Atypical Antipsychotic Augmentation in Major Depressive Disorder: a Meta-Analysis of Placebo-Controlled Randomized Trials

Reviews and Overviews

Atypical Antipsychotic Augmentation in Major Depressive Disorder: A Meta-Analysis of Placebo-Controlled Randomized Trials

J. Craig Nelson, M.D. Objective: The authors sought to deter- trial, depression scale used, response and mine by meta-analysis the efficacy and tol- remission rates, and discontinuation rates George I. Papakostas, M.D. erability of adjunctive atypical antipsy- for any reason or for adverse events. chotic agents in major depressive disorder. Results: Sixteen trials with 3,480 patients Method: Searches were conducted of were pooled using a fixed-effects meta- MEDLINE/PubMed (1966 to January analysis. Adjunctive atypical antipsychot- 2009), the Cochrane database, abstracts ics were significantly more effective than of major psychiatric meetings since 2000, placebo (response: odds ratio=1.69, 95% and online trial registries. Manufacturers CI=1.46–1.95, z=7.00, N=16, p<0.00001; of atypical antipsychotic agents without remission: odds ratio=2.00, 95% CI=1.69– online registries were contacted. Trials se- 2.37, z=8.03, N=16, p<0.00001). Mean lected were acute-phase, parallel-group, odds ratios did not differ among the atyp- double-blind controlled trials with ran- ical agents and were not affected by trial dom assignment to adjunctive atypical duration or method of establishing treat- antipsychotic or placebo. Patients had ment resistance. Discontinuation rates for nonpsychotic unipolar major depressive adverse events were higher for atypical disorder that was resistant to prior antide- agents than for placebo (odds ratio=3.91, pressant treatment. Response, remission, 95% CI=2.68–5.72, z=7.05, N=15, and discontinuation rates were either re- p<0.00001). ported or obtained. Data were extracted by one author and checked by the sec- Conclusions: Atypical antipsychotics are ond. Data included study design, number effective augmentation agents in major of patients, patient characteristics, meth- depressive disorder but are associated ods of establishing treatment resistance, with an increased risk of discontinuation drug doses, duration of the adjunctive due to adverse events.

(Am J Psychiatry 2009; 166:980–991)

Depression is among the most common medical dis- ium was used to augment tricyclic antidepressants, which orders. In the United States approximately 1 in 6 individu- are not currently in widespread use. als will suffer from major depressive disorder during their The use of antipsychotic agents in depression has a long lifetime (1). Even though numerous treatments are avail- history (5), but recognition of the extrapyramidal side ef- able, only about one-third of patients receiving initial an- fects and the risk of tardive dyskinesia discouraged the use tidepressant treatment achieve remission (2). As a conse- of the conventional antipsychotics. However, the advent quence, several treatment strategies have evolved for of a newer generation of antipsychotic agents, which for difficult-to-treat depression. One approach has been to the most part appeared to differ from conventional agents augment antidepressants with an agent not approved for in their side effect and neuropharmacologic profile, rekin- use as monotherapy in major depressive disorder. dled interest in the use of this class of drugs in mood and Augmentation strategies have a long history. Use of anxiety disorders, including as adjunctive therapy. The stimulants and T3 to augment tricyclic antidepressants first report of the use of an atypical antipsychotic for aug- was described four decades ago (3). Subsequently, use of mentation in major depressive disorder appeared in 1999. several other agents for augmentation was described, but Ostroff and Nelson (6) described eight patients whose de- the evidence base comprised a limited number of con- pression had failed to respond to selective serotonin re- trolled trials with small samples. Until recently lithium uptake inhibitors (SSRIs). Rapid effects were noted when augmentation was the best studied augmentation strat- low doses of risperidone were added. This observation egy. Ten placebo-controlled trials were performed, but was soon followed by the first placebo-controlled study. nine of them included no more than 35 patients (4). In 10 Shelton et al. (7) noted that the combination of olanzapine controlled lithium studies, the total number of patients and fluoxetine was more effective than either drug alone was only 269 (4), and in the majority of these studies lith- in a sample of 30 patients with fluoxetine-resistant major

980 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS

FIGURE 1. Results of Search for Articles on Trials of Atypical Antipsychotic Augmentation in Major Depressive Disorder

Medline Search Cochrane Search Review of presentations at major meetings since 2000 and response to requests for information from sponsors

Reports (N=97) Reports (N=69)

Excluded (N=78) Excluded (N=38) Bipolar disorder (N=40) Bipolar disorder (N=8) Schizophrenia (N=25) Schizophrenia (N=16) Other disorder (N=8) Other disorder (N=4) Not treatment for major Not treatment for major depression (N=5) depression (N=10)

Unipolar depression (N=19) Unipolar depression (N=31)

Excluded (N=9) Excluded (N=21) Psychotic Psychotic depression (N=5) depression (N=3) Maintenance trials (N=2) Maintenance trial (N=1) Open-label trials (N=3) Unpublished and Secondary analysis (N=1) Acceleration trials (N=2) nonduplicate poster Open-label trials (N=2) Other (N=4) presentations (N=5) Acceleration trials (N=2) Duplicates (poster presentations) (N=5)

Reports of 11 trials (N=10) Reports of 11 trials (N=10)

Reports of 16 randomized controlled trials (N=15 reports)

depression. Since those initial reports, the number and Given differences in the neuropharmacology of atypical size of studies of augmentation with atypical antipsychot- antipsychotics (14–16), it is not clear that these agents ics has grown rapidly, and in 2008 aripiprazole became the would have similar efficacy in depression or that they first atypical agent and the first pharmacologic treatment would be effective in the same patients. Moreover, the dos- of any type to be approved by the U.S. Food and Drug Ad- ages used in the adjunctive treatment of depression are of- ministration (FDA) for use as an augmentation agent in ten lower than those used in the treatment of psychotic major depressive disorder. disorders, and the prominent pharmacologic effects may With the growth of the adjunctive use of atypical anti- differ between these dosage levels. All of the atypical antipsychotics, interest in their efficacy and tolerability has in- psychotics display 5-HT2 receptor antagonism, although creased. In 2007 Papakostas et al. (8) conducted a system- their relative potency varies. Aripiprazole and ziprasidone atic review and meta-analysis of 10 placebo-controlled also display 5-HT1A partial agonism (15). All of the atypical augmentation trials of atypical antipsychotics for major agents possess dopamine D2 receptor affinity; aripiprazole depressive disorder (N=1,500). The meta-analysis con- acts as a partial agonist at this receptor (17), and the others firmed the efficacy of this strategy. Since that meta-analy- have antagonist properties. The metabolite of quetiapine, sis was published, however, six new trials with 2,007 sub- N-desalkylquetiapine, has recently been shown to have jects have been published or presented at major scientific moderate affinity for the norepinephrine reuptake trans- meetings; these include three large trials of aripiprazole porter (18), but it is not clear that the metabolite achieves (9–11), as well as two large trials of extended-release appreciable plasma concentrations in human subjects. quetiapine (12, 13), that increase the number of subjects Ziprasidone is associated with notable potency at the treated with quetiapine from 109 to 1,028. Because the transporters for dopamine, norepinephrine, and serotonin growth of evidence for the use of atypicals has been so (19). The olanzapine-fluoxetine combination is reported to rapid, an updated review and meta-analysis of the efficacy produce a greater increase in extracellular levels of dopa- and tolerability of these agents is in order. mine and norepinephrine in rat prefrontal cortex than ei-

