Predicting oesophageal cancer progression using genomic information in pre-malignant oesophageal tissues


Year:

Session type:

Sarah Killcoyne1,Eleanor Gregson1,David Wedge2,Rachel de la Rue1,Ahmad Miremadi1,Matthew Eldridge3,Moritz Gerstung4,Rebecca Fitzgerald1
1University of Cambridge,2University of Oxford,3Cancer Research UK Cambridge Institute,4European Bioinformatics Institute

Abstract

Background

Esophageal adenocarcinomas (EAC) have been shown to have high rates of somatic copy number alterations that are likely to drive the progression of the disease. A pre-malignant tissue, known as Barrett’s Esophagus (BE), also develops early somatic copy number changes that appear to be later magnified in EAC suggesting a possible route for new diagnostic tests.

Method

Using a nested case-control cohort of 69 patients diagnosed with BE, and following the standard surveillance protocols, the whole-genome copy number status of the tissue was evaluated in 634 samples from 2-15 years prior to a diagnosis of high-grade dysplasia or cancer. The copy number status of those patients who have progressed to cancer versus those with Barrett’s who have not progressed was used to develop a model which predicted progression per sample.

Results

The model was shown to robustly classify both progressive and non-progressive samples with an AUC of 79% using default cutoffs. At the pathological endpoint (high-grade dysplasia) the model predicted 93% of early cancer samples correctly as progressive, but most important were the predictions for pathologies prior to the endpoint, these predictions included pathological grades prior to current diagnostic guidelines, particularly non-dysplastic BE (see table below).  It was also able to accurately predict patients who would progress 6-8 years prior to diagnosis in 88% of endoscopies. In a separate cohort of 20 patients (158 samples) these prediction rates were validated with a sensitivity of 69% and specificity of 76%.

Pathology (samples)

All Patients (69)

Progressors Only (35)

Validation Cohort (10P+10NP)

BE

77%

69%

74%

Low-grade Dysplasia

78%

77%

93%

Indeterminate

92%

88%

77%

High-grade Dysplasia (Progressors only)

N/A

92%

87%

Conclusion

Genomic copy number information generated from current surveillance protocols can improve early diagnosis in this disease, and enable better management for Barrett’s patients.

Share this abstractTweet about this on TwitterPrint this pageShare on FacebookEmail this to someoneShare on LinkedIn