Understanding Adverse Event Rates: Percentages and Relative Risk in Clinical Trials

When a new drug is tested in clinical trials, one of the most important questions regulators and doctors ask is: how safe is it? It’s not enough to say, "15 out of 100 people had a headache." That number tells you nothing about how long those people were actually taking the drug. If one group took the drug for 3 months and another for 2 years, comparing simple percentages is like comparing apples to oranges. This is why understanding adverse event rates using percentages and relative risk isn’t just technical jargon-it’s critical to knowing whether a treatment is truly safe.

Why Simple Percentages Mislead

For years, the standard way to report adverse events was to calculate the Incidence Rate (IR): the number of people who had an event divided by the total number of people exposed. If 15 out of 100 patients got a rash, you’d say the rate was 15%. Simple. Easy to understand. But deeply flawed.

Here’s the problem: what if 50 of those patients were only on the drug for 14 days, while the other 50 stayed on it for 18 months? The 15% number doesn’t reflect that. Someone who had a rash after two weeks counts the same as someone who had it after 18 months. The result? The true risk gets buried. A 2010 analysis cited by PharmaSUG showed that using IR alone can underestimate actual event rates by 18% to 37% when exposure times vary between groups. That’s not a small error-it’s a dangerous blind spot.

The FDA noticed. In 2023, during a review of a Supplemental Biologics License Application (sBLA), they explicitly requested that the sponsor switch from IR to an exposure-adjusted method. That was a turning point. It signaled that regulators no longer accept simplistic percentages as sufficient evidence of safety.

Enter Exposure-Adjusted Incidence Rate (EAIR)

The industry is shifting toward a more accurate metric: the Exposure-Adjusted Incidence Rate (EAIR). Unlike IR, EAIR accounts for how long each patient was actually exposed to the drug. It’s calculated by dividing the total number of adverse events by the total patient-time at risk-measured in years.

For example, if 20 events occurred across 1,200 patient-years of exposure, the EAIR is 20 ÷ 1,200 = 0.0167 events per patient-year. That’s often expressed as 1.67 events per 100 patient-years. Now you’re not just counting people-you’re counting time. This makes comparisons fairer. If one drug group had longer treatment duration, EAIR adjusts for it. If a patient had multiple events, EAIR counts them all, giving a fuller picture of burden.

This matters most in chronic conditions. Think diabetes, rheumatoid arthritis, or cancer maintenance therapy. Patients stay on treatment for years. A rash that occurs once every 100 patient-years might seem rare-but if 10,000 patients are on the drug for five years, that’s 50,000 patient-years. Suddenly, 1.67 events per 100 patient-years becomes 835 total events. That’s not a minor side effect. That’s a population-level safety signal.

How EIR Differs and When It Falls Short

Another method you’ll see is the Event Incidence Rate (EIR), measured in events per 100 patient-years. It sounds similar to EAIR, but there’s a key difference: EIR often counts the number of events per person, not the number of people affected. That’s fine for rare, one-time events-but problematic for recurring ones.

Imagine two patients on the same drug. Patient A gets diarrhea once. Patient B gets it 12 times in a year. EIR counts both as 12 events. But IR would say two people had diarrhea. EAIR would say 13 events over two patient-years. Each method answers a different question. EIR is good for understanding frequency. But if you want to know how many people are affected, IR still has value. The problem arises when EIR is used as a proxy for risk to individuals-it can inflate perceived danger.

The FDA and EMA don’t require one method over another. But they do require justification. If you use IR, you must explain why exposure time doesn’t matter. If you use EAIR, you must show how you calculated patient-years and handled treatment interruptions. The European Medicines Agency (EMA) is more flexible, but still insists on transparency. The FDA’s 2024 draft guidance on exposure-adjusted analysis is pushing for standardization-and it’s likely to become mandatory in the next few years.

Relative Risk and Why It Matters

Once you have accurate rates, you can calculate relative risk. This is the ratio of the adverse event rate in the treatment group versus the control group. If the EAIR for a drug is 2.1 events per 100 patient-years and the placebo group is 0.8, the relative risk is 2.1 ÷ 0.8 = 2.63. That means patients on the drug are more than twice as likely to experience the event per unit of time.

But here’s the catch: relative risk only makes sense if the underlying rates are accurate. If you use IR instead of EAIR, your relative risk could be completely wrong. A drug might look safer than it is because the treatment group had shorter exposure. Or it might look riskier because the control group was followed longer. Either way, the decision to approve or restrict the drug could be based on faulty math.

Statisticians use confidence intervals to show how certain they are about these numbers. For incidence rates, the Wilson score method with continuity correction is preferred. For relative risk, the Wald method is standard. In R, these are calculated using functions like prop.test and riskratio. These aren’t just technical details-they’re what separate a credible safety profile from a misleading one.

Real-World Impact: When the Numbers Changed Everything

MSD’s safety team reported in 2023 that switching to EAIR revealed previously undetected safety signals in 12% of their reviewed programs. One example: a chronic pain medication. Using IR, the rate of liver enzyme elevations looked low. But when they calculated EAIR, they saw a clear increase in events per patient-year among those on long-term therapy-especially those who’d been on the drug for over 18 months. That signal had been hidden because most patients in the trial were only treated for 6 months.

On the flip side, Roche found that 35% of medical reviewers initially misinterpreted EAIR results. They thought higher numbers meant the drug was more dangerous-without realizing that longer exposure naturally increased event counts. That led to internal training sessions and clearer labeling in safety reports. The lesson? Better methods mean better data-but only if people understand how to read it.

What You Need to Know for Regulatory Submissions

If you’re working in clinical development, here’s what’s non-negotiable:

• Report both IR and EAIR for serious adverse events, especially in Phase 3 trials.
• Use CDISC ADaM datasets to structure exposure and event data properly.
• Document exactly how you calculated patient-years: start date, end date, how interruptions were handled.
• Include confidence intervals for all rates and relative risks.
• Justify your choice of method in your clinical study report.

CDISC’s Oncology Therapeutic Area Guide (v3.0, 2023) now requires both IR and EAIR for serious adverse events. The FDA’s Biostatistics Review Template includes checklists for exposure-adjusted analysis. Non-compliance is no longer an oversight-it’s a rejection risk.

The Future Is Exposure-Adjusted

The global clinical trial safety software market hit $1.84 billion in 2023, growing at over 22% annually. Why? Because regulators are demanding better data. CDISC adoption among top pharma companies is at 89%. Regulatory submissions with exposure-adjusted metrics jumped from 12% in 2020 to 47% in 2023.

By 2027, experts predict 92% of Phase 3 submissions will include EAIR. The FDA’s Sentinel Initiative is even building machine learning tools to detect safety signals using exposure-adjusted metrics-and early tests show a 38% improvement in early detection.

This isn’t just a trend. It’s a fundamental upgrade in how we measure safety. Percentages alone are outdated. Relative risk without exposure adjustment is misleading. The future belongs to those who can calculate, interpret, and communicate risk using time-weighted data.

When you see a drug’s safety data, don’t just look at the percentage. Ask: How long were people exposed? That’s the question that separates noise from real insight.