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 981 META-ANALYSIS OF ATYPICAL ANTIPSYCHOTIC AUGMENTATION IN DEPRESSION

TABLE 1. Placebo-Controlled Atypical Antipsychotic Augmentation Trials in Major Depressive Disorder Atypical Anti- Durationb Rating Authors (Reference) Year Antipsychotic depressant Na (Weeks) Prior Failed Trials Scalec Shelton et al. (7) 2001 Olanzapine Fluoxetine 20 8 Two historical trials and one prospective trial MADRS Shelton et al. (25) 2005 Olanzapine Fluoxetine 288 8 One or more historical trials and one prospec- MADRS tive trial Corya et al. (26) 2006 Olanzapine Fluoxetine 286 12 One or more historical trials and one prospec- MADRS tive trial Thase et al. (27)d 2006 Olanzapine Fluoxetine 206 8 One or more historical trials and one prospec- MADRS tive trial Thase et al. (27)d 2006 Olanzapine Fluoxetine 200 8 One or more historical trials and one prospec- MADRS tive trial Mahmoud et al. (28) 2007 Risperidone Various 268 6 4-week prospective trial HAM-D Reeves et al. (29) 2008 Risperidone Various 23 8 Historical trial of at least 3 weeks MADRS Keitner et al. (30) 2009 Risperidone Various 95 4 One prospective trial of at least 5 weeks or a MADRS clearly documented trial Khullar et al. (31) 2006 Quetiapine SSRI/SNRI 15 8 At least one historical trial of at least 6 weeks MADRS Mattingly et al. (32) 2006 Quetiapine SSRI/SNRI 37 8 One prior trial of at least 4 weeks and at least 6 MADRS weeks in the current trial McIntyre et al. (33) 2006 Quetiapine SSRI/SNRI 58 8 One or more trials of at least 6 weeks HAM-D Earley et al. (12) 2007 Quetiapine SSRI/SNRI 487 6 One historical trial of at least 6 weeks MADRS El-Khalili et al. (13) 2008 Quetiapine SSRI/SNRI 432 8 One historical trial of at least 6 weeks MADRS Berman et al. (9) 2007 Aripiprazole SSRI/SNRI 353 6 1–3 historical trials and one prospective trial MADRS Marcus et al. (10) 2008 Aripiprazole SSRI/SNRI 369 6 1–3 historical trials and one prospective trial MADRS Berman et al. (11) 2008 Aripiprazole SSRI/SNRI 343 6 1–3 historical trials and one prospective trial MADRS a Number of randomized patients with at least one posttreatment rating. b Duration of the acute-phase double-blind controlled trial. c HAM-D=Hamilton Depression Rating Scale; MADRS=Montgomery-Åsberg Depression Rating Scale. d This report included two separate trials of identical design. ther agent alone or other combinations of antipsychotics Trial Selection and SSRIs (20). Each of these effects might contribute to Trials were included if they were acute-phase, parallel-group, clinical efficacy in depression. Alternatively, these effects double-blind, placebo-controlled trials with random assignment to adjunctive treatment with an atypical antipsychotic or placebo. as well as varying levels of α adrenergic, antihistaminic, 1 Patients had to have nonpsychotic major depressive disorder that and antimuscarinic activity of these agents may produce was considered treatment resistant either by history or deter- differences in side effects. mined by a prospective trial. Response, remission, and discontinuation rates either were reported in published articles or poster In our meta-analysis, we hypothesized that adjunctive presentations or were obtained from the sponsor or investigator. treatment with atypical antipsychotics would result in higher rates of response and remission than placebo but Data Extraction that this effect would be offset to some extent by an in- Data were extracted by one of the authors and checked for ac- curacy by the other. Data extracted included study design, num- crease in discontinuation due to adverse events. We fur- ber of patients, patient characteristics, methods used to establish ther hypothesized that sensitivity analyses would reveal treatment resistance, drug dosages, duration of the prospective that adjunctive trials of longer duration would be associ- adjunctive trial, response and remission rates, and rates of discontinuation for any reason and for adverse events. Response was ated with larger drug-placebo differences and that studies defined as an improvement of ≥50% from baseline to endpoint on using a prospective trial to confirm treatment resistance the Hamilton Depression Rating Scale (21) or the Montgomery- would have lower response rates but greater drug-placebo Åsberg Depression Rating Scale (22) in the intent-to-treat sample differences. using the last-observation-carried-forward method. Remission was defined according to each individual trial in the intent-to- treat/last-observation-carried-forward sample. Method Statistical Analysis We searched MEDLINE/PubMed (1966 to January 2009) for Across studies, the numbers of patients who responded, remit- randomized placebo-controlled trials of atypical antipsychotic ted, and dropped out and the number randomly assigned to each agents approved by the FDA for any indication. Search terms were treatment group were statistically combined, using a fixed-effects “major depression” and “aripiprazole or olanzapine or paliperi- meta-analytic model. Effects were expressed as odds ratios with done or quetiapine or risperidone or ziprasidone.” The Cochrane their 95% confidence intervals (CIs), test of significance (Wald’s z), number (N) of contrasts, and p values. Effects were calculated Clinical Trials register was searched using identical terms. Poster for each drug-placebo contrast, for each group of trials using the presentations at major psychiatric meetings since 2000 were same drug, and as meta-analytic summaries for all drugs com- searched, and the manufacturers of atypical antipsychotic medi- bined. A funnel plot in which the standard error of the log odds cations were asked for all published and unpublished reports of ratio against the log of the odds ratio was used to evaluate poten- adjunctive trials in nonpsychotic major depressive disorder. Re- tial publication or retrieval bias. ports from different sources were then compared to identify non- Chi-square tests and the I2 statistic derived from the chi-square duplicative trials. values were used to test for heterogeneity among the studies. I2

982 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS

FIGURE 2. Meta-Analysis of Response Rates of Atypical Antipsychotic Agents in Treatment-Resistant Major Depressive Disordera

Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) Agent and Study n/N n/N (95% CI) (95% CI) Olanzapine studies Shelton et al. (7) 6/10 1/10 13.50 [1.20, 152.21] Shelton et al. (25) 40/146 41/142 0.93 [0.56, 1.55] Corya et al. (26) 100/230 19/56 1.50 [0.81, 2.76] Thase et al. I (27)b 37/102 31/104 1.34 [0.75, 2.40] Thase et al. II (27)b 43/98 30/102 1.88 [1.05, 3.36] Subtotal 586 414 1.39 [1.05, 1.84] Total events: 226 (treatment), 122 (control) Test for heterogeneity: χ²=6.82, df=4 (p=0.15), I²=41.4% Test for overall effect: z=2.30 (p=0.02)

