A randomized, double-blind, placebo-controlled study of maintenance treatment with adjunctive risperidone long-acting therapy in patients with bipolar I disorder who relapse frequently


  • Support for this study was provided by Ortho-McNeil Janssen Scientific Affairs, LLC, Titusville, NJ, USA. LA, WM, JTH, NT, IT, and CB are full-time employees of Ortho-McNeil Janssen Scientific Affairs, LLC. At the time of this study, MK was a full-time employee of Ortho-McNeil Janssen Scientific Affairs, LLC; she is currently an employee of Hoffman-La Roche, Inc. At the time of this study, RM was a full-time employee of Ortho-McNeil Janssen Scientific Affairs, LLC; he is currently an employee of Ethicon, Inc., a Johnson & Johnson company.

Corresponding author:
Wayne Macfadden, M.D.
Ortho-McNeil Janssen Scientific Affairs, LLC
1125 Trenton-Harbourton Road - A32404
Titusville, NJ 08560, USA
Fax: 609-730-3125
E-mail: wmacfadd@its.jnj.com
Clinical trials government registration number: NCT00094926; http://www.clinicaltrials.gov/ct2/show/NCT000;94926?term


Objective:  No large controlled trials have evaluated adjunctive maintenance treatment with long-acting injectable antipsychotics in patients with bipolar disorder. This study assessed whether adjunctive maintenance treatment with risperidone long-acting therapy (RLAT), added to treatment-as-usual (TAU) medications for bipolar disorder, delays relapse in patients with bipolar disorder type I.

Methods:  This study included patients with bipolar disorder type I with ≥ four mood episodes in the 12 months prior to study entry. Following a 16-week, open-label stabilization phase with RLAT plus TAU, remitted patients entered a 52-week, double-blind, placebo-controlled, relapse-prevention phase. Randomized patients continued treatment with adjunctive RLAT (25–50 mg every two weeks) plus TAU (n = 65) or switched to adjunctive placebo injection plus TAU (n = 59). The primary outcome measure was time to relapse to any mood episode.

Results:  Of 240 enrolled patients, 124 entered double-blind treatment. Time to relapse was longer in patients receiving adjunctive RLAT (p = 0.010). Relapse rates were 23.1% (n = 15) with adjunctive RLAT versus 45.8% (n = 27) with adjunctive placebo; relative relapse risk was 2.3-fold higher with adjunctive placebo (p = 0.011). Completion rates were: adjunctive RLAT, 60.0% (n = 39) and adjunctive placebo, 42.4% (n = 25; p = 0.050). Adverse event (AE)-related discontinuations were 4.6% (n = 3) and 1.7% (n = 1), respectively. Common AEs (adjunctive RLAT versus adjunctive placebo) were: tremor (24.6% versus 10.2%), insomnia (20.0% versus 18.6%), muscle rigidity (12.3% versus 5.1%), weight increased (6.2% versus 1.7%), and hypokinesia (7.7% versus 0.0%).

Conclusions:  Adjunctive RLAT significantly delayed time to relapse in patients with bipolar disorder type I who relapse frequently. Safety and tolerability of RLAT were generally consistent with that previously observed.


Bipolar disorder is a chronic mental illness characterized by periodic episodes of mood disturbance (depressive, manic, mixed or hypomanic). It is associated with substantial psychosocial disability, high rates of substance abuse, a greater need for mental health care services and high risk of suicide (1, 2). The burden of bipolar disorder for individual patients can differ markedly, as episode frequency, severity, duration and time between episodes are highly variable. Some patients may experience long periods (four years or more) between episodes (3); others have a more insidious course, reflected by rapid cycling and frequent relapse (4).

Patients who relapse frequently are a difficult-to-treat population associated with a significant health care burden (4, 5). These patients are generally unresponsive to monotherapy with mood stabilizers such as lithium, carbamazepine or valproate (6–10). In the absence of effective monotherapy, combination therapy has become a standard strategy in the management of these patients (6, 10–12), but no consensus regarding their optimal treatment has been established (11). The APA Practice Guidelines acknowledge that current data do not support one combination therapy over another for these patients (13). Further, controlled relapse-prevention studies have not tested adjunctive treatment as it is used in clinical practice—with several concurrent medications for bipolar disorder.

A recent analysis of the Duke University Department of Psychiatry Clinical Management Research Information System (Durham, NC, USA) highlighted the significant public health challenge posed by bipolar patients who have four or more clinically significant events requiring intervention within a 12-month period. These patients utilized more emergency room (ER) and hospital resources, required more psychotropic and nonpsychotropic medications and had more comorbid conditions and higher rates of suicidality than patients who did not relapse as often (14).

For all patients with bipolar disorder, failure to adhere to pharmacologic treatment is common, with reported nonadherence and partial adherence rates ranging from 12% to 64% (1, 15–17). Nonadherence is the most important predictor of relapse and poor prognosis in patients with bipolar disorder; those who are partially adherent have significantly higher rates of psychiatric hospital admission than fully adherent patients (2, 18).

Patients who repeatedly relapse, and thus have poorer treatment response, worse long-term outcome and higher suicide risk (19), represent a group with significant unmet medical needs (19). This study presents the first large-scale controlled data for maintenance treatment with a long-acting antipsychotic therapy, risperidone long-acting therapy (RLAT), in patients with bipolar disorder who relapse frequently.


