Recent “Fraud on the FDA” Court Decision Should Cause CROs to Take Note

November 26, 2007

In the preemption world, “fraud on the FDA” cases are fairly common.  As one court recently used the term, fraud on the FDA means a drug or a medical device company is liable to someone who was injured by their product if that product was approved by FDA based on false information the company submitted and the company has admitted to the Agency that it engaged in fraud.  But drug (and device) companies often do not gather all the data they submit to FDA.  Instead, they frequently have help from Contract Research Organizations (“CROs”), which perform a variety of functions, including overseeing clinical trials.  So what happens when a CRO does something questionable during a clinical trial?

Enter the Wawrzyneks.

Eileen Wawrzynek had spinal surgery in 1999.  During the surgery her doctors used a device – ADCON-L – that had received conditional FDA approval to help prevent scarring.  The surgery did not go well.  Mrs. Wawrzynek had three subsequent surgeries and the Wawrzyneks sued the doctors and the hospital alleging malpractice.  They lost.  Then they sued the medical device manufacturer, Gliatech, and settled.  Then they sued Statprobe Inc., a CRO, in the U.S. District Court for the Eastern District of Pennsylvania.

Very strict procedures needed to be followed during the ADCON-L clinical trials.  In order to ensure the results were not influenced by whether the investigator knew the device had been used (it is rather difficult to create placebo devices) the results consisted of an MRI scan taken of the area where the surgery occurred.  This MRI was then read by a blinded researcher who rated the resulting scar on a scale of 0 (no scar) to 4 (maximal scarring).  This result was to be recorded in pen on a specific sheet.  To be successful, the manufacturer needed to show that ADCON-L prevented scarring and that there was less than a 5% chance that the results shown were due simply to chance (i.e., a p-value less than 0.05).  The preliminary results looked good.  Then things changed.  As new results were analyzed, the p-value skyrocketed to greater than 0.5, meaning no one could really determine whether ADCON-L did anything useful at all.  In fact, scarring between the two groups was about equal, as was the number of people receiving the highest score of 4.  Because only one person was reading all of the MRIs, it made sense to investigate whether he was consistent in his scoring.

The device manufacturer sent two doctors to supervise the researcher.  The researcher still did not know who had received the device, but these two new doctors did.  The researcher re-read each of the prior MRIs, and instead of recording his results in pen, as he had previously done, he called out the number to one of the new (and unblinded) doctors, who recorded it in pencil.  The researcher then signed the results and the CRO re-entered the numbers into the trial database without noting that they had been re-read or were in pencil.

Through the re-read process, numbers changed.  Fewer ADCON-L patients received 4s, and more non-ADCON-L patients did.  These changes were enough to reduce the p-value from 0.66, nowhere near good enough for approval, to 0.01, well below the required 0.05 level.  Not only was the new data now statistically significant, it was good enough for FDA to conditionally approve the device.

The court indicated that this was a pretty straightforward case of fraud on the FDA for the manufacturer, as the manufacturer had already admitted the fraud.  What makes this case interesting is that this admission was enough for the court to hold the CRO liable as well (at least potentially, as the court’s decision in Wawrzynek v. Statprobe was just an order denying summary judgment in favor of the CRO).  As the court stated:

[T]he Court sees no legal theory or compelling policy reason to allow [the CRO] to use [the manufacturer] and its wrongdoing as a shield. Because the FDA found that fraud and wrongdoing occurred during the ADCON-L approval process, the door to [a fraud on the FDA claim] was opened wide enough to allow both [the manufacturer and the CRO] to pass through.

So, what’s a good CRO to do?  In this case, the year before the re-read occurred the CRO had already recognized it was being asked to do some questionable statistical practices by the manufacturer.  Wawrzynek provides strong incentive for a CRO to protect itself the next time a similar situation arises, either removing itself from the study or simply refusing to go along with similar manufacturer practices.