Adequacy of Antidepressant Treatment After Discharge and the Occurrence of Suicidal Acts in Major Depression: A Prospective Study

The patients had a mean age of 38.1 years (SD=11.7, median=36.5, range=18 to 72), 36% were male (N=49), 82% were Caucasian (N=111), and 33% had a high school education or less (N=45). Clinical characteristics of the subjects are shown in Table 1. The subjects were moderately depressed and had prominent suicidal ideation. The follow-up rates were 84% at 3 months (N=114), 76% at 12 months (N=104), and 64% at 24 months (N=87). Twenty-one (15%) of the 136 subjects attempted suicide during the follow-up period. None of the attempts was fatal. More than 50% of the attempts occurred within 5 months of discharge. For the attempters, the median time from hospital discharge until a suicidal act was 132 days (mean=185.1, SD=191.2, range=14–704 days) (Figure 1).

The first model, which included age, sex, lifetime number of suicide attempts at baseline, and interaction between age and number of suicide attempts, showed significant predictive power for suicide attempts during the follow-up period (Table 2). For each lifetime suicide attempt at baseline, the risk of suicide attempts in the follow-up period increased by about 30%; the 95% confidence interval (CI) was 16% to 48%. For each year of age, the risk of a suicide attempt in the future decreased. Thus, older patients were less likely to reattempt suicide than were younger patients who had made the same number of previous suicide attempts. In the second model (Table 3), the risk of a suicide attempt was 7.23 times as high (95% CI=1.7–30.9) for patients who experienced a major depressive episode in the follow-up period. Previous suicide attempts were a significant predictor and increased the risk of an attempt during follow-up in this model also. We found no relationship between the adequacy of antidepressant treatment during the follow-up period and the risk of making a suicide attempt in the 2 years after discharge (Table 3). However, antidepressant treatment during the follow-up period was rated inadequate both during most epochs and on average for the total study group (mean rating=2.1, SD=1.4). Moreover, 18 patients (13%) received no psychopharmacologic treatment during the entire follow-up period. Of those who were treated, the mean rating for treatment was 2.5 (SD=1.2, median=2.5), below the minimum for adequate treatment. Following are examples of mean daily antidepressant doses, in milligrams: desipramine, 177.0 (SD=93.3); bupropion, 284.8 (SD=122.9); fluoxetine, 32.9 (SD=19.2); sertraline, 120.7 (SD=49.8); paroxetine, 42.5 (SD=19.4); nefazodone, 175.0 (SD=35.4); citalopram, 38.3 (SD=20.4); mirtazapine, 88.1 (SD=87.5); and phenelzine, 45.0 (SD=22.5). One subject received 200 mg/day of fluvoxamine. Nine (43%) of the 21 subjects who attempted suicide were receiving adequate antidepressant treatment (i.e., had scores of 3 or higher on the Antidepressant Treatment History Form).

Subjects with a history of suicide attempts at baseline were not treated more vigorously during follow-up than nonattempters (Table 4). Furthermore, the intensity of treatment during periods of major depression in the follow-up phase was similar to the adequacy of treatment during the entire follow-up period, although slightly higher than during the nondepressed periods (Table 4). The presence or absence of comorbid borderline personality disorder was not related to the adequacy of antidepressant treatment; the mean rating of adequacy of treatment was 2.1 (SD=1.5) for subjects with borderline personality disorder and 2.2 (SD=1.4) for those without it (t=–0.39, df=104, p=0.70). When we added borderline personality disorder as a predictor in the survival analysis that included age, sex, and number of previous suicide attempts, borderline personality disorder did not predict suicide attempts during the follow-up period (hazard ratio=1.35, SE=0.25, p=0.23).

