Search

Simple Summary Pathology images are vital for understanding solid cancers. In this study, we created DeepRePath using multi-scale pathology images with two-channel deep learning to predict the prognosis of patients with early-stage lung adenocarcinoma (LUAD). DeepRePath demonstrated that it could predict the recurrence of early-stage LUAD with average area under the curve scores of 0.77 and 0.76 in cohort I and cohort II (external validation set), respectively. Pathological features found to be associated with a high probability of recurrence included tumor necrosis, discohesive tumor cells, and atypical nuclei. In conclusion, DeepRePath can improve the treatment modality for patients with early-stage LUAD through recurrence prediction. Abstract The prognosis of patients with lung adenocarcinoma (LUAD), especially early-stage LUAD, is dependent on clinicopathological features. However, its predictive utility is limited. In this study, we developed and trained a DeepRePath model based on a deep convolutional neural network (CNN) using multi-scale pathology images to predict the prognosis of patients with early-stage LUAD. DeepRePath was pre-trained with 1067 hematoxylin and eosin-stained whole-slide images of LUAD from the Cancer Genome Atlas. DeepRePath was further trained and validated using two separate CNNs and multi-scale pathology images of 393 resected lung cancer specimens from patients with stage I and II LUAD. Of the 393 patients, 95 patients developed recurrence after surgical resection. The DeepRePath model showed average area under the curve (AUC) scores of 0.77 and 0.76 in cohort I and cohort II (external validation set), respectively. Owing to low performance, DeepRePath cannot be used as an automated tool in a clinical setting. When gradient-weighted class activation mapping was used, DeepRePath indicated the association between atypical nuclei, discohesive tumor cells, and tumor necrosis in pathology images showing recurrence. Despite the limitations associated with a relatively small number of patients, the DeepRePath model based on CNNs with transfer learning could predict recurrence after the curative resection of early-stage LUAD using multi-scale pathology images.

To investigate the efficacy of irinotecan-based (IP) and etoposide-based (EP) platinum combinations, and of single-agent chemotherapy, for treatment of extensive-disease small cell lung cancer (ED-SCLC), we performed a large-scale, retrospective, nationwide, cohort study. The population data were extracted from the Health Insurance Review and Assessment Service of Korea database from January 1, 2008, to November 30, 2016. A total of 9,994 patients were allocated to ED-SCLC and analyzed in this study. The primary objectives were to evaluate the survival outcomes of systemic first-line treatments for ED-SCLC. For first-line treatment, patients who received IP showed a better time to first subsequent therapy (TFST) of 8.9 months (95% confidence interval [CI], 8.50–9.40) than those who received EP, who had a TFST of 6.8 months (95% CI, 6.77–6.97, P

Background Isoform 2 of the tight junction protein claudin-18 (CLDN18.2) is a potential target of gastric cancer treatment. A treatment targeting CLDN18.2 has shown promising results in gastric cancer. We investigated the clinical significance of CLDN18.2 and other cell-adherens junction molecules (Rho GTPase-activating protein [RhoGAP] and E-cadherin) in metastatic diffuse-type gastric cancer (mDGC).Methods We evaluated CLDN18.2, RhoGAP, and E-cadherin expression by performing two-plex immunofluorescence and quantitative data analysis of H-scores of 77 consecutive mDGC patients who received first-line, platinum-based chemotherapy between March 2015 and February 2017.Results CLDN18.2 and E-cadherin expression was significantly reduced in patients with peritoneal metastasis (PM) compared with those without PM at the time of diagnosis (P=0.010 and 0.013, respectively), while it was significantly higher in patients who never exhibited PM from diagnosis to death than in those who did (p=0.001 and 0.003, respectively). Meanwhile, CLDN18.2 and E-cadherin were expressed at significantly higher levels in patients with bone metastasis than in those without it (p=0.010 and 0.001, respectively). We identified a positive correlation between the expression of CLDN18.2 and E-cadherin (p < 0.001), RhoGAP and CLDN18.2 (p=0.004), and RhoGAP and E-cadherin (p=0.001). CDLN18.2, RhoGAP, and E-cadherin expression was not associated with chemotherapy response and survival.Conclusions CLDN18.2 expression was reduced in PM but significantly intact in bone metastasis. Furthermore, a positive correlation was identified between the expression of CLDN18.2 and other adherens junction molecules, which has clinical implications for mDGC and PM pathogenesis.

