GI Bleeding on Dabigatran: Controversy Pits Meta-analysis vs Real-World Data

November 18, 2013

CHICAGO, IL — Randomized trial evidence suggests that treatment with dabigatran etexilate (Pradaxa, Boehringer Ingelheim) raises the risk of gastrointestinal (GI) bleeding compared with treatment with warfarin[1], suggests a meta-analysis, with findings entirely opposite those of a recent analysis of dabigatran and warfarin use in the FDA Mini-Sentinel database by several employees of the agency.

The report is published online November 18, 2013 in JAMA: Internal Medicine.

In the new meta-analysis of four dabigatran trials with enough available GI-bleed data, the risk of such events was raised 41% (p<0.001) with the new oral direct thrombin inhibitor compared with the vintage vitamin-K antagonist. The risk increase persisted after dropping any one trial from the meta-analysis, including the RE-LY trial, which overwhelmingly contributed the most patients and events, according to the authors, led by Dr Ilke Sipahi (University Hospitals and Case Medical Center, Cleveland, OH).

The RE-LY trial is well known to have shown more GI bleeds with dabigatran, while intracranial hemorrhage was more common with warfarin (p<0.001 for both differences).

The increased GI-bleed risk in the meta-analysis contrasts with the 54% risk reduction seen for dabigatran vs warfarin in the Mini-Sentinel data, taken from electronic health records routinely monitored by the FDA covering a 14-month period ending at the close of 2011. But that analysis, recently published in the New England Journal of Medicine and reported then by heartwire , didn't adjust for potential confounders.

Sipahi bashed the very publication of those findings. "It is actually very disturbing, because the FDA knows that observational studies are unreliable." Here, the agency is offering observational data to assess a drug's safety: they suggest fewer GI bleeds, yet they are contradicted by "the very robust randomized-trial data about GI bleeding on dabigatran," he said.

On one hand, said Dr Elaine M Hylek (Boston University Medical Center, MA) to heartwire , "I think observational studies are critical, because they give us a glimpse of what is occurring in clinical practice." They can complement randomized controlled trials "if there isn't a lot of selection bias in patients who are preferentially being prescribed a new drug." But her sense, she said, "is that the initial wave of dabigatran use was not [in] a random selection of people." Probably they were lower-risk patients and "expected to have a lower risk of bleeding and strokes."

Strong predictors of elevated bleeding risk in oral anticoagulant trials include older age, prior stroke, and renal dysfunction, according to Hylek. "I have no doubt that there's bleeding with dabigatran. I just think that kidney function must be playing a role there," she said.

"Until we know that there is a greater universe of patients being offered and started on this drug, I think it's going to be hard to definitively conclude that there are discrepancies between the trials and clinical practice."

Sipahi agreed that a "healthy-user bias," including a lesser prevalence of renal dysfunction, may underlie the apparent GI-bleeding risk reduction with dabigatran in the Mini-Sentinel analysis.

His group's meta-analysis included four randomized controlled trials from the literature and FDA and commercial databases that encompassed about 26 000 patients.

Relative Risk (95% CI) for Gastrointestinal Bleeding Events on Dabigatran vs Warfarin, Meta-analysis and Component Trials

End points Dabigatran events (n) Warfarin events (n) RR (95% CI)
Meta-analysis 1387 528 1.41 (1.28–1.55)
RE-LY 1281 452 1.41 (1.27–1.56)
RE-COVER 53 35 1.50 (0.99–2.29)
RE-MEDY 5 8 0.62 (0.20–1.90)
RE-COVER II 48 33 1.47 (0.95–2.27)

In the Mini-Sentinel database, the rate of GI bleeding was 1.6 vs 3.5 events per 100 000 days at risk for dabigatran and warfarin, respectively.

The Mini-Sentinel system "is especially useful for identifying adverse effects that might not be apparent in randomized clinical trials because they are rare, occur in patient groups not included in the trials, occur when used in settings less controlled than randomized trials, or occur in patients taking the medications for periods of time longer than the length of the trial," writes Dr Mitchell H Katz (San Francisco Department of Public Health, CA) in an Editor's Note following the Sipahi et al report[2].

According to Katz, using data from electronic health records "is a smart and efficient method of learning more about medications in real-world settings. However, new data, especially from observational studies, which are prone to confounding and underreporting, must always be judged in the context of biologic plausibility and other data sources."

Sipahi said that "the important message here is that new is not always better in medicine; and when a new drug is discovered, we primarily focus on the good things. Everyone was very excited about the reduced thromboembolism rate with dabigatran 150 mg in the RE-LY trial and focused on that." But one of the costs of using dabigatran at that dosage is an increased GI-bleeding risk, he said. "I think that is extremely important, because we have alternatives to dabigatran."

Neither Sipahi et al nor Katz had disclosures. Hylek disclosed being on the executive steering committees of the ARISTOTLE trial sponsored by Bristol-Myers Squibb and Pfizer and the ORBIT AF registry sponsored by Janssen Pharmaceuticals and being on advisory boards for Bayer, Boehringer-Ingelheim, Bristol-Myer Squibb, Daiichi Sankyo, Johnson & Johnson, Pfizer, and Roche.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.