Every interaction checker answers the first question: does a potential interaction exist? The questions that decide patient care come after: is it clinically meaningful for this patient, what is the mechanism, how strong is the evidence, can it be monitored, should the combination be avoided or merely adjusted, and for how long? A layered workflow answers them in minutes, and SPS itself recommends exactly this layering: BNF or BNFC for the rapid screen, the SmPC's section 4.5 for product-specific authorised information, Stockley's for evidence, mechanism and management, and maintained specialist checkers for defined territories. Here is the workflow with the layers explained.
Which source for which question
BNF and BNFC: the rapid pairwise or multi-medicine screen, comprehensive and concise, the correct first pass every time. emc section 4.5: what the authorised text says for the exact product, occasionally formulation-relevant, always licence-anchored: /blog/how-to-use-emc-smpcs-pils. SPS: practical management of selected common interaction problems, written for working clinicians, DOAC interaction management being a model of the genre. Stockley's: the reference of record for evidence quality, mechanism, clinical importance and management options, the layer that converts "listed interaction" into "understood interaction". Specialist checkers, for their defined territories: HIV and viral hepatitis drug interactions, QT-prolongation risk, reproductive and sexual health, and oncology where a maintained source exists, each authoritative inside its lane and dangerous as a general tool. And pharmacy or a medicines-information service for the unresolved cases: the option list ends with a human, deliberately.
Worked example: apixaban
The DOAC interactions cluster illustrates why mechanism matters more than lists. Strong CYP3A4 and P-glycoprotein inhibitors raise apixaban exposure, a pharmacokinetic problem pointing one direction; strong inducers lower it, the opposite hazard, reduced anticoagulation dressed as tolerance; antiplatelets, NSAIDs and serotonergic antidepressants raise bleeding risk pharmacodynamically without touching drug levels at all. Three different mechanisms, three different management logics, and a checker that flags all of them equally has told you almost nothing; the apixaban clinical page frames the classes with links to SPS's DOAC interaction guidance and the product SmPCs, which is the orientation step before Stockley's or pharmacy settles the specific pair.
Worked example: amoxicillin
Common antibiotics carry a compact interaction set worth knowing by shape: methotrexate, where the combination can raise methotrexate toxicity risk and warrants respect; warfarin, where antibiotics are a classic destabiliser of INR and a prompt for closer checking; allopurinol, associated with more frequent rash; and probenecid, which raises amoxicillin levels by design, historically deliberately. None of these is exotic, all are screenable in the first-pass layer, and the amoxicillin page carries the orientation with its sources; the value of knowing the shapes is speed, the screen confirms what you already suspected rather than surprising you.
The pitfalls that cause real errors
Four, repeatedly observed. Searching only prescription medicines: over-the-counter NSAIDs, herbal products, St John's wort above all, sit outside many patients' medication lists and inside many serious interactions. Treating every listed interaction as equal: lists are sensitivity-optimised; your job is specificity, which is what the mechanism-and-evidence layer exists for. Ignoring context: dose, duration, renal function and timing modulate almost every interaction, and a static pair-check cannot see them. And leaning on an unmaintained specialist checker: a stale interaction database is worse than none, because it answers confidently; check the maintenance status of any specialist tool before trusting it as sole source.
The workflow, assembled
Screen in BNF or BNFC; read 4.5 for the exact products where the flag is live; escalate to SPS for practical management of the common problems and to Stockley's for mechanism, evidence and options; use the specialist checker if, and only if, the territory is theirs; and hand the residue to pharmacy with the context attached, renal function, duration, what you have already found. Minutes, not archaeology, and the answer is defensible at every layer; the full stack this workflow lives inside is the pillar at /blog/best-bnf-alternatives-uk-medicines-resources.
Frequently asked questions
Is there a good free UK interaction checker for patients?
The NHS medicines pages flag key interactions readably, and patients should be actively told to declare over-the-counter and herbal products; patient-grade tools screen, they do not manage, and the pharmacist remains the accessible expert.
How should multi-drug regimens be approached?
Screen the whole list, then think in mechanisms rather than pairs, additive bleeding risk, additive QT load, competing metabolism, which is where polypharmacy interactions actually live; structured medication review is the proper container for it.
When is stopping the interacting drug wrong?
Often: many flagged combinations are managed with monitoring, timing separation or temporary adjustment, and reflexive stopping trades a managed risk for an unmanaged one. The management layer exists precisely because avoid is only one of its verbs.