The tl;dr Pharmacy Journal Club: Rivaroxaban with or without Aspirin in Stable Cardiovascular Disease (COMPASS)
Editor's Note: Eric Long is P4 student from Ferris State University College of Pharmacy based out of Big Rapids, MI. He recently completed his Master of Business Administration (MBA) through the dual-degree program and will be receiving his PharmD. in May 2018. He is interested in the managerial side of pharmacy, especially process improvement and pharmacoeconomics. After graduation he plans to work in a hospital as a clinical pharmacist and eventually own and operate his own pharmacy.
Eric did a great job writing this journal club up, and we're super pumped to share it with you all.
This is our first ever journal club, so let us know what you think about it! Shoot me a line at mail@tldrpharmacy.com. If there's enough interest, we may do more of these journal clubs in the future.
Rivaroxaban with or without Aspirin in Stable Cardiovascular Disease (COMPASS)
WARNING: You are about to embark on a long adventure (with many words). If you realize you've made a mistake, just scroll to the end and look at the mini-outing, affectionately called ‘tl;dr’ to get a brief scoop on this article!
Think of journal club like base-jumping club, but you also determine p-values and confidence intervals at the same time! (Image)
Yes, you read that correctly, this is one of the dreaded things every student hates: Journal Club.
Why call it a club? I thought clubs were fun, right?
Apparently not this one, as I’m told. And you know why? Cause of an evil little thing called statistics!
I hate to break it to you but statistics will never leave your career. So, better to start adopting it and using its powers to make better clinical decisions.
So, yes, this is a journal club, but I'll do it with an entertaining spin. That's the first step of you coming around to Journal Clubs and *shudder* statistics.
Anyway, let’s start the first ever tl;dr pharmacy Journal Club with the COMPASS trial!
Background and Methods of the COMPASS Trial
At the heart of it, this trial is basically trying to challenge aspirin as the king of secondary prevention for cardio events. So these are patients with stable atherosclerotic cardiovascular disease (ASCVD).
The investigators are studying if rivaroxaban alone or in combination with aspirin is better at secondary prevention than our gold standard of aspirin alone.
We're dealing with a randomized, double-blind, double-dummy conducted across 602 centers in 33 countries involving 27,395 patients.
The patients were randomized with a 3-by- 2 partial factorial design. That’s all fine and good except "3-by- 2 partial factorial design?" Wait, what?
Let me explain (and this is a good time to plug the supplementary information. Chances are, if something doesn’t make sense in the trial, the explanation is somewhere in the supplementary information).
The COMPASS investigators are actually completing two studies at the same time.
The one we see right now is discussing Rivaroxaban (Xarelto) with/without Aspirin (ASA) for secondary prevention of cardiac events. But then on top of that they are studying the effects of pantoprazole on the treatment groups already involved.
In other words, patients received either Xarelto, Xarelto + ASA, or ASA. Then, those same patients either received pantoprazole or a placebo. See, two studies in one!
That means there will be 2 versions of the COMPASS trial. And it means that simply Googling "COMPASS trial" may give you the wrong study (trust me, I read the diabetic study of the ACCORD trial instead of the blood pressure portion and boy was I confused).
The very next thing we read is that the study is funded by Bayer.
Strange, huh? I mean, Bayer makes aspirin, so maybe that’s why? But Janssen produces Xarelto, right?
I’m still debating as to what is easier to understand: Europe or Janssen and Bayer’s relationship. (Image)
As it turns out, both companies collaborated to make Xarelto.
Oh, and Janssen is owned by Johnson and Johnson.
Does that clear things up?
Probably not. At least it doesn't for me. But now you know. You’ll win some brownie points and maybe look really cool for recognizing the difference.
If you want a great primer on all things anticoagulation, check this article out.
Study Population
Normally this section goes in the results. I get why (basically it doesn’t fit anywhere else), but it feels wrong still.