This randomized, double-blind (DB), placebo-controlled, international study was conducted from May 2004 to February 2007 at 32 psychiatric centers in the United States and India. The institutional review board or ethics committee at each site approved the protocol. The study was conducted in accordance with current International Conference on Harmonization–Good Clinical Practice guidelines and the Declaration of Helsinki (20).

Patients were enrolled with bipolar disorder type I or type II per Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) criteria (21). However, owing to the very low number of enrolled patients with bipolar disorder type II, this report focuses on the population with bipolar disorder type I. Eligible patients were between 18 and 70 years of age and had experienced four or more mood episodes (defined as an event requiring psychiatric intervention) in the past 12 months. Patients in any phase of bipolar illness (manic, hypomanic, depressed, mixed or euthymic) at study entry were included. Exclusion criteria were a DSM-IV-TR diagnosis of substance dependence; patients judged to be at imminent risk of injury to themselves or others; treatment with carbamazepine, oxcarbazepine, fluoxetine, paroxetine or clozapine within one month of open-label (OL) baseline; significant medical illness; or clinically significant abnormal laboratory or physical examination findings. There were no exclusions based on prior treatment nonresponse. Written informed consent was obtained from patients or their legal representatives before any trial procedures were performed.

Trial design

After a 16-week, OL stabilization phase, eligible patients were randomized into a 52-week, DB, relapse-prevention phase. The stabilization phase consisted of initiating, resuming or changing patients’ treatment-as-usual [(TAU); see below] medications for bipolar disorder at the discretion of the investigators. RLAT was initiated at 25 mg every 2 weeks, with optional dosage increases allowed to 37.5 mg at week 4 and up to 50 mg at week 10, per the investigators’ judgment. Oral antipsychotics that patients were taking prior to the study were continued for 3 weeks after the first RLAT injection (OL baseline). Patients not taking oral antipsychotics received oral risperidone for the first 3 weeks of the OL period.

TAU consisted of any number or combinations of antidepressants, mood stabilizers or anxiolytics, determined for each patient by his or her investigator. RLAT and TAU could be changed or adjusted at any time during the first 12 weeks of the stabilization phase but to be eligible for the DB phase, RLAT and TAU medications and dosages had to be stable for at least 4 weeks prior to randomization. For patients receiving lithium or valproate, plasma concentrations were obtained to assess compliance.

Patients who met the protocol-defined remission criteria (Fig. 1) for at least the last 4 weeks of the stabilization phase were eligible for the 52-week, DB, relapse-prevention phase. Patients were required to complete the entire 16-week OL phase before randomization to the DB phase, and none were randomized sooner than 16 weeks. TAU was continued unchanged and patients were randomized in a 1:1 ratio to continue their same dose RLAT or switch to placebo injection every 2 weeks. Randomization codes were assigned through an interactive voice response system, stratified by site. Placebo was identical in appearance to RLAT.

Figure 1.

 Study flow. RLAT = risperidone long-acting therapy; OL = open-label; TAU = treatment-as-usual; DB = double-blind; YMRS = Young Mania Rating Scale; MADRS = Montgomery-Åsberg Depression Rating Scale; CGI-BP-S = Clinical Global Impressions for Bipolar Disorder-Severity. aOL data for this population to be presented elsewhere. bRelapse determined by investigators, not necessarily the Relapse Monitoring Board.

During the DB relapse-prevention phase, short-term intermittent use of lorazepam 0.5 to 1.0 mg (no more than 3 doses per week and 10 doses per month), or an increase in dose no greater than 20% of the current dose of any drug used as part of the patient’s TAU, was permitted for agitation, irritability, restlessness or hostile behaviors. Study visits during the DB phase occurred every 2 weeks; however, if patients experienced worsening of their symptoms, unscheduled assessments were performed. First, supplemental oral risperidone (1–2 mg/day) was permitted for up to 3 consecutive days to treat worsening symptoms. If adequate relief was not achieved, the investigator could increase the dose of the blinded study medication, add a new psychotropic medication, or significantly (> 20%) increase the dose of one of the medications in the TAU regimen. However, any of these interventions may have led the patient to be considered relapsed if also meeting all other relapse criteria (described below).

Outcome measures

The primary endpoint was time to relapse from randomization in the DB relapse-prevention phase. The investigator considered a patient to have relapsed if the patient met DSM-IV-TR criteria for an acute mood episode in the setting of adequate compliance with oral TAU. Additionally, at least one of the following three conditions was satisfied:

  • (i) Clinical worsening, with the addition of a new mood stabilizer, antidepressant or antipsychotic or a > 20% dose increase of existing oral TAU medication, and meeting the following criteria:
    • (a) Young Mania Rating Scale (YMRS) (22) score > 15 or Montgomery-Åsberg Depression Rating Scale (MADRS) (23) score > 15 and
    • (b) Clinical Global Impressions of Bipolar Disorder-Severity (CGI-BP-S) (24) score ≥ 4 or Clinical Global Impressions of Bipolar Disorder-Change (CGI-BP-C) (24) score ≥ 6 or Global Assessment of Functioning (GAF) score decreased by > 10 points from baseline;
  • (ii) Hospitalization for worsening of manic or depressive symptoms and meeting the following criteria:
    • (a) YMRS score > 15 or MADRS score > 15 and
    • (b) CGI-BP-S score ≥ 4 or CGI-BP-C score ≥ 6 or GAF score decreased by > 10 points from baseline;
  • (iii) Hospitalization for worsening of manic or depressive symptoms and having significant suicidal ideation [e.g., InterSePT Scale for Suicidal Thinking (ISST) (25) revised score > 7].

