Introduction

The PAN-PROMISE symptom score is the first patient-reported outcome (PROM) in acute pancreatitis (AP), and it was developed and validated in a prospective cohort of patients to be used as an endpoint in research. The aim of this study was to assess the performance of the score in a large randomized controlled trial (RCT) and its association with well-established AP endpoints.

Aims & Methods

This is an ancillary study of the WATERFALL trial, where PAN-PROMISE was evaluated at baseline, 24, 48 and 72 hours. The study examined the association between PAN-PROMISE with stablished endpoints in AP: severity, pancreatic and/or peripancreatic fat necrosis (necrosis), infected necrosis, intensive care unit (ICU) admission, persistent organ failure (POF), prolonged hospital stay and mortality. Areas under the ROC curve (AUC) were calculated and used to compare baseline PAN-PROMISE and baseline Bedside Index for Severity in AP (BISAP) for the prediction of these endpoints.

Results

248 patients from the WATERFALL trial were included. A statistically significant association was found between PAN-PROMISE and severity, necrosis, POF, ICU admission, and prolonged hospital stay at all checkpoints (p<0.05). Higher scores were also significantly associated with infected pancreatic necrosis at 24, 48, and 72h and death at baseline and 24 hours. PAN-PROMISE baseline score had a higher AUC than BISAP for severity and necrosis, but the results were not statistically significant.

Conclusion

In the context of a RCT, the PAN-PROMISE score, a patient-centered measure, has been validated with established AP outcomes and as an endpoint for future clinical trials.

Introduction

Acute pancreatitis (AP) is one of the most common and widely increasing gastrointestinal diseases leading to hospitalization without specifically available therapy [1]. It is mainly caused by gallstones, possibly by increasing the pressure in pancreatic duct with refluxing of bile acids (BAs) or changing of serum BA concentration due to bile duct obstruction [2]. So far, few existing reports only mentioned total BA measurements without considering their hydrophilic potential. Our data from experimental work showed that systemic BAs can modulate severity of AP depending on their hydrophobicity and the involvement of cholecystokinin [3]. As a further translational step, we aimed to investigate the role of systemic BAs profiles and analysed their association with the etiology and severity of AP in human.

Aims & Methods

Serum from AP patients and a healthy control group was collected and stored at -80⁰C. A profile of 20 different BAs was quantified by liquid chromatography tandem mass spectrometry-based assay using the AbsoluteIDQBAs kit according to the manufacturer’s instruction. The workflow was managed by MetIDQ and the data was analyzed by Oracle database XE software. Acute Physiology and Chronic Health Evaluation (APACHE) II scores were calculated for each patient at the first and third day of hospitalization and were compared by subtraction of the values (Δ APACHE II). Diagnosis criteria and severity of pancreatitis were assessed according to the Revised Atlanta classification and the APACHE II score.

Results

There were 214 AP patients with mean age of 60±16 (range 26-87) years and 50.4 % were female. Biliary origin was diagnosed in 133 (62.1%) patients and non-biliary pancreatitis (mainly alcoholic AP) in 81 (37.9 %) cases. In non-biliary AP, profiles of both hydrophobic and hydrophilic BAs did not differ from controls. In biliary AP, significantly elevated levels were detected for hydrophobic BAs but not hydrophilic one. The total concentrations of hydrophobic BAs were 33.86±4.51 μM in biliary AP and 8.068±1.98 μM in non-biliary AP (p < 0.0001). Cut-off values were 0.30 μM and 3.78 μM for hydrophilic and hydrophobic BAs, respectively. The total amount of hydrophilic BAs was not different between the groups of deteriorated (Δ APACHE II >0) and ameliorated or stable AP (Δ APACHE II ≤ 0). Conversely, total hydrophobic BAs concentrations were significantly higher in the worsened AP (55.8±12.3 μM) group than in those who recovered or at least did not deteriorate (5.9±0.6 μM). However, ursodeoxycholic acid (UDCA), a hydrophilic BA, was significantly decreased in severe disease and showed a reduced trend in moderately severe AP.
Hydrophobic BAs had more predictive potential with an area under the curve (AUC) of 0.762, a sensitivity of 83.75% and a specificity of 60% at the cut-off value 11.84 μM, whereas hydrophilic BAs were only of limited use for the prediction of AP deterioration, as AUC was 0.57 with a sensitivity of 65% and a specificity of 44% at the cut-off 0.7 μM. Additionally, UDCA concentrations of less than 19 nM can be used to distinguish biliary AP from non-biliary AP with an AUC of 0.723, sensitivity 75%, and specificity 65%. For detection of deterioration, UDCA alone with cut-off concentration 18 nM showed a value with an AUC of 0.68, sensitivity 76%, and specificity 61%.