Knowing the demographics of the study population is key to extrapolating how well it will apply to others. I can’t go over every demographic marker, but I can give an overview of the plurality of patient characteristics.
As with most studies, there is no significant difference in characteristics between study groups. The population is mostly late 60's, white males who come from western Europe, Israel, Australia, or South Africa.
It is very worth noting that nearly 80% of the study population was male...and over 60% of them were caucasian. That's not necessarily a black eye on the study, but you have to keep it in mind when extrapolating the data out to your patient population. In a lot of medical literature, "white males" are very over-representative of the study population.
They are a little overweight and hypertensive (woah, didn’t see that one coming) and have had a past myocardial infarction and coronary artery disease (again, didn’t see that coming...remember, this trial is looking at SECONDARY prevention.).
These patients are also taking an ACE inhibitor or ARB, Beta-Blocker, and a lipid-lowering agent (aka: a statin. Please no gemfibrozil). You will probably recognize that as "standard of care" for anyone with ASCVD.
Eligibility
Patients must have met the criteria for coronary artery disease (CAD), peripheral artery disease, or both (again, this can be found in the supplementary appendix, but the definitions are pretty standard).
Those with CAD under 65 years of age also needed documentation of atherosclerosis involving two vascular beds or two additional risk factors. Essentially, they are trying to make sure these patients have "legit" ASCVD.
The exclusion criteria were:
High bleeding risk
Recent stroke or previous hemorrhage or lacunar stroke
Severe heart failure
Advanced stable kidney disease
Use of dual-antiplatelet therapy
Anticoagulation, or other antithrombotic therapy
Other non-cardiovascular conditions associated with poor prognosis per the investigators (unfortunately, we didn't get more information on this)
Methods
All patients were put through a run-in period, which isn’t stated in the artic….
Wait a minute!
This would be a great time to look at the supplementary information again, in which the run-in is stated as a 30-day period where they received a "placebo" rivaroxaban + ASA.
The idea of the run in was to weed out patients who were unwilling or unable to adhere to the study regimen. It's just another way to remove bias from the study.
It's also worth noting that some patients were enrolled quickly (4 - 14 days) after CABG surgery. These patients didn't go through the 30 day run-in, because the risk of clotting is highest then, and it would be unethical to not give them full anticoagulation.
The patients who properly completed the run-in period would then be moved onto the study itself. Randomized by 1:1:1 ratio, a patient would receive either:
Xarelto 2.5 mg twice a day and Aspirin 100 mg every day
Xarelto 5 mg twice a day alone
Aspirin 100 mg once daily
All of these treatment arms also had placebos involved so that the pill regimen couldn't be distinguished by the observer or the patients (remember, it's a double blind trial).
Patients were followed up at 1 and 6 months, and then every 6 months from there on out (until the trial was terminated early at 23 months...but we'll get to that soon).
Outcomes
As usual, there are about a million outcomes measured in a landmark trial like this. I'll focus on the primary outcome (as that's what the study is designed for), but I'll give a special shout out to the secondary and tertiary outcomes as well.
In this case, the primary outcome is a composite of cardiovascular death, stroke, or myocardial infarction.
The main safety outcome, on the other hand, is major and fatal bleeding via a modified International Society on Thrombosis and Haemostasis (ISTH) criteria. Those include:
Fatal bleeding
Symptomatic bleeding into a critical organ
Bleeding into a surgical site requiring re-operation
Bleeding that led to hospitalization (including presentation to an acute care facility without an overnight stay)
Of note, the investigators differed from ISTH criteria by NOT including bleeding that resulted in an Hgb drop of 2.0 or greater and bleeding that resulted in transfusion of at least 2 units of whole blood/PRBCs in the safety outcome.
BUT, the investigators used "any bleeding that required acute care facility or hospitalization" as a major event. It's pretty safe to say that any bleed requiring transfusion is going to fall into the category of "bleeding that led to hospitalization," which will therefore mean these patients were still included in the safety endpoint.
