The Polypharmacy Problem: What Happens When You Take Too Many Drugs

Consider a common scenario. A patient is prescribed a diuretic for high blood pressure. The diuretic causes leg cramps. A magnesium supplement is added. The supplement interacts with a separate calcium channel blocker, reducing its efficacy. The blood pressure climbs, so a second antihypertensive is added. That drug causes postural hypotension and dizziness. The patient falls and is prescribed a short course of a sedating antihistamine for the pain-related sleep disruption. Two weeks later, they are cognitively foggy and their GP suspects early dementia. A new specialist appointment is booked. A new drug may be coming.

This is the prescribing cascade, and it plays out in consulting rooms across the world every day. It is not caused by incompetence. It is caused by the structural reality of modern multi-specialist medicine: each clinician optimises for their own domain, often without full visibility of what colleagues have already prescribed. The result, at scale, is polypharmacy, one of the most significant and least discussed patient safety problems in contemporary healthcare.

Defining the Problem: How Many Is Too Many?

Polypharmacy is formally defined in the clinical literature as the concurrent use of five or more medications. That threshold is somewhat arbitrary, a convention that evolved from epidemiological research rather than a firm pharmacological boundary, but it has proved useful for identifying populations at elevated risk. A more severe category, hyperpolypharmacy, captures patients on ten or more medications simultaneously. Both categories are increasingly common, and both are overrepresented among older adults.

The scale of the problem is striking. In the United Kingdom, NHS data shows that approximately 15 percent of adults aged 65 and over are prescribed eight or more medications concurrently. In the United States, the picture is similar: data from the National Health and Nutrition Examination Survey indicates that around 39 percent of adults aged 65 and over take five or more prescription drugs, and roughly 20 percent take ten or more. When over-the-counter medications and dietary supplements are included, those figures rise substantially, and those additional agents frequently interact with prescription drugs in ways neither the patient nor the prescriber has accounted for.

It is important to note that not all polypharmacy is inappropriate. A patient with heart failure, type 2 diabetes, chronic kidney disease, and atrial fibrillation may genuinely require eight evidence-based medications to manage those conditions optimally. The goal is never to reduce drug numbers for their own sake, but to ensure that every drug in a patient's regimen has a clear indication, a known benefit-risk balance for that particular patient, and an agreed stopping criterion. The problem arises when prescribing accumulates without that ongoing scrutiny.

The Prescribing Cascade: When Side Effects Become Diagnoses

The prescribing cascade was named and formally described by Jerry Gurwitz and colleagues in the 1990s, though the phenomenon had been observed in clinical practice for decades before it acquired a label. The mechanism is straightforward: an adverse drug reaction is misidentified as a new medical condition, and a second drug is prescribed to treat it. That second drug may cause its own adverse effects, which are treated with a third drug, and so on. Each step in the cascade looks clinically rational in isolation. In aggregate, the patient ends up on a complex polypharmacy regimen for conditions they never had.

Classic examples are well documented. Non-steroidal anti-inflammatory drugs (NSAIDs) raise blood pressure in a meaningful proportion of patients. That blood pressure rise is then treated with an antihypertensive. The NSAID is not identified as the cause. Similarly, metoclopramide, prescribed for nausea, can cause drug-induced parkinsonism. That parkinsonism may be diagnosed as idiopathic Parkinson's disease and treated with levodopa. The dopamine agonist may then cause compulsive behaviours that prompt a psychiatric referral. None of the downstream interventions address the root cause.

This is precisely why root cause medicine frameworks that seek the underlying driver of a symptom before adding a treatment are so valuable in polypharmacy contexts. When a clinician asks "what is causing this symptom?" rather than "what do I prescribe for this symptom?", the prescribing cascade can be interrupted before it begins.

