Malgré des avancées pharmacologiques, la schizophrénie reste difficile à traiter et des résultats insatisfaisants sont encore trop fréquents. Les cibles thérapeutiques de réponse, rémission et rétablissement ont été définies de façon plus uniforme mais il manque encore une mesure convenable « d'efficacité » basée sur les capacités fonctionnelles des patients. De plus, dans ce domaine, il faut avancer en transférant les approches basées sur les mesures de la recherche à la pratique clinique. Savoir si il faut utiliser des antipsychotiques de 1^re ou de 2^e génération et lesquels, fait toujours l'objet de discussions. Un traitement personnalisé nécessite néanmoins d'envisager les symptômes actuels, la comorbidité, les réponses thérapeutiques antérieures, les effets indésirables ainsi que les choix et les attentes du patient. Il faut en outre mettre en balance les objectifs et les effets des traitements à court et à long termes. Alors qu'une réponse aiguë à des antipsychotiques de 1^re génération correctement dosés peut ne pas différer beaucoup d'une réponse à des antipsychotiques de 2^e génération, les bénéfices d'une fréquence moindre d'effets extrapyramidaux, de dyskinésie tardive et, peut-être, de rechute, font pencher la balance du côté des antipsychotiques de 2^e génération. Cependant, si l'on considère les profils individuels des effets secondaires, la distinction dans les antipsychotiques de 1^re et 2^e générations en tant que classes unifiées, ne tient pas et il est nécessaire d'avoir une vision plus différenciée et une sélection thérapeutique plus fine. À ce jour, la clozapine est le seul traitement ayant apporté ses preuves pour les patients résistants, le rôle d'une polythérapeutique antipsychotique et des autres stratégies d'augmentation restant au mieux obscur. L'amélioration des résultats du traitement dans la schizophrénie passe par des efforts de la recherche pour trouver des biomarqueurs de la maladie et de la réponse au traitement (effets thérapeutiques et effets indésirables). Ajoutons que de nouveaux choix de traitement sont nécessaires en ce qui concerne des cibles non dopaminergiques en tenant compte de la diminution des symptômes, de la prévention des rechutes, de l'augmentation de l'efficacité chez les non-répondeurs et de la diminution des effets indésirables majeurs.

Before providing an overview of these results and discussing their implications, it will be useful to set the stage for the shift in prescribing practices and the ensuing costeffectiveness questions that emerged in the 1990s.

Tardive dyskinesia became a particular concern because of its potential severity, persistence, and psychosocial as well as medicolegal consequences. A number of prospective studies were conducted to determine the incidence and risk factors for tardive dyskinesia.^2-4 In general, the risk of new cases of tardive dyskinesia was found to be 5% per year of cumulative drug exposure, with age and early occurring extrapyramidal side effects being two important risk factors. In elderly individuals (though with a different spectrum of diagnoses) receiving antipsychotic medications for the first time the incidence was generally fivefold higher.⁵

With the development of clozapine, the scientific and clinical community became convinced that it was possible to separate the therapeutic effects of antipsychotic medications from their neurologic extrapyramidal effects. Although it took many years after these early observations before clozapine was used on a wide scale for treatment-refractory patients,⁶ the awareness of its lack of propensity to cause extrapyramidal effects provided considerable impetus in drug development and served as a model for “atypicality” (a concept which has outlived its usefulness, but did serve a useful heuristic function).

With the introduction of other “second-generation” antipsychotics (risperidone, olanzapine, quetiapine, etc), there was considerable focus on differences in the propensity to cause extrapyramidal syndrome (EPS) between these agents and first-generation drugs.⁷ Overall, the differences were generally significant, but most often based on comparisons with haloperidol. Although haloperidol was the market leader at that point, making such comparisons somewhat logical, concerns have been raised that the choice of that medication, and its utilization in potentially higher than necessary doses, might have served to accentuate differences in the risk for neurologic adverse effects. The challenge of conducting studies, which take into account and adequately control for the relative dose equivalences of specific medications across a range of illness phases, patient ages, and outcome domains (ie, therapeutic and adverse effects), should not be minimized. In fact, one can easily argue that appropriately validated dose equivalences are generally lacking, and are usually derived from the analysis of large data sets from studies which were not necessarily designed to address these issues.

It appears that in the treatment of acute psychosis among patients with first-episode or recent-onset illness that there are not significant differences in overall response rates of psychotic signs and symptoms with different antipsychotic drugs or drug classes.^16-19 In general, response rates among such patients are quite high.

However, in the maintenance phase of treatment following acute response among first-episode patients, differences do begin to emerge favoring second-generation medications, incuding olanzapine and risperidone,^19-22 as well as amisulpride, quetiapine, and ziprasidone.¹⁹ In the treatment of multiepisode patients the picture becomes more complicated. The enthusiasm with which the second-generation drugs were received was fueled by unmet need, a long period without any new antipsychotics, vigorous marketing, and to some extent “wishful thinking” as clinicians would also like to believe that they have new and better tools with which to help their patients.

Over time as the cost of medications escalated, intense debate ensued about the relative merits of the different drugs and drug classes. Large “effectiveness” studies such as CATIE (Clinical Antipsychotic Trials of Intervention Effectiveness),¹⁷ CUtLASS (Cost Utility of the Latest Antipsychotic Drugs in Schizophrenia),^3,24 and EUFEST (European First Episode Schizophrenia Trial)¹⁹ were intended, to some extent, to clarify this issue.

The data from these trials must be considered along with the data from all other trials which have been the subject of a series of meta-analyses. Single studies, no matter how large, and meta-analyses, no matter how comprehensive, all have their limitations, but it is incumbent upon us to assimilate, objectively integrate and draw relevant conclusions from the evidence, such as it is.

In comparing first and second-generation antipsychotics (FGAs and SGAs),⁹ Leucht et al found four SGAs (amisulpride, clozapine, olanzapine, and risperidone) to be more efficacious than FGAs with effect sizes ranging from small to medium (0.13 for risperidone and 0.52 for clozapine). Leucht et al emphasized that the SGAs which were more efficacious showed these advantages for both of the specific domains of positive and negative symptoms, suggesting that their superiority for negative symptoms does not represent a “core component of atypicality.” As noted previously, all SGAs had fewer EPS than haloperidol, even when the latter was used in doses below 7.5 mg/day.⁹

When Leucht et al⁹ compared their results to those of other meta-analyses by Geddes et al, Davis et al, and the Cochrane Schizophrenia Group, ^25-27 the effect sizes were generally comparable. All meta-analyses - analyzing similar data sets - found that amisulpride, clozapine, olanzapine, and risperidone were significantly more efficacious than FGAs.

When Leucht and colleagues¹⁴ reviewed the head-tohead comparisons of SGAs they reported a similar pattern. Regarding total PANSS score change, olanzapine was more efficacious than aripiprazole, quetiapine, risperidone, and ziprasidone; risperidone was more efficacious than quetiapine and ziprasidone, and amisulpride was not statistically different from olanzapine or risperidone. However, regarding dropouts due to inefficacy, olanzapine was only superior to quetiapine and ziprasidone, two SGAs often dosed inappropriately low in schizophrenia, and amisulpride was more efficacious than ziprasidone. Surprisingly, clozapine only proved to be superior to zotepine and to risperidone in dropouts due to inefficacy. As the authors suggest, possibly inadequate doses of clozapine used in some studies might have contributed to these findings, which are inconsistent with most of the large individual studies.

In comparing these analyses to the results of CATIE and CUtLASS, there are some consistencies. In CATIE,¹⁷ olanzapine was superior in dropout rates due to inefficacy and time on effective treatment (at least when including the 23% of patients who were rerandomized to olanzapine), and clozapine (though given openly) was superior to all other studied SGAs in phase 2.²⁸ In CUtLASS,²⁴ clozapine was superior to other SGAs as well, but in this study, quality of life, and not all-cause discontinuation, was the primary outcome measure. This difference highlights the importance of the choice of the primary outcome measure, particularly in effectiveness studies, where the outcome is supposed to address elements of efficacy, tolerability, patient acceptability, and functioning.

Although the meta-analyses tended to find more support for risperidone's superiority than CATIE did, the dose equivalences used in CATIE have been challenged where the modal dose of risperidone was only 3.9 mg/day.

