Introduction

Accurate prediction of lymph node metastasis risk after endoscopic resection in T1 colorectal cancer (CRC) is essential for determining the treatment strategy since lymph node metastasis occurs in only 10%. Although some histopathological valuables (e.g., lymphovascular invasion or tumor differentiation) or scoring systems for predicting lymph node metastasis have been reported, problems of low inter-observer agreement and evaluator error remain unsolved. To address these issues, we developed and evaluated a novel artificial intelligence (AI) system using whole slide images (WSI) to predict lymph node metastasis in T1 CRC.

Aims & Methods

For the development of the AI, 462 T1 CRCs (including 65 metastasis-positive cases) surgically resected from 2001-2018 were used as training cases. To evaluate the accuracy of AI, 100 consecutive T1 CRCs surgically resected in 2018-2021 were used as test cases. Using Nanozoomer (Hamamatsu Photonics, Hamamatsu, Japan), pathological slides under Hematoxylin-eosin staining were converted to WSI (x40 magnification). Each WSI was divided into small images of 224 x 224 pixels and trained using weakly supervised learning called multiple instance learning (MIL). MIL outputs the risk score (MIL score) for each case using one WSI, and then the data were trained using random forest with the addition of two factors, patient’s sex and tumor location. The area under the curve (AUC) of the receiver operating characteristic curve was calculated and compared in AI model and current Japanese guideline.

Results

The incidence of lymph node metastasis in the test cases was 15% (15/100). Compared to the guideline, the AI (AUC=0.73) demonstrated superior accuracy in predicting lymph node metastasis than the guideline (AUC=0.52). When the sensitivity of both tests was set as 100%, the specificity was found to be 4.7% for the guideline and 35% for the AI. Consequently, the AI could theoretically reduce 26% of over-surgery cases without missing any metastasis-positive cases.

Conclusion

We proposed a new AI system using WSI-based T1 CRC, which can objectively predict the risk of lymph node metastasis. The utilization of this AI system has the potential to minimize the need for unnecessary surgical resection in patients with T1 CRC who have low risk.

Introduction

Ulcerative colitis (UC) is a chronic inflammatory bowel disease affecting 5 million people worldwide, characterized by mucosal inflammation that can extend throughout the colon [1]. Current treatments often have limited efficacy or severe side effects [2]. As, there are no established criteria for selecting the most effective treatment for IBD, innovative preclinical drug testing systems are of major importance.

Aims & Methods

The aim of this study was to evaluate the effects of biological therapy using ulcerative colitis patient-derived personalized co-culture models of colonic epithelial and immune cells under inflammatory conditions.

The effects of upadacitinib (UPA) and ozanimod (OZA) were tested in personalized co-cultures of colonic epithelial cells and macrophages derived from UC patients (n=12) and control (Con) individuals (n=11). Inflammation was induced with inflammatory cytokines TNF-α (10 ng/mL) and IFN-γ (10 ng/mL). Barrier integrity was measured by transepithelial electrical resistance (TEER). Expression of TNF-α, CCL5, and CXCL9 genes was evaluated using qRT-PCR. Secretion of cytokines in the epithelial cells and macrophages culture media was investigated using a flow cytometry-based target protein assay. Data comparison was conducted using the Shapiro-Wilk test and the Wilcoxon rank-sum test. A difference between the values was considered significant when p ≤ 0.05 or padj. ≤ 0.05.

Results

In cell cultures stimulated with inflammatory cytokines, barrier permeability increased (TEER decreased – Con: by 452.03 Ω·cm², UC: 39.3 Ω·cm²), which partially recovered after removal of inflammatory factors (TEER increased – Con: by 65.91 Ω·cm2, UC – only 47.31 Ω·cm²). Epithelial integrity was restored after incubation with UPA and OZA (TEER increased – Con: by 116.82 Ω·cm² (UPA), by 193.77 Ω·cm2 (OZA); UC: by 226.64 Ω·cm2 (UPA), by 188.5 Ω·cm² (OZA)). At the molecular level, overnight stimulation with TNF-α and IFN-γ increased expression of CXCL9 gene in epithelial cells (UC: fold change (FC) = 3064.49 p_adj. = 2.07 x 10^-5; Con: FC = 664.44, p_adj. = 2.00 x 10^-2) and macrophages (UC: FC = 140.47 p_adj. = 0.01; Con: FC = 314.43, p_adj. = 0.01) in both groups. Co-culture treatment with UPA and OZA showed important trend of patient-specific responses to each therapeutic agent, which probably was the most important factor of non-significant values for observed changes in CCL5, CXCL9, TNF-α expression analysing summarized data. The flow cytometry-based target protein assay revealed that TNF-α and IFN-γ can influence inflammatory cytokine secretion in both epithelial cells and macrophages of UC patients and Control individuals, which can be further modified by novel biological therapy agents – UPA and OZA.