Drug-Drug Interactions and the Cytochrome P450 System

The primary reason polypharmacy generates harm at the pharmacological level is drug-drug interactions, and the central machinery through which most of those interactions operate is the cytochrome P450 (CYP450) enzyme system. CYP450 enzymes, located predominantly in the liver but also in the gut wall, are responsible for the oxidative metabolism of roughly 70 to 80 percent of clinically used drugs. They are grouped into families (CYP1, CYP2, CYP3) and subfamilies, with CYP3A4, CYP2D6, CYP2C9, and CYP2C19 handling the majority of drug metabolism.

When two drugs compete for the same CYP enzyme, or when one drug inhibits or induces the enzyme that metabolises another, the plasma concentration of the affected drug shifts in ways the prescriber did not intend. An inhibitor of CYP3A4, such as certain antifungals, macrolide antibiotics, or calcium channel blockers, can dramatically increase the blood levels of co-administered statins metabolised by the same pathway, raising the risk of statin-induced myopathy. An inducer of CYP2C19, such as rifampicin, can reduce plasma concentrations of proton pump inhibitors and antidepressants to subtherapeutic levels. The clinical consequences range from treatment failure to severe toxicity, and they are often invisible unless a clinician specifically checks for the interaction.

Compounding the challenge is genetic variation. The field of pharmacogenomics has established that CYP enzyme activity varies substantially between individuals based on their genotype. CYP2D6, for instance, has over 100 known genetic variants, producing metaboliser phenotypes that range from poor (effectively no enzyme activity) to ultrarapid (multiple gene copies, very high activity). A poor metaboliser given codeine, which requires CYP2D6 to convert to its active form morphine, gets no pain relief. An ultrarapid metaboliser given the same dose gets potentially toxic morphine concentrations. Standard dosing tables assume an average metaboliser. Many patients are not average.

The Beers Criteria: A List That Should Be Better Known

In 1991, geriatrician Mark Beers and colleagues published a consensus-derived list of medications considered potentially inappropriate for use in older adults, specifically those residing in nursing facilities. The Beers Criteria, as they became known, have since been updated multiple times by the American Geriatrics Society (AGS), with the most recent version expanding coverage to all older adults and including new sections on drug-disease interactions, drug-drug interactions, and drugs warranting dose reduction in older patients.

The list includes medications that are widely prescribed and which many patients have been taking for years. First-generation antihistamines such as diphenhydramine (the active ingredient in many over-the-counter sleep aids) appear on the Beers list because of strong anticholinergic effects that increase the risk of confusion, urinary retention, and falls in older patients. Many benzodiazepines are listed because of fall and fracture risk. Certain antipsychotics appear because of excess mortality in older patients with dementia. Several NSAIDs are flagged for gastrointestinal bleeding risk in the context of reduced renal function.

Awareness of the Beers Criteria has improved prescribing practices, but implementation is uneven. Studies consistently show that a significant proportion of older adults in the UK, US, and across Europe remain on at least one Beers list medication, often prescribed by a specialist who may not have been thinking about geriatric appropriateness when initiating it. The 2019 NHS guidance on polypharmacy in Scotland explicitly referenced structured medication review tools as essential practice for patients on multiple medications, but guidance documents do not automatically change what happens in the consulting room.

STOPP and START: A European Complementary Framework

Working in parallel with the Beers Criteria, a group of European geriatricians led by Denis O'Mahony at University College Cork developed the STOPP/START criteria (Screening Tool of Older Person's Prescriptions / Screening Tool to Alert doctors to Right Treatment). First published in 2008 and subsequently updated through several versions, STOPP/START offers two complementary lists: STOPP, which identifies medications that should be considered for discontinuation in older patients, and START, which identifies evidence-based medications that are commonly omitted in older patients who would benefit from them.

That second component is important. Polypharmacy is not only about too many drugs. It also encompasses under-treatment, the failure to prescribe medications with clear benefit because a prescriber is reluctant to add to an already complex regimen. The STOPP/START criteria explicitly address both dimensions, making them arguably more comprehensive than Beers alone for the purpose of structured medication review.

