CliniGA CRO

CliniGA CRO Blog

Interpreting Exposure-Adjusted Safety Rates

When groups differ in observation or treatment duration, raw event counts may not represent equal opportunities for events

Representative research photograph: Interpreting Exposure-Adjusted Safety Rates
Photo: The U.S. Food and Drug Administration · License: Public domain · Cropped and resized.
Published
Updated

When groups differ in observation or treatment duration, raw event counts may not represent equal opportunities for events. Person-time rates can add context, but differ from the proportion of people with an event and do not independently resolve every question about safety.

Define the question and assumptions

Define the risk-window origin and end, together with whether first or recurrent events enter the numerator. Total treatment exposure can differ from time at risk for a particular outcome. Discontinuation reasons and changing risk over time affect interpretation. Justify the summary using the clinical safety question and relevant statistical assumptions rather than selecting whichever measure appears most favourable.

Hypothetical study example

In a hypothetical trial, one arm stops treatment earlier. Its event count is lower, but observation time is also shorter. The team evaluates participant proportions, event rates and discontinuation patterns together instead of presenting fewer events as automatic evidence of better safety without explaining unequal opportunity for observation.

Design review checks

  • Verify that event counts and person-time denominators use corresponding risk windows.
  • State how first events and recurrences contribute to each numerator.
  • Assess the consequences of early discontinuation and different follow-up durations.

Record the decision and its limits

Accepted presentation shows numerator, denominator, unit and window. A rate is not automatically an individual probability in another population. Limited observation cannot exclude uncommon risks. Do not hide differences between summaries; explain which experiences they measure and the remaining uncertainty alongside the clinical context of the observed events.

Frequently asked questions

Is a rate the same as the proportion with an event?
No. Time and counting definitions differ.
Do fewer events establish better safety?
Not automatically. Observation time and discontinuation reasons also require assessment.

Sources

  1. EMA — ICH E9 statistical principles and estimand addendum
  2. EMA — ICH E3 clinical study reports