We describe the pharmacological treatment of schizophrenia and have arranged the manuscript as a simple algorithm which starts from the choice of an antipsychotic drug for an acutely ill patient and concludes with the most important questions about maintenance treatment. In acutely ill patients the choice of drug is mainly based on pragmatic criteria. Among many strategies used for agitated patients, haloperidol plus promethazine is the best examined one. In case of persistent depression or negative symptoms treatment includes antidepressants, and some second-generation antipsychotic drugs (SGAs) have been found somewhat superior to first-generation antipsychotic drugs (FGAs) in these domains. If an antipsychotic is suspected to be ineffective, several factors need to be checked before action is taken. Few trials have addressed strategies such as switching the drug or increasing the dose in case of non-response. Clozapine remains the gold-standard for treatment-refractory patients, while none of the numerous augmentation strategies that have been examined by randomized controlled trials can be generally recommended. Maintenance treatment with antipsychotic drugs effectively reduces relapse rates. Small, not definitive, studies have shown that withdrawing antipsychotics from patients who have been stable for up to 6 yr leads to more relapses than continuing medication. In effect, continuous treatment is more effective than intermittent strategies. The identification of optimum doses for relapse prevention with FGAs has proven difficult, and there is little randomized data on SGAs. Although the randomized evidence on a superiority of depot compared to oral treatment is not ideal, this approach suggests obvious advantages in assuring compliance.
Enormous scientific efforts have been made since the development of chlorpromazine in 1953 to improve the drug treatment of schizophrenia leading to more than 10 000 controlled trials currently summarized in the Cochrane Schizophrenia Group's register (Adams et al.2008). In this context, the current article presents our personal interpretation of the core evidence about the pharmacological treatment of schizophrenia. Other interpretations exist and the article can not compete with national treatment guidelines that have been methodologically developed (e.g. Lehman et al.2004). But what makes our article different is that it attempts to follow the logical steps of clinical care. Arranged as a simple algorithm it starts from the choice of an antipsychotic for acutely ill patients and ends with the most important questions about maintenance treatment.
The report is based on published systematic reviews or relevant key articles on topics not yet covered by a systematic review. We did not perform a literature search for this review but the first author (S.L.) regularly runs Medline searches with the key words ‘antipsychotic* OR schizophreni*’.
Treatment of an acute episode (see Fig. 1 for outline)
Choice of antipsychotic drug and what is the first-line pharmacotherapy
More than 50 yr of psychopharmacological research on antipsychotics have yielded little evidence-based data for the choice among the at least 50 antipsychotic drugs available worldwide (Gaebel & Awad, 1994). Much hope has been placed in pharmacogenetics, but this field is still in its infancy (Stone et al.2010). As all drugs have advantages and disadvantages, it is impossible to define a single first-line compound. In this context we briefly summarize our interpretation of the discussion about second-generation antipsychotics (SGAs) vs. first-generation antipsychotics (FGAs), and which is the best SGA, mainly based on recent systematic reviews (Leucht et al.2008b, 2009a, b) and effectiveness studies CATIE (Lieberman et al.2005) and EUFEST (Kahn et al.2008). We did not include the most recent antipsychotic drugs (paliperidone, asenapine, iloperidone) except for listing their optimum dose ranges as comparative data are still sparse.
We interpret the meta-analyses such that overall clozapine, amisulpride, olanzapine and risperidone may be somewhat more efficacious than FGAs and other SGAs (Davis et al.2003; Geddes et al.2000; Leucht et al. 2009a, b). The magnitude of the efficacy superiority was small, and certainly smaller than differences in the most frequent side-effects. The use of clozapine is restricted to refractory patients due to the increased risk of agranulocytosis (Alvir, 1994). However, careful monitoring of changes in the blood count has almost eliminated agranuloctyosis-induced death (Lieberman, 1998).
Haloperidol produced more extrapyramidal side-effects (EPS) and tardive dyskinesia than SGAs, even when it was used at doses <7.5 mg/d (Leucht et al.2009a). Prophylactic antiparkinson medication reduces haloperidol's EPS (Rosenheck et al.2003). Low-potency FGAs produce fewer EPS than haloperidol, and not more than all SGAs (Leucht et al. 2003b, 2009a). The risk of mid-potency FGAs such as perphenazine, is between that of high- and low-potency FGAs (Klein & Davis, 1969), and it was similar to risperidone according to Hoyberg & Fensbo (1993) . Among the SGAs a meta-analysis suggested that clozapine and quetiapine probably have the lowest and risperidone the highest EPS risk (Rummel-Kluge et al.2010).
