The conventional wisdom: CATIE for schizophrenia

Here is the official line on the design and results of this study: CATIE was a one year study of schizophrenia, in which a large sample of about 1500 patients were recruited at about 50 clinical trial sites (exact numbers are n=1493in 57 sites). They were randomized double-blind to one of five drugs: olanzapine, risperidone, quetiapine, ziprasidone, and perphenazine. The first four agents were second-generation dopamine blockers, and the last agent was a first-generation traditional dopamine blocker. There was no placebo arm.

The basic question was whether second-generation dopamine blockers are more effective than a first-generation dopamine blocker in long-term treatment of schizophrenia. Secondary questions were whether the second-generation agents were more tolerable than the first, and whether any differences within second-generation agents could be seen on effectiveness and tolerability.

The commonly cited interpretation of the results is simple:

All antipsychotics were similar in efficacy and tolerability.

End of story.

Alternative interpretation: Effectiveness

What this simple interpretation overlooks is that they were similarly ineffective, or hardly effective. The main outcome was discontinuation of medication at one year. That’s a very low standard; it doesn’t involve improvement in schizophrenia symptom scores, much less functional improvement, which would be the real test of long-term effectiveness in schizophrenia. But simply asking the question – do patients stay on their drugs and do doctors keep patients on those drugs for up to one year? – the answer was No in about three-quarters of patients (range 64-82% depending on the drug). Since there was no placebo, we can’t know if the one-quarter or so who remained on their antipsychotic would have benefited more than if they had received nothing/placebo. But when three-quarters of patients do poorly with a treatment, the question would seem to rise whether that treatment is effective.

Again, it can be argued that those treatments already are proven effective, but those studies were pharmaceutical industry-sponsored trials for market approval, which are 3 month studies for acute exacerbation benefit using schizophrenia symptom rating scales.

In other words, the outcomes were different. The studies which allow antipsychotics to be on the general market show benefit over three months for acute worsening of symptoms. The CATIE trial tried to test a modest measure of longer term benefit – drug discontinuation up to one year – and it found that most patients did not benefit with antipsychotic treatment of any kind. It could be argued of course that there are maintenance studies of antipsychotics that show benefit over placebo up to one year of treatment, but those studies look at different outcomes than CATIE also, most commonly relapse into an acute exacerbation. They do not show improvement in chronic psychosis over time.

Hence in contrast to the planned comparison of one drug versus another drug, the CATIE trial only answers that question by raising more questions about how effective these drugs are.

The researchers who conducted the study have emphasized that this alternative interpretation should not be taken, but it seems not unreasonable given the lower overall effectiveness rates in the study.

The methodological take-home point is that you can’t change your outcomes when you’re comparing to the literature. CATIE assumed that the outcomes of 3 month trials would extend to a year, and allow drug versus drug comparisons. The outcomes at a year were poor enough, though, to throw into doubt any real long-term efficacy at all.

Alternative interpretation: Tolerability

The other major conventional wisdom interpretation of CATIE was that most drugs had similar tolerability, and specifically that the older antipsychotic perphenazine was similar in its extrapyramidal side effects to the modern antipsychotics.

This matter has been described in more detail elsewhere, where it was shown that this false equivalence was based on a standard mistake: the misuse of p-values. CATIE was not powered statistically to show differences in the lower absolute frequencies seen with extrapyramidal side effects (EPS). So false negative error is present. Using confidence intervals, differences can be shown to exist (with an irrelevance of the concept of “statistical significance), such as lower overall EPS rates with quetiapine compared to other agents.