Drug-drug interactions: a practical guide for prescribers

A drug-drug interaction (DDI) occurs when one medicine changes the exposure to, or the effect of, another. Interactions work through two broad routes: pharmacokinetic, where one drug alters how another is absorbed, distributed, metabolised or excreted, and pharmacodynamic, where two drugs act on the same physiological system and add to or oppose each other. Most clinically important interactions are predictable, so most are preventable. This guide covers the mechanisms, the cytochrome P450 system, the high-risk combinations behind real hospital admissions, who is most vulnerable, and a practical workflow for checking interactions at the point of prescribing.

It is written for clinicians as a reference, not for patients, and it does not replace your formulary or a dedicated interaction resource. Always confirm specific combinations, doses and monitoring against your national formulary, the product label and local guidance.

What is a drug-drug interaction?

A drug-drug interaction is a change in the exposure to, or response to, one drug (the object or "victim") caused by another drug given at the same time (the precipitant or "perpetrator"). The result may be toxicity, loss of effect or, occasionally, a deliberate therapeutic benefit. What matters clinically is not whether an interaction exists on paper but whether it is likely to change the outcome for this patient.

That distinction is the core skill. Interaction databases list thousands of theoretical pairs, many of minor significance. The prescriber's job is to recognise the small number of combinations that genuinely cause harm, understand the mechanism well enough to predict timing and magnitude, and choose the right response: avoid, substitute, adjust the dose, or monitor.

How common are drug interactions, and why do they matter?

Drug interactions are a measurable cause of avoidable hospital admissions. A 2014 systematic review and meta-analysis of 13 studies found that DDIs accounted for a median 1.1% of hospital admissions overall — and a median 22.2% of admissions among patients admitted with an adverse drug reaction (Dechanont et al., Pharmacoepidemiology and Drug Safety).

The same review found NSAIDs were the drugs most often involved in interaction-related admissions, and warfarin the most often involved at outpatient and emergency visits. The underlying driver is polypharmacy: the probability of an interaction rises with the number of drugs taken, and the average 65-year-old takes around five (Cascorbi, Deutsches Ärzteblatt International). The World Health Organization made medication safety in polypharmacy one of three priority areas of its Medication Without Harm challenge, with guidance on prioritising patients for medication review (WHO technical report, 2019).

What are the main types of drug interactions?

There are two main types. Pharmacokinetic interactions change how much drug reaches its site of action, by altering absorption, distribution or transport, metabolism, or excretion. Pharmacodynamic interactions leave drug levels unchanged but combine effects at the receptor, tissue or organ level — additively (two drugs that both cause bleeding) or antagonistically (one drug blunting another).

Pharmacokinetic interactions: the ADME framework

Working through absorption, distribution, metabolism and excretion (ADME) is the most reliable way to anticipate a pharmacokinetic interaction. The examples below are drawn from the Cascorbi review unless stated otherwise:

• Absorption — multivalent cations bind other drugs in the gut. Antacids containing aluminium or magnesium reduce levothyroxine absorption, and fluoroquinolones lose bioavailability when taken with antacids, zinc or iron preparations. Check the product label for how far apart to give them.
• Distribution and transport — the efflux transporter P-glycoprotein limits absorption and speeds elimination of drugs such as digoxin. Verapamil inhibits P-glycoprotein and can markedly raise digoxin levels, while the inducer carbamazepine can render digoxin ineffective. The US FDA lists clarithromycin, erythromycin, itraconazole and verapamil among clinical P-glycoprotein inhibitors (FDA).
• Metabolism — competition for, inhibition of, or induction of cytochrome P450 enzymes. This is the largest and most clinically important group, covered in detail below.
• Excretion — a precipitant can reduce renal clearance of the object drug, as with digoxin and macrolide antibiotics. Renally cleared drugs also need dose adjustment for the patient's glomerular filtration rate, which changes the interaction picture in kidney disease.