In contrast to this partial amelioration of EPS with SGAs there is also a shift in the side-effect profile with most SGAs increasing the risk of metabolic abnormalities (de Hert et al.2009; Newcomer, 2005). Clozapine and olanzapine produced most weight gain and metabolic problems (Allison et al.1999) (the PORT guideline recommends against using them first-line in first-episode patients; Buchanan et al.2010), while ziprasidone and haloperidol caused the least followed by aripiprazole, amisulpride and probably mid-potency FGAs such as perphenazine. Low-potency FGAs, zotepine, sertindole, quetiapine and risperidone probably lie somewhere in the middle (Leucht et al.2009a).
Thioridazine, ziprasidone, sertindole and pimozide are probably associated with most QTc prolongation (Buchanan et al.2010).
Risperidone and amisulpride seem to produce the greatest prolactin increase, even more so than haloperidol (Bushe et al.2008, Kleinberg et al.1999). Aripiprazole, clozapine and quetiapine may be associated with the smallest prolactin increase.
Any creation of a clear hierarchy is difficult. Ideally, studies including all agents would be necessary, but for obvious reasons such studies are not available. Many drugs have not been compared head-to-head and methodologically sound dose-finding studies have not been conducted for almost all first-generation compounds (Davis, 1974). For example, the evidence on sulpiride and perphenazine – drugs that did well in the CATIE (Lieberman et al.2005) and CUtLASS studies (Jones et al.2006) – is scarce and their optimum doses have not been established in appropriate fixed-dose designs (Hartung et al.2005; Omori & Wang, 2009). Furthermore, it is difficult to weigh all these differences in efficacy and side-effects. For many patients weight gain may be unacceptable, while for other patients EPS are undesired. Again other patients may want to receive the most efficacious drug regardless of side-effects.
Recommendation: Drug choice needs to be primarily based on pragmatic criteria:
Prior response to an antipsychotic assuming that previous efficacy predicts efficacy in the current episode.
The avoidance of an antipsychotic that produced side-effects in an earlier episode.
General side-effect profile of the substances (see above) and patient characteristics (e.g. in a patient with diabetes, substances causing weight gain should be avoided).
Slight differences in efficacy.
The intended long-term administration. If it is clear that a depot substance is to be administered for maintenance therapy, the choice of an oral antipsychotic that is available as a depot can facilitate the later change.
If possible, patients' preferences should be taken into account in a shared decision-making process. This may improve compliance and in the end it is the patient who must take the medication. Therefore patients should be involved (Hamann et al.2003).
Sedation strategies in agitated patients
There are many strategies for the treatment of agitated patients such as monotherapy with an antipsychotic (high or low potency, oral or parenteral), monotherapy with a benzodiazepine, combining both, adding a sedating low-potency FGA, rapidly dissolving tablets, liquid formulations, etc. Due to a lack of RCTs a definite recommendation is not possible and probably most options work in some way. Liquid and rapidly dissolving tablets are options for safeguarding drug intake. Parenteral administrations do not lead to a much faster onset of action (Moller et al.1982). A review found no significant difference between intramuscular formulations of haloperidol and chlorpromazine (Leucht et al.2008a). The combination of haloperidol and lorazepam seems to be more efficacious than their monotherapy (Alexander et al.2004; Battaglia et al.1997). The SGAs aripiprazole, olanzapine and ziprasidone are available as intramuscular formulations (Citrome, 2007). Their advantage includes fewer EPS than haloperidol. A general problem is that the SGA studies were conducted for registrational purposes and all participants had to give informed consent. These are not the patients for whom intramuscular medication is indicated (often violent patients who are not willing to take medication). From a methodological perspective four large (200–300 participants), real-world pragmatic trials with intramuscular medication need to be highlighted (Alexander et al.2004; Huf et al.2003, 2007; Raveendran et al.2007). In these the combination of haloperidol and promethazine produced more rapid tranquillization with fewer dystonic reactions than haloperidol alone (Huf et al.2007) and no more EPS than lorazepam alone (Alexander et al.2004). It did outperform olanzapine in secondary outcomes (number of additional injections) with no difference in EPS, probably due to the anticholinergic properties of promethazine (Raveendran, 2007). Only intramuscular midazolam, a highly sedating benzodiazepine was more efficacious than haloperidol plus promethazine (Huf et al.2003) but the mean dose of haloperidol in this study was lower compared to the other listed trials and so a dosing effect leading to the favourable outcome for the benzodiazepine compound cannot be completely ruled out. In the case of chronic, persistent aggressive behaviour clozapine was superior to haloperidol and olanzapine (Krakowski et al.2006). The evidence for other SGAs or mood stabilizers is indirect at best (e.g. derived from non-schizophrenic populations), because RCTs in this specific group of patients have not been conducted.