Accurate prediction of non-small cell lung cancer (NSCLC) prognosis after surgery remains challenging. The Cox proportional hazard (PH) model is widely used, however, there are some limitations associated with it. In this study, we developed novel neural network models called binned time survival analysis (DeepBTS) models using 30 clinico-pathological features of surgically resected NSCLC patients (training cohort, n = 1,022; external validation cohort, n = 298). We employed the root-mean-square error (in the supervised learning model, s- DeepBTS) or negative log-likelihood (in the semi-unsupervised learning model, su-DeepBTS) as the loss function. The su-DeepBTS algorithm achieved better performance (C-index = 0.7306; AUC = 0.7677) than the other models (Cox PH: C-index = 0.7048 and AUC = 0.7390; s-DeepBTS: C-index = 0.7126 and AUC = 0.7420). The top 14 features were selected using su-DeepBTS model as a selector and could distinguish the low- and high-risk groups in the training cohort (p = 1.86 × 10−11) and validation cohort (p = 1.04 × 10−10). When trained with the optimal feature set for each model, the su-DeepBTS model could predict the prognoses of NSCLC better than the traditional model, especially in stage I patients. Follow-up studies using combined radiological, pathological imaging, and genomic data to enhance the performance of our model are ongoing.

Gastrointestinal stromal tumors (GISTs) are the most common mesenchymal tumors of the gastrointestinal tract. A 77-year-old man was referred for the evaluation of general weakness and leukocytosis. Computed tomography showed a 9.5×6.5-cm cavitary lesion with an air-fluid level near the stomach, which was thought to be a perigastric abscess. Upper endoscopy revealed a fistula on the greater curvature at the mid body of the stomach. The margin of the fistula opening was clearly demarcated, and yellow turbid fluid oozing from the fistula was seen. Laparoscopic wedge resection was performed at the perforated area of the stomach. Immunohistochemistry revealed CD117 expression. A diagnosis of intermediate-risk GIST was made. No recurrence was identified within 18 months after the operation. The final diagnosis was perforated gastric GIST communicating with the gastric lumen and presenting as an intra-abdominal abscess.

e16172 Background: Despite recent advancements in the understanding of the molecular biology of biliary tract cancer (BTC), target therapy and immunotherapy have demonstrated only limited efficacy, with cytotoxic systemic therapy still being the most effective treatment in BTC, except for surgery. Thus, this study aimed to analyze the role of DKK1 or β-catenin as a prognostic factor in BTC and determine the clinical association of ß-catenin and DKK1 with CD8+ tumor-infiltrating lymphocyte (TIL). Methods: We used data in The Cancer Genome Atlas (TCGA) Research Network and the clinicopathological data of 145 patients with BTC who had undergone primary radical resection between 2006 and 2016. Immunohistochemistry was performed on formalin-fixed, paraffin-embedded tissue sections, and whole tissue sections of representative tumor samples were used for antigen retrieval. Results: CD8+TIL expression was a significant predictor of favorable overall survival (OS) and recurrence-free survival (RFS) (median OS, 34.9months in TIL-high, 16.7months in TIL-low, P < 0.0001 respectively; median RFS, 27.1months in TIL-high, 10months in TIL-low, P < 0.0001 respectively). Positive ß-catenin expression and high DKK1 expression was also associated with a shorter OS (median OS, 23.95months in positive ß-catenin, 26.1months in negative ß-catenin, P = 0.1009 respectively; median OS, 19.4months in high DKK1, 31.65months in Low DKK1, P = 0.0093 respectively), but not RFS (p = 0.1466, at ß-catenin respectively; p = 0.2924, in DKK1 respectively). In the CD8+TIL-high BTC group, the tumor expression of β-catenin and DKK1 had a significant negative impact on either OS or RFS (p = 0.0146 and p = 0.0112, at ß-catenin respectively; p = 0.0950 and p = 0.3904, in DKK1 respectively). However, in the TIL-low BTC group, there were no differences in OS or RFS according to ß-catenin and DKK1 expression (p = 0.5108 and p = 0.8431, at ß-catenin respectively; p = 0.1127 and p = 0.1095, in DKK1 respectively). Cox regression multivariate analysis demonstrated that CD8+ TIL (hazard ratio [HR], 0.490; 95% confidence interval (CI), 0.303-0.791; p = 0.004) and β-catenin (HR, 1.652; 95% CI, 1.035-2.639; p = 0.036) retained significant association with OS after adjustment for all variables. Conclusions: Among patients with resected BTC, β-catenin and DKK1 protein levels are associated with poor clinical outcomes, whereas high CD8+ TIL levels are associated with good clinical outcomes. This confirms the differential clinical role of Wnt/β-catenin and DKK1 proteins according to TIL expression in BTC.

