Second-Generation Antipsychotic Agents in the Treatment of Acute Mania: A Systematic Review and Meta-analysis of Randomized Controlled Trials

Context  Recommendations of treatment guidelines concerning the use of second-generation antipsychotic (SGA) agents for acute mania vary substantially across committees or working groups. Meta-analyses addressing the use of SGAs in the treatment of acute mania are lacking.

Objective  To conduct a meta-analysis of the efficacy and safety of SGAs in the treatment of acute mania.

Data Sources  Randomized controlled trials comparing SGAs with placebo, first-generation antipsychotic drugs, or mood stabilizers (MSs) in the treatment of acute mania were searched for in the PsiTri and MEDLINE databases (last search: May 2006).

Study Selection  The abstracts, titles, and index terms of studies were searched using the following key words: aripiprazole,amisulpride,clozapine,olanzapine,quetiapine,risperidone,ziprasidone, and zotepine in conjunction with mania,manic, and bipolar.

Data Extraction  Data on efficacy, global dropout, dropout due to adverse events, dropout due to inefficacy, weight gain, rate of somnolence, and extrapyramidal symptoms were extracted and combined in a meta-analysis.

Data Synthesis  A total of 24 studies with 6187 patients were included. The SGAs were significantly more efficacious than placebo. The analysis demonstrated that adding antipsychotic agents to MS treatment was significantly more effective than treatment with MSs alone. The SGAs displayed efficacy comparable with that of MSs. Some SGAs seemed to induce more extrapyramidal symptoms than placebo. The SGAs were also associated with higher rates of somnolence than placebo.

Conclusion  Currently available data suggest that combining SGAs and MSs is the most efficacious treatment of acute mania.

Mood stabilizers (MSs) and first-generation antipsychotic agents have long been the mainstay of treatment of acute mania with and without psychotic features. However, there are reports of first-generation antipsychotics inducing or worsening depressive symptoms in patients with bipolar disorder.¹ Furthermore, patients with bipolar disorder are more susceptible to extrapyramidal symptoms (EPSs) than those with schizophrenia.^2,3 Therefore, first-generation antipsychotics are of limited applicability in the treatment of bipolar disorders.

In recent years, second-generation antipsychotic (SGA) agents have been developed and have proved to be effective in the treatment of bipolar mania. The SGAs do not seem to induce depressive episodes, and recent studies^4,5 revealed that some SGAs may have antidepressant effects.

Fountoulakis et al⁶ recently reviewed treatment guidelines for bipolar disorder. Their investigation revealed that guidelines for the treatment of bipolar disorder vary significantly across committees or specialist groups. In particular for the treatment of acute mania, some guidelines recommend monotherapy with an MS or an SGA drug as first-line treatment, whereas others recommend a combination of an MS and an antipsychotic agent. However, meta-analyses addressing the efficacy and effectiveness of SGAs in the treatment of acute mania are lacking.^7-9

Thus, the aim of this study is to compare the efficacy and safety of (1) SGAs vs placebo, (2) SGAs vs MSs, (3) combination therapy with SGAs plus MSs vs MSs alone, and (4) SGAs vs haloperidol.

