With the approval of HER2-targeted antibody-drug conjugate options for treating patients with breast cancer across different HER2 expression levels, accurate assessment of HER2 expression has become more important than ever. And a recent study may provide a solution to the challenge of accurate immunohistochemical (IHC) HER2 scoring.
Study results presented in a press briefing ahead of the 2025 ASCO Annual Meeting, held on May 21, 2025, suggest that artificial intelligence (AI)-assisted classification may be more effective than unassisted pathologist classification for HER2 scoring of patients with breast cancer, particularly in cases with low levels of expression.1 The accuracy of identifying HER2-low and HER2-ultralow breast cancers was improved with the use of AI. Furthermore, the rate of misclassifications of tumors as HER2-null was reduced with AI assistance.
Marina De Brot, MD, PhD
“Roughly 65% of breast tumors once called HER2‑negative actually demonstrate some level of HER2 expression and belong to subgroups now classified as HER2-low or HER2-ultralow breast cancers. Some of these tumors could be treated with HER2-targeted antibody-drug conjugates, but only if we detect their HER2 expression levels,” stated lead study author and presenter Marina De Brot, MD, PhD, Associate Pathologist and Postgraduate Professor, A.C. Camargo Cancer Center, São Paulo, Brazil, in a press release.
During the press briefing, Dr. De Brot added: “These findings highlight the value of AI systems in biomarker interpretation training, providing pathologists with enhanced decision-making tools at the individual cell level and improving diagnostic precision in HER2 IHC interpretation.”
Study Methodology
An AI-integrated pathologist training platform called ComPath Academy was established to support virtual training masterclasses for HER2 assessment of breast cancer samples. The Breast HER2 Masterclasses included a total of 105 pathologists from 10 countries in Asia, Africa, and South America who assessed 20 digital breast cancer HER2 IHC cases with and without AI assistance.
The pathologists performed 1,940 total readings divided between three exams, with AI assistance for decision support available only during the third exam. In between the exams, the participating pathologists received a lecture on HER2 scoring and discussed the results of the first two exams.
IHC staining applied the HER2 4B5 PATHWAY/VENTANA assay. All readings were compared with ground-truth scores assigned by the consensus of multiple expert pathologists from a central reference center. Classification was conducted in line with ASCO/College of American Pathologists 2023 guidelines on HER2 testing in breast cancer,2 plus newer categories of HER2-ultralow (IHC 0 with membrane staining) and HER2-null (IHC 0 with no membrane staining). Of note, the College of American Pathologists updated its breast cancer guidelines in March 2025 to include HER2-ultralow, referenced as “score 0+ with membrane staining.”3
In the third exam, the AI-integrated platform showed a categorization for each neoplastic cell, a calculation of the total tumor cell count, percentage of each cell class, and HER2 clinical category of the case. If the HER2 scoring was 2+, in situ hybridization was also provided.
The pathologists achieved an accuracy rate with reference scores of 89.1% without AI assistance in the first two exams compared with 96.1% accuracy with AI assistance in the third exam. Concordance among the pathologists was calculated at 0.506 without AI and 0.798 with AI.
Classification of each HER2 clinical category improved from 90.1% without AI to 95.0% with AI assistance. Concordance among pathologists by HER2 clinical category increased from 0.4994 without AI to 0.732 with AI. The rate of misclassification of HER2-low or HER2-ultralow cases as HER2-null was reduced by 24.4% with AI assistance.
Dr. De Brot pointed out during the press briefing that manual scoring sensitivity was lowest in the cases with ultralow or nonexistent levels of HER2 expression. In the first two exams, the predicted classification of HER2-null matched the true classification in only 54.08% of cases and in 50.74% of HER2-ultralow cases. On the other hand, the predicted and true classifications matched in 97.94% of HER2-positive cases. With AI support, the matching rate was 88.24% for HER2-null cases, 93.22% for HER2-ultralow, and 90.29% for HER2-positive.
Clinical Implications
HER2-targeted therapy with antibody-drug conjugates have demonstrated benefit among patients with lower levels of HER2 expression.4 The HER2-directed antibody-drug conjugate trastuzumab deruxtecan has already been approved by the U.S. Food and Drug Association to treat patients with low and ultralow levels of HER2 by IHC testing, in the metastatic setting.5
For patients with breast cancer who have lower levels of HER2 expression to qualify for treatment with these antibody-drug conjugates, their IHC scoring needs to accurately identify even ultralow levels of HER2 expression. With the use of AI assistance, as shown in this study, many more women who have been misclassified as HER2-null could become eligible for treatment with these antibody-drug conjugates.
Julian Hong, MD, MS
“Accurate HER2 scoring is important to ensure that patients receive the best treatment for their breast cancer. This international study shows that an AI-assisted approach improved HER2 scoring, including in situations that would affect treatment decisions,” stated ASCO expert on AI Julian Hong, MD, MS, Associate Professor and Medical Director of Radiation Oncology Informatics at the University of California, San Francisco. “These findings highlight the promising role for AI in oncology, not as a replacement for the physician, but as a powerful tool to help us work smarter and faster to deliver high-quality, more personalized care.”
DISCLOSURE: Funding for this study was provided by AstraZeneca. For disclosures of the study authors, visit meetings.asco.org.
REFERENCES
1. Mulder D, et al: 2025 ASCO Annual Meeting. Abstract 1014. Presented at press briefing May 21, 2025.
2. Wolff AC, et al:J Clin Oncol 41:3867-3872, 2023.
3. College of American Pathologists: Reporting Template for Reporting Results of Biomarker Testing of Specimens From Patients with Carcinoma of the Breast. March 2025. Available at https://documents.cap.org/documents/New-Cancer-Protocols-March-2025/Breast.Bmk_1.6.0.0.REL.CAPCP.pdf. Accessed May 21, 2025.
4. Bardia A, et al: N Engl J Med 391:2110-2122, 2024.
5. U.S. Food and Drug Administration: FDA approves fam-trastuzumab deruxtecan-nxki for unresectable or metastatic HR-positive, HER2-low or HER2-ultralow breast cancer. Updated February 6, 2025. Available at https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-fam-trastuzumab-deruxtecan-nxki-unresectable-or-metastatic-hr-positive-her2-low-or-her2. Accessed May 21, 2025.