Risperidone studies Mahmoud et al. (28) 49/137 33/131 1.65 [0.98, 2.80] Keitner et al. (30) 34/62 11/33 2.43 [1.01, 5.85] Reeves et al. (29) 6/12 4/11 1.75 [0.33, 9.30] Subtotal 211 175 1.83 [1.18, 2.82] Total events: 89 (treatment), 48 (control) Test for heterogeneity: χ²=0.54, df=2 (p=0.76), I²=0% Test for overall effect: z=2.71 (p=0.007)

Quetiapine studies Khullar et al. (31) 4/8 1/7 6.00 [0.48, 75.34] Mattingly et al. (32) 12/242/13 5.50 [1.00, 30.29] McIntyre et al. (33) 14/29 8/29 2.45 [0.82, 7.31] Earley et al. (12) 185/327 74/160 1.51 [1.04, 2.21] El-Khalili et al. (13) 159/289 66/143 1.43 [0.95, 2.13] Subtotal 677 352 1.60 [1.24, 2.08] Total events: 374 (treatment), 151 (control) Test for heterogeneity: χ²=4.04, df=4 (p=0.40), I²=1.0% Test for overall effect: z=3.54 (p=0.0004)

Aripiprazole studies Berman et al. (9) 61/181 41/172 1.62 [1.02, 2.59] Berman et al. (11) 81/174 45/169 2.40 [1.53, 3.77] Marcus et al. (10) 60/185 32/1842.28 [1.40, 3.72] Subtotal 540 525 2.07 [1.58, 2.72] Total events: 202 (treatment), 118 (control) Test for heterogeneity: χ²=1.60, df=2 (p=0.45), I²=0% Test for overall effect: z=5.28 (p<0.00001)

Total 2014 1466 1.69 [1.46, 1.95] Total events: 891 (treatment), 439 (control) Test for heterogeneity: χ²=17.58, df=15 (p=0.29), I²=14.7% Test for overall effect: z=7.00 (p<0.00001) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a Odds ratios for response on drug and placebo are grouped by atypical agent. b This report included two separate trials of identical design. approximates the proportion of total variation in the effect size tribution with the value for degrees of freedom 1 less than the estimates that is due to heterogeneity rather than sampling error number of subgroups (24). Review Manager, version 4.2 (Co- 2 (23). An alpha error <0.20 and an I of at least 50% were taken as chrane Collaboration, Oxford, England), was used for statistical indicators of heterogeneity. In the Results section, the chi-square calculations. test and I2 statistics for heterogeneity follow the 95% CI, z score, N, and p value for the odds ratio unless these values are provided in a figure. Results In sensitivity analyses, we compared subgroups on potential differences between the different agents, duration of the ad- Figure 1 illustrates the search flow. The MEDLINE and junctive trial, and definition of treatment resistance (historical Cochrane searches each found 10 nonduplicative reports or prospective). Differences between two or more subgroups of 11 randomized, double-blind controlled trials of acute- were investigated by subtracting the sum of the heterogeneity chi-square statistics of the subgroups from the overall chi- phase treatment with atypical antipsychotic augmenta- square statistic and comparing the result with a chi-square dis- tion in patients with nonpsychotic major depressive disor-

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 983 META-ANALYSIS OF ATYPICAL ANTIPSYCHOTIC AUGMENTATION IN DEPRESSION

FIGURE 3. Meta-Analysis of Remission Rates of Atypical Antipsychotic Agents in Treatment-Resistant Major Depressive Disordera

Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) Agent and Study n/N n/N (95% CI) (95% CI) Olanzapine studies Shelton et al. (7) 6/10 2/10 6.00 [0.81, 44.35] Shelton et al. (25) 25/146 18/142 1.42 [0.74, 2.74] Corya et al. (26) 69/230 10/56 1.97 [0.94, 4.13] Thase et al. I (27)b 24/102 18/104 1.47 [0.74, 2.91] Thase et al. II (27)b 30/98 16/102 2.37 [1.20, 4.70] Subtotal 586 414 1.83 [1.30, 2.56] Total events: 154 (treatment), 64 (control) Test for heterogeneity: χ²=2.90, df=4 (p=0.57), I²=0% Test for overall effect: z=3.49 (p=0.0005)

Risperidone studies Mahmoud et al. (28) 26/137 12/131 2.32 [1.12, 4.83] Keitner et al. (30) 32/62 8/33 3.33 [1.30, 8.53] Reeves et al. (29) 4/12 2/11 2.25 [0.32, 15.76] Subtotal 211 175 2.63 [1.51, 4.57] Total events: 62 (treatment), 22 (control) Test for heterogeneity: χ²=0.38, df=2 (p=0.83), I²=0% Test for overall effect: z=3.43 (p=0.0006)

Quetiapine studies Khullar et al. (31) 3/8 0/7 9.55 [0.40, 225.19] Mattingly et al. (32) 11/24 2/13 4.65 [0.84, 25.66] McIntyre et al. (33) 9/29 5/29 2.16 [0.62, 7.49] Earley et al. (12) 110/327 38/160 1.63 [1.06, 2.50] El-Khalili et al. (13) 112/289 35/143 1.95 [1.25, 3.06] Subtotal 677 352 1.89 [1.41, 2.54] Total events: 245 (treatment), 80 (control) Test for heterogeneity: χ²=2.61, df=4 (p=0.63), I²=1.0% Test for overall effect: z=4.25 (p<0.0001)

Aripiprazole studies Berman et al. (9) 47/181 27/172 1.88 [1.11, 3.19] Berman et al. (11) 64/174 32/169 2.49 [1.52, 4.08] Marcus et al. (10) 47/185 28/184 1.90 [1.13, 3.19] Subtotal 540 525 2.09 [1.55, 2.81] Total events: 158 (treatment), 87 (control) Test for heterogeneity: χ²=0.77, df=2 (p=0.68), I²=0% Test for overall effect: z=4.88 (p<0.00001)

Total 20141466 2.00 [1.69, 2.37] Total events: 619 (treatment), 253 (control) Test for heterogeneity: χ²=8.16, df=15 (p=0.92), I²=10% Test for overall effect: z=8.03 (p<0.00001) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a Odds ratios for remission on drug and placebo are grouped by atypical agent. b This report included two separate trials of identical design. der patients who had not responded to prior antidepres- The 15 reports of 16 trials included five trials with olan- sant treatment (either by history or in a prospective trial). zapine (7, 25–27), three with risperidone (28–30), five with The search of meeting abstracts and information provided quetiapine (12, 13, 31–33), and three with aripiprazole (9– by the manufacturers of atypical antipsychotics provided 11). No acute-phase, double-blind trials of adjunctive another five unpublished reports that were presented as ziprasidone or paliperidone in major depressive disorder posters. We excluded five trials of patients with psychotic were found. The 16 trials included a total of 3,480 patients, depression, one relapse prevention trial, one secondary of whom 2,014 were randomly assigned to adjunctive analysis, three open-label trials, two acceleration trials, treatment with an atypical antipsychotic and 1,466 to pla- and four other studies that did not investigate efficacy (fo- cebo. Other descriptive features of the trials are listed in cusing instead, for example, on effects on the hypotha- Table 1. In one report for olanzapine (27), discontinuation lamic-pituitary-adrenal axis). rates were presented as combined data from two individ-

