New post up at Medscape Cardiology: Are novel anticoagulants better than warfarin?

For the prevention of stroke in atrial fibrillation, the novel anticoagulant drugs dabigatran (Pradaxa, Boehringer Ingelheim), rivaroxaban (Xarelto, Bayer Pharma/Janssen Pharmaceuticals), apixaban (Eliquis, Pfizer/Bristol-Myers Squibb), and edoxaban (Lixiana, Daiichi-Sankyo) have been sold as both superior and more convenient than warfarin.

But is this true? More than 60,000 patients have been enrolled in randomized controlled clinical trials. Recently, two meta-analyses (studies that combine trials) have been published. That’s a bunch of data.

Screen Shot 2013-12-20 at 6.40.56 AMI had originally set out to explain how these meta-analyses had once and for all demonstrated the drugs’ superior safety and efficacy compared with warfarin. I wanted to know the details. The raw data. In the process of truth-seeking, I made a discovery that had me running around like Archimedes.

Here are two teasers:

  • hype is defined as extravagant or intensive publicity or promotion.
  • When comparing two treatments, there is a huge difference between absolute and relative differences. One obfuscates and the other tells the truth.

The column is up over at Medscape Cardiology. The title of the post is Novel Oral Anticoagulants vs Warfarin: The Truth is Relative.

JMM

Medscape asks you to register with an email. I think it’s worth putting up with these barriers. Not just for my work, but for the superb reporting and other editorial content. Of course, I am biased.

Comments

  1. Jeff says

    Contending with traffic on the way for the next of many finger sticks, I think:
    “Will that seaweed salad I had last evening show up?”

    It’s difficult keeping the vitamin K antagonism in bounds, which is where it has to be to keep the risks within bounds.
    The AHA recommended diet is strong on green vegetables. I’m guilty if I don’t follow it.
    Wait! Now I’m guilty if I do!
    This next INR will indicate if I’m at a bit of a higher risk for ischemic stroke or for a brain bleed – depending on my particular guilt.
    Meal planning becomes self-prescription and moral profile.

    Dr John, you haven’t addressed how long it might take before all this K minimizing calcifies my arteries to some “end point”. It’s a real issue. Worrisome. Did you miss that?

    Oh, and how much inflammation is all this emotional turmoil causing?

    I’m so grateful that I can afford rivaroxaban!

  2. Pete says

    Dr John- I enjoy your posts and appreciate what you do every day. Thanks. I hope my reply doesn’t sound too critical, I guess that is my writing style. This is one afib patient’s view. Well, I’ll take a chance. Here goes…

    You are right. As patients, we want to know the absolute risk of stroke. Most of us have been asking for straight answers since we have been diagnosed. We usually don’t get clear answers. Having said that, I believe your analysis doesn’t really show the true picture either. It doesn’t address what we really want to know: What is our ANNUAL ABSOLUTE risk of stroke in OUR INDIVIDUAL CASE and how much can that risk be REDUCED? We can weigh the costs of each anticoagulant and make our own decisions with our doctors, hopefully without government interference. I have some other concerns. But first, let’s look at the numbers.

    In the trial, they have grouped all afib patients together. But we are not a blended average. We are individuals with different risk factors and drug tolerances. Depending on our CHA2DS2Vasc score, we may have an ANNUAL (emphasis on the word ANNUAL) risk of from 0% to 15.2%. To be fair, let’s use a CHADS score of 1 (when anti-coagulants are considered) as the floor, with an ANNUAL stroke risk of 1.3%. There’s quite a difference between 1.3% and 15.2%. For those with risks 7% and up, stroke reduction percentages are significant. Further, they have grouped all the NOACs together. However, the trials for the 3 NOACs, currently approved, reduced stroke over Coumadin/ Warfarin by varying degrees: -34% P, -12% X, and -21% E. The difference between P and X (22%) is huge for those with moderate to high CHADS scores.

