Specificity of the salivary cortisol dexamethasone suppression test across psychiatric diagnoses

BIOL PSYCHIATRY 1989:25:879-893

879

Specificity of the Salivary Cortisol Dexamethasone Suppression Test Across Psychiatric Diagnoses David L. Copolov, Robert T. Rubin, Geoffrey W. Stuart, Russell E. Poland, Anthony J. Mander, S. P. Sashidharan, Andrew M. Whitehouse, Ivy M. Blackbum, Christopher P. Freeman, and Douglas H. R. Blackwood

One hundred forty-eight psychiatric inpatients, 12 outpatients, and 17 normal controls were given the I .O-mg overnight Dexamethasone Suppression Test (DST), with salivary cortisol concentrations being measured as the dependent variable. Based on the Structured Clinical Interview for DSM-III, the patients were diagnosed as having major depression with melancholia (n = 21), nonmelancholic major depression (n = JO), mania (n = 15), schizophrenia (n = 32), dementia (n = 6), substance dependencelabuse (n = 18), and miscellaneous (n = 18). Neither the melancholic major depressives nor the entire group of major depressives had significantly higher salivary cortisol pre- or postdexamethasone as compared with all the other patients combined, nor did the melancholic patients have significantly higher cortisol than the nonmelancholic depressives. The inpatients as a group had significantly higher pre- and postdexamethasone cortisol values than the normal controls; cortisol values for the outpatients were intermediate between these two groups. Illness severity (in the depressives), length of time in hospital before the DST, and medication regimen were all unrelated to DST outcome. Thus, in this study, the salivary cortisol DST showed little clinical utility in discriminating major depressives with and without melancholia from other patients with a broad range of psychiatric diagnoses. The test did distinguish between hospitalized psychiatric patients and normal control subjects and between depressed inpatients and depressed outpatients, indicating that hospitalization-related variables contributed to DST outcome.

From the MRC Brain Metabolism Unit, University Department of Pharmacology, Edinburgh, Scotland (D.L.C., R.T.R., I.M.B., D.H.R.B.); the Mental Health Research Institute of Victoria, Parkviile, Victoria, Australia (D.L.C., G.W.S.); the Department of Psychiatry, H&r-UCLA Medical Center, Torrance, CA (R.T.R., R.E.P.); the Royal Edinburgh Hospital, Edinburgh, Scotland (A.J.M.; A.M.W.; C.P.F.); and All Saints Hospital, University of Birmingham, England (S.P.S.). Supported in part by NIMH Grants MHZ8380 and MH34471. DC was supported by a Royal Society Commonwealth Bursary and by the U.K. Medical Research Council. R.T.R. was supported by a Fulbright-Hays Senior Research Fellowship, a Senior Visiting Scientist Award from the U.K. Medical Research Council, and by NIMH Research Scientist Award MH47363. Address reprint requests to Dr. Robert Rubin, Division of Biological Psychiatry, Harbor-UCLA Medical Center, Torrance, CA, 90509. Received November 28, 1987; revised July 11, 1988. 1989 Society of Biological Psychiatry

0006-3223/89/$03.50

880

BIOL PSYCHIATRI 1989:3:879-X93

Introduction Some 50% of endogenously depressed patients have heightened hypothalamo-pituitaryadrenal cortical (HPA) activity, as reflected by increased serum adrenocorticotrophic hormone (ACTH) and cortisol concentrations and cortisol resistance in the Dexamethasone Suppression Test (DST) (Carroll 1972, 1982; Carroll et al. 1981; Rubin and Poland 1982. 1983, 1984; Arana et al. 1985; Pfohl et al. 1985; Rubin et al. 1987). There is a continuing lack of consensus about the utility of the DST as a biological marker of endogenous depression, having to do mainly with its specificity. Although some investigators (e.g.. Carroll et al. 198 1) have reported an extremely high specificity of the DST for melancholia. others have indicated an importantly high rate of abnormal DST results in schizophrenia (Berger et al. 1984; Myers, 1984; Sawyer and Jeffries, 1984; Herz et al. 1985), anorexia nervosa and bulimia (Gemer and Gwirtsman, 1981; Hudson et al. 1982; Mitchell et al. 1984), obsessive-compulsive neurosis (Insel et al. 1982). and other conditions (Baumgartner et al. 1986). The issue of DST specificity has become prominent because recent studies have had less of an exclusive focus on melancholic populations than had earlier studies and because there is considerable variation in methodology and design among the reported studies (Checkley and Rush 1983; Rubin et al. 1987). The present investigation was designed to examine the performance of the salivary cortisol DST in a large number of patients with different psychiatric diagnoses, paying close attention to methodology. Saliva was used because it can be collected conveniently and noninvasively, thus permitting cortisol concentrations to be determined a number of times following dexamethasone administration. Although salivary cortisol concentrations are only 4%-60/o of those in plasma (Landon et al. 1984), they are not dependent on salivary flow. Salivary cortisol is not bound to protein and thus correlates closely with the biologically active free plasma cortisol concentration (Riad-Fahmy et al. 1982). Our earlier studies of depressed patients indicated that saliva cortisol concentrations were significantly elevated in DST nonsuppressors compared to suppressors (Poland and Rubin 1982), but that the saliva DST showed no utility as an ancillary diagnostic marker for endogenousimelancholic depression compared to nonendogenous depression (Copolov et al. 1985). In contrast, three groups of investigators have suggested that the salivary cortisol DST provides good diagnostic discrimination between the endogenous/melancholic subtype and other subtypes of depression and might even provide better diagnostic discrimination than the plasma cortisol DST (Ansseau et al. 1984; Hanada et al. 1985; Cook et al. 1986). In the present report, we further examine the performance of the salivary cortisol DST by comparing the test in the major depressives reported previously (Copolov et al. 1985) to the test in patients with other psychiatric diagnoses and in a small group of control subjects.

