Antipsychotic Monotherapy for Major Depressive Disorder: A Systematic Review and Meta-Analysis

Abstract Although several randomized controlled trials (RCTs) have compared the effectiveness, efficacy, and safety of antipsychotic monotherapy (APM) versus placebo in patients with major depressive disorder (MDD), no meta-analysis has examined this topic. We conducted a systematic literature search using MEDLINE and Embase to identify relevant RCTs and performed a meta-analysis to compare the following outcomes between APM and placebo: response and remission rates, study discontinuation due to all causes, lack of efficacy, and adverse events, changes in total scores on depression severity scales, and individual adverse event rates. A total of 13 studies were identified, with 14 comparisons involving 3,197 participants that met the eligibility criteria. There were significant differences between APM and placebo in response and remission rates and changes in the primary depression severity scale in favor of APM, and study discontinuation due to adverse events and several individual adverse events in favor of placebo. No significant difference was observed in discontinuation due to all causes. APM could have antidepressant effects in the acute phase of MDD, although clinicians should be aware of an increased risk of some adverse events.


Introduction
Current clinical guidelines for the treatment of the major depressive disorder (MDD) recommend antidepressants as first-choice pharmacological treatment, particularly selective serotonin reuptake inhibitors, serotonin noradrenaline reuptake inhibitors, and mirtazapine [1]. However, a substantial number of patients do not respond to first-line antidepressants. For instance, in the STAR * D trial, only 36.8 % of patients with MDD achieved remission (defined as a total score of ≤ 5 on the Quick Inventory of Depressive Symptomatology -Self-Report) after the first treatment with citalopram [2], and a meta-analysis showed that 37 % of patients with MDD did not respond (defined as > 50 % improvement in symptoms from baseline on depression severity scales) to first-line treatment with various second-generation antidepressants [3]. When first-line treatment is ineffective, several clinical guidelines recommend augmentation with second-generation antipsychotics as the secondor third-line treatment. However, no clinical guidelines recommend antipsychotic monotherapy (APM) for MDD.
The antidepressant effects of antipsychotics have been reported since the 1960s. For example, thioridazine was reported to be effective for depressive mood in patients with schizophrenia [4,5]. First-generation antipsychotics were used as antidepressants main-ly for mixed anxiety-depressive states [6]. Benzamides, especially sulpiri de, has actually been a treatment option for depression [7,8].

Literature search and study selection
We conducted a systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [26]. The protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO) (CRD42020155338) (https://www.crd.york. ac.uk/prospero/). We performed a systematic literature search (last search: October 29, 2020) on MEDLINE and Embase databases using the following keywords: "depressi * AND (antipsychotic * OR all names of existing antipsychotics [Supplementary Table 1])" and with a limitation of "randomized controlled trial." Two authors (A.N. and K.S.) independently selected studies meeting the following eligibility criteria: (a) an original study (i. e., not a protocol, review, meta-analysis, or secondary analysis); (b) an RCT; (c) a study whose participants were diagnosed with MDD (i. e., not a bipolar disorder or schizoaffective disorder) using standard diagnostic criteria; and (d) a study including treatment arms comprising both APM and placebo arms. In studies involving both patients with MDD and dysthymia, we included only those in which more than 50 % of participants had MDD.
Any disagreements about study selection were resolved by consensus with the lead researcher (H.T.). We used the Cochrane risk of bias tool (available at: https://methods.cochrane.org/bias/resources/rob-2-revised-cochrane-risk-bias-tool-randomized-trials) to assess the risk of bias in each included study.

Data extraction
We extracted the following clinical outcome data from the selected studies: (1) demographic and clinical characteristics of the patients (e. g., age, sex, and duration of illness); (2) information on interventions (e. g., antipsychotic type, dose, and formulation); (3) number of patients who achieved response and remission (response and remission criteria were defined in each study as shown in ▶Table 1); (4) number of patients who discontinued the study due to all causes, lack of efficacy, and adverse events; (5) mean ± standard deviation (SD) of changes from baseline to endpoint in total scores on standard depression severity scales (i. e., the Hamilton Depression Rating Scale [HDRS] [27] and Montgomery-Åsberg Depression Rating Scale [MADRS] [28]) and the Clinical Global Impression -Severity scale (CGI-S); (6) mean ± SD of changes from baseline to endpoint in total scores on self-rating scales that were used in more than one study (i. e., Sheehan Disability Scale (SDS) [29] and Quality of Life Enjoyment and Satisfaction Questionnaire -Short Form [Q-LES-Q-SF] [30]); and (7) number of patients who had a specific adverse event. We calculated the SD from the standard error (SE) if needed. If more than one fixed-dose arm of APM was included, we extracted the data for each dose arm.
Any disagreements about data extraction were resolved by discussion with the lead researcher (H.T.). If the report on the study did not provide sufficient data, we contacted the corresponding author in an attempt to obtain additional information; however, we were not able to obtain any additional data.

