286MO - Radiomics-based prediction of lymphocyte infiltration in IDH-wildtype glioblastoma

Background

In glioblastoma, the benefit of immunotherapy is limited to a small, yet to be defined subgroup. Due to the immunological diversity and as tissue availability remains a challenge, clinically practicable biomarkers are urgently needed.

Methods

Patients treated for newly diagnosed isocitrate dehydrogenase (IDH)-wildtype glioblastoma were included. Tumor-infiltrating lymphocytes (TILs) were analyzed by immunohistochemistry for CD3, CD8 and PD-1 and quantified using tissue phenomics software. Extraction of quantitative radiomics parameters was performed using PyRadiomics on preprocessed preoperative MRI images of (A) contrast-enhancing tumor tissue and (B) T2-hyperintense signal abnormalities as a surrogate for non-enhancing tumor tissue and perifocal edema. TIL parameters were predicted by ElasticNet models, and correlations between predicted and quantified values were used to evaluate prediction accuracy.

Results

In total, 133 patients (39.1% female, median age 62 (range: 20-80) years) were included. O6-methylguanine-methyltransferase (MGMT) promoter methylation status was known in 96 tumors, of whom 42 (43.8%) were methylated. Median percentages of TILs based on total cells were 0.95% (range: 0.08%-5.47%) for CD3+, 0.40% (0.05%-2.78%) for CD8+, and 0.19% (0.02%-0.84%) for PD1+ cells. Median densities were 73.0 cells/mm² (3.1-450.7) for CD3+, 28.0 cells/mm² (5.2-104.4) for CD8+, and 13.2 cells/mm² (1.6-61.4) for PD1+ TILs. In the contrast-enhancing tumor region, PD-1+ TIL densities which were predicted based on radiomics parameters correlated with actual values (r = 0.444, p < 0.001). In contrast, there was no predictive value of radiomic features for CD3+ and CD8+ TIL densities or percentages in contrast-enhancing tumor tissue (r < 0.2, p > 0.05). However, in non-enhancing T2-hyperintense tissue, a weak predictive power for the absolute number of CD3+ (r = 0.315, p = 0.003) and CD8+ (r = 0.274, p = 0.017) TILs was detected.

Conclusions

These provisional results may serve as a proof of concept for further investigations on “immuno-radiomic” parameters as possible immune biomarkers in brain tumors. Validation of these results in an independent cohort is ongoing.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Austrian Federal Ministry for Digital and Economic Affairs; National Foundation for Research, Technology and Development; Christian Doppler Research Association.

Disclosure

M.J. Mair: Financial Interests, Personal, Other: Pierre Fabre. M. Preusser: Financial Interests, Personal, Advisory Board: Bayer, Bristol Myers Squibb, Novartis, Gerson Lehrman, CMC Contrast, Glaxo Smith Kline, Mundipharma, Roche, BMJ Journals, MedMedia, AstraZeneca, AbbVie, Lilly, Medahead, Daiichi Sankyo, Sanofi, Merck Sharp & Dohme, Tocagen, Adastra, Gan & Lee Pharmaceuticals; Financial Interests, Institutional, Research Grant, Clinical Trials and research: Boehringer-Ingelheim, Bristol Myers Squibb, Roche, Daiichi Sankyo, Merck Sharp & Dohme, Novocure, Glaxo Smith Kline, AbbVie; Non-Financial Interests, Leadership Role, Brain Tumor Group Chair: EORTC; Non-Financial Interests, Leadership Role, EANO President: EANO. A.S.S. Berghoff: Financial Interests, Personal, Invited Speaker: Roche, Bristol Myers Squibb, Merck, AstraZeneca; Financial Interests, Personal, Advisory Board: Daiichi Sankyo; Financial Interests, Institutional, Research Grant: Daiichi Sankyo, Roche. All other authors have declared no conflicts of interest.