TCPR: Dr. Jibson, there have been several meta-analyses comparing atypical antipsychotics published over the last few years. They are all long, complicated papers, and the conclusions are often controversial. Can you help us understand this literature?
Dr. Jibson: First, it’s important to distinguish a meta-analysis from a large comparative study. For example, the CATIE trial was a landmark comparative study of several antipsychotics. But it was not a meta-analysis. A meta-analysis is a way of summarizing the results of many different clinical studies.
TCPR: And are there different types of meta-analyses?
Dr. Jibson: There are two kinds of meta-analyses. First, there are those that focus narrowly on a specific question, and are strict in terms of which studies are included in the analysis. The Cochrane reviews are the best example of this. These reviewers typically set the bar high in terms of which studies are “good enough” to qualify for the analysis. This means that they have a high threshold to demonstrate an effect. Consequently, Cochrane reviews often conclude that there is not enough evidence to demonstrate a difference between drugs.
TCPR: Yes, I’ve noticed that. On the other hand, I suppose that when a Cochrane review concludes that one drug is better than another, you can be very certain that this is true.
Dr. Jibson: Yes, and that is one of the advantages of this restrictive approach to doing these analyses. The other type of meta-analysis is a more broad survey of as many studies as you can include. The goal here is to try to gather together data on as many patients as possible. This may allow for more interesting conclusions, although the quality of the studies may not be as high as one would like. These might best be described as hypothesis generating, that is, they may suggest a trend that should be investigated with more rigorous clinical studies. It is important to recognize that a meta-analysis cannot be any better than the studies that go into it. A meta-analysis comprised of poorly designed or poorly conducted studies, for example studies in which drugs are used at inappropriate doses or which lack systematic outcome measures, will not be able to reach a meaningful conclusion. Meta-analyses are useful, however, in some situations in which individual studies may be limited not by quality but by size. For example, a well-conducted study may not have a large enough sample size for group differences to reach statistical significance, but a meta-analysis of several studies together may correct that problem. Second, a group of well-conducted studies may each have an effect size that is too small to be detected in any one study, but a collection of studies that show a small but uniform effect may reach statistical significance. Finally, high quality studies may differ in outcomes and a meta-analysis would allow a weighted average of studies to demonstrate if there were a trend in one direction or another.
TCPR: Can you go through the most important of these studies for schizophrenia?
Dr. Jibson: The largest meta-analysis was the John Davis study of 2003 (Davis JM, et al., Arch Gen Psychiatry 2003; 60:553-64). Davis combined results from 124 clinical trials, including some unpublished material that was submitted to the FDA. He computed effect sizes between second generation antipsychotics (SGAs) and first generation antipsychotics (FGAs). An effect size is a measure of the advantage of one drug over another.
TCPR: And what did the Davis study conclude?
Dr. Jibson: The investigators concluded two things. The first conclusion was expected: clozapine is more effective than FGAs. (They did not compare clozapine with SGAs.) But the second finding was surprising: only two of the newer atypicals – olanzapine and risperidone – separated statistically from FGAs. In 2003, most of us assumed that all atypicals were equally effective, but this finding appeared to challenge this idea.
TCPR: I recall there was some controversy about this conclusion, however.
Dr. Jibson: Yes, there were two criticisms that emerged. First, the dosing used in the studies may have put some drugs at a disadvantage. Studies of risperidone were limited to those using a minimum dose of 4 mg/day, which is equal to the average dose in the community; the minimum allowed dose of olanzapine was 11 mg/day, about half the average community dose of 20 mg/day. But for studies of aripiprazole, which has an average dose in the community of about 20 mg/day, and ziprasidone, which averages around 120 mg/day in the community, no minimum dose was specified, and for quetiapine, the required minimum dose was 150 mg, far below the manufacturer’s recommendation of 400-600 mg/day for schizophrenia. So it may be that these three drugs didn’t separate from FGAs because they were dosed too low.
TCPR: And what was the other problem?
