Results on gastrointestinal tract bleeding in the FDA’s Mini-Sentinel Program are at odds with findings from randomized clinical trials that compared dabigatran and warfarin, a meta-analysis published online Nov. 18 in JAMA Internal Medicine showed.

Ilke Sipahi, MD, Seden Celik, MD, and Nurdan Tozun, MD, all of Acibadem University Medical School in Istanbul, Turkey, took the FDA to task over a report published earlier this year that concluded bleeding rates for dabigatran (Pradaxa, Boehringer Ingelheim) were not higher than rates for warfarin. They based their analysis on insurance claim data and administrative data in the FDA Mini-Sentinel database.   

“The bottom line is observational studies are not reliable,” Sipahi told Cardiovascular Business. “Observational studies are open to many sources of biases and no matter what kind of statistical multivariable modeling, adjustments, propensity matching, etc., that you do, you still have unmeasured confounding. They are very much open to bias and they can produce misleading results.”

Sipahi, Celik and Tozun searched MEDLINE through July 2013 to identify randomized controlled clinical trials that compared dabigatran and warfarin and included gastrointestinal tract bleeding. Their meta-analysis used four trials that enrolled 26,076 patients. They found a significantly increased risk of bleeding with dabigatran (risk ratio 1.41).

The Mini-Sentinel analysis found a lower risk with dabigatran, with a gastrointestinal tract bleeding rate of 1.6 per 100,000 days at risk vs. 3.5 for warfarin.

“We believe that the large number of reported cases of bleeding associated with dabigatran provides a salient example if stimulated reporting,” the FDA team wrote. “In this case, such reporting provided a distorted estimate of the comparative bleeding rates associated with dabigatran and warfarin in clinical practice.”

Sipahi, Celik and Tozun acknowledged that postmarketing surveillance provides a mechanism for monitoring safety issues that may not have been examined during the approval process. But Sipahi questioned the value of using the Mini-Sentinel when the evidence from clinical trials designed specifically to assess efficacy and safety, including gastrointestinal tract bleeding, was already strong.

“If the outcome of interest had not been looked at in a randomized controlled trial then it might make sense,” he said. “Would that lead you to accurate results? I am not sure. Probably not.”

Sipahi cautioned physicians to approach findings from observational studies with caution. He pointed out that physicians now may choose from several other novel anticoagulants with other risk and benefit profiles.

“[Physicians] should look at the totality of the data and the totality of the mortality and morbidity, whatever regimen they pick,” he said, adding that he does not prescribe dabigatran in his practice.