Recommendation: Parenteral administration should usually only be applied in emergency situations when patients refuse medication. Among the many available options the best evidence is available for haloperidol combined with promethazine. If the subsequent oral treatment is planned to be an SGA it can be reasonable to chose its intramuscular formulation for sedation.
Treatment of patients with depressive symptoms
Depressive symptoms are frequently present in acutely ill patients with schizophrenia (Barnes & McPhillips, 1995) and may first improve with antipsychotics alone. Based on limited data some SGAs have shown superior antidepressant properties (Leucht et al.2009a). A Cochrane review found some evidence that antidepressants are effective for post-psychotic depression (Whitehead et al.2003). The risk that an adjunctive antidepressant may provoke psychosis has been judged to be small (Siris, 1991; Whitehead et al.2003).
Recommendation: Depressive symptoms associated with an acute episode should not be automatically treated with an antidepressant. Neuroleptic-induced depressive symptoms might be ruled out by antiparkinson medication or switching to a drug with fewer EPS. Post-psychotic depression may be treated with an antidepressant.
Treatment of patients with negative symptoms
All SGAs and haloperidol improved negative symptoms more than placebo (Leucht et al.2008b). In Leucht et al. (2009a), only four SGAs (amisulpride, clozapine, olanzapine, risperidone) were more efficacious than FGAs for negative symptoms. However, almost all relevant studies were conducted in patients suffering predominantly from positive symptoms. In such patients it is unclear whether it is a superiority for negative symptoms related to primary, or only to secondary, negative symptoms which can stem from positive symptoms, depression or extrapyramidal side-effects. Only studies in patients with primary/predominant negative symptoms could clarify whether SGAs are truly more efficacious. Here, the best evidence is available for low-dose amisulpride (50–300 mg/d) which proved superior compared to placebo in several trials (Leucht et al.2002b). However, comparing SGAs to FGAs only a few RCTs in predominant negative symptoms are available and among these only a single small trial showed superiority of olanzapine compared to haloperidol (Lindenmayer et al.2007). Three meta-analyses showed that adding an antidepressant can be effective for predominant negative symptoms (Rummel et al.2006) or in chronic patients with schizophrenia (Sepehry et al.2007; Singh et al.2010).
Recommendation: Secondary negative symptoms due to EPS should be ruled out (e.g. by antiparkinson medication or changing the antipsychotic). Low-dose amisulpride is the best examined drug in patients with predominant negative symptoms. Adding an antidepressant can be tried.
Factors that need to be ruled out before non-response is assumed
Before assuming non-response and introducing a major change in treatment, a number of items need to be ruled out.
Is the diagnosis of schizophrenia correct?
A psychotic depression/mania or severe personality disturbances can be difficult to distinguish from schizophrenia.
Do side-effects mask a response?
Akathisia can resemble psychotic agitation and parkinsonism can mimic negative symptoms.
