Abstract
INTRODUCTION: Despite great advancements in circulating biomarkers in neuro-oncology, none reliably predict tumor growth and response to dexamethasone (DEX) therapy. Therefore, while most patients receive DEX during their disease course, no objective measures exist to guide DEX dosage and its duration. METHODS: Blood samples from 14 patients were collected before initial radiotherapy and during follow-up. We measured blood BDNF, YKL-40, GFAP, and NSE levels, and paired them with clinical data (KPS, DEX dose) ±7 days, and brain MRI data acquired ±3 days from blood sampling. Edema (EDEM), contrast-enhancement (CE) and necrosis (NEC) were segmented from T2/FLAIR and T1Gd images. Using machine learning, we generated a multivariable model to predict KPS, MRI_CE, MRI_NEC, MRI_EDEM and their combinations. RESULTS: 43 timepoints from 14 patients were used (age ± SD, mean N of samples XX, range, mean follow up ± SD). BDNF, GFAP and NSE (paste these numbers). For KPS prediction, our model (including sex, methylation status, lg2(NSE), GBM diagnosis, DEX dose, and age) had a correlation coefficient of 0.90 (p < 0.0001). MRI_CE+NEC correlated with lg2(GFAP) (r = 0.56, p < 0.0001) and YKL40 (r = 0.50, p = 0.0006); MRI_NEC with lg2(GFAP) (r = 0.53, p = 0.0003) and YKL40 (r = 0.38, p = 0.0123); MRI_EDEM with lg2(GFAP) (r = 0.47, p = 0.0016); and total tumor volume with lg2(GFAP) (r = 0.54, p = 0.0002). CONCLUSIONS: Serum biomarkers, particularly GFAP and YKL40, show promise in predicting tumor volume changes and KPS. However, prospective trials are needed to validate these results in a clinical setting.