748

Article

748 Am J Psychiatry 158:5, May 2001

The Dexamethasone Suppression Testand Suicide Prediction

William Coryell, M.D.

Michael Schlesser, M.D.

Objective: Despite the substantial risksof eventual suicide associated with majordepressive disorder, clinicians lack robustpredictors with which to quantify theserisks. This study compared the validity ofdemographic and historical risk factorswith that of the dexamethasone suppres-sion test (DST), a clinically practical mea-sure of hyperactivity of the hypothalamic-pituitary-adrenal (HPA) axis.

Method: Seventy-eight inpatients withResearch Diagnostic Criteria major de-pressive disorder or schizoaffective disor-der, depressed type, entered a long-termfollow-up study between 1978 and 1981,and, in addition, underwent a 1-mg DST.The number of suicides in this group dur-ing a 15-year follow-up period was deter-mined, and the predictive validity of fourdemographic and historical risk factorsreported in the literature to be consis-tently predictive of suicide in depressed

patients was compared to the predictivevalidity of the DST results.

Results: Thirty-two of the 78 patientshad abnormal DST results. Survival analy-ses showed that the estimated risk foreventual suicide in this group was 26.8%,compared to only 2.9% among patientswho had normal DST results. None of thedemographic and historical risk factorsexamined in the study significantly distin-guished those who later committed sui-cide from those who did not.

Conclusions: In efforts to predict andprevent suicidal behavior in patients withmajor depressive disorder, HPA-axis hy-peractivity, as reflected in DST results,may provide a tool that is considerablymore powerful than the clinical predic-tors currently in use. Research on thepathophysiology of suicidal behavior inmajor depressive disorder should empha-size the HPA axis and its interplay with theserotonin system.

(Am J Psychiatry 2001; 158:748753)

The clinical management of patients with affective dis-orders requires an estimate of their risk for suicide, but theempirical basis for such estimates is tenuous. Of the threeapplicable study designs, one uses vital statistics to de-scribe the demographic characteristics of those in thegeneral population who commit suicide. This approachcan test only a few risk factors, and the results do not gen-eralize to clinical samples because a substantial propor-tion of those who commit suicide do so without seekinghelp for a psychiatric disorder. A second approach identi-fies diagnostically mixed patient groups, most often madeup of patients who have threatened or attempted suicide,and compares those who subsequently complete suicideto those who do not. The diagnostic heterogeneity charac-teristic of these groups seriously limits conclusions. Manypsychiatric disorders substantially increase the risk foreventual suicide, and the factors that predispose an indi-vidual to suicide in one illness may be unimportant in an-other. For example, recent loss and lack of social supportare much more likely to precede suicide among alcoholicsthan they are among individuals who have primary de-pressive disorder (1, 2).

A third design, the follow-up of cohorts with specificdisorders, is more likely to identify risk factors peculiar to

a particular disorder. In a review of the English-languageliterature, we could find only four studies that assessedclinical predictors of completed suicide specificallyamong patients with major affective disorders (36). Thestudies differed considerably in the variables tested, andonly a few predictors emerged with any consistency. Inthree of the four studies, a history of suicide attempts orthe presence of suicidal ideation was more frequentlyfound among those who later committed suicide (35)and, in another three studies, being unmarried and/or liv-ing alone were predictive (3, 4, 6). In two studies, male pa-tients were at higher risk (3, 6).

Although further efforts may identify clinical profiles ofmore value as predictors, a search for relevant biologicalmeasures is clearly warranted. Among the biological ab-normalities tentatively associated with risk for suicide,those involving the hypothalamic-pituitary-adrenal(HPA) axis have shown particular promise. In postmor-tem studies, persons who died by suicide, in comparisonto matched subjects who died by other violent means,had greater adrenal weights (7, 8), fewer binding sites forcorticotropin-releasing factor (CRF) in the frontal cortex(9), and higher levels of CRF in CSF (10). Each of thesefindings associates suicide with HPA-axis hyperactivity.