All published and unpublished randomized controlled trials that assessed the efficacy of SGAs (aripiprazole, amisulpride, clozapine, olanzapine, quetiapine, risperidone, ziprasidone, and zotepine) in the treatment of mania were searched for in the PsiTri database (http://psitri.stakes.fi) (last search: May 2006). PsiTri is a register of controlled trials that compiles the registers of all Cochrane review groups in the field of mental health. The registers of the single Cochrane review groups are compiled by means of regular searches of numerous electronic databases and conference abstract books and hand searches of major journals (the exact search strategies of the individual review groups are listed in The Cochrane Library¹⁰). We also searched MEDLINE. The abstracts, titles, and index terms of studies were searched using the following key words: aripiprazole,amisulpride,clozapine,olanzapine,quetiapine,risperidone,ziprasidone, and zotepine in conjunction with mania,manic, and bipolar. In addition, the reference sections of included articles and key reviews were screened, and the first and last authors (Michael Berk, Charles Bowden, William Carson, Marielle Erdekens, Robert Hirschfeld, Paul Keck, Sumant Khanna, Roger McIntyre, Steven Potkin, Gary Sachs, Mauricio Tohen, Lakshmi Yatham, and John Zajecka) of the included studies and the pharmaceutical companies (AstraZeneca, Eli Lilly, Janssen-Cilag, Bristol-Myers Squibb, and Pfizer) were asked by e-mail between October 1, 2005, and March 31, 2006, whether they were aware of further trials. They were also contacted for the provision of missing data necessary for the meta-analysis. We thank Tohen et al, Yatham et al, McIntyre et al, Smulevich et al, and Bowden et al for sending us additional data. A rating based on the 3 quality categories described in The Cochrane Collaboration Handbook¹¹ was given for each trial: A indicates low risk of bias (adequate allocation concealment); B, moderate risk of bias (some doubt about the results, mainly studies said to be randomized but without an explanation of the method); and C, high risk of bias (clearly inadequate allocation concealment, eg, alternate randomization). Only trials belonging to categories A and B were included. Two of us (H.S. and S.L.) independently extracted data from the trials. Any disagreement was discussed, and the decisions were documented.

The primary outcome of interest was the mean change in the Young Mania Rating Scale (YMRS) score or similar scale scores from baseline to the end point. Further outcome parameters were the rate of response and effectiveness criteria, such as the number of participants leaving the study early (dropouts) for any reason, dropouts due to adverse events, dropouts due to inefficacy, mean weight gain, rate of somnolence, and EPSs. For response, the definition used by the authors of the original studies was adopted by the reviewers. This was generally a reduction of at least 50% on an efficacy scale such as the YMRS.¹²

In a once randomized–analyzed approach (last observation carried forward method) we assumed in the case of dichotomous data that participants who dropped out before completion had no change in their condition unless otherwise stated. Continuous data had to be reported as presented in the original studies without any assumptions about those lost to follow-up.

The outcome data were combined in a meta-analysis. For continuous data the standardized mean difference based on the Hedges adjusted g (a slightly modified version of the Cohen D for correction in the case of small participant numbers below 10)¹³ and its 95% confidence interval (CI) were calculated. When standard deviations were not indicated we either derived them from P values or used the mean standard deviations of the other studies. For dichotomous data, the relative risk (RR), which is defined as the ratio of the risk of an unfavorable outcome among treatment-allocated participants to the corresponding risk of an unfavorable outcome among those in the control group, was estimated again along with its 95% CI. Whereas many meta-analysts preferred to use odds ratios some years ago, it has been shown that the RR is more intuitive¹⁴ and that odds ratios tend to be interpreted as RRs by physicians.¹⁵ This misinterpretation then leads to an overestimated impression of the effect. The random-effects model of DerSimonian and Laird¹⁶ was used in all cases. Random-effects models are, in general, more conservative than fixed-effects models because they take heterogeneity among studies into account, even if this heterogeneity is not statistically significant. Study heterogeneity was sought for by visual inspection of the forest plots and by using a χ² test, which contrasts the RRs of the individual trials with the pooled RR. Significance levels of P<.1 were set a priori to assume the presence of heterogeneity. Results of the pooled analyses, which were statistically significantly heterogeneous, were noted in the results. In the case of significant differences between groups, the number of participants needed to treat (NNT) and the number of participants needed to harm (NNH) were calculated. For this purpose we calculated risk differences (RDs) in addition to RRs. Then, NNT/NNH was derived from the RD by the formula NNT/NNH = 1/RD, with the 95% CIs of NNT/NNH being the inverse of the upper and lower limits of the 95% CI of the RD. Studies with negative results are less likely to be published than studies with significant results. The possibility of such publication bias was examined using the funnel plot method described by Egger and colleagues.¹¹ Owing to the small number of studies, we also tentatively analyzed the antipsychotics as a single group compared with placebo or MSs in the secondary analyses. All the calculations were performed using MetaView, meta-analytic standard software used by The Cochrane Collaboration (Review Manager Version 4.2.8, The Cochrane Collaboration, Oxford, England). The exact formulas were reported there. A P<.05 was considered significant. We conducted 4 comparisons: (1) SGAs vs placebo, (2) SGAs vs MSs, (3) SGAs vs placebo as add-on medication to MSs, and (4) SGAs vs haloperidol. In addition, in each comparison SGAs were entered in an exploratory pooled analysis. The latter results are detailed only in cases in which they were not heterogeneous.