There is better information about the therapeutic dose ranges of SGAs than FGAs because dose-finding studies have been conducted during their registration processes. There is some variability between guidelines, but we present the recommendations of the PORT guideline (Buchanan et al.2010), supplemented by Falkai et al. (2005) or our own evaluation if necessary: amisulpride 400–800 mg/d for predominant positive symptoms and 50–300 mg/d for predominant negative symptoms, aripiprazole 10–30 mg/d, asenapine 10–20 mg/d, clozapine 300–800 mg/d (in treatment-resistant patients), iloperidone 12–24 mg/d, olanzapine 10–20 mg/d, quetiapine 300–800 mg/d, paliperidone 3–15 mg/d, risperidone 2–8 mg/d, sertindole 12–20(24) mg/d, ziprasidone 80–160 mg/d, zotepine 150–250 mg/d. Regarding FGAs, an influential review concluded that very high doses (>700 mg/d chlorpromazine) are generally no more efficacious than moderate doses (Baldessarini et al.1988). A Cochrane review suggested that in uncomplicated patients higher doses than 7.5 mg/d haloperidol do not bring about more efficacy (Waraich et al.2002) and in single RCTs, 4 mg/d (Zimbroff et al.1997) or neuroleptic threshold doses (average 3.3 mg/d) were as effective as higher doses (McEvoy et al.1991). Much less is known about the doses of other FGAs and dose equivalence tables are problematic among others since they assume linear relationships. We orient the interested reader to recent publications on dose equivalence (Andreasen et al.2010; Davis & Chen, 2004; Gardner et al.2010; Kane et al.2003). First-episode and elderly patients often need lower doses. The latter are also more sensitive to side-effects due to reduced metabolism and one third of the dose in younger patients has been suggested as a rule of thumb (Jeste et al.1993).
Was the patient compliant?
Compliance problems are so frequent that they should always be addressed by switching to liquid medication or rapidly dissolving tablets in in-patients; or considering depot medication in outpatients (Leucht & Heres, 2006). Plasma-levels and the determination of a patient's cytochrome P450 status can also be useful.
Sufficient plasma level?
There is not sufficient evidence to allow the exact titration of antipsychotics guided by therapeutic drug monitoring except for clozapine, where it can be useful. Therefore, plasma-level measurements are only indicated in the following situations (for review of therapeutic ranges see Baumann et al. (2004) ):
Suspicion of non-compliance.
Lack of response in spite of taking usually sufficient doses to rule out ultra-rapid excessive metabolization of the antipsychotic due to a polymorphism of the cytochrome P450 enzyme system.
Pronounced side-effects despite the administration of a usual dose to rule out ‘poor metabolizers’ due to too little production of cytochrome P450 enzymes.
Medication interactions, smoking, etc., which can also lead to elevated or lowered plasma levels via effects on the cytochrome P450 system.
Has the patient been treated sufficiently long?
The onset of response to treatment of individual patients is highly variable and recent analyses attempted to resolve some of the heterogeneity by identifying different trajectories (Levine & Leucht, 2010; Levine & Rabinowitz, 2008). Nevertheless, recent meta-analyses rejected the long-held hypothesis that there is a general ‘delay of onset of action’ of several weeks for antipsychotic drugs. The largest part of the drug effect occurred in the first week and became consistently smaller thereafter (Agid et al.2003; Leucht et al.2005a). Two further studies demonstrated that the antipsychotic drug effect can be disentangled from placebo as early as 24 h (Agid et al.2008; Kapur et al.2005).
These findings call into question guidelines that a drug should not be considered ineffective and switched before 3–6 wk (Gaebel et al.2006; Lehman et al.2004; McGorry, 2005; NICE, 2003). Indeed, several studies have now suggested that non-responders might be identified as soon as after 2 wk. Very roughly, those patients who did not have an improvement of at least 20–25% in PANSS or BPRS total score reduction (corresponding to less than minimally improved on the Clinical Global Impressions Scale; Leucht et al.2005b, c; 2006) at 2 wk were unlikely to respond later (Ascher-Svanum et al.2008; Chang et al.2006; Correll et al.2003; Jager et al.2009; Kinon et al.2008a, Leucht et al. 2007a, 2008c, Lin et al.2007). In one analysis, however, only improvement at 4 wk but not at 2 wk predicted later response (Lambert et al.2009). Although only little evidence is available that early switching of medication is effective (see below), keeping the patient on the original drug is likely to lead to a poor outcome. Therefore, changing treatment earlier than the usually proposed 4–6 wk could be justified.
Recommendation: Before non-response is assumed, the diagnosis should be confirmed, and side-effects masking response, non-compliance, insufficient dose and too fast drug metabolism should be ruled out (e.g. by plasma levels). Any major change in treatment should not be considered before absolute non-response to at least 2–4 wk of treatment with a full antipsychotic dose.
Switching the drug or increasing the dose in initial non-responders
Surprisingly few RCTs have examined the effectiveness of switching medication or of a major dose increase.
(a) Switching to a different antipsychotic
To date only one trial examined switching after only 2 wk of virtually no improvement. It found that switching from risperidone to olanzapine was more effective than keeping patients on risperidone, but the superiority was small (Kinon et al.2010).
Kinon et al. (1993) treated 115 participants with 20 mg/d fluphenazine for 4 wk. The 47 non-responders were randomized to double-blind treatment with (a) continuation with 20 mg/d fluphenazine (control group), (b) 80 mg/d fluphenazine (dose increase group) or 20 mg/d haloperidol (switching group). Regardless of the strategy, only four (9%) of these patients responded.
Shalev et al. (1993) randomized 60 patients with acute schizophrenia to haloperidol, perphenazine or chlorpromazine. Non-responders after 4 wk of treatment were randomized twice to open treatment with one of the other two antipsychotics. At the end of the study, the majority of the participants had responded with an overall higher number of patients responding to perphenazine or chlorpromazine than to haloperidol.
Suzuki et al. (2007) randomized 78 patients with schizophrenia to olanzapine, quetiapine or risperidone. After up to 8 wk treatment, the non-responders were twice re-randomized to the remaining compounds. Only 16 patients did not respond to any of the three antipsychotic drugs with olanzapine and risperidone showing comparable efficacy while quetiapine was less effective. As in Shalev et al. (1993) the major limitation of the study was the lack of a control group of patients who stayed on the initial drug to rule out that the improvement was not simply an effect of time.
Stroup et al. (2007)post-hoc analysed those 114 participants who had received perphenazine in CATIE phase I, but discontinued it. Those participants who were randomized to quetiapine or olanzapine in phase II did better than those randomized to risperidone. This was explained by the similar profile of perphenazine and risperidone. Thus, if the antipsychotic drug is changed, it is reasonable to chose a compound with a different receptor-binding profile.
(b) Substantially increasing the dose
The aforementioned study by Kinon et al. (1993) showed no incremental efficacy of increasing fluphenazine from 20 mg/d to 80 mg/d. McEvoy et al. (1991) randomized patients who did not respond to neuroleptic threshold doses of haloperidol (mean 2.3 mg/d) to either continuation of threshold doses or doses up to 10 times greater. The dose increase was not associated with better efficacy. In Kinon et al. (2008b) there was no efficacy difference between 10 mg/d, 20 mg/d and 40 mg/d olanzapine in patients with suboptimal response to previous treatment. Olanzapine (40 mg/d) was, however, slightly better in a severely ill subgroup.
Recommendation: More evidence may be available for switching a drug rather than a massive dose increase beyond the official ranges, although even for the former strategy the evidence is scarce. Individual patients respond only to very high doses and such a history should be considered, but we would not recommend this as a general strategy. If a dose increase beyond the optimum doses was not effective, the dose should be reduced to previous levels. An antipsychotic with a different receptor-binding profile should be chosen in case of a switch. There is no randomized evidence whether patients should be started on a loading dose, a medium dose or titrated slowly upwards. This decision needs to be adapted to the situation (e.g. the acuteness of the symptoms) and the antipsychotic (some require titration or have clearly dose-related side-effects).
Clozapine and other SGAs for treatment resistance
Clozapine had been continuously prescribed in several European countries in the 1970s and 1980s (Naber & Hippius, 1990). It was reintroduced in the USA and other countries after the landmark study of Kane et al. (1988) reporting clozapine to be superior to chlorpromazine in refractory patients. This finding was confirmed by subsequent studies (Kane et al.2001; Rosenheck et al.1997) and meta-analyses (Chakos et al.2001; Essali et al.2009; Wahlbeck et al.1999). Clozapine was superior to other SGAs in CATIE phase II but it was the only open-label arm (McEvoy et al.2006). In the CUtLASS study clozapine was more effective than other SGAs as a group (Lewis et al.2006). A meta-analysis of 28 blinded RCTs did not find significant superiority of clozapine compared to other SGAs (Leucht et al.2009b). We suggested that too low clozapine doses explained this unexpected finding. Nevertheless, we believe that a conclusive clozapine vs. other SGAs trial is still needed.