Dr. Jibson: The second criticism was more global, and more interesting in some ways. Over the past couple of decades, it has become harder and harder for clinical trials to demonstrate a separation between antipsychotics and placebo. I do not believe this is because antipsychotics are ineffective, because we have an enormous body of data to indicate that they are, indeed, effective. The problem seems to be that it is getting harder to recruit patients for studies of antipsychotics, presumably because more and more patients are finding their current treatment satisfactory. So there is a smaller group of patients willing to roll the dice on a placebo vs. active drug study, and the patients who enroll tend to be less responsive to standard treatments.
TCPR: And shortly after the Davis study, a “competing” meta-analysis was published.
Dr. Jibson: Yes, that was the Leucht study, published in Lancet (Leucht S, et al., Lancet 2003; 361,1581-89). This meta-analysis concluded that, in fact, there are no meaningful efficacy differences between the SGAs. But there were critiques of this study as well. It was much smaller than the Davis paper, including only 31 studies, and it included several drugs in the analysis that were not included in the Davis study, including Sertindole, amisulpride, and clozapine.
TCPR: And then the CATIE trial came along and caused quite a splash.
Dr. Jibson: Yes, the CATIE study showed that olanzapine appeared to be more effective than the other SGAs in maintenance treatment of schizophrenia. The main problem with CATIE was that more patients came into the study already on olanzapine than any other drug, potentially enriching the study with olanzapine-responsive patients. Later, when the authors reanalyzed the results by removing any patients who were randomized to the same drug they were already taking, the effect size of olanzapine was cut in half. It was still numerically better than the other SGAs, but it dropped below the 0.05 level of statistical significance to 0.09, leaving only a “trend” toward superior effectiveness for olanzapine, a point not included in the paper’s abstract (Essock SM, et al., Am J Psychiatry 2006; 163:2090-95).
TCPR: What about the most recent meta-analyses?
Dr. Jibson: A recently updated Cochrane review by Hunter (Hunter RH, Cochrane Database Syst Rev 2003; CD000440) compared risperidone with the FGAs. They found that there was slightly better patient acceptance of risperidone, which they attributed to “marginal benefits in terms of clinical improvement” and a side effect profile that “may be better” than haloperidol, but that these advantages were offset by the higher cost. They also made an intriguing observation: earlier studies showed a much clearer advantage for risperidone than later studies. They suggested that this was because most early studies were industry-sponsored, while later studies were publicly sponsored. We know that industry-sponsored studies have a higher probability of showing a benefit of the sponsor’s drug, although the reason for this is not always clear.
TCPR: I think many clinicians are reluctant to prescribe FGAs because of the fear of causing tardive dyskinesia (TD).
Dr. Jibson: That’s a realistic concern, although this has to be weighed against the problems with metabolic side effects of the SGAs. A review by Robert Rosenheck (Rosenheck RA, Br J Psychiatry 2007; 191, 238-45) looked at the cost vs. benefits of choosing SGAs specifically to prevent cases of TD, and found the cost of preventing each case of TD so exorbitant that most public policy officials would conclude that health systems could not afford it.
TCPR: In the end, it seems that the meta-analyses are pretty mixed. How have the studies affected your choices of antipsychotics?
Dr. Jibson: I don’t think these studies support the preferential selection of any one drug for first-line treatment, and I start about equal numbers of patients on each of the atypicals and a few patients on first-generation drugs, usually based on side-effect profile, insurance coverage, need for depot medication, or patient preference. Since the CATIE study, however, I make sure that any patient who has not had a satisfactory response to another drug has had a trial of olanzapine before I consider clozapine treatment. So I’m tending to give olanzapine the benefit of the doubt for a slight edge in effectiveness.
TCPR: We’re starting to see the SGAs becoming available as generics. How will this change clinical practice?
Dr. Jibson: I think that much of the heat of the debate on this issue will dissipate when the atypicals go generic. I predict that it will be similar to what happened in the antidepressant world. When SSRIs were first introduced, there was debate about whether they should be preferentially prescribed over tricyclics, and whether their better side effect profile made up for their increased cost. But once the SSRIs went generic, and the cost of the two classes were the same, everyone lost interest in that debate, and the SSRIs won in terms of popularity. My guess is that the same thing will happen with atypicals.