Limitations
This study was performed at a single academic medical center, which may limit its generalizability. Colchicine was used according to physician preference, so unrecognized confounders may be present even after PSM matching. Data were ascertained by retrospective chart review, which may affect the completeness and accuracy of the data. We could not assess medication side effects or discontinuation rates due to inconsistent reporting of these data. Variable methods were used to detect AF recurrence, which could introduce detection bias for the primary outcome. There was limited statistical power to analyze subgroups, so the differential results in paroxysmal and persistent AF patients should be viewed as hypothesis generating.