Concerning the efficacy of other SGAs compared to FGAs in treatment-resistant patients a meta-analysis published in 2001 concluded that in contrast to clozapine none were convincingly superior (Chakos et al.2001). The evidence base has not substantially changed since. We briefly summarize the studies in treatment-resistant participants in Leucht et al. (2009a) . There was no RCT comparing amisulpride, sertindole or zotepine with FGAs in treatment-resistant schizophrenia. Kane et al. (2007) found no difference between aripiprazole and perphenazine in 300 participants who had not responded to olanzapine or risperidone. Olanzapine was superior to haloperidol in a post-hoc analysis of treatment-resistant patients from a large pivotal trial (Breier & Hamilton, 1999; Tollefson et al.1997). It was also found superior to haloperidol in patients having poorly responded to FGAs in Volavka et al. (2002) . Four smaller studies did not find olanzapine clearly superior (Altamura et al.2002; Buchanan et al.2005; Conley et al.1998; Smith et al.2001). Conley et al. (2005) found no difference between quetiapine, risperidone and fluphenazine, while in Emsley et al. (2000) partial non-responders to fluphenazine did better when they were randomized to quetiapine instead of haloperidol. Wirshing et al. (1999) found a superiority of risperidone compared to haloperidol in the first 4 wk of an 8-wk study, while Conley et al. (2005), See et al. (1999) and Volavka et al. (2002) did not find any clear superiority. Kane et al. (2006) found superiority of ziprasidone compared to chlorpromazine in terms of negative symptoms but not overall symptoms. In our opinion none of these findings are robust enough to generally recommend another SGA than clozapine for refractory patients.
Recommendation: Clozapine remains the gold standard for treatment-refractory patients while for none of the other SGAs is there sufficient evidence available. By most guidelines clozapine is recommended after at least two adequate trials with different antipsychotic drugs.
Among the many augmentation strategies that have been examined, there was no clear effect of benzodiazepines (apart from sedation, Volz et al.2007), beta-blockers (Cheine et al.2001), lithium (Leucht et al.2007b), carbamazepine (Leucht et al.2002a) and valproate (Basan et al.2004). The largest valproate study (249 participants) showed a more rapid onset of improvement in the augmentation group at 2 wk (Casey et al.2003), but even this effect was not replicated in a recent trial with 402 participants (Casey et al.2009). Lamotrigine is a promising adjunct according to a Cochrane review (Premkumar & Pick (2006) ) and a meta-analysis focusing on clozapine non-responders (Tiihonen et al.2009). However, the latter meta-analyses were relatively small and possibly not robust (Trikalinos et al.2004). Attempts to add acetylcholinesterase inhibitors to improve cognitive deficits have not proven effective (Stip et al.2007). The effects of other augmentation strategies such as polyunsaturated fatty acids (omega-3 and omega-6 fatty acids; Joy et al.2006), glutamatergic agents (Tuominen et al.2006), oestrogens (Chua et al.2005), dehydroepiandrosterone (DHEA; Elias & Kumar, 2007), amphetamines (Nolte et al.2004), cyclooxygenase-2 (COX-2) inhibitors such as celecoxib (Akhondzadeh et al.2007; Muller et al.2002), and erythropoietin (Ehrenreich et al.2007) are either still in the experimental stage or inconclusive.
Recommendation: Monotherapy should be the aim, because combinations of medications increase the risk of side-effects and drug–drug interactions. In drug combinations it is often not clear which compound has led to success and which should be discontinued. Compliance with several agents is difficult. Most importantly, there is no convincing evidence for any combination strategy. It might at best be tried for target symptoms (e.g. benzodiazepines for anxiety) and quickly discontinued if ineffective.
Combinations of antipsychotic drugs
Although antipsychotic drugs are frequently combined in clinical practice, several recent systematic reviews did not show a clear benefit. Correll et al. (2009) suggested a better response in the combination group compared to monotherapy, but it remained unclear which of the many combinations were effective, the positive results were mainly based on Chinese studies in which antipsychotics were combined right from the start rather than after non-response (which is current practice), if clozapine was part of the combination treatment, and if the trials lasted >10 wk (confirmed by Paton et al.2007; but not by Taylor & Smith, 2009). Furthermore, there was a possibility of publication bias. Barbui et al. (2009) included only studies in clozapine non-responders and found a positive effect only in randomized open studies, but not in double-blind studies. Recent RCTs suggested some benefits of adding aripiprazole to risperidone for reducing prolactin (Kane et al.2009) and to clozapine for reducing weight (Fleischhacker et al.2008). Another study suggested possible efficacy for aripiprazole augmentation for negative symptoms (Chang et al.2008).
