Skip to content
Parth Ajmera

Drug-drug interactions

#pk-pd

A 65-year-old with heart disease, diabetes, and mild depression might take six to eight medications a day, and those drugs don't act in isolation — they interact, sometimes usefully, sometimes fatally. A drug-drug interaction (DDI) is what happens when one drug changes how another one behaves in the body, and it happens through two distinct mechanisms.

Pharmacokinetic interactions — one drug changes how another is absorbed, metabolised, or excreted. The most common route: one drug blocks or ramps up the liver enzymes (mainly the CYP450 family) that break down another. Block the enzyme, and the second drug accumulates toward toxicity; induce it, and the second drug disappears too fast to work. Grapefruit juice is the standard teaching example — a compound in it inhibits CYP3A4 and can push certain heart medications to dangerous levels.

Pharmacodynamic interactions — two drugs with overlapping effects amplify or cancel each other, independent of blood concentration. Two sedatives together sedate more than either alone; an anticoagulant plus an antiplatelet drug together bleed more than either alone.

DDI science identifies, predicts, and manages these effects — through in vitro lab assays, computational prediction, clinical studies, and regulatory risk assessment — so that by the time a drug is approved, physicians have some idea which combinations are safe.

Where it sits in development

DDI is both a discovery-stage screen and a late-stage regulatory requirement. Early in vitro CYP inhibition assays flag compounds with a severe DDI liability before much money is spent on them; specific clinical DDI studies, testing the new drug against known "probe" drugs, are required before approval; and the DDI section of a drug's label — the part stating known interactions and contraindications — is written from this work and defended to regulators directly.

What's actually settled, and what isn't

CYP450-mediated small-molecule DDI is mature, mechanistically well-characterised science built up over roughly forty years, with detailed regulatory guidance on what data is expected. What's still open:

  • Transporter-mediated interactions — via proteins like P-gp or OATP1B1/1B3 — are meaningfully harder to predict from lab data than enzyme-mediated ones, and the regulatory guidance here is still evolving.
  • Biologic DDIs work through a completely different mechanism: inflammatory disease changes CYP enzyme expression via cytokines, and a monoclonal antibody that neutralises those cytokines can shift the PK of a co-administered small molecule substantially. This is an active concern as combination biologic regimens become more common.
  • Polypharmacy — three or more drugs at once — is essentially unstudied as a combinatorial problem; almost all DDI science examines drugs in pairs, while real patients, especially elderly ones, often aren't on just two.
  • Machine learning DDI prediction from molecular structure alone is an active research area with real limitations — reasonably accurate for known interaction mechanisms, much less so for novel ones.
  • Pediatric DDI is thin, because CYP enzyme expression itself changes substantially through childhood development.

I'd treat any specific dollar figure or percentage attached to DDI-related adverse events elsewhere as someone's estimate rather than a fixed number — it depends heavily on which patient population and which drug classes are being counted.