Methods A total of 177 subjects was studied. One hundred sixty were inpatients (n = 148) or outpatients (n = 12) at the Royal Edinburgh Hospital, and 17 were normal controls. The patients were assessed by 1 of 8 raters using the Structured Clinical Interview for DSMIII (SCID, March 1983 version) (Spitzer 1983), modified to screen for organic brain syndromes, eating disorders, and drug and alcohol abuse and dependence. The 8 raters achieved high interrater reliabilities for this instrument, with an overall kappa coefficient

BIOL PSYCHIATRY

Salivary Cortisoi DST

1989;25:879-893

881

Table 1. Demographic Characteristics and DSM-III Diagnoses of the 177 Subjects Diagnostic group

n

Major depressive disorder

‘71

Mania

15

Schizoph~nia

32

Dementia Substance dependence/abuse

Mi~ll~eous

Normal controls

6 18

18

17

Age (mea@ 42.3 42.1

Sex F M

Percent

DSM-III diagnosis

Pecent

72.5 27.5

Major depressive episode

69.6

44.8 35.8 41.3 33.0

F M F M

57.1 42.9 65.6 34.4

69.0 68.5 37.3 38.7

F M F M

66.3 33.3 16.7 83.3

39.6 51.2

31.3 33.9

F

M

F M

47.1 52.9

47.1 52.9

without melancholia Major depressive episode with melancholia Manic episode Schizop~nia Paranoia Schizophmniform disorder Schizoaffective disorder Atypical psychosis Brief reactive psychosis Acute paranoid disorder Primary degenemtive dementia Alcohol abuse and dependence Alcohol abuse Mixed drug dependence Heroin dependence Adjast~nt disorder Generalized anxiety disorder No diagnosis on Axis I Anorexia nervosa Bulimia Panic disorder Cyclothymia Dysthymia Uncomplicated bereavement No psychiatric illness

30.4 100 53.0 12.5 9.4 9.4 6.3 6.3 3.1 100 50.0 22.2 16.7 11.1 23.5 17.6 17.6 11.8 5.9 5.9 5.9 5.9 5.9 100

of 0.79 (Copolov et al. 1986). Based on the SCID interviews, the patients were classified into 24 diagnostic categories; these were compressed into 6 major diagnostic groups, as indicated in Table 1. Demographic characteristics of the 6 groups also are included in Table 1. Sampling of saliva began at 1l:OO PM, immediately before the administration of dexamethasone elixir (1.0 mg) (Organon, Oss, Netherlands; 0.05 mg/ml) under the supervision of one of two investigators (D.L.C., R.T.R.). Three postdexamethasone salivary samples were collected at 7:OOAM, 3:OOPM, and 11:OOPM the following day. Prior to the collection of each sample, the subjects brushed their teeth and rinsed their mouths ~o~ughly with water. A sialogogue, one or two drops of lemon juice, was applied to each subject’s tongue prior to the collection of several milliliters saliva into a plastic tube. Salivary cortisol was assayed in duplicate by a specific radioimmunoassay (RIA) (Foster and Dunn 1974; Poland and Rubin 1982). Maximum interassay and intraassay coefficients of variation were 13% and 98, respectively.

882

BIOL PSYCHIATR\r 19X9:2.5:879-897

Dl..

cDpolov e1 Bi

Salivary dexamethasone also was measured by RIA (Poland et al. 1987) in the 7:00 samples. (The dexamethasone concentrations in the 3:00 PM and 1l:OO PM samples were too low for accurate quantitation.) Methylene chloride extraction offered no advantage over direct assay of the samples. Maximum inter- and intraassay coefficients of variation were 15% and 100/c, respectively. AM

Statistical Analyses The cortisol concentrations at each time point were compared across the six diagnostic groups by Analysis of Variance (ANOVA). As the cortisol and dexamethasone values were log-distributed, they were log-transformed to achieve a Gaussian distribution. This transformation made the measures of central tendency (means) close to the medians of the transformed data, indicating that the data were relatively uninfluenced by outlier values. Optimum cutpoints for discriminating between diagnostic groups were selected by systematically varying the cutpoint and assessing diagnostic accuracy. Several measures of diagnostic performance were considered. To select the optimum cutpoint for the test, the usual sensitivity and specificity measures were used. As an overall measure, the average of sensitivity and specificity was used. This is a measure of idealized prevalence-free performance (Green and Swets 1966) and is unbiased because it gives both types of diagnostic error equal weight. Once the optimum cutpoint had been selected, the conventional, prevalence-dependent measures of positive and negative predictive value and overall diagnostic efficiency were calculated (Galen and Gambino 1975). These represented, respectively, the proportion of melancholic patients among those above the cutpoint, the proportion of nonmelancholics among those below the cutpoint, and the total proportion of correct diagnoses of melancholia and nonmelancholia. Pearson’s product-moment correlations were used to quantitate the relationship between 7:00 AM dexamethasone concentrations and cortisol concentrations at the different sampling times and the relationship between duration of hospitalization and DST response.

