Positional error in spine SBRT

Renee Finnigan, Brock Lamprecht, Tamara Barry, Kimberley Jones, Joshua Boyd, Andrew Pullar, Bryan Burmeister, Matthew Foote

Stereotactic body radiotherapy (SBRT) for spinal tumours delivers high doses per fraction to targets in close proximity to neural tissue. With steep dose gradients, small changes in position can confer significant dosimetric impact on adjacent structures. We analysed positioning error in consecutively treated patients on a strict image-guidance protocol with online correction in 6 degrees of freedom (6-DOF). Set-up error, residual error post-correction and intra-fraction motion for 30 courses of spinal SBRT in 27 patients were assessed using cone-beam CT. Positional error was corrected in x, y and z translational planes and rotational axes using a robotic couch, applying 2 mm and 2° action levels. Linear mixed-effects model assessed whether positional error was influenced by factors such as vertebral level, immobilisation device and treatment duration. Sixty-two fractions were delivered with 225 image registrations. Median treatment duration was significantly longer for patients treated with static-field intensity-modulated radiotherapy compared with volumetric-modulated arc treatment - 40 min versus 28 min, respectively (P = 0.01). Across all fractions, the median residual positional error after initial correction was greatest in the x translational plane (0.5 mm; 95% confidence interval (CI) 0.3-0.6) and y rotational axis (0.25°; 95% CI 0.1-0.3). Median intra-fraction error was also greatest in the x-plane (0.7 mm; 95% CI 0.5-1.0) and y-axis (0.4°; 95% CI 0.2-0.5). With strict immobilisation, image-guidance and 6-DOF correction, our current practice of applying 3-mm planning margins for target volumes and critical structures appears safe. Lower image-guidance action thresholds plus verification with end-to-end testing would be recommended before further reducing margins.