What FDA Adverse-Event Data Can—and Cannot—Show

An FDA medical-device report can show that somebody reported a suspected death, serious injury, malfunction or other problem involving a device. It can help a clinic operator, clinician, researcher or journalist identify a pattern worth investigating.

It cannot, by itself, show that the device caused the event. It cannot calculate the chance that an event will happen. And a larger report count does not establish that one laser, radiofrequency platform or other aesthetic device is less safe than another.

The short answer — United States, evidence checked July 22, 2026: FDA adverse-event reports are signal-detection records, not a safety league table. Use them to form a question, then test that question against device exposure, clinical studies, recalls, labeling and other evidence.

The governing distinction is simple but consequential: a report is evidence that an event was reported, not proof of what caused it or how often it occurs.

What MAUDE actually contains

The FDA's Manufacturer and User Facility Device Experience database, usually called MAUDE, houses medical device reports submitted by mandatory reporters—manufacturers, importers and device user facilities—and by voluntary reporters including health professionals, patients and consumers.

These are reports of suspected device-associated deaths, serious injuries and malfunctions. Under 21 CFR Part 803, reporting duties differ by actor. For example, the regulation requires device user facilities to report certain deaths and serious injuries, while manufacturers must report certain deaths, serious injuries and malfunctions. The details and deadlines depend on the reporter and event.

That collection model explains both the value and the limits of the data. Reports arrive from different observers with different duties, access to information and follow-up capacity. A narrative may be detailed, sparse, corrected later or partly redacted.

The FDA's MAUDE dataset overview describes the system as passive surveillance and warns that submissions may be incomplete, inaccurate, untimely, unverified or biased. It also says report inclusion may be delayed.

What one report can show

A record can document the contents of a submission. Depending on the report, fields may include:

• event type, such as death, injury or malfunction;
• brand, generic device name, model, product code or manufacturer;
• event and receipt dates;
• device and patient problem terms;
• report source and initial or follow-up status; and
• event, manufacturer and device-evaluation narratives.

Those fields can support a bounded conclusion: “The database contains a report with these recorded characteristics.” They can also reveal recurring language, device components or failure modes that deserve closer review.

They do not establish that every field is complete or independently verified. The FDA states that submission and public release of a report are not necessarily an admission that a product, facility, manufacturer, user or medical professional caused or contributed to the event.

Why a report cannot prove causation

Causation asks whether the device produced the outcome. A MAUDE report usually does not provide the controlled comparison, complete clinical record, verified exposure history or independent device evaluation needed to answer that question.

Several explanations may remain possible. The device may have malfunctioned. The treatment technique, settings, maintenance, patient factors, another product or an unrelated condition may have contributed. Some records contain manufacturer follow-up or evaluation; others do not. Even a detailed narrative remains a report, not a controlled study.

What the report shows: an event and device involvement were described to the reporting system.

What it does not show: that the device caused the event, that the report is complete, or that the same outcome would occur under other conditions.

Why report counts are not incidence rates

An incidence rate needs a numerator and a denominator over a defined period. MAUDE can supply report records that may contribute to a numerator. It does not supply the number of times each device was used, the number of patients treated or a complete count of all events.

The FDA's MAUDE overview identifies two central problems: potential under-reporting and the lack of information about frequency of device use. It therefore says the incidence or prevalence of an event cannot be determined from the reporting system alone.

That means a statement such as “Device A had 40 reports in a year” is incomplete even if the database count is accurate. The device's installed base, treatment volume, time on the market, reporting behavior and follow-up structure are unknown or external to that count.

Why raw counts cannot compare devices

The same denominator problem defeats a simple device ranking. If one platform is used far more often than another, it may generate more reports even if its event rate is lower. A newer device may have less exposure time. Public attention, manufacturer reporting systems, litigation or a safety communication may change the likelihood that an event is reported.