Introduction: Accurate prediction of non-small cell lung cancer (NSCLC) prognosis after surgery is still challenging. Cox proportional hazard model is widely used, but there are some limitations like proportional hazard assumption, computational complexity, and unspecified time component. In this study, we developed the novel neural network models using 30 clinicopathological features of surgically resected NSCLC patients. Materials and Methods: We analyzed a dataset of training cohorts of 1,034 patients and external validation cohort of 303 patients, and compared the novel neural network model and Cox PH model to predict 3-year recurrence of the patient. Results: We developed deep neural network for binned time survival analysis (DeepBTS) algorithm, which use root-mean-squared error (supervised learning model, s- DeepBTS) or negative log-likelihood (semi-unsupervised learning model, su- DeepBTS) as loss function. The su-DeepBTS algorithm showed the improved performance (C-index, 0.7166; AUC, 0.7606) compared to other models (Cox PH: 0.6861 and 0.7185; s-DeepBTS: 0.6926 and 0.7250, respectively). Top 10 features was selected from su-DeepBTS model and su-DeepBTS selector pair: the number of lymph node metastasis, tumor size, histologic subtype, vascular invasion, R0 resection, neoadjuvant /adjuvant treatment, and performance status of patients. These features distinguished low and high-risk group in the training cohort (p = 4.25×10-5) and the validation cohort (p = 1.07×10-11). Conclusions: The su-DeepBTS model using 10 selected features can predict the prognosis of resected NSCLC patients. We plan to diversify the input features by radiologic, pathological imaging and genomic data to integrate multimodal medical information and enhance the performance of this model using deep neural network. Citation Format: Ho Jung An, Bora Lee, Sang Hoon Chun, Ji Hyung Hong, In Sook Woo, Seoree Kim, Joon Won Jeong, Jae Jun Kim, Hyun Woo Lee, Sungsoo Park, Heejoon Jo, Yoon Ho Ko. DeepBTS: Prediction of recurrence-free survival of non-small cell lung cancer using time-binned deep neural network [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2091.

Introduction Histopathological images in high-powered contain quantitative information for tumor cell morphology, but humans could not sufficiently extract prognostic features of tumor cell morphology associated with tumor recurrence of patient. In this study, we developed and trained the DeepRePath model, a deep convolutional neural network (CNN), on histopathological images to predict the recurrence of resected lung adenocarcinoma (LUAD). Methods A total of 244 haematoxylin and eosin stained histopathology slides of resected LUAD patients were used to train and validate the DeepRePath model. We manually segmented a cancer region in a LUAD tissue slide and captured to an image. All images were captured at x400 magnification and split into 13,329 image patches. We trained the DeepRePath model by transfer learning using 13,329 image patches from 244 LUAD patients. Results 95 out of 244 patients developed recurrence after surgical resection. The DeepRePath model showed an average area under the curve (AUC) of 0.78 in 5-fold cross validation on the training cohort. We plan to employ external validation to evaluate the proposed model. In addition, using gradient-weighted class activation mapping (Grad-CAM), we confirmed that DeepRePath identified and used individual tumor cells with atypia to predict the recurrence of the patient. Conclusions DeepRePath model using transfer learning and CNN can accurately predict recurrence after curative resection of LUAD based on atypia of tumor cell. Citation Format: Yoon Ho Ko, Soon Auck Hong, Tae-Jung Kim, Ji Hyung Hong, Sang Hoon Chun, Seoree Kim, Ho Jung An, Sungsoo Park, Won Sang Shim. DeepRePath: Finding prognostic features of tumor cell on the histopathologic image of lung cancer using explainable deep convolutional neural network [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 2093.

