388
GENOMIC MARKERS FOR ENHANCED RISK STRATIFICATION IN BARRETT’S ESOPHAGHUS PATIENTS WITH LOW GRADE DYSPLASIA
Date
May 19, 2024
Introduction
Current risk stratification of Barrett's esophagus (BE) patients, based on histological identification of dysplasia, lacks reliability due to poor interobserver agreement and the limited predictive value of low-grade dysplasia (LGD). This study aims to identify genomic factors enhancing risk stratification for BE patients with a community diagnosis of LGD.
Patients and methods
Progressors to early esophageal adenocarcinoma and non-progressors were identified in a randomized controlled trial screening cohort of community-based LGD patients. Sequencing used a targeted capture-based panel to detect mutations and copy number changes (CNV). Detected mutations, homozygous deletions, and high-level amplifications underwent filtering for likely pathogenic events. We performed logistic regression, covariate analysis, and penalized mixed-effect models. A joint model for survival and mixed effects was applied for spatiotemporal data analysis.
Results
220 samples, comprising 28 progressors with a median time to progression of 1.2 (IQR 0.4-2.2) years and 95 non-progressors with a median progression-free follow-up of 7.9 (IQR 5.9-10.6) years, all with a median C3M5 BE, were analyzed. Multiple factors were associated with progression, including TP53 (p < 0.0001, Hazard Ratio (HR) = 13.39, 95% confidence intervals (CI) 5.64-31.78), chromosomal arm 17p loss (p < 0.0001, HR = 10.24, 95% CI 4.82-21.76), mutational burden (p < 0.001, HR = 1.52, 95% CI 1.21-1.90), and total number of CNVs (p<0.0001, HR = 1.48, 95% CI 1.34-1.64). Several other alterations trended to be associated with progression, including APC mutation and presence of an oncogenic amplification. The combined influence of TP53 and 17p loss enhanced the accuracy of risk prediction. However, high correlation precluded their use as cumulative risk. Patients with samples containing >3 mutations had a very high risk of progression (HR = 10.33, 95% CI 2.22-48.01). Presence of any genetic variant—amplification, deletion, CNV, or mutation— indicated progression risk (p<0.0001, HR = 1.15, 95% CI 1.11-1.21). Samples lacking any genetic variants showed no progression, Figure 1. A combined TP53 and CNV model effectively identifies progression risk with 64% sensitivity, 96% specificity and an AUC of 0.837, distinguishing 91 out of 95 non-progressors, Figure 2.
Conclusion
This study not only reinforces the well-established role of TP53 mutations but also introduces crucial novel genomic markers. The addition of 17p loss emerges as indispensable for enhanced risk assessment. Furthermore, distinct genetic variations, including CNVs and total mutations, individually and collectively signify a significantly higher risk. Intriguingly, patients without any distinctive genetic abnormalities did not progress. A combined genomic model could accurately risk stratify BE patients with a community-based LGD diagnosis.
Current risk stratification of Barrett's esophagus (BE) patients, based on histological identification of dysplasia, lacks reliability due to poor interobserver agreement and the limited predictive value of low-grade dysplasia (LGD). This study aims to identify genomic factors enhancing risk stratification for BE patients with a community diagnosis of LGD.
Patients and methods
Progressors to early esophageal adenocarcinoma and non-progressors were identified in a randomized controlled trial screening cohort of community-based LGD patients. Sequencing used a targeted capture-based panel to detect mutations and copy number changes (CNV). Detected mutations, homozygous deletions, and high-level amplifications underwent filtering for likely pathogenic events. We performed logistic regression, covariate analysis, and penalized mixed-effect models. A joint model for survival and mixed effects was applied for spatiotemporal data analysis.
Results
220 samples, comprising 28 progressors with a median time to progression of 1.2 (IQR 0.4-2.2) years and 95 non-progressors with a median progression-free follow-up of 7.9 (IQR 5.9-10.6) years, all with a median C3M5 BE, were analyzed. Multiple factors were associated with progression, including TP53 (p < 0.0001, Hazard Ratio (HR) = 13.39, 95% confidence intervals (CI) 5.64-31.78), chromosomal arm 17p loss (p < 0.0001, HR = 10.24, 95% CI 4.82-21.76), mutational burden (p < 0.001, HR = 1.52, 95% CI 1.21-1.90), and total number of CNVs (p<0.0001, HR = 1.48, 95% CI 1.34-1.64). Several other alterations trended to be associated with progression, including APC mutation and presence of an oncogenic amplification. The combined influence of TP53 and 17p loss enhanced the accuracy of risk prediction. However, high correlation precluded their use as cumulative risk. Patients with samples containing >3 mutations had a very high risk of progression (HR = 10.33, 95% CI 2.22-48.01). Presence of any genetic variant—amplification, deletion, CNV, or mutation— indicated progression risk (p<0.0001, HR = 1.15, 95% CI 1.11-1.21). Samples lacking any genetic variants showed no progression, Figure 1. A combined TP53 and CNV model effectively identifies progression risk with 64% sensitivity, 96% specificity and an AUC of 0.837, distinguishing 91 out of 95 non-progressors, Figure 2.
Conclusion
This study not only reinforces the well-established role of TP53 mutations but also introduces crucial novel genomic markers. The addition of 17p loss emerges as indispensable for enhanced risk assessment. Furthermore, distinct genetic variations, including CNVs and total mutations, individually and collectively signify a significantly higher risk. Intriguingly, patients without any distinctive genetic abnormalities did not progress. A combined genomic model could accurately risk stratify BE patients with a community-based LGD diagnosis.
FIGURE 1 KM CURVES Association with Malignant Progression
FIGURE 2 Performances combined model
Tracks
Related Products
DOMAIN-SPECIFIC PRETRAINING OF DEEP LEARNING SYSTEMS IN GASTROINTESTINAL ENDOSCOPY IMPROVES PERFORMANCE OVER CURRENT STATE-OF-THE-ART PRETRAINING METHODS
For training deep learning algorithms in medical imaging, application-specific data is often scarce…
ARTIFICIAL INTELLIGENCE: WHERE DOES IT FIT IN NOW?
A variety of new developments in the realms of screening and surveillance of Barrett's esophagus have occurred in recent years. This session will focus on the current evidence base for risk assessment algorithms to enhance screening uptake, new screening tools and obstacles to implementation…
THE TISSUE SYSTEMS PATHOLOGY TEST ENABLES RISK-ALIGNED MANAGEMENT FOR PATIENTS WITH BARRETT’S ESOPHAGUS
INTRODUCTION: The goal of Barrett’s esophagus (BE) surveillance is to detect dysplasia and esophageal adenocarcinoma (EAC) at early, treatable stages…
