Intended for healthcare professionals

Rapid response to:

Analysis

Dabigatran, bleeding, and the regulators

BMJ 2014; 349 doi: https://doi.org/10.1136/bmj.g4517 (Published 23 July 2014) Cite this as: BMJ 2014;349:g4517

Rapid Response:

I read with great interest your review[1] on the regulatory history of dabigatran in the United States and the European Union. It is important to realize that RE-LY never intended to prove whether monitored dabigatran was more efficacious and safer than unmonitored dabigatran, but whether two fixed doses of the drug were non-inferior to INR-adjusted warfarin. In this regards the study was indeed a “positive” one, and to the extent one views p-values as gauges of the strength of evidence, the “p less than 0.001” in RE-LY demonstrate that both doses were substantially non-inferior to warfarin. As you may be aware such low p-values may be used to “waive” the second study that is usually required for approval , a point was also made during the FDA advisory for dabigatran.

Though some secondary analyses of the exposure-response/exposure-safety relationships suggest a potential benefit for pharmacokinetic or pharmacodynamic monitoring of dabigatran it is important to realize that these analyses by both Boehringer Ingelheim (BI)[2] and the FDA[3] (available in the public domain since September 2010) are inherently limited by a number of technical issues. These include the failure to account for intra-individual pharmacodynamic variability, the competing risks of safety (bleed) and efficacy events, uneven discontinuation rates in the dabigatran arms v.s. the warfarin arm and informative missingness of drug concentrations in patients who have an event or dropped out from the study. Even though analyses that address these shortcomings are non-standard, they could improve on the BI models and the independent, nearly identical, ones developed by the FDA. In fact both these analyses are of low quality to merit anything more than a passing consideration. The quality of the BI modeling can be judged by the value of the c-statistic (0.715) quoted by the authors of the subsequent publication[4]. This value barely makes the cutoff for an acceptable predictive model, while the lack of validation in an independent dataset suggests that it would be even less reliable in other settings. Similarly, the FDA models are unable to match the event rates of the RE-LY dataset they were developed on, missing the mark by 6-87% (page 199[3]) depending on the prevailing level of renal function . Given the poor quality of these models I find it premature to use predictions made by them as a justification for regulatory action. Nevertheless, they may be useful hypothesis-generating instruments for the design of further studies of regulatory relevance.

In your paper you also suggest that the regulatory authorities need to agree with the BI on a therapeutic range for dabigatran. In that regards, note that the 48-200 ng/mL range quoted in your paper largely overlaps with the 10th-90th percentile range in the 150mg arm of RE-LY (40-215 ng/mL). In fact the lower end of this range is essentially BI’s estimate for the concentration below which one may expect a partial loss of efficacy(page 157[2]). On the other hand, the upper limit is the simple average of 215 and 185 ng/mL which is the expected concentration of a 300mg twice daily dose that per the FDA’s pharmacometric analyses would be the optimal one for dabigatran (slide 56[5]). Since this range is largely determined by the RE-LY data and given the substantial overlap in plasma levels in patients who had strokes and bleeding, it is likely that drug monitoring will not helpful for most patients receiving the 110mg or the 150mg dose. Hence it might be reasonable to forego drug level testing altogether and opt for pharmacodynamic monitoring e.g. with the Ecarin clotting time (ECT). Validation of this and similar tests against dabigatran concentrations measured by chromatography has been reported in the literature. Furthermore, such tests appear to be as good (or bad) as the dilute thrombin assay (HEMOCLOT), while being available in the US[6–8]. Finally note, that visual inspection of one of the unpublished, yet publicly available analyses, by the FDA suggest that ECT values in patients receiving dabigatran may less variable than the drug level while exhibiting similar or possibly better relationship to bleeding outcomes (Figure 1 found in page 27[3]).

Though important for clinical care, these technicalities should not obscure the major point about “dabigatran and the regulators”, namely the decision of the FDA NOT to approve the 110 mg dose despite BI’s proposal (and subsequent efforts) to do so. In my opinion this is not due to a myopic view exhibited by the FDA but rather a paternalistic attitude towards both patients and practicing physicians. Reading the transcript of the discussions between regulatory officials and external advisors regarding dabigatran one gets the impression that reaching this decision was like a pendulum swinging between the positions of “first do no harm” and “do more good than harm” (page 281[9]). In the end, the principle that “medical practice and medical research are designed to keep things getting better” dictated the decision to approve only the larger dose since it was superior to warfarin but only marginally so compared to the lower and apparently safer 110mg dose. In reaching this decision the FDA explicitly expressed a preference for fewer strokes over an excess of bleeding events. Note that this ranking of outcomes is not shared by many patients who opt not to be treated with warfarin and their physicians, or even the only patient in that advisory (himself on warfarin for atrial fibrillation).

The bigger question that merits an immediate answer is whose preferences about the trade-off between efficacy and safety should regulatory authorities take into account when evaluating novel drugs and devices? In this exceedingly multistakeholder environment for the provision of healthcare and drug development is it acceptable to “keep things getting better” based on the opinions of just regulatory officials and medical authorities without considering the values ofthe medical community at large and their patients? For dabigatran in particular, isn’t it time for the FDA to revisit their decision and approve the 110 mg in accordance with their own pharmacometric and BI analyses for our elderly patients?

Competing interests: No competing interests

05 August 2014
Christos Argyropoulos
Associate Professor Of Nephrology
University of New Mexico, Department of Medicine
901 University Blvd SE, Alqbuquerque, NM, 87114, USA