For the primary analysis, relapse was determined by an independent relapse monitoring board (RMB). The RMB consisted of three international psychiatrists who reviewed, in a blinded fashion, cases of relapse reported by investigators to standardize the assessment of relapse and to support the trial’s scientific validity. Blinded data from investigator-determined relapses (YMRS, MADRS, CGI, GAF and suicidality assessments) were supplied to the RMB along with clinical narratives describing the patients’ symptoms, concomitant medications and the investigators’ determination of the reason for relapse. The RMB also reviewed data from all other patients in the DB phase to determine if any relapses had occurred that the investigators did not identify. The RMB also made a determination of the date of relapse on the basis of all blinded data provided by the investigators. The RMB decision was retrospective and unknown to the investigators, and thus did not impact the investigators’ decisions regarding patient care. Neither the individuals on the RMB nor anyone at their institutions were involved in the conduct of the trial. The RMB decision on relapse superseded the investigator’s decision on relapse for the analysis of the primary endpoint.

Clinicians were trained on rating scales and provided evidence of acceptable inter-rater reliability before the study began. To the extent possible, the same evaluator conducted the efficacy and safety rating scales at each visit. Efficacy rating scales included the YMRS, MADRS, CGI-BP-S and CGI-BP-C. The YMRS, MADRS and CGI-BP were conducted at weeks 0, 4, 8, 12, 14, 16, 22, 28, 34, 42, 48, 54, 60 and 68 (or endpoint).

Safety assessments included regular adverse event (AE) monitoring, clinical laboratory testing (hematology, biochemistry and urinalysis), vital signs, physical examinations, electrocardiograms (ECG), the ISST and movement disorder scale ratings [Simpson-Angus Scale (SAS) (26), the Barnes Akathisia Scale (BAS) (27), and the Abnormal Involuntary Movement Scale (AIMS) (28)]. Laboratory assessments and ECGs were performed at screening and at week 68 (or endpoint). Vital signs were assessed at each visit. A physical examination was performed at screening, week 16 and endpoint. The SAS, BAS and AIMS were performed at weeks 0, 16, 28, 42, 54 and 68 (or endpoint).

Statistical analysis

An intent-to-treat (ITT) population was used for all efficacy and safety analyses in each phase. The OL ITT population included all subjects who received at least one dose of RLAT. The DB, relapse-prevention, ITT population included all randomized patients who received at least one dose of study medication in the DB phase. Baseline for all endpoints in the OL stabilization phase was defined as the measurement taken at OL week 0. The assessment at OL week 16 (the last time point collected in the stabilization phase) was considered the baseline for the DB relapse-prevention phase. For both phases, the baseline for clinical laboratory tests and ECGs was the pretreatment measurement before the OL stabilization phase.

A minimum of 110 randomized patients were necessary to have a 90% chance of detecting a constant hazard ratio of 0.45, assuming that the probability of relapse over the duration of the study was 0.60, based on 2-sided log-rank test statistics and a 5% significance level. Based on an estimated number of dropouts and relapses, 276 patients were sought for this trial.

For the stabilization phase, demographic and baseline characteristics were summarized using descriptive statistics. Treatment differences were evaluated using the Cochran-Mantel-Haenszel (CMH) chi-square test or Fisher exact test for categorical variables and analysis of variance (ANOVA) to test for continuous variables with treatment and pooled centers as factors. The change from OL baseline for efficacy variables during the OL stabilization phase was tested using paired t-tests or signed-rank tests as appropriate.

The primary efficacy variable was the time to relapse to any mood episode, determined by the RMB, during the DB phase. The primary study null hypothesis for time to relapse stated that there was no difference between treatment groups in the distribution of time to relapse. Time to relapse was summarized using Kaplan–Meier (product limit) estimates. A between-group comparison was performed using a log-rank test, controlling for pooled centers.

In addition to the primary analysis, an unstratified log-rank test and a Cox proportional hazards model including treatment and country as factors were performed. Investigator-determined relapses were also evaluated. Differences with respect to the type of relapse episode (manic, depressive or mixed) were examined using Cox regression models. Between-group differences for continuous variables for change scores during the DB phase were assessed using analysis-of-covariance (ANCOVA) models with treatment and pooled center as the main effects and DB baseline score as the covariate, and the within-group differences were evaluated using a paired t-test. All group comparisons from ANOVA and ANCOVA models were based on type III sum of squares. Categorical variables were evaluated using the CMH test stratifying on pooled center or signed-rank tests as appropriate. With the exception of the primary endpoint, last-observation-carried-forward (LOCF) methodology was utilized for all secondary endpoints (YMRS, MADRS, CGI-BP-S and CGI-BP-C) for the week 68 endpoint. All statistical tests were 2-sided, and the nominal type I error was fixed at 0.05. All secondary analyses were supportive in nature. No key secondary endpoints were identified, and the objectives for the OL and DB phases were separate and specific to the phase. Therefore, no adjustments for multiplicity were performed.


