Risk Groups For Survival Research Articles

Systemic immune-related spleen radiomics predict progression-free survival in patients with locally advanced cervical cancer underwent definitive chemoradiotherapy

PurposeSystemic immunity is essential for driving therapeutically induced antitumor immune responses, and the spleen may reflect alterations in systemic immunity. This study aimed to evaluate the predictive value of contrast-enhanced CT-based spleen radiomics for progression-free survival (PFS) in patients with locally advanced cervical cancer (LACC) who underwent definitive chemoradiotherapy (dCRT). Additionally, we investigated the role of spleen radiomics features and changes in spleen volume in assessing systemic immunity.MethodsThis retrospective study included 257 patients with LACC who underwent dCRT. The patients were randomly divided into training and validation groups in a 7:3 ratio. Radiomic features were extracted from CT images obtained before and after dCRT. Radiomic scores (Radscore) were calculated using features selected through least absolute shrinkage and selection operator (LASSO) Cox regression. The percentage change in spleen volume was determined from measurements taken before and after treatment. Independent prognostic factors for PFS were identified through multivariate Cox regression analyses. Model performance was evaluated with the receiver operating characteristic (ROC) curve and the C-index. The Radscore cut-off value, determined from the ROC curve, was used to stratify patients into high- and low-risk survival groups. The Wilcoxon test was used to analyze differences in hematological parameters between different survival risk groups and between different spleen volume change groups. Spearman correlation analysis was used to explore the relationship between spleen volume change and hematological parameters.ResultsIndependent prognostic factors included FIGO stage, pre-treatment neutrophil-to-lymphocyte ratio (pre-NLR), spleen volume change, and Radscore. The radiomics-combined model demonstrated the best predictive performance for PFS in both the training group (AUC: 0.923, C-index: 0.884) and the validation group (AUC: 0.895, C-index: 0.834). Compared to the low-risk group, the high-risk group had higher pre-NLR (p = 0.0054) and post-NLR (p = 0.038). Additionally, compared to the decreased spleen volume group, the increased spleen volume group had lower post-NLR (p = 0.0059) and post-treatment platelet-to-lymphocyte ratio (p < 0.001).ConclusionSpleen radiomics combined with clinical features can effectively predict PFS in patients with LACC after dCRT. Furthermore, spleen radiomics features and changes in spleen volume can reflect alterations in systemic immunity.

Significance of novel PANoptosis genes to predict prognosis and therapy effect in the lung adenocarcinoma

Lung adenocarcinoma (LUAD) is the dominant histotype of non-small cell lung cancer. Panoptosis, a comprehensive form of programmed cell death, is central to carcinogenesis. In this study, the expression of PANoptosis-related genes (PRGs) and their impact on the development, prognosis, tumor microenvironment, and treatment response of patients with lung adenocarcinoma (LUAD) were systematically evaluated. PRGs were selected from The Cancer Genome Atlas database and Genecards dataset using differential expression analysis. The signature of included PRGs was identified using univariate Cox regression analysis and LASSO regression analysis. Additionally, a nomogram was developed that includes signature and clinical information. Kaplan–Meier survival analysis and receiver operating characteristic curves were used to assess the predictive validity of these risk models. Finally, functional analysis of the selected PRGs in signature and analysis of immune landscape were also performed. Preliminary identification of 10 genes related to PANoptosis has significant implications for prognosis. Subsequently, seven related genes were integrated to classify LUAD patients into different survival risk groups. The prognostic risk score generated from the signature and the TNM stage were as independent prognostic factors and were utilized in creating a nomogram plot. Both the features and the nomogram plot showed accurate performance in predicting the overall survival of LUAD patients. The PRGs were enriched in several biological functions and pathways, and stratified studies were conducted on the differences in immune landscape between high-risk and low-risk groups based on their characteristics. Ultimately, our evaluation focused on the differences in drug treatment efficacy between the high-risk and low-risk groups, providing a foundation for future research directions. Potential associations between PRGs and patient prognosis in LUAD have been identified in this study. Potential biomarkers for clinical application could be considered for the prognostic predictors identified in this study.

A Prognostic Model Based on Nutritional Indexes for Patients With Pan-Cancer: A Real-World Cohort Study.

