Search

4061 Background: The recent development of Immune checkpoint inhibitors (ICIs) has improved treatment outcomes for patients with esophageal cancer, although it may initiate autoimmune-related disorders in some patients. Across disease sites, patients who experience irAEs while on therapy with anti-PD-1 and anti-PD-L1 antibodies have been documented to experience improved outcomes as measured by overall response rate (ORR), progression-free survival (PFS) and overall survival (OS). Methods: We reviewed the medical records and the following characteristics of patients were collected: age, gender, Eastern Cooperative Oncology Group performance status (ECOG PS), stage, histology, history of surgery, metastatic sites, therapy line, presence of target lesion according to the response evaluation criteria in solid tumors (RECIST) version 1.1, LDH level before initiating ICIs treatment. We divided the patients treated with ICIs into two groups based on occurrence of irAEs. We compared the efficacy between the irAE - and irAE + groups. Results: There were total 295 patients. Of all the patients, 143 patients (50.2%) suffered from irAEs. The median patient age was 60 years (range, 36–84). Baseline characteristics including gender, ECOG PS, tumor location, history of surgery, histology, number of organs with metastases, site of metastases, therapy line, history of drinking, history of smoking between patients with or without irAE were not significantly different. The most frequent irAEs were Anemia(71/295, 24.07%), Pneumonitis (55/295, 18.64%), Cardiovascular toxicities (35/295, 11.86%). There were no grade 5 adverse events related to immunotherapy. Objective response was observed in 92 patients (31.19%), 54 of the 143 patients with irAEs had objective response (37.76%) in contrast with 38 of the 152 cases without irAEs (25.00%). Median PFS in patients with irAEs was 10.27 months and 6.2 months in those without irAEs (p8, Radiation and Antiangiogenic therapy were positively associated with occurrence of irAEs(p

e20620 Background: This study was to investigate the association between cachexia and prognosis in extensive-stage small-cell lung cancer(ES-SCLC) patients who received first-line immunotherapy. And the prognostic values of some systematic inflammatory and nutritional indexes were also evaluated. Methods: Data of 123 ES-SCLC patients receiving immunotherapy as first-line treatment in one single center from June 2019 to February 2021 were analyzed. Development of cachexia after the initiation of immunotherapy was defined as patients with more than 5% weight loss over past 6 months, or more than 2% weight loss with body mass index (BMI) less than 20 kg/m2. Endpoints included response rate, progression-free survival (PFS) and overall survival (OS). Univariate and multivariate survival analysis were performed to identify the factors correlated with PFS and OS. Results: At evaluation after 6 months of initial treatment, cachexia was presented in 51 patients(41.5%). The treatment response and survival of patients with cachexia were worse than those without. The overall median overall survival (OS) was 14.4 months (95% CI 13.3–15.6). The overall median progression-free survival (PFS) was 8.1 months (95% CI 7.0–9.2). The median PFS was significantly longer in patients without bone metastases at diagnosis(HR = 2.362,95%CI:1.452-3.842;p = 0.001), without cachexia(HR = 1.802,95% CI: 1.121-2.895; p = 0.015), and with lower change of total protein(HR = 1.799,95% CI: 1.084–2.984; p = 0.023). The median OS was significantly shorter in patients with poor ECOG PS (HR = 1.904, 95% CI: 1.017-3.564; p = 0.044), higher baseline SII (HR = 2.135,95% CI: 1.117–4.080; p = 0.022), with cachexia(HR = 2.020,95% CI: 1.075-3.795; p = 0.029), advance M stage(HR = 3.111,95% CI:1.545–6.262; p = 0.001) and higher post-treatment SIRI (HR = 2.135,95% CI: 1.281–4.697; p = 0.007). Conclusions: ES-SCLC patients who develop cachexia after first-line immunotherapy have a poor prognosis. our study also confirms that baseline SII and SIRI at 6 weeks after treatment appears to be promising predictors for OS of ES-SCLC who treated with first-line immunotherapy.

2580 Background: Early prediction of symptomatic pneumonitis (grade ≥ 2) could potentially assist the comprehensive management for non-small cell lung cancer (NSCLC) patients who received radiotherapy combined with immunotherapy. A more accurate integrative model for symptomatic pneumonia prediction using CT images is needed. Methods: This retrospective study contains 243 NSCLC patients (62 symptomatic pneumonitis) who underwent radiotherapy combined with immunotherapy between January 2015 and June 2021. Five-fold cross-validation was performed for training and testing. There were 195 cases in the training set (50 symptomatic pneumonitis) and 48 cases in the validation set (12 symptomatic pneumonitis) in each fold. The deep graph integrative model (DG) was composed of two pre-trained 3D UNet encoders to extract deep features from tumor and lung volumes of pre-treatment CT images, respectively, and a graph attention layer (GAT) for integration and classification. The encoders were fine-tuned using manually segmented tumor and lung CT volumetric patches from the training set. Evaluation measures include area under ROC curve (auc), sensitivity (sen), and specificity (spe). Results: Our new DG achieved auc, sen, and spe of 0.823, 0.767, and 0.810, which outperformed conventional CT radiomics model (auc 0.743, sen 0.620, spe 0.752), 3D UNet based deep radiomics model (auc 0.761, sen 0.746, spe 0.737), and our model without GAT (auc 0.796, sen 0.762, spe 0.782). The improvement was statistically significant (p < 0.001). Conclusions: Our DG model improved symptomatic pneumonia prediction using CT images, which can be used as a tool to effectively improve the safety and personalized treatment of combined radiotherapy with immunotherapy for NSCLC patients.

e21055 Background: PD-L1 is a recommended biomarker for immunotherapy, but Some patients with high pD-L1 expression had poor clinical response. Considering the inconsistent correlation between PD- L1 expression and ICI efficacy, A more sensitive biomarker is needed to predict immunotherapy efficacy. B-cell lymphoma-2 (BCL2L11/BIM) belongs to the Bcl-2 family and is a key factor promoting apoptosis. BIM deletion polymorphism (BIM- del) may lead to BH3 deletion and thus hinder cell apoptosis. Loss and decreased expression of BIM may lead to resistance to tumor therapy. A study on patients with metastatic melanoma found that the high level of BIM in CD8+ T cells suggests a stronger PD-1 and PD-L1 interaction, which predicts clinical benefit in patients with metastatic melanoma receiving anti-PD-1 therapy.This study aims to evaluate the relationship between BIM deletion and PD-1/PD-L1 inhibitors efficacy in advanced NSCLC patients. Methods: Patients diagnosed with advanced stage NSCLC that treated with Immune checkpoint inhibitors (ICIs) at Shandong Cancer Hospital between March 2019 and January 2021 were enrolled in this study. Peripheral blood samples from these patients were collected to detect BIM-del by polymerase chain reaction. Cox proportional hazards models were used to analyze the clinical outcomes of patients with and without BIM-del. Results: A total of 61 patients with advanced NSCLC who received immunotherapy were enrolled, only 18.0% (11/61) patients were detected with BIM-del. Most of the patients were male (73.8%, 45/61), younger than 65 years (72.1%, 44/61), non- smokers (50.8%, 31/61), previous treatment lines 0-1 (90.2%, 55/61),had Eastern Tumor Cooperative group (ECOG) performance status (PS) 0-1 (98.4%, 60/61). There were no associations between the BIM-del and clinical characteristics. There was no difference in the objective response rate between patients with and without BIM-del (27.3% vs 30 %, p = 1.000).BIM-del is associated with shorter progression-free survival (PFS) (5.7 months vs12.0 months, p = 0.001). The median OS had not been reached. Multivariate analysis suggested that BIM-del was an negative independent prognostic factor for PFS in immunotherapy patients with non-small cell lung cancer, but not for OS. Conclusions: BIM-del may be a negative prognostic biomarker for patients with advanced NSCLC treated with immunosuppressive therapy. It is very important to find effective and accurate biomarkers to predict the PD-1/PD-L1 inhibitors Treatment efficacy, and further exploration is needed in the future.

e14561 Background: Cancer patients with immune checkpoint inhibitors (ICIs) treatment often experience unique progression response patterns, accounting for the discordance of short-term and long-term benefits. Identifying the specific characteristics of different quality progression disease (PD) events can maximize the curative effect of ICIs and further improve the prognosis of the patients with immunotherapy. In this study, we established a neutrophil-to-lymphocyte ratio (NLR)-based nomogram model which could predict survival for NSCLC patients with primary resistance to ICIs. Methods: Univariate and multivariate Cox regression analysis were used to evaluate the ability of each parameter to predict overall survival (OS). The regression coefficients obtained in multivariate analysis were visualized in the form of a nomogram, thus a new nomogram and risk classification system were established. The discrimination ability of the new model was evaluated using area under the curve (AUC) in receiver operating characteristic curve (ROC) analysis. Survival curves were made by the Kaplan-Meier method and compared by the Log-rank test. Results: A total of 1235 patients with multiple cancer types who accepted at least one dose of ICIs treatment between 2015 and 2018 at Memorial Sloan Kettering Cancer Center were included in our study. Among them, 624 PD-patients (obtaining PD as their best response) were used to identify potential survival-related characteristics and build a nomogram model. Univariate and multivariate analysis demonstrated that pretreatment NLR, treatment lines and Eastern Cooperative Oncology Group performance status scores were independent prognostic factors and were included in the predictive model. The ROC curves implied the good discrimination ability of the predictive model with the AUC value for 1-year OS was 0.756. And, the calibration curves showed that the predicted value of the 1-year survival rate by the nomogram was in good agreement with the actual observed value of it. Based on the nomogram, we divided PD-patients into high-, med- and low-risk subgroups. The OS of cancer patients in three risk groups were accurately differentiated either in all PD-patients or PD-NSCLC-patients, but not in the patients who responded to ICIs. Importantly, the prognosis of the low-risk PD-patients was significantly better compared with the patients in med- and low-risk groups (P < 0.001). Moreover, in the PD-patients with NSCLC, similar results were also found that patients had a worse OS in the high-risk group (P < 0.001). Interestingly, in contrast to the results in PD-patients, a comparable OS between high-, med- and low-risk groups based on our model was verified in the patients who responded to ICIs. Conclusions: The pretreatment NLR-based nomogram could predict the prognosis for pan-cancer patients with primary resistance to ICIs treatment.

