Changing the Antipsychotic in Early Nonimprovers to Amisulpride or Olanzapine: Randomized, Double-Blind Trial in Patients With Schizophrenia

Abstract

Introduction

Many of the more than 18 000 coded studies in the Cochrane Schizophrenia Group (CSzG) specialized register are on antipsychotic drugs. Nevertheless, the vast majority of these trials are concerned with the general efficacy of these agents. Very few prospective, sequential studies are available that guide decisions, which have to be made in clinical routine. One of the questions that clinicians pose to themselves every day is “If a patient has not responded to the first antipsychotic tried, is switching to another one effective?” While treatment guidelines had stated for a long time that an adequate treatment trial should last at least 4–8 weeks,^{1, 2} evidence has accumulated in the last decade that if patients have not shown any clinically significant improvement after 2 weeks, the majority are unlikely to respond if they are kept on the same antipsychotic.³ But the unresolved question is whether switching the antipsychotic is likely to help or whether early nonimprovers just have a generally unfavorable prognosis irrespective of the drug used.

Concerning previous publications, six small, mainly negative studies were clearly underpowered (14–58 participants randomized) and therefore had only pilot character.^4-8 Hatta et al.⁹ randomized nonresponders to risperidone to either continuation of risperidone or switching to its metabolite paliperidone which has, in essence, the same receptor binding profile so that not much of an efficacy advantage could be expected. In the OPTIMiSE study, patients with schizophrenia who had been treated for four weeks with amisulpride did not benefit from a switch to olanzapine. However, the trial was restricted to first-episode patients with psychosis who responded so well to antipsychotic drugs that there could have been ceiling effects.^{10, 11} In a post-hoc analysis of the CATIE study, patients who had received perphenazine in Phase I were more likely to drop out in Phase II if they were rerandomized to risperidone rather than to the other antipsychotics.¹² This finding was in part explained by some similarities perphenazine and risperidone share in their receptor binding profiles. Finally, in the trial by Kinon et al. (n = 378) there was a statistically significant advantage of switching patients who had not improved after 2 weeks of treatment with risperidone to olanzapine, ¹³ but there were limitations. First, as all patients were initially treated with risperidone before the nonimprovers were randomized to either staying on risperidone or switching to olanzapine, it is not clear whether the switch to a different antipsychotic was effective or whether olanzapine was just a generally more efficacious drug. Finally, olanzapine and risperidone are relatively similar drugs, because both are dopamine D₂/ serotonin 5-HT_2A receptor antagonists. ^{14, 15} Theoretically, it might be a more effective strategy to switch to a drug with a more different receptor binding profile, with the idea that patients who have not responded to one mechanism of action might respond to another one. Therefore, in the current trial, we chose the D₂/5-HT_2A antagonist olanzapine and the selective D₂/D₃ and 5-HT₇ antagonist amisulpride¹⁶ as the drugs between which patients were switched.

Method

Detailed information about the study methodology is described in a separate publication.¹⁷

Patients

Patients aged 18–65 years meeting DSM IV criteria for schizophrenia, schizoaffective disorder or schizophreniform disorder as confirmed by the Mini-International-Neuropsychiatric-Interview-Plus were recruited at 17 German and 15 Romanian sites. To assure that patients had an acute exacerbation the PANSS total score was ≥75 points and at least two items of the PANSS positive component factor cluster“ (P1, G9, P3, P6, P5, G12) were at least 4 (moderate) at both screening and baseline; the CGI-S was ≥4 at inclusion. Moreover, an increase in the level of care (admission to inpatient or daycare, admission from daycare to inpatient care) had to have taken place within 7 days prior to enrollment. Exclusion criteria included the following: documented non-response to a 6–8 week treatment attempt in the past or continuous and compliant treatment with one of the study drugs in the two weeks prior to study entry (olanzapine 10–20mg/d, amisulpride 400–800mg/d), a ≥25% reduction in PANSS-total-score from screening to baseline, intolerance or contraindications to one of the study drugs, imminent risk of suicide or harm for others, depot antipsychotic treatment within one injection cycle, pregnancy, substance dependency (excluding nicotine and caffeine), relevant somatic diseases or abnormal laboratory results. All patients had given written informed consent, and the protocol was approved by respective ethical committees and registered on ClinicalTrials.gov (NCT01029769).

Procedures

Figure 1 illustrates the study design. After a screening phase of up to three days, in Phase I patients were randomized in a 1:1 ratio to 14 days of double-blind treatment (capsules identical in taste, color, and shape) with either amisulpride or olanzapine. Randomization lists using stratification by center and a block size of four were generated with Rancode Professional 3.6 by the “Institut für Medizinische Statistik und Epidemiologie (IMSE)” of the Technical University of Munich. To safeguard allocation concealment investigators called an independent person which held the pergenerated randomization list and allocated a medication kit after a participant was recorded. Doses were titrated within three days to amisulpride 600–800 mg/day and olanzapine 15–20 mg/day to assure that all patients had reached an effective dose. Thereafter, amisulpride and olanzapine doses could be adjusted between 200–800mg/day and 5–20mg/day, respectively, to counter side-effects. Patients who had at least 25% improvement of the PANSS baseline score at the end of Phase I (day 14, so called “early-improvers”) continued double-blind treatment with the same compound in Phase II which lasted 6 weeks. Those patients who had not reached this degree of improvement were rerandomized double-blind to either staying on the same drug or switching to the other drug with the same dose ranges as in Phase I. For this purpose drugs were cross-tapered during week 3. The cutoff <25% PANSS reduction reflects the results of a diagnostic test review demonstrating that <20%–25% PANSS total score reduction from baseline was the optimum predictor of a lack of later symptomatic remission if patients stay on the same antipsychotic.³ At the end of Phase II all participants stopped study drug intake and were observed for another 4 weeks in a safety follow-up. As concomitant medication for agitation, we allowed lorazepam up to 6mg/d) and diazepam up to 60mg/d, for sleep disturbances zolpidem up to 10mg/d, lormetazepam up to 2mg/d, zopiclone up to 7.5 mg/d and temazepam up to 20mg/d, and biperiden up to 6mg/d for extrapyramidal side-effects. Antidepressants and mood-stabilizers could be continued with the same dose if they had been initiated at least four weeks before study start.

Fig. 1.

Study design.

Open in new tabDownload slide

Outcomes

The primary outcome was symptomatic remission at eight weeks according to Andreasen et al.¹⁸ Secondary outcomes were the mean reduction from baseline to endpoint of the PANSS and its subscores, the Clinical Global Impression Scale,¹⁹ the Subjective Well-Being under Neuroleptics Scale²⁰ and the RSM-Scale²¹ for quality of life, the Drug Attitude Scale (DAI-10) as a measure of satisfaction with treatment²², the Personal and Social Performance Scale“ (PSP) ²³ for social functioning, the Simpson-Angus-Scale“ (SAS) ²⁴ and the Barnes-Akathisia-Scale“ (BAS) ²⁵ for extrapyramidal side-effects, and weight gain. A pharmacogenetic analysis will be presented in future reports.

Statistical Analysis

For all efficacy outcomes, an intention to treat approach including all patients who entered Phase II was taken. To compare the primary outcome between the switch and nonswitch groups, a logistic regression model with “remission” as the dependent variable and “switch” of treatment (yes/no) and PANSS-total score at visit 3 as independent variables was used. Multiple imputation (MI) (based upon 20 imputations, primary analysis), last observation carried forward and completers-only analyses were performed. Multiple imputation was performed separately for the switch and nonswitch arms and the imputation model included the PANSS total score from all visits from Phase II baseline (visit 2) onwards, remission at visit 7 and Phase I arm allocation. In addition, a perprotocol analysis, excluding patients with major protocol violations was performed. To compare the quantitative secondary outcomes between the switch and nonswitch groups, mixed models for repeated measurements were fitted to the outcome data recorded at the visits during Phase II. The difference in the frequency of dropouts, adverse events, benzodiazepine and anticholinergic medication use between switchers and stayers was examined with Chi-square tests. The study was designed to assess the switching treatment strategy, but frequencies and means of early improvers are presented in the online appendix. The relative efficacy of amisulpride and olanzapine was also not assessed, except for the question with which drug switchers and stayers were treated in Phase I.

