warfarin — Journal Club (Erudition EM)

A brief rundown of the players:

Dabigatran (Pradaxa) is a direct thrombin inhibitor, FDA approved in October 2010 for the prevention of CVA in patients with Afib. In December 2011, the FDA initiated an investigation into reports of higher than expected rates of serious bleeding events. Rates of dyspepsia and GI bleed are also concerning, and a small but significant increase in MI has been shown in safety outcome data from multiple trials.

Rivaroxaban (Xarelto) is a direct Factor Xa inhibitor, FDA approved July 2011 for DVT prophylaxis, and in November 2011 for prevention of CVA in patients with Afib.

Apixaban (Eliquis), another direct Factor Xa inhibitor, will likely be FDA approved in June 2012 for prevention of CVA in patients with Afib.

This journal club examined the 3 major studies published in NEJM used to support the use of these novel anticoagulants in the prevention of CVA in patients with Afib. Interestingly, these drugs are struggling when studied in ACS; APPRAISE-2 (study of apixaban + antiplatelet therapy after ACS) was stopped early in 2011 due to increase in major bleeding without significant reduction in recurrent ischemic events. One other general note: in all three of these studies, ICH (the bleeding complication that counts) was significantly less with the study drug as compared to warfarin. All 3 of these drugs do ok in these non-inferiority studies-the important question is how they will perform in the real world, and it may be several years before safety conclusions and community effectiveness data give us the answer.

1. Connolly SJ, Ezekowitz MD, Yusuf S, et al. Dabigatran versus Warfarin in Patients with Atrial Fibrillation. N Engl J Med. 2009;361(12):1139-1151. (RE-LY)

Dabigatran 150mg BID or 110mg BID (blinded) vs Dose adjusted warfarin (unblinded)

Randomized, noninferiority trial of 18,113 patients, median follow-up 2 years.

Condition: AF within 6 months prior to randomization + 1 risk factor for CVA. Use of antiplatelets acceptable. Mean CHADS2 score 2.1. Intention to treat.

Primary endpoint: Stroke or systemic embolism

Warfarin 1.7%
Dabigatran 110mg 1.5% (p=0.34)
Dabigatran 150mg 1.1% (p<0.001 for superiority)

Major Bleeding

Warfarin 3.4%
Dabigatran 110mg 2.7% (p=0.003)
Dabigatran 150mg 3.1% (p=0.31)

ICH

Warfarin 0.74%
Dabigatran 110mg 0.23% (p<0.001)
Dabigatran 150mg 0.3% (p<0.001)

Conclusion: Lower dose dabigatran similar to warfarin for preventing stroke/systemic embolism and caused less major bleeding than warfarin. Higher dose dabigatran superior to warfarin for preventing stroke with similar rates of major bleeding to warfarin (trade off with dosing between benefit and bleeding). Both doses of dabigatrin with significantly less ICH than warfarin. Significantly higher rate of GI bleeding in higher dose dabigatran group as compared with warfarin.

Comments: Tony-higher discontinuation rate with dabigatran, likely from dyspepsia rates. What happens with the effectiveness of these drugs in the real world (external validity) when people don’t take them due to side effects, and there is no monitoring such as INR to reinforce compliance? Also likely to see less of a difference between the drugs in motivated patients on warfarin with very well controlled INRs. In real world, INR is maintained in therapeutic range only about 55% of the time. You can improve warfarin safety and effectiveness if you use point of care monitoring and have motivated patients. Brad-bottom line, concern about increased MI and GI bleed rates, and no easy reversal agent.

2. Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus Warfarin in Nonvalvular Atrial Fibrillation. New England Journal of Medicine. 2011;365(10):883-891. (ROCKET)

Rivaroxaban 20mg QD vs Dose adjusted Warfarin

Randomized double blind, noninferiority trial of 14,000 patients, median follow-up 707 days.

Condition: AF within 6 months prior to randomization + at least 2 risk factors for CVA. Mean CHADS2 score 3.5.

Primary endpoint: Stroke and systemic embolism

Warfarin 2.4%
Rivaroxaban 2.1% (ITT p=0.12 for superiority, p<0.001 for noninferiority)

Major Bleeding

Warfarin 3.4%
Rivaroxaban 3.6% (p=0.58)

ICH

Warfarin 0.7%
Rivaroxaban 0.5% (p=0.02)

Conclusion: Rivaroxaban non-inferior to warfarin in the prevention of stroke/systemic embolism with no significant difference in major bleeding events and less ICH.

Comments: Mark-patients in this study with higher CHADS2 scores-more like our patients. Also noted the increased risk for thrombotic events after stopping rivaroxaban. This is also something seen with clopidogrel (Plavix). Brian noted the heavily subsidized authors. Three statistical analyses used in this study: intention to treat, per-protocol and as treated safety analyses. Superiority was achieved only in the as treated safety analysis. It seems like a lot of data torturing to find a way to demonstrate superiority.