Am J Psychiatry 158:5, May 2001 749

WILLIAM CORYELL AND MICHAEL SCHLESSER

The dexamethasone suppression test (DST) offers a clin-ically practical means for detecting such hyperactivity andmay therefore serve to estimate suicide risk. In the proce-dure used in the majority of studies, 1 mg of dexametha-sone is administered orally at 11:00 p.m., and plasma corti-sol levels are determined from blood samples drawn thefollowing day at 8:00 a.m. and 4:00 p.m. A value from eithersample exceeding 5 g/dl indicates failure to suppress cor-tisol and is considered evidence for HPA-axis hyperactivity.

Some studies have, in fact, found that patients with ab-normal DST results were more likely to have recentlymade a suicide attempt (1114) or were more likely tomake future attempts (11, 15). Numerous authors havefailed to find such relationships, though (1620). Therehave been far fewer efforts to test HPA-axis disturbance asa predictor of completed suicide, but the results have beenmuch more consistent.

Carroll et al. (21) identified five suicides from among 250melancholic patients who had undergone the DST andnoted that all had been nonsuppressors of cortisol,whereas this was true of only one-half of the remainingmelancholic patients. Coryell and Schlesser (22) learnedof four suicides from among 205 inpatients with primaryunipolar depression who had undergone the DST whilehospitalized. All of these, but less then half (45.8%) of theremaining patients, had been nonsuppressors. Finally,Norman et al. (23) drew from a large sample of depressedinpatients with DST results and matched the 13 who sub-sequently committed suicide both to other patients whohad attempted suicide before admission and to patientswho had not attempted suicide. Again, approximatelyone-half (54%) of those who later committed suicide werenonsuppressors at 4:00 p.m., compared with 28% of thosewho had not attempted suicide and only 8% of those witha previous attempt.

These studies of completed suicide did not test the rela-tive value of other clinical predictors. This leaves the pos-sibility that DST results served as a proxy for features suchas delusions, overall symptom severity, hopelessness, or ahistory of mania. Moreover, the observation periods werelimited. Carroll et al. (21) did not describe the length of fol-low-up, and the other two studies averaged less than 3years of observation (23, 24).

The earlier report from this center concerning serioussuicide attempts and DST results (15) described patientswho entered the National Institute of Mental Health Col-laborative Study of the Affective DisordersClinicalBranch, a long-term follow-up of patients who met Re-search Diagnostic Criteria (RDC) (25) for major depressivedisorder, mania, or schizoaffective disorder. Although thebaseline assessment for the Collaborative DepressionStudy included only demographic and clinical variables, alarge proportion of those who entered the study as inpa-tients at the University of Iowa center underwent a 1-mgDST as part of their routine clinical assessment. Follow-uphas continued since the original report on suicide at-

tempts, and a number of suicides have occurred in the in-terval. This study examines the relationship between DSTresults and these suicides. Because all subjects underwentstructured phenomenological assessments at baseline,the relative importance of clinical predictors can also beexamined.

Method

Subjects

Between 1978 and 1981, inclusive, patients who sought treat-ment as inpatients or outpatients for conditions that met RDC formajor depressive disorder, mania, or schizoaffective disorderwere recruited into the Collaborative Depression Study. Partici-pants were age 18 or older, white, and English speaking. The cur-rent analysis was restricted to inpatients recruited at the Univer-sity of Iowa center.

Procedures

All subjects provided written informed consent after beinggiven a complete description of the study. Diagnoses were basedon the full Schedule for Affective Disorders and Schizophrenia(SADS) (26), which combined information from direct interviewsand medical records. Follow-up interviews, structured by theLongitudinal Interval Follow-Up Evaluation (27), took place at 6-month intervals for the next 5 years and annually thereafter. Theprotocol did not determine or influence treatment, but ongoingsomatotherapies were methodically monitored and quantified, asdescribed elsewhere (28).

DSTs were not part of the Collaborative Depression Study pro-tocol. The patients described in this report received them in oneof two ways. A formal research protocol, described elsewhere (29),included the DST and overlapped with the period of intake for theCollaborative Depression Study. Other patients received the DSTbecause it was ordered by the treating physician, as was oftendone during the period of intake for the collaborative study. In ei-ther case, the patient took 1 mg of dexamethasone orally at 11:00p.m. and provided blood samples the next day at 8:00 a.m. and/or4:00 p.m. Cortisol levels were determined by a competitive pro-tein-binding assay (30). A cortisol value greater then 5 g/dl in ei-ther postdexamethasone sample indicated nonsuppression ofcortisol. For the patients assessed in the protocol of Schlesser etal. (29), the designation as suppressor or nonsuppressor, but notthe actual postdexamethasone cortisol value, remained availablefor this analysis. The results are therefore limited to this categori-cal grouping.