The aim was to identify the nutritional indexes, construct a prognostic model, and develop a nomogram for predicting individual survival probability in pan-cancers. Nutritional indicators, clinicopathological characteristics, and previous major treatment details of the patients were collected. The enrolled patients were randomly divided into training and validation cohorts. Least absolute shrinkage and selection operator (Lasso) regression cross-validation was used to determine the variables to include in the cox regression model. The training cohort was used to build the prediction model, and the validation cohort was used to further verify the discrimination, calibration, and clinical effectiveness of the model. A total of 2020 patients were included. The median OS was 56.50 months (95% CI, 50.36-62.65 months). In the training cohort of 1425 patients, through Lasso regression cross-validation, 13 characteristics were included in the model. Cox proportional hazards model was developed and visualized as a nomogram. The C-indexes of the model for predicting 1-, 3-, 5-, and 10-year OS were 0.848, 0.826, 0.814, and 0.799 in the training cohort and 0.851, 0.819, 0.814, and 0.801 in the validation cohort. The model showed great calibration in the two cohorts. Patients with a score of less than 274.29 had a better prognosis (training cohort: HR, 6.932; 95% CI, 5.723-8.397; log-rank p < 0.001; validation cohort: HR, 8.429; 95% CI, 6.180-11.497; log-rank p < 0.001). The prognostic model based on the nutritional indexes of pan-cancer can divide patients into different survival risk groups and performed well in the validation cohort.

Abstract 866: Identification of survival-associated epithelial cell markers linked to TROP2 through single-cell and bulk sequencing analysis

Abstract Introduction: Colorectal Cancer (CRC) is known for its aggressiveness and drug resistance. This has increased the interest in antibody-drug conjugates (ADCs) as a possible therapy. In this study, we aim to investigate the effects of the ADC target TROP2, on the level of epithelial cells and its correlation with survival outcomes in CRC patients. Methods: Single-cell RNA (scRNA) data was downloaded from Gene Expression Omnibus (GEO) for CRC patients and evaluated for TROP2 expression. Patients were grouped into low and high-TROP2 based on median TROP2 expression, and pseudo-bulk analysis was used to find the differentially expressed genes (DEGs). The DEGs were used to find the survival association in the TCGA-CRC cohort. A risk score was built using principal component analysis (PCA) on the TROP2-DEGs and cutoff point was chosen based on survival. Kaplan-Meier curve was used to compare overall survival (OS), progression-free survival (PFS), and disease-free survival (DFS) risk groups. Gene Set Variation Analysis (GSVA) was used to identify the enriched pathways between low- and high-TROP2 in GSE using pseudo-bulk by pooling epithelial cells of individual patients. Results: Of the 10424 cells of 15 different types for 9 CRC patients retrieved from GSE146771, TROP2 was significantly expressed by epithelial cells (n = 766) with mean expression of 1.64 (SD: 1.96, P-value: 0.003) using pseudo-bulk analysis. Three significant DEGs were obtained which are: KIF13B, UCHL3, and PLAC8. These were evaluated for survival association in resectable stages (I-III) CRC patients (n = 419) using PCA to group the patients into low-risk (n = 88) and high-risk (n = 331), using -1.4 as the cutoff point. High risk group showed a better survival with a 1-year survival probability of 94% compared to 90% in the low group and a 5-year survival probability of 74% vs. 46% in the low group. The Kaplan-Meier curve showed that OS, PFS &amp; DFS were significantly better in the low-risk group (p-value = 0.004, 0, 04, and 0.024 respectively). TROP2 and PLAC8 were significantly associated with worse prognosis, whereas UCHL3 was associated with better prognosis (p-value = &amp;lt;0.001), however, KIF3B did not significantly affect the prognosis. APC gene mutation and TP53 were higher in the low-risk group (p-value = 0.004, 0.007 respectively). KRAS-signaling, IL-18-signaling, MET activation of PTK2-signaling were upregulated in high-TROP2 group, while MYC targets, MTORC1 signaling, DNA repair, reactive-oxygen-species signaling, and TCR-signaling were upregulated in low-TROP2. Conclusion: In this study we developed a TROP2-associated risk score that significantly predicted OS, PFS, and DFS in patients with resectable CRC by combining single-cell and bulk-RNA-seq datasets and showing predominant expression of TROP2 in epithelial cells. These findings support the development of TROP2 targeting therapeutics in this space. Citation Format: Noor Nedal Al-Bzour, Ayah Nedal Al-Bzour, Azhar Saeed, Lujia Chen, Anwaar Saeed. Identification of survival-associated epithelial cell markers linked to TROP2 through single-cell and bulk sequencing analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 866.

The Role of Geriatric Assessment in the Management of Diffuse Large B-Cell Lymphoma.

The treatment choice for an older patient with diffuse large B-cell lymphoma (DLBCL) depends on many other factors in addition to age, which alone does not reflect the complexity of the aging process. Functional features and comorbidity incidence differ not only between younger and older patients but also among older patients themselves. The comprehensive geriatric assessment (CGA) quickly evaluates fitness status by investigating the patient's different functional areas, degree of autonomy, and presence of comorbidities. Various tools are available to evaluate frailty; which assessment tool to use should be based on the clinical aim. The simplified geriatric assessment (sGA) from the elderly project by the Fondazione Italiana Linfomi, prospectively tested on the largest number of patients, categorizes patients as fit, unfit, or frail, with a decreasing rate of overall survival. The elderly prognostic index (EPI), which combines sGA and IPI scores and hemoglobin level, is the first prognostic score for older patients, with three risk groups for survival. Future GAs should consider new parameters, including sarcopenia, which appears to be inversely related to survival. New tools based on prospective studies can help physicians choose the best treatment in light of the individual patient's characteristics.