Conclusion

In conclusion, treatment with TNF-α and IFN-γ reduced epithelial barrier integrity and induced increased expression of inflammatory cytokines in epithelial cells and macrophages. Biological therapy (UPA or OZA) tended to restore disrupted barrier integrity. However, the response to novel biological therapy agents (UPA and OZA) on gene expression in both cell types and inflammatory cytokine levels in the cell culture media showed varying impacts and effectiveness, thereby revealing individualized responses of subjects to different biological preparations.

References

1. Le Berre C, Honap S, Peyrin-Biroulet L. Ulcerative colitis. Lancet. 2023 Aug 12;402(10401):571-584. doi: 10.1016/S0140-6736(23)00966-2. PMID: 37573077.
2. Singh S, Allegretti JR, Siddique SM, Terdiman JP. AGA Technical Review on the Management of Moderate to Severe Ulcerative Colitis. Gastroenterology. 2020 Apr;158(5):1465-1496.e17. doi: 10.1053/j.gastro.2020.01.007. Epub 2020 Jan 13. PMID: 31945351; PMCID: PMC7117094.

Disclosure

No conflict of interest to declare.

Introduction

Therapeutic management of acute biliary pancreatitis (ABP) includes endoscopic sphincterotomy and cholecystectomy. However, the optimal feeding strategy prior to cholecystectomy—early oral intake or exclusive enteral nutrition—remains debated, particularly given the risk of recurrent biliary events (BEs) before surgery, especially when surgery is delayed.

Aims & Methods

The primary objective was to evaluate the likehood of BEs (biliary colic/migration, stone impaction, cholangitis, recurrent acute pancreatitis, and acute cholecystitis) prior to cholecystectomy, comparing feeding methods: oral vs. exclusive enteral nutrition. The secondary objective was to evaluate cholecystectomy rates during the initial hospitalization and the reasons for non-performance. Data were collected during hospitalization, at 3 months, and at 6 months if surgery had not been performed by month 3. BEs and feeding modalities at those times were documented. Analyses included Cox models and Chi-square tests.

Results

From October 2022 to December 2023, 1,172 patients with ABP were prospectively enrolled by 60 French hospitals, with 1,167 included in the final analysis. Sixty patients (5%) were lost to follow-up, and 34 deaths (3%) were reported. Severe ABP was observed in 27.2% (n=317) of patients. After 48–72 hours, oral feeding was resumed in 62% (n=724), enteral feeding in 16% (n=192), and fasting/parenteral feeding in 21% (n=249). During follow-up, 236 BEs occurred in 172 patients, meaning 15.5% experienced at least one event, with a median onset time of 26 days (range: 9–57 days). Specifically, 6% had at least one episode of biliary colic/migration, 2.1% experienced stone impaction, 2.3% developed cholangitis, 6.1% had recurrent AP (9% of which were severe), and 1.6% developed acute cholecystitis.
Univariate analysis identified several factors associated with a higher risk of BE recurrence (p < 0.10): age ≥52 years (p=0.093), obesity (BMI ≥30, p=0.032), ASA score >2 (p=0.097), total bilirubin ≥60 µmol/L (p=0.010), CRP >150 mg/L, diagnostic delay ≥21 days (p=0.009), and prior sphincterotomy (p=0.136). No difference in BE incidence was observed between severe and non-severe ABP (p=0.497) when analyzing the first 172 BEs.
Over the full follow-up period, despite the 87 BEs analyzed and potential biases, there remained no significant association between nutrition type (oral vs. exclusive enteral nutrition ) and the likelihood of BE occurrence (p=0.447).
In patients with non-severe ABP, cholecystectomy was performed during the initial hospitalization in only 33% of cases. The main reasons for non-performance were structural (e.g., lack of availability or surgeon refusal), accounting for 45% of cases.

Conclusion

BEs are frequent and clinically relevant: 15.5% of patients experienced at least one before surgery, with 6% developing recurrent ABP (9% of which were severe). Cholecystectomy remains underutilized; only one-third of patients with non-severe ABP underwent it during their initial stay, often due to modifiable reasons. Neither ABP severity nor nutritional type-oral vs enteral nutrition (despite potential biases, missing data, and ongoing analysis)-appeared to be associated with the occurrence of BEs prior to surgery. Early cholecystectomy, when possible, remains the only reliable strategy for preventing BEs.

Disclosure

This study received a research grant from Fresenius Kabi.