Results Figure 1 shows the mean log cortisol concentrations for the main diagnostic groups and the normal controls. The major depressive group has been divided into those with and those without melancholia. The absence of the physiological morning cortisol surge at 7:00 AM and the failure of the postdexamethasone 11:OO PM cortisol values to return to their corresponding 11:OO PM predexamethasone levels indicate that suppression of HPA axis activity occurred in all groups of patients, as well as in the controls. As the DST is claimed to be most specific for melancholia (Carroll et al. 1981), the melancholic depressives were compared a priori to the nonmelancholic depressives and to the other patient groups. A two-way ANOVA with Greenhouse-Geisser correction for degrees of freedom on the repeated measures comparisons showed that the melancholic patients did not have higher salivary cortisol concentrations compared with all of the other patients combined, including those with nonmelancholic depression (F1,158 = 1.67, p -=c0.20). There was no significant interaction effect, indicating that the melancholic patients neither had higher predexamethasone cortisol concentrations nor demonstrated any failure of suppression in relation to the other patient groups (F2.77.437.63 = 0.84, p -=c0.50). These findings also applied to the major depressive disorder subgroup as a whole. This subgroup did not have higher cortisol concentrations compared with all the other patients

Salivary

BIOL

Cortisof DST

2300 f+elanchdia III

cl Schiro~hrenie

PsYCImTRY

883

19&9;25:879-893

b=21) b.32)

1500

Pm

q M&ceWnneous(n= 18)

q other Depressbn q Dementia h=B) q Control (rhl7)

2300

Post

Post Tie

(*SO)

/TJ

Mania

bt5)

f Substance Abuse k18) q

I. DST by diagnostic group: Mean log cartisol concentrations by time for the major diagnostic groups. Figure

combined (FL,158 = 0.38, p < 0.60). Again, there was no significant interaction effect, indicating that they suppressed comparably to the other patients (F2.78.4~9.~= 0.30, p < 0.90). Within the group of major depressives, the results were similar; the melancholic patients did not have significantly higher cortisol than the no~el~cholic depressives W1+69= 1.24, p 6 0.301, and there was no signi~c~t interaction effect (F2.71,186.W= 0.61, p < 0.60). h contrast, the normal control subjects had significantly lower salivary cortisol concentrations compared with all of the nondepressed patient groups combined (F1,a6 = 28.8, p < 0.001) and compared to the major depressive disorder group itseff (Flmlw = 26.9, p C 0.001). There were no significant interactions (F~.sB,t4~,32= 1.27, p < 0.50; indicating that the degree of co&sol suppression was ~2.64.274.07 = 0.71, p < 0.80), similar in the controls and in the patients, although the controls started from a considerably lower predexamethasone baseline. The control versus patient comparisons were repeated by Analysis of Covariance (ANCOVA) with age as the covariate. Age contributed significantly to the variance in cortisol con~n~ations (F 1,i74 = 5.98, p < 0.05), due to correlations between age and cortiso~ concentrations ranging from +O. 17 at 390 PM postdexamethasone (p < 0.05) to +0.25 at 1190 PM postdexamethasone (p < 0.01). Nevertheless, age uniquely accounted for only l%-3% of the total variance, in contrast to patient status, which contributed &%-lo%. The effect of psychiatric diagnosis remained highly significant in the ANCOVA both for controls versus all patients (fils174 = 24.34, p < 0.001) and for controls versus depressed patients (Ft,as = 22.03, p < 0.001). To compare the diagnostic accuracy of the salivary DST with previously reported values obtained for this test and for the plasma DST, the points of maximum discrimination between the diagnostic groups were obtained by cutpoint analysis. This was done to compare the major depressives with melancholia to all of the other patients combined

X84

BlOL PSYCHIATKt lYXY:?5~879-X9?

