Background
Being affected by a mental disorder is a significant risk factor for suicide. Therefore, the identification of patients at risk for suicide among patients with mental disorders is of high clinical relevance. Psychological autopsy studies have shown that depression is the most common mental disorder at the time of suicide, occurring in half to two-thirds of cases [
1‐
4]. Unipolar depression has been associated with increased risk for suicide. In fact, suicide accounts for nearly 20% of the deaths in psychiatric samples of depressed patients [
5]. From a clinical point of view, the severity of depressive symptoms has been positively related to suicide [
6].
The presence of psychotic features (delusional ideas and/or hallucinations) during a major depressive episode defines psychotic major depression (PMD) [
7]. Unipolar PMD compared to non-PMD is a more severe form of depression with greater feelings of guilt and more pronounced psychomotor disturbance. In addition, PMD compared to non-PMD patients were found to be more severely cognitively impaired with greater HPA axis abnormalities and with lower serum dopamine
β-hydroxylase activity. Besides, PMD is associated with more brain atrophy and poorer short-term outcome with respect to daily functioning, residual depressive symptoms, mortality, and suicidality (see [
8‐
10] for reviews).
As regards completed suicide (henceforth suicide), studies have reported conflicting results: some comparing suicides between PMD and non-PMD patients have shown that psychosis increases the risk of suicide [
11‐
13], whereas other studies failed to show such an effect [
14‐
19]. However, no study has reported decreased suicide risk in unipolar PMD patients.
A recent systematic review of our group coupled with meta-analysis has shown that the presence of psychosis during a major depressive episode doubles the risk of a suicide attempt both lifetime and in the acute phase of the disorder [
20]. In the same direction, Zalpuri and Rothchild [
21] in a recent systematic review reached the same conclusion. Taking into consideration that a history of suicidal attempt increases the risk of a future suicide at 30–40 times compared with the risk in general population [
22,
23], these findings stress further the need to investigate and quantify the impact of psychosis on suicides in unipolar depressed patients.
Thus, the aim of the current study was to elucidate further through a systematic review coupled with meta-analysis whether the lifetime risk for suicide of unipolar PMD compared to non-PMD patients is increased. An additional purpose of this study was to investigate whether the suicide risk in PMD in comparison to non-PMD patients is associated with (a) age, (b) depression severity, (c) mood congruence of psychotic features, (d) use of more violent methods, and (e) the existence of comorbidity either psychiatric and/or physical.
Discussion
This systematic review and meta-analysis of nine studies, comprising in total 33,873 patients, among them 828 suicide victims, showed that patients with PMD are at 1.21-fold higher lifetime risk of committing suicide than patients with non-PMD. An additional sub-analysis of eight studies comprising 2787 patients, among them 149 suicide victims, after the exclusion of a large study with a huge weight in the analysis (86.62%) revealed a 1.69-fold increased risk for suicide of PMD patients, as well.
Our findings have important clinical implications. Patients with PMD during a major depressive episode experience psychotic phenomena, impulsivity, and strong feelings of guilt, shame, intense anxiety, fear, etc., being in a chaotic mental state and unable to control their actions and suicidal behavior [
53]. Unfortunately, these phenomena and symptoms are frequently underdiagnosed as these patients are often too psychomotorly disturbed and avoid giving adequate verbal information about their psychotic experiences, symptoms, and suicidal thoughts [
65], and thus, their suicidal intention remains frequently overlooked. On the other hand, our study reports alarming findings on suicide risk of PMD patients and highlights the importance of recognizing and treating adequately psychotic depression and assessing patients’ suicidality, as well [
66].
To the best of our knowledge, only the review of Zalpuri and Rothschild [
21], comprising eight studies [
11,
12,
14,
16,
17,
61‐
63] has investigated the effect of psychosis on suicide risk using a systematic research method without a meta-analytic leg though. The authors conclude that “most studies did not find any differences in completed suicides between the two groups; however, in some studies, the psychiatric status of the victims at the time of suicide was unknown”. The systematic part of our study comprised 12 studies (four more studies: 13, 19, 60, 64), while the meta-analytic part, which quantified the effect of psychosis on unipolar depressive patients suicide behavior, included nine studies [
11,
12,
14,
16,
17,
19,
62‐
64]. We also attempted to investigate the associations between PMD’s suicides with age, severity of depression, phase of the disorder, mood congruence, suicidal method used, and comorbidity.
