Girgis and colleagues present a unique 9-year naturalistic follow-up of an initially randomised clozapine v. chlorpromazine trial in first-episode schizophrenia. Reference Girgis, Phillips, Li, Li, Jiang and Wu1 Very few such long antipsychotic drug follow-ups have been published and the trial team must be lauded for their enormous efforts.
First-episode studies find little difference among antipsychotic drugs because there are ceiling effects due to high rates of remission – 78% in Girgis et al’s study. A considerable number of patients with first-episode schizophrenia remitted without drug treatment in the pre-antipsychotic era, Reference Davis and Andriukaitis2 and approximately 20% will not have a second episode. Reference Shepherd, Watt, Falloon and Smeeton3
Nowadays, long-term studies (such as the study by Girgis et al) are analysed by including all patients, whether they remained on their original drug, were switched to another drug or were drug-free. We have likened this approach to an automobile race between Alaska to Argentina where drivers were randomised to very expensive BMWs or the cheapest Ford. Reference Leucht, Kissling and Davis4 If the car broke down during the race, the driver could choose either a BMW (disclosure: BMW is a funder of the Technische Universität München) or a Ford as a replacement car. Many Fords quickly broke down, and the drivers invariable replaced them with BMWs. That only 29 (18%) out of 160 randomised participants remained in Girgis et al’s study in the groups they were originally designated may have diluted the initial drugs’ effects. This said, it is disappointing that even the most efficacious antipsychotic clozapine failed to produce a long-lasting benefit.
We try to put this editorial in a historical perspective. All the previous ‘chlorpromazine me too’ antipsychotics had roughly equal efficacy, and the low-potency antipsychotics were advertised as having few extrapyramidal side-effects (EPS), the high-potency ones as having minimal sedation so you could use them in high doses. The high-potency drug won the media war, the patients losing because the high-potency antipsychotics led to more EPS. The revival of clozapine in 1988 by the landmark study of Kane and colleagues, Reference Kane, Honigfeld, Singer and Meltzer5 who found that 30% of treatment-resistant patients responded to clozapine v. 5% to chlorpromazine, led to enormous enthusiasm about the development of new antipsychotic drugs.
Twenty years later the evidence shows a much more sobering picture: initially, drug companies advertised their drugs as ‘clozapine-like’, i.e. EPS-free, more efficacious for general and treatment-resistant illness, and especially effective for negative symptoms. They promoted these drugs as being 5-HT2 (serotonin) receptor blockers, like clozapine, using this pharmacological property as a justification for calling the drugs ‘atypicals’. We have analysed the degree of dopamine and serotonin blockade in positron emission tomography or single-photon emission computed tomography studies, and found all the antipsychotics to be potent D2 (dopamine receptor) blockers; Reference Stone, Davis, Leucht and Pilowsky6 however, most of the atypical antipsychotics which are potent 5-HT2 blockers are no more efficacious than the ‘typical’ antipsychotics, and amisulpride, which is more effective, Reference Leucht, Kissling and Davis4,Reference Leucht, Corves, Arbter, Engel, Li and Davis7 has no propensity to block 5-HT2 receptors. The pharmaceutical companies also charged very high prices for these drugs, a worldwide market expected to reach $20.8 billion in 2011 (www.reportlinker.com/p0236244/Antipsychotic-Drugs-Technologies-and-Global-Markets.html), cost being the key driver of the debate.
Antipsychotic drugs differ in many respects
Unfortunately, from the first randomised controlled trials onwards, most of the so-called atypicals had about equal efficacy to typicals, even for negative symptoms, and none of those few atypicals with superior efficacy to first-generation antipsychotics reached clozapine’s efficacy. Reference Leucht, Kissling and Davis4,Reference Leucht, Corves, Arbter, Engel, Li and Davis7 Nevertheless, the classification of atypical v. typical antipsychotics obscured differences, because meta-analyses based on hundreds of studies consistently found certain new drugs (e.g. amisulpride, olanzapine, risperidone) to be significantly better than older drugs, although these differences were small (clearly smaller than most differences in side-effects), and there is a debate about the effects of high haloperidol doses in comparator groups. Reference Leucht, Kissling and Davis4,Reference Leucht, Corves, Arbter, Engel, Li and Davis7
Initially, the field was not aware of the weight gain problem, probably due to the preoccupation with EPS and tardive dyskinesia in the 1970s/1980s. But it had been known that that low-potency typicals such as chlorpromazine caused weight gain. In the early papers on clozapine in the 1970s, it was reported that clozapine was associated with weight gain but it was not of great concern, although it should have been owing to its serious medical complications. The weight gain problem received more attention partly due to ‘counter-marketing’ (companies highlighting the problems of their competitor drugs) and the excessive weight gain associated with olanzapine and clozapine. However, some of the atypical drugs cause moderate weight gain (e.g. risperidone, quetiapine) and others minor amounts of weight gain (e.g. ziprasidone, aripiprazole, amisulpride). Reference Leucht, Corves, Arbter, Engel, Li and Davis7 Here again, the over-generalisation of atypicals and typicals created the confusing misinterpretation that all old antipsychotics do not and all new antipsychotics do produce weight gain.