Recommendation: Due to lack of convincing evidence, combining antipsychotic drugs can not be generally recommended (Goodwin et al.2009). If used, drugs with different receptor-binding profiles should be combined. For example, clozapine, which has relatively little antidopaminergic properties, may be combined with selective dopamine receptor antagonists such as amisulpride, sulpiride, aripiprazole or low-dose haloperidol, while e.g. combining olanzapine and quetiapine makes less sense.
Electroconvulsive therapy (ECT) and repetitive transcranial magnetic stimulation (rTMS)
In a Cochrane review, ECT monotherapy was more effective than sham ECT, but less effective than antipsychotic drugs (Tharyan & Adams, 2005). Very limited evidence supported the combination of ECT with antipsychotics. Meta-analyses found medium effect sizes for rTMS for positive symptoms (Aleman et al.2007; Tranulis et al.2008). Fewer studies examined the effect of rTMS applied over frontal areas in order to influence negative symptoms with more equivocal results (Hajak et al.2004; Klein et al.1999). Guidelines consider rTMS to be still in the experimental stage (Gaebel et al.2006; Lehman et al.2004).
Recommendation: ECT is recommended only as a last resort (except for special indications such severe catatonia), but advantageously compared to the other augmentation strategies, it is effective as monotherapy and has a different mechanism of action than antipsychotics.
Maintenance treatment (see Fig. 2 for outline)
In naturalistic studies, only ∼20% of patients with a first episode did not experience another episode within 5 yr (Robinson et al.1999; Shepherd et al.1989). We assume that in multiple-episode patients this proportion is even smaller. Maintenance treatment with antipsychotic drugs clearly reduced relapse rates from 53% to 16% within 9.7 months (Gilbert et al.1995). Intermittent treatment (tapering medication once a patient is in remission and only re-starting it when there are early warning signs) was less efficacious than continuous treatment (Carpenter et al.1990; Herz et al.1991; Jolley & Hirsch, 1990; Pietzcker et al.1993; Schooler et al.1997), even in first-episode patients (Gaebel et al.2010).
Recommendation: As there are no valid indicators that could predict who will not relapse, we follow an early guideline (Kissling et al.1991) that suggested continuous maintenance treatment with antipsychotic drugs for all patients, except for those in whom side-effects outweigh the benefits, those with very mild episodes, and unclear diagnoses.
Choice of the antipsychotic drug in maintenance treatment
In Leucht et al. (2003a), SGAs as a group reduced relapse rates compared to FGAs from 23% to 15%. In an update the various SGAs were not pooled because we felt that they were too different to form a class. Among the single SGAs, only olanzapine, risperidone and, based on a single trial, sertindole were superior to FGAs (Leucht et al.2009a). A clearer advantage includes fewer tardive dyskinesia, which occurred with an annual incidence of 3.0% in SGAs vs. 7.7% with FGAs (see Correll & Schenk, 2008, an update of Correll et al.2004).
Recommendation: For pragmatic reasons, maintenance treatment can be carried out with the antipsychotic drug to which the patient responded in the acute episode as long as there are no important side-effects. However, the higher risk of FGAs for tardive dyskinesia or weight gain-related risks associated with several SGAs should be considered.
Depot formulations are not more efficacious per se (Davis et al. 1994), but they could have advantages in terms of improved compliance, although the randomized evidence is limited. A summary of Cochrane reviews found little supportive evidence but included short-term studies and studies in in-patients – both factors that may have limited the compliance-improving effect (Adams et al.2000). Indeed, a recent meta-analysis including only long-term RCTs in outpatients found a superiority of depot compared to oral treatment, but there were limitations (Leucht et al.in press). Interested readers might refer to a comprehensive supplement on treatment with depot in the British Journal of Psychiatry (Patel et al.2009).
Recommendation: Despite the currently limited evidence, we believe that depot medication has obvious advantages such as assured medication and the awareness of when a patient stopped treatment (Kane et al.2003).