Relapse or recurrence of major depression increased the risk of suicide attempt during the 2 years after discharge from the hospital, underlining the importance of optimal maintenance antidepressant treatment as a suicide prevention strategy. We were unable to demonstrate that pharmacotherapy of major depression protected patients against suicide attempts in this naturalistic prospective study. One explanation is the overall undertreatment of depression, which undermines the statistical power to measure the benefit of treatment. We found that nine (43%) of the 21 follow-up suicide attempts occurred while patients were receiving adequate treatment. Of those 21 patients, four were depressed at the time of their attempts despite adequate treatment, suggesting that treatment resistance is important. However, little attention has been given in the literature to the role of treatment-resistant depression in suicidal behavior (27). The significance of the inadequacy of the treatment in this study is heightened by the fact that 54% of our inpatient group reported having made a suicide attempt before study entry and thus were at high risk for future suicide attempts. Nonetheless, the suicide attempters were not treated more aggressively than nonattempters. The undertreatment was not explained by the presence of borderline personality disorder, which could have caused diagnostic complications. The presence of borderline personality disorder had no effect on treatment adequacy. Moreover, in the previous attempters, the appearance of a major depressive episode during follow-up did not trigger more intense pharmacologic treatment.

The small number of suicide attempts during follow-up and possible confounds, such as psychotherapy received during follow-up, may have hampered detection of an effect of antidepressant treatment on suicide attempt rates during the follow-up period. Studies directly examining the effects of antidepressant medication on suicidal behavior suggest a role for somatic treatments in the prevention of suicidal acts. In a meta-analysis by Montgomery et al. (28), paroxetine reduced the frequency of completed suicide and suicidal thoughts more than did placebo or other active antidepressant treatment, but it did not reduce the number of suicide attempts. Similarly, in the National Institute of Mental Health collaborative depression study (N=643) there was 56% less risk for suicidal acts among patients taking fluoxetine than among patients receiving no somatic therapy, but this finding was not statistically significant, possibly because of clinical differences in illness severity and suicide history in the subjects treated with fluoxetine (29). In addition, of the subjects treated with fluoxetine, the proportion with suicidal behavior was significantly reduced from 38.9% before fluoxetine treatment to 3.8% during fluoxetine treatment. In contrast, Khan et al. (30) recently reported a meta-analysis that showed that placebo and active antidepressant treatment did not differ in their effects on rates of suicide attempts and completions. However, they studied subjects entered into phase II or III trials for new antidepressants that were ultimately approved by the Food and Drug Administration, and the subjects were selected so as to exclude the comorbidity or acute suicidality often seen in clinical populations. The strongest evidence for psychopharmacologic protection against suicidal behavior comes from controlled and naturalistic treatment studies of lithium and suicidal behavior (31–33). These studies suggest that lithium has an antisuicidal effect independent of its mood-stabilizing properties. Because many of the aforementioned studies were carried out in lithium clinics, where blood serum lithium levels are routinely monitored and noncompliance easily identified, it is possible that part of the robustness of lithium’s antisuicidal properties is due to compliance and not necessarily specific to lithium. At least one study has shown such an effect (34). In the current study, whether inadequate treatment was due to patients’ difficulties with compliance was not determined. Nonetheless, these studies support the notion that pharmacotherapy can have antisuicidal effects.

Improved identification of major depression and more widespread prescription of antidepressant medications appear to be associated with declining rates of completed suicide as well. In Hungary and Sweden, suicide rates show a negative correlation with rates of treated depression (35) and antidepressant prescription rates (36). The Gotland study (37) showed an increase in antidepressant prescription rates and a reduction of suicide rates after an educational program for general practitioners about the diagnosis and treatment of mood disorders. Conversely, lack of treatment of major depression has been associated with increased suicide risk. In Sweden, Isacsson et al. (36) reported that suicide risk among untreated depressed patients was 1.8 times higher than for those treated with antidepressants. In addition, postmortem studies have shown that major depression in suicide victims is infrequently diagnosed and ineffectively treated before death. Isacsson et al. (38) examined data from toxicological screens of 5,281 suicide victims in Sweden and detected antidepressants in only 12.4% of the men and 26.2% of the women. In the San Diego Study (39), half of the depressed suicide victims had consulted a physician in the 90 days before suicide, yet fewer than half of those were given prescriptions for antidepressants. Furthermore, toxicological tests revealed an absence of antidepressants in more than half of subjects for whom they were prescribed. Similarly, in Finland, Isometsä et al. (40) found that although 45% of suicide victims with major depression were receiving psychiatric treatment at the time of death, only 6% had received antidepressants or electroconvulsive therapy in adequate doses.

