Background/Aims: Inaccurate prediction of lymph node metastasis (LNM) may lead to unnecessary surgery following endoscopic resection of T1 colorectal cancer (CRC). We aimed to validate the usefulness of artificial intelligence (AI) models for predicting LNM in patients with T1 CRC. Methods: We analyzed the clinical data, laboratory results, pathological reports, and endoscopic findings of patients who underwent radical surgery for T1 CRC. We developed AI models to predict LNM using four algorithms: regularized logistic regression classifier (RLRC), random forest classifier (RFC), CatBoost classifier (CBC), and the voting classifier (VC). Four histological factors and four endoscopic findings were included to develop AI models. Areas under the receiver operating characteristics curves (AUROCs) were measured to distinguish AI model performance in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines. Results: Among 1,386 patients with T1 CRC, 173 patients (12.5%) had LNM. The AUROC values of the RLRC, RFC, CBC, and VC models for LNM prediction were significantly higher (0.673, 0.640, 0.679, and 0.677, respectively) than the 0.525 suggested in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines (vs RLRC, p<0.001; vs RFC, p=0.001; vs CBC, p<0.001; vs VC, p<0.001). The AUROC value was similar between T1 colon versus T1 rectal cancers (0.718 vs 0.615, p=0.700). The AUROC value was also similar between the initial endoscopic resection and initial surgery groups (0.581 vs 0.746, p=0.845). Conclusions AI models trained on the basis of endoscopic findings and pathological features performed well in predicting LNM in patients with T1 CRC regardless of tumor location and initial treatment method.

Background/Aims: Inaccurate prediction of lymph node metastasis (LNM) may lead to unnecessary surgery following endoscopic resection of T1 colorectal cancer (CRC). We aimed to validate the usefulness of artificial intelligence (AI) models for predicting LNM in patients with T1 CRC. Methods: We analyzed the clinical data, laboratory results, pathological reports, and endoscopic findings of patients who underwent radical surgery for T1 CRC. We developed AI models to predict LNM using four algorithms: regularized logistic regression classifier (RLRC), random forest classifier (RFC), CatBoost classifier (CBC), and the voting classifier (VC). Four histological factors and four endoscopic findings were included to develop AI models. Areas under the receiver operating characteristics curves (AUROCs) were measured to distinguish AI model performance in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines. Results: Among 1,386 patients with T1 CRC, 173 patients (12.5%) had LNM. The AUROC values of the RLRC, RFC, CBC, and VC models for LNM prediction were significantly higher (0.673, 0.640, 0.679, and 0.677, respectively) than the 0.525 suggested in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines (vs RLRC, p<0.001; vs RFC, p=0.001; vs CBC, p<0.001; vs VC, p<0.001). The AUROC value was similar between T1 colon versus T1 rectal cancers (0.718 vs 0.615, p=0.700). The AUROC value was also similar between the initial endoscopic resection and initial surgery groups (0.581 vs 0.746, p=0.845). Conclusions AI models trained on the basis of endoscopic findings and pathological features performed well in predicting LNM in patients with T1 CRC regardless of tumor location and initial treatment method.

Background/Aims: Inaccurate prediction of lymph node metastasis (LNM) may lead to unnecessary surgery following endoscopic resection of T1 colorectal cancer (CRC). We aimed to validate the usefulness of artificial intelligence (AI) models for predicting LNM in patients with T1 CRC. Methods: We analyzed the clinical data, laboratory results, pathological reports, and endoscopic findings of patients who underwent radical surgery for T1 CRC. We developed AI models to predict LNM using four algorithms: regularized logistic regression classifier (RLRC), random forest classifier (RFC), CatBoost classifier (CBC), and the voting classifier (VC). Four histological factors and four endoscopic findings were included to develop AI models. Areas under the receiver operating characteristics curves (AUROCs) were measured to distinguish AI model performance in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines. Results: Among 1,386 patients with T1 CRC, 173 patients (12.5%) had LNM. The AUROC values of the RLRC, RFC, CBC, and VC models for LNM prediction were significantly higher (0.673, 0.640, 0.679, and 0.677, respectively) than the 0.525 suggested in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines (vs RLRC, p<0.001; vs RFC, p=0.001; vs CBC, p<0.001; vs VC, p<0.001). The AUROC value was similar between T1 colon versus T1 rectal cancers (0.718 vs 0.615, p=0.700). The AUROC value was also similar between the initial endoscopic resection and initial surgery groups (0.581 vs 0.746, p=0.845). Conclusions AI models trained on the basis of endoscopic findings and pathological features performed well in predicting LNM in patients with T1 CRC regardless of tumor location and initial treatment method.

