aPatients can be in more than 1 category for reason of exclusion. Reasons for exclusion from the specific analysis population and the number of patients in that category are given beneath the total of the specific analysis population.
A, Change of Positive and Negative Syndrome Scale (PANSS) negative symptom factor score from baseline (mean [SE]). All patients were receiving primary antipsychotic treatment. B, Response rates at week 8. C, Clinical Global Impression–Global Improvement for negative symptoms at week 8. D, Personal and Social Performance Scale scores at baseline and week 8. All data shown are for the per-protocol population (placebo, n = 61; 10-mg–dose group, n = 60; 30-mg–dose group, n = 57; and 60-mg–dose group, n = 53).
The basis for modeling was occupancies observed in the thalamus, pons, and cerebellum in healthy volunteers after multiple doses of bitopertin.48
eTable 1. Patient Disposition and Safety
eTable 2. Primary and Secondary End Points for the ITT Population
Daniel Umbricht, Daniela Alberati, Meret Martin-Facklam, Edilio Borroni, Eriene A. Youssef, Michael Ostland, Tanya L. Wallace, Frédéric Knoflach, Ernest Dorflinger, Joseph G. Wettstein, Alexander Bausch, George Garibaldi, Luca Santarelli. Effect of Bitopertin, a Glycine Reuptake Inhibitor, on Negative Symptoms of SchizophreniaA Randomized, Double-Blind, Proof-of-Concept Study. JAMA Psychiatry. 2014;71(6):637–646. doi:10.1001/jamapsychiatry.2014.163
In schizophrenia, the severity of negative symptoms is a key predictor of long-term disability. Deficient signaling through the N-methyl-D-aspartate receptor is hypothesized to underlie many signs and symptoms associated with schizophrenia in particular negative symptoms. Glycine acts as an N-methyl-D-aspartate receptor coagonist. Blockade of the glycine transporter type 1 to inhibit glycine reuptake and elevate synaptic glycine concentrations represents an effective strategy to enhance N-methyl-D-aspartate receptor transmission.
To determine the efficacy and safety of bitopertin (RG1678), a glycine reuptake inhibitor, in patients with schizophrenia and predominant negative symptoms who were stable while taking an antipsychotic treatment.
Design, Setting, and Participants
This randomized, double-blind, placebo-controlled, phase 2 proof-of-concept trial involved 323 patients with schizophrenia and predominant negative symptoms across 66 sites worldwide.
Bitopertin (10, 30, or 60 mg/d) or placebo added to standard antipsychotic therapy for a treatment duration of 8 weeks.
Main Outcomes and Measures
Change from baseline in the Positive and Negative Syndrome Scale negative factor score.
In the per-protocol population, 8 weeks of treatment with bitopertin was associated with a significant reduction of negative symptoms in the 10-mg/d (mean [SE] reduction in negative symptoms score, −25% [2%]; P = .049) and 30-mg/d (mean [SE], −25% [2%]; P = .03) bitopertin groups, a significantly higher response rate and a trend toward improved functioning in the 10-mg/d group when compared with placebo (mean [SE], −19% [2%]). Results reached trend-level significance in the intent-to-treat population. Estimates of bitopertin binding to glycine transporter type 1 showed that low to medium levels of occupancy yielded optimal efficacy in patients, consistent with findings in preclinical assays.
Conclusions and Relevance
Bitopertin-mediated glycine reuptake inhibition may represent a novel treatment option for schizophrenia, with the potential to address negative symptoms.