(“other diagnoses”) and to the major depressives without melancholia (“other depression”), AS the data were log-distributed, cutpoints spaced at equal intervals on a log scale were used. (Raw cortisol values also are shown in the figures.) The percentages of the correct diagnosis of melancholic patients (sensitivity) and nonmelancholic patients (specificity) at several cortisol cutpoints are shown in Figures 2 and 3. Also shown is the average of sensitivity and specificity, as described above. As indicated in Figures 2 and 3, the discrimination between melancholic major depression and nonmelancholic depression or other psychiatric illness was relatively poor. The best criterion value and time for discriminating between melancholia and other psychiatric illness at a reasonable (>50%) level of sensitivity was at a cutpoint of 1.6 ng/ml (log cutpoint 0.2) at 3:OO PM (Figure 2). This gave a sensitivity for melancholia of 52%, but a specificity of only 70% in relation to all of the other psychiatric patients. At this cutpoint, the positive predictive value was only 2 I%, the negative predictive value 91%. and the diagnostic efficiency 68%. The best cutpoint for discriminating between melancholic and nonmelancholic depression was 1.3 ng/ml (log cutpoint 0.1) at 3:00 PM (Figure 3). Sensitivity was 57%, positive predictive value was 41%, negative predictive value 79%, and diagnostic efficiency 63%. To determine the effect of inpatient status on salivary cortisol, a comparison among the depressed inpatients, depressed outpatients, and normal controls was carried out by unweighted ANOVA (Figure 4). Salivary cortisol concentrations in the 61 depressed inpatients were higher than those in the 10 depressed outpatients, particularly at 3:00 PM and 11:OO PM postdexamethasone. This difference was almost significant as a main effect (F 1.6Y = 3.79, p < 0.06) and was particularly apparent as an interaction (F2.67,L84.01 = 4.98, p < O.Ol), indicating that the degree of cortisol suppression among the outpatients was considerably greater. These differences could not be attributed to illness severity, as the mean Hamilton depression rating scale score in the inpatients (20.5 2 3.06) was not significantly different from that in the outpatients (18.2 ? 1.3 1). The inpatients also had significantly higher salivary cortisol concentrations over the four measurements than the normal controls (F, .7h = 40.46, p < O.OOOl), and there was a near-significant interaction (FZ,X3,2,5,,3 = 2.47, p < 0.07). The average cortisol concentrations over the four measurements for the outpatients also were higher than those for the controls, but not quite significantly so (F1,2s = 3.84. p < 0.07), and there was no interaction (F2.39,59 66 = 0.86, p < 0.50). Haskett et al. (1983), Coccaro et al. (1984), and Roy-Byrne et al. (1984) suggested that the number of false-positive DSTs might be decreased by postponing the administration of the DST until at least several days after hospital admission. We therefore examined the relationship between the length of hospitalization prior to DST administration and mean salivary cortisol values at each time point; there were no significant correlations. To further examine the possibility of early (day 1 or 2) effects of hospitalization, the data were analyzed by dividing the test results into two groups at cutpoints of 1, 2, 3, etc., days following admission; again, no significant differences occurred. However, as in 70% of cases the DST was administered within 1 week of admission, and in 90% of cases within 2 weeks of admission, no firm conclusions about the effect of delaying administration of the DST to hospitalized patients beyond 2 weeks can be drawn from our data. A number of medications, including barbiturates, diphenylhydantoin, carbamazepine, methaqualone, and high-dose benzodiazepines, have been reported to interfere with the plasma cortisol DST (Carroll 1986). In order to determine the influence of drugs on the

BIOL PSYCHIATRY 1989;25:879-893

Salivary Cortisol DST

23~00

PRE- DEXAMETHASONE

CUTPOINT 6

-.I

3

.I

.s

7~00 POST- DEXAMETHASONE CUTPOINT

(LOGlo CORTISOL)(ng/ml) 0

..z

.6 .7 .8 91

RAW CORTISOL

.z

13 I.8

.a

.I

3 25 3

4

.8

CUTPOINT

-.I

VALUES (“g/ml)

2

.4

‘\

\

,/’

\

/’

0

5

-0.7.8.91

.*

131.6

.I

3 3.5 3

6

I

.8

5 6 7

23~00 POST- DEXAMETHASONE CUTPOINT

(LOGlo CORTISOL)(ng/ml)

-.2 0

.I

(LOGlo CORTISOL)(ng/ml)

4

RAW CORTISOL VALUES (ng/ml)

6

(LOGlo CORTISOL)(ng/ml)

.B 1

100 \\

-.I

.3

,551

15:OO POST- DEXAMETHASONE

6

885

/---’

i *0 10

I

i

,/

i._.,

_-’

‘\ ‘\._

‘L.

“\,

0 .3

.I

.6 .d.1.8.01

CORTISOL

1.31.0

3 3.5 3

.3

3 0

KEY:

.I

.5 .0.1.*.91

1.31.0

3 3.5 3

4501

RAW CORTISOL VALUES (ng/ml)

VALUES (ng/ml)

__ _ _ __

Sensitivity

-

Average

-----_

of Sensitivity

Specificity and Specificity

Figure 2. Melancholia vs. other diagnoses: Cutpoint analysis of the DST; melancholies versus other diagnoses. The sensitivity (proportion of melancholies above cutpoint) and specificity (proportion of other diagnoses below cutpoint) are plotted for different cutpoints, equally spaced on a log scale at the top of the graph. The corresponding cortisol concentrations are indicated at the bottom of the graph. Also plotted is the prevalence-free measure-overall diagnostic confidence (see text).

886

BIOL FSYCHIATRt IU89:?5:R79-X9?

D.t..

PRE- DEXAMETHASONE

23~00

c’opolov et al

7:OO POST-DEXAMETHASONE

i i \

I)’

/

\

I

‘.

’ j

i’

I’ ___*--- ,’

RAW

CORTISOL

VALUES

(ng/nl)

15:OO POST- DEXAMETHASONE

3

4 RAW

5 6.7891

1316

CORTISOL

KEY:

2 253

VALUES

4

23:OO POST- DEXAMETHASONE

$67

3

~..____

4

RAW

(ng/ml)

---

/

S
- __ _ _ .

of Sensitivity

ant!

5

67891

CORTISOL

I.316

2 253

VALIJIB

*

561

(n&d)

specifxity Specificity

Figure 3. Melancholia vs. other depression: Cutpoint analysis of the DST; melancholies versus other depression. See Figure 2 legend for details.

