Introduction

Machine learning algorithms (MLA) trained on capsule endoscopy (CE) images have shown high accuracy in detecting signs of villous atrophy in patients with coeliac disease. However, using MLA to quantify coeliac disease extent and the proportion of small bowel affected has not been studied.

Aims & Methods

We analysed CE videos of adult patients (≥18 years) with biopsy-proven coeliac disease from a large prospectively maintained database. Demographic, clinical, and biochemical data were collected for all patients. The extent of coeliac disease and the proportion of small bowel affected were calculated for each video by an expert human reader and by a trained MLA. The consistency of agreement between the expert human and the MLA was estimated by intraclass correlation coefficients (ICCs) using a two-way mixed-effects model.

Results

CE videos of 75 patients with CD (median age 59 years [IQR 41.5-68 years], 60% female) were included. The median gastric passage time and small bowel transit time were 16 (IQR 98– 50.5) and 277 (IQR 217.5 – 252) minutes, respectively. Macroscopic features of CD, including scalloping, fissuring, mosaicism, nodularity, were identified beyond the first tertile in 34.7% of patients (extensive disease). The overall agreement between the expert human and MLA for the proportion of small bowel affected was good (ICC 0.799, 95% CI 0.572-0.893). Patients with extensive small bowel CD were significantly more likely to have anaemia or ferritin deficiency than those with proximal disease (p<0.01 and 0.026 respectively).

Conclusion

Trained MLA can accurately quantify the extent of coeliac disease and the proportion of small bowel involvement on CE, showing a good agreement with expert human assessment. This machine learning quantification of coeliac disease extent allows closer monitoring for nutritional deficiencies associated with extensive disease and provides an objective and standardised endpoint on response to therapy in clinical trials. Larger training and validation datasets are required to optimise the performance of the MLA.

Introduction

Patients with inflammatory bowel diseases (IBD) frequently experience fatigue, affecting up to 80% of those with active disease and approximately 50% with quiescent disease (1). The exact cause of IBD-associated fatigue is often unknown, making clinical management very challenging. Previous studies have suggested the potential involvement of infectious triggers and host immune responses in the development of fatigue.

Aims & Methods

In this study, we aimed to explore whether patients with quiescent IBD reporting fatigue exhibit specific systemic antibody responses, which could provide insight into potential immune reactivities underlying fatigue.
Systemic antibody epitope repertoires were profiled leveraging PhIP-Seq against 344,000 rationally selected peptide antigens. Fatigue severity was assessed on a 10-point Likert scale, ranging from 1 (no fatigue) to 10 (highest fatigue severity). Quiescent IBD was defined as clinical (Harvey-Bradshaw Index [HBI] <5 or Simple Clinical Colitis Activity Index [SCCAI] <2.5) and biochemical remission (C-reactive protein [CRP] <5 mg/L) at time of sampling. Multivariable logistic regression analyses, allowing adjustment for potential confounding factors e.g. age, sex, and smoking, were performed to identify associations between fatigue and systemic antibody responses.

Results

In total, 327 patients with IBD (156 with Crohn’s disease [CD]; 171 ulcerative colitis [UC]) were profiled. A total of 153 different antibody-bound peptides were identified in association with fatigue considering nominal significance (nominal P-value <0.05), in the absence of significant associations after adjustment for multiple comparisons. Among antibodies nominally significantly associated with fatigue, 56 antibody-bound peptides (36.6%) were found in lower frequency among highly fatigued patients (fourth quartile, Q4), while 97 peptides (63.4%) were identified as more frequent in highly fatigued patients compared to those with lower fatigue scores (<Q4). Among highly fatigued patients, elevated antibody responses were primarily directed towards viral antigens, with by far the strongest signal against several antigens from Epstein-Barr virus (EBV), mostly directed at the EBV nuclear antigens 1 (EBNA-1) and 3B (EBNA-3B). In addition, elevated antibody responses were observed against bacterial antigens, including functional proteins from Streptococcus and Staphylococcus species. Most findings remained materially unchanged when we repeated analyses solely in patients with quiescent IBD, although antibody responses against EBV were altered, e.g. showing lower responses directed against the BLRF/BMRF/EA-D pathway. Furthermore, these antibody signatures were distinct from those recently identified in association with chronic fatigue syndrome (CFS) (2).

