While tumor-associated collagen signature (TACS) is known to independently predict outcomes in breast cancer, it is not yet established whether a full collagen profile—including TACS, TACS-derived microscopic collagen characteristics (TCMF1), and nuclear features associated with TACS (TCMF2)—can enhance the prognostic accuracy of the existing tumor–node–metastasis (TNM) staging system. A total of 941 breast cancer patients were included across three cohorts: training (n = 355), internal validation (n = 334), and external validation (n = 252). Tumor-associated collagen signature (TACS) and TACS-derived microscopic features (TCMF1) were assessed using multiphoton microscopy (MPM), while TACS-derived nuclear features (TCMF2) were extracted from hematoxylin and eosin-stained images aligned with the MPM images. These features were combined linearly to generate a comprehensive collagen signature score, which was then used to refine TNM staging into stage I (II and III)/low-risk and stage I (II and III)/high-risk categories.
Patients with low-risk collagen signatures who were originally classified as stage II or III were effectively “downstaged,” while those with stage I tumors and high-risk collagen signatures were “upstaged.” Integrating the complete collagen signature into the TNM system markedly improved risk stratification. In the modified staging, stage II–new showed hazard ratios of 8.655, 6.136, and 4.699, and stage III–new had hazard ratios of 14.855, 11.201, and 13.245 across the training, internal, and external validation cohorts, respectively, compared to stage I–new. In contrast, the conventional TNM system exhibited lower hazard ratios for stage II (1.642, 1.853, and 1.371) and stage III (4.131, 4.283, and 3.711) in the same cohorts. Moreover, the modified system outperformed the traditional TNM staging in predictive accuracy, achieving AUCs of 0.843 versus 0.683 in the training cohort, 0.792 versus 0.661 in the internal validation cohort, and 0.793 versus 0.646 in the external validation cohort. The comprehensive collagen signature independently predicts survival outcomes in breast cancer and provides additional insights into the tumor’s biological behavior, enhancing the prognostic value of current staging systems.