A total of 271 patients with bipolar disorder type I were screened; 240 patients met the enrollment criteria and entered the OL stabilization phase (Fig. 1). Demographic and clinical characteristics of the OL population are listed in Table 1. At OL baseline, 8.8% of the patients were not prescribed or taking any medication for bipolar disorder, 30.8% were taking 1 medication, 39.6% were taking 2 medications, and 20.8% were taking ≥ 3 medications. Mood stabilizers were being taken by 85.4%, and antidepressants by 26.7% of patients. Oral antipsychotics were taken by 32.9% of patients prior to OL baseline. In the stabilization phase, the RLAT modal dose was 25 mg in 190 (79.2%) patients, 37.5 mg in 47 (19.6%) patients, and 50 mg in 3 (1.3%) patients.

Table 1.   Demographics and patient characteristics at baseline
 Open-label baseline (n = 240)Open-label baseline for patients who entered the double-blind relapse-prevention phase (n = 124)
Adjunctive RLAT (n = 65)Adjunctive placebo (n = 59)p
  1. p-values reflect pairwise comparisons of risperidone long-acting therapy (RLAT) and placebo groups.

Age, years    
 Mean (SD)38.4 (11.6)40.0 (11.8)37.6 (12.0)0.261
 Median [range]38 [18–70]41 [18–61]36 [18–63] 
Gender, n (%)   0.960
 Male144 (60.0)46 (70.8)43 (72.9)
 Female96 (40.0)19 (29.2)16 (27.1)
Race, n (%)   0.902
 Caucasian54 (22.5)6 (9.2)6 (10.2)
 Hispanic3 (1.3)1 (1.5)0 (0)
 Black23 (9.6)5 (7.7)3 (5.1)
 Other (Indian)158 (65.8)53 (81.5)50 (84.7)
Age at onset of bipolar diagnosis, years, mean (SD)26.4 (9.9)28.3 (9.4)24.6 (8.4)0.023
No. of mood episodes in the past 12 months    
 Mean (SD)6.5 (5.9)5.2 (1.4)5.2 (2.8)0.701
 Median [range]5.0 [4–60]5.0 [4– 9]4.0 [4–24] 
Most recent episode prior to study enrollment, n (%)   0.390
 Depressed70 (29.2)23 (35.4)14 (23.7)
 Manic131 (54.6)31 (47.7)37 (62.7)
 Mixed20 (8.3)5 (7.7)3 (5.1)
 Hypomanic19 (7.9)6 (9.2)5 (8.5)
Time since most recent episode, weeks, mean (SD)6.9 (5.0)7.1 (4.7)7.0 (4.9)0.951

At the end of the stabilization phase, 51.7% (n = 124) of patients with bipolar disorder type I entered the DB relapse-prevention phase (Fig. 1). The OL and DB populations at baseline differed in terms of the proportion of male patients and the racial/ethnic composition. The demographic and clinical characteristics in the DB population randomized to adjunctive RLAT or adjunctive placebo treatment were similar (between-group p values ≥ 0.05), except for age at which patients were diagnosed with bipolar disorder type I (Table 1).

TAU medications and dosages and doses of adjunctive RLAT or placebo could not change during DB treatment, and reflect those used to stabilize patients during at least the last 4 weeks of the stabilization phase. TAU medications were comparable between the two randomized groups (Table 2). Doses of adjunctive RLAT or placebo could not change during DB treatment. The RLAT modal dose during the DB phase was 25 mg in 44 patients (67.7%), 37.5 mg in 18 patients (27.7%), and 50 mg in 3 patients (4.6%) in the adjunctive RLAT group. In the DB phase, short-term use of lorazepam for agitation, irritability, restlessness or hostile behaviors was used by 9.2% of patients in the adjunctive RLAT group and by 18.6% in the adjunctive placebo group. Brief, intermittent use of oral risperidone for symptom worsening during the DB phase (as permitted by protocol) occurred in 33.8% of the adjunctive RLAT group and 37.3% of the adjunctive placebo group, for a mean duration of 4.3 (4.2) and 5.0 (3.5) days, respectively. Overall, study completion rates were 60.0% with adjunctive RLAT and 42.4% with adjunctive placebo (p = 0.050; Fig. 1).

Table 2.   Use of mood stabilizers, antidepressants and anxiolytics at double-blind baseline of the relapse-prevention phase
 Adjunctive RLAT (n = 65) n (%)Adjunctive placebo (n = 59) n (%)
  1. RLAT = risperidone long-acting therapy.

  2. aA patient taking > one prior medication within a class and subclass was counted once within the class and subclass. Classes include mood stabilizers, antidepressants and anxiolytics.

  3. bPercentages in the subclass do not equal the percentage of the class due to polypharmacy.

  4. cDivalproex, valproate acid, sodium and semisodium.

  5. dBupropion HCL, mirtazapine, trazodone.

