Evaluation of the HER2 amplification status in oesophageal adenocarcinoma by conventional and automated FISH: a tissue microarray study

Authors:
Publisher website
Google Scholar
Abstract
Introduction The manual fluorescence in situ hybridisation (FISH) Human Epidermal Growth Factor Receptor 2 (HER2)/CEP17 testing method is frequently used, however, it is time consuming and liable to subjectivity. Automation of FISH might increase the throughput and accuracy. Aim To evaluate the agreement between automated and conventional FISH with regard to a reference test silver-enhanced in situ hybridization (SISH) for HER2 amplification, as well as its prognostic significance. Material and methods 154 oesophageal adenocarcinomas were included in a tissue microarray. HER2/CEP17 gene amplification was assessed by automated FISH and was compared with conventional HER2/CEP17 testing methods. Results 46.8% of patients showed HER2 amplified tumours by automated FISH (ratio ≥1.8) compared with 18.1% by conventional FISH. A high automated HER2/CEP17 ratio (≥1.8) was significantly associated with worse survival (HR 1.731; 95% CI 1.075 to 2.786; p=0.024). However, agreement between automated and conventional FISH was only 72.2% and 71.4% between automated FISH and SISH, compared with 94.6% for conventional FISH/SISH. Therefore, thresholds for HER2/CEP17 amplification were sequentially raised from HER2/CEP17 ratio 1.8 till 5.0. A HER2/CEP17 ratio threshold of ≥3.6 had similar prognostic significance as conventional FISH (HR 1.880; 95% CI 1.060 to 3.332; p=0.031 vs HR 1.828; 95% CI 1.102 to 3.033; p=0.020), yielded comparable amplification rates as conventional FISH (14.3% vs 18.1%) and comparable agreement to SISH/Immunohistochemistry (IHC). Conclusions Automation of HER2 FISH analysis in oesophageal cancer has not been performed before. Automated HER2 is feasible, but it seems that the HER2/CEP17 threshold should be adjusted to ≥3.6 to arrive at best comparability with other methods and prognostic value.

This publication has 12 references indexed in Scilit: