Impact of immobilisation and image guidance protocol on planning target volume margins for supine craniospinal irradiation.

Background: The setup errors during supine-CSI (sCSI) using single or dual immobilisation (SM, DM) subsets from two institutions were reviewed to determine if DM consistently decreased the required planning target volumes (PTV) margins and to identify the optimal image guidance environments. Materials and methods: Ours and a sister institutional cohort, each with a subset of SM or DM sCSI and daily 3-dimensional online image verification sets, were reviewed for the cranial and spinal regions translational shifts. Using descriptive statistics, scatter plots and independent sample Mann-Whitney test we compared shifts in each direction for two subsets in each cohort deriving PTV margins (Van Herk: VH, Strooms: St recipes) for the cranial and spinal regions. Three image guidance (IG) protocols were simulated for two regions on the combined cohort with SM and DM subsets to identify the most optimal option with the smallest PTV margin. The IG protocols: 3F, 5F and 5FB where the systematic error correction was done using the average error from the first three, five and in the cranium alone (applied to both the cranium and spine, otherwise) for the first five set-ups, respectively. Results: 6968 image sets for 179 patients showed DM could consistently reduce the PTV margin (VH/St) for the cranium from 6/5 to 4/3.5 (31.8/30.8%) and 6/4 to 4/3.5 mm (30.5/16.8%) for primary and validation cohort, respectively. Similarly, for the spine it was 10/8.5 to 6/5.5 (38.6/38.4%) and 9/7.7 to 7/6 (21.6/21.4%), respectively. The "5F-IG" resulted in the smallest margins for both the cranial (3 mm) and spinal region (5 mm) for DM with estimated 95% CTV coverage probability. Conclusion: DM with 5F-IG would significantly reduce the required PTV margins for sCSI.