Patients taking a number of bipolar medicationsa
 None4 (6.2)2 (3.4)
 1 medication31 (47.7)33 (55.9)
 2 medications22 (33.8)19 (32.2)
 ≥ 3 medications8 (12.3)5 (8.5)
Patients taking medications by number of bipolar medication classesa
 144 (67.7)47 (79.7)
 216 (24.6)9 (15.3)
 31 (1.5)1 (1.7)
Mood stabilizersb59 (90.8)56 (94.9)
 Valproic acid derivativesc42 (64.6)42 (71.2)
 Lithium and lithium carbonate33 (50.8)30 (50.8)
 Lamotrigine2 (3.1)2 (3.4)
 Topiramate1 (1.5)0 (0)
Antidepressantsb19 (29.2)11 (18.6)
 Selective serotonin reuptake inhibitors15 (23.1)6 (10.4)
 Serotonin norepinephrine reuptake inhibitors1 (1.5)1 (1.7)
 Tricyclic antidepressants2 (3.1)5 (8.5)
 Otherd4 (6.2)0 (0)
Anxiolyticsb16 (24.6)10 (17.0)
 Alprazolam3 (4.6)4 (6.8)
 Chlordiazepoxide1 (1.5)0 (0)
 Clonazepam3 (4.6)1 (1.7)
 Diazepam5 (7.7)2 (3.4)
 Lorazepam5 (7.7)3 (5.1)


OL stabilization phase.  Significant improvements were observed in the mean YMRS total score, MADRS total score and CGI-BP scores during this phase (all p values < 0.001, t-test) (Table 3).

Table 3.   Change in efficacy measures from open-label baseline to endpoint in the stabilization phase and from double-blind baseline to endpoint in the relapse-prevention phase
 Open-label stabilization phase (n = 240)Double-blind relapse-prevention phase (n = 124)
Adjunctive RLAT (n = 65)Adjunctive placebo (n = 59)
  1. Values listed as mean (SD).

  2. RLAT = risperidone long-acting therapy; YMRS = Young Mania Rating Scale; MADRS = Montgomery-Åsberg Depression Rating Scale; CGI-BP-S = Clinical Global Impressions for Bipolar Disorder-Severity; CGI-BP-C = Clinical Global Impressions for Bipolar Disorder-Change.

  3. ap ≤ 0.001 for change from open-label (OL) baseline to OL endpoint, paired t-test.

  4. bp ≤ 0.05 for change between groups.

  5. cp ≤ 0.001 for improvement at OL endpoint.

 Baseline14.2 (12.1)1.6 (2.8)1.4 (2.7)
 Change from baseline−9.7 (12.9)a1.7 (7.2)b7.6 (13.5)
 Baseline10.4 (10.3)2.1 (2.6)2.1 (3.2)
 Change from baseline−3.6 (11.2)a1.7 (7.0)2.5 (8.8)
 Baseline3.6 (1.2)1.4 (0.6)1.3 (0.6)
 Change from baseline−1.4 (1.6)a0.3 (1.3)b1.1 (1.9)
 Baseline2.9 (1.5)1.2 (0.5)1.2 (0.6)
 Change from baseline−1.3 (1.6)a0.2 (1.0)b0.9 (1.6)
 Baseline2.2 (1.4)1.2 (0.5)1.3 (0.6)
 Change from baseline−0.3 (1.5)a−0.2 (1.0)0.1 (1.4)
 Overall at endpoint2.5 (1.5)c4.1 (1.2)b4.6 (1.3)
 Mania at endpoint2.6 (1.4)c4.0 (1.0)b4.5 (1.2)
 Depression at endpoint3.5 (1.4)c4.0 (0.9)4.1 (0.9)

DB relapse-prevention phase.  Adjunctive RLAT treatment was associated with a significant delay in the time to relapse of any mood episode compared with adjunctive placebo treatment (p = 0.010, log-rank test) (Fig. 2). Relapse rates were 23.1% (n = 15) with adjunctive RLAT treatment and 45.8% (n = 27) with adjunctive placebo. The relative risk of relapse was 2.3-fold higher with adjunctive placebo compared with adjunctive RLAT [p = 0.011, chi-square (Cox regression)].

Figure 2.

 Kaplan–Meier estimation of the distribution of time to relapse based on independent Relapse Monitoring Board decision. RLAT = risperidone long-acting therapy; DB = double blind.

Overall, 19 patients in the total DB population relapsed to a depressive episode [adjunctive RLAT, n = 8 (12.3%); adjunctive placebo, n = 11 (18.6%)], 17 patients relapsed to a manic episode [adjunctive RLAT, n = 5 (7.7%); adjunctive placebo, n = 12 (20.3%)], and 6 relapsed to a mixed episode [adjunctive RLAT, n = 2 (3.1%); adjunctive placebo, n = 4 (6.8%)]. The overall observed risk ratio of relapse, 2.3, between treatment groups did not differ among relapse episode types [p = 0.581, chi-square (Cox regression, competing risks)]. Despite numerical differences in the most recent episode type prior to OL baseline in patients randomized to adjunctive RLAT, the time to relapse to any mood episode remained longer with adjunctive RLAT (p = 0.009) when controlled for the most recent episode type using a Cox regression model.

The RMB identified 42 relapses and the investigators identified 36 relapses. Relapse rates determined by the investigators were 20.0% (n = 13) with adjunctive RLAT and 39.0% (n = 23) with adjunctive placebo. Consistent with the primary analysis, time to relapse as determined by investigators also was significantly delayed (p = 0.024, log-rank test) with adjunctive RLAT. Mean YMRS, MADRS and CGI-BP-S scores during the DB phase are shown in Figure 3 and Table 3.

Figure 3.