Pharmacodynamic interactions: additive and opposing effects

Pharmacodynamic interactions are easy to overlook because no interaction checker is needed to see them — only a moment's thought about what each drug does. Classic examples from the Cascorbi review:

• Bleeding — an NSAID added to a vitamin K antagonist is additive; SSRIs raise the bleeding risk with vitamin K antagonists to a similar degree as NSAIDs.
• Antagonism — ibuprofen can blunt the antiplatelet effect of aspirin.
• QT prolongation — combining a fluoroquinolone with a macrolide can prolong the QT interval.
• Serotonin toxicity — adding other serotonergic drugs such as tramadol or triptans to an SSRI increases the risk of serotonin syndrome.
• Hyperkalaemia — an ACE inhibitor combined with a potassium-sparing diuretic such as amiloride can cause life-threatening hyperkalaemia.

How do CYP450 drug interactions work?

The cytochrome P450 (CYP) family has more than 50 enzymes, but six of them metabolise around 90% of drugs, with CYP3A4 and CYP2D6 the most significant (Lynch and Price, American Family Physician). A CYP interaction occurs when a precipitant inhibits an enzyme (raising levels of drugs it clears) or induces it (lowering them).

Inhibition vs induction: why timing matters

The two mechanisms behave very differently over time, and getting the timing right is what makes monitoring useful. Per the AAFP review:

• Inhibition is fast and dose-dependent. Its effect begins quickly once the inhibitor reaches adequate levels. Sertraline, for example, is a mild CYP2D6 inhibitor at 50 mg but a potent one at 200 mg.
• Induction is delayed. Inducers work by increasing enzyme production, so the effect builds over days and depends on the inducer's half-life — within about 24 hours for rifampicin (rifampin), but up to a week for phenobarbital.
• Offset matters too. When an inducer is stopped, enzyme activity falls back gradually and levels of the object drug can rise — a common trap when a patient finishes a course of an enzyme-inducing drug. Long half-life inhibitors are similar: interactions with SSRIs should be expected for several days after the last dose (Cascorbi).

The US FDA grades inhibitors by their effect on sensitive substrates: a strong inhibitor increases the area under the curve (AUC) by five-fold or more, a moderate one by two- to five-fold. Its clinical index lists clarithromycin and itraconazole as strong CYP3A inhibitors, and carbamazepine, phenytoin and rifampicin as strong CYP3A inducers that cut substrate exposure by 80% or more (FDA table of substrates, inhibitors and inducers).

Prodrugs: when inhibition causes treatment failure

For prodrugs, the usual logic reverses: inhibiting the activating enzyme reduces the effect. Clopidogrel needs CYP2C19 for activation; tamoxifen depends on CYP2D6 to form endoxifen, which some SSRIs can inhibit; and CYP2D6 inhibition reduces conversion of codeine to morphine (Cascorbi). Tramadol and losartan are further examples of drugs that need metabolic activation (AAFP). The danger here is silent: no toxicity, just a treatment that quietly stops working.

Genetics and grapefruit

Pharmacogenetics adds a second layer. Roughly 7% of white patients and 2–7% of Black patients are CYP2D6 poor metabolisers, and about one in five Asian patients is a CYP2C19 poor metaboliser (AAFP). In these patients a standard dose can behave like an overdose — or, for a prodrug, like no dose at all.

Grapefruit is the best-known food interaction. Its furanocoumarins irreversibly inactivate intestinal CYP3A4; more than 85 drugs are known or predicted to interact, 43 of them with potentially serious consequences, and a single 200 mL glass can be enough to raise drug exposure meaningfully. In one study cited by the review, simvastatin exposure reached 330% of control after three days of grapefruit juice (Bailey et al., CMAJ). Ask about it — patients rarely volunteer their breakfast.

Which drug interactions are most dangerous?

The most dangerous interactions involve drugs where a small change in exposure causes serious harm — anticoagulants, digoxin, sulfonylureas, potassium-raising drugs, statins and drugs affecting the QT interval or serotonin. Population data show these combinations drive real admissions, and that choosing a non-interacting alternative removes the excess risk.

A landmark Ontario study of adults aged 66 and over compared patients admitted with drug toxicity against matched controls (Juurlink et al., JAMA 2003). In the week before admission:

• Patients on glibenclamide (glyburide) admitted with hypoglycaemia were far more likely to have received co-trimoxazole (adjusted odds ratio 6.6).
• Patients on digoxin admitted with digoxin toxicity were far more likely to have received clarithromycin (adjusted OR 11.7).
• Patients on an ACE inhibitor admitted with hyperkalaemia were far more likely to have received a potassium-sparing diuretic (adjusted OR 20.3).

