCVX, evaluated on the basis of simple and partial correlations, the proportion of variance explained or the number of correctly classified patients, is
too weak, and the resulting predictor pattern not clear enough, to describe the clinical picture of the typical suppressor or nonsuppressor within our sample of depressive patients. Our modest findings are not in contradiction with the results of related works in this field. Miller and Nelson (1987) present a table which contains ‘Some Reported Dexamethasone Suppression Test-Symptom Associations (p. 772)‘. This overview of 12 studies impressively makes clear that a continuous symptom profile for depressive nonsuppressors obviously does not exist. The rather moderate correlations and partly contradictory associations between psychopathslogical symptoms and suppression/ nonsuppressionstatus do not promise much success for further investigations. The approach of a stress-activation model for DST on a biological level (e.g. AbouSaleh, 1985; Mellsop et al., 1985) seems to be more fruitful for future research.
Abou-Saleh. M.T. (1985) Dexamethasone suppression tests in psychiatry: Is there 3 place for an integrated hypothesis’? Psych&r. Dev. 3. 275-306. Coppen. A. and Metcalfe, M. (1987) WHO collaborative study: The dexamcthasone suppression test in depression. Br. J. Psychintty 150. 45c)-462. Hamilton, M. (19671 Development of a rating scale for primary depressive iilness. Br. .I. Clin. Sot. Psychnl. 6, 27% 2%. Mellsop, G.W., Huttun, J.D. and Delahunt. J.W. (1985) Dexamethasone suppression test as a simple measure of stress’? Br. Med. J., 2961, 1804-1086. Miller, K.B. and Nelson. J.C. (19871 Does the Dexamethasone Suppressir;n Test relate to subtypes, factors, symptoms, or severity’? Arch. Gen. Psychiatry, 44, 769-774. Sartorius, N., Davidian. H., Err&erg, G.. Fenton. F.R.. Fujii, I.. Gastprlr, M.. Gulbinat. W.. Jahtensky, A., Kielhotz. P.. Lehmann, HI.. Naraghi, M.. Shimizu, M., Shinfuku, N. and Takahashi, R. (1983) Depressive disorders in different cultures. yeport on the WHO Collaborative Study on Standardized Assessment of Dupreshlve Disorder. World Health Orgsnizatiun. Geneva. Spirzer. R.. Endicott. J. and Rnhbins. E. (1978) Research diagnostic criteria: rationale and reliubility, Arch. Gen. Psychiatry, 35, 773-782. Swade. C., Metcalfe, M., Cuppen. A.. Mrndlewicz, J. and Linkowski, P. (1987) Seasonal variations in the Dexamethasone Suppression Test, J. Affect. Disord.. 13, Y-11.
73
Appendix I
Standardized Assc~sment of Depressive Disorder s (SADD) (A) Symptoms and Signs item Description
Sadness, depressed mood Joyless, inability to enjoy Hopelessness Anxiety and/or
tension
Aggression Irritahilty lack of energy Disruption of social functioning Desire to be alone Subjective experience of slowness and retardation of thought Indecisiveness Lack of self-confidence Loss of interest Loss of ability to concentrate Subjective experience of loss of memory Early awakening inability to fall asleep Fitful. restless sleep Hypersomnia Lack of appetite (B) Psychiatric History 11~m Description
4x 49
Affective disorders in blood relatives Other disorders in blond relatives Early separation Stressful event: in childhood Psychopathological symptoms in childhood Psychopathological symptoms in adolescence Psychopathological symptoms. syndromes and disorders in adult life other than past or present affective disturbances Pronounced traits of premorbid personality Continual psychic stress
St1
Precipitating stress
Item
Description
21 22
Change of body weight Constipation Feelings of pressure Other somatic signs and symptoms Other psychological symptoms Decrease of libido Change of perception of time
23 24 25 26 27 28 29 30
Suicidal ideas
36 37 38 39 4U
Feelings and/or ideas of guilt and self-reproach Ideas of insufficiency, inadequacy and worthlessness Hypochondriasis Ideas of impoverishment Ideas of ref’erence and/or per:;ecution Other delusions Disorders of perception: illusions/hallucinations Psychomotor retardation Psychomotor restlessness and agitation Diurnal fluctuation: Worse in the morning Diurnal fluctuation: Worse in rhe evening Physical disease or infirmity
Item
Description
51
No. of previous depressive episodes No. of years since 1. depressive episode No. of previous manic episodes No. of years since 1. manic episode No. of intervals free of symptoms Type of onset of current episode Duration of present condition
31
3’ 33
34 35
52 53
54 5s 56 57
Items of par1 (A) can be rated for the time interval of last month. and if no positive rating can be done for “any time in this episodr’. The irltensity of part (Al-items is ratehle as (a) absent, (b) present mild, (c) present crrntinously or in severe form, and (d) unknown. if the interviewer is not sure or information is not sufficient. Items 31 to !&Ican be rated as (a) absent or (b) present.
Appendix 2
This report was prepared on behalf of the collaborating investigators from the WI-IQ designated centrcs in this study at: MRC Neumps~~l~jatry Research Laburatory, Epsum (Lead Centre):
A. Coppen, M.T. Abou-Sal&, M. Metcalfe, 9. Harwood, C. Swade, P. Milln, J. Baily, Fsychiatrisclae UniL-ersithtsklinik Basd, Basel: P. JSielholz, M. Gastpar, G. Gastpar, Chigrres Unirersitaires tie Bruxelles, Brztsselss J. Mendlewicz, P. Linkowski, National Imtittite for- Nerrous and
Menrnl LXsuases, Brldupest: M. Arato, Z. Rhimer, Serr*ice Hopitn?ro-elrril’~l.sitL7irede Psychiatric, C’nsablcr~m: D. Moussaoui, K. Chihabeddirle, M. Touhani. Psyclmchcrrtist~~ Jmfitube, Coperthoget~: 0.9. Rafaelsen, I-I. Dam, A. Gjerris, L’pzir*ersi@of Califamiu, Inine: W.E. Bunney Jr., D. Garber, ,a/. Zetin, Ming G~cwge)LcMedical College, Lucknow: f3.B. Sethi, J.K. Trivedi, A. Bhiman, Abl C’PticnResenrch Centrc far Menfai Health, A.M.S., MOSCQW; ME. Vartanian, B.S. Beliayev, E.I.
Psychiatri.yche Kh%ik und Poliklinik der Unir’c sitiit Miinchen, Munich: I-I. Hippius, H. KIcin, Nagasaki UMzrsity School of Medicine, Nagasaki: R. Takahashi, Y. Nakane, S. Yoshimoto, Hokkaido University School of Medicine, Sap~~oro: L Yamashita, I. Koyama, T. Qhmori, Uvliw&y Hospi~ni, Utrecht: H.M. van Praag, D. Ceulemans, W.M.A. Verhoeven, Dicision of Mental Health, World Health Organization, Geneva, Switzerlmd: P.V. jMorozov, T. Ngo-Khac.
Minsker,
Sv.K. Eukanina,