 Change in Young Mania Rating Scale (YMRS), Montgomery-Åsberg Depression Rating Scale (MADRS), and Clinical Global Impressions for Bipolar Disorder-Severity (CGI-BP-S) total scores during the study in patients randomized to adjunctive risperidone long-acting therapy (RLAT) or adjunctive placebo treatment [last observation carried forward (LOCF) analysis from weeks 16 to 68]. DB BL = Week 16; baseline for double-blind (DB) relapse-prevention phase. Observed cases analysis from weeks 0 to 16. *p < 0.05; between-group differences in mean change from baseline using an ANCOVA model with effects for group, pooled center and baseline.


The most common treatment-emergent AEs in the DB phase were tremor, insomnia, muscle rigidity and mania (Table 4). Three patients (4.6%) in the adjunctive RLAT group and one (1.7%) in the adjunctive placebo group discontinued the study due to AEs. At least one serious AE occurred in 9 adjunctive RLAT-treated patients (13.8%) and 13 adjunctive placebo-treated patients (22.0%); most were related to underlying illness. Three deaths were reported during the trial: 1 in a patient randomized to adjunctive RLAT (cause of death was hypertensive heart disease, judged to be unrelated to study drug by the investigator) and 2 in patients randomized to adjunctive placebo treatment. One of the latter subjects died in a motor vehicle accident and the other subject relapsed and was subsequently treated with OL RLAT. He committed suicide approximately three months after leaving the study, which was approximately four weeks after the last RLAT dose.

Table 4.   Summary of treatment-emergent adverse events in ≥ 5% of subjects in either treatment group during the double-blind relapse-prevention phase
Adverse eventaTAU plus adjunctive RLAT (n = 65) n (%)TAU plus adjunctive placebo (n = 59) n (%)
  1. TAU = treatment-as-usual; RLAT= risperidone long-acting therapy.

  2. aMedDRA preferred terms.

Tremor16 (24.6)6 (10.2)
Insomnia13 (20.0)11 (18.6)
Muscle rigidity8 (12.3)3 (5.1)
Weight increased4 (6.2)1 (1.7)
Hypokinesia5 (7.7)0 (0)
Pyrexia3 (4.6)5 (8.5)
Akathisia3 (4.6)4 (6.8)
Upper respiratory tract infection 4 (6.2) 1 (1.7)
Sedation4 (6.2)0 (0)
Mania3 (4.6)8 (13.6)
Dizziness3 (4.6)4 (6.8)
Suicidal ideation1 (1.5)3 (5.1)
Headache0 (0)5 (8.5)
Depression0 (0)5 (8.5)
Fatigue0 (0)4 (5.1)
Nausea0 (0)3 (5.1)
Injury0 (0)3 (5.1)

No evidence of treatment-emergent suicidality was observed. Suicidal ideation was reported in 1 patient (1.5%) in the adjunctive RLAT group and in 3 patients (5.1%) in the adjunctive placebo group; a suicide attempt was reported in 1 patient in the adjunctive RLAT group. At DB baseline, ISST severity scores on scale items were low (between 0 and 1); endpoint changes in ISST score were minimal and not statistically significant between the treatment arms.

The proportion of patients with at least one extrapyramidal symptom (EPS)-related AE was 30.8% in the adjunctive RLAT group and 16.9% in the adjunctive placebo group. Severity of EPS, assessed by mean AIMS, BAS and SAS scores, was low in both treatment groups at DB baseline and remained low at endpoint, with no significant between-group differences in change scores.

Potential prolactin-related AEs were reported by 4 patients (6.2%) in the adjunctive RLAT group (1 patient, erectile dysfunction and libido decreased; 2 patients, amenorrhea; 1 patient, menstrual disorder) and 3 patients (5.1%) in the adjunctive placebo group (1 patient each: erectile dysfunction, loss of libido, amenorrhea). At DB baseline, mean [standard deviation (SD)] prolactin levels were 24.4 (29.1) ng/mL and 36.4 (52.2) ng/mL in the adjunctive RLAT and adjunctive placebo groups, respectively. The mean (SD) change from baseline to DB endpoint was 14.3 (36.1) ng/mL versus −20.9 (52.5) ng/mL, respectively (p < 0.001, ANCOVA).

Mean changes in other laboratory values and vital signs were similar between groups, including those for glucose, cholesterol and triglycerides. At OL baseline, the mean (SD) weight was 67.5 (18.2) kg; the least squares (LS) mean (SE) change at OL endpoint was 2.6 (0.5) kg. An AE of ‘weight increased’ was reported by 6.2% of patients in the RLAT plus TAU group and by 1.7% in the adjunctive placebo group. At DB baseline, the mean (SD) weight was 69.8 (17.5) kg in the RLAT plus TAU group and 70.3 (19.9) kg in the adjunctive placebo group. At endpoint, there was a mean increase of 0.7 (5.3) kg with adjunctive RLAT and a decrease of 2.0 (4.5) kg with adjunctive placebo (p = 0.002, ANCOVA). The percentage of patients with an increase ≥ 7% in body weight at endpoint was 28.1% with adjunctive RLAT and 31.0% with adjunctive placebo. There was no significant between-group difference in ECG data, including QTc intervals, for the adjunctive RLAT and the adjunctive placebo groups.


