Monday, March 02, 2009

Internal Documents Suggest that Seroquel Data Were Not Presented Accurately

A document dated March 9, 2000 titled "BPRS meta-analysis" shows that AstraZeneca, maker of the antipsychotic drug quetiapine (Seroquel), knew fully that its drug did not relieve schizophrenia symptoms to the same extent as its older, generic competitor haloperidol (Haldol). The document provides results of a meta-analysis, a statistical analysis that combines the results of several individual studies. The authors used the Brief Psychiatric Rating Scale (BPRS) as their main measure of efficacy. The BPRS rates a variety of psychiatric symptoms relevant to schizophrenia. More details on the BPRS can be seen here. A total of ten clinical trials were included in the meta-analysis, which variously compared Seroquel to placebo, Haldol, and several other antipsychotic medications. Four trials compared Seroquel to Haldol. Several subscales of the BPRS were included in the analysis.

When examining the amount of change on the BPRS, Seroquel consistently outperformed placebo, both on the BPRS total score and on several of the BPRS subscales. However, in several analyses, Seroquel was outperformed by Haldol and by risperidone (Risperdal; Janssen's antipsychotic). The document states: "Against 'all doses' of Seroquel, each of the three significant p-values generated was in favour of Haloperidol (Total BPRS, Factor V, and Hostility Cluster). There was no evidence of significant differences between the treatments when Haloperidol was compared to high-dose Seroquel." This is a plain admission that Haldol outperformed Seroquel on several outcomes, but that high dose Seroquel yielded approximately equivalent results to Haldol. Only one trial compared risperidone to quetiapine and the results clearly favored risperidone. The document stated: "Comparisons against Risperidone using all doses of Seroquel showed significant improvements for Risperidone on total BPRS, Factor V scores, and the Hostility Cluster. Against high-dose Seroquel only, the Anxiety item, Factor I, and Mood cluster scores were also significantly in favor of Risperidone." Risperidone beat Seroquel, and did so by a wider margin when a high doses of Seroquel was used.

The author of the document, Rob Hemmings, summarizes the results in a table, which appears below. It is described as such: "The following table is an attempt to simplify the claims that could be obtained from these results. A ✔ is entered for those comparisons where we have a statistically significant benefit, be it with 'all doses' or with high dose Seroquel... A x marks those comparisons where a comparator has demonstrated significant superiority compared to Seroquel."
The table demonstrates that according to an analysis by AstraZeneca employees, Seroquel is only shown to outperform placebo, whereas Seroquel is shown to demonstrate poorer efficacy than several other medications.

Under the heading "Conclusions," the document states, in part:
In terms of generating positive claims for Seroquel, these analyses seem somewhat disappointing. Although some trends in favour of Seroquel were observed in the Factor I and Mood cluster items, there was no evidence in these analyses of a significant benefit for using Seroquel over any of the active agents assessed."
The internal analysis clearly indicates that, based on several clinical trials, Seroquel offered no benefits over the competition in terms of reducing schizophrenia symptoms. Indeed, other drugs tended to outperform Seroquel.

How Can These Data be Managed? Shortly after the internal meta-analysis was completed, AstraZeneca employees discussed how to handle the negative results. An AstraZeneca publications manager, John Tumas, wrote in an email
The data don't look good. I don't know how we can get a paper out of this. My guess is that we all (including Schulz) saw the good stuff, ie the meta-analysis of responder rates that showed we were superior to placebo and haloperidol and then thought further analyses would be supportive and that a paper was in order. What seems to be the case is that we were only highlighting the good stuff and that our own analysis support the "view out there" that we are less effective than haloperidol and our competitors.
It would appear that an earlier analysis provided positive results which did not hold up during the internal meta-analysis. "Schulz" almost certainly refers to Dr. Charles Schulz, a psychiatrist at the University of Minnesota. In a press release from the year 2000, Dr. Schulz was quoted:
I hope that our findings help physicians better understand the dramatic benefits of newer medications like SEROQUEL because, if they do, we may be able to help ensure patients receive these medications first. The data suggest that SEROQUEL is an effective first- choice antipsychotic.
This press release was based on Schulz's presentation at the American Psychiatric Association convention in May 2000. The email from John Tumas discussed earlier noted that a group at AstraZeneca needed to meet soon "because Schulz needs to get a draft ready for APA and he needs any additional analyses we can give him well before then." It is unclear if Schulz ever received the analyses that showed Seroquel was less effective than Haldol. Regardless, in the press release, he was also quoted as saying: "Almost 50 years later, however, many patients are still taking these medications [such as Haldol], even though more effective treatments like Seroquel exist." While he was stumping for Seroquel in a press release, AstraZeneca's internal data painted a completely different picture.

Schulz, in his role as primary author, would typically be expected to demonstrate a solid understanding of the data underlying his presentation. It raises troubling questions when an independent academic author presents results that are in direct opposition to the underlying data. Such issues have been mentioned previously on this site.

The documents regarding Seroquel are available at Furious Seasons. Reporting on other facets of the documents can be found at the St. Petersburg Times, Bloomberg, New York Times, and the Wall Street Journal.

No comments: