Investigation of immunohistochemical marker expression in breast cancer of varying severity and construction of a predictive model.

Background: Breast cancer is one of the most common malignant tumors in women, associated with high morbidity and mortality; immunohistochemical markers play a critical role in its diagnosis, prognostic evaluation, and treatment selection, and this study aims to evaluate the associations between clinicopathological/immunohistochemical markers and stage-defined breast cancer severity at presentation. Methods: We conducted a retrospective cohort study of consecutive breast cancer patients treated at Qinghai University Affiliated Hospital (1 January 2019-1 May 2024). Candidate predictors in the primary model included age, body mass index (BMI), estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2), Ki-67 antigen (Ki-67), epidermal growth factor receptor (EGFR), and cytokeratin 5/6 (CK5/6). Because axillary lymph-node status is structurally related to stage, it was excluded from the primary model and evaluated only in a sensitivity analysis. Associations with advanced disease were assessed by univariable and multivariable logistic regression. A nomogram was built from independent predictors and internally validated for discrimination [area under the curve (AUC)] and calibration. Results: In univariable analyses, older age, higher BMI, axillary lymph-node metastasis, HER2 positivity, high Ki-67 expression, and EGFR positivity were associated with higher odds of severe disease, whereas ER positivity was associated with lower odds. In the revised primary multivariable model excluding nodal status, age remained independently associated with severe disease [odds ratio (OR) =1.03; 95% confidence interval (CI): 1.01-1.06; P=0.01], while BMI and EGFR showed borderline positive associations, and ER was no longer independently significant. In the sensitivity model including nodal status, axillary lymph-node metastasis remained the strongest predictor, and EGFR positivity remained independently associated with severity. The primary model showed moderate discrimination after internal validation (apparent AUC =0.713; optimism-corrected AUC =0.679) with acceptable calibration. Conclusions: Clinicopathologic and immunohistochemical markers correlate with advanced presentation. In the revised primary model, age was independently associated with severe disease, while ER and EGFR showed non-significant or borderline associations after adjustment. In the sensitivity analysis, axillary lymph-node metastasis remained the strongest predictor, and EGFR positivity remained independently associated with severity. The nomogram demonstrated moderate discrimination after internal validation and requires external validation before clinical application.