Results

Patient Characteristics at the Beginning of Phases I and II

A total of 355 patients were enrolled. Five patients withdrew their consent and one did not fulfill the inclusion criteria. Three hundred and forty-nine patients were randomized in study phase I. Twenty-two patients of one site which violated GCP in other trials were excluded (a sensitivity analysis including this site did not lead to notably different results, see Appendix). Baseline characteristics are shown in table 1. On average patients were markedly ill (mean CGI of 5.2, PANSS total score of 99.8), had a paranoid form of schizophrenia, had been ill for 10.6 years, were aged 39.8 years, 84% lived independently, 24% were married or had a partner, 39% had a regular occupation, and 47% were women.

A total of 42 (12%) of patients dropped out during Phase I (see figure 3). Of 282 patients entering Phase II 142 (50%) patients were “early non-improvers” (≤ 25% PANSS total score reduction from baseline). Baseline characteristics at the stage of the second randomization are summarized in table 2 ().

Fig. 2.

Percentages of subjects in remission at different time points and study endpoint.

Open in new tabDownload slide

Fig. 3.

CONSORT diagram.

Open in new tabDownload slide

Treatment Effects During Phase II

Primary Outcome Significantly more “switchers” (41 of 60 completers, 68%) than “stayers” (25 of 55 completers, 46%) were in symptomatic remission at endpoint (multiple imputation results: OR 3.01, 95% CI (1.35, 6.72), p = .01, see figure 2 and table 3). Consistent results were obtained in the perprotocol analysis (OR 3.18 (1.38, 7.35), p = .01, 40 of 59 (68%) switchers versus 25 of 53 (47%) stayers) and the last-observation-carried-forward analysis (OR 3.17 [1.55, 6.49]), p = .002), 43 of 70 (61%) switchers versus 27 of 72 (38%) stayers). Table 2 shows that 14 of 142 (10%) early nonimprovers were already in remission at week 2. Excluding these patients in a post-hoc sensitivity analysis did not notably change the results of the primary analysis (35 of 54 (65%) of the “switcher” completers were in remission versus 19 of 49 (39%) of “stayer” completers, multiple imputation OR 3.26 (1.41, 7.55), p = .01). There was no significant interaction between switch/staying and the country in which patients were recruited and switch/staying (p = .95) or baseline PANSS score (p = .22). Additional analysis showed no interaction between initial treatment (ie, starting with olanzapine or amisulpride) and switch/stay (p = .85) suggesting that the effectiveness of switching the antipsychotic does not depend on the actual drug the patients were initially assigned in the study phase I.

Secondary Efficacy Outcomes “Switchers” tended to improve more than “stayers” in most efficacy-related outcomes, but this was not significant, except for the PANSS negative subscore (p = .04) and the CGIs (p = .02, table 3). There were also no significant differences between “switchers” and “stayers” in the number of patients dropping out for any reason, side-effects and inefficacy (Chi-square tests, all >0.05).

At the end of eight weeks of treatment 73.6% of the early improvers met criteria for symptomatic remission (please see online supplement table 11 for additional outcomes of early improvers).

Medication use There was no clear difference between stayers and switchers in mean olanzapine and amisulpride doses, benzodiazepine doses (with and without z-drugs) and concomitant anticholinergic use at the end of Phase II (table 3).

Side-effects There was one suicide that occurred more than 4 weeks after the patient had completed the study (last study drug amisulpride), and one death in Phase I of a patient randomized to olanzapine after the patient had dropped out. Both deaths were considered unrelated to the study drugs. A summary of the other 21 serious adverse events (SAEs) is presented in the supplement.

There were no significant differences between “switchers” and “stayers” in terms of the Barnes Akathisia scale and Simpson Angus scale in Phase II (table 3). Single adverse events that occurred in Phase II with a frequency of at least 5% in either group are presented in table 4.

Discussion

Patients with schizophrenia who had less than a 25% reduction of the PANSS from baseline after two weeks of treatment with amisulpride or olanzapine were more likely to attain symptomatic remission after another six weeks of antipsychotic treatment if they switched to the other drug. This result was independent of the initial drug used, and the switching strategy was safe. Differences in secondary efficacy outcomes were less clear.