Data analysis
We performed a meta-analysis using Review Manager (RevMan) version 5.4. We combined and compared the outcome data between APM and placebo, followed by a subgroup analysis for each antipsychotic. When more than one fixed-dose arm of APM was included in a study, the data in the highest-and second-highest-dose arms were included as the main analysis and sensitivity analysis, respectively, because the highest-dose arm was expected to show the highest therapeutic effect. The primary outcome was the response rate. For depression severity scales such as the HDRS and MADRS, we used the scale defined as the primary outcome in each study for the main analysis and combined the data on each scale (i. e., the HDRS or MADRS) for a sensitivity analysis. For individual adverse events, we used the data on each adverse event noted in two or more studies and listed under the exact same term. We did not evaluate individual adverse events by grouping similar categories (e. g., sedation and fatigue) to avoid double-counting overlapped adverse events. For dichotomous and continuous outcomes, pooled estimates of risk ratios (RRs) and standardized mean differences or mean differences, respectively, were calculated with two-sided 95 % confidence intervals (CIs) using a random-effects model. We calculated the number needed to treat (NNT) for the response and remission rates and the number needed to harm (NNH) for individual adverse events based on pooled estimates of absolute risk reduction. Study heterogeneities were quantified using the I 2 statistic, with an I 2 ≥ 50 % indicating significant heterogeneity. All effect sizes with P < 0.05 were considered significant.
Publication bias was assessed using visual inspection of funnel plots for each outcome. Plots without obvious asymmetry indicated a low possibility of significant publication bias.
Finally, we assessed the overall quality of the evidence regarding the effects of APM versus placebo on each clinical outcome according to the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) Handbook (available at http:// gdt.guidelinedevelopment.org/app/handbook/handbook.html).

Included studies
A total of 13 studies with 14 comparisons involving 3,197 participants (n = 1,818 for the APM group; n = 1,379 for the placebo group) that met the eligibility criteria were identified (Supplementary Figure 1). The characteristics of these studies are summarized in ▶ Table 1 and Supplementary Table 2. Only two studies were published before 2000 [9,20]. Because one study included two separate comparisons [12], a total of 14 comparisons were included in the meta-analysis. All studies were conducted in a doubleblind fashion and examined an oral antipsychotic formulation. Eleven and two studies investigated second-and first-generation antipsychotics, respectively. Almost all studies lasted for 12 weeks or shorter: 6 weeks (three studies), 8 weeks (four studies), 10 weeks (two studies), 11 weeks (one study), and 12 weeks (two studies). One study had a short intervention period of 1 week [10] and one study had a long intervention period of 6 months [9]. In all but two studies, the participants were in their 40s and 50s on average. All studies included patients who had the acute phase of MDD (i. e., not relapse prevention studies) and three studies included patients who had MDD with specific conditions (i. e., psychotic features [12], treatment resistance [13], and mixed state [11]), one study included patients with comorbidity (i. e., fibromyalgia syndrome [18]), and one study included patients with MDD or dysthymia [9]. Mean depression severity scores ranged from 19.9 to 26.6 on the HDRS and from 24.2 to 33.3 on the MADRS, which correspond to moderate-to-severe depression [31][32][33]. All except for three studies [9,13,20], including the two earliest published, defined response as a ≥ 50 % reduction in total scores on the primary depression severity scale. Ten studies defined remission as lower than a certain score on the primary depression scale. Only two studies examining quetiapine included more than one fixed-dose arm of APM [14,15].

Risk of bias
The results of risk of the bias assessment are shown in Supplementary Figure 2. The risks of blinding the outcome assessment and allocation concealment were not clear in almost all studies. The risks of selective reporting were high in about half of the studies because several studies reported only statistical significance without detailed data.

Meta-analysis
Response rate A significant difference was found in response rates between the APM and placebo groups in favor of the APM group (11 comparisons, n = 2,448, RR = 1.50, 95 % CI = 1.29 to 1.74, P < 0.001, I 2 = 61 %; ▶ Fig. 1). The NNT was 5. Evidence quality was moderate.