A total of 24 studies dealing with all the SGAs except zotepine and amisulpride were included (eTable 1 and eTable 2). These studies could be classified according to 4 different comparisons (Table 1): (1) SGAs vs placebo,^17-28 (2) SGAs vs MSs,^22,29-32 (3) SGAs vs placebo as add-on to MSs,^33-38 and (4) SGAs vs haloperidol.^{23,26,32,39,40} Four studies^22,23,26,32 conducted 3-branch examinations and could be used in 2 comparisons each. Assessment of manic symptoms was performed using the YMRS (18 trials), the Mania Rating Scale (3 trials), and the Mania Scale (1 trial).

The baseline mania scores were similar in all the trials except 2 studies with more²⁵ or less³³ severely manic patients. The duration of most studies was 3 weeks; however, 3 studies investigated a 4-week period^21,31,32 and 2 a 6-week period.^33,40 Four trials^23,26,30,37 investigated a 12-week period but also evaluated treatment outcomes after 3 weeks. The 3-week data were used for the analysis.

Four trials^22-24,35 investigated purely manic patients, 4 studies^26,31,32,34 did not report the types of manic episodes, and all the other trials examined patients with purely manic symptoms (45%-97%) and patients with mixed symptoms (3%-55%). Each of these trials was matched for episode type. Seven studies^{22,23,25,26,34,35,39} excluded patients with rapid cycling, 12 studies^{17,18,24,27,28,31-33,36-38,40} did not report data on this aspect, and 5 trials^19-21,29,30 included 16% to 61% of patients with a rapid cycling course.

Given the small number of studies, the use of funnel plots (a method based on symmetry) was appropriate only for SGAs vs placebo. The plots on the primary efficacy outcomes did not suggest publication bias. The plot on dropouts regardless of reason was the only asymmetrical one, but it remains unclear whether a study was unpublished in case an SGA failed to prove superiority in terms of dropout rate.

Twelve trials compared the effects of aripiprazole,^17-19 olanzapine,^20,21 quetiapine,^22,23 risperidone,^24-26 and ziprasidone^27,28 vs placebo in the treatment of acute mania (Table 1). Figure 1 displays the results of the primary outcome (YMRS score changes), and Table 2 gives the pooled results of the secondary outcome parameters.

Each individual SGA agent was significantly superior to placebo in treating acute manic symptoms (Figure 1). Response rates were significantly higher in the aripiprazole, olanzapine, risperidone, and ziprasidone trials but not in the quetiapine trials.

The analysis revealed a significantly lower global dropout rate in patients treated with olanzapine and risperidone but not with aripiprazole, quetiapine, and ziprasidone. Dropout due to adverse events did not differ between treatments.

Except for aripiprazole, the dropout rate due to inefficacy was lower for SGAs and for the pooled data compared with placebo.

Weight gain was significantly greater in patients treated with olanzapine and quetiapine but not with the other SGAs.

All the SGAs exhibited significantly higher rates of somnolence (Figure 2).

The incidence of EPSs was significantly higher in the aripiprazole (NNH, 13; 95% CI, 9-20) and risperidone trials and in the pooled analysis of all SGAs (Figure 3). In addition, increased EPS rates were found for ziprasidone. Although this difference was not significant (P = .06), the RD was (NNH, 11; 95% CI, 7-33). The results were heterogeneous in the risperidone trials and in the pooled analysis (χ² = 4.98; P = .03).

