Striking the Balance: Tailored Stereotactic Body Radiation Therapy for Osseous Spinal Metastases and Outcome Prediction.

Abstract

Stereotactic body radiation therapy (SBRT) is increasingly used for spinal metastases. Contouring for SBRT typically follows the International Spine Radiosurgery Consortium (ISRC) guidelines, leading to large planning target volumes (PTV). We compared minimal-volume contouring, which yields a more focused PTV, with conventional contouring and developed predictive models for adverse outcomes. We analyzed all patients treated for spinal metastases at our center over the past decade. After exploratory and correlation analyses, we compared the contouring methods for adverse outcomes and PTV and spinal cord dose distributions. Logit and support vector machine (SVM) models were used for outcome prediction. Among 121 patients (235 lesions), 147 lesions received minimal-volume contouring, 67 followed ISRC recommendations, and 21 were unassigned. The median prescription dose was 20 Gy (15-22 Gy) for single-session, and 24 Gy (19.5-30 Gy) for hypofractionated treatments. The median PTV was 11.1 cm³ (0.2-173.7 cm³). Complications correlated significantly with pre-SBRT Karnofsky status (Cramer V = 0.277). Local control rates were comparable (80.5% vs 78.4%), but complication rates were lower with minimal contouring (15.1% vs 27.8%). Minimal contouring resulted in significantly higher median PTV biologically effective dose (BED) (75.9 Gy vs 57.5 Gy) and lower maximum spinal cord BED (72.7 Gy vs 81.9 Gy, both P < .001). Logit models identified mean PTV (P = .001) and vertebral dose (P < .001) as significant predictors of vertebral fractures but showed limited accuracy (49.3% for local recurrence, 73.3% for fractures). SVM classifiers outperformed logit models, achieving higher accuracy (89.9% for recurrence, 92.0% for fractures) and improved positive predictive values (83.7% and 86.7%, respectively). Minimal-volume target delineation based on thin-layer magnetic resonance imaging and prostate-specific membrane antigen/fluorodeoxyglucose-positron emission tomography-computed tomography/magnetic resonance imaging may be as effective as contouring according to the ISRC guidelines, if those imaging modalities are available. Further studies are warranted to assess minimal and expanded volumes. SVM models show promising potential for predicting patient outcomes, warranting further exploration.