This is an important distinguishing point and you’ll see why soon.
Here's my cursory shout out to the secondary outcomes.
A composite of ischemic stroke, MI, acute limb ischemia, or death from coronary heart disease
A composite of ischemic stroke, MI, acute limb ischemia, or CV death
Death from any cause
As you can see. Those are nearly identical to each other.
And FWIW, the tertiary outcomes include
Individual components of the primary and secondary outcomes
Hospitalization for CV causes
Revascularization
Limb amputation
Stent thrombosis
Angina
Heart failure
VTE
Resuscitated cardiac arrest
New diagnosis of cancer
I think this is where that phrase “pick your battles” or something like that comes in handy.
Anyway, the net clinical benefit outcome (i.e. the combo of the primary objective + safety) was the composite of CV death, stroke, MI, fatal bleeding, or symptomatic bleeding into a critical organ.
A good rule of thumb to remember here is that if the primary outcome isn’t reached, then the secondary and tertiary outcomes are ignored.
Statistics
(Check out this super sweet overview of statistics if you need a quick refresher).
The first thing you are told in the statistics section is power. In our case, the power was set at 90%, which is a little higher than normal (which means this test is 90% likely to detect a difference if a difference truly exists. The gold standard of power in clinical research is 80%, so by going for 90% the authors are setting a more strict criteria.).
To reach power, 27,400 patients needed to be enrolled in order to detect a 20% reduction in the primary objective for the two Xarelto groups. Uh-oh, some of you may remember that only 27,395 patients were enrolled, so we didn’t achieve our power, did we?
Yes….and no. A quick and easy trick about power is to look at the p-value of the primary outcome. If it is proven significant, then that means power is irrelevant.
Editor's Note: That depends on who you ask. Most clinicians will say that "significant p-value means power is meaningless" as a rule of thumb. But if you ask a statistician about it (even ones that are specialists in clinical research) they would disagree. Check out the papers here and here to read more about it if you're interested).
You can think of power as a “planning” statistic insofar that it is useful to guide the study towards sound statistical analysis. Power is important but only when the primary outcome isn’t achieved.
There were a couple of interim analyses during this study: one at 50% and one at 75% of the total event outcomes achieved. The purpose of these analyses was to make sure that the study was still ethical, meaning the treatment group wasn’t doing so well (or so poorly) that it is now inhumane to withhold treatment from the control group.
To determine if the trial should be stopped early, a modified version of the Haybittle-Peto rule was used. This rule basically sets a higher burden of statistical proof (for example a p value of 0.001) for each interim analysis. If that level of difference is achieved in an interim analysis, then the trial is stopped. You can read more about it in the link above.
For this trial, they modified the Haybittle-Peto rule by changing the standard deviations required for each interim analysis.
Finally, both Kaplan-Meier and Cox hazard ratios were used (check this article out for a more thorough walk-through on hazard ratios).
Just remember this: if you are looking at hazard ratios or odds ratios, having a 1 in your confidence interval means that there is no increased or decreased chance of the event happening compared to the control group.
Or, said another way, it means the result is not statistically significant.
Or said another way, having a 1 in your OR or hazard ratio means that the result is no different than the control.
In addition, the lower the hazard ratio (i.e. below 1) the bigger the difference between the groups (and the more pronounced the effect).
As a quick reminder, remember that when you're NOT dealing with ratios (OR, hazard, etc...), having 0 in your confidence interval means the result is not statistically significant. Make sure you look for the difference in papers you read (and tests that you take).
Kaplan-Meier and Cox hazard ratios are similar in what they measure but there are some nuances between the two.
The way I think of it is that Kaplan-Meier shows the survival of one group versus another over a period of time and Cox hazard ratios look at specific variables that contribute to survival time and the likelihood of one group having that variable occurring versus another group.
Results
Let's get this out of the way right off the bat. They stopped this trial after 23 months because of superiority in the Xarelto + ASA group.
