Journal ofAffectiL)e Disorders, 27 (1993) 117-121 0 1993 Elsevier Science Publishers B.V. All rights reserved
117 01650327/93/$06.00
JAD 00967
Suicidal behavior in bipolar and unipolar affective disorders: a meta-analysis David Lester Center for the Study of Suicide, Blackwood, NJ, USA (Received (Revised (Accepted
11 September 1992) 9 November.1992) 12 November 1992)
Summary A meta-analysis of studies of suicidal behavior in patients with bipolar and unipolar affective disorders found two possible trends: an excess of subsequent completed suicide in unipolar patients and an excess of subsequent attempted suicide in bipolar patients. Suggestions were made for future research.
Key words: Suicide;
Bipolar
affective
disorder;
Unipolar
Introduction It has long been known that patients with affective disorders have a high risk of subsequent suicide. Pitts and Winokur (1964) reviewed research up until that time and estimated that 15% of the deaths of patients with affective disorders were from suicide. More recently, research has focussed on differences in the suicidality of patients with the different categories of affective disorder, and the results of studies comparing bipolar and unipolar affective disorder appear to be in conflict (Lester, 1992). The present paper seeks to explore the research on this topic by (1) searching for all relevant studies comparing bipolars and unipolars; (2) distinguishing between the
Correspondence Suicide, RR41, USA.
to: David Lester, Center for the Study of 5 Stonegate Court, Blackwood, NJ 08012,
affective
disorder;
Meta-analysis
type of suicidal behavior (fatal suicidal acts, nonfatal suicidal acts and suicidal ideation) and its timing (before or after psychiatric evaluation); and (3) carrying out a meta-analysis (Rosthenal, 1984) in which the results of each study are put into a standardized format so that the results can be averaged. Method A search was made of Index Medicus and Biological Abstracts for all studies comparing the suicidal behavior of bipolar and unipolar affective disorder patients. The bibliographies of the articles located were examined for references to further studies. Only studies in which the data were reported fully were utilized in the present metaanalysis. (For example, Black (1989) claimed that unipolars have a higher suicide rate than bipolars, but he did not report the data in sufficient detail that a table of results could be recon-
118
strutted.) Care was taken to identify studies reporting data from the same set of subjects, but this was possible only when the sample sizes were identical. The studies included provided adequate data from the construction of 2-by-2 contingency tables (bipolar/unipolar versus suicidal/nonsuicidal) but, as can be seen from Table 1, often did not described the sample adequately. Frequently, the sex composition of the sample was not stated, nor were data on age reported. The follow-up period was, however, always reported. All of the studies were of inpatient and outpatient psychiatric patients, save for Morrison (1982) who studied patients in a private practice and Spalt (1980, 1983) who studied college students. The statistical data Morrison’s study are not presented in detail, and the present meta-analysis used a a contingency table reconstructed from the report. However, the table of data does not meet the same criteria as those used in all of the other studies. The results of each study were cast into 2-by-2 contingency tables with bipolar versus unipolar and suicidal versus nonsuicidal. A phi correlation coefficient was calculated for each table (a Pearson correlation calculated on binary data) since phi correlations can be averaged using Fisher’s technique (Guilford and Fruchter, 1973). Results Eight studies were located with data on subsequent completed suicide (see Table 1). One study found no suicidal behavior, while six of the remaining seven found a higher incidence of completed suicide in the unipolars (two-tailed binomial P = 0.012), but the single deviant result was based on such a large sample size that the overall average Pearson correlation failed to reach statistical significance (r = -0.014, df= 10,404, t = 1.43). However, this study was the only study with poor data presentation among the studies surveyed. The study is worth noting, though, since it was one of the few not using psychiatric inpatients or outpatients (it used patients seen in a private practice) and suggests, therefore, the usefulness of exploring the source of subjects on the results.