A strength of the study are the two randomizations within two weeks, a design that has been rarely used in psychiatry. The initial randomization to amisulpride or olanzapine ruled out that any superiority of later switching was due to a general superiority of the drug patients were switched to rather than the switching process itself. This was a limitation of the study by Kinon et al. ¹³ in which all patients had initially received risperidone. Their design did not rule out that olanzapine – to which half of the early nonresponders were randomized - was just a generally more efficacious drug. In contrast to Kinon et al.,¹³ who used the PANSS total score change as the primary outcome, we used symptomatic remission²⁶ which is a clinically more interpretable outcome. Despite recent efforts to understand the clinical meaning of the PANSS total score,^{27, 28} this scale remains difficult to translate to practice. Multiple analyses have shown that symptomatic remission, defined as maximally mild presence of eight core symptoms of schizophrenia,²⁶ has clinical validity (for review see Lambert et al.²⁹).

Given the high drop-out rates in current antipsychotic drug trials, which increase over time³⁰ and cause major problems for statistical analysis, such a switching design could also be used in drug development.³¹ Indeed, one study showed that patients who did not respond to two weeks of treatment with 80mg/day lurasidone benefitted from an increase to 160mg/day compared to staying on 80mg/day.³² Finally, in contrast to a number of mainly negative pilot trials^4-8 our study was comparably large and it was industry independent (Eli Lilly donated olanzapine medication but had no influence on the design or interpretation of the study).

With amisulpride and olanzapine, we chose two agents with very different mechanisms of action: a selective D₂/D₃, 5-HT₇ receptor antagonist^{16, 33} and a combined D₂/5HT_2A receptor antagonist,^{14, 15} respectively. In theory, switching patients who have not improved to a drug with a different receptor binding profile makes more sense than to a drug with a similar profile. The assumption is that if one mechanism of action was not effective for a given patient, the patient might benefit from a drug that uses a different mechanism of action. Having this said we must nevertheless consider that the shared mechanism of antipsychotic action may derive from the D2-antagonism. The relevance of the other receptor binding activities for overall antipsychotic efficacy of the two molecules is not fully understood yet. Still the “Clinical Trials of Antipsychotic Effectiveness” (CATIE) study provides some more evidence supporting this rationale: patients who had discontinued perphenazine in CATIE Phase I and were then rerandomized to risperidone in Phase II were less likely to continue the study than those who were randomized to quetiapine or olanzapine.¹² This effect was explained by the similar receptor binding profiles of perphenazine and risperidone. In case a drug with a mode of antipsychotic action clearly distinctive from D2-antagonism became available, switching between two receptor-wise nonoverlapping antipsychotic treatment strategies may become a target for a future trial.

The switching strategy appeared to be as safe as keeping patients on their initial compound, because there were no significant differences in side-effects and dropouts between the two groups (except more insomnia among switchers, p = .03). Studies have not found major differences between different cross-titration strategies,³⁴ and the chosen cross-over period of eight days is probably common in clinical practice.

The trial is not without limitations. The most important issue is that as figure 2 shows the switchers tended to slightly worsen on average during week 3 and outperformed the stayers only at the last visit (week 8). While it is plausible that switching may first lead to a certain destabilization, only a longer study could demonstrate whether the superiority at week 8 would persist, vanish or increase beyond this duration. Also, patients with poor response to antipsychotic treatment in the past were not excluded by definition from study participation. Instead, we demanded that an increase in the level of care (admission to inpatient or day care, admission from day care to inpatient care) had to have taken place within 7 days prior to enrollment. This would translate into some dynamic in the disease course and minimize recruitment of poor responders. Fortunately, the overall good response rate in phase 1 is unlikely to reflect a high share of patients considered as treatment resistant. Of note, our trial should by no means advocate a further delay of switching to clozapine in case of overt treatment-refractoriness. In addition, our trial does not have the statistical power to encourage switching in first-episode patients. In fact, the OPTIMSE trial cited in the introduction did not support a switching strategy in first-episode patients poorly responding to the initial treatment. ¹⁰ The superiority in remission was not corroborated by significant differences in secondary efficacy related outcomes, except the CGI and the PANSS negative subscore, making the results of the study inconclusive. It should be noted, however, that the remission items of the PANSS were not rated separately from the other PANSS items making a rater bias unlikely, although incomplete blinding due to guessing from the different side-effect profiles of amisulpride and olanzapine cannot be fully ruled out. Although the remission items are part of the PANSS, remission is a different outcome than the PANSS total score. Statistical power issues can also have played a role. The estimated 4 points difference in PANSS total score reduction between switchers and stayers is about the same as that in the study by Kinon et al.¹³ (3.5 points) where it was statistically significant, possibly due to the larger sample size (628 patients enrolled).

