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

When you hear that a drug caused headaches in 15% of patients, it sounds simple. But what if one group took the drug for 3 months and another took it for 2 years? That 15% doesn’t tell the full story. In clinical trials, how you measure adverse events changes everything - and the FDA is pushing for better methods.

Why Simple Percentages Mislead

The most common way to report adverse events is the incidence rate (IR). It’s just the number of people who had an event divided by the total number of people in the study. If 15 out of 100 people got a rash, you say, "15% incidence." But here’s the problem: IR ignores how long people were exposed to the drug. In a trial where one group got the treatment for 6 weeks and another for 24 months, the longer-exposed group will naturally have more events - not because the drug is more dangerous, but because they had more time to experience side effects. A 2010 analysis found IR underestimates true risk by 18% to 37% in trials with uneven exposure times. That’s not just a small error - it’s a dangerous blind spot.

Enter Patient-Years: The EIR Method

To fix this, statisticians started using event incidence rate adjusted by patient-years (EIR). Instead of counting people, you count time. One patient taking a drug for one year = one patient-year. If five patients each took the drug for 2 years, that’s 10 patient-years. If 15 adverse events happened across those 10 patient-years, the rate is 150 events per 100 patient-years.

This method works well for recurring events - like nausea or dizziness - that happen more than once per person. It gives you a sense of frequency, not just who was affected. JMP Clinical and other regulatory tools now calculate EIR automatically using treatment start and end dates (TRTSDTM and TRTEDTM). But EIR has its own flaw: it counts events, not people. If one person gets 10 headaches, that’s 10 events. That inflates the rate, making the drug look riskier than it is for most users.

The FDA’s New Standard: EAIR

In 2023, the FDA requested that a biologics company switch from IR to exposure-adjusted incidence rate (EAIR) in a regulatory submission. That was a turning point. EAIR doesn’t just adjust for time - it also accounts for how many times an event occurs per person and handles treatment interruptions.

EAIR is calculated by dividing the total number of events by the total exposure time, but with careful handling of when people stopped and restarted treatment. If someone drops out after 3 months, their exposure stops there. If they come back later, their time resumes. This matters because real patients don’t take drugs perfectly - they miss doses, pause treatment, or switch due to side effects.

MSD’s safety team found that switching to EAIR uncovered safety signals in 12% of their chronic therapy programs that IR had missed. That’s not noise - that’s real risk hiding in plain sight.

Relative Risk and Why It’s Not Enough

You might think, "Just compare the rates between groups - that’s relative risk." But relative risk only works if both groups had the same exposure time. If Group A had 100 patient-years and 12 events (12 per 100 PY), and Group B had 200 patient-years and 20 events (10 per 100 PY), the relative risk is 1.2. But if you used IR instead - Group A: 12/50 = 24%, Group B: 20/50 = 40% - you’d say Group B had 67% higher risk. That’s wrong. The real difference is 20%, not 67%.

The FDA and EMA now require clear justification for which method you use. The International Council for Harmonisation (ICH) E9(R1) guidelines, updated in 2020, say you must consider exposure time and treatment discontinuation. But they don’t tell you which formula to pick. That’s up to you - and your data.

What the Industry Is Doing

The shift isn’t theoretical. In 2020, only 12% of FDA submissions included exposure-adjusted metrics. By 2023, that jumped to 47%. CDISC, the global standard for clinical data, now requires both IR and EAIR for serious adverse events in oncology trials (v3.0, 2023). The global clinical trial safety software market hit $1.84 billion in 2023 - growing at 22.7% a year - because companies can’t afford to get this wrong.

But adoption isn’t smooth. A 2024 PhUSE survey found that 68% of companies now calculate EAIR alongside IR, but 42% ran into formatting problems when submitting to regulators. SAS programmers spent 3.2 times longer building EAIR analyses than IR ones. Common errors? Incorrect event dates (28%), ignoring treatment breaks (19%), and miscalculating patient-years (23%).

Roche reported that 35% of medical reviewers didn’t understand EAIR at first. They thought higher numbers meant worse safety - not realizing it was just a rate per time unit. Training became essential.

Competing Risks and the Hidden Danger

There’s another layer most people ignore: competing risks. If a patient dies before having a heart attack, you never see the heart attack. But if you use the standard Kaplan-Meier method to estimate risk, you treat death as if it’s just a lost observation - not a competing event. That inflates the estimated risk of the heart attack.

A 2025 study in Frontiers in Applied Mathematics and Statistics showed that using cumulative hazard ratio estimation - which separates death from the event of interest - improved accuracy by 22% when competing events happened in more than 15% of cases. This isn’t niche math. It’s critical for drugs used in elderly or cancer patients, where death is a real possibility.

What You Need to Know

- IR is simple but misleading if exposure times vary. Use it only for short-term trials with uniform dosing.

- EIR is good for recurring events but overstates risk if one person has many events.

- EAIR is the gold standard for long-term, complex trials. It accounts for time, recurrence, and interruptions.

- Always justify your method. Regulators now expect it.

- Use standardized tools. The PhUSE GitHub repo has free SAS macros that cut programming errors by 83%.

- Train your team. If reviewers don’t understand EAIR, your data won’t get through.

What’s Next?

The FDA’s 2024 draft guidance on exposure-adjusted analysis proposes a single, standardized EAIR formula. Public comments closed in October 2024. By 2027, experts predict 92% of Phase 3 submissions will include EAIR. The FDA’s Sentinel Initiative is even building AI tools to auto-detect safety signals using these methods - with 38% better early detection than old systems.

This isn’t about fancy statistics. It’s about getting the truth right. A drug that causes 10 headaches per 100 patient-years is very different from one that causes 10 headaches in 10 people over 3 months. One is manageable. The other is a red flag.

The numbers matter. But only if you calculate them the right way.