Comparison of Proton Versus Photon SBRT for Treatment of Spinal Metastases Using Variable RBE Models

largely due to concerns of increased risk of spinal cord injury given the challenges of end of range relative biological effectiveness (RBE). Although the 1.1 RBE constant for proton beam has been adopted for clinical use, data indicate that proton RBE is variable and dependent on technical-, tissue-, and patient factors. To better understand the safety of proton SBRT for spinal metastasis, this dosimetric analysis compares plans using photon robotic techniques and proton therapy accounting for RBE-weighted dose (D_{RBE}). Materials and Methods: Nine patients with spinal metastasis were selected to be representative of a broad range of complex clinical practice (3 cervical, 3 thoracic, 3 lumbar) that are uniquely challenging to treat with SBRT were identified. Each vertebral level contained a case with paraspinal extension, a reirradiation case, and a case with high-grade epidural disease (Bilsky grade >= 1c) given that such complex cases in current practice often require target volume under-coverage with photon SBRT (PH-SBRT) in order to meet organ at risk (OAR) dose constraints. All selected patients were treated with PH-SBRT using a robotic system to a prescription dose of 30 Gy in 5 fractions despite our institutional preference for further dose escalation, because further dose escalation was not feasible in the original planning process while keeping normal tissues below acceptable dose constraints. To see if superior target coverage could be achieved with proton treatment, comparative intensity modulated proton therapy (IMPT) plans were generated with the same prescription dose as what was clinically delivered using the 1.1 RBE constant. Dose escalated IMPT plans were then generated to 45 Gy(RBE) in 5 fractions. Variable RBE models (Carabe, McNamara, and Wedenberg) were then utilized to generate RBE-weighted dose D_{RBE} distribution for 30 Gy(RBE) and 45 Gy(RBE) plans using the alpha/beta value (which was 3.76 in this study), physical dose, linear energy transfer (LET) value, and dose per fraction parameters. Proton plans used the robust optimization parameters of +/- 3.5% range and 2-mm setup uncertainties. Planning target volume (PTV) coverage and OARs sparing were compared using the Wilcoxon signed-rank test. Results: Planning target volume coverage was significantly improved when comparing PH-SBRT at 30 Gy in 5 fractions (median: 25 Gy) to IMPT at 30 Gy[RBE] in 5 fractions (median: 30.3 Gy[RBE], P = .02) and 45 Gy (RBE) in 5 fractions (median 35.6 Gy[RBE], P = .001). Maximum dose of the spinal cord (cord Dmax) was significantly lower with IMPT at 30 Gy(RBE) (median: 17.6 Gy[RBE], P = .04) and 45 Gy(RBE) (median: 16.1 Gy[RBE], P = .04) compared to conventional plan at 30 Gy (median: 18 Gy). Spinal cord expansion (cord + 2 mm) maximum dose did not change in both photon (median 21.5 Gy) and proton plans (median 22.5, P = .27). Other OARs were better spared with the same and dose-escalated proton plans. No difference was seen in cord Dmax when comparing the PH-SBRT at 30 Gy to D_{RBE} at 30 and 45 Gy(RBE) using Carabe-, McNamara-, or Wedenberg models. However, for spinal cord expansion (cord + 2 mm), there was significant difference between PH-SBRT and D_{RBE} at 30 Gy(RBE) and 45 Gy(RBE) in 5 fractions using Carabe- (median: 25.4 Gy[RBE], P = .002), McNamara- (median: 25.1 Gy[RBE], P = .003), or Wedenberg (median: 24.8 Gy [RBE], P = .0001) models. The average increase in the spinal cord expansion maximum dose using these models compared to the fixed RBE plans was 5.3%. Conclusion: We report the first dosimetric analysis of proton SBRT for spine metastasis using variable RBE dose models. Compared to photon SBRT, IMPT may provide improved target coverage and better spare adjacent OARs, though fixed RBE models can underestimate the maximum dose to adjacent OARs. Future prospective studies are needed to validate these results.