Background: HER2 is the only validated predictive biomarker in gastro-oesophageal carcinoma (GOC). However, several factors, such as heterogeneity in protein expression, shortage of evaluable tumour tissue and need for quick target assessment, underline the usefulness of a pre-screening tool in order to anticipate HER2 status.
Methods: Data from 723 consecutive GOC analysed for HER2 at four Italian Institutions were collected. HER2 positivity was defined as 3+ by immunohistochemistry (IHC) or 2+ with gene amplification by in situ hybridisation (ISH). A multivariate logistic regression model was built using data from 413 cases, whereas 310 patients served as validation cohort. C-index, visual inspection of the calibration plot, Brier score and Spiegelhalter z-test were used to assess the performance of the nomogram.
Results: HER2 positive rate was 17.4%. Four variables were retained after adjustment in the final model: grading, Lauren's histotype, pathologic material analysed (surgical specimen/biopsy) and site of tissue collection (primary tumour/metastases). Visual inspection of the calibration plot revealed a very good overlap between predicted and observed probabilities, with a Brier score of 0.101 and a non-significant Spiegelhalter z-test (P = 0.319). C-index resulted in 0.827 (95%CI 0.741-0.913).
Conclusion: A simple nomogram based on always-available pathologic information accurately predicts the probability of HER2 positivity in GOC.