Conclusion

Fatigue in IBD is not reflected by distinct antibody epitope repertoires. However, our study suggests a potential involvement of viral antigens, particularly EBV, in IBD-associated fatigue. Although larger confirmatory studies are needed to validate these findings, this study shows that a technology like PhIP-Seq may represent a valuable strategy to approach the investigation of immune responses underlying complex symptoms such as fatigue.

References

1. D’Silva A, Fox DE, Nasser Y, Vallance JK, Quinn RR, Ronksley PE, et al. Prevalence and Risk Factors for Fatigue in Adults With Inflammatory Bowel Disease: A Systematic Review With Meta-Analysis. Clinical Gastroenterology and Hepatology. 2022;20(5):995-1009.e7. doi:10.1016/j.cgh.2021.06.034

2. Vogl T, Kalka IN, Klompus S, Leviatan S, Weinberger A, Segal E. Systemic antibody responses against human microbiota flagellins are overrepresented in chronic fatigue syndrome patients. Sci Adv. 2022;8(38). doi:10.1126/sciadv.abq2422

Clinical Case Summary

Secondary aortoenteric fistula (AEF) is an abnormal connection between the aorta and gastrointestinal tract (GT) resulting from reconstructive surgery for an abdominal aortic aneurysm (AAA), including open repair surgery and endovascular treatment. Despite the low incidence (1.6 to 4%) [1–4], it is life-threatening and has a high mortality rate (24 to 45.8%) [5–6].
A 79- year-old man reported to the emergency department of our hospital with overt bleeding from the stoma. The patient had a history of open Dacron graft repair for a ruptured AAA eight months previosly, which became complicated on the 5th postoperative day by sigmoid ishemia that required an emergency Hartmann procedure.
His blood pressure was 90/60 mmHg, heart rate 93/minute and hemoglobin level 7.5 g/dL.
The patient was admitted to our department. He underwent upper and lower endoscopy and no bleeding lesions were found. After endoscopy the patient presented another episode of overt bleeding from the stoma and we decided to perform Small-bowel capsule endoscopy (SBCE) which showed activ bleeding within the proximal ileum. The contrast-enhanced CT demostrated an aneurysm sac enlargement, a close proximity of the small bowel loops with the aneurysm sac in the pelvic cavity and a suspected leakage at the distal part of the aneurysm sac.
The patient was therefore admitted to the Surgical Derpatment and underwent an emergency endovascular repair of the leakage caused by distal detachment of the aorto-aortic prosthesis using a bifurcated stant graft. No more bleeding from the stoma occurred and the patient was discharged three days later with a diagnosis of distal aortic anastomosis leakage with aorto-ileal fistula.
The duplex ultrasound of the graft after one, three and six months showed no sign of endoleak.
This case shows that the SBCE may be useful in patients with history of recostructive surgery for AAA who manifest overt bleeding from the GT and no lesions are found in upper and lower endoscopy.

References

1.Wilson WR, Bower TC, Creager MA, Amin-Hanjani S, O’Gara PT, Lockhart PB, Darouiche RO, Ramlawi B, Derdeyn CP, Bolger AF, et al. Vascular Graft Infections, Mycotic Aneurysms, and Endovascular Infections: A Scientific Statement From the American Heart Association. Circulation. 2016;134(20): e412–60
2.Champion MC, Sullivan S, Coles J, Goldbach M, Watson WC. Aortoenteric fistula. Incidence, presentation recognition, and management. Ann Surg. 1982;195(3):314–7
3.Hallett JW Jr, Marshall DM, Petterson TM, Gray DT, Bower TC, Cherry KJ Jr, Gloviczki P, Pairolero PC. Graft-related complications after abdominal aortic aneurysm repair: reassurance from a 36-year population-based experience. J Vasc Surg. 1997;25(2):277–84. discussion 285-276
4.Chenu C, Marcheix B, Barcelo C, Rousseau H. Aorto-enteric fistula after endovascular abdominal aortic aneurysm repair: case report and review. Eur J Vasc Endovasc Surg. 2009;37(4):401–6
5.Malik MU, Ucbilek E, Sherwal AS. Critical gastrointestinal bleed due to secondary aortoenteric fistula. J Community Hosp Intern Med Perspect. 2015;5(6):29677
6.Gnus J, Ferenc S, Koscielna M, Paprocka-Borowicz M, Dawidczyk P, Dziewiszek M, Witkiewicz W. Secondary Aortoenteric Fistula After Abdominal Aortic Graft Implementation in Our Own Material. Adv Clin Exp Med. 2016;25(6):1265–71