    So let’s get back to the study. Your summary of the study doesn’t reveal the CHADS scores of the patients. Let’s say the average fell right in between CHADS scores 1 and 2, which I suspect is the case. That would be an ANNUAL stroke risk of about 3.4% (average of 2.8% and 4.0%). That means, for the average of the group, not anti-coagulated, 96.6% would not have a stroke in one year. If they took Coumadin/Warfarin, their risk would drop about 70% (I’ve seen as low as 67% used), to about the 1.2% absolute risk in the study (98.8% were spared having a stroke). So, for that group, Coumadin/Warfarin only dropped the absolute risk by a mere 2.2%. But you wouldn’t tell patients to not take Coumadin/Warfarin because the absolute risk difference is only slightly over 2% would you? Of course not. You would look at the CHA2DS2Vasc score of the patient and weigh the benefit versus the bleed risk. So the ABSOLUTE numbers can be tricky.

    Now let’s look at the afib patient with a 15.2% ANNUAL risk of stroke. If the patient took Coumadin/Warfarin (70% reduction), they would reduce their ANNUAL stroke risk by 10.64% to 4.56%. And if the patient took NOAC P (34% more effective than C/W = 93.8% reduction: 1.34 times .70), the risk would be reduced to 3%. That’s over a 1.5% ABSOLUTE risk reduction (pretty close to a 2.2% number in paragraph 1 above, right?)

    With estimates of 4-5 million patients with afib and the expected “epidemic” of baby boomers that will get afib in the coming years, we are talking about tens of thousands of people that could be spared the devastating consequences of stroke. Also, it would reduce the costs of stroke that the anticoagulant comparisons never mention. Plus the costs of monitoring. And the dietary restrictions. And the issue of and cost of more brain bleeds (any brain bleed, a medicine-caused major complication often resulting in death, is a big deal). And then there those that can’t take Coumadin/Warfarin and who need alternatives. No, we’re not a blended average.

    I also have to mention that Coumadin/Warfarin has not been improved since it was first used for anticoagulation, almost 60 years ago. The NOACs are an improvement, both in efficacy and patient-friendliness. Those improvements come from research and development. That research and development needs to be supported by the products being integrated into practice. Why is this important? To further advances in treatment.

    I’m betting that the drug companies will virtually eliminate stroke risk for afib patients, before the medical field will virtually eliminate/cure afib. And, oh, the costs of ablation are astronomical. A couple (or more) lifetimes of the costs of the NOACs (granted, ablation improves QOL and potentially lessens stroke and heart failure risk). And, what? Ablation is “successful” (loose definition) only about 80% of the time and remission falls off by 8% each year? And patients usually still have to be on anticoagulants? I know, I know. .. research and development is improving on the percentages every day.

    That’s my point. More expensive, developing technologies have to be integrated into medical practice if treatment is to improve. No money, no incentive. The drug companies are competing fiercely for our anticoagulant business. They do that by improving quality and pricing (advertising too, I admit). I believe that is called the free enterprise system. Sure, there are savings in the short term to be had by using Coumadin/Warfarin for lower risk patients. But at a cost to advancement of the development of the anticoagulants. God, I hope my insurance company and the government don’t read this blog. Thanks for listening.

    • jane says

      Regarding your estimated 3.4% annual stroke risk – I doubt it. At the time when my husband was suffering atrial flutter, he had a putative CHADS2 score of 2. It was preached at us that his annual stroke risk was 4%. I discovered that this oft-quoted number derives from a study of Medicare patients whose minimum age was 65 and whose *mean* age was in the low 80s! My husband was at that time about 50. Most of the people in that study who had a score of 2 would have gotten one point from age, which is a more potent risk factor than any of the other one-point factors, and many of these people were VERY old. Those who had a score of 2 without having a point for age were still ca. 15-20 years older than the hubby, and stroke risk is said to double per decade of age. Therefore, if their annual risk was 4% there was no way that his was equally high.

      There are other issues with that number; for one thing, he was getting a point for hypertension even though his BP was drugged down to a pretty consistent 80/50, and there is no justification whatsoever in the Framingham numbers for treating people with such low post-treatment BPs as being at high stroke risk. For another, stroke rates are dropping over time, due to things like reduced trans fat consumption, so studies more than a decade or two old overestimate the stroke risk today’s citizens face. But the hubby’s -ologists could not seem to grasp any of the mathematical problems with applying the Medicare number to a 50-year-old; those numbers were sacred to them.