Crucially, drugs with similar indications but no known interaction — amoxicillin, cefuroxime and indapamide — showed no increased risk. That is the practical lesson: when an interacting option has a safe alternative, the safest interaction is the one you never create.

Other combinations worth committing to memory:

• Warfarin with fluconazole, amiodarone or metronidazole — CYP-mediated rises in INR and bleeding risk (AAFP). Add SSRIs and NSAIDs to the list of drugs that increase bleeding with vitamin K antagonists.
• The "triple whammy" — a diuretic plus an ACE inhibitor or angiotensin receptor blocker plus an NSAID. In a UK cohort of nearly 490,000 antihypertensive users, triple therapy was associated with a 31% higher rate of acute kidney injury (rate ratio 1.31), rising to 1.82 in the first 30 days; double combinations were not associated with increased risk (Lapi et al., BMJ 2013).
• Simvastatin with potent CYP3A4 inhibitors — higher statin levels raise the risk of myopathy (AAFP).
• QT-prolonging combinations — metabolic inhibition causing life-threatening arrhythmias led to the withdrawal of terfenadine, astemizole and cisapride from the US market (AAFP).
• Serotonergic stacking — SSRIs with tramadol or triptans (Cascorbi).

Which patients are most at risk of drug interactions?

Risk concentrates in older adults on multiple medicines, patients with renal impairment, people taking narrow-margin drugs such as anticoagulants or digoxin, and genetic poor metabolisers. Patients who see several prescribers, or buy over-the-counter and herbal products, add hidden exposures that no single prescriber can see without asking.

• Polypharmacy — interaction probability rises with each added medicine (Cascorbi; WHO).
• Older age — more medicines and less physiological reserve; all three Juurlink cohorts were aged 66 and over.
• Renal impairment — reduced clearance of renally excreted drugs, which need dose adjustment for the glomerular filtration rate (Cascorbi).
• Narrow-margin drugs — anticoagulants, digoxin, sulfonylureas and antiepileptics, where a modest change in exposure is clinically significant.
• Transitions of care — admissions, discharges and new specialist prescriptions are when a regimen changes fastest and is least likely to be reconciled.

How should prescribers check for drug interactions?

Check every new prescription against a complete, reconciled medication list — including over-the-counter, herbal and recently stopped drugs — using a reliable interaction reference. Then judge clinical relevance: the mechanism, the likely timing, the patient's risk factors, and whether a safer alternative exists. Document the decision and the monitoring plan.

A practical seven-step workflow:

• 1. Reconcile the list. Prescribed, over-the-counter, herbal and supplement products, plus anything stopped in the last few weeks (an inducer's effect outlasts the last dose).
• 2. Flag the high-risk objects. Anticoagulants, antiplatelets, digoxin, sulfonylureas and insulin, antiepileptics, immunosuppressants, statins, and drugs that raise potassium or prolong the QT interval.
• 3. Identify the mechanism. Pharmacokinetic (which enzyme or transporter, inhibition or induction?) or pharmacodynamic (which shared effect?). The mechanism predicts the timing and the right monitoring.
• 4. Look for a non-interacting alternative. As the Juurlink data show, substituting an equivalent drug is often the simplest and safest option.
• 5. If the combination is necessary, adjust and monitor. Reduce the dose where guidance supports it, and time monitoring to the mechanism — INR, potassium and creatinine, drug levels or ECG.
• 6. Counsel the patient. Explain what to avoid (grapefruit, NSAIDs bought over the counter) and which symptoms should prompt early review.
• 7. Document it. Record the interaction, your reasoning and the monitoring plan in the Plan section of the note, so the next clinician understands why the combination was accepted.

Writing that reasoning down is part of safe prescribing, not paperwork for its own sake. Our guide to how to write a SOAP note covers how to tie every plan item back to a documented problem — including a medication decision like this one.

Do drug interaction checkers work, and what is alert fatigue?

Electronic interaction checkers help, but they are blunt. A review of 17 studies found drug safety alerts were overridden by clinicians in 49% to 96% of cases — often justifiably, because many alerts are low in specificity or clinically irrelevant (van der Sijs et al., JAMIA). This is alert fatigue.