In most cases, raters learned of suicides as they attempted toreach the patient for the next follow-up interview. In all cases ofdeath, raters sought to obtain descriptions of the circumstancesfrom informants, from the death certificate, and from pertinentmedical records. They then made a judgment of whether thedeath was due to suicide. Raters were not formally blinded to theDST results, and the cortisol values from tests ordered outside ofthe Schlesser et al. (29) protocol were recorded in the hospitalcharts. These results were not, however, among the variables as-sessed by raters in the Collaborative Depression Study.

Statistical Analysis

Three potential predictors of suicidemale sex, living alone,and the presence of a suicide attempt during the index illness ep-isode before intakewere selected on the basis of the previouslydescribed literature review. The presence of hopelessness in theweek preceding intake was also considered because it has been aparticularly robust predictor of suicide in diagnostically mixedsamples (31, 32). Hopelessness was considered present if the


DST AND SUICIDE

SADS item that quantified discouragement, pessimism, andhopelessness (item 244) was scored 4, 5, or 6 (moderate, se-vere, or extreme) for the worst week of the index episode. Be-cause some reports have found polarity to be predictive of even-tual suicide (33, 34), this variable was tested as well.

Each potential predictor was dichotomized, and the two result-ing groups compared by using nonparametric Kaplan-Meier pro-cedures (35). These procedures included censored data that re-sulted from loss to follow-up to derive risk estimates and tocompute log-rank tests for group comparisons. Alpha was set at0.05. The results presented here also include the simple propor-tions of those followed who were known to have committed sui-cide. The variables that generated a p value less than 0.1 were en-tered into a regression analysis in which risk for suicide was thedependent variable. Baseline differences between suppressorsand nonsuppressors in the proportion of subjects with individualrisk factors were analyzed by using two-by-two chi-square testswith continuity adjustments.

Results

Of the 246 probands who entered the Collaborative De-pression Study at the Iowa site, 83 underwent a DST within1 week of admission. Excluding the 13 patients known tohave died during follow-up, 61 (87.1%) completed at least2 years of follow-up, and 44 (62.9%) completed at least 15years. Some follow-up information was available for 78patients. These 78 differed from the remaining 151 pa-tients followed at the Iowa site in three of the baseline vari-ables listed in Table 1 and Table 2: those who received aDST were less likely to have psychotic features (N=12[15.4%] versus N=41 [27.2%]) (2=4.0, df=1, p



The fact that suppressors and nonsuppressors had cu-mulative probabilities of suicide that continued to divergeacross 15 years of follow-up indicates that the suicide riskimplied by a positive DST result is an enduring one. Thesimplest view of this finding presumes that the propensityto HPA-axis hyperactivity during a given major depressiveepisode is a characteristic of an individuals lifetime ill-ness. Although the few efforts to study the stability of DSTresults over time have shown only modest test-retest reli-ability (24, 36), high correlations have been demonstratedbetween postdexamethasone cortisol levels obtained dur-ing separate hospitalizations (24).

The eight suicides in this report include three of the fivesuicides described by Coryell and Schlesser (22). Thepresent study is therefore an extension, rather than a rep-lication, of the earlier study. Thus, three nonoverlappingstudies have associated nonsuppression of cortisol withmarkedly higher risks for subsequent suicide among pa-tients hospitalized for depression. The current study isalso an extension of an earlier analysis of this cohort inwhich nonsuppressors were significantly more likely tomake suicide attempts that were considered psychologi-cally serious during follow-up (15).

This is a surprising degree of consistency given the tem-poral instability of DST results demonstrated in earlierstudies (24, 36). It is also noteworthy that nonaffective dis-orders such as schizophrenia, alcoholism, and drug depen-dence also carry substantial suicide risks. DST results in

these disorders, even in the context of a superimposed de-pressive disorder, may not have the same predictive valueas they do in primary depressive disorder. This possibilitycould not be assessed in the current study group, given itslimited size and composition. Thus, the use of DST resultsto quantify risk for eventual suicide should, pending con-trary evidence, be confined to patients for whom major de-pressive disorder has been the dominant illness over time.

TABLE 2. Potential Predictors of Completed Suicide During Follow-Up in 78 Depressed Patients With a Baseline Dexam-ethasone Suppression Test (DST) Result

Patients WhoCommitted Suicide Risk Estimate Analysis

Predictor N N % % SE Log-Rank 2 df pSex 0.11 1 0.74

Male 26 3 11.5 17.4 9.7Female 52 5 9.6 10.8 4.6

Age at intake relative to median of 32 years 0.04 1 0.85At or below 39 4 10.3 14.1 7.1Above 39 4 10.3 11.7 5.6

Living alone 1.85 1 0.17Yes 16 0 0.0 0.0 0.0No 62 8 12.9 15.4 5.2

Feelings of hopelessness 0.94 1 0.33Yes 22 1 4.5 4.6 4.4No 56 7 12.5 15.9 5.8

Hamilton Depression Rating Scale score relative to median score of 27 0.53 1 0.47Below 35 4 11.4 11.9 5.6At or above 42 3 7.1 8.9 4.9

Serious suicide attempt in index episode 1.34 1 0.25Yes 16 3 18.8 21.4 11.0No 62 5 8.1 10.1 4.5

Polarity 0.01 2 0.99Nonbipolar 59 6 10.2 11.5 4.4Bipolar II 9 1 11.1 12.5 11.7Bipolar I 10 1 10.0 25.0 21.6

Delusional 1.57 1 0.21Yes 12 0 0.0 0.0 0.0No 66 8 12.1 15.0 5.1

DST result 7.87 1 0.005Nonsuppression 32 7 21.9 26.8 9.3Suppression 46 1 2.2 2.9 2.9

FIGURE 1. Cumulative Probabilities of Suicide Among De-pressed Patients by Baseline Dexamethasone SuppressionTest Resultsa

a Log-rank 2=7.9, df=1, p=0.005.

Perc

en

t W

ith

Co

mp

lete

d S

uic

ide

Time From Intake (years)

0

5

10

15

20

25

30

Inta

ke 151413121110987654321

Nonsuppressors (N=32)

Suppressors (N=46)


DST AND SUICIDE

The inclusion criteria used in the Collaborative Depres-sion Study are also relevant to generalizability. No AfricanAmericans were included, and the patients studied at theIowa center were otherwise more culturally homogenousthan patient samples in many other settings. More re-search will be necessary to determine whether these find-ings can be applied in diverse populations.

The DST is subject to a number of confounds that couldnot be rigorously controlled in this data set. These includea variety of medications and medical illnesses. Althoughthe DST was generally not administered in circumstancesthought to invalidate the results, the presence or absenceof these factors was not adequately documented in thesedata. Such factors as plasma level variability (37, 38) andthe stress associated with hospital admission (39) havealso been thought to influence DST results, but it has notbeen clearly shown that the control of these variables im-proves the diagnostic performance of the DST. In as muchas such variables are confounds, their influence wouldtend to produce false negative results, and this was clearlynot the outcome here.

Evidence for an association between nonsuppression ofcortisol and suicide risk seems strong. That HPA-axis hy-peractivity is a far more robust predictor than other clini-cal variables clearly needs independent replication. Giventhe widespread use of the DST in the early 1980s, there aremany potential cohorts that could be followed to furthertest this finding.

These findings also underscore the importance of theHPA axis in the pathophysiology of suicide. Although theserotonin system has been the major focus of biologicalresearch on suicide (40, 41), there is now substantial evi-dence that HPA-axis abnormalities may underlie seroto-nin abnormalities in the genesis of suicidal behavior (42).Further attention to the interplay between these systemsin suicidal behavior is clearly warranted. Whichever sys-tem eventually proves to be the more etiologically funda-mental, it appears that, at present, HPA-axis hyperactivityis the more clinically accessible, and therefore more clini-cally useful, of the two.

Received June 16, 2000; revisions received Oct. 12 and Nov. 20,2000; accepted Nov. 28, 2000. From the Department of Psychiatry,University of Iowa College of Medicine; and Dallas NeuropsychiatryAssociates, Dallas, Tex. Address reprint requests to Dr. Coryell, De-partment of Psychiatry, University of Iowa College of Medicine, 2-205MEB, Iowa City, IA 52242-1000.

