Room: Exhibit Hall | Forum 7
Purpose: To derive spatial, anatomically-localised associations between planned dose and treatment-induced toxicity following pelvic radiotherapy, and to establish consistency across diverse external data sets.
Methods: Radiotherapy planning data was obtained from three prospective prostate radiotherapy trials, RADAR (N=754), RT01 (N=399) and CHHiP (N=259), together with participant follow-up information describing genitourinary (GU) and gastrointestinal (GI) toxicities. CT images were used to deform patient geometries to a common template and the deformations applied to planned dose distributions summed as equivalent dose. Voxel-level associations were assessed against GI and GU toxicities via Cox regression, a multiple-comparisons permutation test and using feature reduction via regularization in a multi-voxel model. Such associations were formed across the three data sets, as well as for the combined data, and the spatial patterns compared.
Results: Multiple spatial regions were found where increasing dose was associated with both increased and decreased rates of toxicity. Consistency between data sets was variable, with the beam arrangements employed within each related clinical trial having a strong influence over derived associations. For GU toxicity, dysuria consistently showed increased incidence with dose to the distal urethra, and haematuria with dose to the posterior bladder for RADAR and combined data, though not for CHHiP and RT01. For GI toxicity, RADAR and RT01 revealed the influence of rectal dose at the level of the prostate for bleeding, whilst for tenesmus, RADAR indicated a strong influence of dose to the peri-prostatic space and CHHiP indicated a strong influence of lateral beams inferior to the prostate.
Conclusion: Spatial (voxel level) dose-response assessment has revealed anatomically-localised associations. Such associations are dependent on the extent of diversity in the underlying patient dosimetric data, and show variable consistency between disparate data sets. Increased availability of planned, and delivered, radiotherapy dose information is needed to facilitate this type of approach.
Funding Support, Disclosures, and Conflict of Interest: Supported by the Australian Government National Health and Medical Research Council (APP1077788).