366P - Artificial intelligence (AI)-powered assessment of complete and intense human epidermal growth factor receptor 2 (HER2)-positive tumor cell proportion in breast cancer: Predicting fluorescence in situ hybridization (FISH) positivity and response to HER2-targeted therapy

Background

The level of expression of HER2 is important in predicting response to HER2-targeted therapies. HER2 expression level is classified in groups (3+ positive, 2+ equivocal, and 1+ or 0 negative) or quantified as an H-score based on the proportion of tumor cell intensity (complete and intense [strong positive], weak to moderate complete [moderate positive], incomplete and faint/barely perceptible [weak positive]). If the case is classified as 2+ equivocal, an additional test such as Fluorescence In Situ Hybridization (FISH) test is performed to confirm the positivity. In this study, we aim to figure out that the complete and intense HER2-positive tumor cell proportion can predict both FISH positivity and response to HER2-targeted therapy, compared to the H-score.

Methods

Lunit SCOPE HER2, an AI-powered HER2 analyzer that identifies and quantifies HER2 IHC according to staining intensity of tumor cells, was used to evaluate the HER2 status of breast cancer cases (n=205) from Kyung Hee University Hospital. They consisted of 41 3+, 163 2+, and 1 1+, as interpreted by pathologists. Of these, 171 had FISH results, and 44 received HER2-targeted therapy as neoadjuvant (n=25) or as palliative (n=19). The proportion of strong positive tumor cells or the H-score by the AI-powered HER2 analyzer was assessed to predict clinical outcomes.

Results

Among all cases, FISH positivity was 20.5%, pathologic complete response (pCR) after neoadjuvant therapy was 40.0%, and complete/partial response during palliative therapy was 68.4%. The proportion of strong positive tumor cells was a better predictor of FISH positivity and response compared to H-score (Area under the curve: 0.784 vs. 0.688 for FISH, and 0.872 vs. 0.859 for response). Prediction of pCR was equivalent for both (0.773). The DeLong test showed a significant difference in the prediction of FISH positivity (p=0.023).

Conclusions

This study showed that AI analysis of breast cancer HER2 expression, specifically the proportion of strong positive tumor cells, can predict FISH and therapy response better than the H-score.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Lunit Inc.

Disclosure

S.I. Cho, H. Song, S. Park, S. Pereira, D. Yoo: Financial Interests, Personal, Full or part-time Employment: Lunit Inc.; Financial Interests, Personal, Stocks/Shares: Lunit Inc. S. Shin, S. Kim: Financial Interests, Personal, Full or part-time Employment: Lunit Inc. All other authors have declared no conflicts of interest.