      By the way, he discontinued warfarin against “medical advice” and shortly got rid of his other CHADS2 “point” by, among other things, consuming a plant-based diet. Since all the involved specialists were devoted to the dogma that You Can Never Get Better, the idea that lifelong warfarin might sabotage his best hope of actually getting better was incomprehensible and irrelevant to them. I had tried to explain to a few different MDs that there is no such thing as a diet that is healthful, palatable, affordable AND invariant year-round – pick three; you can’t have all four – and they just did not get it.

  3. jane says

    You make excellent points in this column. Of course, the same consideration is true of the decision to treat at all. There are a subset of people with AF who get anticoagulation though their actual annual stroke risk was probably 1% to 1.5% – and for them, there’s less than a 1% chance that they will benefit from any treatment at all. (In fact, as you doubtless are aware, two studies have shown that warfarin does more harm than good in patients with a CHADS2 score of 0 and very little good in patients with a score of 1 – ignoring the fact that not all higher scores are equally meaningful.)

    If you have been scrutinizing these studies, I wonder if you might be able to comment on an issue I’ve wondered about. While warfarin and novel anticoagulants are clearly superior to low-dose aspirin, there’s little direct data comparing warfarin to high-dose aspirin, but some published evidence that high-dose aspirin is superior to low-dose aspirin in AF. Based on those limited data, Bayesian analysis suggests that high-dose aspirin might not be much inferior to warfarin. Yet this is routinely ignored and treated as unworthy even of further research.

    Now, the large study comparing apixaban to aspirin allowed doctors to dose aspirin or placebo aspirin as they chose (by giving multiple small pills), and only 6% got high (325 mg) or near-high doses. The authors chose not to address the issue or provide statistics, but those taking high-dose aspirin had what one might think was a clinically significant lower stroke rate than those on low-dose aspirin. The funny thing was that the people who were taking apixaban and placebo high-dose aspirin (3-4 dummy pills per day) also had lower event rates than people who were taking the exact same dose of apixaban and placebo low-dose aspirin (1 dummy pill per day). This permitted it to be said that apixaban was superior to aspirin whether the aspirin was low-dose or high-dose. But if the people taking apixaban and placebo high-dose aspirin had had the same event rates as those on apixaban and placebo low-dose aspirin, apixaban might not have looked impressively different from high-dose aspirin. I would kind of like to know not only whether the difference between the two aspirin groups was [statistically] significant, but whether the difference between the two apixaban groups was, because if so, it seems, pardon my paranoia, a bit JUPITEResque.

  4. Pete says

    Jane: Sounds good to me what you are saying. I had to use something to compare the NOACs with Coumadin/Warfarin, so the CHA2DS2Vasc numbers are as good as anything. One shortcoming of the CHADS system that becomes apparent to any thinking person is that all the risk factors are counted as 1 or 2 points… no 1.5, .8, or 1.2… so what’s the chance that every risk factor is exactly 1 or 2? Probably they aren’t all exactly the same. It is just a guide. I’ve also heard left atrial size and Utah score, neither included in CHADS, can affect stroke risk. And then they say that 35% of afib patients will have a stroke in their lifetime. That means 65% won’t. So you could “luck out” by doing nothing. I don’t think we are getting any guarantees here. Only guesstimates.

    • jane says

      I also find the selling of CHADS2-VASC iffy. Being able to dish out two points instead of one for age seems fine – but you can’t tell people with a CHADS2-VASC of 2 based on being 62 that they are at higher risk of stroke than the Medicare population with a CHADS2 of 1 who were all over 65! The European idea that every woman with AF should be on anticoagulants for life is also fishy given that the excess sex-associated risk is in large studies modest in the first year and non-existent thereafter except in the very old. And if you tell the woman with a CHADS2 of 2 that her risk is “even higher than” the standard line from a study using people of both sexes, you must also logically tell the male with the same CHADS2 score that he should presume his own risk to be lower than that figure But no doctor does that. The effort is to pile on more and more risk factors until everyone is at above average risk.