There were no overall differences in the symptom severity of EPS measures using the Simpson Angus Scale or the Extrapyramidal Symptom Rating Scale in the aripiprazole, olanzapine, risperidone, and ziprasidone trials. Akathisia, however, assessed using the Barnes Akathisia Scale, proved to be significantly more pronounced in patients treated with aripiprazole and ziprasidone.

Five studies investigated olanzapine, quetiapine, and risperidone vs the MSs valproate sodium^29,30 or lithium^22,31,32 (Table 1). Figure 4 displays the results of the primary outcome (YMRS score changes), and Table 3 gives the pooled results of the secondary outcome parameters.

Olanzapine compared with valproate showed greater symptom improvement (Figure 4). In no other trials were differences between the comparative treatments found. All the trials together indicated a trend for superiority of SGAs compared with MSs. Response rates were reported in 2 trials only.^22,24 In the olanzapine vs valproate comparison, patients treated with olanzapine showed a higher response rate. In the quetiapine vs lithium comparison, no difference was observed. As to the global dropout rate and the dropout rates due to adverse events or inefficacy, no differences between SGAs and MSs could be discerned.

Patients treated with olanzapine and quetiapine had greater weight gain and a greater rate of somnolence than those treated with lithium or valproate (data for risperidone were not available). In these studies, the rates of EPS were not reported.

The 6 studies included in this analysis investigated olanzapine,³³ quetiapine,^34,35 risperidone,^36,37 and ziprasidone³⁸ vs placebo as add-on medication to the MSs lithium,^33-38 valproate,^33-37 and carbamazepine³⁷ (Table 1). Three of these studies^33-35 investigated patients who did not fully respond to MS monotherapy after 7, 14, or 28 days. Two more studies^36,37 included 43% and 64% of patients, respectively, with partial response to monotherapy with MSs. One trial³⁸ did not report previous treatment. Figure 5 displays the results of the primary outcome (YMRS score changes), and Table 4 gives the pooled results of the secondary outcome parameters.

Compared with placebo as add-on medication to MSs, statistically significant superiority in improving manic symptoms was found for olanzapine, quetiapine, and risperidone but not for ziprasidone (Figure 5). Considered as a group, the SGAs were significantly superior.

The percentage of patients with a response was much higher in groups of patients who received add-on treatment with olanzapine and quetiapine but not with risperidone (data for ziprasidone were not available). Analysis of all the trials showed a significant advantage for combination therapy.

The global dropout rate was significantly lower in patients treated with MSs plus quetiapine or risperidone than in those treated with MSs plus placebo. No difference was found for olanzapine and ziprasidone. Analysis of all the trials showed a significantly reduced global dropout rate in patients treated with combination therapy.

In studies with quetiapine, risperidone, and ziprasidone, adverse event dropout rates were not different; they were, however, higher for olanzapine than for placebo add-on treatment. There was no overall difference between the active treatment and placebo groups.

Regarding the dropout rate due to inefficacy, a significant advantage for combination therapy was shown in the olanzapine study but not for quetiapine and risperidone (data for ziprasidone were not available). The combined dropout rate due to inefficacy was significantly lower in patients treated with combination therapy.

Mean weight change was increased in patients treated with olanzapine, risperidone, and quetiapine (data for ziprasidone were not available). The rate of somnolence was significantly higher in patients treated with olanzapine, quetiapine, and ziprasidone but not with risperidone. The pooled analysis revealed a significantly higher rate of somnolence in patients treated with MSs plus SGAs.

Data on EPS rates were reported only in the risperidone and ziprasidone trials. The incidence of EPSs was higher with ziprasidone than with placebo but not with risperidone vs placebo.

We included 2 studies investigating aripiprazole³⁹ and olanzapine⁴⁰ vs haloperidol and the branches of 3 further studies analyzing quetiapine²³ and risperidone^26,32 vs haloperidol (Table 1). Figure 6 displays the results of the primary outcome (YMRS score changes), and Table 5 gives the pooled results of the secondary outcome parameters.