Introduction: To retrospectively assess hyperprogressive disease(HPD) in patients who underwent immune checkpoint blockade therapy and explore the heterogenous nature of tumor microenvironment of hyperprogressive patients using multiplex immunohistochemistry. Method: We retrospectively reviewed medical records of non-small cell lung cancer patients (n=243) treated with anti-PD-1 or anti-PD-L1 monotherapy at 5 institutes of St. Mary's Hospitals between January 2014 and May 2018. HPD was defined as a tumor growth kinetics ratio (TGKr) exceeding two during anti-PD-1 / PD-L1 therapy and a time-to-failure (TTF) of less than 2 months. We also analyzed the association of Neutrophil-lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR) and CRP-albumin ratio (CAR) with HPD. To evaluate the difference of immune composition in TME by multiplex IHC(mIHC). Panel 1/2 composed of T cell markers and co-inhibitory signal markers. Panel 3/4 examined macrophages, NK cells and dendritic cells. Results: Overall, 231 patients were included. Overall, 26 patients (11.3%) met the criteria for hyperprogression. Multiplex IHX data quantitated immune subset number present intra-tumor vs the tumor stroma. there was a difference in immune cell composition between the HPD, poor responders (PRs) and good responders (GRs). High levels of CD4 + Teff cells, FOXP3 Treg cells were associated with HPD(p < 0.052, Conclusion: This study was important in investigating the HPD-related immune prognostic factors in NSCLC patients in korea. According to these data, HPD is a unique biologic process distinct from a non HPD-PR. Multiplex IHC can be used not only to predict HPD, but also to understand the dynamicity and the complexed immune compositional change within TME. Furthermore, based on these findings, we are undergoing study the TME by serially collected pre- and post-treatment tissue samples, it may be helpful to shed light on the mechanisms of HPD. Citation Format: Seoree Kim, Sang Hoon Chun, Sang-Yeob Kim, Junyoung Seo, Chan Kwon Jung, Jinhyoung Kang. Multiplex immunohistochemistry accurately defines the immune compositional change of tumor microenvironment to predict hyperprogressive disease [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3876.