P13.06.A RADIOMIC FEATURES FOR RISK-STRATIFICATION IN PATIENTS WITH BRAIN METASTASES

Abstract BACKGROUND Recently, non-invasive characterization of brain tumors on MRI has emerged as a promising field of research. The identification of quantitative imaging biomarkers, also known as radiomics, may complement molecular characterization and thereby improve clinical management of neuro-oncologic patients. In this study we aimed to identify imaging predictors with improved performance over clinical parameters in order to stratify patients with brain metastases into high and low risk groups for overall survival (OS). MATERIAL AND METHODS 422 patients (allocated in a 3:1 ratio to a discovery [n=317] and test [n= 105] set) with first diagnosis of brain metastases from different primary tumors from two neuro-oncologic centers were included. In each patient, eight clinical features (age, gender, KPS, systemic disease status, presence of extracranial metastases, number of cerebral metastases, primary tumor and available individual prognostic molecular status eg HER2, BRAF,⋯) were gathered and a total of 321 radiomic MRI features (including shape, first-order and higher-order features) from cerebral MRI (contrast T1-weighted and apparent diffusion coefficient maps) were extracted. Radiomic and clinical features of patients in the discovery set were subjected to different machine learning models in order to classify patients into low- and high-risk groups for OS. By performing a comprehensive grid search the best machine learning model according to the macro-evaluated score and accuracy was identified and evaluated on the testing set. In addition, a subgroup analysis, based on the primary tumor entity was done and confusion matrices were calculated in order to evaluate final predictions. RESULTS With an extra trees classifier including all clinical and 30 radiomic features, we were able to stratify patients into a high- and low-risk group for OS (test set: macro-evaluated = 0.60, accuracy = 0.67). The best performing model was a gradient boosting model only including clinical features (test set: macro-evaluated = 0.62, accuracy = 0.72), while the radiomic features alone led to the poorest results (test set: macro-evaluated = 0.52, accuracy = 0.63). Interestingly, in the subgroup of melanoma patients, the predictive power of radiomic features outperformed the clinical and the combined (radiomics and clinical) features. With a prediction solely based on radiomic features, 80% and 67% of patients were correctly classified into the low- and the high-risk group for OS, respectively. CONCLUSION In conclusion, we found that in the entire study population of patients with brain metastases radiomic features did not allow for a better prediction of the clinical outcome compared to the clinical parameters alone. However, in the subgroup of melanoma patients the predictive power of radiomic features alone was superior compared to clinical features or all features combined.

HLA amino acid Mismatch-Based risk stratification of kidney allograft failure using a novel Machine learning algorithm

ObjectiveWhile associations between HLA antigen-level mismatches (Ag-MM) and kidney allograft failure are well established, HLA amino acid-level mismatches (AA-MM) have been less explored. Ag-MM fails to consider the substantial variability in the number of MMs at polymorphic amino acid (AA) sites within any given Ag-MM category, which may conceal variable impact on allorecognition. In this study we aim to develop a novel Feature Inclusion Bin Evolver for Risk Stratification (FIBERS) and apply it to automatically discover bins of HLA amino acid mismatches that stratify donor-recipient pairs into low versus high graft survival risk groups. MethodsUsing data from the Scientific Registry of Transplant Recipients, we applied FIBERS on a multiethnic population of 166,574 kidney transplants between 2000 and 2017. FIBERS was applied (1) across all HLA-A, B, C, DRB1, and DQB1 locus AA-MMs with comparison to 0-ABDR Ag-MM risk stratification, (2) on AA-MMs within each HLA locus individually, and (3) using cross validation to evaluate FIBERS generalizability. The predictive power of graft failure risk stratification was evaluated while adjusting for donor/recipient characteristics and HLA-A, B, C, DRB1, and DQB1 Ag-MMs as covariates. ResultsFIBERS’s best-performing bin (on AA-MMs across all loci) added significant predictive power (hazard ratio = 1.10, Bonferroni adj. p < 0.001) in stratifying graft failure risk (where low-risk is defined as zero AA-MMs and high-risk is one or more AA-MMs) even after adjusting for Ag-MMs and donor/recipient covariates. The best bin also categorized more than twice as many patients to the low-risk category, compared to traditional 0-ABDR Ag mismatching (∼24.4% vs ∼ 9.1%). When HLA loci were binned individually, the bin for DRB1 exhibited the strongest risk stratification; relative to zero AA-MM, one or more MMs in the bin yielded HR = 1.11, p < 0.005 in a fully adjusted Cox model. AA-MMs at HLA-DRB1 peptide contact sites contributed most to incremental risk of graft failure. Additionally, FIBERS points to possible risk associated with HLA-DQB1 AA-MMs at positions that determine specificity of peptide anchor residues and HLA-DQ heterodimer stability. ConclusionFIBERS’s performance suggests potential for discovery of HLA immunogenetics-based risk stratification of kidney graft failure that outperforms traditional assessment.