I'll talk about that more a bit later, but I just wanted to quickly address the elephant in the room.
As you might remember, the primary outcome in this study was a composite. Often in clinical literature, when you see a composite outcome, it's a fancy way of saying, “We don’t have enough statistically significant results in any one category, so we’ll lump them together and, voila, significance!”
That doesn’t sound right, when you first hear it, does it? But it depends on WHAT is lumped together. If you have outcomes that are related (such as stroke and MI), then it makes sense to do a composite outcome.
But if you had a composite outcome of less related disease states (such as stroke and UTI), then it would be a problem.
Another potential issue with composite outcomes is in how the authors present the data. Let's again say that we have some study where we are determining if Drug X can help prevent a composite outcome of "stroke and MI."
That's all well and good. But what if the treatment being studied did REALLY well at preventing stroke, but wasn't helpful at all for MI?
If the authors don't present the data for both subsets (or if you're not paying attention as a reader), you won't notice that minor detail. And that can have big implications on clinical practice. You'll end up using Drug X for MI prevention, when it really didn't do anything for these patients.
Your job as a reader of clinical literature is to make sure the endpoints studied make sense. Composite outcomes are useful (as long as the outcomes make sense together) because they can demonstrate a "real world" treatment effect.
Now that we have straightened out, does the composite outcome in this study make sense? The authors lumped cardiovascular death, stroke, and myocardial infarction together.
Nod of approval, thumbs up, the “go-ahead”…they all mean the same. I’m okay with composites, so here is a composite of thumbs ups *ba-dum- psh*(Image)
In my opinion, two of those are morbidities and one is a mortality, so the group isn’t completely homogeneous. That being said, they are related enough (and are commonly used together as composites in literature), so they get the nod.
Do I like using composites? Not really, because I would love to see significances for each individual outcome. But I understand the need for them.
Since we can justify that the composite is reasonable, we next look at the actual results.
In the Xarelto + Aspirin group, the primary outcome was met and was statistically significant (the hazard ratio was 0.76 with a p-value of < 0.001 and a confidence interval of 0.66 - 0.86).
In simple language, that means that we are 95% certain the true value of the hazard ratio in the Xarelto + Aspirin group, versus aspirin alone, lies somewhere between 0.66 to 0.86.
Let's go to even simpler terms (this is tl;dr pharmacy, after all). Remember that the hazard ratio is a measure of the effect of an intervention on some outcome over time.
Specifically here, we are looking at the rates of stroke, MI, or CV death over time in patients that take Xarelto + aspirin (and comparing those rates to patients that take aspirin alone).
And finally, remember that a hazard ratio of 1 means the 2 treatment groups are equal (because it's a ratio). This means that a hazard ratio of > 1 is a negative treatment effect (those patients do worse). And a hazard ratio of < 1 is a positive treatment effect (those patients do better).
Our hazard ratio is 0.76. So that means that, according to this study, taking Xarelto + aspirin is more effective than taking aspirin alone at preventing stroke, MI, and CV death.
HOW much better? That's what the confidence interval is for. Our confidence interval was 0.66 - 0.86. I like to subtract this number from 1 and convert it to a percent to get a "real world" version of result.
1 - 0.66 = 34%
1 - 0.86 = 14%
If you start a patient on Xarelto + aspirin for SECONDARY prevention, they will be 14 - 34% less likely to have a second stroke, MI, or CV death compared to if you gave them aspirin alone (which is our gold standard for secondary prevention). The median percent reduction is the point estimate for the hazard ratio (so 1 - 0.76 = 24%).
Okay, great! We’re done, right? Nope, but we’re getting close.
We have a significant primary outcome but we have to look at one more part: the adverse events. What good is a significant primary outcome if the person is 10x more likely to have severe internal bleeding?
Just as you may expect, adding an anti-coagulant will increase your bleeding risk more than with just aspirin alone.