Background/Aims: Inaccurate prediction of lymph node metastasis (LNM) may lead to unnecessary surgery following endoscopic resection of T1 colorectal cancer (CRC). We aimed to validate the usefulness of artificial intelligence (AI) models for predicting LNM in patients with T1 CRC. Methods: We analyzed the clinical data, laboratory results, pathological reports, and endoscopic findings of patients who underwent radical surgery for T1 CRC. We developed AI models to predict LNM using four algorithms: regularized logistic regression classifier (RLRC), random forest classifier (RFC), CatBoost classifier (CBC), and the voting classifier (VC). Four histological factors and four endoscopic findings were included to develop AI models. Areas under the receiver operating characteristics curves (AUROCs) were measured to distinguish AI model performance in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines. Results: Among 1,386 patients with T1 CRC, 173 patients (12.5%) had LNM. The AUROC values of the RLRC, RFC, CBC, and VC models for LNM prediction were significantly higher (0.673, 0.640, 0.679, and 0.677, respectively) than the 0.525 suggested in accordance with the Japanese Society for Cancer of the Colon and Rectum guidelines (vs RLRC, p<0.001; vs RFC, p=0.001; vs CBC, p<0.001; vs VC, p<0.001). The AUROC value was similar between T1 colon versus T1 rectal cancers (0.718 vs 0.615, p=0.700). The AUROC value was also similar between the initial endoscopic resection and initial surgery groups (0.581 vs 0.746, p=0.845). Conclusions AI models trained on the basis of endoscopic findings and pathological features performed well in predicting LNM in patients with T1 CRC regardless of tumor location and initial treatment method.

Purpose: Programmed death-1 blockade with pembrolizumab has shown promising activity in relapsed/refractory (R/R) extranodal natural killer/T-cell lymphoma (NKTCL), but studies are limited, with small patient numbers. Materials and Methods: Thirteen institutes involved with the Consortium for Improving Survival of Lymphoma, a Korean lymphoma study group, collected the clinical data of 59 patients treated with pembrolizumab as salvage therapy between 2016 and 2022. Results: The median age of the patients was 60 years (range, 22 to 87 years), and 76.3% had advanced Ann Abor stage disease. Pembrolizumab was given to 35.6%, 40.7%, and 23.7% of the patients as second-, third-, and fourth- or higher-line chemotherapy, respectively. The overall response rate was 40.7%, with 28.8% having complete response. The estimated 2-year progression-free survival (PFS) and overall survival rates for all patients were 21.5% and 28.7%, respectively; for responders, the rates were 53.0% and 60.7%, respectively. Although not statistically significant, Eastern Cooperative Oncology Group performance status ≥ 2 (hazard ratio [HR], 1.91; 95% confidence interval [95% CI], 0.93 to 3.94; p=0.078) and stage III or IV disease (HR, 2.59; 95% CI, 0.96 to 6.96; p=0.060) were associated with a trend toward shorter PFS in multivariate analysis. Grade 3 or 4 adverse events (AEs) were noted in 12 patients (20.3%); neutropenia (10.2%), fatigue (6.8%), and pneumonitis (5.1%) were most common AEs. Conclusion In conclusion, while pembrolizumab had a modest effect on patients with R/R NKTCL, it may be a useful salvage therapy for patients with localized disease and good performance status.

Purpose: Diffuse large B-cell lymphoma (DLBCL) is the most common hematologic malignancy worldwide. Although substantial improvement has been achieved by the frontline rituximab-based chemoimmunotherapy, up to 40%-50% of patients will eventually have relapsed or refractory disease, whose prognosis is extremely dismal. Materials and Methods: We have carried out two prospective cohort studies that include over 1,500 DLBCL patients treated with rituximab plus CHOP (#NCT01202448 and #NCT02474550). In the current report, we describe the outcomes of refractory DLBCL patients. Patients were defined to have refractory DLBCL if they met one of the followings, not achieving at least partial response after 4 or more cycles of R-CHOP; not achieving at least partial response after 2 or more cycles of salvage therapy; progressive disease within 12 months after autologous stem cell transplantation. Results: Among 1,581 patients, a total of 260 patients met the criteria for the refractory disease after a median time to progression of 9.1 months. The objective response rate of salvage treatment was 26.4%, and the complete response rate was 9.6%. The median overall survival (OS) was 7.5 months (95% confidence interval, 6.4 to 8.6), and the 2-year survival rate was 22.1%±2.8%. The median OS for each refractory category was not significantly different (p=0.529). Conclusion In line with the previous studies, the outcomes of refractory DLBCL patients were extremely poor, which necessitates novel approaches for this population.