Supported in part by NIMH grants MH-51324 and MH-25416.

References

1. Barraclough B, Bunch J, Nelson B, Sainsbury P: A hundredcases of suicide: clinical aspects. Br J Psychiatry 1974; 125:355373

2. Murphy GE, Wetzel RD, Robins E, McEvoy L: Multiple risk factorspredict suicide in alcoholism. Arch Gen Psychiatry 1992; 49:459463

3. Barraclough BM, Pallis DJ: Depression followed by suicide: acomparison of depressed suicides with living depressives. Psy-chol Med 1975; 5:5561

4. Roy A: Suicide in depressives. Compr Psychiatry 1983; 24:487491

5. Fawcett J, Scheftner WA, Fogg L, Clark DC, Young MA, HedekerD, Gibbons R: Time-related predictors of suicide in major affec-tive disorder. Am J Psychiatry 1990; 147:11891194

6. Newman SC, Bland RC: Suicide risk varies by subtype of affec-tive disorder. Acta Psychiatr Scand 1991; 83:420426

7. Dorovini-Zis K, Zis AP: Increased adrenal weight in victims of vi-olent suicide. Am J Psychiatry 1987; 144:12141215

8. Szigethy E, Conwell Y, Forbes NT, Cox C, Caine ED: Adrenalweight and morphology in victims of completed suicide. BiolPsychiatry 1994; 36:374380

9. Nemeroff CB, Owens MJ, Bissette G, Andorn AC, Stanley M: Re-duced corticotropin releasing factor binding sites in the frontalcortex of suicide victims. Arch Gen Psychiatry 1988; 45:577579

10. Arato M, Banki CM, Bissette G, Nemeroff CB: Elevated CSF CRFin suicide victims. Biol Psychiatry 1989; 25:355359

11. Targum SD, Rosen L, Capodanno AE: The dexamethasone sup-pression test in suicidal patients with unipolar depression. AmJ Psychiatry 1983; 140:877879

12. Banki C, Arato M, Papp Z, Kurcz M: Biochemical markers in sui-cidal patients. J Affect Disord 1984; 6:341350

13. Lopez-Ibor JJ, Saiz-Ruiz J, Pereze de los Cobos JC: Biological cor-relations of suicide and aggressivity in major depressions (withmelancholia): 5-hydroxyindoleacetic acid and cortisol in cere-bral spinal fluid, dexamethasone suppression test and thera-peutic response to 5-hydroxytryptophan. Neuropsychobiology1985; 14:6774

14. Pfeffer CR, Stokes P, Shindledecker R: Suicidal behavior and hy-pothalamic-pituitary-adrenocortical axis indices in child psy-chiatric inpatients. Biol Psychiatry 1991; 29:909917

15. Coryell W: DST abnormality as a predictor of course in majordepression. J Affect Disord 1990; 19:163169

16. Brown RP, Mason B, Stoll P, Brizer D, Kocsis J, Stokes PE, MannJJ: Adrenocortical function and suicidal behavior in depressivedisorders. Psychiatry Res 1986; 17:317323

17. Secunda SK, Cross CK, Koslow S, Katz MM, Kocsis J, Maas JW,Landis H: Biochemistry and suicidal behavior in depressed pa-tients. Biol Psychiatry 1986; 21:756767

18. Ayuso-Gutierrez J, Cabranes J, Garcia-Camba E, Almoguera I: Pi-tuitary-adrenal disinhibition and suicide attempts in depressedpatients. Biol Psychiatry 1987; 22:14091412

19. Schmidtke A, Fleckenstein P, Beckmann H: The dexametha-sone suppression test and suicide attempts. Acta PsychiatrScand 1989; 79:276282

20. Roy A: Hypothalamic-pituitary-adrenal axis function and sui-cidal behavior in depression. Biol Psychiatry 1992; 32:812816

21. Carroll BJ, Greden JF, Feinberg M: Suicide, neuroendocrine dys-function and CSF 5-HIAA concentrations in depression, in Re-cent Advances in Neuropsychopharmacology: Proceedings ofthe 12th CINP Congress. Edited by Angrist B. Oxford, UK, Perga-mon Press, 1980, pp 307313

22. Coryell W, Schlesser MA: Suicide and the dexamethasone sup-pression test in unipolar depression. Am J Psychiatry 1981;138:11201121

23. Norman WH, Brown WA, Miller IW, Keitner GI, Overholser JC:The dexamethasone suppression test and completed suicide.Acta Psychiatr Scand 1990; 81:120125

24. Coryell W, Schlesser MA: Dexamethasone suppression test re-sponse in major depression: stability across hospitalizations.Psychiatry Res 1983; 179189

25. Spitzer RL, Endicott J, Robins E: Research Diagnostic Criteria(RDC) for a Selected Group of Functional Disorders, 3rd ed.



