Accutane-PET Study Discredited in Court Opinion
It’s one thing for a physician to serve as a medical expert for plaintiffs in court cases. It’s very much another when plaintiffs actually commission a physician to conduct a study—the results of which are intended to support their litigation. (Think Andrew Wakefield and the associated antivaccination mess.)
Now, according to a newly released court opinion (courtesy of Jim Beck at the Drug and Device Law blog), it is revealed that plaintiffs in the Accutane litigation* specifically commissioned Emory psychiatrist J. Douglas Bremner to conduct a PET study, which led to the conclusion that Accutane reduces metabolism in “a brain area known to mediate symptoms of depression.” The results of the uncontrolled study were published in a 2005 issue of the peer-reviewed American Journal of Psychiatry,** and for the purposes of the court case, “Bremner issued an expert report opining that Accutane can cause depression and suicide.”
If this conflict of interest isn’t sufficiently disturbing, the court opinion reveals further that Bremner’s commissioned data were highly suspect.
During a lengthy pretrial hearing, in which Bremner was questioned about the PET study by defendant’s counsel,
Bremner was repeatedly confronted with problems in the PET study, including missing data, inaccurate data, and deviations from the methodology he claimed to have followed. As a result…the court permitted Bremner to re-work his study data and issue a supplemental expert report and allowed defendant to re-depose him. When the hearing resumed, Bremner admitted that certain underlying data, known as “Bmax numbers” which had been used to make critical calculations in the study, [were] not retrievable from its computerized format, and some of the data concerning individual study subjects [were] still inaccurate.
(I could be wrong here, but Bmax, in this context, appears to be a maximum basal metabolic rate of the brain area in question—a necessary reference standard from which to calculate any metabolic changes.)
The opinion goes on to state,
[T]hat Bremner did not actually use the methodology he claimed to have used. Although his PET scan article was peer-reviewed, he admitted that he did not in fact follow the steps described in the article.
Significantly, contrary to representation made in the article, [Bremner] did not get before-and-after Skindex questionnaires from many of the subjects. Those questionnaires were designed to elicit the extent to which the subjects might be worried about their acne. This was relevant because some scientists were of the view that worrying, as well as depression, could affect activity in the orbital frontal cortex.
Note: The orbitofrontal cortex is the brain area in question. Reduced volume of the orbitofrontal cortex has been implicated in major depressive disorder. According to Bremner’s published article, the Skindex questionnaire was “administered before and after treatment.” The implication is that it was administered to all participating subjects.
The opinion continues,
Bremner also could not document much of the data on which his published results were based. Further, he admitted that some of the statistical analysis was inaccurate. For example, in the October 2, 2006 hearing session, Bremner admitted that, for each study participant, comparing the activity in the orbital frontal cortex with the activity in the whole brain revealed no difference between the subject who took Accutane and those who took antibiotics.
Retreating from the results claimed in his 2005 article, he testified at the hearing that the “absolute metabolic rates” for the two groups was significantly different, and contended that was the key finding of the PET study. However, Bremner claimed that he could not produce the source data for that analysis because the “Bmax” numbers used to calculate those metabolic rates was on an optical computer drive that could not be opened.
Further, while he admitted that some of the Bmax numbers he used in his calculations were inaccurate, he could not check the accuracy of the remaining numbers because the original data could not be retrieved.
An expert’s scientific peers cannot fairly judge the expert’s written work, including whether it is worthy of publication, if his article does not accurately represent either the underlying data or what the author did to produce his results. We agree with the trial judge that, in essence, Bremner’s study was not “soundly and reliably generated.”
While this information indicates that Bremner and his coauthors should have retracted the PET study from publication in the AJP, the shaken peer-review process is sustained by a mere notice of correction. At the end of the Letters to the Editor section in a 2008 issue of the AJP is this short paragraph in reduced type:
In the article “Functional Brain Imaging Alterations in Acne Patients Treated With Isotretinoin,” by J. Douglas Bremner et al. (Am J Psychiatry 2005; 162:983-991), only seven subjects in each treatment group completed the Skindex posttreatment. The secondary analysis that included whole brain metabolism before and after treatment did not reach significance on re-analysis.
In the meantime, the authors of at least 42 articles, according to a Google search, have cited Bremner’s PET study. These authors include Bremner himself, who wrote in 2008, “Administration of [Accutane] (but not antibiotic) was associated with a 16% decrease in brain metabolism in the orbitofrontal cortex after four months of treatment.”
Also a blogger, Bremner, in an ironic turn, charged conflicts of interest among Roche’s key opinion leaders who believe that Accutane does not cause depression.
PET = positron emission tomography.
* Specifically Palazzolo v Hoffman-LaRoche, in which a mother argued that Accutane caused the suicide of her teenage son. Since July 2009, Roche’s trade name for the drug is Roaccutane.
** Published disclosures do reveal that the PET study was “[s]upported by funding from Liam Grant, director of the Roaccutane Action Group (80%), and by lawyers involved in Accutane litigation (20%).” The amount of money that Grant provided is not clear; however, information provided by Bremner here indicates that Grant, whose son committed suicide while taking Accutane, contacted Bremner directly and gave an “unrestricted grant” to Emory for the PET study. A 2005 article in USA Today indicates that Grant spent about $1 million to fund the Emory PET study and a rodent study at the University of Massachusetts.
Unrelated, generic PET image from Wikipedia.