clinicaltrials.gov Identifier: NCT00616798
Persistent negative symptoms and cognitive deficits represent major independent factors determining long-term disability and successful social and occupational rehabilitation of patients with schizophrenia.1- 5 Antipsychotics exert limited effects on cognitive deficits or persistent negative symptoms,3,4 which may explain why as few as 30% of patients treated with second-generation antipsychotics (SGAs) achieve functional or symptomatic remission and even fewer attain full recovery.1,6
In healthy volunteers, subanesthetic doses of the N-methyl-D-aspartate receptor (NMDAR) antagonists phencyclidine and ketamine induce psychotomimetic effects resembling positive and negative symptoms and cognitive deficits of schizophrenia,7- 12 and they exacerbate both positive and negative symptoms in patients with schizophrenia.9,10,13,14 In contrast, dopaminergic agents, such as amphetamine, or agonists at the 2A serotonin (5-hydroxytryptamine type 2A) receptor, such as psilocybin, only induce effects mimicking positive symptoms of schizophrenia.15,16 Thus, reduced NMDAR signaling may not only contribute to positive symptoms, but also uniquely be a root cause of negative symptoms of schizophrenia. Therefore, agents that enhance NMDAR transmission may impact not only positive, but also negative symptoms of schizophrenia.
Activation of the NMDAR requires the binding of both glutamate and glycine.17 Transmission of NMDAR can thus be enhanced either directly with glycine site agonists, such as glycine and D-serine; partial agonists, such as D-cycloserine; or indirectly by increasing the synaptic glycine concentration through inhibition of the glycine transporter type 1 (GlyT1), which maintains intrasynaptic glycine at a subsaturating level.18 Clinical trials of glycine site agonists or the glycine reuptake inhibitor sarcosine in patients with chronic schizophrenia showed favorable effects particularly on the negative symptom domain.19- 27 However, 2 large subsequent trials—the Cognitive and Negative Symptoms in Schizophrenia Trial28 and the study by Weiser et al29 specifically evaluating the efficacy of glycine and D-cycloserine and D-serine, respectively, as adjunct therapy for the treatment of negative symptoms—did not replicate these findings. Nevertheless, 3 meta-analyses concluded that this pharmacological approach had beneficial effects on negative symptoms30- 32 and on all symptom domains of schizophrenia.31,32
Bitopertin (RG1678) is a potent and selective inhibitor of GlyT1 (IC50, 25 nM).33 Preclinical studies showed effects consistent with enhanced NMDAR signaling and normalization of induced dopaminergic and glutamatergic abnormalities.34 We conducted a phase 2 proof-of-concept (PoC) study to assess whether bitopertin could ameliorate negative symptoms of schizophrenia. We used a trial design following the recommendations by the National Institute of Mental Health–Measurement and Treatment Research to Improve Cognition in Schizophrenia consensus statement on negative symptoms35,36 and in which bitopertin was added to stable SGA treatment.
This randomized, placebo-controlled, double-blind, phase 2 study was conducted at 66 sites in Brazil, France, Germany, Hungary, Japan, Mexico, Poland, Russia, and the United States, following International Conference on Harmonization Guidelines for Good Clinical Practice (ClinicalTrials.gov identifier: NCT00616798). The protocol was approved by the health authorities of each country and respective ethics committees of each site. All patients gave written informed consent.
The inclusion and exclusion criteria were defined in a manner to ensure the recruitment of patients with predominant negative symptoms but with few positive symptoms and other potential confounders of negative symptoms (eg, extrapyramidal symptoms [EPS] and depression).36 Key inclusion criteria were (1) age 18 to 60 years; (2) total score of greater than or equal to 40 on the sum of the 14 items constituting the Positive and Negative Syndrome Scale (PANSS) negative symptoms and disorganized thought/cognition factors37; (3) a score of less than or equal to 28 on the sum of the 8 items of the PANSS positive symptoms factor and a score of 4 on 2 or fewer of the items P1, P3, P6, and G9 and none with a score of 5; and (4) taking 2 or fewer antipsychotics, with the primary antipsychotic being an SGA and the total dose of all antipsychotics not exceeding 6 mg of risperidone equivalents. Key exclusion criteria were (1) a score of 4 or more on the PANSS item G6 (depression); (2) any movement disorder due to antipsychotic treatment not currently controlled with anti-EPS treatment (for a complete listing of all inclusion and exclusion criteria, see eAppendix in Supplement); and (3) clozapine treatment within the last 3 months. Demographic information was collected, including race and ethnicity, by patient self-report according to the categories specified by the Food and Drug Administration.
After screening and a 4-week run-in period to document clinical stability without medication or dosing changes, patients were randomized to receive once-daily placebo or bitopertin, 10, 30, or 60 mg/d, for 8 weeks, followed by a 4-week follow-up period (Figure 1). All patients continued to receive their respective antipsychotic treatment throughout the study without change.
Randomization was centralized using an interactive voice response system. A medication (or randomization) list was generated at Roche, and the interactive voice response system randomly assigned a treatment to each enrolled patient, ensuring a best possible balance among treatment groups overall and for each stratification factor (baseline PANSS negative symptom factor score [NSFS, ≤24 vs >24], primary antipsychotic at randomization [olanzapine vs quetiapine vs aripiprazole vs others], and study center). The medication list was not available to the study center, monitors, or any project team members. The results of the following laboratory tests were not sent to either the sites or the sponsor to reduce the potential risk for unblinding: hemoglobin (Hb), hematocrit, erythrocyte count, reticulocyte count, mean red blood cell volume, mean cellular Hb, mean corpuscular Hb concentration, serum iron, ferritin, and soluble transferrin receptor.
Patients were assessed by trained and certified raters using the PANSS,38 the Clinical Global Impression–Severity of Illness (CGI-S), and the CGI–Global Improvement (CGI-I)39 for overall psychopathology, as well as the CGI-S and CGI-I of Negative Symptoms (CGI-I-N)40 for negative symptoms. Functioning was evaluated with the Personal and Social Performance (PSP)41 scale and quality of life was assessed with the patient-administered Self-Report of Quality of Life Scale (SQLS)–Revision 4.42,43 Cognitive functioning was assessed with a computerized test battery (CNS Vital Signs).44
Patients were assessed at screening; 2 weeks before baseline; baseline; and weeks 1, 2, 4, 6, 8, 10, and 12. The primary efficacy measure was the change from baseline in the PANSS NSFS37 (ie, the sum of PANSS items N1, N2, N3, N4, N6, G7, and G16). Secondary efficacy measures were the CGI-I-N score and the percentage of patients with a clinical response defined a priori as 20% improvement from baseline in the PANSS NSFS. Additional measures included changes in PANSS total score and the remaining PANSS factor scores37; clinical global scores (CGI-S and CGI-I); cognitive functioning based on the computerized cognitive battery composite and domain scores; functional status (mean change in PSP41 score); and quality of life (change in SQLS42 [total score and vitality and psychosocial domain scores]). The Barnes Akathisia Rating Scale,45 Simpson-Angus Scale,46 and Abnormal Involuntary Movements Scale39 were used to assess EPS.
Safety evaluations included adverse events (AEs), laboratory tests, electrocardiogram and vital signs, urine drug screen, and pregnancy tests. Because treatment with a glycine reuptake inhibitor potentially affects hematopoiesis, Hb levels and other parameters of hematopoiesis were assessed at each visit.
A mixed-model repeated-measure analysis was used for variables with repeated scheduled measurements. The primary mixed-model repeated-measure model included fixed-effect terms for treatment group, visit week (categorical and nested within patient), baseline score, treatment-by-visit interaction, and interaction between baseline score and visit week. Regional effects were explored in a second model. An unstructured covariance matrix was used to model the repeated-measures covariance. An analysis of covariance was used for variables with a single scheduled postbaseline measurement. Dichotomous variables (ie, responders) were analyzed using the Cochran-Mantel-Haenszel Test including geographical region of the study centers as a stratification factor. Categorical variables (CGI-I and CGI-I-N scores) were analyzed using the Van Elteren Test, adjusting for region. Because this was an exploratory study, adjustment for multiple testing was not applied. Nominal P values for 2-tailed tests without any adjustment are presented.
Because the main analysis of the key efficacy variables used a mixed-effects model, no imputation for missing data was applied. Missing data for the primary end point led to exclusion from the per-protocol (PP) analysis.
