The prognostic impact of PD-L1 and CD8 expression in anal cancer patients treated with chemoradiotherapy.

Background: Programmed death-ligand 1 (PD-L1) expression has been shown to be prognostic in many cancer types and used in consideration of checkpoint inhibitor immunotherapy. However, there are very limited and conflicting data on the prognostic impact of PD-L1 in patients with anal squamous cell carcinoma (ASCC). The objectives of this study were to measure the expression of PD-L1 and CD8 in patients with ASCC treated with radical chemoradiotherapy (CRT) and to correlate tumor expression with progression-free survival (PFS) and overall survival (OS). Methods: Ninety-nine patients with ASCC treated with primary CRT at two tertiary care cancer centers between 2000 and 2013, with available pre-treatment tumors, were included. Tissue microarrays (TMAs) from pre-treatment tumor specimens were stained for PD-L1 and CD8. PD-L1 expression in the tumor and stroma was quantified using HALO image analysis software, and results were interpreted using quantitative methods. The density of CD8 cells within the tumor was interpreted by a trained pathologist semi-quantitatively, using a 0-4 scoring system. Kaplan-Meier analysis with log-rank was used to determine the significance in the association of tumor markers with PFS and OS. Cox multivariate analysis was used to explore independent predictors of PFS and OS. Results: Of the 99 patients, 63 (64%) had sufficient tumor samples available for full analysis. CD8 high status was documented in 32 of 63 (50.8%) % of cases. PD-L1 expression was positive in 88.9% of cases. Approximately half the patients had tumor PD-L1 ≥ 5%. Patients with tumor PD-L1 ≥ 5% had better OS vs those with lower expression, HR=0.32 (95% CI 0.11-0.87), p=0.027; 10 years OS: 84% for tumor PD-L1 ≥ 5% vs 49% for PD-L1 < 5%. PD-L1 expression was not associated with PFS. On multivariate analysis, tumor PD-L1 ≥ 5% showed a trend to statistical significance for better OS, HR=0.55 (95% CI 0.12- 1.00), p=0.052. Conclusions: Tumor PD-L1≥5% is associated with OS in patients with ASCC treated with CRT. PD-L1 expression status using this unique cut-point warrants further validation for prognostication in patients with this disease. Future studies are required to determine the benefit of alternative treatment strategies based on PD-L1 status.

Bench to bedside radiosensitizer development strategy for newly diagnosed glioblastoma.

Glioblastoma is the most common primary brain malignancy and carries with it a poor prognosis. New agents are urgently needed, however nearly all Phase III trials of GBM patients of the past 25 years have failed to demonstrate improvement in outcomes. In 2019, the National Cancer Institute Clinical Trials and Translational Research Advisory Committee (CTAC) Glioblastoma Working Group (GBM WG) identified 5 broad areas of research thought to be important in the development of new herapeutics for GBM. Among those was optimizing radioresponse for GBM in situ. One such strategy to increase radiation efficacy is the addition of a radiosensitizer to improve the therapeutic ratio by enhancing tumor sensitivity while ideally having minimal to no effect on normal tissue. Historically the majority of trials using radiosensitizers have been unsuccessful, but they provide important guidance in what is required to develop agents more efficiently. Improved target selection is essential for a drug to provide maximal benefit, and once that target is identified it must be validated through pre-clinical studies. Careful selection of appropriate in vitro and in vivo models to demonstrate increased radiosensitivity and suitable bioavailability are then necessary to prove that a drug warrants advancement to clinical investigation. Once investigational agents are validated pre-clinically, patient trials require consistency both in terms of planning study design as well as reporting efficacy and toxicity in order to assess the potential benefit of the drug. Through this paper we hope to outline strategies for developing effective radiosensitizers against GBM using as models the examples of XPO1 inhibitors and HDAC inhibitors developed from our own lab.

Radiosensitizers in the temozolomide era for newly diagnosed glioblastoma.