Reduction in manic symptoms was similar for aripiprazole and risperidone compared with haloperidol. However, olanzapine and quetiapine showed a significantly lower improvement in YMRS scores (Figure 6). Overall there were no significant differences in mean YMRS score changes between patients treated with an SGA or haloperidol. However, the overall analysis was significantly heterogeneous (χ² = 13.0; P = .01) owing to the different results between the individual SGAs. The response rates did not differ between SGAs and haloperidol.

The analysis revealed a significantly lower global dropout rate in patients treated with aripiprazole and a trend toward a higher rate in patients treated with quetiapine. For olanzapine and risperidone, no difference was observed.

The dropout rate due to adverse events was significantly lower for aripiprazole. No differences were found in the other trials.

No differences in the dropout rate due to inefficacy were revealed for olanzapine, quetiapine, or risperidone compared with haloperidol. The dropout rate due to inefficacy was higher with aripiprazole.

Only 3 studies^26,39,40 reported data on weight change. In olanzapine-treated patients, the mean weight change was significantly greater than in haloperidol-treated patients but not for aripiprazole or risperidone (data on quetiapine were not available).

Regarding olanzapine, the rate of somnolence was significantly higher compared with that of haloperidol. It did not differ in the quetiapine and risperidone trials (data for aripiprazole were not available). The pooled analysis revealed that the rate of somnolence was significantly higher in patients treated with SGAs.

Four trials^27-30 reported the number of patients with at least 1 EPS. The analysis revealed a significantly higher incidence of EPSs in patients treated with haloperidol compared with all SGAs, taken either singly or as a group. Depressive symptoms improved more with aripiprazole treatment compared with haloperidol but not with olanzapine, quetiapine, or risperidone. In the pooled analysis, however, depressive symptoms improved more with SGAs.

To our knowledge, this is the first broad meta-analysis of efficacy, effectiveness, and adverse effects of SGAs in the treatment of acute mania. Its results deserve careful reflection. To draw firm conclusions concerning the overall benefits of SGAs, it is not enough simply to consider efficacy data, such as a reduction in symptoms in mania rating scales. Effectiveness criteria, which include dropout rates for any reason and due to adverse events, probably reflect the most valuable outcome parameters for clinical practice.

The SGAs are significantly more efficacious than placebo in the treatment of acute mania, as indicated by greater reductions in mania rating scores. Except for quetiapine, the superiority of SGAs is emphasized by higher response rates and, except for aripiprazole, lower dropout rates due to inefficacy.

The comparison of SGAs as a group with MSs as a group showed a certain trend toward the superiority of SGAs. This result was mainly due to the significant superiority of olanzapine in reducing manic symptoms. No differences were found for any other drugs or in any secondary outcome criteria.

Adding SGAs to MSs clearly increased the efficacy compared with monotherapy with MSs alone. Results of each single SGA drug, however, have to be discussed in detail. Olanzapine, for example, showed higher response rates and lower rates of dropout due to inefficacy but higher rates of dropout due to adverse events. Thus, the potential advantages of olanzapine in higher efficacy are counteracted by a higher rate of adverse effects, which limits clinical effectiveness. For ziprasidone, in contrast, there is no proof of higher efficacy as an add-on treatment to MSs. These results are disputable because in 5 trials^33-37 patients with partial responses to monotherapy were included. These studies addressed more the question of whether an add-on treatment of an SGA to an MS in patients who were partial responders or nonresponders is more helpful than continuing them on their first medication rather than the efficacy of combination treatment. Therefore, we are reluctant to generalize the results of this comparison. However, these results are, in a way, remarkable because they are not in accordance with many clinical guidelines.⁶ Some guidelines recommend as first-choice treatment monotherapy with an MS^41-47 or an SGA,^45-48 whereas others recommend combination treatment with MSs and SGAs,^49,50 especially in the case of severe manic episodes. Before any definitive recommendation of a combination therapy, pharmacoeconomic cost-benefit analyses are required. The selected studies do not provide any information on this question. In a recently published review⁵¹ the researchers were unable to draw any firm conclusions because of the limited availability of meaningful data.

