Primary Care Corner with Geoffrey Modest MD: Rivaroxaban, Is It Really Better Than Warfarin for Non-valvular Atrial Fib?

By Dr. Geoffrey Modest

The BMJ spotlight team (actually Deborah Cohen, associate editor) just released another investigation into poorly-regulated drug-company-supported studies which have led to large changes in clinical practice and large profits for drug companies (see BMJ 2016;352:i575). This one is a challenge to the ROCKET-AF study (see N Engl J Med 2011;365:883-91), which was the single study leading to FDA approval for rivaroxaban, a​ direct thrombin inhibitor, by showing that although it was as good as warfarin in patients with non-valvular atrial fibrillation in preventing strokes, there were fewer episodes of intracranial and fatal bleeds. Of note, rivaroxaban is now the world’s best-selling new anticoagulant. However, it turns out that the ROCKET-AF study relied on defective point-of-care INR machines which gave falsely low INR levels, thereby leading the researchers to increase the warfarin dosing and perhaps predisposing patients to more bleeding.


  • BMJ found that the INR devices used had been recalled in December 2014 because the FDA issued their highest level recall notice finding that these devices could give “clinically significantly lower” INR values. The device manufacturer Alere had received 18,924 reports of malfunction and 14 serious injuries, dating back to 2002 (prior to the ROCKET-AF trial starting)
  • September 2015: BMJ asked ROCKET investigators about the device recall. No direct response, just one from the drug company that they were unaware of the recall.
  • The European Medicines Agency (EMA, the European equivalent of the FDA) in April 2015 did not know these devices were used in the ROCKET-AF study, and called for an independent investigation into direct oral anticoagulants. The company claimed that the results of ROCKET-AF were still valid after conducting some sensitivity analyses. No independent investigation has happened to date.
  • But, as per Harlan Krumholz, a cardiologist at Yale, “the study should be considered of uncertain validity until a more thorough review can be  done”, and to have “an investigation by an independent group of experts to quickly determine if there are grounds for retraction [of the study]”.
  • In looking back at the original FDA approval for rivaroxaban, 2 primary reviewers had expressed concern about the warfarin control and recommended that the drug not be approved, stating “ROCKET provides inadequate information to assess the relative safety and efficacy of [rivaroxaban] in patients whose warfarin administration can be well-controlled”
  • BMJ asked the ROCKET researchers if those patients who had major bleeds had had laboratory INRs checked. None responded.
  • It turns out that at the 12-week and 24-week time-periods, there were simultaneous measurements of INR by device and a central lab, though this data has not been released.
  • There was a rebuttal by the investigators to the above BMJ charges (see DOI: 10.1056/NEJMc1515842). The investigators did post-hoc analyses using the FDA recall criteria [medical conditions that the FDA thought might lower the INR value of the device: anemia with hematocrit >30%, and conditions which raised fibrinogen levels (acute inflammation, chronic inflammatory conditions, severe infection, cancer or end-stage renal disease]. Results of the post-hoc analysis were mixed:
    • ​In those without recall conditions: no significant difference in strokes with rivaroxaban vs warfarin; more intracranial hemorrhages, fatal or critical organ bleeding with warfarin; more GI bleeding overall with rivaroxaban
    • In those with recall conditions (and presumably on unnecessarily higher doses of warfarin): decreased stroke in those on rivaroxaban but only in the per-protocol patients; no difference in intracranial hemorrhages, fatal/critical organ bleeding; more GI bleeding with rivaroxaban.

So, these articles bring up several issues:

  • It is a little unsettling to me that rivaroxaban was approved based on only one drug company-sponsored study
  • I’m not sure what to make of the post-hoc analysis by the investigators. Surprisingly, they seem to show fewer adverse events in those without the FDA-designated recall conditions, though interestingly in those patients with those conditions, rivaroxaban​seemed a bit worse on clinical bleeding. But are those recall conditions complete? Many older studies, for example, found a relationship between smoking and high fibrinogen levels (to the point that many posited that the main mechanism of smoking causing heart disease was mediated by fibrinogen, and that stopping smoking quickly reversed much of the increased CAD risk because smoking’s effect on fibrinogen levels reversed so quickly). Smoking was not on the recall list. Are there other unknown conditions which lead to aberrancies in the device INRs?Or are there other unknown biases which can happen in a post-hoc analysis? It would be interesting to know the relationship between the laboratory INR measurements at the 12 and 24-week times and the results of the point-of-care devices. And what the correlation was on an individual basis between these discrepancies and the positive and negative clinical outcomes. And, I would strongly support an independent review.
  • A prior BMJ issue and several other studies have suggested that dabigatran also may have many more adverse events than reported and that there were drug company shenanigans; in particular the drug company covered up evidence suggesting that there needed to be blood monitoring, despite the main selling point for the drug being that no drug monitoring was necessary, as opposed to warfarin and the need to follow INRs, (see As a result of the BMJ and other articles about dabigratan, at the end of 2015 both the EMA and FDA held meetings to see if there was a need to monitor blood levels of direct anticoagulants and adjust the dose as necessary to maximize benefits/minimize harms.
  • This raises the issue that the FDA historically has been less aggressive in regulating devices than meds. Typically, once one device is accepted, there is much less overview on similar products (new devices just need to be “substantially equivalent” or similar to ones already on the market). And, the issue of rivaroxaban highlights that approval of meds may well require studies using not-so-well-regulated devices which could potentially dramatically affect the results. And, this is compounded by the selection bias favoring the drugs: positive studies of drugs are published at a much higher rate than negative studies. And once a drug company has a positive study (as with ROCKET-AF), there is little incentive to do further confirmatory studies after a drug is approved.
  • So, yet again, I am very hesitant to prescribe these newer meds, given huge conflicts-of-interest, the large numbers of cases of drug companies hiding results they do not like, the decreasing authority and efficacy of the FDA in monitoring new drugs (e.g., see ), and the increasing reliance on post-marketing surveillance, which seems to happen only pretty rarely (e.g., see ). I continue to use warfarin as my preferred anticoagulant, and I use the close monitoring of INRs (which in rock stable patients, can actually be done on an every 2-3 month basis) as a means to check-in with these patients, who are typically medically-complex anyway.

(Visited 24 times, 1 visits today)