A novel nomogram integrated with PDL1 and CEA to predict the prognosis of patients with gastric cancer.

This study aimed to develop a prognostic nomogram for patients with gastric cancer (GC) based on the levels of programmed death 1 ligand 1 (PDL1) and carcinoembryonic antigen (CEA). The nomogram was developed using data from a primary cohort of 247 patients who had been clinicopathologically diagnosed with GC, as well as a validation cohort of 63 patients. Furthermore, the nomogram divided the patients into three different risk groups for overall survival (OS)-the low-risk, middle-risk, and high-risk groups. Univariate and multivariate Cox hazard analyses were used to determine all of the factors included in the model. Decision curve analysis and receiver operating characteristic (ROC) curves were used to assess the accuracy of the nomogram. The Kaplan-Meier survival analysis revealed that metastasis stage, clinical stage, and CEA and PDL1 levels were predictors for progress-free survival (PFS) and OS of patients with GC. Metastasis stage, clinical stage, and CEA and PDL1 levels were found to be independent risk factors for the PFS and OS of patients with GC in a multivariate analysis, and the nomogram was based on these factors. The concordance index of the nomogram was 0.763 [95% confidence interval (CI) 0.740-0.787]. The area under the concentration-time curve of the nomogram model was 0.81 (95% CI 0.780-0.900). According to the decision curve analysis and ROC curves, the nomogram model had a higher overall net efficiency in forecasting OS than clinical stage, CEA and PDL1 levels. In conclusion, we proposed a novel nomogram that integrated PDL1 and CEA, and the proposed nomogram provided more accurate and useful prognostic predictions for patients with GC.

Predictive power of deep-learning segmentation based prognostication model in non-small cell lung cancer.

The study aims to create a model to predict survival outcomes for non-small cell lung cancer (NSCLC) after treatment with stereotactic body radiotherapy (SBRT) using deep-learning segmentation based prognostication (DESEP). The DESEP model was trained using imaging from 108 patients with NSCLC with various clinical stages and treatment histories. The model generated predictions based on unsupervised features learned by a deep-segmentation network from computed tomography imaging to categorize patients into high and low risk groups for overall survival (DESEP-predicted-OS), disease specific survival (DESEP-predicted-DSS), and local progression free survival (DESEP-predicted-LPFS). Serial assessments were also performed using auto-segmentation based volumetric RECISTv1.1 and computer-based unidimensional RECISTv1.1 patients was performed. There was a concordance between the DESEP-predicted-LPFS risk category and manually calculated RECISTv1.1 (φ=0.544, p=0.001). Neither the auto-segmentation based volumetric RECISTv1.1 nor the computer-based unidimensional RECISTv1.1 correlated with manual RECISTv1.1 (p=0.081 and p=0.144, respectively). While manual RECISTv1.1 correlated with LPFS (HR=6.97,3.51-13.85, c=0.70, p<0.001), it could not provide insight regarding DSS (p=0.942) or OS (p=0.662). In contrast, the DESEP-predicted methods were predictive of LPFS (HR=3.58, 1.66-7.18, c=0.60, p<0.001), OS (HR=6.31, 3.65-10.93, c=0.71, p<0.001) and DSS (HR=9.25, 4.50-19.02, c=0.69, p<0.001). The promising results of the DESEP model were reproduced for the independent, external datasets of Stanford University, classifying survival and 'dead' group in their Kaplan-Meyer curves (p = 0.019). Deep-learning segmentation based prognostication can predict LPFS as well as OS, and DSS after SBRT for NSCLC. It can be used in conjunction with current standard of care, manual RECISTv1.1 to provide additional insights regarding DSS and OS in NSCLC patients receiving SBRT. While current standard of care, manual RECISTv1.1 correlated with local progression free survival (LPFS) (HR=6.97,3.51-13.85, c=0.70, p<0.001), it could not provide insight regarding disease specific survival (DSS) (p=0.942) or overall survival (OS) (p=0.662). In contrast, the deep-learning segmentation based prognostication (DESEP)-predicted methods were predictive of LPFS (HR=3.58, 1.66-7.18, c=0.60, p<0.001), OS (HR=6.31, 3.65-10.93, c=0.71, p<0.001) and DSS (HR=9.25, 4.50-19.02, c=0.69, p<0.001). DESEP can be used in conjunction with current standard of care, manual RECISTv1.1 to provide additional insights regarding DSS and OS in NSCLC patients.