The SGAs showed no superiority in improving manic symptoms compared with haloperidol. Results for the individual SGAs were diverse. Olanzapine and quetiapine reduced manic symptoms less effectively than haloperidol. In addition, quetiapine showed a lower rate of response and a higher rate of global dropout. Aripiprazole was less efficacious in terms of a higher rate of dropouts due to inefficacy, but effectiveness criteria such as rates of global dropout and dropout due to adverse events were superior compared with haloperidol. These findings are surprising because in meta-analyses in schizophrenia olanzapine has been consistently shown to be more effective than haloperidol, and quetiapine proved to be as effective as haloperidol.^52-54 Haloperidol-treated patients, however, showed a higher rate of dropout due to adverse events and higher rates of EPSs, which limits its use. Depressive symptoms improve less with haloperidol than with SGAs. Depressive symptoms were reported only as mean reductions in depression rating scale scores, and no study reported the number of patients who switched to depression. Therefore, we could not clarify whether SGAs improve depressive symptoms more than haloperidol or whether haloperidol more frequently leads to a switch into a full episode of major depression.

Adverse effects might hamper the clinical effectiveness of an antipsychotic agent despite its efficacy. As far as the data of the included studies have been reported, we analyzed the 3 important adverse events of antipsychotic drug treatment: weight gain, somnolence, and EPSs. In many studies, however, data on the rate of adverse events were reported incompletely. The SGAs are not alike, and this “class” of drugs is heterogeneous within itself.⁵⁵ The results require a balanced evaluation.

Mean weight gain was significantly greater in olanzapine- and quetiapine-treated patients, as is known from trials in schizophrenia. In contrast to the treatment of schizophrenia, however, very few data are available for bipolar disorder concerning metabolic effects.^56,57

Rates of somnolence were increased with SGA treatment. It was not only higher in olanzapine- and quetiapine-treated patients but also in aripiprazole-, risperidone-, and ziprasidone-treated patients compared with placebo. The severity of sedation was not indicated in the included studies, which limits the interpretation of the present results. In the treatment of acute mania, somnolence can be a welcome effect that can calm agitated patients.

Findings concerning EPS rates are difficult to interpret, and EPS data were inconsistently reported, particularly in comparisons between SGAs and MSs alone or in combination. Only olanzapine and quetiapine had no evidence of increased EPS rates. In the placebo-controlled trials, a higher incidence of EPSs was observed for aripiprazole, risperidone, and ziprasidone; however, increases in EPS rating scales marginally failed to reach statistical significance. In the 2 trials in which risperidone was used in addition to an MS, no higher incidence of EPS was found compared with a treatment of MSs plus placebo. Aripiprazole-treated patients additionally showed significantly higher akathisia scores compared with placebo. No further data on aripiprazole were available compared with MSs. Treatment with ziprasidone also revealed increased scores on akathisia rating scales compared with placebo and a higher incidence of EPSs in combination with an MS compared with an MS plus placebo.

These results may open the discussion on whether some SGAs might be more prone to induce EPSs in patients with bipolar disorder. We conclude that at least some SGAs are more likely to generate EPSs compared with placebo. These results become more evident when looking at the incidence rates rather than the changes in rating scale scores. It seems to us that the incidence rate is of more clinical relevance.

In this regard, patients with bipolar disorder may differ from those with schizophrenia. For all SGAs investigated in schizophrenia trials, Leucht et al^52,58 did not find evidence of EPS rates higher than for placebo. A recent study⁶ reanalyzed data on EPSs in olanzapine trials in patients with schizophrenia and bipolar disorder. This study revealed a higher incidence of EPSs in haloperidol-treated patients with bipolar disorder compared with haloperidol-treated patients with schizophrenia. They did not find a difference between olanzapine-treated patients with bipolar disorder and schizophrenia.