02/06/10 addendum: At his blog, Bremner responded today to Jim Beck’s post and this post. Bremner harshly and wrongly, in my opinion, describes our discussions of the court opinion as “an outpouring of hate.” (I think “criticism” is a better descriptor.) He also casts aspersions at Beck’s and my respective alliances with the pharmaceutical industry. (Beck makes it no secret that he represents drug companies in related litigation. I make it no secret that I have worked in pharmaceutical marketing.* While this information is important to know, it doesn’t make our criticisms or concerns about plaintiff-sponsored research necessarily baseless.)
Bremner continues by denying that his PET study was “commissioned for the litigation,” despite the fact that the court opinion states, on page 3, “There is no dispute that the study was commissioned specifically for use in this litigation.” This statement clearly indicates that opposing parties in the litigation did not contest this point, despite the fact that Bremner disagrees with it.
Bremner then fails to support his disagreement. He merely affirms that 80% of the PET study was funded by an Accutane plaintiff (Liam Grant) and writes that Roche wouldn’t perform a study (presumably a PET study) or supply medication for a study. Bremner’s allegation that Roche wouldn’t participate in an Accutane-PET study hardly refutes the statement that Bremner’s PET study was commissioned by the plaintiffs. And the fact that the drug industry supports a lot of clinical research, much of it under FDA scrutiny as Jim Beck stresses, doesn’t mean that a) a physician is justified in accepting plaintiff money to perform research (whether intended for use in their litigation); b) a physician who accepts plaintiff money for research is immune to plaintiff bias; or c) accepting plaintiff money somehow makes up for any potential bias in pharma-funded research.
Bremner then addresses the study’s missing data (which he says were erroneously labeled “Bmax” by the court). According to Bremner, these data were eventually retrieved, albeit after “a court deadline had passed”; however, Bremner does not address the court’s disturbing statement, “some of the data concerning individual study subjects [were] still inaccurate.”
Bremner minimizes the court’s allegation that he did not follow the study’s described methodology with respect to the use of the Skindex questionnaire, by stating that “[i]t was by no means the primary outcome of the study” and that “there was no correlation between this item and brain function.” He alleges that the questionnaire was added “late” to the study protocol at the request of a “dermatologist who was later found to be a paid consultant to Roche.” The implication is that Roche, through this dermatologist, somehow tried to manipulate Bremner’s PET study. (It should be noted that the authors of the study do not include a dermatologist, as far as I can tell; so if a dermatologist had significant input into the study protocol, this should have been revealed in the study article.)
The Skindex questionnaire (if a PubMed search is any indication) appears to be a widely used, quality-of-life measure in dermatologic studies. The court wrote that the consistent use of the Skindex questionnaire “was relevant because some scientists were of the view that worrying, as well as depression, could affect activity in the orbital frontal cortex.” The article by Bremner et al indicates that the Skindex questionnaire was 1 of 3 components of the study’s behavioral assessment (the other 2 being the Hamilton Depression Rating Scale and a clinician-administered acne questionnaire).
In any event, Bremner et al wrote that the Skindex questionnaire was “administered before and after treatment” in his 28-person PET study. This information, it turns out, was misleading, if not outrightly false. The questionnaire was administered to only half of the enrollees before and after treatment, according to the 2008 correction.
Bremner does acknowledge that “some data entry errors were found” in his study but that a “re-analysis of the study resulted in no change in the conclusion.” This latter statement doesn’t appear to be entirely true, at least according to the court opinion and to the 2008 correction.
In their 2005 AJP article, Bremner et al wrote,
Administration of [Accutane] but not antibiotic was associated with decreased brain metabolism in the orbitofrontal cortex after 4 months of treatment…This effect was seen for both absolute metabolism…and for the ratio of orbitofrontal to whole brain metabolism (F=4.64, df=1, 110, p<0.05). A secondary analysis included pretreatment whole brain metabolism in the model and also showed greater reductions in orbitofrontal metabolism after treatment in the isotretinoin group than in the antibiotic group (F=9.66, df=1, 104, p=0.002).
But the court wrote,
[I]n the October 2, 2006 hearing session, Bremner admitted that, for each study participant, comparing the activity in the orbital frontal cortex with the activity in the whole brain revealed no difference between the subject who took Accutane and those who took antibiotics.
And the 2008 correction states,
The secondary analysis that included whole brain metabolism before and after treatment did not reach significance on re-analysis.
Bremner concludes his rebuttal by stating that “it doesn’t matter” and provides a link to a 2008 review article he cowrote. In this article, Bremner proposes how retinoids (like Accutane) may be involved in the neurobiology of affective disorders and cites 2 suggestive mice studies. Among the clinical studies described by Bremner, none of which is well-controlled, the results are mixed. Some suggest a link between Accutane use and depression, and others do not.
I have no vested interest in whether Accutane causes depression or suicide; however, the data, as they currently exist and as Bremner presents them, are not abundantly compelling.** Furthermore, I maintain that plaintiff-sponsored research is disturbing and represents a potentially significant conflict of interest for the investigator who accepts plaintiff funds to perform related studies—particularly studies that may be used to the advantage of the plaintiff in ongoing litigation.
Whether the results of Bremner’s PET study, which the court found problematic, are valid can only be determined by their reproducibility—ideally in a randomized, double-blind trial.
* However, during the last 6 years, I’ve worked in continuing medical education (CME); during the last 2 years, this has been in a freelance capacity. I’m most certainly not paid by anyone to blog.
** Which is why plaintiffs are interested in funding research.