ideasIdea
Post-marketing safety monitoring stratified by ancestry and sex, with label updates
Once a drug is in wide use, real-world records could be checked routinely for whether side effects differ by ancestry or sex, since trials were too small in those groups to notice. Findings would go into the label.
Regulators run active pharmacovigilance on linked EHR, claims and registry data for new oncology drugs, pre-specifying analyses of serious adverse events and discontinuation by self-reported race, genetic ancestry where available, and sex, with a defined signal threshold that triggers label review. Sentinel-type systems exist in the US and EU but rarely stratify this way.
Hypothesis
Stratified surveillance will detect subgroup-specific toxicity signals for several agents within three years of approval that were not visible in pivotal trials, leading to label changes.
Rationale
Under-enrolment means subgroup safety is unknown at approval; real-world use quickly generates the numbers, but only if the analysis is planned.
What would test it
Apply stratified surveillance to the ten most-used oncology drugs approved in the last five years and count actionable signals.
Maturity
speculative
Who has to act
regulator
Cost to try
Medium ($1M to $50M)
Years to first evidence
3
Bottlenecks it attacks
- Trials do not represent the people who get cancer · Older, Black, Hispanic, Asian, rural, poor and multimorbid patients are under-represented, so results may not apply to them.
- Weak real-world evidence and registries · We do not reliably know what happens to patients after approval, so we cannot tell which drugs deliver in practice.