984 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS

FIGURE 4. Meta-Analysis of Rates of Discontinuation for Any Reason for Atypical Antipsychotic Agents in Treatment-Resis- tant Major Depressive Disordera

Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) Agent and Study n/N n/N (95% CI) (95% CI) Olanzapine studies Shelton et al. (7) 1/10 3/10 0.26 [0.02, 3.06] Shelton et al. (25) 30/146 28/142 1.05 [0.59, 1.87] Corya et al. (26) 60/243 12/60 1.31 [0.65, 2.63] Thase et al. (27) 52/200 41/206 1.41 [0.89, 2.25] Subtotal 599 418 1.23 [0.90, 1.69] Total events: 143 (treatment), 84 (control) Test for heterogeneity: χ²=2.18, df=3 (p=0.54), I²=0% Test for overall effect: z=1.29 (p=0.20)

Risperidone studies Mahmoud et al. (28) 26/141 16/133 1.65 [0.84, 3.24] Keitner et al. (30) 8/62 7/33 0.55 [0.18, 1.68] Reeves et al. (29) 1/12 4/11 0.16 [0.01, 1.73] Subtotal 215 177 1.08 [0.63, 1.86] Total events: 35 (treatment), 27 (control) Test for heterogeneity: χ²=5.41, df=2 (p=0.07), I²=63.0% Test for overall effect: z=0.27 (p=0.79)

Quetiapine studies Khullar et al. (31) 1/8 1/7 0.86 [0.04, 16.85] Mattingly et al. (32) 3/242/13 0.79 [0.11, 5.43] McIntyre et al. (33) 11/29 13/29 0.75 [0.26, 2.14] Earley et al. (12) 51/330 16/161 1.66 [0.91, 3.01] El-Khalili et al. (13) 78/297 23/148 1.94 [1.16, 3.24] Subtotal 688 358 1.59 [1.12, 2.25] Total events: 144 (treatment), 55 (control) Test for heterogeneity: χ²=3.22, df=4 (p=0.52), I²=1.0% Test for overall effect: z=2.57 (p=0.01)

Aripiprazole studies Berman et al. (9) 22/182 16/176 1.38 [0.70, 2.71] Berman et al. (11) 30/177 22/172 1.39 [0.77, 2.52] Marcus et al. (10) 29/191 28/190 1.04 [0.59, 1.82] Subtotal 550 538 1.24 [0.87, 1.76] Total events: 81 (treatment), 66 (control) Test for heterogeneity: χ²=0.63, df=2 (p=0.73), I²=0% Test for overall effect: z=1.20 (p=0.23)

Total 2052 1491 1.30 [1.09, 1.57] Total events: 403 (treatment), 232 (control) Test for heterogeneity: χ²=13.04, df=14 (p=0.52), I²=0% Test for overall effect: z=2.85 (p=0.004) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a Odds ratios for rates of discontinuation for any reason on drug and placebo are grouped by atypical agent. ual trials, so the two trials were presented as a single trial overall pooled response rate for treatment with an atypical in the meta-analysis of discontinuation rates. agent was 44.2%, compared with 29.9% for placebo. Odds ratios for the trials grouped by atypical agent var- Response and Remission ied from 1.39 (95% CI=1.05–1.84) to 2.07 (95% CI=1.58– The meta-analyses for response and remission are sum- 2.72), with considerable overlap among the confidence in- marized in Figures 2 and 3. The odds ratio for response tervals. Testing the difference between the atypical sub- with drug versus placebo was 1.69 (95% CI=1.46–1.95, z= groups indicated no significant differences (χ2=4.58, df=3, 7.00, N=16, p<0.00001). The risk difference by meta-analy- p=0.21). sis was 0.12 (95% CI=0.08–0.15, z=7.14, N=16, p<0.00001). The odds ratio for remission was 2.00 (95% CI=1.69– The risk difference translates into a number needed to 2.37, z=8.03, N=16, p<0.00001). The risk difference for re- treat of nine. The test for heterogeneity indicated a lack of mission was 0.12 (95% CI=0.09–0.15), indicating a number heterogeneity (χ2=17.58, df=15, p=0.29, I2=14.7%). The needed to treat of nine. The test for heterogeneity was not

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 985 META-ANALYSIS OF ATYPICAL ANTIPSYCHOTIC AUGMENTATION IN DEPRESSION

FIGURE 5. Meta-Analysis of Rates of Discontinuation Due to Adverse Events for Atypical Antipsychotic Agents in Treatment- Resistant Major Depressive Disordera

Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) Agent and Study n/N n/N (95% CI) (95% CI) Olanzapine studies Shelton et al. (7) 0/10 0/10 Not estimable Shelton et al. (25) 10/146 4/142 2.54 [0.78, 8.28] Corya et al. (26) 29/243 3/60 2.57 [0.76, 8.76] Thase et al. (27) 27/200 5/206 6.27 [2.36, 16.64] Subtotal 599 418 3.85 [2.03, 7.29] Total events: 66 (treatment), 12 (control) Test for heterogeneity: χ²=1.85, df=2 (p=0.40), I²=0% Test for overall effect: z=4.13 (p<0.0001)

Risperidone studies Mahmoud et al. (28) 8/141 3/133 2.61 [0.68, 10.04] Keitner et al. (30) 6/62 1/33 3.43 [0.39, 29.76] Reeves et al. (29) 0/12 0/11 Not estimable Subtotal 215 177 2.84 [0.91, 8.91] Total events: 14 (treatment), 4 (control) Test for heterogeneity: χ²=0.04, df=1 (p=0.83), I²=0% Test for overall effect: z=1.79 (p=0.07)

Quetiapine studies Khullar et al. (31) 0/8 0/7 Not estimable Mattingly et al. (32) 0/1 0/1 Not estimable McIntyre et al. (33) 8/29 2/29 5.14 [0.99, 26.81] Earley et al. (12) 30/330 6/161 2.58 [1.05, 6.34] El-Khalili et al. (13) 43/297 1/148 24.89 [3.39, 182.60] Subtotal 665 346 5.52 [2.71, 11.24] Total events: 81 (treatment), 9 (control) Test for heterogeneity: χ²=4.95, df=2 (p=0.08), I²=59.6% Test for overall effect: z=4.71 (p<0.00001)