e21096 Background: Small cell lung cancer (SCLC) is a highly proliferative and rapidly growing tumor with poor prognosis, even in limited disease (LD), and a timely and accurately high intensity therapy is necessary. During the concurrent chemoradiotherapy (CCRT), etoposide with platinum (EP) regimen is recommended, however, irinotecan with cisplatin (IP) regimen is also effective. Therefore, we performed a large-scale, retrospective and nationwide cohort study to analyze the efficacy of CCRT in patients with LD-SCLC Methods: The population data were extracted from the Health Insurance Review and Assessment Service of Korea database from January 1, 2008, to November 30, 2016. Among 14,490 patients with SCLC patients who received chemotherapy, a total of 4,496 patients with LD-SCLC were analyzed. The primary objectives were to evaluate the survival outcomes of CCRT for LD-SCLC. Results: Patients who received EP-CCRT (n = 4,187) had better PFS of 11.233 months (95% confidence interval (CI) 10.900 – 11.667, P = 0.0477) and prolonged OS of 22.233 months (95% CI 21.233 – 23.333, P < 0.0001) than those who received IP-CCRT (n = 259; PFS = 9.567 months, 95% CI = 8.500 – 10.667; OS = 16.433 months, 95% CI = 14.467 – 18.333; P = 0.0477). After the failure of CCRT, we observed that dual chemotherapy (n = 925; OS = 9.133 months, 95% CI = 8.400 - 10.100) has better survival benefit (P < 0.0001) than single agent chemotherapy (n = 815; OS = 7.500 months, 95% CI, 6.933 – 8.133). IP chemotherapy demonstrated a better OS (9.567 months, 95% CI, 8.667 – 10.333 vs. 7.100 months, 95% CI, 5.100 – 10.400, P = 0.0170, respectively) than EP regimen in the platinum-resistant group; patients with PFS within 6 months after CCRT. However, in platinum sensitive group; patients who recurred after 6 months, the clinical benefit of EP regimen was superior to IP regimen (OS = 17.183 months, 95% CI, 13.033 – 25.633 vs. 6.617 months, 95% CI, 5.433 – 7.767; P < 0.0001). Cox proportional hazards regression analysis demonstrated that age, EP-CCRT, and dyslipidemia retained significant associations with OS after adjusting for all variables. Conclusions: In Korean population, concurrent thoracic radiotherapy with EP regimen had significantly favorable effects on OS and PFS. In addition, after the failure of CCRT, combination chemotherapy had clinical benefits over single agents. Among combination chemotherapy, IP combination regimen showed significantly favorable effects on OS and PFS, especially in the platinum-resistant group.

Introduction Although immune checkpoint blockades (ICBs) therapy can lead to favorable and durable results by reinvigorating the anti-tumor immune response in some patients, many other patients experience poor prognosis and even tumor overgrowth can be seen in real practice. So, we performed retrospectively hyperprogression-related studies on patients who underwent immunotherapy to find the predictable clinical factor. We also revealed the difference of immune composition in TME that are associated with HPD by multiplex IHC Method We retrospectively evaluated the medical records of NSCLC patients (n=243) who treated with anti-PD-1/anti-PD-L1 therapy at 5 institutes of St. Mary’s Hospitals between January 2014 to June 2018. Results A total of 231 patients were included in the TGK analysis. Median age was 64.2 years with a majority being male (74.9%) and smokers (70.1%, n=162). Among them, 88.9% were heavy smokers with over 20 pack per years (n=144). Most of the ICB drugs were used in the second and above lines, but the proportion used in the fourth or higher heavy treated patients was 16%(n=37). The rate of over-growth was 33.3% (n=77) among these patients, the rate of hyperprogression was 32.4% (n=25). The proportion of patients harboring oncogenic driver mutation such as EGFR alteration was 18.6%. Univariate analyses indicated that HPD was significantly associated with oncogenic drive mutant status compared with non-HPD (18.6% [8 of 43] vs 9.04% [17 of 188]; P= 0.001). No significant differences were observed according to age, number of previous lines of therapy, or presence of more than 3 metastatic sites before ICBs. Multivariate analyses with a multivariate logistic regression model demonstrated that presence of less than 3 metastatic sites (odds ratio [OR], 4.769; 95% [CI], 1.384-16.433; P < .013) was an independent predictor for HPD. We also analyzed the association of NLR, PLR and CAR with HPD. Serologic markers post 6 weeks, i.e., at the time of first response evaluation, showed prognostic value for the most significant hyperprogression after ICB use. Kaplan-Meier OS estimates showed that there was a clear trend toward worse outcome for the patients with HPD (median OS, 5.6 months;95% CI, 4.5-10.3; P < 0.001) compared with the patients with non-HPD disease progression (median OS, 7.4 months; 95% CI, 15.1-19.6; P = 0.003) The multiplex IHC results showed that the percentage of T-cell marker expression of Tumor and Stroma in each slide was different between good responder and HPD. Conclusion Even though the improved recognition of hyperprogression, the etiology of HPD remains unclear, apart from the pseudoprogression. However, we observed that some clinical factors and some serologic biomarker can be used to predict HPD. Furthermore, we undergo TMB and multiplex IHC to understand mechanism of hyperprogression. Citation Format: Seoree Kim, Sang-Yeob Kim, Jin Hyoung Kang. A new pattern called Hyperprogression when using Immune checkpoint blockers in real world [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5004.