This is the first large, multisite, controlled study to examine the role of a long-acting atypical antipsychotic in the maintenance treatment of patients with bipolar disorder type I. During the stabilization phase, treatment with RLAT plus individualized TAU was associated with significant improvements in mood symptoms and overall clinical status. During the DB relapse-prevention phase, adjunctive treatment with RLAT significantly delayed time to relapse to a mood episode compared with adjunctive placebo and generally maintained the clinical improvement attained during the stabilization phase. The number needed to treat (NNT), based on the percentage of relapsed patients and the treatment duration of this study, is 4 (95% confidence limit: 3–12), suggesting that for every 4 patients receiving RLAT plus TAU, relapse will be delayed in at least 1 patient. This NNT compares favorably with a range of treatments for other serious illnesses (29–34).

Although AEs may reflect, in part, the safety profile of TAU medications in addition to RLAT, results suggest that adjunctive RLAT treatment was generally well tolerated, with a safety profile consistent with that previously observed in short- and long-term studies in patients with schizophrenia or schizoaffective disorder treated with RLAT, without emergence of new safety issues (35–38). AE rates that were higher with adjunctive RLAT versus placebo included tremor, muscle rigidity, hypokinesia and increased weight. AEs higher with placebo were generally related to underlying illness: mania, depression and fatigue. During DB treatment, the mean prolactin level increased with adjunctive RLAT but decreased with adjunctive placebo. However, a similar percentage of patients reported AEs potentially related to prolactin: adjunctive RLAT (6.2%) and adjunctive placebo (5.1%).

Although patients with bipolar disorder type I and II were enrolled, only 35 (12.7%) had bipolar disorder type II; 15 (10.8%) entered the DB phase. Therefore, results for patients with type I only were reported here. Full data results may be found on clinicaltrials.gov; identifier: NCT00094926. The exclusion of these patients does not affect the overall study results or their interpretation.

Other clinical data support the efficacy of risperidone in patients with mood disorders. Oral risperidone, alone or as adjunctive treatment, is efficacious for the treatment of bipolar mania (39–43). The benefits observed in this study may reflect both the efficacy of risperidone for mood symptoms and the benefits of assured medication delivery in a population that is often nonadherent with their medication. Commonly cited advantages for long-acting injectable medications are improved adherence and more rapid detection of nonadherence (44).

The population in this study differed from that described by the DSM-IV-TR course specifier for rapid cycling, which applies to those bipolar patients whose “episodes are demarcated either by partial or full remission for at least two months or a switch to an episode of opposite polarity (e.g., major depressive episode to manic episode)” (21). Thus, the studied population is inclusive and overlaps with patients with a rapid cycling course, but cannot, however, be assumed to be identical with those patients. The rapid-cycling specifier raises several difficulties. For example, intensive monitoring is required to precisely determine the periods between episodes and clear demarcation of episodes may be lacking, or the periods between episodes may be less than two months. Some of the most severely ill patients (e.g., those who experience ultra-rapid or ultradian cycling) would not fulfill these stringent criteria, yet they may be patients of particular interest in this category. The population in this study, which was not required to meet the requirement for a ≥ 2-month interval of remission between episodes, is more easily identifiable by clinicians and represents a broader spectrum of patients who relapse frequently, a population for whom safe and effective maintenance treatment remains a significant unmet medical need. Studies have demonstrated that patients with bipolar disorder requiring intervention for four or more mood episodes in the previous 12 months have greater morbidity, an elevated rate of substance abuse and higher rates of nonresponse with traditional mood-stabilizing drugs than patients who experience fewer relapses (4, 45). This study suggests that adding RLAT to maintenance treatment may be beneficial in patients who relapse often.

Use of TAU was individualized, and patients could enter the study in any phase or with any degree of symptoms, aspects of the study design that reflect the management of patients in clinical practice. Prior treatment nonresponse, including nonresponse to oral risperidone, did not exclude patients from this study. This approach was used to provide more generalizable results and avoid potential bias in favor of RLAT.

In this study, the OL stabilization phase was 4 months long, and the required period of sustained remission prior to randomization was 4 weeks. These time periods were chosen because this patient population relapses not less than once every 3 months (on average), and the cycles of mood episodes are typically less than 3 months. Therefore, an OL treatment period of 4 months was sufficient to establish the treatment response and meet the objectives of OL stabilization, which were to stabilize the patients’ mood disorders and induce remission of mood symptoms with adjunctive RLAT. Patients who did not remit by week 16 and who continued OL RLAT generally remitted over the course of a separate 36-week, OL extension phase (results to be reported elsewhere).

Time to relapse and relapse rates were calculated for both investigator and RMB-determined relapses. The primary analysis, however, was based on RMB-determined relapse. The RMB, a panel of international experts with extensive academic and clinical experience, was used to reduce potential site-based bias and diagnostic idiosyncracies and to support the scientific validity and generalizability of this trial. The significant results (delaying time to any mood episode) were similar for relapses and dates of relapse specified by either group. Of the 42 investigator-determined relapses, 2 were judged by the RMB not to be relapses and therefore were not counted as relapse events in the calculation of the primary analysis. Conversely, 8 patients not identified by the investigators as having relapsed were determined to have met relapse criteria by the RMB. Difference in relapse determination between investigators and RMB may reflect the differences in the primary function of the two entities. Investigators are guided by dictates of patient care, which may be influenced by acute needs or social issues, whereas the RMB provided a determination based on the threshold criteria specified in the protocol, with a longitudinal perspective on clinical relevance.