Aripiprazole studies Berman et al. (9) 6/182 4/176 1.47 [0.41, 5.29] Berman et al. (11) 11/177 3/172 3.73 [1.02, 13.62] Marcus et al. (10) 7/191 2/190 3.58 [0.73, 17.44] Subtotal 550 538 2.68 [1.23, 5.81] Total events: 24 (treatment), 9 (control) Test for heterogeneity: χ²=1.23, df=2 (p=0.54), I²=0% Test for overall effect: z=2.49 (p=0.01) Total 2029 1479 Total events: 185 (treatment), 34 (control) 3.91 [2.68, 5.72] Test for heterogeneity: χ²=8.73, df=10 (p=0.56), I²=0% Test for overall effect: z=7.05 (p<0.00001) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a Odds ratios for discontinuation rates due to adverse events on drug and placebo are grouped by atypical agent. significant (χ2=8.16, df=15, p=0.92, I2=0%). The pooled re- geneity was not significant (χ2=13.04, df=14, p=0.52, I2= mission rates were 30.7% for atypical antipsychotics, com- 0%). The odds ratios for rates of discontinuation for any pared with 17.2% for placebo. reason did not differ significantly among the drug sub- The odds ratios for remission varied from 1.83 to 2.63 groups (χ2=1.60, df=3, p=0.66). The pooled actual rates of among the drug subgroups, with overlapping confidence discontinuation for any reason were 19.6% in the atypical intervals. Testing the difference between the atypical sub- treatment group and 15.5% in the placebo group. groups indicated no significant differences (χ2=1.50, df=3, The odds ratio for the rate of discontinuation for ad- p=0.68). verse events with drug and placebo was 3.91 (95% CI= 2.68–5.72, z=7.05, N=15, p<0.00001) (Figure 5). The test for Discontinuation Rates heterogeneity was not significant (χ2=8.73, df=10, p=0.56, The odds ratio comparing drug and placebo groups for I2=0%). The risk difference was 0.06 (95% CI=0.05–0.08), rates of discontinuation for any reason was 1.30 (95% CI= resulting in a number needed to harm of 17. The pooled 1.09–1.57, z=2.85, p=0.004) (Figure 4). The test for hetero- adverse event discontinuation rates were 9.1% in the atyp-

986 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS

FIGURE 6. Odds Ratios of Response Rates on Drug and Placebo in Atypical Antipsychotic Augmentation Trials, Grouped by Duration of the Acute-Phase Trial

Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) Trial Duration and Study n/N n/N (95% CI) (95% CI) 4 weeks’ duration Keitner et al. (30) 34/62 11/33 2.43 [1.01, 5.85] Subtotal 62 33 2.43 [1.01, 5.85] Total events: 34 (treatment), 11 (control) Test for heterogeneity: not applicable Test for overall effect: z=1.98 (p=0.05)

6 weeks’ duration Mahmoud et al. (28) 49/137 33/131 1.65 [0.98, 2.80] Berman et al. (9) 61/181 41/172 1.62 [1.02, 2.59] Earley et al. (12) 185/327 74/160 1.51 [1.04, 2.21] Berman et al. (11) 81/174 45/169 2.40 [1.53, 3.77] Marcus et al. (10) 60/185 32/184 2.28 [1.40, 3.72] Subtotal 1004816 1.83 [1.50, 2.25] Total events: 436 (treatment), 225 (control) Test for heterogeneity: χ²=3.50, df=4 (p=0.48), I²=0% Test for overall effect: z=5.86 (p<0.00001)

8 weeks’ duration Shelton et al. (7) 6/10 1/10 13.50 [1.20, 152.21] Shelton et al. (25) 40/146 41/142 0.93 [0.56, 1.55] Khullar et al. (31) 4/8 1/7 6.00 [0.48, 75.34] Mattingly et al. (32) 12/242/13 5.50 [1.00, 30.29] McIntyre et al. (33) 14/29 8/29 2.45 [0.82, 7.31] Thase et al. I (27)a 37/102 31/104 1.34 [0.75, 2.40] Thase et al. II (27)a 43/98 30/102 1.88 [1.05, 3.36] El-Khalili et al. (13) 159/289 66/143 1.43 [0.95, 2.13] Reeves et al. (29) 6/12 4/11 1.75 [0.33, 9.30] Subtotal 718 561 1.50 [1.19, 1.89] Total events: 321 (treatment), 184 (control) Test for heterogeneity: χ²=11.45, df=8 (p=0.18), I²=30.1% Test for overall effect: z=3.38 (p=0.0007)

12 weeks’ duration Corya et al. (26) 100/230 19/56 1.50 [0.81, 2.76] Subtotal 230 56 1.50 [0.81, 2.76] Total events: 100 (treatment), 19 (control) Test for heterogeneity: not applicable Test for overall effect: z=1.30 (p=0.20)

Total 2014 1466 1.69 [1.46, 1.95] Total events: 891 (treatment), 439 (control) Test for heterogeneity: χ²=17.58, df=15 (p=0.29), I²=14.7% Test for overall effect: z=7.00 (p<0.00001) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a This report included two separate trials of identical design. ical antipsychotic group and 2.3% in the placebo group. We also examined whether the odds ratios for response The odds ratios for the subgroups varied from 2.68 to 5.52. differed in trials using history of drug failure rather than a Testing the difference between the atypical subgroups in- prospective trial to establish treatment resistance (Figure dicated no significant differences (χ2=0.66, df=3, p=0.88). 7). Ten of the trials required that patients have failed to re- Trial duration varied from 4 to 12 weeks. Duration of the spond in one prospective trial, and six required that pa- trial did not significantly affect the odds ratio for response tients have a history of nonresponse or failed to respond in to drug and placebo (Figure 6). The odds ratios for trials of the current trial. The odds ratio of the 10 trials requiring a 4, 6, 8, and 12 weeks were 2.43 (95% CI=1.01–5.85), 1.83 failed prospective trial was 1.73 (95% CI=1.45–2.06), which (95% CI=1.50–2.25), 1.50 (95% CI=1.19–1.89), and 1.50 did not differ from the trials using only historical data, in (0.81–2.76), respectively, and did not differ significantly which the odds ratio was 1.61 (95% CI=1.24–2.08). The (χ2=2.63, df=3, p=0.45). This analysis, however, was limited magnitude of the drug-placebo difference, estimated by because only one trial had a duration of 4 weeks and only the risk difference from the meta-analysis, was identical one of 12 weeks. using the two methods, at 0.12. However, the pooled re-

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 987 META-ANALYSIS OF ATYPICAL ANTIPSYCHOTIC AUGMENTATION IN DEPRESSION

FIGURE 7. Odds Ratios of Response Rates on Drug and Placebo in Atypical Antipsychotic Augmentation Trials, Grouped by Method of Establishing Treatment Resistance