Duration of relapse prevention with antipsychotic drugs
There is little evidence to guide us about the ideal duration of antipsychotic relapse prevention. An early consensus conference suggested keeping first-episode patients on maintenance medication for at least 1–2 yr and multiple-episode patients for at least 5 yr. In severe cases the prophylaxis might even be indefinite (Kissling et al.1991). This recommendation was based on a few, small, controlled (but not all randomized), withdrawal trials. Multiple-episode patients who had been stable for up to 5 yr had significantly higher relapse rates when their medication was withdrawn compared to those staying on drugs (Cheung, 1981; Johnson, 1976, 1979; Odejide & Aderounmu, 1982). A later study found the same result in patients who had been stable for 6 yr (Sampath et al.1992). In first-episode patients the duration was restricted to 2 yr, because this was the longest study duration here (Chen et al.2010, Crow et al.1986; Hogarty & Ulrich, 1998; Kane, 1982). In the last 20 yr we have gained little new evidence and this remains a gap. But even a gradual symptom-guided tapering (and discontinuation if feasible) of antipsychotic treatment after 6 months of symptom absence led to doubled relapse rates compared to maintenance treatment in first-episode patients in a recent trial (Wunderink et al.2007). It should also be noted that lack of evidence beyond 6 yr (or 2 yr for first-episode patients) does not mean that the risk for relapse goes down. Empirically, multiple-episode patients stable for 2 yr relapsed at the same rates as patients newly entered into a relapse study (Hogarty et al.1976). Many national guidelines have adopted the early suggestions. Their recommendations range between 1 and 2 yr for first-episode patients, and between 2 and 5 yr for multiple-episode patients (Gaebel et al.2005).
Recommendation: First-episode patients should usually be kept on maintenance medication for at least 1–2 yr and multiple-episode patients for at least 5 yr.
The optimal dosage in maintenance treatment
It is unclear whether lower doses than those usually applied in the acute phase are sufficient for maintenance treatment. Studies in the 1980s and 1990s attempted to identify doses of depot FGAs that were high enough to prevent relapses, but low enough to minimize side-effects. Unfortunately, the definition of an ideal dose was not clear cut and the lowest doses led to the highest relapse rates although the differences were not always statistically significant (Hogarty & McEvoy, 1988; Johnson et al.1987; Kane et al.1983, 2002; Marder et al.1984; Schooler et al.1997). In a systematic review, higher dosages of conventional antipsychotics than 375 mg/d chlorpromazine equivalent did not produce additional effectiveness in maintenance therapy (Bollini, 1994). In another review, dosages between 50 mg/d and 100 mg/d chlorpromazine equivalents led to more relapses than doses between 200 and 500 chlorpromazine equivalents (Barbui et al.1996). Only a few dose-finding studies on relapse prevention with SGAs are available. Arato et al. (2002) found no difference between 40 mg/d ziprasidone and higher doses. Simpson et al. (2006) reported a trend superiority of risperidone Consta 50 mg bi-weekly compared to 25 mg bi-weekly, and Wang et al. (2010) found the lowest relapse rate when the risperidone dose that was effective in the acute phase was maintained compared to 50% and 25% reductions. In a systematic review including SGAs and FGAs, standard doses were clearly more effective than ‘very low’ doses, while the comparisons of relapses with ‘low’ doses was only of borderline statistical significance (Uchida et al.2009). On the other hand, it may not be so important to lower the dose of SGAs, since within their official dose ranges many side-effects are not very dose related.
Recommendation: Given the uncertainty whether lower doses are as effective as standard doses, we recommend keeping the dose of the acute phase as long as there are no important side-effects. Nevertheless, due to various pressures, psychiatrists sometimes tend to give very high doses or combination treatments in the acute phase. The maintenance phase should definitely be used for reducing such excessive dosing. Any dose reduction should be performed very slowly to avoid withdrawal effects and rebound psychoses.
Statement of Interest
Stefan Leucht has received speaker/consultancy/advisory board honoraria from Sanofi-Aventis, Bristol–Myers Squibb, Eli Lilly, Essex Pharma, GlaxoSmithKline, Janssen/Johnson and Johnson, Lundbeck and Pfizer including the fees and travel expenses for attending these functions; he has received funding for research projects from Eli Lilly and Sanofi-Aventis. Stephan Heres has received honoraria from Janssen-Cilag, Sanofi-Aventis, Pharmastar, and Johnson & Johnson; he has accepted travel or hospitality payment from Janssen-Cilag, Sanofi-Aventis, Johnson & Johnson, Pfizer, Bristol–Myers Squibb, AstraZeneca, Lundbeck, Novartis, and Eli Lilly and Company. Werner Kissling has received speaker and/or advisory board/consultancy honoraria from Janssen, Sanofi-Aventis, Johnson & Johnson, Pfizer, Bayer, Bristol–Myers Squibb, AstraZeneca, Lundbeck, Novartis and Eli Lilly.