e21051 Background: Tumor-infiltrating B cells accompanying several widely used B cell-related biomarkers, such as CD19 and CD20, have inconsistent clinical prognostic values in non-small-cell lung cancer (NSCLC) patients. Considering only one B cell-related biomarker could not distinguish the anti-tumor and pro-tumor B cell subsets in tumor, we aimed to construct a more accurate B-cells related gene pairs (BRGPs) prognostic model and evaluate its potential predictive ability to immunotherapy in NSCLC patients. Methods: Using public single-cell RNA sequencing data, the B cells-related genes (BRGs) in NSCLC samples were identified. The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) datasets were utilized to construct the BRGPs model which was not affected by the technical bias of different platforms. With no dependence upon specific gene expression levels, a novel signature based on BRGPs was established in this study. In addition, the prognostic value and immunotherapeutic response for this signature with regard to the TME components and potential molecular mechanism were explored. Results: Based on the TCGA database, we built a novel prognostic signature of 23 BRGPs comprising 28 unique BRGs. This risk model showed significant power in distinguishing good or poor prognosis and could serve as an independent prognostic factor for NSCLC patients in the TCGA cohort. The prognostic accuracy of the model was further verified in the GSE31210 dataset from the GEO database. Likewise, the prognostic value of the risk score for the BRGPs model was demonstrated in the GEO cohorts by the univariate and multivariate cox regression analysis. In addition, the risk model was significantly associated with sex, TNM stage, immune score, tumor purity and various tumor-infiltrating immune cells. GSEA analysis indicated that low-risk group enriched with several immune-related pathways, such as activation of immune response, antigen receptor mediated signaling pathway, B cell activation and B cell mediated immunity, whereas several proliferation-related pathways, such as nuclear chromosome segregation, sister chromatid segregation and mitotic sister chromatid segregation were most enriched in high-risk group. Besides, the tumor mutational burden (TMB) score rather than CD274(PD-L1 mRNA) expression was positively correlated with the risk score (P＜0.001; P = 0.94, respectively). NSCLC patients with high-risk exhibited significantly higher TMB score compared with low-risk patients (P < 0.001). Correspondingly, we demonstrated that immune checkpoint blockade therapy may be more efficacious in high-risk group NSCLC patients according to TIDE method (P＜0.01). Conclusions: This novel BRGPs model can assess the prognosis of patients with NSCLC, and may be helpful to guide immune checkpoint inhibitors treatment in our clinical practice.

e20591 Background: Alectinib and thoracic radiotherapy (TRT) are important modalities in the management of ALK-rearranged non-small cell lung cancer. We therefore aimed to assess the incidence and risk factors of pneumonitis when these two treatments were combined. Methods: Patients with ALK-rearranged non-small cell lung cancer, receiving alectinib and TRT from January 2018 to December 2020 were reviewed, and the clinical and dosimetry data were collected. Logistic regression analyses were performed to evaluate risk factors associated with treatment-related pneumonitis (TRP). The prediction ability of dosimetry parameters for pneumonitis were examined by receiver-operating characteristic (ROC) curve analyses. Results: Of the 25 enrolled patients, 16 (64%) developed pneumonitis, and 12 (48%) developed grade 2 or higher pneumonitis. Logistics regression analyses revealed chemotherapy (OR = 8.750, 95% CI 1.207 - 63.428, p = 0.032), higher ipsilateral V5 (OR = 1.060, 95%CI 1.002 - 1.122, p = 0.043) and total lung V5 (OR = 1.083, 95%CI 1.000 - 1.174, p = 0.05) to be risk factors for TRP. Total lung V10, total lung V5, and ipsilateral V5 were strong predictors of TRP among the dosimetry parameters (AUC = 0.800, 0.786 and 0.781, respectively). After developing pneumonitis, 52% of patients recovered or improved, but one patient died due to respiratory failure. Conclusions: The combined use of alectinib and TRT significantly increased the risk of pneumonitis. Clinicians should consider the elevated risks and related dosimetric factors when deciding on combination treatment for ALK-rearranged NSCLC patients.