The overall results of our main meta-analysis of nine studies showed an increased 1.21-fold lifetime risk of suicide for PMD patients compared to non-PMD. We consider this finding as very interesting: so far, the results of large and influential studies did not show any differences between the two groups of patients in the long term [
14,
16,
17], and thus, the above-mentioned comment of Zalpuri and Rothschild [
21] seemed to be justified. Furthermore, the finding of our main meta-analysis attempts to answer whether psychosis is a risk factor for suicide in the context of unipolar depression, a question which is often encountered in the literature. We consider this research question as quite general in nature. In the current meta-analysis, two different sorts of studies were included: two studies [
11,
63] investigated the suicide risk of PMD patients during the acute phase of the disorder and the remaining seven studies presented the long-term suicide risk of PMD patients. In the Roose et al.’s [
11] study, PMD patients were found to be 5.3 times as likely to commit suicide compared to non-PMD patients. The patients of this study committed suicide while hospitalized in the inpatient setting or after having recently eloped or while out on pass. Therefore, they were thoroughly assessed for their psychopathological status at the time of suicide. In addition, Wolfersdorf et al. [
63] following the same methodology with the above-mentioned study found only very small and inconclusive numbers of suicide victims. The results of Roose et al.’s [
11] study are in line with the findings of our recently published meta-analysis of 20 studies [
20] which found a twofold increase of suicide attempt during the acute phase. The authors of the Roose et al.’s [
11] study attempt to answer the question “is the risk for suicide of PMD compared to non-PMD patients increased during a major depressive episode?” We consider this question as the most crucial in clinical practice and their results are alarming at the highest degree. Of note, the weight of the Roose et al.’s [
11] study in the meta-analysis was extremely low (0.58%) and Wolfersdorf et al. [
63] 0.62%, as well, and thus, practically, the pooled OR of our main analysis is extracted from the remaining seven studies. Surprisingly, the results of the Roose et al. [
11] have not been replicated in a large study yet, and this certainly constitutes a major weak point in PMD’s suicidal behavior research.
The remaining seven studies reported data on short- [
62] or long-term time period [
12,
14,
16,
17,
19,
64]. These studies assessed initially patients as either PMD or non-PMD, without providing information over the presence of psychosis in patients’ depressive episodes during the course of the disorder, with the exception of the Leadholm et al.’s [
17] study which was a registered-based nation-wide study and followed patients within a period of 17 years providing data over the number of patients with psychotic or non-psychotic episodes or with both. These studies followed up their patients without providing separate risk values for suicide for the acute phase of the disorder or the remission though, attempting thus to answer the question “is the risk for suicide of PMD compared to non-PMD patients increased in the long-term?”. Given the much lower fixed-effect pooled OR of our main meta-analysis in comparison to the Roose et al. [
11] study, there is an “attenuation” in the risk of suicide between the study of the acute phase [
11] to the studies offering lifetime data.
In an attempt to interpret this previously mentioned difference, we should take into consideration several parameters: these studies consisted of initially hospitalized patients, gathering information about suicides from national data bases [
14,
17], or by interviewing living probands and first-degree relatives even many years after death [
16]. In these studies, retrospective in nature, the patients’ psychopathological state at the time of suicide remains unknown, which constitutes a major limitation. Even studies which closely followed initially admitted patient in short- [
62], or long-term [
19] follow-up do not provide an accurate diagnosis at the time of suicide. In addition, patients on long-term follow-up receive antidepressants and for long periods of time are in remission.
Furthermore, the diagnoses of PMD and/or non-PMD are not stable: in the Kessing [
60] study, only 50% of patients with a single PMD episode re-presented the same symptoms in their second admission, and in the Ruggero et al.’s [
50] study, PMD patients shifted to other diagnoses at a rate over 40% across 10 years. Coryell and Tsuang [
16] determined a switch to bipolarity at a rate of about 12.5% for both PMD and non-PMD patients and Maj et al. [
19] at 10.1%. Moreover, we should take into consideration that, in the course of the disorder, patients present not only PMD but also non-PMD episodes [
67]. Finally, the Leadholm et al. [
17] and the Suominen et al. [
12] studies used ICD-10 criteria for PMD that include beyond delusions or hallucinations depressive stupor, as well.
