P-glycoprotein inhibitors: the list, the substrates that matter, and the inducers

P-glycoprotein is an ATP-driven efflux pump encoded by the ABCB1 gene, expressed on the apical surface of cells in the intestine, the proximal renal tubule, the biliary canaliculus and the endothelium of the blood-brain barrier. In each of those locations it moves its substrates in the same direction: out of the body or out of a protected compartment. In the gut it limits how much of an oral drug is absorbed. In the kidney and bile it promotes elimination. At the blood-brain barrier it keeps certain drugs out of the central nervous system.

That is why P-gp inhibition can matter even when CYP metabolism is untouched. Dabigatran and digoxin are not meaningfully metabolized by CYP enzymes, so a pure CYP interaction check clears them, yet both are P-gp substrates whose levels rise when the pump is blocked. The interaction is real, it is just invisible to a mental model built only around enzymes. The reverse is the loperamide problem: block P-gp at the blood-brain barrier and a drug that is normally excluded from the brain gets in.

The substrates that change management are a short list. Digoxin is the classic probe substrate and has a narrow therapeutic index, so a P-gp inhibitor added to a stable digoxin regimen can push it toward toxicity. The direct oral anticoagulants are all P-gp substrates: dabigatran is the most P-gp dependent and edoxaban labeling carries specific dose reductions with certain P-gp inhibitors, while rivaroxaban and apixaban are affected mainly when a drug inhibits both P-gp and CYP3A4. Colchicine is a P-gp and CYP3A4 substrate, and the combination of a strong dual inhibitor with colchicine has caused fatal toxicity, especially in renal or hepatic impairment.

Beyond those, loperamide, fexofenadine, aliskiren, sirolimus, tacrolimus, the statins to a degree, and some chemotherapy agents are P-gp substrates. Fexofenadine and digoxin are used as probe substrates in interaction studies precisely because they are almost purely dependent on the transporter. The practical takeaway is not to memorize the whole list but to recognize the handful whose exposure change actually alters a decision: digoxin, the DOACs, colchicine and loperamide.

The first is colchicine with a strong dual P-gp and CYP3A4 inhibitor, such as clarithromycin, itraconazole or ritonavir. Colchicine has no early warning: the toxicity presents late as multi-organ failure, and the labeling gives specific dose reductions or contraindicates the combination outright in renal or hepatic impairment. This is the P-gp interaction most likely to be fatal and the one worth checking every time colchicine is prescribed.

The second is a DOAC with a P-gp inhibitor, which raises anticoagulant exposure and bleeding risk. Dabigatran and edoxaban labeling name specific inhibitors and specific dose adjustments, and the interaction is worse when renal function is reduced because both the drug and the inhibitor effect stack on top of impaired clearance. The third is loperamide with a P-gp inhibitor, which lets loperamide cross the blood-brain barrier and reach the central nervous system, producing opioid effects it normally cannot; high-dose loperamide also carries a QT and cardiac arrhythmia risk that P-gp inhibition amplifies.

Inducers do the opposite of inhibitors: they increase P-gp expression, so more drug is pumped out and substrate exposure falls, sometimes below the therapeutic range. Rifampin is the strongest and also induces CYP3A4, which is why it lowers the levels of so many drugs at once. Carbamazepine, phenytoin, phenobarbital and St John's wort are the other common ones, and St John's wort is the one patients do not report because they do not consider a supplement a medication.

The clinical consequence is loss of efficacy rather than toxicity, which makes it quieter and often only noticed when a treatment stops working. A patient on dabigatran or edoxaban who starts rifampin can drift toward subtherapeutic anticoagulation, and the DOAC labeling generally advises against combining them for that reason. Induction also has a lag: it builds over one to two weeks as new transporter is synthesized and persists for a similar period after the inducer is stopped, so the interaction window extends past the obvious dates.

A memorized list of P-gp inhibitors is a starting point, but it fails at the exact place the risk concentrates: the drugs that inhibit both P-gp and CYP3A4. Clarithromycin, itraconazole, ketoconazole, ritonavir, cobicistat, cyclosporine and dronedarone all hit both pathways, so a substrate that happens to depend on both, colchicine and the DOACs above all, sees a multiplied effect that a single-pathway mnemonic underestimates. The mnemonic tells you a drug is on the list; it does not tell you the patient in front of you is also on a substrate that both pathways clear.

That is the case for checking the actual regimen rather than recalling a list. The interaction that matters is not "is this a P-gp inhibitor" in the abstract, it is "does this patient take both a dual inhibitor and a dual substrate, and is their renal function low enough to make it worse." Those three facts live in three different places, and pulling them together at the moment of prescribing is the work a point-of-care check is meant to do.

Prescriber.io evaluates the whole regimen rather than isolated pairs, so it surfaces a P-gp interaction even when there is no CYP interaction to find, which is exactly the blind spot for digoxin and dabigatran. When a drug inhibits both P-gp and CYP3A4, the check flags the combined effect on dual substrates like colchicine and the DOACs, and it folds the patient's renal and hepatic considerations into the same card rather than leaving them for a separate lookup.

It is decision-support for licensed US clinicians and does not prescribe. Each flag comes with its mechanism and a named source, alongside the contraindication and allergy check and guideline-based alternatives, in one card. You review the flags, decide what the patient actually needs, verify against the current labeling, and sign.