8563 Background: Impower133 and CASPIAN studies showed that PD-L1 antibody combined with first-line chemotherapy could prolong the overall survival. Previous studies have shown that radiotherapy could potentially promote tumor antigen presentation and reverse immunosuppressive microenvironment in tumor. The purpose of this study was to explore the efficacy and safety of SHR-1316 (PD-L1 antibody) combined with chemotherapy and sequential chest radiotherapy as first-line therapy for ES-SCLC. Methods: Key inclusion factors were 18-75 years old, histologically or cytologically confirmed ES-SCLC, ECOG performance status 0-1, no previous systematic treatment. Patients (pts) included in this study received 4̃6 cycles of SHR-1316 (20mg/kg, D1, q3w) combined with EP/EC (cisplatin 75mg/m² D1-3 q3w or carboplatin AUC = 5, D1 q3w; etoposide 100mg/m2, D1-5, q3w), sequentially SHR-1316 combined with chest radiotherapy (≥3Gy*10f or ≥2Gy*25f, involved- field irradiation), and then entered the maintenance treatment stage until disease progression or intolerable side effects. The main endpoints included PFS, ORR and safety. Results: From October 2020 to December 2021, 31 pts with ES-SCLC received at least one dose of SHR-1316. The median age was 64 (range: 37-75) with 25(80.6%) male, 22(71%) former smokers and 31 (100%) ECOG performance status 1. 17 (54.8%) pts were with brain metastasis, 8 (25.8%) pts with liver metastasis, 8 (25.8%) pts with kidney/adrenal metastasis, 7(22.6%) pts with bone metastasis. At the data cutoff date, 15 pts remained on treatment, the average number of treatment cycles was 5.19. 24 pts had at least one 1 post-treatment tumor assessment. The median PFS was 7.56 months, the confirmed ORR and DCR in all pts were 50% (12/24) and 87.5% (21/24), respectively. The confirmed ORR and DCR in pts with brain metastasis were 38.5% (5/13) and 76.9% (10/13), and were 63.6% (7/11) and 100% (11/11) in pts without brain metastasis. In pts received chest radiotherapy, the confirmed ORR and DCR were 80% (8/10) and 100% (10/10). During the study period, 23 (74.2%) pts had adverse drug reactions, and 16 (51.6%) pts had grade 3 or 4 adverse drug reactions, including 12 (38.7%) neutropenia 8 (25.8%) leukopenia, 2 (6.5%) thrombocytopenia, 2 (6.5%) anemia, 1(3.2%) lymphocytopenia, 1(3.2%) amylase increased. No grade 5 adverse drug reaction was observed. Conclusions: SHR-1316 combined with chemotherapy and sequential chest radiotherapy as first-line therapy for ES-SCLC showed promising efficacy and acceptable safety. It is worthy of further clinical exploration. Clinical trial information: NCT04562337.

e20519 Background: Immune checkpoint inhibitors have revolutionized the patterns of tumor treatment with significantly improved survival. However, it’s worth noting that the objective response rate of PD-1/PD-L1 inhibitors monotherapy is just 20% in non-small cell lung cancer (NSCLC) patients, and biomarkers for the prognosis of patients receiving PD-1/PD-L1 inhibitors are urgently needed. LILRB2 (ILT4) is known as an immunosuppressive molecule and promotes the development and progression of NSCLC. However, the relationship between LILRB2 expression and immune microenvironment remains unclear in patients with lung adenocarcinoma. Methods: Expression profile and corresponding clinicopathological data of patients with lung adenocarcinoma were obtained from The Cancer Genome Atlas (TCGA) database. TIMER 2.0 was used to investigate the relationship between LILRB2 expression and immune infiltrates, including T cell subset, macrophage and so on. Cox proportional hazard model was also performed to evaluate the role LILRB2 expression and immune infiltrations in prognostic prediction of patients with lung adenocarcinoma. Results: 515 patients with lung adenocarcinoma were included in analysis from TCGA dataset. LILRB2 expression was found to have significant negative correlation with CD8+ T cell infiltration (P = 0.0012), and positive correlation with CD4+ T cell infiltration (P < 0.001). There were also significant positive correlations between LILRB2 expression and Treg (P = 0.0083), cancer associated fibroblast (P < 0.001), monocyte (P < 0.001), M2 macrophage (P < 0.001) infiltration, which was regarded as the marker of immunosuppressive microenvironment. Notably, there were also significant association between the expression of LILRB2 and PD-1 (P < 0.001) and PD-L1(P < 0.001). In terms of survival analysis, high level of Treg (HR = 0.86, P = 0.03) and M2 macrophage (HR = 0.84, P = 0.017) infiltration was associated with poor prognosis of patients with lung adenocarcinoma. However, the level of CD8+T cell (HR = 0.93, P = 0.298), CD4+ T cell (HR = 0.933, P = 0.321), and LILRB2 expression (P = 0.56) was not associated with the survival of patient with lung adenocarcinoma. Conclusions: The overexpression of LILRB2 is significantly correlated with immunosuppressive immune infiltration and microenvironment in lung adenocarcinoma and might be the potential predictive biomarkers for the PD-1/PD-L1 inhibitors.