We conclude that patients who have not even minimally improved to a two weeks treatment with a dose at the upper limit of the officially licensed dose range of an antipsychotic may more frequently be in remission after an additional six weeks if they are switched to a drug with a different mechanism of action. This switching strategy appears to be safe. However, as a clear superiority was present only at 8 weeks and as most secondary efficacy outcomes did not corroborate the primary outcome remission, an additional longer study is needed.

Acknowledgment

We thank the “Münchner Studienzentrum” and Tangent Data for independent monitoring and database management. In the last three years Stephan Heres has received speaker honoraria from Johnson & Johnson and Otsuka/Lundbeck; services as consultant in clinical trials were provided for the companies TEVA, ROVI and KYE. Stefan Leucht honoraria as a consultant and/or advisor and/or for lectures from Alkermes, Angelini, Eisai, Gedeon Richter, Janssen, Lundbeck, Lundbeck Institute, Merck Sharpp and Dome, Otsuka, Recordati, Rovi, Sanofi Aventis, TEVA, Medichem, Mitsubishi; Berthold Langguth received consulting honoraria and speakers’ fees from ANM, AstraZeneca, Autifony Therapeutics, Decibel Therapeutics, Desyncra, Frequency Therapeutics, Lundbeck, Merz, MagVenture, Neurolite, Neuromod, Novartis, Pfizer, Schwabe and Servier; and travel and accommodation payments from Eli Lilly , Lundbeck, Servier, and Pfizer; Valentin Matei honoraria from Servier Pharma, Angelini Pharmac Inc., Janssen Pharmaceutica NV, Hofmann-La Roche, Astra Zeneca, GlaxoSmithKline, Teva; Richard Musil speakers’ fees from Otsuka Pharmaceuticals and Lundbeck, and research funding as principal investigator in clinical trials for Böhringer-Ingelheim, Otsuka Pharmaceuticals, Teva, Oryzon, Novartis and Emalex Biosciences; Claus Wolff-Menzler honoraria from Janssen und LivaNova, Ion Papava lecture honoraria from Terapia S.A., Johnson & Johnson, Ever Neuropharma GMBH, S.C. Eli Lilly, Angelini Pharmaceuticals, travel or hospitality payment from Johnson & Johnson, Janssen-Cilag, Bristol-Myers-Squibb, Gerot Lannach, Angelini Pharmaceuticals, Lundbeck and Terapia S.A; Liana Dehelean has received lecture honoraria from Janssen-Cilag, Johnson & Johnson, Eli Lilly and company, Astra Zeneca, Servier Pharma, Lundbeck A/S, Bristol-Myers-Squibb and Angelini Pharmaceuticals. Dr. Liana Dehelean has accepted travel or hospitality payment from Janssen-Cilag, Johnson & Johnson, Astra Zeneca, Eli Lilly and company, H. Lundbeck A/S, Sanofi-Aventis, Servier Pharma, Bristol-Myers-Squibb, Novartis, Pfizer Inc, Gedeon Richter Plc, and Angelini Pharmaceuticals, Liana Dehelean participated in clinical trials sponsored or supported by Eli Lilly and company, Janssen Cilag, Johnson & Johnson, Servier Pharma, and Pfizer Inc.

Funding

Supported by a grant from the “German Federal Ministry for Education and Research (Bundesministerium für Bildung und Forschung, BMBF, grant number 01KG0910). Eli Lilly provided olanzapine medication. The funders had no influence on the protocol, data analysis, publication or other aspects of the trial.

References