e20633 Background: Although immune checkpoint blockades (ICBs) therapy can lead to favorable and durable results by reinvigorating the anti-tumor immune response in some patients, many other patients experience poor prognosis and even tumor overgrowth can be seen in real practice. We aimed to assess these hyperprogressive disease (HPD) in patients who underwent ICB therapy. Methods: We retrospectively reviewed medical records of non-small cell lung cancer patients (n = 243) treated with anti-PD-1 or anti-PD-L1 monotherapy. HPD was defined as a tumor growth kinetics ratio > 2 during anti-PD-1/PD-L1 therapy and a time-to-failure of less than 2 months. We analyzed the association of Neutrophil-lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR) and CRP-albumin ratio (CAR) with HPD and the difference of immune composition in TME by multiplex IHC. Results: Overall, 231 patients were included. The median age was 64.2 years; most patients were male (74.9%) and smokers (70.1%, n = 162). 25 patients (10.8%) met the criteria for HPD. Oncogenic drive mutant status was significantly associated with HPD (18.6% vs 9.04%; P = 0.001). Kaplan Meier overall survival estimates showed a clear trend toward worse outcomes for patients with HPD (median, 5.6 months; P < 0.001) than for those without (median, 7.4 months). We also analyzed the association of NLR, PLR and CAR with HPD. Serologic markers at the time of first response evaluation (post 6 weeks, i.e.,) showed a prognostic value for HPD after ICB use ( P < 0.001, 0.008, 0.017). In multiplex IHC, there was no difference in T cell related immune marker of CD 3/4/8/45RO and FOXP3, dendritic cell, NK cell and other co-inhibitory signal markers between two groups. However, the key point was the M2 polarized tendency in the HPD group when the macrophage M1 / 2 markers of CD68, CD163, CD206 and CCR7 were observed. This suggests that changes in the AXL pathway and EMT features associated with reinvigorating anti-tumor immunity in M2 polarized cells may lead to changes in poor prognosis to HPD. Conclusions: Despite improved recognition of HPD, its etiology and predisposing factors remain unclear, but it is clear that the prognosis becomes poor when it occurs. We observed that some serologic biomarkers (NLR, PLR and CAR. i.e.,) can be used to predict HPD. Multiplex IHC can be used not only to predict HPD, but also to validate its mechanism. Furthermore, we undergo whole exon sequencing and multiplex IHC to understand mechanism of Hyperprogression.

Ovarian tumors are reported in 1 of every 200 pregnancies and may cause serious problems such as torsion, infarction, and obstruction of vaginal delivery. These tumors have typically been removed by elective surgery, preferably in the second trimester. Because abdominal surgery significantly stresses both the mother and fetus, this study was planned to determine the optimal management of ovarian tumors in 89 pregnant women with tumors requiring surgery. Surgery was emergent because of torsion of the tumor in 36 cases (group A) and elective in 53 cases (group B). The 2 groups were similar in maternal age, parity, and gestational age at the time of surgery. Emergency surgery was done in the first, second, and third trimesters in 22, 5, and 9 women, respectively. Elective surgery was done in the second trimester in two thirds of cases and in the first trimester in the others. Mean birth weights were similar for the 2 groups, but preterm births were significantly more frequent in group A (22% vs. 4%). There was no group difference in gestational age at preterm delivery. Tumors were of comparable size in the 2 groups, and there was no difference in their location. Most tumors were 6 to 10 cm in size. Nearly 10% of women in both groups had bilateral ovarian tumors. There were no perinatal deaths, and no difference was noted in rates of vaginal or cesarean delivery. Dermoid cyst was the most common histologic diagnosis, accounting for 36% of group A and 45% of group B cases. Two group A tumors but none of those in group B were malignant. In this series, the risk of adverse pregnancy outcomes was not greater when surgery was delayed until symptoms of ovarian tumor developed rather than being done electively. The authors favor a conservative approach to pregnant women who have an ovarian tumor.