Remission rate
A significant difference was shown in remission rates between the APM and placebo groups in favor of the APM group (10 comparisons, n = 2,312, RR = 1.57, 95 % CI = 1.26 to 1.95, P < 0.001, I 2 = 55 %; ▶ Fig. 1). The NNT was 8. There was no obvious publication bias. Evidence quality was moderate.

Study discontinuation
No significant difference was found in the study discontinuation due to all causes between the APM and placebo groups (14 comparisons, n = 2,684, RR = 1.04, 95 % CI = 0.89 to 1.22, P = 0.62, I 2 = 40 %; ▶Fig. 2). On the other hand, there were significant differences in the study discontinuation due to lack of efficacy in favor of the APM group (12 comparisons, n = 2,609, RR = 0.47, 95 % CI = 0.31 to 0.70, P < 0.001, I 2 = 20 %; ▶Fig. 2) and the study discontinuation due to adverse events in favor of the placebo group (13 comparisons, n = 2,662, RR = 2.47, 95 % CI = 1.86 to 3.29, P < 0.001, I 2 = 0 %; ▶ Fig. 2). There were no obvious publication biases. Evidence quality was moderate for the study discontinuation due to all causes and high for the study discontinuation due to lack of efficacy and adverse events.

Continued
This document was downloaded for personal use only. Unauthorized distribution is strictly prohibited.
were found (Supplementary Table 3). There were no obvious publication biases. The overall quality of evidence was moderate to high.
Sensitivity analyses of studies using quetiapine (150 mg/day) Two studies examining quetiapine included fixed-dose arms of 300 mg/day and 150 mg/day; therefore, we performed sensitivity analyses of studies using 150 mg/day quetiapine. The significant differences in response rate, remission rate, and study discontinuation were unchanged (Supplementary Table 4). In terms of individual adverse events, the significant differences disappeared for constipation and dizziness, and the risk became significantly higher in the APM group than in the placebo group for myalgia (data not shown).
We were not able to perform sensitivity analyses on depression severity changes because the studies did not show SD.

