Apixaban drug interactions: the list, the dose rules, and what changes with the other DOACs

Clarithromycin is the one that reaches the labeled threshold, because it is a strong inhibitor of both CYP3A4 and P-glycoprotein, which is exactly the combination apixaban labeling singles out. Erythromycin inhibits CYP3A4 but is not classified as a strong dual inhibitor, and azithromycin is not a meaningful inhibitor at all, which makes it the straightforward choice when a macrolide is needed in an anticoagulated patient.

The wider point is that the warfarin reflex does not transfer. Trimethoprim/sulfamethoxazole, metronidazole and the fluoroquinolones dominate warfarin interaction lists because of CYP2C9 and gut flora effects, and none of that machinery applies to apixaban. What does carry across is bleeding risk from the acute illness itself, from reduced oral intake, and from any NSAID taken for the associated fever or pain.

Apixaban is cleared by two independent routes: CYP3A4 metabolism and P-glycoprotein efflux in the gut wall. Blocking one leaves the other working, and exposure rises only modestly. Blocking both at once removes the redundancy, and exposure rises far more than either mechanism alone would suggest. That is why the labeling is written around strong dual inhibitors specifically rather than around CYP3A4 inhibitors generally.

The practical consequence is a frequent source of unnecessary alarm. Diltiazem, verapamil and amiodarone all appear on interaction screens with apixaban, and none of them is a strong dual inhibitor, so no dose change is directed. A clinician who reduces the apixaban dose every time an interaction alert fires will under-anticoagulate a lot of patients. Knowing which alerts carry a labeled action and which do not is the skill the alert itself does not supply.

Rifampin, carbamazepine, phenytoin, phenobarbital and St John's wort induce both CYP3A4 and P-glycoprotein, and the result is roughly a halving of apixaban exposure. This is the more dangerous direction, and the less noticed one. There is no INR to fall, no symptom to report, and no alert after the prescription is written. The patient simply becomes less anticoagulated than everyone believes, and the first sign is a stroke or a clot.

The timing compounds it. Induction takes several days to develop and two to four weeks to wear off, so the vulnerable window extends well past the last dose of the inducer. When a rifamycin is genuinely required, for tuberculosis or a prosthetic joint infection, the anticoagulation strategy usually needs to be rethought rather than adjusted, and that decision belongs to the treating team against the current labeling.

Almost entirely. Warfarin is a CYP2C9 substrate whose effect also depends on vitamin K supplied by gut flora, so azoles, sulfonamides, metronidazole, macrolides, fluoroquinolones, diet and any antibiotic that disturbs the microbiome all move the INR. Apixaban touches none of that. It is a CYP3A4 and P-gp substrate with a short half life and no dietary vitamin K dependence, so the interaction list is an order of magnitude shorter. Our full breakdown of the older drug is on the warfarin drug interactions page.

Two things do carry across identically. Pharmacodynamic bleeding risk from NSAIDs, antiplatelets and SSRIs is the same on either drug, and here there is no routine coagulation test to provide even false reassurance. And renal function matters more with the DOACs than with warfarin, because it is a direct dosing input rather than a background consideration. An interaction check that ignores creatinine clearance is checking half the question.

DOAC safety is really three questions asked at once: is there an interaction that changes the dose, is there a pharmacodynamic bleeding risk that no dose change will fix, and does this patient's renal function still support the dose they are on. Answering them separately, in three different references, is how the third one gets skipped.

Prescriber.io returns all three in one card: the interaction with its mechanism and cited source, the contraindication and bleeding risk flags, and the renal and hepatic dose considerations for the agent in question, with guideline-based alternatives where they exist. It is decision-support for licensed US clinicians, not autonomous prescribing. You review each flag, verify against the current prescribing information, and sign.