Identification of gene profiles related to the development of oral cancer using a deep learning technique

BackgroundOral cancer (OC) is a debilitating disease that can affect the quality of life of these patients adversely. Oral premalignant lesion patients have a high risk of developing OC. Therefore, identifying robust survival subgroups among them may significantly improve patient therapy and care. This study aimed to identify prognostic biomarkers that predict the time-to-development of OC and survival stratification for patients using state-of-the-art machine learning and deep learning.MethodsGene expression profiles (29,096 probes) related to 86 patients from the GSE26549 dataset from the GEO repository were used. An autoencoder deep learning neural network model was used to extract features. We also used a univariate Cox regression model to select significant features obtained from the deep learning method (P < 0.05). High-risk and low-risk groups were then identified using a hierarchical clustering technique based on 100 encoded features (the number of units of the encoding layer, i.e., bottleneck of the network) from autoencoder and selected by Cox proportional hazards model and a supervised random forest (RF) classifier was used to identify gene profiles related to subtypes of OC from the original 29,096 probes.ResultsAmong 100 encoded features extracted by autoencoder, seventy features were significantly related to time-to-OC-development, based on the univariate Cox model, which was used as the inputs for the clustering of patients. Two survival risk groups were identified (P value of log-rank test = 0.003) and were used as the labels for supervised classification. The overall accuracy of the RF classifier was 0.916 over the test set, yielded 21 top genes (FUT8-DDR2-ATM-CD247-ETS1-ZEB2-COL5A2-GMAP7-CDH1-COL11A2-COL3A1-AHR-COL2A1-CHORDC1-PTP4A3-COL1A2-CCR2-PDGFRB-COL1A1-FERMT2-PIK3CB) associated with time to developing OC, selected among the original 29,096 probes.ConclusionsUsing deep learning, our study identified prominent transcriptional biomarkers in determining high-risk patients for developing oral cancer, which may be prognostic as significant targets for OC therapy. The identified genes may serve as potential targets for oral cancer chemoprevention. Additional validation of these biomarkers in experimental prospective and retrospective studies will launch them in OC clinics.

Identification of a novel anoikis‐related gene signature to predict prognosis and tumor microenvironment in lung adenocarcinoma

Lung adenocarcinoma (LUAD) is the most prevalent histotype of non-small cell lung cancer. Anoikis, an alternative form of programmed cell death, plays a pivotal role in cancer invasion and metastasis, preventing the detached cancer cells from readhering to other substrates for abnormal proliferation. The aim of this study was to conduct a comprehensive analyses of the prognostic implications of anoikis-related genes (ARGs) in LUAD. ARGs were selected from The Cancer Genome Atlas (TCGA) database and Genecards dataset using differential expression analysis. The signature incorporating ARGs was identified using univariate Cox regression analysis and LASSO regression analysis. Furthermore, a nomogram containing the signature and clinical information was developed through univariate and multivariate Cox regression analysis. Kaplan-Meier survival analysis and receiver operating characteristic (ROC) curves were applied to evaluate the predictive validity of these risk models. Finally, functional analysis of the selected ARGs in signature and analysis of immune landscape were also conducted. A 16-gene signature was integrated to stratify LUAD patients into different survival risk groups. The prognostic risk score generated from the signature and TNM stage were identified as independent prognostic factors and utilized to develop a nomogram. Both the signature and the nomogram showed satisfactory prediction performance in predicting overall survival (OS) of LUAD patients. The ARGs were enriched in several biological functions and signaling pathways. Finally, differences of immune landscape were investigated among the high- and low-risk groups stratified by the signature. This study revealed potential relationships between ARGs and prognosis of LUAD. The prognostic predictors identified in present study could be utilized as potential biomarkers for clinical applications.

An interpretable machine learning prognostic system for risk stratification in oropharyngeal cancer

BackgroundThe optimal management of oropharyngeal squamous cell carcinoma (OPSCC) includes both surgical and non-surgical, that is, (chemo)radiotherapy treatment options and their combinations. These approaches carry a risk of specific treatment-related side effects. HPV-positive OPSCC has been reported to be more sensitive to (chemo)radiotherapy-based treatment modalities. Objectives: This study aims to demonstrate how machine learning can aid in classifying OPSCC patients into risk groups (low-chance or high-chance) for overall survival. We examined the input variables using permutation feature importance. Furthermore, we provided explanations and interpretations using the Local Interpretable Model Agnostic Explanations (LIME) and SHapley Additive Explanation (SHAP) frameworks. Methods: The machine learning model for 3164 OPSCC patients was built using data obtained from the Surveillance, Epidemiology, and End Results (SEER) program database. A total of five variants of tree-based machine learning algorithms (voting ensemble, light GBM, XGBoost, Random Forest, and Extreme Random Trees) were used to divide the patients into risk groups. The developed model with the best predictive performance was temporally validated with a different cohort. Results: The voting ensemble machine learning algorithm showed an accuracy of 88.3%, Mathews’ correlation coefficient of 0.72, and weighted area under curve of 0.93, when temporally validated. Human papillomavirus (HPV) status, age of the patients, T stage, marital status, N stage, and the treatment modality (surgery with postoperative radiotherapy) were found to have the most significant effects on the ability of the machine learning model to predict overall survival. Similarly, for the individual patients with SHAP framework, HPV status, gender, and treatment modality (surgery with postoperative radiotherapy) were the input features that improved the model’s prediction. Conclusion: The proposed stratification of OPSCC patients into risk groups by machine learning techniques can provide accurate predictions and thus aid clinicians in administering early and personalized interventions. Clinicians could utilize the predicted risk with the explanations offered by the SHAP and LIME frameworks to understand previously undetected relationships between prognostic variables to make informed clinical decisions and effective interventions.