New York, New York State Psychiatric Institute, Biometrics Re-search, 1978

26. Endicott J, Spitzer RL: A diagnostic interview: the Schedule forAffective Disorders and Schizophrenia. Arch Gen Psychiatry1978; 35:837844

27. Keller MB, Lavori PW, Friedman B, Nielsen E, Endicott J, Mc-Donald-Scott P, Andreasen NC: The Longitudinal Interval Fol-low-Up Evaluation: a comprehensive method for assessing out-come in prospective longitudinal studies. Arch Gen Psychiatry1987; 44:540548

28. Keller MB, Lavori PW, Mueller TI, Endicott J, Coryell W, Hirsch-feld RM, Shea T: Time to recovery, chronicity, and levels of psy-chopathology in major depression. Arch Gen Psychiatry 1992;49:809816

29. Schlesser MA, Winokur G, Sherman BM: Hypothalamic-pitu-itary-adrenal axis activity in depressive illness. Arch Gen Psychi-atry 1980; 37:737743

30. Murphy BEP, Patee CJ: Determination of plasma corticoids bycompetitive protein-binding analysis using gel filtration. Endo-crinology 1964; 24:919923

31. Beck AT, Steer RA, Kovacs M, Garrison B: Hopelessness andeventual suicide: a ten-year prospective study of patients hos-pitalized with suicidal ideation. Am J Psychiatry 1985; 142:559563

32. Beck AT, Brown G, Berchick RJ, Stewart BL, Steer RA: Relation-ship between hopelessness and ultimate suicide: a replicationwith psychiatric outpatients. Am J Psychiatry 1990; 147:190195

33. Morrison JR: Suicide in a psychiatric practice population. J ClinPsychiatry 1982; 43:348352

34. Black D, Winokur G, Nasrallah A: Suicide in subtypes of majoraffective disorder. Arch Gen Psychiatry 1987; 44:878880

35. Kalbfleisch JD, Prentice RL: The Statistical Analysis of FailureTime Data. New York, John Wiley & Sons, 1980, pp 1215

36. Coryell W, Smith R, Cook B, Moucharafieh S, Dunner F, HouseD: Serial dexamethasone suppression test results during anti-depressant therapy: relationship to diagnosis and clinicalchange. Psychiatry Res 1983; 10:165174

37. Johnson GF, Hunt GE, Caterson I: Plasma dexamethasone andthe dexamethasone suppression test. J Affect Disord 1988; 15:93100

38. OSullivan B, Hunt G, Caterson I: The plasma dexamethasonewindow: evidence supporting its usefulness to validate dexam-ethasone suppression test results. Biol Psychiatry 1989; 25:739754

39. Szadoczky E, Rihmer Z, Arato M: Influence of hospital admis-sion on DST results in patients with panic disorder (letter). AmJ Psychiatry 1986; 143:13151316

40. Mann JJ, McBride PA, Brown RP, Linnoila M, Leon AC, DeMeo M,Mieczkowski T, Myers JE, Stanley M: Relationship between cen-tral and peripheral serotonin indexes in depressed and suicidalpsychiatric inpatients. Arch Gen Psychiatry 1992; 49:442446

41. Mann JJ, Malone KM, Sweeney JA, Brown RP, Linnoila M, Stan-ley B, Stanley M: Attempted suicide characteristics and cere-brospinal fluid amine metabolites in depressed inpatients.Neuropsychopharmacology 1996; 15:576586

42. Lopez JF, Vazquez DM, Chalmers DT, Watson SJ: Regulation of5-HT receptors and the hypothalamic-pituitary-adrenal axis.Ann NY Acad Sci 1997; 836:106134

Documents

748