Prespecified analyses included PP and intent-to-treat (ITT) populations. The ITT population comprised all patients receiving at least 1 dose of the study drug and having at least 1 assessment of the primary efficacy variable after baseline. All data were analyzed by randomized treatment group regardless of actual treatment received. The PP population comprised a subset of the ITT population that received study medication as randomized, completed 8 weeks of double-blind treatment (ie, treatment ≥49 days), were compliant (ie, took 80%-120% of the study tablets as established by pill count), had valid primary efficacy variable scores (PANSS NSFS) at baseline and all treatment assessments after baseline (weeks 1, 2, 4, 6, and 8), and had negative urine drug screening results during the treatment period (Table 1). The safety population consisted of all patients who received at least 1 dose of the study medication, whether withdrawn prematurely or not.
Glycine transporter type 1 occupancies were estimated by applying individual exposure data to an exposure-occupancy model developed in a positron-emission tomography occupancy study of bitopertin and [11C]RO5013853, a proprietary ligand for GlyT1, in 18 healthy volunteers.47,48
Of 343 patients recruited, 323 completed a prospective run-in period and were randomized to the 4 treatment arms (Figure 1). Demographic characteristics, symptoms, and primary antipsychotic treatments were well matched across the 4 arms for all patient populations analyzed. As intended by the inclusion criteria (see the Methods section), patients showed considerably higher levels of negative than positive symptoms at baseline, fulfilling the definition of patients with predominant negative symptoms (Table 1).37 In addition, baseline scores for EPS and depressive symptoms were low in all treatment groups (Table 1). Similar percentages of patients in the 4 treatment arms completed the study (Figure 1; eTable 1 in Supplement). More patients in the bitopertin 30-mg/d and 60-mg/d groups (9% and 10%, respectively) withdrew for safety and tolerability reasons than in the placebo and bitopertin 10-mg/d groups (1% each) (Figure 1; eTable 1 in Supplement). Patients were categorized into ITT and PP analysis populations (Figure 1). Because the primary goal of this PoC study was specifically to determine the efficacy of bitopertin against negative symptoms, findings from the PP population were considered most relevant. However, results for the ITT population are also presented, with full details in eTable 2 in Supplement.
In the PP population, the reduction of the PANSS NSFS score37 was significantly greater in the 10-mg/d (mean [SE], −25% [2%]; P = .049) and 30-mg/d (mean [SE], −25% [2%]; P = .03) groups than in the placebo group (mean [SE], −19% [2%]) (Figure 2A, Table 2). The mean [SE] reductions in the 60-mg group and in the placebo group were comparable (placebo: −19% [2%]; 60 mg: −20% [2%]). Effect sizes (ESs) for the change in NSFS for 10 mg/d and 30 mg/d were 0.37 and 0.40, respectively. The corresponding dose groups in the ITT population also showed reductions in the PANSS NSFS that reached trend-level significance (mean [SE], 10 mg: −25% [2%], P = .07; 30 mg, −24% [2%], P = .09; 60 mg, −19% [2%]; placebo: 20% [2%] [eTable 2 in Supplement]; ES for 10 mg, 0.29; ES for 30 mg, 0.27).
In the PP population, the response criterion (≥20% improvement in the PANSS NSFS) was met by significantly more patients treated with 10 mg/d than with placebo (65% vs 43%, P = .01), equating to a number needed to treat of 5. The difference in response rates between the 10-mg/d and placebo groups reached trend-level significance in the ITT population (58% vs 44%, P = .08; eTable 2 in Supplement). Patients in the PP population receiving 30 mg/d showed a statistical trend for a greater response rate (60%, P = .09), while there was no difference in response rates between the 60-mg/d and placebo groups (both 43%) (Figure 2B).