However, there were no differences between the individually studied SGAs and the FGA comparator perphenazine in CATIE,¹⁷ as well as between the class of clinician's choice SGAs and FGAs (mainly consisting of sulpiride, which some consider to be an SGA) in CUtLASS.²³

The large effectiveness trials have been discussed in great detail elsewhere. Despite the rigorousness with which we try to design trials, there are always compromises in both design and execution that are unavoidable, particularly when multiple outcome measures are included and multiple questions are addressed simultaneously.

A particularly important issue which impacts all of the effectiveness trials was stressed by Kraemer et al.²⁹ Whether a study is designed to establish superiority of one treatment or groups of treatments over others or to establish equivalence is critical. As Kraemer et al suggest, if a study is designed to demonstrate one treatment's superiority, then statistically nonsignificant results should not be assumed to be evidence of “equivalence.” To test this, a true noninferiority design is needed that generally requires larger samples.

The inevitable conclusions from these data are that drugs and drug classes are heterogeneous, and that we should not assume commonalities based on anything except appropriate comparisons. It is also obvious that every drug involves its own risks and benefits, and that clinicians have to evaluate data and make decisions based on the individual patients' presentation, history, sensitivities, preferences, responses, adverse effects, etc (Table I).

This serves as a segue into the next section of this discussion, which focuses not so much on which drug to choose, but how to conceptualize and evaluate response (both therapeutic and adverse) in order to inform treatment decisions (which may or may not involve changing medication). It is our firm belief that the real challenges and opportunities in treating patients with schizophrenia lie in how treatments are managed, evaluated, and potentially altered, rather than which drug one chooses for an initial trial.

As with all treatment planning, formulating and tracking treatment goals and outcomes is important (Figure 1).

These data have reinvigorated the effort to use early response/nonresponse as a predictor of subsequent response.^43,44 Correll et al⁴⁵ were the first to attempt to predict nonresponse at 4 weeks using the change of symptoms at 1 week in a sensitivity-specificity analysis in 131 patients receiving uniform treatment with fluphenazine. When Leucht et al⁴⁶ conducted a receiver-operator analysis to answer this question, a response of less than 20% improvement on the total Brief Psychiatric Rating Scale⁴⁷ (BPRS) best predicted nonresponse at 4 weeks. Chang et al⁴⁸ reported similar results in 123 patients treated with risperidone, and Leucht et al⁴⁹ replicated their earlier findings in 1996 patients from pooled olanzapine clinical trials.

Kinon et al⁵⁰ and Ascher-Svanum et al⁵¹ reported on post-hoc analyses of 1077 patients who had participated in a series of double-blind trials involving olanzapine and found that a less than 20% reduction in PANSS scores at 2 weeks was associated with good predictive power to identify patients unlikely to respond by 12 weeks. Patients with poor early response were also found to be more likely to discontinue from the trial and their cost of care was significantly higher than those with more robust early response.⁵¹

Kinon et al⁵² conducted a prospective study of 630 patients treated with risperidone (2 to 6 mg/day). The 192 patients who completed 2 weeks and experienced a less than 20% improvement on the PANSS at 2 weeks were then randomized to either continue on risperidone or be switched to olanzapine 10 to 20 mg/day for an additional 10 weeks. This first prospective study of its kind replicated the results of the retrospective, post hoc analysis described previously. The initial nonresponders were unlikely to respond when given up to 10 weeks additional time on the same medication. The authors found a very modest gain in switching to olanzapine; however, specific subgroups of patients might be identified who benefit more dramatically from a switch, ie, those who were more severely ill.

Kinon et al⁵³ reported on 115 newly admitted schizophrenia patients treated for 4 weeks with fluphenazine 20 mg/day. Those who failed to achieve the a priori defined remission level of positive symptoms (ie, no more than mild positive symptoms) were randomly assigned, double-blind, to continue on the same treatment, have the dose quadrupled, or be switched to haloperidol at an equivalent (20 mg/day) dose and followed for an additional 4 weeks. There were no significant differences in the outcomes between the three groups and, overall, only an additional 9% of patients met remission criteria at 8 weeks. This design needs to be applied with second-generation medications and ideally should also include polypharmacy and clozapine as comparators.

There are several reports indicating a high prevalence of antipsychotic polypharmacy in recent clinical practice.⁵⁵ Despite widespread use, there is relatively little evidence that this strategy is helpful, particularly when clozapine is not involved.