Treatment Resistance Subcategory Treatment Control Odds Ratio (Fixed) Odds Ratio (Fixed) and Study n/N n/N (95% CI) (95% CI) Historical failure only Khullar et al. (31) 4/8 1/7 6.00 [0.48, 75.34] Mattingly et al. (32) 12/242/13 5.50 [1.00, 30.29] McIntyre et al. (33) 14/29 8/29 2.45 [0.82, 7.31] Earley et al. (12) 185/327 74/160 1.51 [1.04, 2.21] El-Khalili et al. (13) 159/289 66/143 1.43 [0.95, 2.13] Reeves et al. (29) 6/12 4/11 1.75 [0.33, 9.30] Subtotal 689 363 1.61 [1.24, 2.08] Total events: 380 (treatment), 155 (control) Test for heterogeneity: χ²=4.05, df=5 (p=0.54), I²=0% Test for overall effect: z=3.60 (p=0.0003)

One prospective trial Mahmoud et al. (28) 49/137 33/131 1.65 [0.98, 2.80] Shelton et al. (7) 6/10 1/10 13.50 [1.20, 152.21] Shelton et al. (25) 40/146 41/142 0.93 [0.56, 1.55] Corya et al. (26) 100/230 19/56 1.50 [0.81, 2.76] Thase et al. I (27)a 37/102 31/104 1.34 [0.75, 2.40] Thase et al. II (27)a 43/98 30/102 1.88 [1.05, 3.36] Berman et al. (9) 61/181 41/172 1.62 [1.02, 2.59] Berman et al. (11) 81/174 45/169 2.40 [1.53, 3.77] Keitner et al. (30) 34/62 11/33 2.43 [1.01, 5.85] Marcus et al. (10) 60/185 32/184 2.28 [1.40, 3.72] Subtotal 1325 1103 1.73 [1.45, 2.06] Total events: 511 (treatment), 284 (control) Test for heterogeneity: χ²=13.30, df=9 (p=0.15), I²=32.3% Test for overall effect: z=6.02 (p<0.00001)

Total 2014 1466 1.69 [1.46, 1.95] Total events: 891 (treatment), 439 (control) Test for heterogeneity: χ²=17.58, df=15 (p=0.29), I²=14.7% Test for overall effect: z=6.02 (p<0.00001) 0.10.2 0.5 1 2 5 10 Favors Favors Control Treatment a This report included two separate trials of identical design. sponse rates in trials requiring a failed prospective trial unaffected (1.64 [95% CI=1.42–191] compared with 1.69 (38.6% [511/1,325] and 25.8% [284/1,103] for drug and [95% CI=1.46–196]). placebo, respectively) were considerably lower than response rates in trials using historical data (55.2% [380/ Discussion 689] and 42.7% [155/363], respectively), suggesting that patients who did not respond in prospective treatment In this systematic review and meta-analysis, 15 reports were more treatment resistant. All but two of the trials of 16 adjunctive trials were pooled. Augmentation with continued with the original antidepressant during the ad- atypical antipsychotic agents was significantly more effec- junctive trials. Two of the olanzapine trials switched to tive than placebo for response and remission. The odds ra- olanzapine plus fluoxetine after using a different antide- tio for remission was 2.00, with a number needed to treat pressant during the prospective trial. When these two tri- of nine. No significant differences in efficacy were noted als were excluded from the analysis, the odds ratio for among the different atypical agents (olanzapine, risperi- trials requiring nonresponse in a prospective trial was done, quetiapine, and aripiprazole). The relative efficacy slightly higher at 1.93 (95% CI=1.58–2.36) but still not sig- of the atypical agents compared with placebo in augmen- nificantly different from that in the trials requiring histori- tation did not appear to be influenced by the duration of cal nonresponse. the adjunctive trial or by whether treatment resistance A funnel plot of the odds ratios for response plotted was determined using a prospective trial or historical non- against standard error (log odds ratio) was asymmetric, response. with three of four small studies (N<25 per arm) showing At present, this body of evidence is considerably larger high odds ratios (Figure 8). When these four trials were ex- than that for any other augmentation strategy in the treat- cluded, however, the odds ratio for response was relatively ment of major depressive disorder. For example, the 10

988 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS controlled trials of lithium cited earlier (4) included only FIGURE 8. Funnel Plot of the Odds Ratios for Response Ver- 269 patients, whereas these 16 trials of atypical antipsy- sus the Standard Error (Log Odds Ratio) in Trials of Atypical Antipsychotic Augmentation in Major Depressive Disorder chotics included 3,480 patients. In addition, the early trials of lithium and T3 augmentation often required minimal 0.0 evidence of treatment resistance, for example, 4 weeks of failure to respond to the initial antidepressant. The only 0.4 placebo-controlled lithium augmentation trial to retro- 0.8 spectively and prospectively establish treatment resistance (34) failed to find lithium effective. All of the trials of 1.2 atypicals required evidence of treatment resistance, and

for several of the trials, patients had failed to respond to SE(logratio)odds 1.6 more than one antidepressant. Although the funnel plot of the odds ratios was asym- 0.1 0.2 0.5 1 2 5 10 metric for the smaller studies and thus suggestive of pub- Odds Ratio (fixed) lication or retrieval bias, exclusion of the smaller studies from the meta-analysis had essentially no effect on the mentation agents (e.g., lithium and thyroid) even though odds ratio. After a thorough literature search as well as the efficacy data are stronger for the atypicals. Finally, little contact with the manufacturers of all the atypical agents is known about efficacy and safety of the atypicals during approved for use in the United States, we are reasonably continuation and maintenance treatment in major de- confident that we have included all of the larger studies pression. that have been completed. An asymmetric funnel plot is only suggestive of bias. It might be argued that smaller There are limitations to this analysis. There may be studies with a limited number of sites and careful patient other studies with different results that were not published selection may produce more valid results. or presented and thus were not included in our analysis. We used a fixed-effects model for the meta-analysis be- Definitions of treatment resistance are still evolving, and cause the patient samples were reasonably homogeneous, there is no standard for studies such as these. Yet the the study designs were similar, and all studies used either methods used to define treatment resistance in these trials the Hamilton Depression Rating Scale or the Montgom- of augmentation with atypical agents were more rigorous ery-Åsberg Depression Rating Scale to rate response. Tests than those of previous augmentation trials. While the total for heterogeneity were not statistically significant. Had we number of patients included in these trials (N=3,480) is employed a random-effects model, the results for re- fairly large, the number of trials on which this trial-level sponse (odds ratio=1.69, 95% CI=1.43–2.00, z=6.21, data analysis is based is limited. This limitation is even p<0.00001) and remission (odds ratio=1.99, 95 CI=1.68– more important when comparing subgroups. While ris- 2.36, z=7.95, p<0.00001) would have been nearly identical peridone was studied in three trials, the samples of two of to those obtained using a fixed-effects model. the trials, 23 and 95, were relatively small, and the total While the efficacy of the atypical agents for adjunctive number studied was 386. The other three agents were therapy in major depressive disorder appears fairly well studied in at least 1,000 patients. While adjunctive olanza- established, there are other considerations. The rate of pine appears to be effective, the trials all examined ad- discontinuation due to adverse events was significantly junctive treatment with fluoxetine. The efficacy of olanza- higher for the atypical group (9.1%) than the placebo pine combined with other antidepressants has not been group (2.3%). The risk difference by meta-analysis was well studied. 0.06, with a number needed to harm of 17. While the dis- In addition to factors limiting this analysis, there are continuation rates did not differ significantly among the several limitations to the existing clinical literature regard- agents, rates of specific side effects may be quite different. ing the use of atypicals in depression. For each of these In addition, during continuing treatment there may be agents, limited data are available for establishing an effec- other secondary effects that in the aggregate affect tolera- tive dosage range during adjunctive treatment of depres- bility and patient acceptance. The atypical agents also are sion. Even if efficacy in groups of patients is similar among associated with a variety of relatively serious adverse ef- these agents, it is not clear that patients of the same profile fects, such as metabolic syndrome, extrapyramidal symp- are responding. This would seem unlikely given the differ- toms, and rare but serious symptoms such as tardive dys- ences in neuropharmacology among these agents. Indi- kinesia and neuroleptic malignant syndrome. As a vidual patients may respond better to one agent than to consequence the risk-benefit ratio appears to be different another; yet methods to tailor treatment to the patient from that of several alternative treatments for major de- have not been established. Given the cost and safety issues pressive disorder. Because of these risks, the Texas Algo- for the atypical agents, it will be important to compare rithm Group (M.H. Trivedi et al., unpublished 2007 data) their efficacy and safety with those of other, less expensive placed the atypical antipsychotics below some other aug- augmentation and combination strategies in depression.