887

BIOL PSYCHIATRY 1989;25:879-893

Salivary Cortisol DST

i abc

h ab

k abc

i _

Figure4. DSTbyinpatientstatus:Meanlog cortisol concentrationsby time for depressed inpatients, depressed outpatients, and normal control subjects.

salivary cortisol DST, we categorized the patients into those receiving or not receiving

the following classes of drugs: tricyclic antidepressants, oral antipsychotics, depot antipsychotics, sedatives, lithium, monoamine oxidase inhibitors, anticholinergics, and others (this last group included antibiotics, oral contraceptives, antianginal agents, and vitamin supplements). Table 2 lists the numbers of patients receiving each class of drugs. ANOVAs at each of the four time points comparing salivary cortisol concentrations between patients taking each class of drugs and all other patients revealed no significant effect of any drug class on either overall mean cortisol or cortisol suppression postdexamethasone. However, these analyses were confounded by diagnosis, in that patients often were on other drugs as well as the drug of interest. To obtain a more valid analysis, we next compared depressed patients receiving tricyclic antidepressants alone (n = 17) with drug-free depressives (n = 22) and schizophrenic patients receiving antipsychotics alone (n = 11) with drug-free schizophrenics (n = 7). Neither analysis showed a significant effect of medication on overall mean salivary cortisol concentrations or on cortisol suppression postdexamethasone. However, there was a trend for antipsychotic drugs to lower mean cortisol (F ,.,6 = 3.37,p < 0.10) without affecting cortisol suppression by dexamethasone. No other single drug class by single diagnosis analysis could be validly undertaken because of the small number of patients in each group.

Table 2. Numbers of Patients Taking Different Classes of Drugs at the Time of the Salivary Cortisol DST, Grouped According to Whether the Drug Was Taken Alone or in Combination with Others Drug class

Alone

Tricyclic antidepressants Monoamine oxidase inhibitors Antipsychotics Sedatives Lithium Anticholinergics Others No current medication

20

1 21 13 0 0 10 -

In combination 13 5 33 21 14 4 17 -

Total 33 6 54 34 14 4 27 65

888

BIOL PSYCHIATRY 19X9:25:879-89?

D.1,. Copolov et al

In order to study the relationship between the salivary dexamethasone concentrations and salivary cortisol concentrations after dexamethasone, we calculated correlations between 11:OO PM predexamethasone and 7:00 AM postdexamethasone cortisol concentrations, 7:00 AM dexamethasone and 7:00 AM cortisol concentrations, and the multiple correlation of 7:00 AM dexamethasone and 1l:OO PM predexamethasone cortisol with 7:00 AM cortisol. The best predictor of 7:00 AM postdexamethasone salivary cortisol was the 11:OOPM predexamethasone cortisol value (r = 0.62, Y’ = 0.38). The correlation between 7:00 AM dexamethasone and 7:00 AM cortisol was weak, although in the expected negative direction (r = - 0.17). The addition of 7:oO AM dexamethasone to the function resulted in a nonsignificant increase in r of only 0.03 (multiple R = 0.65).

Discussion The present study was undertaken to determine the clinical utility of the salivary cortisol DST for discriminating major depressives with and without melancholia from other patients with a broad range of psychiatric diagnoses. Our findings cast doubt on the usefulness of the saliva DST in this context. The findings are consistent with those of our earlier report on the major depressive group, which demonstrated that postdexamethasone salivary cortisol failed to significantly discriminate among four groups categorized by increasing frequency of endogenous symptomatology (Copolov et al. 1986). Two main analyses were performed on the data. The ANOVA on the log-transformed salivary cortisol concentrations demonstrated that the cortisol values were not significantly different across the diagnostic groups. The cutpoint analyses indicated that a distinct subpopulation of DST nonsuppressors did not exist among the DSM-111 melancholic depressives. Our method for determining the optimal cutpoint, i.e., averaging sensitivity and specificity, was based on the fact that the selection of diagnostic criteria in clinical tests involves a trade-off between the rate at which a finding is present in the target population (sensitivity) versus the rate at which it is absent among control subjects (specificity). Furthermore, this method does not depend on the prevalence of the target population, in contrast to the predictive value of a test. Consistent with the ANOVA on the log-transformed cortisol values, the cutpoint analysis clearly differentiated the melancholies from the normal control subjects. For example, with a cortisol criterion value of 1.0 ng/ml at 7:00 AM, the sensitivity of the test for melancholia was 71%, and the specificity in relation to normal controls was 100%. However, when the melancholies were compared to the other patient groups, the cutpoint analyses provided little support for the clinical utility of the salivary cortisol DST. At the optimal “melancholies versus all other patients” cutpoint of 1.6 ng/ml (log cutpoint 0.2) at 3:00 PM (Figure 2), the sensitivity (52%) was similar to that previously reported for the plasma cortisol DST (Carroll et al. 1981; Berger et al. 1984), but the specificity (63%) was considerably lower than the 80%-95% reported for the plasma cortisol DST (Holsboer et al. 1980; Schlesser et al. 1980; Carroll et al. 1981; Rush et al. 1982; Peselow et al. 1983) and the 75%-900/c reported previously for the salivary cortisol DST (Ansseau et al. 1984; Hanada et al. 1985; Cook et al. 1986). However, other investigators (Cot-yell et al. 1982; Meltzer et al. 1982) have reported a lower specificity (60%-70%) for the plasma cortisol DST, which is close to the 52% reported herein. Of importance is the fact that at the optimal cortisol criterion value of 1.6 ng/ml at 3:OO PM, the positive predictive value of the salivary cortisol DST was only 21%, indicating that 79% of all nonsuppressed tests occurred in patients who were diagnosed as