Based on the CGI-I-N scale, significantly more patients in the 10-mg group in both analysis populations were rated much or very much improved than in the placebo group (PP population: 36.6% vs 23.0%, P = .03, Figure 2C, Table 2; ITT population: 33.4% vs 19.50%, P = .02, eTable 2 in Supplement). In PP patients receiving 30 mg/d, the proportion of patients rated as much improved/very much improved reached trend-level significance compared with those receiving placebo (P = .06) (Table 2). Differences between the 60-mg/d and the placebo groups were not significant (Table 2).
Although the decreases from baseline in the PANSS total and other factor scores were generally greater in the 10-mg/d and 30-mg/d groups than in the placebo group for all patient populations analyzed, they did not differ significantly (eTable 2 in Supplement). Reductions from baseline in the PANSS total and other factor scores were similar in the 60-mg/d and placebo groups in both analysis populations (Table 2; eTable 2 in Supplement).
The greatest effect of bitopertin on patient function as assessed with the PSP scale was observed in the 10-mg/d group of the PP population (mean [SE] change in total PSP score, 10 mg: 8.76 [1.15]; placebo: 5.96 [1.13]; P = .07; Figure 2D). The corresponding difference was not significant in the ITT population (P = .15) (eTable 2 in Supplement). Changes from baseline PSP total score for the other 2 bitopertin dose groups did not differ from those that took placebo, regardless of patient population analyzed.
Total and domain scores on the SQLS improved from baseline across all treatment groups but no significant differences between bitopertin and placebo were observed (data not shown). The results of the effects on cognition as measured by the CNS Vital Signs battery were negative.
Drug-related AEs showed a dose-dependent increase (18%, 23%, and 31% of AEs in patients receiving 10-mg, 30-mg, and 60-mg doses, respectively) (eTable 1 in Supplement). Similarly, the number of patients withdrawing from the trial owing to AEs was higher (9%) in the 30-mg and 60-mg groups than in the placebo and 10-mg groups (1%). Common AEs (incidence ≥5%) included somnolence, dizziness, and headache (eTable 2 in Supplement). The safety profiles of the 10-mg/d and placebo groups were similar. There were no drug-related effects on weight; vital signs; laboratory parameters, including glucose, triglycerides, and cholesterol; or on electrocardiogram parameters including PR intervals, R-R intervals, QRS, and QTc. The percentages of patients with a more than 2-g/dL reduction in Hb at any time during treatment were 4%, 5%, 10%, and 21% for the placebo and bitopertin 10-mg, 30-mg, and 60-mg groups, respectively. Serious AEs occurred in 1 patient in the 10-mg group, 2 patients in the 30-mg group, and 3 patients in the 60-mg group (eTable 1 in Supplement).
The mean (SD) estimated occupancy at GlyT1 for the 10-mg/d group was 47% (10%), with the 10th and 90th exposure percentiles equaling occupancies of 36% and 57%; for the 30-mg/d dose group, it was 67% (7%), with the 10th and 90th exposure percentiles equaling occupancies of 59% and 75%; and for 60-mg/d, it was 77% (5%), with the 10th and 90th exposure percentiles equaling estimated occupancies of 69% and 82% (Figure 3). Thus, moderate occupancy was associated with the strongest clinical effect.
In this PoC trial of patients with stable schizophrenia and predominant negative symptoms, bitopertin significantly reduced negative symptoms in patients completing a full course of 8 weeks of treatment. Given the inclusion and exclusion criteria, it is plausible to conclude that the observed improvement of negative symptoms by bitopertin reflects a primary effect on negative symptoms and does not result indirectly from an improvement in other symptom domains. Overall, 10 mg/d of bitopertin was most efficacious in reducing negative symptoms across continuous and categorical efficacy parameters and in both analysis populations, while 30 mg per day produced a similar, but somewhat weaker, effect. In patients receiving bitopertin, 10 mg per day, we observed a trend effect on personal and social functioning, consistent with previous evidence that negative symptom severity is a key factor driving functional disability.1,2,4,5 Bitopertin, 60 mg per day, did not differ from placebo on any outcome measure. The question of whether doses bracketing 10 mg per day, such as 5 mg per day or 20 mg per day, might offer even more benefit is currently being evaluated in the ongoing phase 3 trials.