e16071 Background: Distant metastasis with an incidence of 25% in esophageal cancer(EC) represents a poor prognosis. However, there was few study for prediction of distant metastasis in EC, due to unsatisfactory specificity of clinical factors and lack of reliable biomarkers. Methods: Two hundred and ninety-nine patients were enrolled and randomly assigned to a training cohort(n = 207) and a validation cohort(n = 92). Logistic univariate and multivariate regression analyses were used to identify clinical independent predictive factors and construct a clinical nomogram. Radiomic features were extracted from contrast-enhanced CT performed before treatment, and Lasso regression was used to screen the optimal features, which were developed a radiomics signature subsequently. Four machine learning algorithms were used to establish radiomics models respectively based on the screened features. The joint nomogram incorporating radiomics signature and clinical independent predictors was developed by logical regression algorithm. All models were further validated by discrimination,caliberation, reclassification and clinical usefulness. Results: The joint nomogram had a better performance [AUC(95%CI): 0.827(0.742-0.912)] than clinical nomogram [AUC(95%CI): 0.731(0.626-0.836)]and radiomics predictive models[AUC(95%CI): 0.747(0.642-0.851),SVM algorithms]. Caliberation curve, and decision curve analysis also revealed joint nomogram outperformed the other models. Compared with the clinical nomogram, net reclassification Improvement(NRI) of the joint nomogram was improved by 0.114(0.075, 0.345),and integrated discrimination Improvement (IDI) was improved by 0.071(0.030-0.112), P= 0.001. Conclusions: We constructed and validated the first joint nomogram for distant metastasis in EC based on radiomics signature and clinical independent predictive factors, which could help clinicians to identify patients with high risk of distant metastasis.

e12594 Background: Triple-negative breast cancer (TNBC) is the subtype of breast cancer with the worst prognosis. There is no reliable model for survival prediction of TNBC patients. The traditional Cox regression analysis with poor prediction power cannot satisfy the clinical needs. The purpose was to establish a deep learning model and develop a new prognostic system for TNBC patients. Methods: This study collected data of TNBC patients from the Surveillance, Epidemiology, and End Results (SEER) program between 2010 and 2016. 70% were used to develop the deep learning model, 15% were used as the validation set, and 15% as the independent testing set. Then the concordance-index (c-index) and Brier score (IBS) were calculated and compared with the Cox regression analysis and random forest. Finally, according to the classification of the deep survival model, an individualized prognosis system was established. Results: A total of 37,818 patients were enrolled in this study. In the validation set, the c-index of the deep learning was 0.799, which was better than the traditional Cox regression model (0.774) and random forest (0.763). The independent testing set further proved the robustness of the deep survival model (c-index 0.788). The new prognosis system based on the deep survival model reached an area under the curve (AUC) of 0.805, which was better than the Tumor, Node, Metastases (TNM) staging system (0.771). Conclusions: Deep learning model had better prediction power than the Cox regression analysis and the random forest. The established prognosis system can better predict prognosis and aid individual risk stratification for TNBC patients patients.

e21032 Background: Previous studies suggested that MET exon 14 ( METex14) mutation regarding as a distinct subset was sensitive to MET-inhibitors, but poorly response to immunotherapy. Conversly, MET non-exon-14 (non-ex14) mutations including those undetermined functions and affecting the kinase or extracellular domains were found to be associated with the resistance to MET-inhibitors. However, therapeutic strategies for MET-non-ex14 mutant cancer are still largely unknown, and the relationship between MET-non-ex14 mutations and the efficacy of immune checkpoint inhibitors (ICIs) has never been reported. Using two public ICIs-treated cohorts, we aimed to assess the role of MET mutations including both METex14 and MET-non-ex14 mutations in NSCLC patients undergoing ICIs therapy. Methods: A total of 385 ICIs-treated NSCLC patients were enrolled to our study. MET mutations were defined as any nonsynonymous mutations, and we divided them into METex14 and MET-non-ex14 mutation subsets according to the mutated-position on MET. Kruskal-Wallis test was used to analyze the difference of tumor mutational burden (TMB) score, and the Fisher’s exact test was applied to compare the rates of durable clinical benefit (DCB). Log-rank test was used to analyze the differences between Kaplan-Meier survival curves. Results: In the entire cohort, we found that 17 patients (17/385, 4.4%) had MET mutations, most of which were pure METex14 mutations (10/17, 58.8%). The median TMB of patients in the entire NSCLC cohort was 6.89 mut/Mb. MET-non-ex14 mutant patients (7/385, 1.8%) possessed a significantly higher TMB than METex14-mutant (10/385, 2.6%) and MET wild-type (368/385, 95.6%) sub-cohorts, respectively (median TMB, 17.92 mut/Mb versus 4.17 mut/Mb, p = 0.008; 17.92 mut/Mb versus 6.96 mut/Mb, p = 0.01, respectively). DCB was more common in patients harbored MET-non-ex14 mutations than patients with METex14 mutations and MET wild-type (66.7% versus 14.3%, p = 0.103; 66.7% versus 29.9%, p = 0.075, respectively). We found that patients with MET-non-ex14 mutations had a numerically longer progression free survival (PFS) than those with METex14 mutations and MET wild-type (p = 0.169). Moreover, the PFS was significantly longer in MET-non-ex14-mutant subgroup than patients with METex14 mutations (median PFS, 9.1 versus 2.1 months, p = 0.025). Correspondingly, the overall survival (OS) was significantly longer in MET-non-ex14-mutant subgroup than their wild-type counterparts (median OS, not reached versus 11 months, p = 0.039). Additionally, patients with MET-non-ex14 mutations exhibited relatively better OS versus METex14-mutant patients (median OS, not reached versus 18 months, p = 0.175). Conclusions: MET-non-ex14 mutations were associated with higher TMB, improved DCB rate, and could act as a favorable prognostic biomarker in ICIs-treated NSCLC patients.

