Summary
A medRxiv preprint analysing 2017–2023 claims data reports that 0.43% of mifepristone uses in its medication-abortion cohort were followed by a serious adverse event within 45 days. The retrospective study classified outcomes using claims-based criteria aligned with FDA seriousness criteria.
A retrospective analysis of national multipayer claims data found that 3,306 mifepristone uses in a medication-abortion cohort were followed by at least one serious adverse event within 45 days. That corresponds to an absolute risk of 0.43% per use, according to a preprint posted on medRxiv on October 5, 2026.
The study examined medical and pharmacy claims from 2017 through 2023. Its results describe safety outcomes identified through claims, rather than effectiveness estimates; the abstract reports no effectiveness results.
How the researchers defined the cohort and serious events
The researchers first identified 869,409 mifepristone uses among 698,780 people, including uses for medication abortion and other indications. They then restricted the analysis to uses with additional evidence of medication abortion and excluded evidence of other indications. This left 775,931 uses among 626,443 people.
The distinction between uses and people matters: the reported 0.43% is a rate per use, not a percentage of unique patients. Researchers classified potential safety events and healthcare use within 45 days of mifepristone use. For an event to count as serious, the claims had to show consequences consistent with FDA seriousness criteria. Because claims data record diagnoses, treatments and healthcare use, the researchers inferred outcomes and their severity from those codes.
Serious events reported within 45 days
In the medication-abortion cohort, the study reported serious bleeding or haemorrhage after 0.24% of uses, serious infection or sepsis after 0.05%, serious ectopic-pregnancy outcomes after 0.05%, and serious hypersensitivity or anaphylaxis after 0.01%. These are the individual categories reported in the abstract; it does not explain whether categories could overlap.
The authors describe serious adverse events overall as rare and say that applying established, clinically grounded criteria markedly reduced the number of events classified as serious. The result therefore depends in part on how researchers define a serious event and what consequences are visible in claims.
This is a preprint based on retrospective claims data. Its 45-day estimate captures serious outcomes identified through the study’s claims-based definitions during that window. The findings provide a large routine-care safety estimate, while the abstract’s results do not quantify medication-abortion effectiveness.