To examine and quantify set-up errors in patient positioning in head-and-neck radiotherapy and to investigate the impact of the choice of reference isocentre—on the patient neck or patient skull—on the magnitude of set-up errors.

Set-up position corrections obtained using online kV 2D/2D matching were recorded automatically for every treatment fraction. 3,413 treatment records for 117 patients treated with volumetric modulated arc therapy during 2013 and 2014 on a single treatment machine in our clinic were analysed. In 79 treatment plans the reference isocentre was set to the patient skull, and in 47 to the neck.

Standard deviation of group systematic error in the vertical, longitudinal and lateral direction and the couch rotation were found to be 2·5 mm, 2·1 mm, 1·9 mm and 0·43° (skull) and 2·5 mm, 1·8 mm, 1·7 mm and 0·49° (neck), respectively. Random error of the vertical, longitudinal, lateral and rotational position correction was 1·8 mm, 1·5 mm, 1·6 mm and 0·62° (skull) and 1·9 mm, 1·6 mm, 1·5 mm and 0·60° (neck), respectively. Positional shifts in different directions were found to be uncorrelated.

Neither reference isocentre set-up shows a clear advantage over the other in terms of interfraction set-up error.