To provide evidence towards a quantitative response assessment framework incorporating MRI-based linear measurements for spinal metastasis that predicts outcome following stereotactic body radiation therapy (SBRT). Adult patients with de novo spinal metastases treated with SBRT between 2008 and 2018 were retrospectively assessed. The metastatic lesions involving the pedicles, articular processes, lamina, transverse process, spinous process and vertebral body at leach level were measured separately using linear measurements on pre- and all post-SBRT MRIs. The outcome was segment-specific progression (SSP) using SPINO guidelines which was dated to the first clinical documentation of progression, or the date of the associated MRI if imaging was the reason for progression. Random forest analysis for variable selection and recursive partitioning analysis for SSP probability prediction were used. Five Hundred Ninety-three spinal levels (323 patients) from 4081 MRIs were evaluated. The appearance of new T1 hypointensity and increase in Bilsky grade had an odds ratio (OR) of 33.5 and 15.5 for SSP, respectively. Compared to baseline, an increase of > 3mm in any lesion dimension, combined with a 1.67-fold increase in area, had an OR of 4.6 for SSP. The sensitivity, specificity, positive predictive value, negative predictive value, balanced accuracy and area under the curve of the training model were 96.7%, 89.6%, 28.6%, 99.8%, 93.2% and 0.905 and of the test model were 91.3%, 89.3%, 27.1% 99.6%, 90.3% and 0.933, respectively. With further refinement and validation in prospective multicentre studies, MRI-based linear measurements can help predict response assessment in SBRT-treated spinal metastases.
Read full abstract