Correll et al⁵⁶ reviewed 19 randomized trials including 1216 subjects. Although the overall results could be interpreted as suggesting an advantage for combination therapy in comparison with monotherapy, the factors which appeared to contribute to this effect limit the conclusions that can be drawn for the present context. Combination therapy was more likely to be efficacious when administered from the start of treatment rather than waiting to identify poor or partial responders and when clozapine was involved. Studies which took place in China and studies which lasted longer than 10 weeks were also more likely to be positive. This leaves the question unresolved as to whether or not adding a second antipsychotic is likely to be helpful when a patient fails to derive an adequate response from an initial trial of monotherapy with drugs other than clozapine.

Lamotrigine has been examined in a Cochrane review including five studies involving 537 participants.⁶⁷ Some evidence of efficacy was found, but the results were not considered robust. Tiihonen et al⁶⁸ have reported evidence from a post-hoc analysis of patients from five trials involving lamotrigine augmentation of clozapine which suggested some significant effects, but interpretation remains difficult because of different designs and inclusion criteria. Further well-designed studies involving lamotrigine would be very valuable.

There is little question about the indications for continuing antipsychotic medication on an indefinite basis, except perhaps in patients who have only experienced one episode. Even there, however, relapse rates are 82% after 5 years,³⁷ and discontinuing medication is associated with a five times higher risk of relapse than staying on medication. This does suggest, however, that a small subgroup of patients might remain free of relapse, but at present we have no means to identify such individuals prior to making the decision to stop antipsychotic maintenance therapy.

The choice of medication takes on particular importance when long-term treatment is the focus, as the benefitto-risk ratio may change substantially. Some drugs are associated with greater or lesser degrees of specific longterm risks, eg, tardive dyskinesia, weight gain, type 2 diabetes, dyslipidemia, etc. Risk not only varies by drug, but of course also from individual to individual. At present, taking a good history and appropriate ongoing monitoring is the best strategy to identify particular risk profiles, but it is hoped that in the not-too-distant future pharmacogenetics might help in informing choice of optimum treatment(s).⁷²

Most clinicians know from experience that changing medications can be a destabilizing event. In fact, even in the CATIE phase 1 trial, patients rerandomized in a double-blind fashion to the same antipsychotic stayed in the study longer than those who were switched to a different antipsychotic.⁷⁵ Unless patients require a switch for an acute destabilization or life-threatening adverse event, clinicians need to consider the current psychosocial situation, level of support and symptomatic status of the primary disorder and of comorbid conditions when planning for a change in the medication regimen. In addition, consideration of the pharmacologic profiles of the pre-switch and the post-switch antipsychotic can also be helpful to predict the potential emergence of rebound phenomena that can complicate the switch process. The potential for rebound and withdrawal phenomena is greatest when the pre- and post-switch antipsychotics differ considerably regarding binding affinity for specific receptors (ie, pharmacodynamic dopaminergic, histaminergic, or cholinergic rebound) and/or when they differ considerably regarding their respective half-life (ie, pharmacokinetic dopamine rebound when the new antipsychotic has a much longer half-life and the prior antipsychotic is discontinued too quickly.)⁷⁶

Histaminergic rebound is characterized by a reversal of the antihistaminergic anxiolytic, calming, sleep-inducing, and EPS-reducing effects, potentially resulting in rebound insomnia, anxiety, agitation, EPS, and restlessness. Similarly, cholinergic rebound is characterized by agitation, confusion and EPS, and dopaminergic rebound can manifest as worsening or newly emerging agitation, aggression, psychosis, mania, akathisia, or withdrawal dyskinesia.

There is an ongoing debate as to whether and which first- or second-generation antipsychotics should be used, and in which particular patients. However, an individualized treatment strategy needs to consider both current symptoms, comorbid conditions, past therapeutic and adverse effect response, and patient choice and expectations. Moreover, acute and long-term goals and medication effects need to be balanced. While the acute response to appropriately dosed first-generation antipsychotics may not differ much from second-generation antipsychotics, advantages of lower rates of extrapyramidal side effects, tardive dyskinesia, and, possibly, relapse may favor second-generation antipsychotics. However, when considering individual adverse effect profiles, the differentiation into first- and second-generation antipsychotics as unified classes cannot be supported, and a more differentiated view and treatment selection is required.