Salivary Cortisol DST

BIOL PSYCHIATRY 1989;25:879-893

889

nonmelancholic. In a nonhospital setting, in which the prevalence of melancholia would be lower, the diagnostic efficiency would decrease accordingly; this casts doubt on the clinical utility of the test not only in hospital, but especially in outpatient or office settings. As mentioned earlier, three studies of the salivary cortisol DST have suggested that it indeed has potential clinical usefulness in the diagnosis of melancholia (Ansseau et al. 1984; Hanada et al. 1985; Cook et al. 1986). Ansseau et al. (1984) and Hanada et al. (1985) used comparatively small numbers of patients (n = 26 and n = 43, respectively) and either no or only a few non-affectively disordered patients. In contrast, Cook et al. (1986) studied 178 patients with a wide variety of psychiatric diagnoses. At their optimal cutpoint, the salivary DST had a sensitivity of 58%, a specificity of 82%, and a positive predictive value of 28%. The average of their sensitivity and specificity was 70%, compared to the 58% average in our patients and a chance level of 50%. Given the relative diagnostic nonspecificity of high salivary cortisol concentrations in our study, it is important to determine if factors other than diagnosis contributed to any differences. One of the clearest factors was that of hospi~lization. Depressed inpatients had significantly higher cortisol values than depressed outpatients, who, in turn, had salivary cortisol concentrations similar to those of the normal controls. Hospitalization results in a number of stresses, including separation from family and adjustment to a new environment, and other studies have demonstrated that salivary cortisol, as well as plasma cortisol, is a stress-responsive hormone (Stahl and Domer 1982; Jones et al. 1986). Our data suggest that the hospitalization effect on salivary cortisol is enduing and cannot be obviated by delaying the DST for several days after admission. The range of postadmission days prior to the DST in our study is similar to those of Haskett et al. (1983) and Coccaro et al. (1984). These studies demonstrated that factors associated with hospitalization can give spuriously high plasma DST nonsuppression rates early, but not later, in the postadmission phase. Roy-Byrne et al. (1984) also found a statistically nonsignificant trend toward more cortisol nonsuppression in the first 2 days of hospitalization compared to days 4-6 following admission. Another factor that theoretically could influence salivary cortisol concentrations is psychotropic medication. Our finding of only small, nonsignificant differences in salivary cortisol between patients on various types of medication adds support to the findings of Cook et al. (1986) and is consistent with the findings of Vining and McGinley (1984) that the diffusion of cortisol across the salivary gland acinar cells is so rapid that the salivary cortisol concentration is independent of salivary flow rate and thus is unaffected by drugs with prominent parasympatholytic (anticholinergic) effects. Nevertheless, most of our analyses of possible drug effects were confounded by diagnosis and by the fact that many patients were taking drugs of more than one type. Therefore, from our data, we can only ten~tive~y conclude that psychotropic drugs in their usual clinical doses do not interfere with the salivary cortisol DST. Several studies have reported an inverse relationship between plasma dexamethasone concentrations and plasma cortisol concentrations postdexamethasone (Arana et al. 1984; Holsboer et al. 1984; Johnson et al. 1984; Berger et al. 1985; Holsboer et al. 1986; Morris et al. 1986; Poland et al. 1987). Our data, the first to be reported on the relationship between salivary dexame~asone and salivary cortisol, indicate that the salivary dexamethasone concentration 8 hrs after administration of dexamethasone elixir (I .O mg) did not significantly affect the strong positive correlation between the 7:00 AM postdexamethasone salivary cortisol concentration and its best predictor, the 11:OOPM predexamethasone salivary cortisol concentration. The strength of this correlation (r = -t 0.62)

D.L. Copolov et al.

is similar to that which we reported for pre- and postdexamethasone plasma cortisol concentrations (r = + 0.56) in a separate group of depressed patients (Rubin et al. 1987). In that study as well, plasma dexamethasone concentrations did not contribute significantly to the correlation (Poland et al. 19873. However, none of these findings exclude the possibility that peak dexamethasone concentrations might be a significant factor in determining the cortisol response. In conclusion, the salivary cortisol DST clearly distinguished between hospitalized psychiatric patients and normal control subjects and between depressed inpatients and depressed outpatients, providing evidence that hospitalization-related variables contributed to DST outcome. The nature of the hospitalization effect was probably stress-related; it did not seem to be primarily a consequence of administration of psychotropic medication. the severity of depression in the depressed patients, salivary dexamethasone concentrations 8 hr after dexamethasone administration, or length of admission prior to the DST. Although the salivary DST has several advantages over the plasma DST in the investigation of HPA activity in psychiatric patients, especially for ease of sample collection, its promise as a laboratory measure to distinguish melancholies from other psychiatric patients is unfortunately not supported by our data. We wish to thank the medical and nursing staffs of the Royal Edinburgh Hospital for their help and cooperation.

References Ansseau M, Sulon J. Doumont A, Cerfontaine JL, Legros JJ, Sodoyez JC, Demey-Ponsart E (1984): Use of saliva cortisol in the Dexamethasone Suppression Test. Psychiatry Res 13:203211. Arana GW. Workman RJ. Baldessarini RJ (1984): Association between low plasma levels of dexamethasone and elevated levels of cortisol in psychiatric patients given dexamethasone. Am J Pswhiat~ 141:1619-1620. Arana GW, Baldessarini RJ, Omsteen M ( 1985): The Dexamethasone Suppression Test for diagnosis and prognosis in psychiatry: Commentary and review. Arch Gen Psychiatry 42:1193-1204. Baumgartner A, GrIf K-J, Kiirten 1 ( 1986): Serial Dexamethasone Suppression Tests in psychiatric illness: Part I. A study in schizophrenia and mania. Psychiatry Res 18:9-23. Berger M, Pirke K-M, Doerr P, Krieg J-C, von Zerssen D (1984): The limited utility of the Dexamethasone Suppression Test for the diagnostic process in psychiatry. Br J Psychiatry 145:372-382.