Without comparative data available, the relevance of the observed ESs (difference from placebo for change in NSFS score for 10-mg ES: 0.37 and for 30-mg ES: 0.40) is difficult to gauge. The substantial improvement observed in the placebo group is in line with placebo responses reported in previous studies of patients with negative symptoms49,50 and may result from the intense interaction with the patient by the study staff in a clinical trial setting.
An inverted U-shaped dose-response relationship was apparent because 60 mg did not differ from placebo across all outcome measures. Consistent with these clinical observations, bitopertin showed a similar inverted U-shaped dose-response relationship in a variety of preclinical models, including cognitive tests in monkeys (T.L.W. and D.A., unpublished observations, July 2009), reversal of amphetamine-induced hyperlocomotion in mice, and long-term potentiation in rat hippocampal slices.34 We estimated that the clinically efficacious doses of 10 mg/d and 30 mg/d achieved mean GlyT1 occupancies of 47% and 67%, respectively. Similarly, in the preclinical studies, low-to-moderate occupancy was associated with maximal pharmacologic activity (D.A. and E.B.; unpublished observation; rat experiments, September 2008; nonhuman primates, September 2010). The consistency of the clinical and preclinical observations supports the conclusion that a low-to-moderate level of GlyT1 inhibition is optimal for best clinical efficacy.
A potential explanation for these findings is the observation that high levels of glycine result in agonist-dependent internalization of NMDARs, an effect that is blocked by inhibiting clathrin-mediated endocytosis.51,52 Presumably, excessive inhibition of GlyT1 may have similar effects on NMDAR availability, eclipsing the glycine-induced enhancement of NMDAR signaling. Alternatively, an excessive enhancement of gamma-aminobutyric acidergic drive at high levels of GlyT1 inhibition due to increased NMDAR activity on gamma-aminobutyric acidergic neurons in local circuits may tip the inhibitory/excitatory balance in favor of inhibition.
Safety and tolerability outcomes with 10 mg/d were similar to those with placebo. The addition of bitopertin to antipsychotics did not increase the risks associated with metabolic parameters or QTc prolongation. Glycine transporter type 1 inhibition by bitopertin could impair Hb synthesis because glycine is transported into erythroblasts by GlyT1 for heme biosynthesis.53 Indeed, we observed a modest, dose-dependent decrease in Hb concentration that was reversible on treatment cessation.
As an exploratory PoC study, this trial had the limitations of a medium sample size and a relatively short duration. This may explain the relatively modest ESs observed in this trial and the absence of appreciable effects on quality of life as assessed by the patients themselves. Indeed, an 8-week trial duration may be insufficient to fully evaluate the therapeutic potential of a treatment for negative symptoms and the related impact on functioning and quality of life.32 Furthermore, given the exploratory nature of this PoC study, we did not apply corrections for multiple comparisons in our analyses, which may result in an increased chance of a false-positive finding. In addition, this PoC study evaluated the effects of bitopertin on predominant negative symptoms only. However, the NMDA hypothesis would also postulate an amelioration of positive symptoms as a consequence of improved NMDAR signaling. Currently, phase 3 trials evaluating the effect of adjunctive treatment with bitopertin on positive symptoms that are suboptimally responsive to SGAs (ClinicalTrials.gov Identifiers: NCT01235559 and NCT01235585) will provide a detailed assessment of the effects of bitopertin on persistent positive symptoms in patients with chronic schizophrenia. Studies are exploring longer treatment duration and will provide a more detailed characterization of the therapeutic potential of bitopertin for the treatment of schizophrenia and associated social and occupational disabilities (ClinicalTrials.gov Identifiers: NCT01192906, NCT01192880, and NCT01192867).
Corresponding Author: Luca Santarelli, MD, F. Hoffmann–La Roche Ltd, B 74/R.3W.300, 4070 Basel, Switzerland (firstname.lastname@example.org).
