Donald W. Black, M.D., M.S.
George Winokur, M.D.
Emmanuel Mohandoss, M.S.
Robert F. Woolson, Ph.D.,
and Amelia Nasrallah, M.A.

Ann Clin Psych 1989;1:165-173

This article reports mortality risk among 1076 Iowans with major affective disorders (705 primary unipolar, and 152 bipolar depressives) compared to that of the general population. Patients were divided into four treatment groups depending on primary mode of therapy during the index admission; the groups included electroconvulsive therapy, adequate antidepressants, inadequate antidepressants, and neither treatment. All patients in the sample had an increased risk for an early death. A high risk for suicide was found for patients within each individual treatment group during the follow-up, especially the first 2 years when 69.4% (n= 25) of total suicides occurred. There were no significant differences in the risk for suicides, or deaths from all causes combined, among patients in the four treatment groups. Furthermore, mortality did not differ between patients having a lifetime history of ECT and patients never having had ECT. We conclude from a short-term follow-up of depressives that mode of therapy received in the hospital has minimal influence on subsequent mortality, including suicide.

Do modern psychiatric treatments help prevent suicide? These provocative questions have been asked repeatedly since effective treatments for the major psychiatric illnesses were developed beginning with electroconvulsive therapy (ECT). Few answers have been provided. Although several early studies on convulsive therapies (ECT or metrazol) were encouraging [1-3], other reports were not. More recently, two studies found lower death rates in depressives [6] and schizoaffectives [7] treated with ECT, but suicide rates were unchanged. Three additional studies since 1976 have not shown ECT to reduce suicide rates in depressives, either [8-10].

Confounding the effect of somatic treatment on death rates has been the independent trend in general mortality and suicide rates. In the past, both natural and unnatural causes of death were highly excessive [11, 12, 14], but now death from suicides and accidents is primarily responsible for the excess [13, 14]. Death from natural causes in psychiatric patients has been declining, however, most likely due to improvements in the availability and efficacy of general medical care, and deinstitutionalization, and may no longer be excessive [11, 13, 15]. Because natural causes of death may no longer be excessive, any protective effect that ECT may have had in the past in these deaths may now be unimportant. Any effect that ECT might have on preventing suicide could still be critical, however. Also of concern is whether antidepressants, particularly tricyclics, might actually increase death rates, due either to their demonstrated effects on vascular and cardiac conduction at both therapeutic and supratherapeutic dosages [16, 17].

We adjusted for length of follow-up because study subjects were not all followed for the same amount of time. For example, a person followed 10 years would have a greater cumulative risk for mortality than someone followed 1 year. This method is more fully described elsewhere [22].

Expected and observed numbers of deaths were compared using the Freeman-Tukey-corrected chi square. The Freeman-Tukey correction was used because it is more conservative than the regular chi square and many of our expected numbers were so small. Standardized mortality ratios (SMRs) were calculated and represent the ratio of observed to expected mortality. An SMR greater than I means that observed death exceeds expectation. Ninety- five percent confidence limits were calculated for the SMRs using Byar's method [26].

RESULTS

Of 1076 patients, 372 (34.6%) received ECT, 180 (16.7%) received adequate antidepressants, 317 (29.5%) received inadequate antidepressants, and 207 (19.2%) received neither ECT nor antidepressants during the index hospitalization. Using a four-way chi square, there were significant differences among the groups on age, marital status, prior episodes, prior suicide attempts, precipitating factors, delusions, and recovery at discharge. There were no differences in sex or suicidal ideations. Patients receiving ECT were older than the others, were more likely to be married (probably because of their advanced age), tended to have more delusions, and were less likely to have attempted suicide. Patients within the two antidepressant groups were similar except that patients receiving adequate antidepressants were more likely to have had prior episodes of illness. The group of patients receiving neither treatment differed from the other groups. These patients were younger, were less likely to be married, nearly two thirds had reported factors precipitating their depressions, nearly one-half had prior suicide attempts, and few were reported as receiving drug prophylaxis. At hospital discharge, patients receiving ECT were more likely to have recovered than patients in the other treatment groups.

Thirty-six suicides were identified in the record-linkage and comprise 3.3% of the study sample (Table 1). The following percentage of the total sample size for each diagnostic group committed suicide: ECT 3.2, adequate antidepressant 2.8, inadequate antidepressant 3.5, and neither treatment 3.9. There were no significant differences for the unadjusted (crude) suicide rates among the treatment groups (x2 = 0.944, df = 3). Suicides as a percentage of the total deceased were ECT 23.5, adequate antidepressant 33.3, inadequate antidepressant 50.0, and neither treatment 53.3.

Table 2 shows the distribution of the 103 deaths by treatment group and portion of follow-up. Forty (38.8%) deaths occurred during the first 2 years of the follow-up. During this portion of the follow-up, general (all cause) mortality was significantly excessive compared with expectation for the groups receiving ECT and inadequate antidepressants. For the entire follow-up period, the mortality is excessive for the groups receiving ECT or neither treatment. There are no significant differences between SMRs among the four treatment groups, as demonstrated by overlapping confidence intervals.