e16093 Background: Lymph node (LN) metastasis is the most important factor for decision making in esophageal squamous cell carcinoma (ESCC). A more accurate prediction model for LN metastatic status in ESCC patients is needed. Methods: In this retrospective study, 397 ESCC patients who took Contrast-Enhanced CT (CECT) within 15 days before surgery between October 2013 and November 2018 were collected. There are 924 (798 negative and 126 positive) LNs with pathologically confirmed status after surgery. All LNs were randomly divided into a training set (n = 663) and validation set (n = 185). Data augmentation including shifting and rotation was performed in the training set, resulting in 1326 negative and 1140 positive LN samples. The GACNN model was trained over CT volumetric patches centred at manually segmented LN samples. GACNN was composed of a 3D UNet encoder to extract deep features, and a graph attention layer to integrate morphological features extracted from segmented LN. The model was validated using the validation set (135 negative and 50 positive) and measured by area under ROC curve (auc), sensitivity (sen), and specificity (spe). Results: GACNN achieved better auc, sen, and spe of 0.802, 0.765, and 0.826, when compared to 3 other models including CT radiomics model (auc 0.733, sen 0.689, spe 0.765), 3D UNet encoder (auc 0.778, sen 0.722, spe 0.767), and our model without morphological features (auc 0.796, sen 0.754, spe 0.803). The improvement was statistically significant (p < 0.001). Conclusions: Our prediction model improved the prediction of LN metastasis, which has the potential to assist LN metastasis risk evaluation and personalized treatment planning in ESCC patients for surgery or radiotherapy.

e20591 Background: The main aim of this study was to evaluate the efficiency of albumin-bound paclitaxel (nab-PTX) plus PD-1/PD-L1 inhibitor versus second-line chemotherapy in small cell lung cancer (SCLC) patients who have failure to the first-line standard treatment. Methods: We retrospectively collected patients’ data from medical records between January 2015 to July 2020 in Shandong Cancer Hospital and Institute. Consecutive 42 patients who were treated with nab-PTX plus PD-1/PD-L1 inhibitor were enrolled and compared with, 126 patients who received second-line chemotherapy (1:3 matched with patient and tumor characteristics). Progress free survival (PFS), overall survival (OS), objective response rate (ORR) and disease control rate (DCR) were evaluated for each group. Results: Patients treated with nab-PTX plus immunotherapy group and second-line chemotherapy group achieved the median PFS of 5.6 months and 3.3 months ( p = 0.043), respectively. The median OS were 7.7 months and 6.3 months ( p = 0.021), respectively. The ORR and DCR were also higher in nab-PTX plus PD-1/PD-L1 inhibitor group (ORR: 33.3% vs 20.6%, p = 0.094; DCR: 61.9% vs 41.3%, p = 0.020, respectively). The most common incidences of grade ≥3 adverse events were leukopenia and neutropenia, there were no significance difference between the two groups. Conclusions: Albumin-bound paclitaxel plus PD-1/PD-L1 inhibitor conferred higher ORR and DCR, and improved PFS and OS in SCLC patients failed with first-line treatment. Further prospective and randomized trial that directly compares the treatments is urgently warranted.

e12625 Background: Neutrophil-to-lymphocyte ratio (NLR) has been found to offer clear predictive utility for the overall survival (OS) and pathologic complete response (pCR) in breast cancer (BC) patients receiving neoadjuvant chemotherapy(NAC). However, previous studies mainly focused on pre-NLR. The aim of this study was to explore the role of pre-/post-NLR towards OS and longitudinal NLR kineticsonpCR for BC patients undergoing NAC. And we also tried to build a nomogram for OS prediction based on these parameters. Methods: In this study, we retrospectively collected 501 female patients with locally advanced BC receiving 4-8 cycles of NAC from 2009 to 2018. Clinicopathological characteristics, NLR at pre-, mid-(every two cycles of NAC) and post-treatment were collected. The primary endpoint was OS. Among the patients, 421 patients with available pre- and post-NLR were included in the survival analysis. These patients were randomly divided into a training cohort (n = 224) and a validation cohort (n = 197). Multivariate survival model was built by including all the significant prognostic factors from the univariate analysis in the training cohort, and a nomogram model was established by “R” version 3.4.3. The performance of the model was further tested in the validation cohort by the concordance index. The second endpoint was pCR. Longitudinal analysis of NLR was performed using a mixed-effects regression model to predict pCR among 176 patients who finished 8 cycles of NAC. Results: The median follow-up time was 43.2 months for 421 patients. In the training cohort, multivariate analysis revealed that ER status, clinical node stage , pCR , pre-NLR, and post-NLR (all Ps＜0.05) were independent predictors of OS. Nomogram for OS prediction was established by combining all these significant factors. The C-indexes of the nomogram were 0.764 and 0.693, respectively in the training and validation cohort. In the longitudinal analysis, patients achieved pCR experienced a reduction of NLR every 2 cycles (Coef = -0.032, std error = 0.014, P = 0.024). Conclusions: This study demonstrated the prognostic value of pre-NLR and post-NLR towards BC patients received NAC. Based on that , a novel nomogram was established to predict the 3- and 5- year OS for BC patients. And we also found patients who experienced a decline of NLR during NAC seems to be more likely to achieve pCR from chemotherapy. Routine assessment of NLR may be an easy and affordable tool for defining prognosis.

