A New Predictive Model for Radiation Necrosis Risk Based on PTV-Enriched Blood Inflammatory Biomarkers in Patients With Brain Metastases Treated With Stereotactic Radiosurgery.

PURPOSE: Radiation necrosis (RN) is an adverse event following stereotactic radiosurgery (SRS) for brain metastases (BMs). The planning target volume (PTV) size is a potential yet suboptimal predictor of RN, unlike V12, which remains independent of the number of fractions delivered. Prior authors demonstrated that radiation-induced brain injury can be traced in peripheral blood. This study aimed to identify predictive biomarkers of symptomatic RN at the time of SRS to develop a risk prediction model based on inflammatory proteomic plasma markers and the PTV as dosimetric surrogate.

METHODS AND MATERIALS: A retrospective study design was conducted using plasma samples from patients with BMs treated with SRS (single fraction) or fractionated stereotactic radiation therapy (3-6 fractions). The Olink 96 Target Immuno-Oncology Panel was used to analyze 92 related human proteins using oligonucleotide-bound antibodies and real-time polymerase chain reaction. Statistical analyses included receiver operating characteristic curves and multivariable Cox proportional hazards regression to evaluate clinical parameters, PTV, and blood biomarkers to predict RN risk. Multiplex immunophenotyping was performed on available RN tissue samples to assess immune infiltration.

RESULTS

A total of 47 patients with BMs were analyzed (21 cases with RN and 26 without). Cox regression analysis identified PTV and the inflammation-related blood biomarkers MUC-16 and CXCL11 as independent predictors of RN. A high Necrosis Predictive Index combining PTV with MUC-16 and CXCL11 was significantly associated with a higher risk (hazard ratio = 2.543 [1.615-4.005]; P < .0001) of RN development. Receiver operating characteristic analysis demonstrated that the Necrosis Predictive Index effectively distinguished patients with RN, achieving an area under the curve of 0.808. An exploratory independent analysis found that baseline levels of other proinflammatory blood biomarkers (ie, CD8a and IL-8) were also associated with RN. Tissue analysis confirmed a proinflammatory microenvironment in RN compared with tumor recurrence.

CONCLUSIONS

This hypothesis-generating study suggests the combination of PTV, CXCL11, and MUC-16 may predict the development of RN, warranting further validation.