Monday, May 15, 2017

Making Decisions with the Evidence: A Patient on DAPT after a Drug Eluting Stent is Admitted with Bleeding

One of my frustrations from the last post was the dichotomization of the results of trials to either state that the therapy "works" or "doesn't work".  Even ignoring the nuances of the statistical design and the experimental structure and assuming that the results are true, very important information is lost in the dichotomization that is essential for making decisions about the therapy in question.  Namely:
  • the patient population that was included/excluded in the trial
  • whether the primary outcome is one that is meaningful (trials of fibrinolytics for complicated parapneumonic effusions are bedeviled by this problem)
  • the magnitude of the absolute difference in outcomes (positive utilities)
  • the magnitude of the absolute difference in costs, side effects (negative utilities)
You can see that what I'm trying to do here it apply expected utility theory (EUT) to medical decisions, something that I have tried to do, following the example of others, for over a decade now.  It is oftentimes a fraught endeavor because assigning values to the outcomes, especially relative to one another is a very difficult task.  However, in some cases the data can provide reassuring evidence that a particular course can be taken without excessive worry.

Recently I admitted a patient 9 months into a 12 month course of Dual Anti-Platelet Therapy (DAPT) after drug-eluting coronary stent placement.  He had bleeding that abated with 12 hours of observation and holding DAPT (ASA 81 mg and clopidogrel/Plavix).  The question of course is, do we restart DAPT, and if so when, and what are we losing in terms of the benefit of DAPT during the hiatus?


Fortunately, excellent data are available to guide this set of analyses and decisions.  The DAPT trial from the NEJM in 2014 enrolled almost 10,000 patients and randomized them to DAPT for 30 days versus 90 days.  The co-primary outcomes were - well wait.  You can't have a co-primary outcome.  You cannot go to Vegas and place bets on two numbers and expect to pay for only one bet.  There are several ways to deal with this as a reader of the literature and none of them is satisfactory.  In this case I think we have an out.  The first outcome is "stent thrombosis" which appears to be not very patient-centered (the one that patients care about), but the second one, a composite of MI, stroke, and death is patient-centered.  The absolute difference in stent thrombosis between the groups was 1%, and highly statistically significant (I'll leave clinical significance for you to judge - the study was funded by more than 8 drug and stent manufacturers).  The difference in the composite second co-primary endpoint was 1.5% and also statistically significant.  When we dissect the composite we find that it was driven entirely by MI, with a 2.0% reduction, also individually statistically significant.  So, it is reasonable to infer that the outcomes are being driven by stent thrombosis and related myocardial infarctions.  The results are neat and convenient - we can use a 2% annual reduction in myocardial infarction as out estimate for the patient-centered benefit of DAPT.

Assuming that the benefit of DAPT is constant over time (it probably declines over the first year), our patient, who was stented 9 months ago, has captured 75% of the 2% benefit, or 1.5%, leaving 0.5% remaining.  For the remaining 3 months, his daily risk of MI is increased by 0.000056 as a result of being off DAPT.  I think we can agree that this maximal upper estimate (given the expected decline of benefit of DAPT with time) is one that we can safely neglect for several days or weeks.  Indeed, after a bleeding episode leading to hospitalization, considering these numbers, it would be reasonable to say "you have captured the majority of the benefit of DAPT and can now stop it altogether."  Or, alternatively, one could restart the clopidogrel in a week or so with the a priori plan to stop it altogether if any further bleeding occurs.

In addition to the use of simple math to evaluate the propriety of a treatment course, this case highlights the issue of "competing risks".  The original risk, of stent thrombosis/MI at the time of insertion took the highest priority and outweighed the risk of bleeding.  Nine months later, most of that risk is gone, while the cumulative risk of clopidogrel continues and is consummated.  All therapeutic decisions must be evaluated on the basis of competing risks of omission and commission, and periodically re-evaluated because the relative risks and benefits change over time.
 

No comments:

Post a Comment