The rate of completers varied among different studies and trial arms. Furthermore, data on person-days of exposure were usually lacking. Thus, an unbalanced exposure of treatments cannot be excluded. In all the trials, antipsychotic agents and MSs were prescribed in commonly used dose ranges. The effect of psychotic features on efficacy were reported in 14 studies. Treatment efficacy was not different in patients with vs without psychotic symptoms in all but 2 studies.^23,29 Only 2 placebo-controlled studies reported outcome of treatment in patients with rapid cycling course. The studies concluded that aripiprazole¹⁹ and olanzapine²⁰ are more efficacious than placebo. The effect of pure manic or mixed episode type on the outcome was also only marginally reported. Placebo-controlled trials showed that antipsychotic drugs are efficacious in manic and mixed episode types. One study reported that a combination of olanzapine with MSs was more efficacious than MSs alone in patients with mixed episodes.³³ Other studies found no difference in efficacy between episode types.^34,36 Only 1 study³³ reported the time to respond. In this study, median response time was significantly shorter in patients treated with a combination of olanzapine and MSs (18 days) compared with patients treated with MSs alone (28 days).

A limitation of this review is that most of the trials were sponsored by the pharmaceutical industry and were conducted to gain regulatory approval for the treatment of acute mania. We found only 2 studies lacking industry support.^31,32 Therefore, the possibility of a sponsor bias induced in favor of their product cannot be excluded.⁵⁹

Furthermore, the statistical power varied among the 4 categories of comparisons. The greatest number of studies (n = 12) and patients (n = 2827) was available for the first comparison (SGAs vs placebo). Five studies with 636 patients were included in the second comparison (SGAs vs MSs), 6 studies with 1395 patients in the third comparison (SGAs vs placebo as add-on medication to MSs), and 5 studies with 1329 patients in the last comparison (SGAs vs haloperidol). Owing to the low number of patients and trials in 3 comparisons, we also tentatively analyzed the SGAs as a single group vs the comparison treatments. The exploratory pooling procedure seems justified because efficacy results were rather homogenous, in particular in comparisons 2 and 4. Concerning all investigated adverse events, results were much more heterogeneous, indicating that SGAs differed substantially in tolerability.⁵⁵

In conclusion, this meta-analysis found that SGA agents as add-on medication to MSs are highly superior to MSs alone in improving acute manic symptoms, as indicated by greater reductions in mania scores, higher response rates, and fewer dropouts due to inefficacy. However, effectiveness criteria should also be included in treatment decisions. Adverse effects such as somnolence, weight gain, and EPS have an impact on treatment adherence. Based on the results reported herein, combination treatment with an SGA and an MS should be the treatment of choice, in particular for severe manic episodes.

Correspondence: Harald Scherk, MD, Department of Psychiatry and Psychotherapy, Georg-August University Goettingen, Von-Siebold-Str 5, 37075 Goettingen, Germany (hscherk@uni-goettingen.de).

Submitted for Publication: February 20, 2006; final revision received September 8, 2006; accepted September 20, 2006.

Financial Disclosure: Dr Scherk has received speech honoraria from AstraZeneca and Eli Lilly and has accepted travel or hospitality not related to a speaking engagement from Eli Lilly. Dr Pajonk has been a consultant for AstraZeneca, Eli Lilly, Janssen, Novartis, and Wyeth and has acted as an expert witness for AstraZeneca, Eli Lilly, and Janssen. Dr Pajonk receives research funding from AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Janssen, Novartis, Pfizer, Sanofi-Synthelabo, and Wyeth. In addition, he is a member of a speakers' bureau for AstraZeneca and Janssen and has accepted paid speaking engagements in industry-sponsored symposia from AstraZeneca, Eli Lilly, Janssen, and Pfizer and travel or hospitality not related to a speaking engagement from AstraZeneca, Eli Lilly, and Janssen. Dr Leucht has received speech or consultancy honoraria from Sanofi-Aventis, Bristol-Myers Squibb, Eli Lilly, Janssen, Johnson & Johnson, Lundbeck, and Pfizer and research support from Sanofi-Aventis and Eli Lilly.

Additional Information: The online-only eTable 1 and eTable 2 are available.