It should also be underlined that Schatzberg and Rothschild [
68] presented data in favor of poorer short-term outcome (< 2 years) of PMD patients compared to non-PMD patients; however, their long-term outcome (> 2–5 years) was found to be similar to their non-PMD counterparts. Studies that compared the course and outcome of PMD to non-PMD patients have found PMD patients to present more enduring depressive episodes, more depressive episodes, fewer weeks in remission, higher frequency of relapse of recurrence, higher number of hospitalizations, more psychological impairment, higher rates of alcoholism, and a twofold greater risk of death [see, for review,
8,
68,
69]. However, other studies found similar rates between the two groups with regard to the number of depressive episodes and hospitalizations [
70] and global functioning at the end of the follow-up period [
19,
71]. We consider that the suicide risk in PMD follows a similar pattern to the impact of psychosis in the short- or long-term course of the disorder as presented by Schatzberg and Rothschild [
68]. Consequently, these follow-up studies might present an attenuation in PMD’s suicide OR due to the long periods of patients in remission, use of antidepressants, lack of accurate diagnosis at the time of suicide, PMD’s diagnostic instability, and the use of ICD-10 criteria.
At this point, we should comment that the Leadholm et al.’s [
17] study is by far the largest study included in the meta-analysis, carried out in Denmark, following up the course of 34,671 PMD and non-PMD patients, with 755 of them committing suicide, without finding any statistically significant differences between the two groups. Nevertheless, the inclusion of this interesting study raises two methodological questions. First, its weight in the meta-analysis is huge (86.62%), and thus, its overall results tend to determine the findings of the meta-analysis, at a crucial degree. Furthermore, the PMD and non-PMD diagnoses were not mutually exclusive, leading thus to the inclusion of 3585 patients in both PMD and non-PMD groups, among them 76 who committed suicide. To adjust for any effects due to patients appearing in both groups, in our main meta-analysis, we removed the numbers of patients with both diagnoses. We speculate that this removal was necessary not only to avoid the inclusion of 3585 patients in both groups, but also to reduce the probability of the type II error. More specifically, the ICD-10 criteria for psychosis include depressive stupor and tend either way to categorize, in the PMD group, a number of patients with no real psychosis, but with only severe psychomotor retardation. When in the PMD group we included patients with both ICD-10 PMD and non-PMD episodes, this tendency to type II error might have become even stronger. We consider that, for this reason, our main meta-analysis showed both the fixed and random-effects being statistically significant, whereas, in the last one, the fixed-effect models remained statistically significant, but the random-effects were marginally not significant.
Owning to the previously mentioned considerations, we performed a sub-analysis excluding the above-mentioned large study and including the remaining eight studies. PMD patients were found to be 1.69 times more likely to commit suicide compared to non-PMD ones. This sensitivity analysis comprised studies of four countries (USA, Germany, Italy, and Finland) with even more alarming results. The strong point of this sensitivity analysis lies on the low heterogeneity of the included studies with respect to research question. On the other hand, the total number of included patients was reduced, and this constitutes its weak point.
Furthermore, it should be noted that the three excluded studies, namely the Isometsä et al. [
61] psychological autopsy study and the population-based Brådvik et al. [
13] study presented data which showed a preponderance of PMD patients among suicide victims. The third study, in particular the Kessing [
60] one, did not report statistically significant findings, but its sample was part of the Leadholm et al. [
17] study.
Suicides in association with age and depression severity
Our search did not isolate any studies concerning PMD’s suicides in adolescence and in old age. Only Schneider et al. [
64] and Suominen et al. [
12] included some patients over 11 or 16 years, respectively, and an undetermined number of elderly people, as well. In general, we consider that all studies included patients within the age range of adulthood.