e21667 Background: Previous studies suggested that the efficacy of osimertinib was better in patients with epidermal growth factor receptor ( EGFR) exon 19 deletion (ex19del) than exon 21 Leu858Arg (L858R). However, the differential clinical outcomes of osimertinib in previous tyrosine kinase inhibitors (TKIs) treated non-small-cell lung cancer (NSCLC) patients with brain metastases (BrMs) according to the two different common EGFR genotypes remains undetermined. Methods: We retrospectively collected EGFR mutant (ex19del and L858R) NSCLC patients with BrMs and received osimertinib treatment between Jan 2016 and Feb 2019 in our hospital. Other eligible standards including Eastern Cooperative Oncology Group (ECOG) performance status (PS) 0-2, osimertinib 80mg monotherapy and neurologic symptoms existing was permitted to enroll in our study. Results: 78 patients with EGFR mutant NSCLC and BrMs were included in the study. EGFR mutation types were ex19del in 55% of patients (43/78) and L858R in 45% of patients (35/78). All of patients received at least one prior EGFR-TKI therapy, among which 46% (36/78) had gefitinib, 49% (38/78) had erlotinib, 19% (15/78) had icotinib. T790M status at osimertinib initiation was positive in 81% of patients (63/78), negative in 4% (3/78), unknown in 15% (12/78). The systemic objective response rate (ORR) was 46%. There was no apparent difference in systemic ORR (49% vs 43%, P = 0.598) between ex19del group and L858R group. The median progression free survival (PFS) and overall survival (OS) were 7.0 (range 4.7-9.3) and 18.8 (range 14.0-23.5) months for all patients. The median PFS in the ex19del and L858R subgroups was 9.1 (range 6.0-12.1) months and 5.3 (range 3.4-7.2) months, respectively ( P = 0.031). The median OS in the ex19del and L858R subgroups was 26.0 (range 13.7-38.3) months and 13.9 (range 7.7-20.2) months, respectively ( P = 0.033). Multivariate analysis identified L858R and ECOG PS 2 as poor independent predictors to the PFS and OS. Conclusions: The efficacy of osimertinib was associated with EGFR genotypes in pre-TKI treated NSCLC patients with BrMs, and L858R maybe an independent bad factor for long-term outcomes of osimertinib.

4553 Background: We aimed to propose a risk model based on MM-Atten-CNN for predicting esophageal fistula in patients with esophageal cancer (EC) from computerized tomography (CT) -based radiomics. Methods: EC patients who didn’t received esophageal surgery between July 2014 and August 2019 were collected. Of these, 186 patients (cases) who developed esophageal fistula were enrolled and compared with 372 controls (1:2 matched with the diagnosis time of EC, sex, marriage, and race). All 558 patients were divided into training set (n = 390) and validation set (n = 168) randomly. The MM-Atten-CNN risk model was trained over 2D slices from nine views of planes, where there were three patches of contextual CT, segmented tumor and neighbouring information in each view. In the training set (130 cases and 260 controls), data augmentation was performed including pixel shifting [-10, -5, +5, +10] and rotation [-10, +10]. In total, there were (130+260) *16*2 = 12480 subjects used for training. Finally, the risk model was validated in the validation set (56 cases and 112 controls) and measured by accuracy (acc), sensitivity (sen), and specificity (spe). Results: The developed risk model achieved (acc, sen, spe) of (0.839, 0.807, 0.926), which were more predictive for the occurrence of esophageal fistula when compared to CNN models using single coronal view (acc 0.763, sen 0.581, spe 0.837), multi-view 2D contextual CT slices (acc 0.779, sen 0.656, spe 0.896), and 3D CNN using contextual CT volumes (acc 0.781, sen 0.689, spe 0.852). Conclusions: MM-Atten-CNN CT-based model improved the performance of esophageal fistula risk prediction, which has the potential to assist individualized stratification and treatment planning in EC patients.