Berger M, Pirke K-M, Krieg J-C, von Zerssen D (1985): The effect of weight loss and of inappropriate plasma dexamethasone levels on the Dexamethasone Suppression Test. Psychiatry Res 15:351-360. Brown WA, Haier RJ, Quails CB (1980): Dexamethasone Suppression Test identifies subtypes of depression which respond to different antidepressants. Lancer i:928-929. Carroll BJ (1972): Control of plasma cortisol levels in depression: Studies with the Dexamethasone Suppression Test. In Davies B, Carroll BJ, Mowbray RM (eds), Depressive Illness: Some Research Studies. Springfield, IL: Charles C Thomas, pp 87-148. Carroll BJ ( 1982): The Dexamethasone

Suppression Test for melancholia.

Br J Psychiatry

140:292-

304.

Carroll BJ (1986): Informed 47:1(Suppl):lO-12.

use of the Dexamethasone

Suppression

Test. J Clin Psychiatry

Carroll BJ, Schroeder K, Mukhopadhyay S, &eden JF, Feinberg M, Ritchie J, Tarika J (1980): Plasma dexamethasone concentrations and cortisol suppression response in patients with endogenous depression. J Clin Endocrinol Metab 5 11433-437. Carroll BJ. Feinberg M, Greden JF. Tarika J. Albala AA, Haskett RF. James NMT,Kronfol

Z,

BIOL PSYCHIATRY 1989m879-893

Salivary Cortisol DST

891

L&r N, Steiner M, de Vigne JP, Young E (1981): A specific laboratory test for the diagnosis of melancholia: Standardization, validation and clinical utility. Arch Gen Psychiatry 38: 15-22. Checkley SA, Rush AJ (1983): Functional indices of biological disturbance. In Angst J (ed), The Origins of Depressjon. Current Concepts and Appr~ches. Berlin: Sponger-Verlag, pp 425445. Coccaro EF, Prudic T, Rothpearl A, Numberg HG (1984): Effect of hospital admission on DST results, Am J Psychiatry 141:982-985. Cook N, Harris B, Walker R, Hailwood R, Jones E, Johns S, Riad-Fahmy D (1986): Clinical utility of the Dexamethasone Suppression Test assessed by plasma and salivary cortisol determinations. P~chi~t~ Res 18143-150. Copolov D, Rubin RT, Mander AJ, Sashidharan SP, Whitehouse AM, Blackbum I, Freeman CP, Lane L, Poland RE (1985): Pre- and postdexamethasone salivary cortisol concentrations in major depression. Psychoneuroendocrinology 10:461-467. Copolov DL, Rubin RT, Mander AJ, Sashidharan SP, Whitehouse AM, Blackburn IM, Freeman CP, Blackwood DHR (1986): DSM-III melancholia: Do the criteria accurately and reliably distinguish endogenous pattern depression? J A&ct Disord 10: 191-202. Coryell W, Gaffney G, Burkhardt PE (1982): DSM-III melancholia and the primary-secondary distinction: A comparison of concurrent validity by means of the Dexamethasone Suppression Test. Am J Psychiatry 139: 120-122. Foster LB, Dunn RT (1974): Single antibody technique for radioimmunoassay of cortisol in unextracted serum or plasma. Clin Chem 20:365-368. Gaien RS, Gambino R (1975): Blot Diagnoses. New York: Wiley.

Nor~lj~:

The Predj~tjve Value and E~cien~y of Medical

Gemer RH, Gwirtsman HE (1981): Abnormalities of Dexamethasone Suppression Test and urinary MHPG in anorexia nervosa. Am J Psychiatry 138:650-653. Green PM, Swets JA (1966): Signal Detection Theory and Psychophysics. New York: Wiley. Hanada K, Yamada N, Kazutaka S, Takahashi K, Takahashi S (1985): Direct mdioimmunoassay of cortisol in saliva and its appIication to the ~xame~asone Suppression Test in affective disorders. Psychoneuroe~docrinoiogy IO:193-201. Haskett RF, Zis AP, Albala AA, Carroll BJ (1983): DST performance during first 48 hours of admission. Presented at the 38th Annual Meeting, Society of Biological Psychiatry, New York (abstr 100). Herz MI, Fava G, Molnar G, Edwards L (1985): The Dexamethasone Suppression Test in newly hospi~lized schizoph~nic patients. Am J ~~chiu~~ 142:127-129. Holsboer F (1983): The Dexamethasone Suppression Test in depressed patients: Clinical and biochemical aspects. J Steroid Biochem 19:251-257. Holsboer F, Bender W, Benkert 0, Klein HE, Schmauss M (1980): Diagnostic value of Dexamethasone Suppression Test in depression. Lancer ii:760. Holsboer F, Haack 0, Gerken A, Vecsei P (1984): Plasma dexamethasone concentrations and differential suppression response of cortisol and co~icoste~ne in depressives and controls. Biot Psychiatry 19:28 l-29 1. Holsboer F, Wiedemann K, Boll E (1986): Shortened dexamethasone half-life in depressed dexamethasone nonsuppressors. Arch Gen Psychiatry 43:813-815. Hudson JI, Laffer PS, Pope HG (1982): Bulimia related to affective disorder by family history and response to the Dexamethasone Suppression Test. Am J Psychiaw 139:685-687. Insel TR, Kahn NH, Guttmacher LB, Cohen RM, Murphy DL (1982): The Dexamethasone Suppression Test in patients with primary obsessive-compulsive disorder. Psychiatry Res 6: 153-160. Johnson GE, Hunt G, Kerr K, Caterson I (1984): Dexamethasone Suppression Test (DST) and plasma dexamethasone levels in depressed patients. Psychiatry Res 3:305-313. Jones KJ, Copolov DL, Outch KH (1986): Type A, test performance and salivary cortisol. J Psychosom Res 30:699-707.