Conclusion

In patients, the sum of hydrophobic but not total hydrophilic BAs associated with the etiology and severity of AP. Serum levels of UDCA, a particular type of hydrophilic BAs, inversely associated with the severity of AP, which suggests a protective role.

References

[1]. Iannuzzi, J.P.; King, J.A.; Leong, J.H.; Quan, J.; Windsor, J.W.; Tanyingoh, D.; Coward, S.; Forbes, N.; Heitman, S.J.; Shaheen, A.A.; et al. Global Incidence of Acute Pancreatitis Is Increasing Over Time: A Systematic Review and Meta-Analysis. Gastroenterology 2022, 162, 122-134, doi:10.1053/j.gastro.2021.09.043
[2]. Lerch, M.M.; Aghdassi, A.A. The role of bile acids in gallstone-induced pancreatitis. Gastroenterology 2010, 138, 429-433, doi:10.1053/j.gastro.2009.12.012
[3]. Tran, Q.T.; Sendler, M.; Wiese, M.L.; Doller, J.; Zierke, L.; Gischke, M.; Glaubitz, J.; Tran, V.H.; Lalk, M.; Bornscheuer, U.T.; Weiss F.U; Lerch, M.M and Aghdassi A.A. Systemic Bile Acids Affect the Severity of Acute Pancreatitis in Mice Depending on Their Hydrophobicity and the Disease Pathogenesis. International journal of molecular sciences 2022, 23, 13592. doi: 10.3390/ijms232113592

Disclosure

The authors declared no Conflict of Interest.

Introduction

The vedolizumab clinical decision support tool (VDZ-CDST) was developed and validated as a clinical prediction tool for vedolizumab (VDZ) use in Crohn’s disease (CD). It has demonstrated prognostic value with the ability to stratify patients by probability of clinical and endoscopic response to VDZ at week 26. We examined the applicability of the tool in a real-world dataset of VDZ treated CD patients.

Aims & Methods

We performed a retrospective data collection and analysis at a single tertiary centre on patients who had started VDZ between January 2015 to February 2022. We collected baseline data to calculate the CDST: prior exposure to tumour necrosis factor (TNF) antagonists, prior surgery for CD, prior fistulating disease, baseline albumin and C reactive protein (CRP). Our main endpoints were treatment persistence at 12 and 24 months. We also retrieved the following parameters of disease activity at 12 month follow up: CRP, faecal calprotectin (FCP), mucosal healing (defined as ulcer free endoscopy) and progression to surgery. Chi square for trend was used for 12 month persistence and Kaplan-Meier curves with log rank test for trend was used for 24 month persistence. ANOVA was used for comparing continuous variables.

Results

130 patients were included. 74.6% had prior TNF antagonist exposure, median baseline CRP was 4 mg/L, and median albumin was 43 g/L. Using the CDST at baseline, 3 groups were defined – 9 (7%) were low, 51 (39%) were intermediate, and 70 (54%) had a high probability of treatment response. One year treatment persistence rates were 56%, 69%, and 80% in the low, intermediate and high probability groups respectively (p=0.048). Kaplan-Meier analysis demonstrate increasing rates of persistence at 24 months in keeping with CDST-defined probability of response (p=0.035). There were also statistically significant differences in outcomes across the 3 groups, with lower CRP at 12 months, higher rates of mucosal healing and lower rates of progression to surgery in the high probability group versus the low and intermediate groups.

Conclusion

The prognostic value of the VDZ-CDST appears valid in a real-world CD cohort with high rates of prior TNF antagonist exposure. The tool has demonstrated predictive ability for long-term treatment persistence as well as measures of biochemical and endoscopic activity. Our study is consistent with a growing body of published evidence on the applicability of CDST and suggests a role for CDST in guiding therapeutic decision making in routine clinical practice.

