Letter to the FDA regarding research misconduct

If you have read my September 2006 post, "Gambling and Dopamine Agonists: Where's the Beef?" you are already familiar with my criticism of the FDA-sponsored "study" that mined the FDA's Adverse [Drug] Event Reporting System (AERS) database and found what it claimed was an exorbitantly high reporting ratio for pramipexole (a dopamine agonist) and gambling. What they failed to reveal was the profoundly stimulating effect that publicity clearly had on the reporting ratio, i.e., 38 out of the 39 reports they attribute to this reporting ratio were registered after the publicizing of the first study that claimed to have found an association.

Here is the letter.

My complaint concerns the research letter to the editor entitled “Association Between Pathologic Gambling and Parkinsonian Therapy as Detected in the Food and Drug Administration Adverse Event Database” by Ana Szarfman, MD, PhD, P. Murali Doraiswamy, MD, Joseph M. Tonning, MD, MPH, and Jonathan G. Levine, PhD published in the Archives of Neurology in February of 2006.

In their letter, Szarfman et al say they found 39 reports linking Mirapex to gambling in the FDA's AERS database. They then go on to apply the Multi-item Gamma Poisson Shrinker (MGPS) statistical algorithm to the reporting rates of certain drugs in conjunction with gambling, and come up with an exorbitantly high adjusted reporting ratio for Mirapex and gambling.

However, what Szarfman et al failed to reveal, never mind take into account when interpreting the reporting rates, was that *all but one* of the reports in question regarding Mirapex were registered *after* a study purporting an association between it and gambling was publicized heavily by the American Academy of Neurologists in August of 2003 (Stacy, Driver-Dunckley. Neurology, August 2003).

Just to put this in context, in the six years Mirapex was on the market prior to the publicizing of Stacy’s 2003 study, there was one report concerning it and gambling (there were three such reports for levodopa-carbidopa, the “’gold standard” of Parkinson’s therapies, during the same time period), and in the year and a half after the publicizing of Stacy’s study and before the end point of Szarfman et al’s presented data, there were 38 reports for gambling and Mirapex - that is a 149-fold increase annually.

Research misconduct consisting of omission is classified as falsification and is defined in CFR 42 93.103(b) as follows:

“Falsification is [intentionally]... omitting data or results such that the research is not accurately represented in the research record.”

Given that without the reports generated by the publicity of the first study, there are virtually no reports involving Mirapex and gambling, and therefore no exorbitant reporting ratio, this omission clearly results in the inaccurate representation of the research in the research record.

That leaves the question of intent.

In the absence of some sort of admission on the part of an author, the determination of intent in the context of omission can only be inferred from the answers to the following questions:

1. was the author aware of the relevance of the omitted data (precludes ignorance);
2. was the author in possession of or did the author have access to the omitted data (precludes lacking access to the data); and
3. did the omission of the data require that the research record be written differently than it would have been had the omitted data been included (precludes oversight)

The answer to the first question is yes, as is evidenced by the following recommendation made by Dr. Szarfman and her co-authors in their 2005 paper entitled “Perspectives on the Use of Data Mining in Pharmacovigilance”:

"Publicity resulting from advertising, litigation or regulatory actions (e.g. "Dear healthcare provider" letters and product withdrawals) may result in increased reporting and can generate higher-than-expected relative RRs.[reporting ratios] Relative RRs should be examined over time in hopes of detecting these influences, although there are no definitive criteria for using data-mining techniques to reliably identify such effects."

The answer to the second question is yes, as well, as not only is every entry into the AERS database dated, but Dr. Szarfman herself responded to my inquiry regarding the dates of the reports by saying “this information is publicly available through NTIS,” which is true.

Finally, the answer to the third question is obviously also yes - if the data in question had been included, a completely different letter would have been required.

And while the authors do make mention of some level of stimulated reporting, they fail to reveal the extent to which the stimulated reporting contributed to the reporting ratio in question. Given the skyrocketing in reporting activity (from 0.17 reports per year to 25 reports per year) after the first study received such publicity, it would be impossible not to acknowledge that the stimulated reporting *was* the MGPS score, and that in fact without it, there would be no MGPS score for Mirapex at all – unless one omitted that information, of course.

So - the lead author on the study was aware of the relevance of the omitted data; and, just in case the fact that every AERS record is dated weren’t evidence enough that she was in possession of the omitted data, her response to me proves that she both knew where to find the data and that it was accessible to her. Finally, including the omitted data would have necessitated the reporting of a failure to find a reporting ratio of any note rather than the success that was reported – in other words, the omission could not have been an oversight, since it necessitated (or facilitated) the writing of a completely different letter.

Therefore, there is a preponderance of evidence that the omission was intentional.

As an aside, oddly, they also attribute the stimulated reporting to the reporting to the AERS of the cases outlined in Stacy’s 2003 study, which is clearly nonsensical, since such reports are not publicized.

I would provide you with some background as to why this misinformation is damaging to people with Parkinson’s, but I get the sense that it is really not relevant to the evaluation of an allegation of scientific misconduct.

I have attached a bar chart illustrating the sudden influx of reports related to Mirapex (pramipexole) upon the publicizing of the 2003 study by Stacy, et al (it also shows the reports related to Sinemet (levodopa-carbidopa) for a little context.)



On another note, the only one of the authors who makes any disclosure is Dr Doraiswamy, who merely states that he has received research grants and consulting or speaking honoraria from “several pharmaceutical companies.”

Clearly, the fact that Dr. Doraiswamy has had a long standing relationship with Novartis (including serving on an advisory board and their speaking bureau) is relevant, given that Novartis owns 33% of Roche, which makes the European version of Sinemet (levodopa and a decarboxylase inhibitor); and given that Novartis’ generic arm, Sandoz, makes levodopa/carbidopa; and Novartis/Orion market/make the newest iteration of levodopa, Stalevo, all of which are competitors of Mirapex. Also omitted are his past relationships with Merck/Bristol Myers Squibb, maker/marketer of Sinemet and Sinemet CR, which are also levodopa products and therefore Mirapex’s major competitors.

I will be happy to provide my sources for the above – and anything else in this letter, including the information I obtained from the AERS under the FOI act – please let me know if I should.

I would appreciate it if you would tell me exactly what I should expect going forward in terms of transparency and sharing of information (for example, if an inquiry is initiated, how updated will I be kept regarding what is going on, or should I expect no information until the matter is resolved? And will I be told who else, if anyone, will review my complaint?)

Thank you for your time.