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 989 META-ANALYSIS OF ATYPICAL ANTIPSYCHOTIC AUGMENTATION IN DEPRESSION

Finally, all of the trials we reviewed were of short duration. Meltzer HY: A novel augmentation strategy for treating resis- Long-term efficacy and safety data are sorely needed. In tant major depression. Am J Psychiatry 2001; 158:131–134 8. Papakostas GI, Shelton RC, Smith J, Fava M: Augmentation of addition, even if continuing adjunctive treatment is antidepressants with atypical antipsychotic medications for shown to reduce relapse rates, because of the increased treatment-resistant major depressive disorder: a meta-analy- side effect burden and cost of continuing two agents, there sis. J Clin Psychiatry 2007; 68:826–831 will be additional questions. For example, if a patient re- 9. Berman RM, Marcus RN, Swanink R, McQuade RD, Carson WH, mits with adjunctive treatment, can the initial antidepres- Corey-Lisle PK, Khan A: The efficacy and safety of aripiprazole sant be discontinued? If so, when and in which patients? as adjunctive therapy in major depressive disorder: a multicenter, randomized, double-blind, placebo-controlled study. J Alternatively, can individuals be identified who will re- Clin Psychiatry 2007; 68:843–853 main in remission if the atypical agent is withdrawn? 10. Marcus R, McQuade R, Carson W, Hennicken D, Fava M, Simon These questions deserve study. J, Trivedi M, Thase M, Berman R: The efficacy and safety of aripiprazole as adjunctive therapy in major depressive disorder: a Received March 3, 2009; revision received April 23, 2009; accepted second multicenter, randomized, double-blind placebo-con- June 4, 2009 (doi: 10.1176/appi.ajp.2009.09030312). From the De- trolled study. J Clin Psychopharmacol 2008; 28:156–165 partment of Psychiatry, University of California San Francisco; Massa- 11. Berman RB, Fava M, Thase ME, Trivedi MH, Swanink R, Mc- chusetts General Hospital, Boston; and Harvard Medical School, Bos- Quade RD, Carson WH, Adson D, Taylor L, Hazel J, Marcus RN: ton. Address correspondence and reprint requests to Dr. Nelson, The third consecutive, positive, double-blind, placebo con- Department of Psychiatry, University of California San Francisco, 401 trolled trial of aripiprazole augmentation in the treatment of Parnassus Ave., Box 0984-F, San Francisco, CA 94143; major depression, in American College of Neuropsychophar- [email protected] (e-mail). macology 2008 Annual Meeting Abstracts (Scottsdale, Ariz, Dec Dr. Nelson has received grant support from or served as a speaker 7–11, 2008). Nashville, Tenn, ACNP, 2008 on advisory boards, or as a consultant for Abbott Laboratories, Aca- 12.Earley W, McIntyre A, Bauer M, Pretorius HW, Shelton R, dia Pharmaceuticals, AstraZeneca, Biovail, Bristol-Myers Squibb, Cor- cept, Cyberonics, Eli Lilly, Forest Pharmaceuticals, GlaxoSmithKline, Lindgren P, Brecher M: Efficacy and tolerability of extended re- Health Resources and Services Administration, Janssen Pharmaceu- lease quetiapine fumarate (quetiapine extended release) as tica, Medtronics, Merck, NIMH, Novartis Pharmaceuticals, Organon, add-on to antidepressants in patients with major depressive Orexigen, Otsuka, Pfizer U.S. Pharmaceuticals Group, Sepracor, Shire, disorder (MDD): results from a double-blind, randomized, and Sierra Neuropharmaceuticals. Dr. Papakostas has received grant phase III study, in American College of Neuropsychopharma- support from or served as a speaker, on advisory boards, or as a con- cology 2007 Annual Meeting Abstracts (Boca Raton, Fla, Dec 9– sultant for Bristol-Myers Squibb, Eli Lilly, Evotec AG, GlaxoSmithKline, 13, 2007). Nashville, TN, ACNP, 2007 Inflabloc Pharmaceuticals, Jazz Pharmaceuticals, Lundbeck, NIMH, 13. El-Khalili N, Joyce M, Atkinson S, Buynak R, Datto C, Lindgren P, Otsuka, PAMLAB LLC, Pfizer U.S. Pharmaceuticals Group, Pierre Fabre Laboratories, Precision Human Biolaboratories, Shire Pharmaceuti- Eriksson H, Brecher M: Adjunctive extended-release quetiapine cals, Titan Pharmaceuticals, and Wyeth. fumarate (quetiapine-extended release) in patients with major No external funding was received for the study design, trial search, depressive disorder and inadequate antidepressant response, data analysis, interpretation of the data, writing of the paper, or the in American Psychiatric Association 2008 Annual Meeting: New decision to submit for publication. Research Abstracts (Washington, DC, May 3–8, 2008). Washing- ton, DC, APA, 2008 14. Reynolds GP: Receptor mechanisms in the treatment of schizo- References phrenia. J Psychopharmacology 2004; 18:340–345 15. Stahl SM, Shayegan DK: The psychopharmacology of ziprasi- 1. Kessler RC, Berglund P, Demler O, Jin R, Koretz D, Merikangas done: receptor-binding properties and real-world psychiatric KR, Rush AJ, Walters EE, Wang PS: The epidemiology of major practice. J Clin Psychiatry 2003; 64(suppl 19):6–12 depressive disorder: results from the National Comorbidity 16. Farah A: Atypicality of atypical antipsychotics. Prim Care Com- Survey replication (NCS-R). JAMA 2003; 289:3095–3105 panion. J Clin Psychiatry 2005; 7:268–274 2. Rush AJ, Trivedi MH, Wisniewski SR, Nierenberg AA, Stewart JW, 17. Tadori Y, Forbes RA, McQuade RD, Kikuchi T: Characterization Warden D, Niederehe G, Thase ME, Lavori PW, Lebowitz BD, of aripiprazole partial agonist activity at human dopamine McGrath PJ, Rosenbaum JF, Sackeim HA, Kupfer DJ, Luther J, D(3) receptors. Eur J Pharmacol 2008; 597:27–33 Fava M: Acute and longer-term outcomes in depressed outpa- 18. Jensen NH, Rodriguiz RM, Caron MG, Wetsel WC, Rothman RB, tients requiring one or several treatment steps: a STAR*D re- Roth BL: N-desalkylquetiapine, a potent norepinephrine report. Am J Psychiatry 2006; 163:1905–1917 uptake inhibitor and partial 5-HT1A agonist, as a putative me- 3. Nelson JC: Augmentation strategies for treatment of unipolar diator of quetiapine’s antidepressant activity. Neuropsycho- major depression, in Mood Disorders: Systematic Medication pharmacology 2008; 33:2303–2312 Management, vol 25, Modern Problems of Pharmacopsychia- 19. Tatsumi M, Jansen K, Blakely RD, Richelson E: Pharmacologic try. Edited by Rush AJ. Basel, Karger, 1997, pp 34–55 profile of neuroleptics at human monoamine transporters. Eur 4. Crossley NA, Bauer M: Acceleration and augmentation of anti- J Pharmacol 1999; 368:277–283 depressants with lithium for depressive disorders: two meta- 20. Zhang W, Perry KW, Wong DT, Potts BD, Bao J, Tollefson GD, By- analyses of randomized, placebo-controlled trials. J Clin Psychi- master FP: Synergistic effects of olanzapine and other antipsy- atry 2007; 68:935–940 chotic agents in combination with fluoxetine on norepineph- 5. Nelson JC: The use of antipsychotic drugs in the treatment of rine and dopamine release in rat prefrontal cortex. depression, in Treating Resistant Depression. Edited by Zohar J, Neuropsychopharmacology 2000; 23:250–262 Belmaker RH. New York, PMA Publishing, 1987, pp 131–146 21. Hamilton M: A rating scale for depression. J Neurol Neurosurg 6. Ostroff RB, Nelson JC: Risperidone augmentation of selective Psychiatry 1960; 23:56–61 serotonin reuptake inhibitors in major depression. J Clin Psy- 22. Montgomery SA, Åsberg M: A new depression scale designed to chiatry 1999; 60:256–259 be sensitive to change. Br J Psychiatry 1979; 134:382–389 7. Shelton RC, Tollefson GD, Tohen M, Stahl S, Gannon KS, Jacobs 23. Higgins JP, Thompson SG: Quantifying heterogeneity in a meta- TG, Buras WR, Bymaster FP, Zhang W, Spencer KA, Feldman PD, analysis. Stat Med 2002; 21:1539–1558

