P04.24 Apparent diffusion coefficient histogram analysis in recurrent glioblastoma predicts overall survival after the first course of bevacizumab and lomustine

摘要

PURPOSE: Early prediction of response to bevacizumab based on imaging parameters is important but difficult. We investigated the value of Apparent Diffusion Coefficient (ADC) histogram analysis to predict overall survival (OS) after the first course of bevacizumab in patients with recurrent glioblastoma. METHODS: Patients were recruited from the multicenter Dutch BELOB trial (n=148), in which patients had been randomized into 3 treatment arms: bevacizumab-only, lomustine-only, or both. Volumes of interest (VOIs) of enhancing and non-enhancing tumor were drawn on resp. T1w post-contrast and T2w-FLAIR 3D images. ADC maps were derived from diffusion weighted images (DWI) scanned with b=0 and b=1000 s/mm² (slice thickness 3mm, in-plane resolution 2mm). Registration of ADC-maps, VOIs and FLAIR images to the T1w post-contrast images was performed with custom scripts in Matlab using Elastix. Unimodal ADC-histogram parameters were extracted from the enhancing and non-enhancing VOIs. In case of multiple lesions in a single patient, VOIs were combined. VOIs <0.200 cm³ were excluded from analysis. Exploratory analysis of histogram parameters showed that change in minimum ADC from baseline to first follow-up (i.e. 6 weeks) was the most promising parameter. Patients were dichotomized using multiple cut-offs of percentage change (decrease) in minimum ADC and differences in OS were calculated from the first follow-up until death from any cause. Analyses were performed for all patients together, for both bevacizumab-treatment arms together, and for all treatment arms separately. RESULTS: Data from 118 patients was considered for analysis, consisting of 108 enhancing and 99 non-enhancing lesions. In enhancing lesions of all treatment arms together, decrease of minimum ADC using a cut-off value between 20 and 30% was associated with significantly better OS. The lowest HR was found with 27.5% (HR= .572; 95% CI .382-.858). When looking at both bevacizumab-treated arms, significantly better OS was associated with a decrease in minimum ADC using a cut-off value between 10 and 30%. The lowest HR was found with 27.5% (HR= .452; .272-.751). These findings appeared to be driven by the bevacizumab plus lomustine treatment arm, with significantly better OS in those with a decrease of minimum ADC using a cut-off value between 5 and 27.5%. The lowest HR was found with 27.5% (HR= .287; 95% CI .137-.603). The bevacizumab-only and lomustine-only treated patients did not show better OS in associated with a decrease of minimum ADC. No significant associations were found for the non-enhancing lesions. CONCLUSION: Enhancing lesions in patients with recurrent glioblastoma treated with bevacizumab and lomustine that show a decrease in minimum ADC of at least 5% have a significantly better OS than those without.