Larynx cancer survival model developed through open-source federated learning

IntroductionFederated learning has the potential to perfrom analysis on decentralised data; however, there are some obstacles to survival analyses as there is a risk of data leakage. This study demonstrates how to perform a stratified Cox regression survival analysis specifically designed to avoid data leakage using federated learning on larynx cancer patients from centres in three different countries. MethodsData were obtained from 1821 larynx cancer patients treated with radiotherapy in three centres. Tumour volume was available for all 786 of the included patients. Parameter selection among eleven clinical and radiotherapy parameters were performed using best subset selection and cross-validation through the federated learning system, AusCAT.After parameter selection, β regression coefficients were estimated using bootstrap. Calibration plots were generated at 2 and 5-years survival, and inner and outer risk groups’ Kaplan-Meier curves were compared to the Cox model prediction. ResultsThe best performing Cox model included log(GTV), performance status, age, smoking, haemoglobin and N-classification; however, the simplest model with similar statistical prediction power included log(GTV) and performance status only. The Harrell C-indices for the simplest model were for Odense, Christie and Liverpool 0.75[0.71–0.78], 0.65[0.59–0.71], and 0.69[0.59–0.77], respectively. The values are slightly higher for the full model with C-index 0.77[0.74–0.80], 0.67[0.62–0.73] and 0.71[0.61–0.80], respectively. Smoking during treatment has the same hazard as a ten-years older nonsmoking patient. ConclusionWithout any patient-specific data leaving the hospitals, a stratified Cox regression model based on data from centres in three countries was developed without data leakage risks. The overall survival model is primarily driven by tumour volume and performance status.

Predictors of survival in elderly patients with coronavirus disease 2019 admitted to the hospital: derivation and validation of the FLAMINCOV score

ObjectiveTo identify predictors of 30-day survival in elderly patients with coronavirus disease 2019 (COVID-19). MethodsRetrospective cohort study including patients with COVID-19 aged ≥65 years hospitalized in six European sites (January 2020 to May 2021). Data on demographics, comorbidities, clinical characteristics, and outcomes were collected. A predictive score (FLAMINCOV) was developed using logistic regression. Regression coefficients were used to calculate the score. External validation was performed in a cohort including elderly patients from a major COVID-19 centre in Israel. Discrimination was evaluated using the area under the receiver operating characteristic curve (AUC) in the derivation and validation cohorts. Survival risk groups based on the score were derived and applied to the validation cohort. ResultsAmong 3010 patients included in the derivation cohort, 30-day survival was 74.5% (2242/3010). The intensive care unit admission rate was 7.6% (228/3010). The model predicting survival included independent functional status (OR, 4.87; 95% CI, 3.93–6.03), a oxygen saturation to fraction of inspired oxygen (SpO2/FiO2) ratio of >235 (OR, 3.75; 95% CI, 3.04–4.63), a C-reactive protein level of <14 mg/dL (OR, 2.41; 95% CI, 1.91–3.04), a creatinine level of <1.3 (OR, 2.02; 95% CI, 1.62–2.52) mg/dL, and absence of fever (OR, 1.34; 95% CI, 1.09–1.66). The score was validated in 1174 patients. The FLAMINCOV score ranges from 0 to 15 and showed good discrimination in the derivation (AUC, 0.79; 95% CI, 0.77–0.81; p < 0.001) and validation cohorts (AUC, 0.79; 95% CI, 0.76–0.81; p < 0.001). Thirty-day survival ranged from 39.4% (203/515) to 95.3% (634/665) across four risk groups according to score quartiles in the derivation cohort. Similar proportions were observed in the validation set. DiscussionThe FLAMINCOV score identifying elderly with higher or lower chances of survival may allow better triage and management, including intensive care unit admission/exclusion.