Two of these studies examined the differences for men and women separately. Black et al. (1987) found an excess of subsequent completed suicide in unipolars in both men (r = + 0.035) and women (r = +0.104), but Perris and D’Elia (1966) found the excess in men (r = t-0.041) but not in women (r = -0.034). Black et al. (1988) found the excess of subsequent completed suicide in both psychotic and nonpsychotic subjects (r values of = + 0.093 and + 0.050, respectively). Three studies were found on subsequent attempted suicide, two of which found a higher incidence in bipolars and one of which found a higher incidence in unipolars (see Table 1). The overall average Pearson correlation was - 0.081, df = 589, t = 1.97, two-tailed P = 0.05). Eight studies were found which studied prior suicide attempts (see Table 1). Five studies found a higher incidence of prior suicide attempts in the bipolars and three in the unipolars. The Pearson correlations averaged to +0.006 (df= 1639, t = 0.25), not significantly different from zero. Two of these studies studied the association in men and women separately. Linkowski et al. (1985) found a greater percentage of prior suicide attempts in the unipolars in both men and women (r values of + 0.097 and t-0.089, respectively), while Stallone et al. (1980) found a lesser percentage of prior suicide attempts in the unipolars for men and no difference for women (r values of -0.010 and 0.000, respectively). Linkowski et al. (1985) found more prior violent suicide attempts in bipolars (r = - 0.035) but more nonviolent suicide attempts in unipolars (r = + 0.144). One study (Till and Kapamadzija (1983; see Table 1) studied prior and subsequent attempted suicide combined and found an excess of suicidal behavior in the unipolars. Stallone et al. (1980; see Table 1) found a higher incidence of prior suicidal ideation in unipolars, and McGlashan (1984; see Table 1) found a higher incidence of subsequent suicidal ideation in unipolars. One study examined prior suicidal threats (see Table 11, reporting that unipolars had a higher incidence of such threats. Two studies compared subsequent completed suicide in Type I bipolars (with mania) and Type II bipolars (without mania) with conflicting results (see Table 2). Three studies examined prior
119 TABLE
1
Suicidal
behavior
Study
in bipolar
and unipolar
Sample size
Percent male
patients Mean
Z-by-2 Table
age
bipolar
unipolar
Phi coefficient
Type of patient
Follow-up period (years)
Subsequent completed suicide Angst et al. 258 (1979) 1291 Black et al. (1987) 1593 Black et al. (1988) Dingman and 66 McGlashan (1986) 158 Martin et al. (1985) ca. 6330 Morrison (1982) Perris and D’Elia 797 (1966) Roy-Byrne et al. 87 (1988)
Prior and subsequent Till and Kapamadzija (1983) Prior suicidal ideation Stallone et al. (1980)
3 96 7 579 7 579
16 143 24 681 34 979
+ 0.132
inpatients
12-16
+ 0.072
inpatients
2-14
+ 0.066 B
inpatients
2-14
suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal
1 19 0 19 12 1167 3 117 2 65
8 38 0 139 7 5 147 20 657 1 19
+0.166
inpatients
7-33
outpatients
6-12 O-12.5
+ 0.010
private practice inpatients
+ 0.046
inpatients
l-3
suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal
14 55 8 82 1 18
31 272 2 71 14 30
- 0.120
2
- 0.127
inpatients & outpatients inpatients
+ 0.286
inpatients
suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal
2 17 20 25 65 179 21 111 39 28 4 7 6 36 55 102
7 31 2 21 174 295 27 241 10 10 13 24 4 39 17 31
+0.102
inpatients
- 0.362
inpatients
+ 0.105
inpatients
- 0.084
inpatients & outpatients inpatients
+ 0.003
college students referrals college students referals inpatients
41
suicidal nonsuicidal
13 28
53 67
+0.110
inpatients
?
suicidal nonsuicidal
111 46
35 13
+ 0.021
inpatients
63
39
?
?
?
?
?
?
?
?
?
51
?
59
40
35
40
?
?
?
?
57
?
?