Statistics Rant about noninferiority studies. We commonly read superiority studies-attempts to show that drug A is better than drug B. Why do a noninferiority study? It’s a logical choice when you can’t use a placebo arm, and want to show drug A is as “good as” drug B. Also, choosing the noninferiority margin, or how close the results for drug A need to be to drug B in order to conclude that they are similar, is arbitrary. Pick a margin, and if you’re dealing with FDA approval, you just need to pick a margin that the FDA will swallow in order to approve your drug. For the apixaban study, if apixaban was shown to be at least 50% as good as wafarin, it was judged noninferior. In general the requirements for noninferiority studies are much more lax than for superiority studies, making them attractive to Big Pharma. Also, in the end, if you able to demonstrate and report superiority, your results looks even better. As far as analyzing results, in randomized studies trying to demonstrate superiority, intention to treat (ITT) is the best approach. With non-inferiority studies, it’s actually preferable to do both ITT and per-protocol analyses, as per-protocol analyses bias against demonstrating significant differences (is a stronger analysis). When in doubt though, as Erik always stresses, look at effect size (number needed to treat, absolute risk reduction), the comparison between benefits and harms, and expense…not statistics, and certainly not p values.

3. Granger CB, Alexander JH, McMurray JJV, et al. Apixaban versus Warfarin in Patients with Atrial Fibrillation. New England Journal of Medicine. 2011;365(11):981-992. (ARISTOTLE)

Apixaban 5mg BID vs Dose adjusted Warfarin

Randomized double blind noninferiority trial of 18,201 patients, median duration of followup 1.8 years. Mean CHADS2 score 2.1.

Condition: AF within 6 months prior to randomization + at least 1 risk factor for CVA

Primary endpoint: Stroke or systemic embolism

Apixaban 1.3%
Warfarin 1.6% (p<0.001 for noninferiority, p=0.01 for superiority)

Major Bleeding

Apixaban 2.1%
Warfarin 3.1% (p<0.001)

ICH

Apixaban 0.2%
Warfarin 0.5% (p<0.001)

Conclusion: Apixaban superior to warfarin in preventing stroke/systemic embolism with less major bleeding and lower rates of ICH.

Comments: Paarul-questioned the costs of these agents. Paarul and Pikul both felt the drug is promising, but have concerns about external validity, as there were significant inclusion/exclusion criteria. Dabigatran costs 8 $/day, warfarin is 30 cents. Need to factor in costs of INR testing, and as Mary pointed out, also costs of bridging with enoxaparin (Lovenox) if INR subtherapeutic. Christine pointed out that the study stressed relative risk reduction, which can sound good, but is less appropriate than using absolute risk reduction. Kelly wondered about the outcomes of the patients who bled-not discussed in the paper. Harwood pointed out that although results superior to warfarin, small effect size: NNT to prevent stroke/embolism 303, NNT to prevent a major bleed 104, NNT to save a life 238, and outside of carefully monitored study setting, real world effect size will likely be even less.

In my humble opinion, this is the strongest study of the three. Then again, apixaban isn’t even FDA approved yet-unlike iPhones, there are potentially big downsides to being the early adopter for new drugs…meaningful safety conclusions are still years in the future.

BONUS!! AVERROES - Apixaban in Patients with Atrial Fibrillation (this was a background article, but interesting as it compared Apixaban to ASA; relevant for patients who aren’t candidates for warfarin)

Apixaban 5mg BID vs ASA

5,600 patients

Condition: AF within 6 months prior to randomization + 1 risk factor for CVA

Primary endpoint: Stroke and systemic embolism

ASA 3.9%
Apixaban 1.7% (p<0.001)

Major Bleeding

ASA 1.2%
Apixaban 1.4% (p=0.33)

ICH

ASA 0.3%
Apixaban 0.4% (p=0.83)

Conclusion: Apixaban superior to aspirin with similar rate of major bleeding and ICH. Study stopped early due to benefit of apixaban over aspirin.

Our question was: given a generally healthy asymptomatic patient on warfarin for afib who is sent to the ED because of an elevated INR of 8 but who has heme + brown stool and a Hgb of 11, what is the appropriate intervention (hold warfarin only, or also treat with vitamin k, and if so, at what dose/route)? What are the risks of bleeding, and what are the thrombotic risks?

First, as a reminder, our patient has a CHADS score of 1. This represents an annual stroke risk of 2.8% for patients with afib (see below), so warfarin treatment is reasonable. Unfortunately, even in warfarin clinics, INRs are outside the therapeutic range one third to one half of the time. It’s well accepted that bleeding risk increases with increasing INR, especially when INR >4. What is our patient’s risk of bleeding?

