Dosimetric Comparison and Clinical Implications of Dose Calculation Algorithms for Stereotactic Body Radiation Therapy in Spine Metastases

Abstract

Dose calculation algorithms for radiation oncology plans differ in their ability to handle patient specific anatomy, tissue heterogeneity, and artificial implants, which is especially crucial in spine stereotactic body radiation therapy (SBRT) as high ablative dose radiation is directed at a heterogeneous volume, often with metal. In addition, due to the sharp dose gradient required, a slight difference in dose calculation can result in overdosing of the cord or underdosing of the tumor. Better understanding of the different calculation algorithms are needed in order to apply existing historical dose objectives to the modern era. Evaluate the dosimetric differences in three commonly used calculation algorithms—collapsed cone convolution (CCC), anisotropic analytic algorithm (AAA), and Acuros XB (AXB)—for SBRT in spine tumors with and without metal implants. We retrospectively analyzed 27 consecutive patients who have received SBRT for spine metastases, eight of whom had metal hardware within the target. Fourteen plans initially calculated using CCC were recalculated with AAA and AXB while preserving the total monitor units. An additional 13 plans previously calculated with AXB were recalculated using AAA in a similar fashion. Target coverage and the maximum dose for the spinal cord were examined. Doses are expressed relative to AXB as the standard. The maximum dose to the spinal cord was overestimated with AAA by +3.4% [95% CI: -2.8%, 9.6%] and CCC by +0.34% [-7.98%, 8.66%] compared to AXB. Patients with metal hardware in the target volume showed a larger degree of overestimation: AAA by +4.1% with metal vs. +3.0% without; CCC by +1.2% with metal vs. -0.3% underestimation for those without. The maximum dose to the target was overestimated with AAA by +1.2% [-3.4%, 5.9%], whereas CCC underestimated by -4.9% [-11.0%, 1.3%]. However, for patients with metal in the target, AAA overestimated by +0.7% compared to +1.5% for those without. CCC underestimation was much larger with metal, -7.0% compared to -3.2% without metal. Both algorithms overestimated the dose covering 95% of the target volume: AAA by +2.9% [-.8%, 6.6%] and CCC by +3.9% [-1.8%, 9.6%]. The degree of overestimation was larger in patients with metal; AAA: +4.9% overestimation with metal vs. +2.1% without; CCC: +6.3% overestimation with metal vs. +2.2% without. As more accurate dose calculation algorithms (AXB) are implemented clinically, the dosimetric review of target and normal organs need to be re-examined. Historical algorithms overestimate the minimum target coverage compared with AXB, indicating potential under-dosing of those plans generated with CCC and AAA. Additionally, spinal cord tolerances established using historically less accurate algorithms need to be validated in plans calculated with AXB. The differences in the calculated doses are exaggerated in patients with metal implants, warranting additional attention.