Risk-group Classification by Recursive Partitioning Analysis of Patients Affected by Oligometastatic Renal Cancer Treated with Stereotactic Radiotherapy

AimsDue to the absence of consensus on metastases-directed treatment in kidney cancer, we conducted an analysis of patients treated with stereotactic radiotherapy (SRT) on cranial or extracranial metastases to classify them in survival class risk according to pre-treatment characteristics. Materials and methodsWe included oligometastatic kidney cancer patients treated with SRT on up to five metastases. Concomitant systemic treatment was allowed. End points included overall survival and the binary classification tree approach with recursive partitioning analysis was applied to stratify patients into overall survival risk groups. ResultsIn total, 129 patients were treated on 242 metastases. The brain was the most common site (34.71%), followed by lung (25.62%). With a median follow-up of 19.4 months, 1- and 3-year overall survival were 82.62 and 55.11%. The recursive partitioning analysis identified four prognostic classes. Class 1 included patients aged ≤ 65 years treated on extracranial metastases, with 3-year overall survival of 82.66%. Class 2 included patients aged > 65 years, without history of metastatic bone disease, treated on extracranial metastases, with a 3-year overall survival of 67.91%. Patients aged > 65 years and a history of bone disease, treated on extracranial metastases, were classified as class 3, with a 3-year overall survival of 37.50%. Class 4 included patients treated on brain metastases, with a 3-year overall survival of 9.70%. ConclusionWe produced a stratification model that can predict survival of oligometastatic kidney cancer patients treated with metastases-directed SRT. Site of disease, patient’s age and presence of bone disease can help clinicians in the decision-making process.

Prognostic risk stratification of gliomas using deep learning in digital pathology images.

BackgroundEvaluation of tumor-tissue images stained with hematoxylin and eosin (H&E) is pivotal in diagnosis, yet only a fraction of the rich phenotypic information is considered for clinical care. Here, we propose a survival deep learning (SDL) framework to extract this information to predict glioma survival.MethodsDigitized whole slide images were downloaded from The Cancer Genome Atlas (TCGA) for 766 diffuse glioma patients, including isocitrate dehydrogenase (IDH)-mutant/1p19q-codeleted oligodendroglioma, IDH-mutant/1p19q-intact astrocytoma, and IDH-wildtype astrocytoma/glioblastoma. Our SDL framework employs a residual convolutional neural network with a survival model to predict patient risk from H&E-stained whole-slide images. We used statistical sampling techniques and randomized the transformation of images to address challenges in learning from histology images. The SDL risk score was evaluated in traditional and recursive partitioning (RPA) survival models.ResultsThe SDL risk score demonstrated substantial univariate prognostic power (median concordance index of 0.79 [se: 0.01]). After adjusting for age and World Health Organization 2016 subtype, the SDL risk score was significantly associated with overall survival (OS; hazard ratio = 2.45; 95% CI: 2.01 to 3.00). Four distinct survival risk groups were characterized by RPA based on SDL risk score, IDH status, and age with markedly different median OS ranging from 1.03 years to 14.14 years.ConclusionsThe present study highlights the independent prognostic power of the SDL risk score for objective and accurate prediction of glioma outcomes. Further, we show that the RPA delineation of patient-specific risk scores and clinical prognostic factors can successfully demarcate the OS of glioma patients.

The Predictive Power of a Deep-Learning Segmentation Based Prognostication (DESEP) Model in Non-Small Cell Lung Cancer

Previous work has reported that imaging features captured by a deep learning tumor segmentation algorithm correlates with overall and disease-specific survival, even without direct supervision using survival information. The current study aims to create a model to predict survival outcomes for non-small cell lung cancer (NSCLC) after treatment with stereotactic body radiotherapy (SBRT) using deep-learning segmentation-based prognostication (DESEP).The DESEP model was trained using imaging from 108 patients with NSCLC with various clinical stages and treatment histories. The model generated predictions based on unsupervised features learned by a deep-segmentation network from computed tomography imaging to categorize patients into high and low risk groups for overall survival (DESEP-predicted-OS), disease specific survival (DESEP-predicted-DSS), and local progression free survival (DESEP-predicted-LPFS). Serial assessments were also performed using auto-segmentation based volumetric RECIST and computer-based unidimensional RECIST methods.There was a high concordance between DESEP-predicted-LPFS risk category and manually calculated RECISTv1.1 (φ = 0.544, P = 0.001). Neither the auto-segmentation based volumetric RECIST nor the computer-based unidimensional RECIST correlated with RECISTv1.1 (P = 0.081 and P = 0.144, respectively). While RECISTv1.1 correlated with LPFS (HR = 6.97,3.51-13.85, c = 0.70, P < 0.001), it could not provide insight regarding DSS (P = 0.942) or OS (P = 0.662). In contrast, the DESEP-predicted methods were predictive of LPFS (HR = 3.58, 1.66-7.18, c = 0.60, P < 0.001), OS (HR = 6.31, 3.65-10.93, c = 0.71, P < 0.001) and DSS (HR = 9.25, 4.50-19.02, c = 0.69, P < 0.001).Deep-learning segmentation-based prognostication can predict LPFS as well as OS, and DSS after SBRT for NSCLC. It can be used in conjunction with RECISTv1.1 to provide additional insights regarding DSS and OS in NSCLC patients receiving SBRT.