Glioblastoma (GBM) is a challenging diagnosis with almost universally poor prognosis. Though the survival advantage of postoperative radiation (RT) is well established, around 90% of patients will fail in the RT field. The high likelihood of local failure suggests the efficacy of RT needs to be improved to improve clinical outcomes. Radiosensitizers are an established method of enhancing RT cell killing through the addition of a pharmaceutical agent. Though the majority of trials using radiosensitizers have historically been unsuccessful, there continues to be interest with a variety of approaches having been employed. Epidermal growth factor receptor inhibitors, histone deacetylase inhibitors, antiangiogenic agents, and a number of other molecularly targeted agents have all been investigated as potential methods of radiosensitization in the temozolomide era. Outcomes have varied both in terms of toxicity and survival, but some agents such as valproic acid and bortezomib have demonstrated promising results. However, reporting of results in phase 2 trials in newly diagnosed GBM have been inconsistent, with no standard in reporting progression-free survival and toxicity. There is a pressing need for investigation of new agents; however, nearly all phase 3 trials of GBM patients of the past 25 years have demonstrated no improvement in outcomes. One proposed explanation for this is the selection of agents lacking sufficient preclinical data and/or based on poorly designed phase 2 trials. Radiosensitization may represent a viable strategy for improving GBM outcomes in newly diagnosed patients, and further investigation using agents with promising phase 2 data is warranted.

Automated glioma grading on conventional MRI images using deep convolutional neural networks.

PURPOSE: Gliomas are the most common primary tumor of the brain and are classified into grades I-IV of the World Health Organization (WHO), based on their invasively histological appearance. Gliomas grading plays an important role to determine the treatment plan and prognosis prediction. In this study we propose two novel methods for automatic, non-invasively distinguishing low-grade (Grades II and III) glioma (LGG) and high-grade (grade IV) glioma (HGG) on conventional MRI images by using deep convolutional neural networks (CNNs). METHODS: All MRI images have been preprocessed first by rigid image registration and intensity inhomogeneity correction. Both proposed methods consist of two steps: (a) three-dimensional (3D) brain tumor segmentation based on a modification of the popular U-Net model; (b) tumor classification on segmented brain tumor. In the first method, the slice with largest area of tumor is determined and the state-of-the-art mask R-CNN model is employed for tumor grading. To improve the performance of the grading model, a two-dimensional (2D) data augmentation has been implemented to increase both the amount and the diversity of the training images. In the second method, denoted as 3DConvNet, a 3D volumetric CNNs is applied directly on bounding image regions of segmented tumor for classification, which can fully leverage the 3D spatial contextual information of volumetric image data. RESULTS: The proposed schemes were evaluated on The Cancer Imaging Archive (TCIA) low grade glioma (LGG) data, and the Multimodal Brain Tumor Image Segmentation (BraTS) Benchmark 2018 training datasets with fivefold cross validation. All data are divided into training, validation, and test sets. Based on biopsy-proven ground truth, the performance metrics of sensitivity, specificity, and accuracy are measured on the test sets. The results are 0.935 (sensitivity), 0.972 (specificity), and 0.963 (accuracy) for the 2D Mask R-CNN based method, and 0.947 (sensitivity), 0.968 (specificity), and 0.971 (accuracy) for the 3DConvNet method, respectively. In regard to efficiency, for 3D brain tumor segmentation, the program takes around ten and a half hours for training with 300 epochs on BraTS 2018 dataset and takes only around 50 s for testing of a typical image with a size of 160 × 216 × 176. For 2D Mask R-CNN based tumor grading, the program takes around 4 h for training with around 60 000 iterations, and around 1 s for testing of a 2D slice image with size of 128 × 128. For 3DConvNet based tumor grading, the program takes around 2 h for training with 10 000 iterations, and 0.25 s for testing of a 3D cropped image with size of 64 × 64 × 64, using a DELL PRECISION Tower T7910, with two NVIDIA Titan Xp GPUs. CONCLUSIONS: Two effective glioma grading methods on conventional MRI images using deep convolutional neural networks have been developed. Our methods are fully automated without manual specification of region-of-interests and selection of slices for model training, which are common in traditional machine learning based brain tumor grading methods. This methodology may play a crucial role in selecting effective treatment options and survival predictions without the need for surgical biopsy.