  5. Pete says

    Jane: I just realized I used the wrong chart. On CHA2DS2Vasc, a score of 1 is only a 1.3% risk. I should have said the score averaged between 3 and 4 in the trial. But now that I think of it, you are right. With CHA2DS2Vasc, they recommend considering an anticoagulant with a score of 1 (that carries a 1.3% risk) whereas it used to be (under CHADS2) that a score of 1 gave you a 2.8% risk. Because they increased the risk factors from 5 to 8, producing scores that max out at 6 and 10, there are more factors to get you on anticoagulants. I wonder if stroke reduction really is reduced 70-93% if your risk is only 1.3%. It’s kind of curious that the highest risk under CHADS2 is 18.2, and under CHA2DS2Vasc it is only 15.2, even with all the new risk factors.

    • jane says

      There was a study from a large California database in which, with just a slight weighting of hemorrhagic stroke to account for its greater severity, people with a CHADS2 score of 0 suffered net harm from warfarin – with only strokes counted, not non-stroke bleeds or the effects of coercion to keep eating the Standard American Diet. No risk reduction at all. For CHADS2 of 1, the reduction was pathetically small and not statistically significant.

  6. says

    John – I loved your post on Medscape about NOACs vs Warfarin! It is equally impressive reviewing the above detailed comments from your readers (!) – but I think it all comes down to an informed consent decision in which absolute risk, relative risk, convenience factors AND cost all come into play. To me – it helps tremendously knowing that for practical purposes (ie, within 99% absolute risk) – Warfarin and NOACs perform the same. To me – I also get VERY skeptical when Pharmaceutical companies who stand to make billions “spin” results. I therefore find it hard to trust what they state. I also wonder if as-yet-to-be-discovered side effects are awaiting down the road for the NOACs … That said – there seems to be no right and no wrong (which is clearly a different message than literature being put forth by those who stand to make billions).

    • says

      Jeff – I have reviewed the article by Azoulay et al that you refer to in the European Heart Journal – and my “take” is that I am VERY skeptical of the conclusion reached for a number of reasons. These include: i) This is not a randomized prospective trial – but rather a retrospective look at a large data base collected from another time period (1993-to-2008) in another country (UK) – without specification that I can see as to the dosing or the specific regimen for Coumadin that was used – nor even IF the study involved inpatients or outpatients !!! ; and ii) One or more of the authors is sponsored by makers of one of the NOACs (Novel Oral Anticoagulant Drugs).

      The write-up by Todd Neale in Medpage Today (that your link goes to) tells all. This is a quote from Todd Neale’s writeup: The results from the Azoulay Eur Heart J study “adds fuel to the fire that the newer novel oral anticoagulant therapies are superior based upon their mechanisms and based upon the speed at which they achieve therapeutic anticoagulation levels in patients with nonvalvular atrial fibrillation”. So – Keep in mind that there are BIG BUCKS to be made by “studies” such as the Azoulay one (that is sponsored by the maker of one of the NOAC drugs) which suggest more “problems” with Coumadin, ergo “preference” for use of one of the super-expensive new NOAC agents …

      As to the pharmacology involved – It IS possible to see a “hypercoagulable state” soon after starting coumadin IF the patient is actively clotting – and IF a higher initial dosing regimen is used (as used to be done in the past when 10-10-10 mg initiation regimens were routinely used). This is because the half-life of Factor VII is slightly longer than the half-life of Protein C. But the difference is only by ~2 hours – so risk of a “hypercoagulable state” would seem smalll (if not negligible) IF baseline risk of a stroke on any given day is low (ie, patient not actively clotting, as in chronic AFib) and IF Coumadin is more gradually started (as is the case with 5-5-5mg regimens used currently – but which were not used in 1993 …).

      BOTTOM LINE: It is reasonable (if not strongly advisable) to overlap initiation of Coumadin for in-patients with active thromboembolic disease (ie, pulmonary embolism, DVT, stroke) until the Coumadin is therapeutic and until any possible hypercoagulable period has passed. But there seems to be no need to do this for chronic AFib when Coumadin is started as an outpatient (which may be the most common situation). There is NO evidence to my reading from the Azoulay article that you cite at this point that convinces me otherwise.