The duration of untreated ventricular fibrillation (VF) is of paramount importance for CPR success. Moreover, therapeutic interventions taking into account the interval between cardiac arrest onset and initiation of CPR improve outcome. This study was performed to investigate whether VF feature analysis could be used to estimate the duration of VF in patients with out-of-hospital cardiac arrest. Demographic data recorded according to the Utstein guidelines and ECG recordings of 376 cardiac arrest patients from three European areas were analysed. Ten features in the time and frequency domain derived from different sub-bands of the initial VF ECG (n=127) were evaluated. The correlation between VF ECG features and cardiac arrest times was investigated using Pearson's correlation coefficient in a subset of 40 patients with reliably estimated downtimes and artefact-free initial VF tracings. No significant correlation (p<.05) between any of the VF ECG features and downtime could be found. The duration of cardiac arrest could not be estimated reliably from human VF ECG single feature analysis.