An evaluation model of chemotherapy efficacy in breast cancer based on combined serum CA15-3, CEA, and CYFRA 21 - 1.

[BACKGROUND AND AIM] Breast cancer remains one of the most prevalent malignancies in women worldwide, and chemotherapy plays a crucial role in systemic treatment. However, accurate assessment of therapeutic efficacy remains a clinical challenge. Traditional radiographic evaluation is time-consuming and sometimes limited in reflecting biological response. Serum tumor markers, including carbohydrate antigen 15 − 3 (CA15-3), carcinoembryonic antigen (CEA), and cytokeratin fragment 21 − 1 (CYFRA 21 − 1), have been reported to correlate with tumor burden and treatment response. This study aimed to construct a retrospective evaluation model based on combined serum CA15-3, CEA, and CYFRA 21 − 1 for predicting chemotherapy efficacy in breast cancer patients.

[METHODS] A total of 352 breast cancer patients who received standardized chemotherapy between January 2020 and December 2024 were retrospectively enrolled. Serum levels of CA15-3, CEA, and CYFRA 21 − 1 were collected at baseline and after two cycles of chemotherapy. The associations between tumor marker dynamics and clinical response (complete response, partial response, stable disease, progressive disease) were analyzed using logistic regression. A composite evaluation model (CCC model) was developed, and receiver operating characteristic (ROC) curves were applied to assess its discriminative performance compared with single markers.

[RESULTS] Among the 352 patients, 198 (56.3%) achieved an objective response (CR + PR), while 154 (43.7%) had non-response (SD + PD). Post-chemotherapy levels of CA15-3, CEA, and CYFRA 21 − 1 were significantly lower in the response group compared with the non-response group (all  < 0.001). Logistic regression revealed that decreases in CA15-3 (OR = 1.046, 95% CI: 1.018–1.075,  < 0.001), CEA (OR = 1.037, 95% CI: 1.012–1.064,  = 0.002), and CYFRA 21 − 1 (OR = 1.059, 95% CI: 1.022–1.097,  = 0.001) were independent predictors of chemotherapy efficacy. The CCC model integrating the three markers achieved an area under the ROC curve (AUC) of 0.921 (95% CI: 0.894–0.949), which was superior to single markers (CA15-3: 0.812, CEA: 0.773, CYFRA 21 − 1: 0.756).

[CONCLUSIONS] The combined serum tumor marker model (CA15-3, CEA, and CYFRA 21 − 1) provides a robust, simple, and cost-effective method for retrospective evaluation of chemotherapy efficacy in breast cancer patients. This CCC model can serve as a valuable adjunct to imaging modalities, enabling more precise treatment monitoring and potentially guiding timely therapeutic adjustments.

[SUPPLEMENTARY INFORMATION] The online version contains supplementary material available at 10.1186/s12885-026-15687-y.