Submitted for Publication: August 30, 2013; final revision received January 24, 2014; accepted January 24, 2014.
Published Online: April 2, 2014. doi:10.1001/jamapsychiatry.2014.163.
Author Contributions: Dr Umbricht 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.
Study concept and design: Alberati, Martin-Facklam, Borroni, Knoflach, Dorflinger, Wettstein, Garibaldi, Santarelli.
Acquisition, analysis, or interpretation of data: Umbricht, Martin-Facklam, Borroni, Youssef, Ostland, Wallace, Bausch, Santarelli.
Drafting of the manuscript: Umbricht, Wallace, Knoflach, Wettstein, Garibaldi, Santarelli.
Critical revision of the manuscript for important intellectual content: Alberati, Martin-Facklam, Borroni, Youssef, Ostland, Dorflinger, Bausch, Garibaldi, Santarelli.
Statistical analysis: Ostland.
Administrative, technical, or material support: Youssef, Wallace, Dorflinger, Garibaldi.
Study supervision: Umbricht, Borroni, Youssef, Bausch.
Conflict of Interest Disclosures: Drs Umbricht, Alberati, Martin-Facklam, Borroni, Youssef, Ostland, Knoflach, Dorflinger, Wettstein, Garibaldi, and Santarelli are employees of F. Hoffmann–La Roche Ltd. Dr Wallace was an employee of F. Hoffmann–La Roche Ltd during the conduct of the experiments in nonhuman primates. Dr Bausch was an employee of F. Hoffmann–La Roche Ltd during the conduct of the proof-of-concept study.
Funding/Support: This study was sponsored by F. Hoffman–La Roche Ltd.
Role of the Sponsor: F. Hoffmann–La Roche Ltd had a role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Donald Goff, MD (Department of Psychiatry, New York University, New York), and Marc Laruelle, MD, PhD (Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut), for valuable comments on earlier versions of this manuscript. They did not receive compensation for their contributions. We also thank the investigators of the phase 2 study for recruiting and assessing patients: Csaba Almasi, Aleksander Araszkiewicz, Alla Avedisova, Istvæn Bitter, Przemyslaw Bogacki, Rodrigo Bressan, David Brown, Miranda Chakos, Læszl Csekey, Marek Cwiakala, Daniel Dassa, Natalia Dobrovolskaya, Helio Elkis, Tetsuro Enomoto, Donald Garcia, Hamilton Grabowski, Andrey Gribanov, Jun Ishigooka, Nakao Iwata, Mieczyslaw Janiszewski, Lala Kasimova, Hiroaki Kawasaki, Mary Knesevich, Sændor Koffler, Alex Kopelowicz, Alexsander Kotsubinsky, Jelena Kunovac, Tamas Kurimay, Toshihide Kuroki, Ichiro Kusumi, Mark Lerman, Andreas Mahler, Margarita Morozova, Joaquim Mota Neto, Tadashi Murakami, Judit Nagy, Henry Nasrallah, Nikolay Neznanov, Claus Normann, Mark Novitsky, Youtaro Numachi, Toshinari Odawara, Tetsuro Ohmori, Irismar Reis de Oliveira, Alfonso Ontiveros, György Ostorharics-Horvæth, Hiroki Ozawa, Høctor Pinedo, Philippe Raymondet, Robert Reisenberg, Sandra Ruschel, Margot Schmitz, Gunther Schumann, Kazumasa Shiozaki, Rajinder Shiwach, George Simpson, Joachim Springub, Klaus Steinwachs, Jaroslaw Strzelec, Louise Thurman, Gabor Vincze, David Walling, Ryszard Wardenski, Kauser Yakhin, Reiji Yoshimura, and Marcel Zins-Ritter. All investigators were compensated for their work according to the contracts made for their participation in the study. Editorial assistance for this manuscript was provided by ApotheCom and Archimed Medical Communication AG, Switzerland.