Sample size was revised using a log rank test with a two-sided alternative hypothesis. For a significance level of 5% (two-tailed), assuming a hazard ratio of 2.414 calculated, i.e based on an effect size similar to the effect size of oxygenotherapy in a delay of 60 seconds (Rheims et al., 2019), 40 events should be observed to reject the null hypothesis in 80% of cases,. A proportion of 89% of patients being expected to recover SpO2 ≥ 90% 120 seconds after the end of the seizure and considering a proportion of unusauble records for technical reasons of 20%, at least 54 patients should be randomized.
Assuming that about 10% of patients with drug-resistant partial epilepsy who undergo long-term video-EEG monitoring develop at least one partial secondary generalized tonic-clonic seizure, a total of 554 patients will be included in the study.
