We used the median in the discontinued treatment arm from eligible RCTs as the baseline risk and estimated an annualized mortality of 0.7%.

The optimal information size was not met and the wide CI around absolute estimates does not rule out substantial benefit or substantial harm, and therefore we downrated the quality by 2 levels for impression.

We used the median in the discontinued treatment arm from eligible RCTs as the baseline risk and estimated an annualized mortality of 1.4%.

A meta-analysis¹⁸ reported that the cumulative incidence of recurrent VTE was 16% at 2 years for patients with unprovoked VTE, and because the median follow-up of our included studies was 24 months, we calculated the baseline risk to be 8% per patient-year. Assuming that 45% of the VTE events were PE and 55% were DVT,¹⁷ we estimated that, for patients with unprovoked VTE, the risk for PE recurrence was 3.6 per 100 patient-years and DVT recurrence was 4.4 per 100 patient-years.

The CI around absolute estimates does not rule out substantial benefit or substantial harm, and therefore we downrated the quality by 1 level for impression.

Although I² = 85%, the treatment effects were all in favor of indefinite antithrombotic therapy, and we therefore did not downrate the quality for inconsistency.

We used the median in the discontinued treatment arm from eligible RCTs as the baseline risk and estimated the annualized risk for bleeding to be 0.4%.