A randomised controlled trial by O'Mahony and colleagues published in Age and Ageing demonstrated that pharmacist-led medication reviews using STOPP/START reduced the rate of adverse drug reactions in hospitalised older patients. The intervention arm showed a statistically significant reduction in falls, delirium, and length of hospital stay compared to usual care. These are not trivial outcomes. Falls in older adults are a leading cause of injury-related death and long-term disability, and delirium during hospitalisation is an independent predictor of dementia progression and institutionalisation.

The Karolinska Deprescribing Evidence

The most rigorous evidence base for deprescribing, the systematic and planned withdrawal of medications that are no longer appropriate, has come in significant part from Scandinavian research. A series of studies from the Karolinska Institute in Stockholm, led by researchers including Johan Fastbom and Maria Eriksdotter, examined the relationship between polypharmacy and clinical outcomes in Swedish older adult populations using the Swedish national drug registry, one of the most complete and longitudinally rich pharmaceutical databases in the world.

A landmark study from the Karolinska group, published in JAMA Internal Medicine, examined outcomes in nursing home residents following structured medication reviews that led to significant deprescribing. The findings were striking: residents in whom medications were reduced experienced improvements in cognition, reduced rates of falls, and lower all-cause mortality compared to control residents who continued their existing regimens. The study was observational, and causal inference is complex in this population, but the direction and magnitude of the associations were consistent with the hypothesis that some of the deterioration commonly attributed to ageing or disease progression in older adults is actually iatrogenic, caused by the drugs prescribed to help them.

The deprescribing movement has since developed structured frameworks, the most widely used being the GRADE-informed deprescribing guidelines being developed through the Canadian Deprescribing Network and collaborators internationally. These guidelines take a patient-centred approach, emphasising that the right time to stop a medication is when the burden exceeds the benefit for that individual patient, accounting for their prognosis, functional goals, and values. A cancer patient with a limited prognosis and dysphagia may have very different priorities than a community-dwelling 75-year-old with a 20-year life expectancy. Deprescribing conversations must be individualised.

AI and Clinical Decision Support: A New Layer of Protection

The cognitive demands of managing a patient on twelve medications, accounting for their renal function, hepatic metabolism, body weight, genetic polymorphisms, and the interaction potential of every drug pair in the regimen, are beyond unaided human capacity. This is precisely the gap that artificial intelligence is beginning to fill. Modern AI clinical decision support systems are being trained to identify high-risk drug combinations in real time, cross-referencing a patient's current medication list against interaction databases, pharmacogenomic profiles where available, and the patient's active diagnoses and laboratory values.

Companies including Tabula Rasa Healthcare and DrFirst have developed proprietary medication risk scoring engines that go beyond simple drug-drug interaction checking to produce composite risk scores that account for each drug's contribution to the overall pharmacokinetic burden of a regimen. Tabula Rasa's MedWise Risk Score, for instance, calculates which drugs are competing for inhibitory or inductive control over the CYP450 enzymes in a patient's full regimen, then quantifies the net drug interaction burden as a single number that can be tracked over time and used to identify the specific drugs most driving risk.

These tools are at their most effective when embedded at the point of prescribing, where a pharmacist or physician is about to add a new drug to an existing regimen, or during structured medication review. The challenge is alert fatigue: if a clinical decision support system fires warnings for every theoretical interaction in a complex regimen, clinicians quickly learn to dismiss them. The most effective AI-powered tools are those that prioritise by clinical significance, present brief and actionable guidance rather than lengthy interaction monographs, and are calibrated to suppress low-severity alerts that do not require immediate clinical action. Getting this balance right is as much a human factors challenge as a machine learning one.

Looking forward, the integration of pharmacogenomic testing into routine clinical care offers the prospect of truly personalised polypharmacy management. When a prescriber knows in advance that a patient is a CYP2D6 poor metaboliser, they can avoid drugs that depend on that enzyme for efficacy or require it for detoxification, before the drug is ever prescribed rather than after an adverse reaction has occurred. Several health systems, including Vanderbilt University Medical Centre's PREDICT programme and Mayo Clinic's RIGHT Protocol, have demonstrated that embedding pharmacogenomic results directly into the electronic health record at the point of prescribing is feasible and reduces the rate of adverse drug events.