The prevalence of altered body image in patients with primary brain tumors: an understudied population.

PURPOSE: Body image (BI) is an important issue for cancer patients, as patients with BI concerns are susceptible to depression, anxiety, difficulty coping, and poor quality of life (QoL). While this concern has been documented in patients with other malignancies, no data exists of this QoL issue in patients with primary brain tumors (PBT). METHODS: A cross-sectional survey of 100 PBT patients was conducted on an IRB approved prospective protocol using structured questionnaires. Participants completed the body image scale (BIS), Appearance Scheme Inventory Revised (ASI-R), MD Anderson Symptom Inventory Brain Tumor (MDASI-BT), and Patient-Reported Outcomes Measurement Information System (PROMIS) Depression, Anxiety, and Psychosocial Impact Positive measures. RESULTS: The prevalence of clinically significant body image dissatisfaction (BIS ≥ 10) was 28% (95% CI 19-37%), median BIS score was 5 (range 0-27). The median ASI-R composite score was 2.9 (range 1.5-4.7). BIS was significantly correlated with the ASI-R (r = 0.53, 95% CI 0.37 to 0.65). The mean PROMIS Depression score was 48.4 (SD = 8.9), PROMIS Anxiety score was 49.4 (SD = 9.9), and PROMIS Psychosocial Illness Impact Positive score was 48.9 (SD = 9.7). BIS was significantly correlated with age, and trended with BMI and sex. The PROMIS Psychosocial Illness Impact Positive and PROMIS Anxiety scores were the most strongly related to BIS. CONCLUSIONS: This study, the first to explore altered body image in PBT patients, revealed clinically significant body image dissatisfaction in nearly 1/3 of patients, similar to other malignancies. These findings underscore the potential contribution of disease and treatment-related body image concerns on psychosocial wellbeing in patients with PBT.

A Serum Proteomic Signature Predicting Survival in Patients with Glioblastoma.

PURPOSE: Glioblastoma (GBM) is the most common form of brain tumor and has a uniformly poor prognosis. Development of prognostic biomarkers in easily accessible serum samples have the potential to improve the outcomes of patients with GBM through personalized therapy planning. MATERIAL/METHODS: In this study pre-treatment serum samples from 30 patients newly diagnosed with GBM were evaluated using a 40-protein multiplex ELISA platform. Analysis of potentially relevant gene targets using The Cancer Genome Atlas database was done using the Glioblastoma Bio Discovery Portal (GBM-BioDP). A ten-biomarker subgroup of clinically relevant molecules was selected using a functional grouping analysis of the 40 plex genes with two genes selected from each group on the basis of degree of variance, lack of co-linearity with other biomarkers and clinical interest. A Multivariate Cox proportional hazard approach was used to analyze the relationship between overall survival (OS), gene expression, and resection status as covariates. RESULTS: Thirty of 40 of the MSD molecules mapped to known genes within TCGA and separated the patient cohort into two main clusters centered predominantly around a grouping of classical and proneural versus the mesenchymal subtype as classified by Verhaak. Using the values for the 30 proteins in a prognostic index (PI) demonstrated that patients in the entire cohort with a PI below the median lived longer than those patients with a PI above the median (HR 1.8, p=0.001) even when stratified by both age and MGMT status. This finding was also consistent within each Verhaak subclass and highly significant (range p=0.0001-0.011). Additionally, a subset of ten proteins including, CRP, SAA, VCAM1, VEGF, MDC, TNFA, IL7, IL8, IL10, IL16 were found to have prognostic value within the TCGA database and a positive correlation with overall survival in GBM patients who had received gross tumor resection followed by conventional radiation therapy and temozolomide treatment concurrent with the addition of valproic acid. CONCLUSION: These findings demonstrate that proteomic approaches to the development of prognostic assays for treatment of GBM may hold potential clinical value.