68
?
?
713
36
47
400
32
51
87
59
40
48
48
23
85
?
?
205
45
?
Subsequent attempted suicide 372 Coryell et al. (1987) Dunner et al. 163 (1976) 63 McGlashan (1984) Prior attempted suicide Dingman and McGlashan (1988) Dunner et al. (1976) Linkowski et al. (1985) Perugi et al. (1988) Roy-Byrne et al. (1988) Spalt (1980) Spalt (1983) Stallone et al. (1984)
suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal
30
attempted suicide 161 39
205
45
_ - 0.063
- 0.070 -0.011 - 0.077
h
1-15
l-9 15
120 TABLE
1 (continued)
Study
Sample size
Subsequent McGlashan, (1984)
Percent male
suicidal ideation 63 ?
Prior suicidal threats 48 Spalt (1980)
48
Mean
2-by-2 Table
age
bipolar
unipolar
suicidal nonsuicidal
5 14
30 14
+ 0.387
inpatients
23
suicidal nonsuicidal
1 10
10 27
+0.179
college student referrals
suicide attempts in the the two types of bipolars and again found conflicting results (see Table 2). In a single study, Stallone et al. (1980) found that Type II bipolars had more prior suicidal ideation than Type I bipolars (see Table 2). Discussion The present meta-analysis of suicidal behavior in patients with bipolar and unipolar affective disorders has revealed some interesting trends. First, studies of subsequent suicidal behavior suggests that completed suicide may be more com2
Suicidal
behavior
Study
in type I and type II bipolar
Sample size
Type of patient
?
a The studies by Black et al. (1987) and Black et al. (1988) may have overlapping b These appear to be the same subjects as reported in Dingman and McGlashan
TABLE
Phi coefficient
patients
2-by-2 table
Phi coefficient
Type I
Type II
Subsequent suicide attempts 83 suicidal Coryell et al. nonsuicidal (1987) 65 suicidal Dunner et al. nonsuicidal (1976)
7 29 10 29
7 40 9 16
- 0.060
Prior suicide attempts 98 Dunner et al. (1976) 132 Perugi et al. (1988) Stallone et al. 12.5 (1980)
suicidal nonsuicidal suicidal nonsuicidal suicidal nonsuicidal
4 68 8 17 24 48
4 22 13 94 24 29
+0.158
Prior suicidal ideation Stallone et al. 125 (1980)
suicidal nonsuicidal
44 28
41 12
+ 0.172
+ 0.090
- 0.213 + 0.121
subjects. (1988) and McGlashan
Follow-up period (years)
(1984).
mon in unipolars than in bipolars. Of the seven studies reviewed in Table 1, six found this excess. The sole dissenting study was the only one not using psychiatric patients in clinics, reporting instead data from a large private practice sample. The size of the sample was so great that it dominated the average of the phi correlation coefficients, but the data presentation in this report was poor. However, the discrepant result does suggest the importance in future research of investigating the effect of the type of sample on the results. Studies of subsequent attempted suicide produced a significant overall correlation coefficient indicating an excess in patients with bipolar affective disorders. However, of the three studies, only two had the association in this direction. These two trends indicate that more research is needed on this issue before we can draw reliable conclusions. Studies of prior attempts at suicide were evenly split on the direction of the difference, and the average correlation coefficient was not significantly different from zero. Thus, studies of subsequent behavior would seem to hold more promise than studies of prior suicidal behavior. There were too few studies of other suicidal behaviors for reliable conclusions to be drawn. Furthermore, studies of Type I bipolar patients and Type II bipolar patients have resulted in inconsistent results. Since different investigators do not always the same definition of Type I and Type II bipolar patients, the effect of definition must be examined in future studies.