Purpose: We conducted a nationwide, multicenter, prospective registry study for newly diagnosed patients with peripheral T-cell lymphoma (PTCL) to better define the clinical characteristics, treatment patterns, survival outcomes, and the role of upfront autologous stem cell transplantation (ASCT) in these patients. Materials and Methods: Patients with PTCL receiving chemotherapy with curative intent were registered and prospectively monitored. All patients were pathologically diagnosed with PTCL. Results: A total of 191 patients with PTCL were enrolled in this prospective registry study. PTCL, not otherwise specified (PTCL-NOS) was the most common pathologic subtype (n=80, 41.9%), followed by angioimmunoblastic T-cell lymphoma (AITL) (n=60, 31.4%). With a median follow-up duration of 3.9 years, the 3-year progression-free survival (PFS) and overall survival (OS) rates were 39.5% and 60.4%, respectively. The role of upfront ASCT was evaluated in patients who were considered transplant-eligible (n=59). ASCT was performed as an upfront consolidative treatment in 32 (54.2%) of these patients. There were no significant differences in PFS and OS between the ASCT and non-ASCT groups for all patients (n=59) and for patients with PTCL-NOS (n=26). However, in patients with AITL, the ASCT group was associated with significantly better PFS than the non-ASCT group, although there was no significant difference in OS. Conclusion The current study demonstrated that the survival outcomes with the current treatment options remain poor for patients with PTCL-NOS. Upfront ASCT may provide a survival benefit for patients with AITL, but not PTCL-NOS.

Background: Brentuximab vedotin (BV), a potent antibody-drug conjugate, targets the CD30 antigen.In Korea, BV has been approved for the treatment of relapsed or refractory Hodgkin lymphoma (HL), anaplastic large-cell lymphoma (ALCL), and cutaneous T-cell lymphomas, including mycosis fungoides (MF). However, there are limited data reflecting real-world experiences with BV treatment for HL, ALCL, and MF. Methods: This was a multicenter, non-interventional registry study of the efficacy and safety of BV in patients with relapsed or refractory CD30-positive lymphoma (CISL1803/BRAVO).Outcomes were determined based on the occurrence of relapse or progression and overall survival after BV treatment. Results: A total of 85 patients were enrolled in this study. The median number of BV cycles was 10 (range, 2‒16) in the patients with HL. The objective response rate (ORR) of patients with HL to BV was 85.4% (41/48), comprising 27 complete responses (CRs) and 14 partial responses (PRs). The ORR of ALCL was 88% (22/25), consisting of 17 CRs and five PRs, whereas the ORR of MF was 92% (11/12). At the median follow-up of 44.6 months after BV treatment, the median post-BV progression-free survival of HL, ALCL, and MF patients was 23.6 months, 29.0 months, and 16.7 months, respectively (P =0.641). The most common side effect of BV was peripheral neuropathy; 22 patients (25.9%, 22/85) experienced peripheral neuropathy (all grades). Conclusion The treatment outcomes of patients with relapsed or refractory CD30-positive lymphoma improved with BV treatment, and the safety profile was manageable.

Objective: Peritonitis is a life-threatening, emergent surgical disease with very high mortality and morbidity. Currently, there are insufficient Korean studies using the P-POSSUM (Portsmouth-Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity) and the Mannheim Peritonitis Index (MPI) as risk prediction models for nontraumatic peritonitis patients who visit the emergency room. Methods: This retrospective study was carried out on 196 cases of non-traumatic peritonitis in a single emergency center from January 2015 to December 2019. Receiver operating characteristic (ROC) curves were obtained and the area under the ROC curve (AUC) was compared using both P-POSSUM and MPI. The observed mortality and expected mortality for P-POSSUM were compared using the goodness of fit assessed using the Hosmer-Lemeshow equation. Results: Diastolic blood pressure, blood urea nitrogen, potassium, length of stay, and intensive care unit admissions were significantly different between survivors and non-survivors. The AUC was 0.812 for P-POSSUM and 0.646 for MPI. The observed-to-expected mortality ratio for P-POSSUM indicated fewer than expected deaths in all quintiles of risk and this was more pronounced, especially when the expected mortality was over 60%. Conclusion In non-traumatic peritonitis patients, P-POSSUM was more useful in predicting risk than the MPI score. However, P-POSSUM overestimated the risk in high-risk patients. Although the MPI score is only somewhat useful for predicting mortality in patients with non-traumatic peritonitis, it is useful as an adjuvant.