Design: Scoping review (PRISMA-ScR) of clinical reports and observational datasets
Population: 30 studies of patients taking antiseizure medications with DOACs
Primary outcome: DOAC concentrations and clinical thrombotic or bleeding outcomes
Effect: Enzyme inducers consistently linked to subtherapeutic DOAC levels and failures; dose doubling ineffective in several cases

A scoping review in Therapeutic Drug Monitoring, published on 23 October, collected 30 clinical reports and observational datasets on antiseizure medications given with direct oral anticoagulants (DOACs).

Enzyme-inducing drugs, mainly carbamazepine, phenobarbital, primidone and oxcarbazepine, were consistently associated with subtherapeutic DOAC concentrations and serious failures such as thromboembolism. In several cases doubling the DOAC dose did not prevent events. Most data concerned rivaroxaban and apixaban; edoxaban data were sparse. Evidence for non-inducing drugs such as levetiracetam and valproate was limited and variable.

Phenobarbital and carbamazepine remain common in India because of cost, so this combination is not rare. Where anticoagulation is needed in someone on an enzyme inducer, the options are to switch the antiseizure drug to a non-inducer with neurology input, or to use warfarin with INR monitoring.

The evidence is case-based and prone to publication bias, but the direction is consistent and the mechanism well understood.

Avoid combining a DOAC with carbamazepine, phenobarbital, primidone or phenytoin where possible.
Do not rely on doubling the DOAC dose to overcome induction.
Discuss switching to a non-inducing antiseizure drug with neurology, or use warfarin with INR checks.
Remember induction persists for about two weeks after stopping the inducer.
Consider DOAC level measurement where available if the combination cannot be avoided.

Why it matters

Dose escalation, the intuitive fix, did not protect patients in these reports.

The statistics, in plain English

This is a scoping review of case reports and observational data, so it cannot give a rate of failure. Its strength is consistency: the same direction of effect across many reports, with a known mechanism.