892

BIOL PSYCHIATR'L 1989;25:879-893

11.1.. Copolov et al

Landon J, Smith DS, Perry LA, AI-Ansari AAK (1984): The assay of salivary cortisol. In Read GF. Riad-Fahmy D, Walker RF, Griffiths K (ed). Immunoassuys ofSteroids in Saliva. Cardiff: Alpha Omega Publishing, pp 300-307. Meltzer HY. Tricou BJ, Robertson A, Piyaka SK (1982): Effect of dexamethasone prolactin and cortisol levels in psychiatric patients. Am J Psychiatry 139:763-768.

on plasma

Mendlewicz J, Charles G, Franckson JM (1982): The Dexamethasone Suppression Test in affective disorders: Relationship to clinical and genetic subgroups. Br J Psychiatry 141:464-470. Mitchell JE, Pyle RL, Hatsukami D, Boutacoff Ll (1984): The Dexamethasone in patients with bulimia. J Clin Psychiatry 45508-51 I.

Suppression

Test

Morris H. Carr V, Gilliland J, Hooper M (1986): Dexamethasone concentrations and the Dexamethasone Suppression Test in psychiatric disorders. Br J Psychiatry 148:66-69. Myers ED (1984): Serial Dexamethasone Am J Psychiatry 141:904-905.

Suppression Tests in male chronic schizophrenic

Peselow ED, Goldring N, Fieve RR, Wright R (1983): The Dexamethasone Suppression depressed outpatients and normal control subjects. Am J Psychiatry 140:245-247. Pfohl B, Sherman B, Schlechte J, Stone R (1985): Pituitary/adrenal psychiatric depression. Arch Gen Psychiatry 42:897-903. Poland RE. Rubin RT (1982): Saliva cortisol levels following endogenously depressed patients. Life Sci 30: 177-l 8 I

patients. Test in

axis rhythm disturbances

dexamethasone

administration

in in

Poland RE, Rubin RT, Lane LA, Hart PJ, Lesser IM (1987): Neuroendocrine aspects of primary endogenous depression II. Serum dexamethasone concentrations and hypothalamo-pituitaradrenal cortical activity as determinants of the Dexamethasone Suppression Test. Arch Gen Psychiatry 44~790-795. Riad-Fahmy D, Read GF, Walker RF, Griffiths K (1982): Steroids in saliva for assessing endocrine function. Endocrinol Rev 3:367-395. Roy-Byrne P, Gwirtsman H, Stembach H, Gemer RH (1984): Effects of acute hospitalization the Dexamethasone Suppression and TRH Stimulation tests. Biof Psychiatry 17:41-48.

on

Rubin RT, Poland RE (1982): The chroncendocrinology of endogenous depression. In Mtiller EE, MacLeod RM (eds). Neuroendocrine Perspectives, vol. I. Amsterdam: Elsevier, pp 305-337. Rubin RT, Poland RE (1983): Neuroendocrine function in depression. In Angst J (ed), The Origins of Depression: Current Concepts and Approaches. Berlin: Springer-Verlag, pp 205-220. Rubin RT, Poland RE (1984): The Dexamethasone Suppression Test in depression: Advantages and limitations. In Burrows GD, Norman TR. Maguire KP (eds). Biological Psychiatry: Recent Studies. London: John Libbey, pp 76-83. Rubin RT, Poland RE, Lesser IM, Winston RA, Blodgett ALN (1987): Neuroendocrine aspects of primary endogenous depression 1. Cortisol secretory dynamics in patients and matched controls. Arch Gen Psychiatc 44:328-336. Rush AJ, Giles DE, Roffwarg HG, Parker CR (1982): Sleep EEG and Dexamethasone Suppression Test findings in outpatients with unipolar major depressive disorder. Biol Psychiatry 17:327334. Sawyer J, Jeffries JJ (1984): The Dexamethasone chiutr?, 45:39Y-402.

Suppression

Test in schizophrenia.

Schlesser MA, Winokur G, Sherman BM (1980): Hypothalamic-pituitary-adrenal depressive illness. Arch Gen Psychiatry 37:737-743. Shapiro MF, Lehman AF (1983): The diagnosis of depression analysis of the literature on the Dexamethasone Suppression 720.

J Clin Psy-

axis activity in

in different clinical settings. An Test. J Nerv Ment Dis 171:714-

Spitzer RL (1983): Psychiatric diagnosis: Are clinicians still necessary? Compr Psychiatry 24:399411.

Salivary Cortisol DST

EIOL PSYCHIATRY ~98~25:8?9-S93

893

Stahl F, Diimer G (1982): Responses of salivary cortisol levels ta stress-situations. Endokrinologie 80:158-162. Vining RF, McGinley RA (1984): Transport of steroids from blood to saliva. In Read GF, RiadFahmy D, Walker RF, Griffiths K (eds), ~rnrn~n~~s~s of Steroids in Saliva. Cardiff: Alpha Omega ~blishing, pp 56-63.