AbstractBackgroundVitamin‐K antagonists (VKAs) have widely been replaced by non‐VKA oral anticoagulants (NOACs). This includes Austria, Germany and Switzerland, where as VKA, instead of warfarin, the much longer‐acting phenprocoumon is used, which was not compared to NOACs in clinical trials.MethodsUsing administrative data from a large German health insurance, we included all anticoagulation‐naïve patients with a first prescription of a NOAC or VKA between 2012 and 2020. We analysed overall survival, major adverse cardiac and cerebrovascular events, major thromboembolic events and major bleeding.ResultsOverall, 570,137 patients were included (apixaban: 26.9%, dabigatran: 4.6%, edoxaban: 8.8%, rivaroxaban: 39.1% and VKA: 20.7% of these 99.4% phenprocoumon). In the primary analysis using a 1:1 propensity score matching‐cohort (PSM‐cohort), a significantly higher overall mortality was found for apixaban, edoxaban and rivaroxaban (all p < 0.001) but not for dabigatran (p = 0.13) compared to VKA. In this PSM‐cohort, 5‐year mortality was 22.7% for apixaban versus 12.7% for VKA, 19.5% for edoxaban versus 11.4% for VKA, 16.0% for rivaroxaban versus 12.3% for VKA (all p < 0.001) and 13.0% for dabigatran versus 12.8% for VKA (p = 0.06). The observed effect was confirmed in sensitivity analyses using un‐weighted and three different weighted Fine–Gray regression models on the basis of the entire cohort.ConclusionsIn this large real‐world analysis, apixaban, edoxaban and rivaroxaban, but not dabigatran, were associated with worse survival compared to VKA. These findings, consistent with a few other studies including phenprocoumon, cast profound doubts on the unreflected, general use of NOACs. Randomized trials should assess whether phenprocoumon might actually be superior to NOACs.