e21066 Background: The immune status of tumor microenvironment is extremely complex. One single immune feature cannot reflect the integral immune status and its prognostic value was limited. We postulated that the immune signature based on multiple immuno-features could markedly improve the prediction of post-chemoradiotherapeutic survival in inoperable locally advanced non-small-cell lung cancer (LA-NSCLC) patients. Methods: In this study, 100 patients who were diagnosed as inoperable LA-NSCLC between January 2005 and January 2016 were analyzed. A 5-immune feature-based signature was then constructed using the nested repeat 10-fold cross validation with LASSO Cox regression model. Nomograms were then established for predicting prognosis. Results: Immune signature combining 5 immuno-features were significantly associated with OS and PFS (P = 0.002 and P = 0.014, respectively) in patients with inoperable LA-NSCLC, and at a cutoff of -0.198 stratified patients into two groups with 5-year OS rates of 39.8% and 8.8%, and 2-year PFS rates of 22.2% and 5.5% for the high- and low-immune signature groups, respectively. Using immune signature, we proposed immune signature nomograms, which were better than the traditional TNM staging system in terms of discriminating ability (OS: 0.692 vs. 0.588; PFS: 0.672 vs. 0.586, respectively) or net weight classification (OS: 32.96%; PFS: 9.22%), suggesting that immune signature plays a complementary role in the prognosis prediction of patients with inoperable LA-NSCLC. Conclusions: Multiple immune features based immune signature could effectively predict recurrence and survival of inoperable LA-NSCLC patients, and complemented the prognostic value of the TNM staging system.

e21030 Background: The clinical benefits of immunotherapy in patients with stage I non-small cell lung cancer (NSCLC) is still controversial. Immune status plays critical role in the development and progression of NSCLC, and is associated with the patient survival outcomes. The analysis of immune features is thus valuable for the determination of immunotherapy. However, one single immune feature cannot reflect the complex immune status, and its prognostic value is extremely limited. In this study, we aimed to construct an immunoscore classifier based on multiple immuno-genes to predict the prognosis of patients with early NSCLC. Methods: A total of 522 patients with stage I NSCLC were included in this study. All patients' follow-up records and gene expression data were completely preserved. A least absolute shrinkage and selection operator (LASSO) algorithm was used to screen immune-related genes, and a COX proportional hazard regression model was used to construct the immunoscore classifier based on multiple immune-genes. Besides, the net reclassification improvement (NRI) calculation and concordance index (C-index) were applied to quantify the improvement of usefulness added by the immunoscore classifier compared to TNM staging system. Results: The immunoscore classifier including CCL5, CD8A, CXCL9, HLA-DQA1, LAG3, STAT1, and CD276 was significantly correlated with OS (HR: 2.785 CI: 1.809-4.289 P < 0.001) in patients with stage I NSCLC. With the optimal cut-off value of 4.32, all patients can be divided into a low-risk immune group and a high-risk immune group. The 10-year survival rates of the two groups were 36.8% and 12.3%, respectively. Besides, the immunoscore classifier was superior to the traditional TNM staging system in terms of distinguishing ability (C-index improvement by 0.075) and net reclassification ability (NRI improvement by 11.29%), indicating that the immunoscore classifier plays an important role in improving prognostic value. Conclusions: Multiple immune-genes based immunoscore classifiers can effectively predict the prognosis of patients with stage I NSCLC, and is significantly superior to the traditional TNM staging system in terms of prediction effectiveness and accuracy. As a new assessment tool, the immunoscore classifier may be helpful for determining the immune status of patients with stage I NSCLC and screening patients suitable for subsequent immunotherapy.

e21669 Background: In the new era of immunotherapy, the regimen based on bevacizumab is still one of the standard options for treatment of advanced non-small cell lung cancer (NSCLC) patients without driver mutations. However, the prognostic factors for bevacizumab are still missing. We aimed to determine the integrative value of computed tomography (CT), epigenetic modifications, clinicopathological and systemic inflammatory factors for predicting the survival of advanced NSCLC patients with bevacizumab. Methods: Clinicopathological parameters, dynamic systemic inflammatory factors, radiomics features, and DNA methylation profiling in advanced non-squamous NSCLC patients receiving first- or second- line bevacizumab plus chemotherapy were included in this study. The prognostic radiomics signature were constructed by least absolute shrinkage and selection operator (LASSO) Cox analysis. A multi-omics prediction nomogram for progression-free survival (PFS) based on radiomics, clinicopathological and systemic inflammatory features was established and independently validated. Furthermore, radiomics signature-related DNA methylation were submitted to functional enrichment analysis. Results: Total of 272 patients were included in analysis, 224 in training cohort and 48 in validation cohort. Five radiomics features, including Information Measure Corr1, Inverse Variance, Local Std Max, Gauss Area, Spherical Disproportion, were finally selected to construct radiomics signature with the AUC of 0.71. Smoking history, anatomical feature, liver metastasis, LDH4, NLR2 and radiomics signature were found to be independent prognostic factors for PFS. A multi-omics nomogram was developed based on these features in training cohort with the C-index of 0.76, and external validated with the C-index of 0.75. The tissue slices of 20 patients receiving bevacizumab were used for DNA methylation profiling. Functional enrichment analysis indicated widespread and statistically significant associations between radiomics features and DNA methylation changes which involved in several pathways related to angiogenesis and immune system, such as Notch signaling pathway, small GTPase Rho signal transduction, TGFβ signal transduction. Conclusions: This multi-omics nomogram integrating radiomics, clinicopathological, systemics inflammatory features improved the prediction of PFS in advanced NSCLC patients receiving bevacizumab. And the radiomics signature were found to be related to angiogenesis and immune status.