Millions Use GLP-1 Drugs — But Safety Data Remains Scarce
Prescriptions for GLP-1 receptor agonists — the class of medications that includes semaglutide, sold under the brand names Ozempic and Wegovy — have climbed at a pace that few drug launches in modern history can match. According to tracking data from IQVIA, a health data analytics firm, GLP-1 prescriptions in the United States more than doubled between 2021 and 2023, with millions of Americans now on these medications for type 2 diabetes management or weight loss. The Centers for Medicare and Medicaid Services has flagged the drug class as one of the fastest-growing cost drivers in the federal health program.
That scale of adoption makes the next fact both surprising and consequential: systematic, population-level data on how often GLP-1 drugs cause serious harm remains thin. When physicians, researchers, and now plaintiffs' attorneys ask how frequently these medications produce life-altering adverse events — vision loss, severe gastrointestinal damage, or other significant injuries — the honest scientific answer is that no one can say with confidence. This is not a fringe concern raised by skeptics. It is a structural problem in post-market drug surveillance that experts have flagged for years, now sharply illuminated by the scale of GLP-1 adoption.
The Gaps in Reporting Serious Adverse Events
The primary federal mechanism for tracking drug harm after approval is the FDA Adverse Event Reporting System, known as FAERS. The database is large — containing tens of millions of reports accumulated over decades — and it does capture signals that regulators act on. But FAERS operates on a voluntary basis. Healthcare providers, patients, and manufacturers can submit reports, but none except manufacturers of certain products are legally required to do so for every event.
Read next Medicaid Work Requirements Strand Cancer SurvivorsThe consequence is systematic underreporting. Studies published in journals including Drug Safety and Pharmacoepidemiology and Drug Safety have estimated that FAERS captures somewhere between 1 and 10 percent of actual adverse events, depending on the severity and type of reaction. Serious, hospitalization-level events tend to have better capture rates than mild reactions — but the gap between observed and actual frequency remains enormous.
For a drug taken by a small population over a short period, this gap might be manageable. For a drug class prescribed to millions of people across multiple years, the uncaptured portion of the iceberg grows proportionally. A pharmacovigilance researcher at an academic medical center studying drug safety signals faces a fundamental statistical challenge: with voluntary reporting, what looks like a rare signal might simply be a commonly unreported one, and separating those two possibilities demands study designs that take years and significant funding to execute.
This is what makes the current moment with GLP-1 drugs serious side effects surveillance particularly fraught. The drugs are widely used, the theoretical mechanisms for several serious harms are biologically plausible, and the real-world incidence data needed to answer basic safety questions does not yet exist at the scale the public health situation demands.
Blindness, Pancreatitis, and Other Serious Concerns
Two GLP-1 drugs serious side effects have attracted particular scientific and legal attention: a specific form of vision loss called non-arteritic anterior ischemic optic neuropathy, known as NAION, and pancreatitis.
NAION involves a sudden reduction in blood supply to the optic nerve, causing partial or complete blindness, typically in one eye. It is not a common condition in the general population, which makes any potential drug-associated signal notable. Researchers have proposed mechanistic pathways by which GLP-1 agonists might affect vascular tone in ways that could, in theory, influence optic nerve perfusion — but moving from biological plausibility to confirmed causation requires the kind of large-scale epidemiological data that does not yet exist in clean form.
Pancreatitis, inflammation of the pancreas, has appeared in FAERS reports associated with GLP-1 drugs for years, and the FDA has required label updates reflecting that association. Whether the drug class actually causes acute pancreatitis at a clinically meaningful rate above background, or whether patients with metabolic conditions — who are already at elevated risk for pancreatic problems — are simply more likely to be on these medications, has been debated in the literature without full resolution.
Data gaps cut in both directions here. The absence of robust incidence data does not confirm that serious harms are occurring at high rates. It also does not rule it out. That epistemic uncertainty is uncomfortable, but it is the honest position the evidence currently supports.
Legal Battles Highlight the Evidence Problem
The evidentiary void is no longer just a scientific inconvenience. It is now a focal point in litigation. Families like the Engels, who have been associated with legal action over GLP-1 drug injuries, are bringing claims that require courts to assess causation — a question that hinges on epidemiological frequency data that is, at best, incomplete.
Plaintiffs' attorneys in product liability cases must typically demonstrate general causation: that a drug is capable of causing the harm alleged, at a frequency that exceeds background rates. Defendants, in turn, can point to the absence of conclusive population-level evidence. When the reporting system is voluntary and widely acknowledged to undercount, both sides can use the same data gap to support opposing conclusions. Courts are not well-equipped to adjudicate epidemiological adequacy disputes, and this dynamic is likely to result in inconsistent outcomes across jurisdictions — a poor substitute for the clear signal that rigorous surveillance would provide.
The structure of pharmaceutical litigation in the United States has historically provided some pressure on manufacturers to investigate and disclose risk. But litigation is a slow, expensive, and adversarial mechanism for generating safety knowledge. It does not scale to the millions of people currently on GLP-1 therapy.
What Regulators and Researchers Are — and Aren't — Doing
The FDA does have post-market surveillance tools beyond FAERS. The agency's Sentinel System, a distributed network of electronic health records and insurance claims data covering more than 100 million Americans, can run near-real-time epidemiological queries on specific drug-outcome pairs. Sentinel has been used to study GLP-1 drugs, and some analyses have been published or are underway.
The challenge is bandwidth and prioritization. Sentinel can run a finite number of studies at any time, and with GLP-1 drugs prescribed for at least two distinct indications — diabetes and obesity — across highly heterogeneous populations, the number of safety questions that could be asked far exceeds current capacity to answer them systematically.
Academic pharmacovigilance researchers note a structural problem: the more successful a drug is commercially, the larger the gap between the scientific resources devoted to understanding its post-market safety and the scale of real-world exposure. Blockbuster drugs generate enormous revenue for manufacturers and enormous prescription volumes, but they do not automatically generate the epidemiological infrastructure needed to detect rare serious outcomes across diverse populations.
What Patients Should Know When Taking GLP-1 Medications
None of this means patients should stop or avoid GLP-1 drugs absent medical guidance suggesting otherwise. For many people with type 2 diabetes or significant obesity, these medications have demonstrated meaningful clinical benefits in large randomized trials. The risk-benefit calculation is a medical judgment that belongs to patients and their physicians.
What the current surveillance situation does suggest is that patients taking these medications have a concrete interest in the quality of the data being generated about them. Reporting adverse events to FAERS directly — a process patients can do through the agency's MedWatch portal — contributes to the signal detection system, even knowing that system's limitations. Patients experiencing serious or unexpected symptoms should document those experiences with their healthcare providers, creating clinical records that, in aggregate, constitute a form of informal surveillance.
The broader problem — the structural underpowering of post-market safety systems for widely used drugs — is a policy issue that predates GLP-1 medications and will outlast them. The volume of people now on these therapies makes it one of the clearest current examples of a known gap with direct public health stakes. Improved mandatory reporting, expanded funding for the Sentinel network, and investment in independent pharmacovigilance research are the kinds of solutions researchers and patient advocates have proposed for years. Whether the scale of GLP-1 adoption finally creates the political will to pursue them is a question the evidence alone cannot answer.
Related coverage
Source: NPR Topics: News