Discussion
This meta-analysis revealed significant differences between APM and placebo in response and remission rates and reduction in the primary depression severity scale in favor of APM. Regarding the study discontinuation due to all causes, there was no significant difference between APM and placebo for MDD in the meta-analysis, which may be attributed to the findings that APM was superior ▶Fig. This document was downloaded for personal use only. Unauthorized distribution is strictly prohibited.
and inferior to placebo for the study discontinuation due to lack of efficacy and adverse events, respectively. In the subgroup analysis of only studies examining quetiapine, significant effectiveness and efficacy of quetiapine were found, which further supports the findings in the previous meta-analysis and pooled analyses [22][23][24][25]. The present meta-analysis included additional RCTs and evaluated the risks of individual adverse events, making it superior to past meta-analysis/pooled analyses. APM showed significantly higher response and remission rates than placebo in this meta-analysis. The NNTs for response and remission rates were 5 and 8, consistent with a previous meta-analysis [34]. The present study found that APM significantly improved depressive symptoms by 3.3 points on the HDRS-17 and 7.5 points on the MADRS in patients with moderate-to-severe MDD compared with placebo. A recent network meta-analysis showed that 21 antidepressants were more effective for the acute treatment of MDD than placebo, with mirtazapine, duloxetine, and venlafaxine show-ing the highest efficacy [35]. We looked at the score changes in depression rating scales in each included study investigating these three antidepressants because only odds ratios were reported in the network meta-analysis. Interestingly, we found that the score improvements in the HDRS-17 or MADRS for mirtazapine, duloxetine, or venlafaxine were almost the same as those for APM in the present meta-analysis [36][37][38][39][40][41][42]. Moreover, two RCTs included in the present meta-analysis demonstrated no differences in efficacy between quetiapine and antidepressants (i. e., duloxetine and escitalopram) [15,19]. Another RCT also indicated the non-inferiority of amisulpride to paroxetine in terms of efficacy and safety for MDD [43]. These suggest that APM, especially quetiapine, may be a useful treatment for MDD, at least in terms of efficacy, compared with standard treatment with antidepressants.
Compared with placebo, APM significantly increased the risks of study discontinuation due to adverse events. This is consistent with a recent network meta-analysis showing significant differenc-▶Fig. 2.20 [1.46, 3.30] es in study discontinuation due to adverse events between individual antidepressants and placebo in favor of placebo (odds ratios ranged from 1.64 of vortioxetine to 4.44 of clomipramine) [35]. Also, APM was associated with significantly increased risks of individual adverse events such as EPS, somnolence, and anticholinergic and metabolic side effects. In the subgroup analyses, no clear differences were found in most adverse events such as EPS, somnolence, and metabolic side effects among individual antipsychotics, which may be attributed to reductions in the statistical powers in the subgroup analyses. On the other hand, anticholinergic side effects such as constipation and dry mouth were less frequent for amisulpride and sulpiride, which have little or no anticholinergic ▶Fig.  effect, than for other antipsychotics. One review reported that quetiapine extended-release significantly increases the risk of the study discontinuation due to adverse events versus placebo in patients with bipolar depression or MDD, but not in those with schizophrenia [44]. Other reviews also indicated that quetiapine is associated with significantly higher risks of EPS and somnolence than placebo in patients with bipolar disorder, particularly in those with bipolar depression, but not in those with schizophrenia [45][46][47]. These findings are consistent with the results of the current meta-analysis. In terms of metabolic side effects, there was a significant difference in only increased appetite between APM and placebo in favor of placebo, but no significant difference was observed in weight gain. The results may be attributed to the small number of included RCTs conducted for a short duration (i. e., ≤ 12 weeks) because weight gain follows an appetite increase. Nonetheless, clinicians need to closely monitor metabolic side effects in patients treated with APM, given that a recent review determined that APM causes significant weight gain in patients with bipolar disorder compared with placebo, regardless of treatment duration [47]. It should be noted that the present meta-analysis did not evaluate serious adverse events associated with antipsychotics, such as tardive dyskinesia and neuroleptic malignant syndrome, because ▶Fig. 2 Study discontinuation; (b) Study discontinuation due to lack of efficacy Test for subgroup differences: Chi 2 = 4.82, df = 5 (P = 0.44); I 2 = 0 % these relatively rare adverse events were not reported in the included studies. Considering that mood disorders are a risk factor for tardive dyskinesia [48,49], clinicians should also continue to pay attention to these potentially serious adverse events.
The present study has several limitations. First, although this meta-analysis included several antipsychotics, almost half of the studies examined quetiapine and only one study examined each of the other antipsychotics.Thus, the results may have been largely influenced by studies examining quetiapine. Second, the treatment duration in almost all studies was 12 weeks or shorter; therefore, we could not evaluate the long-term effectiveness, efficacy, and safety of APM for MDD. This is especially relevant for depression in light of the chronic and recurrent nature of the illness. Third, because all patients had moderate-to-severe depression, the findings of this meta-analysis cannot be applied to mild-to-moderate depression or the most severe or psychotic depression. Fourth, a small number of studies contributed to the meta-analysis of depressive symptom severity assessed using rating scales (i. e., the HDRS and MADRS), which may have resulted in insufficient statistical power, particularly for the sensitivity analyses. Fifth, some of the included studies included patients who had MDD with specific conditions. This heterogeneity of studies may have influenced the results. However, the additional meta-analysis of studies that included MDD without any other conditions confirmed unchanged significant differences in response rates, remission rates, and changes in the primary depression severity scale in favor of APM, and study discontinuation due to adverse events and several individual adverse events in favor of placebo (data not shown). Sixth, adverse events have probably not been systematically assessed in the included studies. Furthermore, in addition to the risk of blinding of outcome assessment, blinding of participants may not have been assured for patients who experienced adverse events more specific to an-tipsychotics such as EPS, weight gain, and sedation. Seventh, there are risks to over-or under-estimation of the response and remission rates by dividing continuous variables (i. e., scores on rating scales) into the dichotomous variables [50], although in this metaanalysis, significant differences were also found in depression severity scales in favor of APM. Lastly, this meta-analysis included several studies that had a high reporting bias. However, the sensitivity analyses of the studies with a low reporting bias in the additional meta-analysis found similar results (data not shown).
In conclusion, this meta-analysis of 13 RCTs showed the antidepressant effects of APM in the acute phase of MDD. However, clinicians should be aware of the increased risks of some adverse events and carefully consider if APM should be used for patients with MDD. Given that no clear relationship between antipsychotic dose and effectiveness, efficacy, or safety was observed and that patients with mood disorders are sensitive to adverse events, clinicians should use the lowest effective dose of antipsychotics and carefully monitor the occurrence of adverse events. Further studies are ▶Table 2 Depression severity change. needed to examine the efficacy and safety of second-generation antipsychotics other than quetiapine for MDD and elucidate the mechanism of their antidepressant effects.

Author contribution
Dr. Takeuchi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data availability
The data that support the findings of this study are available from the corresponding author (H.T.) on reasonable request. The corresponding author had full access to all the data in the study.