This naturalistic prospective study of major depression and suicidal acts during a 2-year follow-up period after hospitalization for a major depressive episode showed clinically serious undertreatment of depression. This was so even for patients with a history of suicide attempts, who are at a higher risk for future suicidal acts. In fact, we found that for each lifetime suicide attempt at baseline, the risk of suicide attempts in the follow-up period increased by about 30%. Poor treatment of depression during follow-up has been reported (41–43). In our previous retrospective study of patients seeking treatment at our center for major depression (44), we found that they were strikingly undertreated, regardless of their past history of suicide attempts. The present prospective study demonstrates a continuation of that practice of undertreatment after discharge from inpatient hospitalization for major depression, even for those at high risk for suicidal behavior. These findings are in agreement with the results of a study of 73 patients hospitalized after suicide attempts (28). Using a similar statistical analytic technique but a different patient group, Leon et al. (29) determined that for each additional prior suicide attempt, the risk of a suicide attempt in the follow-up period increased by 32%. Thus, both of these studies reveal that the hazard for suicidal acts in the follow-up period increases as a function of the number of suicide attempts before the index hospitalization, and they indicate that clinicians should obtain this information to assess risk (45). We have also confirmed our retrospective finding that suicide attempters and nonattempters are undertreated (44).

The crucial issue remains unresolved, namely, why is depression undertreated and how can this be corrected? The National Depressive and Manic-Depressive Association consensus statement on the treatment of depression (46) proposes several reasons for undertreatment. They include patient-based reasons, such as reluctance to seek treatment due to failure to recognize symptoms, fear of stigma, and limited access to health care, and provider-based factors, such as poor professional school education about depression, failure to consider psychotherapeutic approaches, lack of parity in insurance coverage for psychiatric illness, and prescription of inadequate doses of antidepressants for inadequate durations (46). Nonetheless, undertreatment of depression may explain the relatively small decline in the suicide rate in the United States, compared with Sweden (36, 47, 48), despite the development of newer, better tolerated, and safer antidepressant medications in the United States. Indeed, in 1995 antidepressant prescriptions were filled for about 33 per 1,000 persons in Sweden (48), 38% more than in the United States, where prescriptions were filled for approximately 24 per 1,000 persons (data from National Prescription Audit, a commercial database maintained by Information Marketing Systems Health America, Plymouth Meeting, Pa.). Perhaps Sweden’s government-funded national health care system imposes fewer obstacles to payment for mental health care than the U.S. system.

A major depressive episode during the follow-up period emerges as the most robust predictor of suicide attempts in our study and emphasizes the need for better diagnosis and treatment of major depression. The efficacy of educational programs should be tested in the United States and elsewhere, targeting not only general practitioners but also psychiatrists, other medical specialists, and primary care physicians. Future studies of antidepressants and their effects on suicidal acts should also include controlled treatment for high-risk patients, with monitoring of blood levels (where appropriate), so that the adequacy of antidepressant prescriptions can be ensured and the effect of pharmacologic treatment on suicidal behavior can be assessed.

Received Aug. 15, 2001; revisions received Jan. 22 and April 25, 2002; accepted May 7, 2002. From the New York State Psychiatric Institute and the Department of Psychiatry, Columbia University College of Physicians and Surgeons. Address reprint requests to Dr. Oquendo, New York State Psychiatric Institute, Unit 42, 1051 Riverside Dr., New York, NY 10032; [email protected]. Supported by NIMH grants MH-48514 and MH-62185. The authors thank Batsheva Halberstam for editorial assistance.