Background: Treatment approaches differ for isolated in-breast tumor recurrence (representing treatment failure) and new primary breast tumors (representing high etiologic risk). However, methods for distinguishing recurrences from second primaries (based on radiographic and histologic criteria) are subject to error. Gold-standard genomic datasets for assessing classification accuracy have been lacking.
Methods: To identify the scope of misclassification, we performed DNA sequencing of 1,200 genes in 108 participants with synchronous or metachronous second breast cancer in the Carolina Breast Cancer Study (CBCS3). DNA sequencing data included 87 second tumors and 42 paired first and second breast cancers in the same patient (14 contralateral and 28 ipsilateral). Recurrence status was classified based on mutations and copy number, accounting for site-specific mutation probabilities. DNA-based recurrence versus second primary classifications were compared with clinical and Surveillance, Epidemiology and End Results (SEER) classifications based on laterality, histology, quadrant, and latency.
Results: Twenty-two of 28 ipsilateral tumor pairs (79%) shared at least one mutation and were classified as recurrences. Pathologist classifications of ipsilateral second tumors were 79% accurate [95% confidence interval (CI), 60%-90%) with higher sensitivity (95%, 95% CI = CI, 78%-99%) and lower specificity (17%, 95% CI, 3%-56%). SEER classifications were 82% accurate (95% CI, 64%-92%) with lower sensitivity (77%, 95% CI, 57%-90%) and higher specificity (100%, 95% CI, 61%-100%).
Conclusions: Although most genomically defined recurrences are captured by clinical recurrence definitions, new primaries are frequently misclassified.
Impact: Given overtreatment harms, genomic methods may support treatment de-escalation for second breast cancers.
©2025 American Association for Cancer Research.