Specifically, patients taking both Xarelto and aspirin had a 70% increased chance of major bleeds (as defined by the authors) compared to ASA alone. Of course, that 70% is relative risk. The absolute risk of major bleeds was 1.9% in the ASA group and 3.1% in the Xarelto + ASA group.
Editor's note: Honestly, I'm just excited they even included absolute risk in their results. Most often you have to calculate that for yourself.
With that in mind, I’m not too worried about the bleeding rates for a few reasons.
One, this study used a very broad definition of major bleeding. On top of the ISTH definition (minus the blood transfusions or the Hgb dropping by 2.0 or greater), the authors decided that if a patient went to the hospital, then they had a major bleed.
Now come on! That means that even if they hit their head and went to the ER, had a quick negative CT scan, and went on home in time to watch Top Chef, that they were still included as "major bleeds."
And even with that, the absolute risk of major bleed for rivaroxaban + ASA was 3.1% compared to 1.9% for ASA alone. So just over 1% of an increase.
Stomachable? Shakespeare would be proud! (Image)
Secondly, the only subtype of the major bleeding category that had significance (i.e., more likely to happen in the Xarelto + ASA vs ASA alone group) was the super ambiguous ‘other major bleeding’.
All-in-all, the side effect profile isn’t perfect, but within the range of the scientific term “stomachable”.
Strengths and Weaknesses
This is probably my favorite part of any journal club! It’s literally a license to say why the study was bad and why you shouldn’t trust a single part!
Well, okay, maybe it’s not that cynical, but it is a good time to take a look at the study and see what worked well and what did not work well.
The “low-hanging” strengths of the fruit tree are the easy ones: *monotone, drab voice*.
It was a:
multi-center
double-blind
double-dummy
randomized control-trial
that utilized a run-in phase
had a huge number of participants
and aimed for very high power
Whew, got that out of the way.
These aren’t bad strengths (actually, they're VERY good strengths), they’re just the go-to for almost any example of clinical literature. Strengths that are unique to each study are the fascinating ones to look at.
In this case, the author’s definition of major bleeding was overly broad. I thought it was great they used such an all-encompassing (and possibly overly-cautious) definition of major bleeding.
Very often in literature, you'll see this profound clinical effect on the primary endpoint, but the safety data is much less desirable. This study was almost "too" careful and conservative regarding bleeding events, and that's a good thing.
Because while it would be great so see rivaroxaban + aspirin improve our outcome in secondary prevention, it doesn't help us if all of the patients are hospitalized and/or die from major GI bleeds.
Of course, there are weaknesses to this trial. Yes, it was funded by the drug company. This isn't inherently bad, it's just something to be aware of.
The key to determining how much of a weakness Bayer's involvement is is to figure how exactly HOW involved Bayer was. Many trials are funded by the drug company in that they only provide research drug and maybe some additional money for other research costs. As long as the authors are transparent about this, it's "barely" a weakness.
This study, however, saw a lot more involvement from Bayer. We are lucky that the authors were forthcoming about this, because often they'll just say "This study was funded by INSERT DRUG COMPANY HERE" and leave it at that. To their credit, the authors of this study mention that Bayer representatives were part of the steering committee that "were responsible for the development of the protocol...and for the conduct and oversight of the study."
That basically means that Bayer (probably) had a hand in every step of the study analysis. Patient selection criteria, data collection, data interpretation, statistical analysis, and so on. Even though an independent committee reviewed everything, that opens the door for a MUCH bigger insertion of pro-Bayer bias.
Again, this is not an accusatory statement. It's just a thought to keep in the back of your mind when reading studies like this.
Another issue with this study is in the study groups. Remember, there was
Xarelto 2.5 mg BID + ASA 100 mg Daily
Xarelto 5 mg BID
ASA 100 mg Daily
The issue is that we don't readily have 2.5 and 5 mg tablet strengths of Xarelto available in the United States, nor do we have 100 mg of ASA.