What remained was a clear advantage of the new drugs in terms of EPS, which with some restrictions is still true. A drug with a high EPS risk would probably nowadays no longer get approved. Of course, most studies used the high-potency typical antipsychotic haloperidol as a comparator. This drug, however, produced more EPS than atypicals even in low doses below 5 mg/day (see discussions in Leucht et al Reference Leucht, Kissling and Davis4,Reference Leucht, Corves, Arbter, Engel, Li and Davis7 ). The EPS risk of the various atypicals differs (e.g. risperidone produces more EPS than most other second-generation antipsychotics Reference Rummel-Kluge, Komossa, Schwarz, Hunger, Schmid and Kissling8 ), but these differences are relatively small. Low-potency typical antipsychotics also produce relatively few EPS and not necessarily more than atypicals, Reference Leucht, Corves, Arbter, Engel, Li and Davis7 but they are all sedating. There are a few old drugs such as sulpiride and some old mid-potency antipsychotics such as perphenazine, which clinically surprisingly did not produce more EPS than quetiapine in the CATIE (Clinical Antipsychotic Trials of Intervention Effectiveness) study. Reference Lieberman, Stroup, McEvoy, Swartz, Rosenheck and Perkins9 Some atypical antipsychotics (e.g. amisulpride, risperidone) increase prolactin levels more than typical antipsychotics, but almost all other atypicals increase prolactin levels less than typicals. Some atypicals (e.g. sertindole, ziprasidone) and some typicals (e.g. thioridazine, pimozide) prolong the QTc interval. There are differences in sedation, as well.
CATIE, CUtLASS and EUFEST
Finally, the controversial effectiveness studies CATIE, Reference Lieberman, Stroup, McEvoy, Swartz, Rosenheck and Perkins9 Cost Utility of the Latest Antipsychotic drugs in Schizophrenia Study (CUtLASS) Reference Jones, Barnes, Davies, Dunn, Lloyd and Hayhurst10 and European First-Episode Schizophrenia Trial (EUFEST) Reference Kahn, Fleischhacker, Boter, Davidson, Vergouwe and Keet11 fuelled the discussion, because they were again interpreted as showing no differences, although the side-effect results were virtually the same as in the previous meta-analyses: even concerning efficacy we see a similar pattern (for discussion, see Leucht et al Reference Leucht, Kissling and Davis4 ). The CUtLASS Reference Jones, Barnes, Davies, Dunn, Lloyd and Hayhurst10 trial is an extreme case because groups of drugs were compared assuming that all atypical drugs were similar. The EUFEST Reference Kahn, Fleischhacker, Boter, Davidson, Vergouwe and Keet11 study found a markedly lower rate of treatment discontinuation with amisulpride and olanzapine and in the last observation carried forward results of the Positive and Negative Syndrome Scale (PANSS) compared with haloperidol. These results may be clouded by bias present in open studies. Reference Leucht, Corves, Arbter, Engel, Li and Davis7 The CATIE study found olanzapine was the most effective in terms of rates of discontinuation (hazard ratios between 0.41 to 0.59) and time to good response, Reference Lieberman, Stroup, McEvoy, Swartz, Rosenheck and Perkins9 although other statistical analyses failed to find differences in efficacy. High discontinuation rates (e.g. 74% in CATIE), which differ in amount and reasons between groups, are an important problem of many long-term studies. At some degree of attrition due to poor efficacy, patients are no longer randomised and even the best statistical method cannot account for these enormous discontinuation rates. Although we have shown that pharmaceutical companies often put spin on the interpretation of their studies, Reference Heres, Davis, Maino, Jetzinger, Kissling and Leucht12 two analyses did not detect a sponsor bias in the actual efficacy data. Reference Leucht, Corves, Arbter, Engel, Li and Davis7,Reference Davis, Chen and Glick13
Our personal interpretation of this debate is that the classification of antipsychotics into atypical and typical has led to a lot of confusion because these drugs differ in many properties such as efficacy, side-effects and cost (some are now generic), mechanism of action (amisulpride is a selective dopamine antagonist and aripiprazole is a partial dopamine agonist), and therefore the classification might be abandoned. We agree that apart from clozapine no new drug was a breakthrough in efficacy. Their (amisulpride, olanzapine, risperidone) effect sizes compared with typical antipsychotics were small (between 0.13 and 0.31), although the numbers needed to treat for one additional responder were actually not bad (between 6 and 15). Reference Leucht, Corves, Arbter, Engel, Li and Davis7 But to say that all drugs are the same and that therefore it does not matter which drug is given to the patient is wrong, and reverting back to high-dose haloperidol, the previous primary treatment in many countries, would not help patients.
We feel, as there are many real differences among drugs, that the physician should adapt the treatment accordingly to the individual patient through a shared decision-making process. Patients have different preferences and at the end it is they who must take the medication. Some patients want to avoid weight gain or EPS or sexual side-effects, and others want to receive the most efficacious compound. The best antipsychotic drug will not work if the patient does not take it, so there is a role for depot formulations. We feel it is important to empower the patient to make informed decisions about which drug and dose to take as well as the route of delivery, which may lead to better feedback and adherence, and although evidence is not perfect (in some cases only suggestive), we think the cumulative effect would be to get a better result with more patients.