PWE-074 Combination of nuclear textural analysis and image cytometric DNA analysis is more accurate than image cytometric DNA ploidy analysis alone for risk stratification in Barrett's oesophagus

Abstract

Introduction Dysplasia is a marker of cancer risk in Barrett9s oesophagus (BE), but this risk is variable and diagnosis is subject to inter-observer variability. Cancer risk in dysplastic and non-dysplastic BE is increased when large scale genomic instability (DNA ploidy) is present, with published data showing the incidence of aneuploidy in patients with HGD is 67%. Nuclear textural analysis (Nucleotyping) is a new method that uses ultra-high-resolution digital images of stained nuclei to classify DNA content, chromatin structure and organisation both quantitatively and qualitatively. This information is derived from the same sections used for image cytometric DNA ploidy analysis (ICDA). Our aim was to evaluate nucleotyping as a prognostic biomarker in Barrett9s oesophagus, and compare with ICDA. Methods 120 patients with BE were studied. 60 patients had high-grade dysplasia or intramucosal cancer and went on to ablative therapy or surgery. A further 60 patients had specialised intestinal metaplasia only on two consecutive endoscopies with a median of 51 months follow-up (IQR=25–120 months). The two groups were then randomly assigned to a training set (30 HGD, 30 non-dysplastic) and a blinded test set. 40 micron sections were cut from formalin fixed paraffin embedded (FFPE) biopsies and a Feulgen stained nuclear monolayer was made as previously described for ICDA. Results Analysis of the training set demonstrated a DNA ploidy rate of 64% in the HGD group vs 0% in non-dysplastic. By comparison, nuclear textural analysis was abnormal in 71%. A correct classification rate (CCR) of 84% was achieved if the two features were combined. The test set was blindly analysed using the same textural features. DNA ploidy rate was 61% in the HGD group vs 0% in non-dysplastic group. Nucleotyping was again more accurate at 78%, and combining the two features gave a CCR of 84%. When all patients were analysed, sensitivity and specifity of combined nucleotyping with ICDA for detecting dysplasia was 75% and 94%, respectively. Conclusion Nucleotyping yields additional prognostic information vs image cytometry alone on the same nuclear monolayers. The textural features used to differentiate normal from dysplastic tissue were similar to those used in studies of other early cancers. These data demonstrate that nucleotyping is a promising additional test for separating at-risk groups in Barrett9s oesophagus. This work is supported by the CRUK Experimental Cancer Medicine Centre at University College London and the NIHR Comprehensive Biomedical Research Centre at UCLH.