Twenty-five (69.4%) suicides occurred during the first 2 years of follow-up (Table 3). This was particularly obvious in the inadequate antidepressant group; 10 of 11 suicides occurred during this phase of follow-up. SMRs in each treatment category are about 15% in long-term follow-ups, the percentage of suicides is much higher in a short-term study [23, 281. Further, the study shows no significant differences in death rates among four treatment groups.

Literature Survey

The purpose of this study was to determine whether specific treatment categories were associated with a differential risk for suicide, which had been suggested by early studies. Ziskind and colleagues followed 197 patients with affective psychoses, mostly manic-depression, for a mean of 40 months. Eighty-eight had received convulsive therapy (ECT or metrazol); 109 had refused convulsive therapy, had symptoms too mild to warrant the treatment, or had a contraindication to ECT. There were 13 deaths in the control patients, with 9 by suicide compared with 3 deaths with 1 suicide in the convulsive therapy patients. Huston and Locher [21 compared patients with involutional psychosis treated with ECT with those receiving conservative therapies. None of the patients in the convulsive therapy groups committed suicide but 13% of those receiving conservative therapies did. Unfortunately, the follow-up periods differed for the different groups complicating subsequent data interpretation. In a later report on manic-depressive illness [3], these same authors found that ECT-treated patients followed for a mean of 36 months had a 1% suicide rate while the control patients followed for a mean of 82 months had a 7% suicide rate. Milstein and co-workers [10] recently reexamined these studies and noted that although they are suggestive of a beneficial effect, only the data from Ziskind and colleagues [1] yielded a statistically significant finding with a Fisher's exact probability of 0.029. The results of the current study are different from the results of Ziskind and colleagues [1] and Huston and Locher [2], but are consistent with the conclusions of Eastwood and Peacocke [8], Babigian and Guttmacher [91, and Milstein and colleagues [10]. We found significant, but similar risk for suicide among patients in all treatment categories. Suicide as a percentage of the total number of patients who died differs between groups but this is easily explained on the basis of age. Patients in the ECT group were older than the patients in other groups and would be more likely to die from natural causes. SMRs in the different treatment groups (Table 3) might suggest that ECT and adequate antidepressants might carry a lower risk for suicide, but this could not be demonstrated statistically. However, expected values as small as those reported in Table 3 make any conclusions tentative. Clearly, one must be cautious in basing interpretations on such small numbers.

Since there was no difference in suicide rates during the first 2 years after hospital discharge, was it possible that a shorter follow-up might reveal a difference between groups? Perhaps the protective effect of a treatment is shorter lived. We looked at this possibility (Table 4), but found no discernible trend in the pattern of suicides. There was also no apparent association between treatment group and follow-up interval.

Concerns and Caveats

Because ECT and antidepressants are efficacious [32, 33], we might expect them to lower suicide rates, at least in a short-term follow-up. However, during a relatively short follow-up we found no statistically significant difference in general mortality or suicide rates among treatment groups, nor did a lifetime history of ECT appear to have any influence on mortality. One would like to believe that effective treatments would decrease the likelihood of an early death. Possibly, suicide rates would have been much higher had somatic treatments not been used at all. We cannot state categorically that ECT or antidepressants failed to prevent suicides; our methods may not be sensitive enough to measure the effect of adequate treatment on individual patients. As Murphy has observed, "a saved life is a statistical non-event" [34, p. 573]. It is reassuring, however, that death rates are not higher in the antidepressant groups because antidepressants are associated with cardiovascular disturbances, are reported to lead to sudden death, and can be lethal in overdose.

A potential criticism of the analysis is that the treatment groups are inherently unequal, so that it may be inappropriate to draw comparisons among them. As treatment selection was left to the clinician, group assignment was nonrandom and, as we noted earlier, the four groups differ substantially In many respects, including age, age of onset, psychosis, suicide attempts, and aftercare. We have attempted to correct for some of these inequalities. First, we restricted our sample to diagnostic groups shown to have similar suicide rates. We deleted front our sample medically ill persons. We also used SMRs to correct for age, sex, and duration of follow-up. Regardless, some may argue that even with appropriate statistical safeguards, the groups remain unequal, that patients receiving ECT are simply sicker than patients receiving antidepressants, or no somatic treatment, and that such patients may have higher suicide rates. It is curious that the ECT group had fewer prior suicide attempts. Review of the literature, however, shows that suicide in depressed persons has not been clearly associated with clinical symptoms [35, 36], diagnostic subtype [21, 36], or psychosis [37, 38]. According to our results, suicide may also be unrelated to treatment.

We must emphasize that the results pertain to a highly select sample-depressives hospitalized at a tertiary care facility. The results should not be generalized to other groups, including outpatients. Indeed, most psychiatric patients are not hospitalized [39]. In particular, the results should not be generalized to never-treated outpatients. By virtue of entering the hospital, all our patients received "treatment," although some may not have had an active biologic therapy. Other treatments are provided in hospitals, including individual or group psychotherapy, milieu therapy, and activities and occupational therapies, whose effect on mortality cannot easily be assessed. Unfortunately, we also had no way to control for prophylactic treatments. This important variable may well affect mortality rates. Future studies will need to address this issue, as it is clear that drug prophylaxis affects relapse rates among the mood disorders.