Garcia DA, Regan S, Crowther M, Hylek EM. The risk of hemorrhage among patients with warfarin-associated coagulopathy. J Am Coll Cardiol. Feb 21 2006; 47:804-808

evaluated a cohort of 979 patients with a first episode of INR >5. During 30 day followup, 0.96% of patients with an INR >5<9 experienced major hemorrhage. When the INR >/= to 9, the risk of bleeding increased to 9.5%. Only 9% of patients in this study received vitamin k (mostly 2.5 mg or less). So, a good place to start: there’s about a 1% 30 day risk of major bleeding when your INR >5 <9. There were no ICHs in this study. In the Aiyagari background article on ICH/warfarin, a range of 0.3-0.6% per year risk of ICH while on warfarin is quoted, but as Harwood pointed out, the risk of ICH skyrockets in the elderly.

So, how should we protect our patient? The meta-analysis,

Dezee KJ, et al. Treatment of excessive anticoagulation with phytonadione (vitamin K): a meta-analysis. Arch Intern Med. Feb 27 2006; 166:391-397

evaluated all of the RCTs and prospective nonrandomized trials using vitamin K to treat patients without major hemorrhage with an INR >4 on oral anticoagulants, from 1985-2004. Unfortunately, as they admit, the quality of studies available was fair to poor, leading to a concern of “garbage in, garbage out.” There were a few limited conclusions. Subcutaneous vitamin K is no better than placebo. Just don’t use it (and this is also part of the 2008 ACCP guidelines). Oral and IV vitamin K worked about equally as well in this meta-analysis, but with the risks of anaphylactoid reactions with IV vitamin K, you should choose oral if the patient isn’t having major bleeding. Most (about 80%) patients with INR <10 who receive oral vitamin K (2.5 mg or less) will have an INR < 4 in 24 hours. That’s fine, but does giving vitamin K prevent bleeding? Does giving vitamin K cause an increased risk of thrombosis? Sorry, but no answers from this meta-analysis due to the fair/poor quality of the included studies.

So, our third article came out in 2009. It’s a large study of over 700 non-bleeding patients with INRs of 4.5 to 10, trying to look whether or not giving low dose vitamin K (1.25 mg) decreases bleeding risk.

Crowther MA, et al. Oral vitamin K versus placebo to correct excessive anticoagulation in patients receiving warfarin: a randomized trial. Ann Intern Med. Mar 3 2009; 150:293-300

At 90 days, this study demonstrated no significant difference in the rates of bleeding between patients receiving 1.25 mg vitamin K vs. those receiving placebo. The study was under-powered to detect a difference in major bleeds, although interestingly, more major bleeds occurred in the vitamin K group (likely chance, but the major bleeds all happened >30 days out-could there be issues with re-starting warfarin and maintaining a therapeutic INR after receiving vitamin K?). Although death and thromboembolism were rare and rates were balanced between the 2 groups, these were secondary outcomes and no firm conclusions are possible. Erik Kulstad made several important points about the methodology: there are issues of external validity (ability to apply study conclusions to your own patients): patients enrolled in studies are different-they are compliant, in general healthier, and in this case had amazing followup with multiple phone calls, likely identifying problems/bleeding more frequently than in an unselected population. Also, we are missing an explanation of who and how many patients were not included in the study-who are the ineligible and missed patients?

So, coming back to our patient….after much discussion, we took votes on 2 different patients: first, our PICO patient, but without heme positive stools. Everyone in the room agreed to send him home, and all except one would not give him vitamin K. When the patient had heme + stools (occult GI bleed, no melena, and stable/asymptomatic), still most would send him home, although now many would given low dose vitamin K (1-2.5 mg, orally). Again, as Harwood pointed out, you only need to change one variable to dramatically change risk and outcome and disposition (old, or melena, or poor social situation, or INR 11). Still, we do see a ton of patients like our PICO guy, and I believe this JC supports being more comfortable with doing less than many us (me included) may have done in the past. Interestingly, in the most recent ACCP guidelines from 2008, although the table looks similar to 2004, the doses of vitamin K recommended are half of the 2004 recs for patients without significant bleeding (both INR 5-9, and INR >9).

Last point: we chose the PICO patient deliberately: the heme + stools and hgb 11 means that he isn’t really exactly like any of the patients in our studies. There is no “right answer” on how to treat him. Involving the patient in his medical decisions is the “third circle” of EBM (evidence, clinical experience, and patient values). Maybe he is really afraid of a thrombotic stroke or MI. Maybe he doesn’t have insurance and isn’t going to get his INR rechecked tomorrow. You get the idea.

CHADS score for background:

	Condition	Points
C	Congestive heart failure	1
H	Hypertension (or treated hypertension)	1
A	Age >75 years	1
D	Diabetes	1
S₂	Prior Stroke or TIA	2

Annual Stroke Risk
CHADS₂ Score	Stroke Risk %	95% CI
0	1.9	1.2–3.0
1	2.8	2.0–3.8
2	4.0	3.1–5.1
3	5.9	4.6–7.3
4	8.5	6.3–11.1
5	12.5	8.2–17.5
6	18.2	10.5–27.4