Drug Interactions: Interaction Decisions: Using Evidence to Weigh Risks

Drs. Horn and Hansten are both professorsof pharmacy at the University ofWashington School of Pharmacy. Foran electronic version of this article,including references if any, visitwww.hanstenandhorn.com.

In 1877, the English mathematicianWilliam K. Clifford wrote a remarkableessay that outlived all of hismathematical work—it is called TheEthics of Belief. In the essay, he proposesa hypothetical story of a shipowner whose old and rickety ship wasabout to carry a group of emigrantsto the New World. At first, the shipowner thought perhaps he should havethe ship overhauled, but he knew thatthe repairs would be expensive. Themore he thought about it, the more heconvinced himself that the ship wouldprobably survive the trip. After all, ithad safely made many trips before, andProvidence would certainly watch overthese unhappy people seeking a betterlife. So he stifled his doubts, and asClifford says, the ship owner "got hisinsurance money when she went downin mid-ocean and told no tales."

Most people would have no difficultycondemning the ship owner for callousdisregard of the lives of the emigrantswho went down with his ship. Then,however, Clifford proposes a much moredifficult question: What if the ship madethat voyage, and many others as well,safely? Does that remove the guilt of theship owner? "Not one jot," says Clifford,because a decision is right or wrong foreverbased on the evidence available atthe time the decision was made.

At first, this may sound preposterous—that the outcome is irrelevant tothe soundness of a decision. Upon reflection,however, Clifford appears to havea point. Clifford's argument can also beapplied to drug interactions. Suppose apatient is stabilized on simvastatin, andthe prescriber decides to give the patient10 days of clarithromycin for an infection.Because clarithromycin inhibits CYP3A4(and P-glycoprotein), simvastatin plasmaconcentrations are likely to substantiallyincrease, leading to life-threateningmyopathyin some patients.

Let us assume, however, that thepatient in question has only a few minormuscle aches from the interaction, andthey subside quickly after the course ofclarithromycin. By Clifford's reasoning,the lack of a severe reaction is irrelevantbecause the prescriber had "no right tobelieve" the combination was safe, andthe patient was unnecessarily put at risk.(Azithromycin does not interact withsimvastatin and in most cases could beused as an alternative, or the simvastatincould simply have been discontinuedduring the clarithromycin therapy.)

The fact that most patients whoconcurrently receive simvastatin andclarithromycin do not have severe reactionsalso is irrelevant, because it is notpossible to determine ahead of timewhich patients will develop seriousadverse outcomes from the interaction.Therefore, even if severe myopathy onlyoccurred in 1 in 100 patients on the combination,it would be unwise to subjectpatients to this risk.

One could argue, therefore, that placingpatients at increased risk of anadverse drug interaction—when therisk is clearly avoidable—is ethicallyand professionally indefensible no matterwhat the outcome in a particularpatient. The fact that patients are regularlyplaced at such risk, however, ismore an indictment of our alerting systemsfor drug interactions, rather thanrepresenting lack of concern by prescribersand pharmacists. Hence, currentefforts under way to improve druginteraction alerting systems are a muchneeded step toward the goal of reducingadverse drug interactions.

Clifford concludes his essay with somestrong words: "To sum up: it is wrongalways, everywhere, and for anyone, tobelieve anything upon insufficient evidence."To believe, for example, that itis safe to give simvastatin with CYP3A4inhibitors—just because we have notpersonally observed any adverse effectsfrom such combinations—is not goodpractice. The published clinical evidenceexists, but it is not reliably reaching thehealth professionals who are making thedecisions. We need to improve both thedrug interaction detection systems andthe information provided to pharmacistsand prescribers if we are to reduce therisk of adverse drug interactions.