So unless (and until?) Bayer makes a new formulation of Xarelto (possibly co-formulated with 100 mg of aspirin), it will be really hard to generalize this study to US patients.
Of course, a quick glance at the Bayer pipeline shows a cryptic Phase 3 study of rivaroxaban in preventing major cardiac events, so maybe we have that coming our way soon. But, how much will it cost?
Discussion
Finally, let’s put it all together. Good study? Bad study? This is where you get to really interject your opinion and show what you value in a good study, which is always bound to be different from person-to-person.
Overall, I thought it was a pretty good study. It was a well-designed and showed promising benefit to adding on Xarelto 2.5 mg BID to a typical once-a-day ASA (so much that they even stopped the trial for superiority in the Xarelto group after 23 months of follow-up). Also a sample size of over 27,000 people ain't bad.
The primary outcome was achieved while not significantly increasing the risk of major bleeding beyond something within the realm of acceptable for a person on an anti-coagulant and anti-platelet at the same time.
However, even though I like the study, I'm not sure it has a sustainable place in therapy right now. From a pharmacoeconomic perspective the implications are huge. Xarelto is already an expensive drug and like I mentioned above, they don't make a 2.5 mg or a 5 mg version.
Will Bayer look to extend the patent with a new product? I’m not sure, but I venture to guess "probably."
But even if the new indication is warranted, the number-needed-to-treat (NNT) is quite high at 77 patients.
Hold on, what’s a NNT? Fortunately, there is an article here that goes into that. But in brief, 4.1% of patients in the study group compared to 5.4% of patients in the control group hit the primary outcome. That's a difference of 1.3%.
Absolute Risk Reduction = 1.3%
NNT is the inverse of ARR.
And therefore, NNT = 1 / 0.013 = 76.9 patients = 77 patients (because you just can't have 9 tenths of a patient).
That means 77 patients will need to be treated with twice a day with Xarelto 2.5 mg (in addition to 100 mg of ASA daily) for about 23 months to avoid one additional case of CV death, stroke, or myocardial infarction (and any one of those events can be fatal, so death might not even be prevented).
And, because this is secondary prevention, the patients won't just be taking it for 23 months. They'll be "lifers," taking this medication all the way until their twilight years.
In addition, the confidence interval is sort of wide at 0.66-0.86. As I said above, that means that the hazard ratio could be anywhere between 14 and 34% depending on your patient.
And let's highlight again that this trial over-emphasized white males (by a lot). So if your patient is an African American, or a Southeast Asian, or a female, do these results still apply?
And all of that is BEFORE we talk about how much Xarelto costs (and before we talk about the fantasy 2.5 mg strength of Xarelto that doesn't exist in the US...or the 100 mg strength of ASA that doesn't exist in the US).
A bottle of 30 Xarelto 10 mg tablets costs around $450. Even with patient coupons and assistance programs, SOMEONE is paying for that.
And a final note...let's talk about compliance.
In the control regimen, you have patients taking 1 pill of ASA once a day.
In the study regimen, you have patients taking 3 total pills per day (1 tab of Xarelto BID, and 1 ASA daily). Depending on who your patient is, that pill burden may lead to compliance issues.
I mean, they felt it necessary to do a 30 day run in period to qualify the patients that ended up being enrolled in the study. That's not exactly a "real world" scenario, is it?
COMPASS Trial Conclusion (aka: tl;dr)
So, in conclusion...
This study was well designed and conducted. It's a good study, it really is.
It's not the first time we've looked at different alternatives to secondary prevention than aspirin alone. And frankly, it looks like this is a reasonable choice. The trial showed clear efficacy, and they still hit acceptable safety measures even though they used a (possibly) overly strict criteria of "major bleed."
Yet...
Real life is not a clinical trial. And there are going to be cost and compliance issues with this proposed regimen of Xarelto 2.5 mg BID + ASA 100 mg daily.
It seems like there is a clinical benefit worth noting. But does that make it right for your patient? That'll be up to you to decide as a clinician.