121
This meta-analysis indicates that studies of subsequent completed and attempted suicide are needed in order to see whether the trends identified in this analysis can be confirmed. Such studies must present their data in such a way that meta-analyses are possible, and fuller descriptions of the samples studied is crucial. References Angst, J., Felder, W. and Frey, R. (1979) The course of unipolar and bipolar effective disorders. In: M. Schou and E. StrGmgren (Eds.) Origin, prevention and treatment of affective disorders. London: Academic, pp. 215-226. Black, D.W. (1989) The Iowa record-linkage experience. Suicide & Life-Threatening Beha. 19, 78-98. Black, D.W., Winokur, G. and Nasrallah, A. (1987) Suicide in subtypes of major affective disorder. Arch. Gen. Psychiatry 44, 878-880. Black, D.W., Winokur, G. and Nasrallah, A. (1988) Effect of psychosis on suicide risk in 1593 patients with unipolar and bipolar affective disorders. Am. J. Psychiatry 145, 849-852. Coryell, W., Andreasen, N.C., Endicott, J. and Keller, M. (1987) The significance of past mania or hypomania in the course and outcome of major depression. Am. J. Psychiatry 144, 309-31s. Dingman, C.W. and McGlashan, T.H. (1986) Discriminating characteristics of suicides. Acta Psychiatrica Stand. 74, 91-97. Dingman, C.W. and McGlashan, T.H. (1988) Characteristics of patients with serious suicidal intentions who ultimately commit suicide. Hospital Community Psychiatry 39, 295299. Dunner, D.L., Gershon, E.S. and Goodwin, F.K. (1976) Heritable factors in the severity of affective illness. Biological Psychiatry 11, 31-42.
Guilford, J.P. and Fruchter, B. (1973) Fundamental Statistics in Psychology and Education. New York: McGraw-Hill. Lester, D. (1992) Why people kill themselves. Springfield, IL: Charles Thomas. Linkowski, P., de Maertelaar, V. and Mendlewicz, J. (19851 Suicidal behavior in major depressive illness. Acta Psychiatr. Stand. 72, 233-238. Martin, R.L., Cloninger, R., Guze, S.B. and Clayton, P.J. (1985) Mortality in a follow-up of 500 psychiatric outpatients. Arch. Gen. Psychiatry 42, 58-66. McGlashan, T.H. (1984) The Chestnut Lodge follow-up study. Arch. Gen. Psychiatry 41, 573-601. Morrison, J.R. (1982) Suicide in a psychiatric private practice. J. Clin. Psychiatry 43, 348-353. Perris, C. and D’Elia, G. (1966) A study of bipolar (manic-depressive) and unipolar recurrent depressive psychoses. Acta Psychiatr. Stand. 42, Suppl. 194, 172-189. Perugi, G., Musetti, L., Pezzica, P., Piagentini, F., Cassano, G.and Akiskal, H. (1988) Suicide attempts in primary major depressive subtypes. Psychiatria Fennica 95-102. Pitts, F.N. and Winokur, G. (1964) Affective disorder. J. Nervous Mental Dis. 139, 176-181. Rosenthal, R. (1984) Meta-analytic procedures for social research. Beverly Hills, CA: Sage. Roy-Byrne, P.P., Post, R.M., Hambrick, D.D., Leverich, G.S. and Rosoff, AS. (1988) Suicide and course of illness in major affective disorder. J. Affect. Disord. 15, l-8. Spalt, L. (1980) Suicide behavior and depression in university student referrals. Psychiatr. Quart. 52, 235-239. Spalt, L. (1983) Suicide attempts and nonpatient college students. In: J.P. Soubrier and J. Vedrinne (Eds.), Depression and Suicide. Paris: Pergamon, pp. 627-631. Stallone, F., Dunner, D.L., Ahearn, J. and Fieve, R.R. (1980) Statistical predictions of suicide in depression. Compr. Psychiatry 21, 381-387. Till, E. and Kapamadzija, B. (18083) Endogenous depressions and suicidal behavior. In: J.P. Soubrier and J. Vedrinne (Eds.), Depression and Suicide. Paris: Pergamon, pp. 235238.