The FDA is explicit: MAUDE data alone cannot establish event rates, evaluate changes in rates over time or compare event rates between devices. The number of reports should not be used in isolation to conclude how severe or frequent a device problem is.

This is the strongest reasonable objection to using MAUDE as a comparative tool: the database is large and searchable, so counting feels objective. The count may indeed be reproducible. The interpretation is still invalid without comparable exposure and stronger controls.

Watch for related and follow-up reports

A raw result row should not automatically be treated as one unique patient, procedure or underlying event. Reports can come through different source pathways, and records may identify initial and follow-up submissions. That creates a risk that related submissions are counted as independent events.

Before counting, compare the available report identifiers, manufacturer report numbers, dates, device details, source fields, initial/follow-up labels and narratives. Similarity does not prove duplication, and missing fields may prevent a clean determination. The defensible output may be a range or an unresolved related-report flag rather than a single “case” count.

How to search MAUDE without fooling yourself

Start with an exact device identity, not a treatment category. “Laser” is too broad: FDA-regulated lasers include devices used far outside medical aesthetics. A broad search can therefore retrieve technically matching but irrelevant ophthalmic, surgical or other records.

1. Identify the device. Record the exact brand, manufacturer, model and FDA product code from reliable device documentation or an FDA device record.
2. Set one bounded period. Use the FDA-received date and record the dates exactly.
3. Search more than one identifier. Trade names, company names and product descriptions vary. The FDA warns that those variations affect search results.
4. Export or log the query. Preserve the search terms, filters, access date and result set.
5. Read records, not just totals. Check event type, source, follow-up status, device identity and narratives.
6. Mark uncertainty. Separate confirmed matches, possible matches, possible related reports and exclusions.

A reproducible openFDA example

The openFDA device adverse-event API exposes MAUDE data in a machine-readable format. This example asks for reports received during 2025 where the generic device name contains “laser”:

https://api.fda.gov/device/event.json?search=date_received:[20250101+TO+20251231]+AND+device.generic_name:laser&limit=1

Shows: how to fix a receipt-date window, search one field and retrieve a record with API metadata. The endpoint returned successfully when checked on July 22, 2026.

Does not show: an aesthetic-only result set, a complete event count, a causal relationship or an incidence rate. The broad term can retrieve non-aesthetic lasers. For a substantive review, replace it with validated identifiers for the exact device and document every inclusion and exclusion rule.

The API is a different access route to MAUDE-derived data, not an independent confirming dataset. Agreement between the MAUDE interface and openFDA should not be described as two-source corroboration.

How device problems are reported

Mandatory reporting is governed by federal law and depends on the actor. Part 803 sets requirements for manufacturers, importers and defined device user facilities. A physician's office is excluded from the regulation's definition of a “device user facility,” although another facility category or actor may have a separate reporting duty.

Patients, consumers and health professionals can also report problems voluntarily through MedWatch, the FDA's medical product safety reporting program. Whether a specific clinic or person has a mandatory duty is a legal and fact-specific question; the presence of a voluntary pathway does not answer it.

What stronger evidence looks like

A MAUDE signal becomes more informative when tested against evidence designed for the question. Depending on the claim, that may include:

• device labeling, safety communications, recalls and FDA regulatory records;
• well-designed clinical studies with defined populations and denominators;
• active surveillance or registries with known exposure;
• independent device evaluation and clinical-record review; and
• consistent findings across sources that do not merely reuse the same MAUDE reports.

No single evidence type answers every question. A trial may have controlled follow-up but limited sample size or generalizability. A registry may capture practice conditions but retain confounding. A recall describes a regulatory action, not the incidence of every possible outcome. The method must match the claim.

The bottom line

FDA adverse-event data is valuable because it can expose suspected device problems that merit investigation. Its value disappears when a report is turned into proof of causation or a report count is turned into a safety rate.

For a defensible review, define the exact device, date window and question; preserve the query; inspect individual records; flag related or follow-up reports; and compare the signal with evidence that has an exposure denominator and a design capable of answering the claim.