Editorial · methodology

How our drug interaction checker works

Where the interaction data comes from, how a severity is assigned, what the Drug Interaction Checker can and cannot detect, and how the records are kept current and reviewed.

Principle

The checker never invents, infers or estimates an interaction. Every result is a record written from a named official document, and every record keeps the document, its section and its country. No language model and no rule of our own decides whether an interaction exists, how serious it is, what its mechanism is or how it should be managed: those come from the cited source or they are absent. When no record matches a pair, the checker says so in exactly these words: No known interaction identified in the available data, followed by a reminder that absence from the dataset is not evidence of safety.

Where the interaction information comes from

Each record paraphrases what one or more official documents say about a pair of medicines; nothing is quoted verbatim and no proprietary interaction database is reproduced or scraped. The documents are the medicine information of official health bodies, national formularies and regulators, listed as sources on every card. The data currently indexed:

Publishers cited
BNF (NICE) (77), NHS (75), MedlinePlus (NIH) (18), NICE (4), UK Government (2)
Records
72 interaction records for 39 medicines with pages, 14 brand and combination products, 73 named substances and 27 drug classes
Scope
drug-drug (medicine-to-medicine); food, alcohol and supplement interactions live on each medication page
Data version
2026-09-12 · last updated
Review status
Facts checked against the cited sources; not yet clinically reviewed (72)
Licence of the records
The written records are CC BY 4.0 like the rest of the site's content; the cited documents remain their publishers' and are linked, not reproduced.

A record names the medicine on this site it belongs to and either another medicine on this site, a drug class (when the source states the interaction for the whole class) or a named substance without a page yet. Brand names resolve to their generic medicine; combination products are split into their active ingredients; a drug class is applied only to its members, as the source states.

How severity is assigned

A record carries one of six source states, assigned only when the cited document's wording supports it. Where two sources disagree, both are kept as evidence and the record stays at the weaker state, flagged for review, until a person resolves it. The checker shows every state as one of five display tiers, mapped as below. The mapping adds no information: "avoid" and "contraindicated" share the top tier because both mean the source says the medicines should not be taken together; a record the source documents without any management wording is shown as not graded, never as minor. No state maps to minor: the sources we use describe management rather than grading harm, and we do not label an interaction minor by inference.

Source stateAssigned when the cited document saysDisplay tier
Contraindicated in official information CONTRAINDICATEDcontraindicated / must not be used together / do not takeContraindicated or avoid
Official guidance: avoid the combination AVOID_COMBINATIONavoid / not recommended / do not take without adviceContraindicated or avoid
Specialist supervision or close monitoring SPECIALIST_OR_CLOSE_MONITORINGonly under specialist supervision, or for specific indications with close monitoringMajor
Monitoring or dose adjustment advised MONITOR_OR_ADJUSTmonitor (a level, the INR, kidney function…) or adjust a doseModerate
Interaction documented INTERACTION_DOCUMENTEDlisted as an interaction to tell a doctor or pharmacist about, without a specific management instructionSeverity not graded
No severity assigned by the sources NO_SEVERITY_ASSIGNEDdescribes an effect but gives no management wording (no severity document needed)Severity not graded
Minor a source grades the interaction as of limited clinical significance (no current source state)Minor

What the checker can and cannot detect

How the data is updated

Records are edited when the cited guidance changes and the dataset carries the date of its last change, shown on the checker. An automated job fetches every cited URL weekly and on every content change and fails the build on a dead link, so a record can never cite a document that has disappeared. The compiler that builds the site rejects any record without a source, any severity state without the document that supports it, and any link to a medicine, class, test or organ that does not exist.

Review status

Every record shows its review status on its card. At present all records are facts checked against the cited sources; not yet clinically reviewed: the editorial team has verified each record against the documents it cites, but no independent clinician or pharmacist has yet reviewed them. We say so rather than imply a review that has not happened; see the medical review policy. Before a clinical release the records are tested against a clinician-approved validation set of representative pairs, including contraindicated, major, no-interaction, combination-product and brand-plus-generic cases.

Privacy

The medicines you add are not stored by the site and are not sent to any third party. They appear only in the page address so a check can be shared, and that address is reported to analytics without its query string. No account is needed.

Open the Drug Interaction Checker