Real-World Evidence Sources for Drug Safety: How Registries and Claims Data Keep Medications Safe

When a new drug hits the market, we assume it’s safe. But clinical trials only involve a few thousand people over a few years. What happens when millions start taking it? That’s where real-world evidence comes in. Registries and claims data are two of the most powerful tools we have to catch rare side effects, track long-term outcomes, and make sure drugs stay safe after approval. These aren’t just backup systems-they’re now central to how regulators and drug makers monitor safety in the real world.

What Real-World Evidence Actually Means

Real-world evidence (RWE) isn’t lab data or controlled trial results. It’s what happens when drugs are used by real people in everyday life. Think of it as watching how a car performs on actual roads, not just a test track. The U.S. Food and Drug Administration (FDA) officially defined RWE in 2018 as clinical evidence drawn from real-world data (RWD)-information collected outside clinical trials. That includes hospital records, insurance claims, patient surveys, and registries tracking specific diseases or treatments.

Before 2016, regulators mostly relied on clinical trials. But those trials leave gaps. They often exclude older patients, people with multiple conditions, or those taking other medications. Real-world data fills those gaps. And since the 21st Century Cures Act passed in 2016, the FDA has increasingly used RWE to support drug approvals and safety updates. Between 2017 and 2021, the agency approved 12 drugs or new uses where RWE played a key role. Five of those relied directly on claims or registry data.

How Disease and Product Registries Work

Registries are like living databases for specific patient groups. Some track people with a disease-like cystic fibrosis or Parkinson’s. Others track people using a specific drug, like a new cancer therapy. These aren’t random collections of records. They’re carefully designed systems that collect standardized information: diagnosis codes, lab results, treatment changes, side effects, and even how patients feel day to day.

Take the Cystic Fibrosis Foundation Patient Registry. It tracks over 30,000 patients in the U.S. When the drug ivacaftor was approved, clinical trials showed it worked well. But the registry spotted something unexpected: patients with rare CFTR gene mutations had stronger responses than anyone predicted. That insight changed how doctors prescribed the drug.

Disease registries vary in size. Some are small, run by a single hospital with a few hundred patients. Others, like the SEER cancer registry, cover nearly half the U.S. population. They’re rich in detail-87% of lab values and imaging results are recorded, compared to just 52% in claims data. But they’re expensive. Setting up a registry takes 18 to 24 months and costs between $1.2 million and $2.5 million. Annual upkeep runs $300,000 to $600,000. And because they rely on voluntary participation, only 60-80% of eligible patients join. That can introduce bias.

Claims Data: The Power of Millions of Records

Claims data is what insurance companies collect when they pay for care. Every time someone visits a doctor, fills a prescription, or goes to the hospital, a claim is generated. These claims include diagnosis codes (ICD-10), procedure codes (CPT), and drug codes (NDC). They don’t have detailed lab results or patient stories-but they have scale.

IBM MarketScan tracks 200 million lives. Optum covers 100 million. Truven Health has 150 million. Medicare claims alone cover over 60 million Americans. That’s enough data to find rare side effects that clinical trials would never catch. For example, in 2015, the FDA analyzed 1.2 million Medicare records over five years to check if the Parkinson’s drug entacapone increased heart risks. They found no link.

Claims data is also long-lasting. Medicare records go back 15+ years. That’s crucial for spotting problems that take years to show up-like bone fractures from long-term osteoporosis drugs or liver damage from chronic pain meds. But claims data has blind spots. Only 45-60% of lab values are recorded. Patient-reported symptoms like fatigue or brain fog? Almost never captured. And coding errors happen. The Agency for Healthcare Research and Quality estimates 15-20% of diagnosis codes are wrong.

Registries vs. Claims Data: When to Use Which

Here’s the simple breakdown:

• Use registries when you need deep clinical detail-like tracking how a drug affects patients with rare mutations, or measuring quality of life over time.
• Use claims data when you need big numbers-like detecting a side effect that occurs in 1 in 10,000 patients, or monitoring safety across an entire population.

For rare events, claims data needs about 1 million records to be confident. Registries can do it with 500,000 because their data is cleaner and more complete. A 2021 FDA paper showed registries are 37% better at capturing long-term outcomes than claims data alone. But claims data wins on reach.

The FDA used claims data to approve the cancer drug palbociclib for new patient groups in 2019. But they also used registry data from transplant patients to support a safety update for tacrolimus in 2021. The best approach? Combine them. The International Council for Harmonisation (ICH) found that using both sources cuts false safety signals by 40%. That’s huge-because every false alarm wastes time and creates unnecessary fear.

Regulators Are Taking Notice

The FDA doesn’t just accept RWE-it’s building systems around it. The Sentinel Initiative, launched in 2008, connects 11 major healthcare systems and 3 claims processors to monitor 300 million patient records. In 2022, the FDA reviewed 107 RWE submissions. In 2018, that number was 29.

The European Medicines Agency (EMA) launched Darwin EU in 2021 to do the same across Europe. By late 2023, it connected 32 databases across 15 countries, covering 120 million people. These aren’t pilot projects. They’re infrastructure.

Dr. Amy Abernethy, former FDA deputy commissioner, said registry studies can provide evidence nearly as strong as randomized trials for safety questions. Dr. Janet Woodcock, former head of FDA’s drug evaluation unit, called claims databases “indispensable” for spotting rare risks.

But it’s not all smooth sailing. Dr. Joseph Ross from Yale warns that claims data alone can’t prove cause-and-effect. He found that 22% of safety signals from claims data turned out to be false positives once doctors reviewed actual medical records. That’s why regulators now require statistical methods to fix common biases-like “immortal time bias,” where patients are incorrectly counted as being at risk before they even started the drug. Proper methods reduce that bias by 35-50%.

What’s Changing in 2024 and Beyond

The field is moving fast. In January 2024, the FDA released new draft guidance requiring registries to maintain at least 80% data completeness on key variables. That’s a big step toward standardization.

New tech is helping too. AI tools now scan millions of records to flag unusual patterns. A 2024 study in JAMA Network Open showed AI cut false safety signals by 28%. Novartis is even combining claims data with wearable device readings-like heart rate and activity levels-to monitor heart failure patients on Entresto.

The FDA’s REAL program, launched in 2023, aims to standardize registry data collection for 20 priority diseases by 2026. The first focus? Rare diseases. Why? Because traditional trials can’t recruit enough patients. Registries are the only way to understand safety in these groups.

Why This Matters to Patients and Doctors

You might think this is all behind-the-scenes stuff. But it’s not. Every time your doctor prescribes a new medication, they’re relying on data collected from people just like you. If a drug causes a rare liver problem in 1 in 50,000 users, it’s likely that a registry or claims database caught it before the warning went out.

For patients with chronic conditions-diabetes, cancer, autoimmune diseases-this means better, safer treatments. For doctors, it means more confidence in prescribing. And for everyone, it means drugs are being watched longer and more closely than ever before.

The old model-approve a drug, then wait for complaints-is gone. Today’s system is proactive, data-driven, and built on the real experiences of millions of patients. Registries and claims data aren’t perfect. But together, they’re the most powerful safety net we’ve ever had for medications.

Real-world evidence is no longer a side note in drug safety. It’s the backbone of modern pharmacovigilance. Registries and claims data-despite their flaws-are giving us a clearer, more complete picture of how medications affect real people. And as technology improves and global systems like Darwin EU expand, we’ll get even better at catching risks before they become crises.