References

1. Dulai PS, Boland BS, Singh S, Chaudrey K, Koliani-Pace JL, Kochhar G, Parikh MP, Shmidt E, Hartke J, Chilukuri P, Meserve J. Development and validation of a scoring system to predict outcomes of vedolizumab treatment in patients with Crohn’s disease. Gastroenterology. 2018 Sep 1;155(3):687-95.
2. Dulai PS, Amiot A, Peyrin-Biroulet L, Jairath V, Serrero M, Filippi J, Singh S, Pariente B, Loftus EV Jr, Roblin X, Kane S, Buisson A, Siegel CA, Bouhnik Y, Sandborn WJ, Lasch K, Rosario M, Feagan BG, Bojic D, Trang-Poisson C, Shen B, Altwegg R, Sands BE, Colombel JF, Carbonnel F; GETAID OBSERV-IBD, VICTORY Cohorts Collaboration*. A clinical decision support tool may help to optimise vedolizumab therapy in Crohn's disease. Aliment Pharmacol Ther. 2020 Mar;51(5):553-564. doi: 10.1111/apt.15609. Epub 2019 Dec 22. Erratum in: Aliment Pharmacol Ther. 2021 Apr;53(8):963. PMID: 31867766; PMCID: PMC7028036.

Introduction

Local endoscopic resection for early rectal cancer is increasingly used. For tumors extending into the muscularis propria (T2 carcinoma) a local resection is a suboptimal oncological treatment. Discrimination between T1 and T2+ rectal cancer is therefore important. Optical diagnosis lacks specificity, MRI. EUS has sufficient resolution to visualize bowel wall layers, but interpretation is difficult and subjective. In this pilot study we investigated whether automated imaging analysis (AI), which segemtates the bowel wall layers and the tumor, could be an aid to the performance of EUS in staging early rectal cancer.

Aims & Methods

In this single-center study, L-EUS media, optical and pathological data were prospectively collected from patients referred for local resection. EUS images were manually segmented by FtB. This dataset was used to train a Convoluted Neural Network (CNN) to assign a label (background, lesion / mucosa, submucosa, and muscularis propria) to every pixel. We used an U-Net architecture. CNN performance was increased by data augmentation using affine random transformations.
For the training of our CNN we used cross validation to divide the data set in a training, testing and validation set. We trained our CNN 16 times with every time a different testing set of 5 patients (and a different training and validation set of resp. 70 and 5 patients), this way we could use the entire data set to investigate the CNN performance.
Segmentation success of the trained CNNs was validated by calculating the Dice similarity index (DSI), the mean surface distance (MSD) and Hausdorff distance (HDD) between the automatic and manual segmentation.
In addition, the results of AI analyses were assessed from a clinical point of view, considering it as an aid for EUS interpretation. Overtly misinterpretated images were considered useles and were excluded in the clinical evaluation. Finally, the results were tested against pathology.

Results

We included 80 cases in whom 373 EUS images were manually segmented. The median MSD and HDD were resp. 3.2 and 17.7 pixel for the tumor, 3.4 and 24.7 pixel for the (sub)mucosa and 2.6 and 20.0 for the muscularis propria. The median DSI values for the tumor, (sub)mucosa, and muscularis propria were 0.82, 0.57 and 0.59 respectively. These values reflect good agreement between manual and automatic segmentation.
Clinical evaluated of AI images revealed usefull segmentations in 81% of cases. The sensitivity of automated segmentations for infiltration in the muscular wall was 83.0%; the specificity was 65%, and the accuracy 73%. These were static images only.

Conclusion

We found encouraging results with automatic segmentation based on relatively few static EUS images for the staging of early rectal cancer. Our results support exploration of automatic EUS segmentation to stage early rectal cancer, and improvement is expected when this tool is applied real-time on dynamic EUS imaging.

References

Kuwahara T, Hara K, Mizuno N, Haba S, Okuno N, Koda H, et al. Current status of artificial intelligence analysis for endoscopic ultrasonography. Dig Endosc. 2021;33:298–305.
Ronneberger O, Fischer P, Brox T. U-Net: Convolutional Networks for Biomedical Image Segmentation. Int Conf Med image Comput Comput Interv. 2015;234–41