Methodology

Data Source

Drug Fact Check uses data from the FDA Adverse Event Reporting System (FAERS), accessed through the openFDA Drug Adverse Events API. The openFDA API is a public, RESTful API maintained by the FDA that provides access to structured FAERS data without requiring registration or API keys (though we use an API key for higher rate limits).

FAERS contains adverse event reports submitted to the FDA from 1969 to the present, though our data primarily covers recent years where electronic reporting has made the data more complete and standardized. The FDA updates the openFDA dataset quarterly.

We are not affiliated with or endorsed by the FDA, openFDA, or any government agency. The openFDA API provides the data under the openFDA terms of service.

Drug Selection Criteria

Not every drug in the FAERS database appears on Drug Fact Check. We apply minimum report thresholds to ensure the data displayed is statistically meaningful and useful:

• Minimum report count: Drugs must have a sufficient number of adverse event reports to be included. This filters out drugs with very few reports where individual cases could be misleading.
• Name normalization: Drug names in FAERS can be inconsistent — brand names, generic names, misspellings, and abbreviations may all appear. We normalize drug names using openFDA's drug identification fields to consolidate reports under a single drug entry.
• Active ingredient matching: Where possible, we link brand names to their generic (active ingredient) names using openFDA's drug label data.

Reaction Aggregation

Adverse reactions in FAERS are coded using MedDRA (Medical Dictionary for Regulatory Activities) Preferred Terms. For each drug, we count how many reports mention each reaction term.

Important considerations:

• A single report can list multiple reactions. If a patient reports experiencing both nausea and headache while taking a drug, both reactions are counted separately.
• Reactions are counted at the Preferred Term level, which is the standard level of specificity in MedDRA. More general (System Organ Class) and more specific (Lowest Level Term) groupings are not displayed.
• Reactions are sorted by frequency (most reported first) on each drug page.
• The "% of Total" column on drug pages shows each reaction's count as a percentage of all reaction mentions for that drug, not as a percentage of total reports.

Drug Class Classification

Drugs are grouped into therapeutic classes using the Established Pharmacologic Class (EPC) information available through the openFDA API. EPC classifications come from FDA drug labeling and describe a drug's mechanism of action, physiologic effect, or chemical structure.

Some drugs belong to multiple pharmacologic classes and will appear on multiple drug class pages. Drugs without an EPC designation in the openFDA data are not assigned to any class on this site.

Severity Metrics

We calculate several metrics to help users understand the severity profile of adverse events for each drug:

• Serious reports: Reports flagged as "serious" in FAERS, meaning the outcome involved death, hospitalization, disability, a life-threatening event, a congenital anomaly, or required medical intervention to prevent one of these.
• Death reports: The subset of serious reports where death was listed as an outcome. This does not indicate causation — the patient may have died from their underlying condition or an unrelated cause.
• Death report ratio: Death reports divided by total reports, expressed as a percentage. This is color-coded as Low (under 1%, green), Moderate (1-5%, yellow), or High (over 5%, red).

These metrics provide relative context but should not be interpreted as measures of drug safety without considering the drug's therapeutic area, patient population, and total prescribing volume.

Demographic Data

Where available, we display a breakdown of reports by patient sex (male and female). This data comes directly from FAERS reports and has several limitations:

• Sex is not reported in all cases. The demographic totals may be significantly lower than the total report count.
• FAERS only captures binary sex categories (male/female) as reported on the form.
• Differences in reporting rates between sexes may reflect prescribing patterns, biological differences in drug response, or differences in reporting behavior rather than true differences in adverse event rates.
• Age and weight data are available in raw FAERS data but are not displayed on this site due to high rates of missing or inconsistent values.

Known Limitations

FAERS data has well-documented limitations that users should understand when interpreting the information on this site:

Voluntary Reporting

Most adverse event reports are submitted voluntarily by patients and healthcare professionals. Studies estimate that only 1-10% of adverse events are actually reported to the FDA. This means the data on this site represents a fraction of total adverse events and cannot be used to calculate true incidence rates.

No Proven Causation

An adverse event report does not prove that the drug caused the reported event. The patient may have had a pre-existing condition, been taking other medications, or experienced a coincidental health problem. FAERS is a signal detection tool, not a proof of causation.

Reporting Bias

Several factors systematically affect which events get reported:

• Serious events are reported more frequently than mild ones
• Drugs that receive media attention generate more reports
• Newly approved drugs may be under closer scrutiny
• Manufacturer-mandated reporting captures more events for serious outcomes
• Widely prescribed drugs accumulate more reports simply due to larger patient populations

Duplicate Reports

The same adverse event may be reported multiple times by different parties — the patient, their doctor, and the drug manufacturer might all submit reports for a single incident. While the FDA attempts to de-duplicate reports, some duplicates remain in the database.

Data Completeness

Many FAERS reports have missing fields. Patient demographics, concomitant medications, and detailed event descriptions are frequently incomplete. The data we display is only as complete as the underlying reports.

No Denominator Data

FAERS does not include information about how many people took a drug without experiencing an adverse event. Without this denominator, it is impossible to calculate true adverse event rates from FAERS data alone.

Data Processing Pipeline

Our data pipeline works as follows:

1. Query the openFDA Drug Adverse Events API for drugs meeting our inclusion criteria
2. Normalize drug names and link brand names to generic equivalents
3. Aggregate reaction counts, severity metrics, and demographic data for each drug
4. Build reverse lookups from reactions to drugs
5. Classify drugs into pharmacologic classes using EPC data
6. Generate static pages for each drug, reaction, and drug class

This entire site is statically generated, meaning all pages are pre-built at deployment time. No user queries hit the openFDA API in real time.