Several limitations should be noted. Although this study demonstrated a delay in time to relapse into any mood episode during treatment with RLAT, this study was not powered to conclusively demonstrate prevention of particular types of mood episodes (mania, depression, mixed states). Also, the study was not stratified by mood episode type immediately prior to entry to evaluate the relative benefit of RLAT on the most recent episode type. TAU also was not stratified; however, the use of TAU was generally similar between treatment groups. This study does not demonstrate what, if any, proportion of the benefit was due solely to the delivery of treatment via long-acting formulation, as there was no active oral adjunctive antipsychotic treatment comparison group. However, in a population known to exhibit low adherence with medications (including oral antipsychotics), and in whom low adherence is associated with poor outcomes, likely compliance benefits associated with a long-acting injectable formulation may play a role in the overall treatment effect. Moreover, 32.9% of patients entering the trial had been prescribed oral antipsychotic at baseline but still experienced at least four episodes in the past 12 months, indicating that oral formulations were not effective for these patients. Finally, lower rates of substance use and comorbidity were reported by patients and investigators than have been previously reported for patients with bipolar disorder (1, 2). This may in part reflect cultural influences on the incidence and reporting of other psychiatric symptoms and drug or alcohol use.

This is the first large, multisite, controlled trial to study a long-acting atypical antipsychotic in patients with bipolar disorder type I and the first in patients who relapse frequently. Adding maintenance RLAT to TAU delayed mood episode relapses in these patients and was generally well tolerated. The risk:benefit ratio for adjunctive RLAT should be evaluated individually and may not be advantageous for all patients with bipolar disorder. In consideration of the overall illness burden associated with bipolar disorder type I, however, adjunctive RLAT may be beneficial for many, especially those who experience frequent relapses.


The authors wish to thank the study investigators and the Relapse Monitoring Board.

Relapse Monitoring Board (RMB)

RMB Chairperson: Ravi Anand, MD, Anand Pharma Consulting, St. Moritz, Switzerland; Sumant Khanna, MD, PhD, MAMS, MRCPsych, Consultant Psychiatrist and Consultant on Clinical Trials, New Delhi, India; Mark Hyman Rapaport, MD, Department of Psychiatry and Behavioral Neurosciences at Cedars-Sinai Medical Center and Department of Psychiatry at the David Geffen School of Medicine at UCLA, Los Angeles, CA, USA.

Study Investigators

Caleb Adler: Psychiatric Professional Services, Inc., Cincinnati, OH, USA; Mohammed Bari: Synergy Clinical Research Center, National City, CA, USA; Jason Baron: MedLabs Research of Houston, Inc., Houston, TX, USA; Ronald Brenner: Neurobehavioral Research, Inc., Cedarhurst, NY, USA; Shiv Gautam: SMS Hospital, Government Medical College, Jaipur, Raj, India; Ramanath Gopalan: Comprehensive NeuroScience of N. Virginia, Arlington, VA, USA; Howard Hassman: CNS Research Institute, Clementon, NJ, USA; Ram Jiloha: G.B. Pant Hospital, JN Marg, Delhi, Del, India; Arifulla Khan: Northwest Clinical Research Center, Bellevue, WA, USA; Arun Kishore: Government Medical College & Chest Hospital, Thrissur, Kerala, India; Mark Lerman: Alexian Center for Psychiatric Research, Hoffman Estates, IL, USA; Xiaohua Li: University of Alabama at Birmingham, Birmingham, AL, USA; Adam Lowy: Comprehensive NeuroScience, Inc., Washington, DC, USA; Morteza Marandi: Comprehensive NeuroScience, Cerritos, CA, USA; Rick Mofsen: Clinical Research, Inc., St. Louis, MO, USA; S. Nambi: Institute of Mental Health, Kilpauk, Chennai, TN, India; Michael Plopper: Sharp Mesa Vista Hospital, San Diego, CA, USA; Steven Potkin: UCI Medical-Neuropsychiatric Center, Orange, CA, USA; G. Prasad Rao: Asha Hospital, Institute of Medical Psychology, Banjara Hills, Hyderabad, AP, India; Janardhan Reddy: NIMHANS, Wilson Gardens, Bangalore, Kar, India; Alok Sarin: VIMHANS, Nehru Nagar, Delhi, Del, India; R. Sathianathen: Madras Medical College, Periyar EVER Sallai, Park Town, Chennai, TN, India; P.S.V.N. Sharma: Kasturba Medical College & Hospital, Manipal, Kar, India; Mikel Thomas: CTT Research, Inc., Prairie Village, KS, USA; J.K. Trivedi: King George’s Medical College, Lucknow, UP, India; Vihang Vahia: Dr. RN Cooper Hospital & GS Medical College, Juhu, Vileparle, Mumbai, Mah, India; K. Srinivasan: St. John’s Medical College Hospital, Bangalore, Kar, India; T.P. Sudhakar: SV Medical College, Tirupati, AP, India; Deborah Yurgelun-Todd: McLean Hospital, Belmont, MA, USA.

The authors also wish to acknowledge Georges Gharabawi, MD, for his assistance in the design of the study; and the writing and editing assistance provided by Mariana Ovnic, PhD, and Helix Medical Communications (funding supported by Ortho-McNeil Janssen Scientific Affairs, LLC, Titusville, NJ, USA) in the development and submission of this manuscript.