Adverse Effects of Cholinesterase Inhibitors in Dementia, According to the Pharmacovigilance Databases of the United-States and Canada

If two drugs treat the same disease, work by the same mechanism, and are prescribed interchangeably, does it matter which one your doctor chooses? For cholinesterase inhibitors in dementia, Ali and colleagues examined the adverse event records of two countries and found that the answer is yes. The difference shows up in the starkest possible outcome: death. Start with the scale of the problem. More than 35 million people worldwide live with dementia, and about 7.7 million new cases are diagnosed every year. Alzheimer's disease accounts for roughly 60 to 80 percent of those cases. The number of people affected is projected to reach 115 million by 2050, with the sharpest rises, around 300 percent, in developing countries. The drugs used to treat Alzheimer's are symptomatic, not curative. They aim to slow cognitive decline and extend patient autonomy by raising the availability of the neurotransmitter acetylcholine at synapses. They do this by inhibiting the enzyme that breaks acetylcholine down. The three main drugs in this class are donepezil, rivastigmine, and galantamine. Donepezil and galantamine are reversible, non-covalent inhibitors. They bind acetylcholinesterase, slow it down, and let it go. Rivastigmine is different. It's classified as a pseudo-irreversible inhibitor, meaning it produces an extremely low reactivation rate of the enzyme, holding on so tightly that its effects can last far longer than expected. Rivastigmine is also the only drug in the class available as a transdermal patch — a feature that, as we'll see, turns out to matter quite a bit. Clinical trials established that all three drugs have comparable efficacy and broadly similar safety profiles. But trials are finite. They run for months in carefully selected patients, and rare or delayed adverse reactions often don't emerge until a drug is out in the real world, prescribed to millions of people with complicated medical histories. That's exactly what post-marketing surveillance is designed to catch. Ali and colleagues analyzed two national pharmacovigilance databases: the U.S. Food and Drug Administration Adverse Event Reporting System, known as FAERS, and the Canada Vigilance Adverse Reaction Database, or CVARD. These systems collect voluntary adverse event reports from clinicians, patients, and caregivers. A critical limitation applies immediately: they record what gets reported, not everything that happens. There are no prescription denominators or absolute incidence rates — just signals. The tool the researchers used to extract those signals is called a Reporting Odds Ratio, or ROR. Think of it this way: the ROR compares how often a specific bad outcome is reported for one drug against how often that same outcome is reported for comparable drugs in the same database. An ROR of 1 means no difference. An ROR above 1 means the outcome shows up disproportionately for that drug. From FAERS, the team extracted 9,877 reports for the three cholinesterase inhibitors out of more than 3.6 million total reports in the database. From CVARD, they extracted 2,247 reports. These are the datasets that produced the signal. And the signal is hard to dismiss. In FAERS, rivastigmine was associated with a reporting odds ratio for death of 3.42, with a confidence interval from 2.94 to 3.98, and a p-value below 0.0001. In plain terms, death was reported about three and a half times as often for rivastigmine relative to the other drugs in the class. That alone is striking. What makes it even more striking is that the same analysis, run independently on an entirely separate national database, returned almost the same number. In CVARD, the ROR for death with rivastigmine was 3.67, with a confidence interval from 1.92 to 7.00, and a p-value of 0.001. Two countries, two databases, and two sets of reports collected through different systems over different time periods produced the same signal. Look at the raw counts and the picture sharpens. In FAERS, rivastigmine accounted for 5,918 total reports, and 995 of those — nearly 17 percent — involved death. Donepezil had 95 death reports, and galantamine had 126. Donepezil's ROR for death was 0.26, and galantamine's was 0.51 — both significantly below the class average. Rivastigmine's signal wasn't just elevated; it was elevated while the other two drugs ran in the opposite direction. A secondary signal also emerged, and it's worth pausing on because it points in a completely different direction. Donepezil was strongly associated with rhabdomyolysis, which is a breakdown of muscle tissue that can lead to kidney failure. In FAERS, donepezil had 94 rhabdomyolysis reports and an ROR of 16.87, with a confidence interval from 10.39 to 27.41, and a p-value below 0.0001. The paper discusses a supporting case report of acute renal failure linked to donepezil through this mechanism. That's a separate safety concern, specific to donepezil, and a reminder that no drug in this class is without its risks. So why rivastigmine? Ali and colleagues offer two non-exclusive explanations, and they are careful not to overstate them. The first is pharmacological. Rivastigmine's pseudo-irreversible binding could prolong cholinergic effects beyond what the prescribing dose anticipates. If the enzyme stays blocked longer than expected, the downstream consequences — slowed heart rate, gastrointestinal distress, and respiratory depression — can accumulate. The second explanation is practical, and arguably more immediate: the patch. Transdermal drug delivery introduces a specific category of medication error that oral pills simply don't have. You can't accidentally take six pills without noticing. With patches, errors happen. Ali and colleagues cite a 2007 FDA and Novartis warning about improper Exelon patch use, and they reference a published fatal case described by Henrik Lövborg, Anna Jönsson, and Staffan Hägg — an eighty-seven-year-old patient whose healthcare providers applied six rivastigmine patches on two consecutive days. The patient developed nausea, vomiting, renal failure, electrolyte disturbances, and died. That case is not an anomaly in the literature; it's an illustration of a documented failure mode. When a patch is invisible under clothing, when a new shift of caregivers doesn't know the old one was applied, or when a patient in cognitive decline applies their own patch without remembering they already did, the drug accumulates in ways that oral dosing simply doesn't permit. The paper also acknowledges a potential confound: rivastigmine may be prescribed preferentially in more advanced Alzheimer's disease, and patients with more advanced disease carry higher baseline mortality risk. That's a real concern. A pharmacovigilance signal can't untangle the drug's effect from the severity of the condition it's treating. This is why Ali and colleagues consistently use the word "association" and are explicit that causality cannot be established from these databases alone. What this study does is quantify a signal that holds across two independent national reporting systems, identify a plausible pharmacological and practical mechanism, and place the finding on record. That's exactly what post-marketing surveillance is for. Clinical trials weren't designed to catch a mortality signal in a frail, elderly population over years of real-world use. FAERS and CVARD, imperfect as they are, exist to catch precisely this. The practical implications are measured. Cholinesterase inhibitors remain important treatments for Alzheimer's disease, and the paper is not an argument against rivastigmine. It's an argument for informed prescribing. When choosing between these three agents, clinicians should factor in caregiver reliability, the specific risks associated with transdermal dosing, and individual patient vulnerability. The patch offers real advantages, such as better gastrointestinal tolerability and steady drug delivery, but those advantages come with a failure mode that requires active management. Meanwhile, the rhabdomyolysis signal for donepezil warrants its own attention, particularly in patients with conditions that already compromise muscle or kidney function. What future work would look like is clear: prospective, head-to-head comparisons with prescription data attached, so researchers can calculate actual incidence rates rather than reporting odds ratios. That would let the field separate the drug's contribution from the disease's. Until then, this study gives prescribers and regulators two convergent national signals pointing in the same direction. That convergence, across different databases, different populations, and different time periods, is not something to set aside. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.