The investigation of the contribution of depression severity to PMD patients’ suicides was between the purposes of our study. At this point, it should be noted that our main analysis is conducted between patients suffering from severe depression. Leadholm et al. [
17] and Suominen et al. [
12] compared both groups of patients suffering from severe depression and the weight of these two studies in the main meta-analysis is over 90%. As, we analytically mention in our Results, the clinical characteristics of Black et al. [
14], Coryell and Tsuang [
16] and Schneider et al. [
64] studies, which do not offer data on depression severity and occupy an additional 7.55% weight in the main meta-analysis, imply the inclusion of patients suffering to a great extent of severe depression, as well. The remaining studies share the same clinical characteristics but offer only a small number of individuals in the analysis. In other words, our main meta-analysis may show that the existence of psychosis in the context of severe depression elevates the risk of suicide. In addition, the Suominen et al.’s [
12] study displayed data on suicides across different degrees of depression severity, and as we calculated in the Results, this study clearly showed that the greater the depression severity, the higher the risk for suicide. Moreover, it showed that psychosis in the context of severe depression doubles the risk of suicide. However, not all studies are in agreement: Kessing [
60] comparing PMD patients to all non-PMD patients (ICD-10 criteria), regardless of depression severity, did not find any statistical significant differences in terms of suicides. Thus, our meta-analysis reports that PMD patients manifest elevated risk of suicide even when they are compared with non-PMD patients who suffer from severe depression. However, the ICD-10 categorization of PMD only in the context of severe depression and the inclusion of depressive stupor in the criteria of psychosis constitute obstacles on the further elucidation of suicide risk, PMD and depression severity association.
Suicide and method used (violent or not), mood congruence, and comorbidity
The systematic review leg of this study revealed a psychological autopsy study [
61] of all suicides committed within 1 year in Finland, which found that PMD patients were more likely to commit violent suicides than non-PMD patients. In the Suominen et al.’s [
12] study, patients with severe depression used a more violent method; however, the authors do not provide data separately on the suicides committed by PMD and/or non-PMD patients. As regards suicidal attempts, Hori et al. [
35] have reported significantly higher levels of violent suicidal attempts in PMD patients, and Coryell et al. [
15] found that “non-PMD patients were twice as likely to have made attempts judged to be medically and psychologically non- serious”. In the same vein, Lyness et al. [
38] have shown that patients with PMD committed more “severe” suicidal attempts in comparison to non-PMD patients. On the other hand, in a study of our group [
18], no differences were found between the two groups in terms of the method used. Thus, more research is warranted to clarify the effect PMD exerts on suicide method used.
In respect to the mood congruence of psychosis on suicide risk, Kessing [
60] found no differences in the risk of suicide between PMD patients with or without mood congruent psychotic features. In the Maj et al. [
19] and in the Schneider et al. [
64] studies, the numbers of PMD patients with or without mood congruent psychosis who committed suicide are too small to be conclusive. In addition, our recent meta-analysis [
20] on suicidal attempts found the evidence scarce on this topic. Moreover, rather the high prevalence of mixed psychotic features in PMD, 58% in the Burch et al.’s [
72] study underlines the restrictions that research work can face to clarify the differential effect of mood congruent vs. mood incongruent psychosis on suicide risk evaluating the two concepts in a categorical way.
Surprisingly, only two studies [
17,
61] assessed a number of parameters regarding psychiatric or physical comorbidity such as anxiety, psychoactive substance use, physical illness, or personality disorder but found only limited associations: in the Leadholm et al.’s [
17] study, PMD patients were less likely to committing suicide if they suffered from an organic mental disorder. Consequently, the evidence at this point is restricted to the findings of the afore-mentioned two studies, and thus, no conclusions on this topic can be drawn [
73].
Limitations
In addition to the methodological issues discussed above regarding the studies included, such as lack of accurate diagnosis at the time of suicide, PMD’s diagnostic instability, and the use of ICD-10 criteria, further limitations of this study are the search of relevant articles only in English literature and the inclusion of studies carried out in different time periods with possibly different general trends in suicidal behavior. On the other hand, we performed a large search in English literature and in the so-called “gray literature” as well, including, in our meta-analysis, 9 studies with a large number of patients. Furthermore, we consider the clarification of the research questions as a crucial contribution to the field.