Reading a Drug Safety Signal: What FAERS and Disproportionality Can and Cannot Tell You

FAERS is the FDA's database of adverse-event reports, required from manufacturers and submitted voluntarily by health professionals and patients, and a disproportionality signal flags when a drug and an event are reported together more often than expected across the database. That is a reason to investigate, not evidence that the drug caused the event, because the reports are unverified, incomplete, and shaped by publicity and reporting habits. FAERS also cannot tell you how often an event happens, since there is no reliable count of how many people took the drug.

What FAERS is, and what it is not

The FDA Adverse Event Reporting System, FAERS, is a large database of reports describing suspected harms from drugs and biologic products. Health professionals, patients, and manufacturers submit them. Manufacturers are required to forward reports they receive, while reports from the public are voluntary. Anyone can search a public dashboard.

The crucial thing to hold onto is what a report is. It is someone's account that a person took a drug and something bad happened afterward. The FDA does not require proof that the drug caused the event, and reports are often incomplete. A report is a signal that something might be worth looking at, nothing more.

How a disproportionality signal is built

With millions of reports, the FDA and researchers look for drug and event pairs that appear together more often than you would expect if there were no association at all. This is disproportionality analysis. Measures such as the reporting odds ratio and the proportional reporting ratio compare how often a given event is reported for one drug against how often it is reported for all other drugs.

When a pair stands out, it is called a signal. The simpler measures are sensitive but not specific, meaning they catch a lot, including many false alarms. More conservative Bayesian methods exist to temper that tendency. Whatever the method, a signal is a statistical flag, a prompt to investigate, not a conclusion.

Why the numbers can mislead

The reports feeding these calculations are not a random sample of what happens in the world, and that is the heart of the problem. Reporting rises when a drug is new, when a lawsuit is in the news, or when a warning is added, all of which draw attention to certain events. This is reporting bias, and it can manufacture a signal out of publicity rather than pharmacology.

There is a deeper limit. FAERS has no denominator. It records reports of events but not how many people took the drug, so it cannot tell you how often an event actually occurs. The FDA is explicit that FAERS data cannot be used to calculate incidence in the population. A large count of reports can reflect a widely used drug, not a dangerous one.

What a signal actually triggers

Inside the FDA, a disproportionality signal is a beginning. Safety reviewers treat it as a hypothesis and then bring other evidence to bear: the clinical details of the reports, biological plausibility, published studies, and where available, controlled data from trials or large healthcare databases. Only after that fuller assessment might a signal lead to a label change, a safety communication, or a request for further study.

This is why a single alarming statistic pulled from the public dashboard, or from a paper that reports nothing but a disproportionality number, should be read with caution. The signal is real as a prompt. It is not, by itself, a measure of how much a drug should worry anyone.

How to read a FAERS-based claim

When you see a headline or a study built on FAERS, ask a few questions. Is the finding a disproportionality signal, or is it being described as if the drug caused the events? Could reporting bias explain the pattern, a recent warning, a lawsuit, a burst of media coverage? And is anyone quoting a rate or incidence, which FAERS cannot legitimately provide?

Used as designed, FAERS is a valuable early-warning net that has surfaced genuine safety problems too rare or too delayed for trials to catch. The error is treating its signals as verdicts. A careful reader keeps the two apart, and asks what happened after the signal, because that is where the real evidence gets made.