Elevated LncRNA TRERNA1 correlated with activation of HIF-1α predicts poor prognosis in hepatocellular carcinoma

BackgroundTranslation regulatory long non-coding RNA 1 (TRERNA1) has been reported to be upregulated in several cancers and accelerate tumor metastasis by inducing epithelial-to-mesenchymal transition (EMT). However, it remains unclear how TRERNA1 is upregulated and whether the upregulation of TRERNA1 influences the prognosis of HCC patients. In this study, we aimed to investigate the prognostic value of TRERNA1 in HCC and the regulatory mechanism underlying TRERNA1 upregulation. MethodsIn situ hybridization was adopted to analyze the expression level of TRERNA1, and immunohistochemistry technique was employed to evaluate the expression level of HIF-1α and E-cadherin. χ2 test was used to assess the association between TRERNA1 level and clinicopathological characteristics of HCC cases, whereas Kaplan-Meier survival analysis, ROC curve analysis and Cox proportional hazards regression model were performed to evaluate the prognostic significance of TRERNA1 expression level in HCC. ResultsTRERNA1 was substantially upregulated in HCC tissues, which was accompanied by aberrant decreased expression of E-cadherin. Elevated TRERNA1 level was correlated with high tumor grade, high recurrence rate and poor survival in patients with HCC. Moreover, TRERNA1 expression level was positively correlated with the activation of HIF-1α. Importantly, high TRERNA1 expression level can be an independent risk factor for poor prognosis in HCC, especially combining elevated TRERNA1 and HIF-1α with decreased E-cadherin predicted a worst prognosis of patients with HCC. ConclusionTRERNA1 is not only a biomarker for predicting poor prognosis in HCC patients, but also can categorize HCC patients into different risk groups for survival when combined with HIF-1α and E-cadherin.

Abstract SY05-03: Radiomics of immune therapy in NSCLC

Abstract Checkpoint blockade immunotherapy provides improved long-term survival in a subset of advanced stage NSCLC patients. Currently, predictive biomarkers of immunotherapy response are an unmet clinical need. CT and PET/CT images are routinely obtained during workup of NSCLC patients, and we hypothesize that quantitative analysis of these images (radiomics) can fill this need. In a study of CT radiomics, 332 NSCLC patients treated with checkpoint blockade were curated into independent training (N = 180), validation (N = 90) and test cohorts (N = 62) for radiomic analyses. The most informative clinical and radiomic features were subjected to Classification and Regression Tree (CART) analysis which identified novel cut-points to stratified patients into four different survival risk-groups. The identified very-high risk group was associated with extremely poor OS in both test ([HR] = 5.35, 95% CI= 2.14 - 13.36) and validation (HR = 13.81 95% CI: 2.58 - 73.93) cohorts when compared to the low-risk group (HRs = 1.00). Similar findings were observed for 3-year PFS (0% HR = 1.00 vs. 29.8% HR = 3.95). The most predictive radiomic feature was the GLCM inverse difference, which was representative of a set of nine features that were related to tumor heterogeneity. For further validation, this feature was then tested against four additional independent NSCLC data sets (total N = 446) who were treated with surgery, radio- or chemo-therapies, showing that it was significantly associated with OS suggesting a pan-marker. The biological underpinnings of the GLCM inverse difference were investigated using gene expression data and was associated with the hypoxia-related carbonic anhydrase, CA-IX, and confirmed by immunohistochemistry. In a separate study of PET/CT radiomics, we curated data sets from 837 NSCLC patients from four international institutions. Two residual-convolutional-network (ResCNN) models were generated to predict EGFR mutation and PDL1 positive status in training (N=429) and validation (N=187) cohorts and were subsequently tested on external cohorts to predict EGFR mutation status (N=65) and PDL1 expression status (N=85). The EGFR and PDL1 scores demonstrated high accuracy to predict EGFR mutation status with AUCs of 0.86, 0.83, and 0.81, and in prediction of PDL1 positive status with AUCs of 0.89, 0.84, and 0.82 for training, validation, and external test cohorts, respectively. The generated EGFR and PDL1 deep learning scores were subsequently used to predict PFS in patients treated with EGFR TKI (N=67) and chekpoint blockade (N=149). For patients with high EGFR score, EGFR-TKI treatment was significantly associated with longer PFS in patients with a low PDL1 score (p=0.013); for patients with a high PDL1 score, checkpoint blockade was significantly associated with longer PFS (p&amp;lt;0.001) in particular those patients with a low EGFR score. Therefore, the combination of EGFR and PDL1 scores, generated from the same PET-CT scan can be used as a non-invasive Decision Support Tool for NSCLC. Citation Format: Wei Mu, Ilke Tunali, Matthew B. Schabath, Robert J. Gillies. Radiomics of immune therapy in NSCLC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr SY05-03.

Prognostic Index for Survival in Adult Patients with Philadelphia-Negative B-Cell Acute Lymphoblastic Leukemia (ALL)

Prognostic Index for Survival in Adult Patients with Philadelphia-Negative B-Cell Acute Lymphoblastic Leukemia (ALL)