990 ajp.psychiatryonline.org Am J Psychiatry 166:9, September 2009 NELSON AND PAPAKOSTAS

24. Deeks J, Altman D, Bradburn M: Statistical methods for exam- ized, double-blind, placebo controlled pilot study. J Clin Psychi- ining heterogeneity and combining results from several studies atry 2008; 69:1228–1336 in meta-analysis, in Systematic Reviews in Health Care: Meta- 30. Keitner GI, Garlow SJ, Ryan CE, Ninan PT, Solomon DA, Nemer- Analysis in Context, 2nd ed. Edited by Egger M, Davey-Smith G, off CB, Keller MB: A randomized, placebo-controlled trial of ris- Altman D. London, BMJ Publishing Group, 2001, pp 285–312 peridone augmentation for patients with difficult-to-treat uni- 25. Shelton RC, Williamson DJ, Corya SA, Sanger TM, Van Campen polar, non-psychotic major depression. J Psychiatr Res 2009; LE, Case M, Briggs SD, Tollefson GD: Olanzapine/fluoxetine 43:205–214 combination for treatment-resistant depression: a controlled 31. Khullar A, Chokka P, Fullerton D, McKenna S, Blackamn A: A study of SSRI and nortriptyline resistance. J Clin Psychiatry double-blind, randomized, placebo-controlled study of que- 2005; 66:1289–1297 tiapine as augmentation therapy to SSRI/SNRI agents in the treatment of non-psychotic unipolar depression with residual 26. Corya SA, Williamson DJ, Sanger TM, Briggs SD, Case M, Tollef- symptoms, in American Psychiatric Association 2006 Annual son GD: A randomized, double-blind comparison of olanza- Meeting: New Research Abstracts (Toronto, Canada, May 20– pine/fluoxetine combination, olanzapine, fluoxetine, and ven- 25, 2006). Washington, DC, APA, 2006 lafaxine in treatment-resistant depression. Depress Anxiety 32. Mattingly G, Ilivicky H, Canale J, Anderson R: Quetiapine com- 2006; 23:364–372 bination for treatment-resistant depression, in American Psy- 27. Thase M, Corya SA, Osuntokun O, Case M, Henley DB, Sanger chiatric Association 2006 Annual Meeting: New Research Ab- TM, Watson SB, Dube S: A randomized, double-blind compari- stracts (Toronto, Canada, May 20–25, 2006). Washington, DC, son of olanzapine/fluoxetine combination, olanzapine, and APA, 2006 fluoxetine in treatment-resistant major depressive disorder. J 33. McIntyre A, Gendron A, McIntyre A: Quetiapine adjunct to se- Clin Psychiatry 2007; 68:224–236 lective serotonin reuptake inhibitors or venlafaxine in patients 28. Mahmoud RA, Pandina G, Turkoz I, Kosik-Gonzalez C, Canuso with major depression, comorbid anxiety, and residual depres- CM, Kujawa MJ, Gharabawi-Garibaldi GM: Risperdione for treat- sive symptoms: a randomized, placebo-controlled pilot study. ment-refractory major depressive disorder. Ann Intern Med Depress Anxiety 2007; 24:487–494 2007; 147:593–602 34. Nierenberg AA, Papakostas GI, Petersen T, Montoya HD, Worth- 29. Reeves H, Batra S, May RS, Zhang R, Dahl DC, Li X: Efficacy of ington JJ, Tedlow J, Alpert JE, Fava M: Lithium augmentation of risperidone augmentation to antidepressant in the manage- nortriptyline for subjects resistant to multiple antidepressants. ment of suicidality in major depressive disorder: a random- J Clin Psychopharmacol 2003; 23:92–95

Am J Psychiatry 166:9, September 2009 ajp.psychiatryonline.org 991