Isocentre shifts of approximately 1 to 4 mm in this patient cohort representing set-up accuracy are within the expected range of extracranial targets in the head and neck [4, 9, 11–16, 18, 19, 24]. Higher precision for the Scotchcast or thermoplastic systems have been reported before [15, 17, 24], however it needs to be mentioned that mostly intracranial targets were evaluated in systems allowing position correction in 3DOF. Treatment of extracranial targets with these immobilization systems has been investigated resulting in less accurate positioning of more distal as compared to intracranial targets [18, 19]. This is supported by the clinicians' experience in everyday clinical routine in the conventional radiotherapy.
Reproducibility of fixation devices can be analyzed by evaluation of standard deviations of respective set-up corrections. In our cohort, this was evaluated by the root mean square of all patients' standard deviations or the centered distributions as described above.
In view of the higher rigidity of Scotchcast masks as opposed to thermoplastic head masks, higher reproducibility of the Scotchcast immobilization would initially be expected. This is supported by our data for 3 and 6 DOF except for the vertical component. The Scotchcast mask's rigid shell does not seem to allow significant motion in both the vertical and lateral direction but does allow some motion in the longitudinal direction. Thermoplastic head masks on the other hand immobilize the patient between headrest and thermoplastic layer with very little motion in the vertical direction. Less restriction apparently occurs in the lateral and longitudinal direction.
Scotchcast and thermoplastic (including shoulder fixation) masks were shown to immobilize head and neck cancer patients equally well if considering 3 DOF position correction only. Higher discrepancies were found when comparing these systems in 6 DOF. While these statistically significant differences could not be attributed to the patients' age distribution in the two immobilization groups, overall differences (Scotchcast and thermoplastic immobilization) were higher in the 3 DOF position correction versus 6 DOF which is supported by Spadea et al .
This difference was maintained presuming our traditional action level of 3 mm in fractionated head and neck treatments. Albeit isocentre localization was similar in 3 DOF and 6 DOF, target volumes usually extended more caudally in the 3 DOF (IMRT) as compared to 6 DOF (C12). Therapists had to consider adequate target position over a higher volume therefore making the best possible compromise for positioning while only the more cranial part (CTV1) of the CTV2 had to be considered in carbon ion therapy.
We are aware different imaging modalities were used for position verification in 3 DOF (MV-CBCT) as opposed to 6 DOF (orthogonal x-rays). However, various investigations have already been carried out addressing the issue of imaging modality for position verification suggesting orthogonal x-rays to be equivalent to CBCT for the determination of setup errors [2, 15].
Also, we have analyzed significantly higher numbers of position checks in 3 DOF than in 6 DOF. This however, is due to the nature of our treatment regimen applying mostly 8 fractions of carbon ion therapy followed by approximately 25 fractions of IMRT for reported indications in head and neck malignancies.
The Scotchcast mask was shown to require lower absolute interfractional set-up corrections; hence, this fixation system appears superior for lesions in the vicinity of small critical structures such as optic nerves or the optic chiasm where the highest possible reproducibility is required.
In a rigid body setup such as our head and neck patients, optimal translational corrections were found to be dependent on whether or not rotations were included in the registration and position correction . In standard treatments, where treatment tables commonly only allow corrections in 3 DOF without rotation correction capability, optimal corrections for translational shifts are dependent on registration landmarks. Therefore, it is recommended in rigid registrations to choose landmarks approximately coincident with the treatment site. Hence, when our therapists need to match the more extensive target volumes for IMRT following carbon ion treatment, compromises need to be made at the cranial/caudal edge of the target. Our findings practically illustrate these theoretical considerations of Murphy .