The authors argued that the alerting system itself creates conditions for error: low specificity, unclear information and unnecessary workflow interruptions train clinicians to click past warnings, including the important ones. Electronic prescribing alerts do help catch interactions early (Cascorbi), but they work best when they are tiered by severity and specific to the patient — and when the prescriber treats them as a prompt to think, not a box to dismiss.

Where AI drug-safety review fits

The weakness of a generic checker is that it sees drug pairs, not patients. A context-aware review that reads the actual medication list, allergies and the plan being written is a more useful model than a pair-by-pair lookup. Shifaa AI, the AI clinical assistant coming soon and now in private beta, includes a drug-safety review that checks interactions, allergies, contraindications and dosing against the patient's medication list before you sign the prescription. See the full feature overview.

The boundary is the same as for any clinical AI. It is decision support: it flags potential problems, and you decide — then confirm against your formulary. A general chatbot has no such safety layer, which is one of the reasons we argue in ChatGPT for doctors that purpose-built tools are different. In polypharmacy-heavy specialties such as internal medicine and cardiology, this kind of check earns its place at every prescription.

Frequently asked questions

What are the two main types of drug-drug interactions?

Pharmacokinetic interactions change how much of a drug reaches its target by altering absorption, distribution, metabolism or excretion — CYP450 inhibition and induction are the commonest. Pharmacodynamic interactions leave drug levels unchanged but combine effects, such as two drugs that both increase bleeding risk or prolong the QT interval.

What is the difference between a CYP450 inhibitor and an inducer?

An inhibitor reduces enzyme activity, raising levels of drugs cleared by that enzyme; its effect starts quickly and is dose-dependent. An inducer increases enzyme production, lowering levels of those drugs; its effect builds over days to a week and fades gradually after it is stopped.

Which drug interactions most often lead to hospital admission?

In a 2014 meta-analysis, NSAIDs were most often involved in interaction-related admissions and warfarin in interaction-related outpatient and emergency visits. Population studies also link co-trimoxazole with sulfonylurea hypoglycaemia, clarithromycin with digoxin toxicity, and potassium-sparing diuretics with hyperkalaemia on ACE inhibitors.

Can an AI drug interaction checker replace clinical judgement?

No. An AI drug-safety review can check a full medication list quickly and flag potential interactions, but it is decision support. The prescriber weighs the clinical context, decides whether to avoid, substitute, adjust or monitor, and confirms against the formulary.

References

• Cascorbi I. Drug interactions — principles, examples and clinical consequences. Deutsches Ärzteblatt International, 2012;109(33–34):546–556. pmc.ncbi.nlm.nih.gov/articles/PMC3444856
• Lynch T, Price A. The effect of cytochrome P450 metabolism on drug response, interactions, and adverse effects. American Family Physician, 2007;76(3):391–396. aafp.org
• US Food and Drug Administration. Drug development and drug interactions: table of substrates, inhibitors and inducers. fda.gov
• Juurlink DN, Mamdani M, Kopp A, Laupacis A, Redelmeier DA. Drug-drug interactions among elderly patients hospitalized for drug toxicity. JAMA, 2003;289(13):1652–1658. doi.org/10.1001/jama.289.13.1652
• Lapi F, Azoulay L, Yin H, Nessim SJ, Suissa S. Concurrent use of diuretics, ACE inhibitors, and angiotensin receptor blockers with NSAIDs and risk of acute kidney injury: nested case-control study. BMJ, 2013;346:e8525. pmc.ncbi.nlm.nih.gov/articles/PMC3541472
• Bailey DG, Dresser G, Arnold JMO. Grapefruit–medication interactions: forbidden fruit or avoidable consequences? CMAJ, 2013;185(4):309–316. pmc.ncbi.nlm.nih.gov/articles/PMC3589309
• Dechanont S, Maphanta S, Butthum B, Kongkaew C. Hospital admissions/visits associated with drug-drug interactions: a systematic review and meta-analysis. Pharmacoepidemiology and Drug Safety, 2014;23(5):489–497. doi.org/10.1002/pds.3592
• van der Sijs H, Aarts J, Vulto A, Berg M. Overriding of drug safety alerts in computerized physician order entry. Journal of the American Medical Informatics Association, 2006;13(2):138–147. pmc.ncbi.nlm.nih.gov/articles/PMC1447540
• World Health Organization. Medication safety in polypharmacy: technical report (Medication Without Harm). 2019. who.int