Breast cancer is one of the most frequently occurring malignancies in China, with a steadily increasing incidence (1). Although the development of comprehensive therapy has reduced the mortality rate of breast cancer patients (2), regional and distant recurrences remain a major threat to breast cancer patients (3, 4). Serum tumor markers have been widely used as non-invasive tools for measuring treatment response, early diagnosis of recurrence and predicting prognosis. In breast cancer, the most widely used serum tumor markers are cancer antigen 15-3 (CA 15-3) and carcinoembryonic antigen (CEA). Although CA 15-3 and CEA are not currently recommended as markers for breast cancer screening or therapeutic response monitoring according to the American Society of Clinical Oncology guidelines (5), CA 15-3 and CEA are the markers most widely used for surveillance purposes and monitoring of treatment response in clinical practice.

Previous studies have suggested that elevated serum levels of CA 15-3 and CEA preoperatively were significantly associated with tumor size, axillary node metastasis and advanced stage in the multivariate analysis; in addition, patients with elevated CA 15-3 and CEA levels exhibited a worse prognosis compared to those with normal levels of CA 15-3 and CEA, even in stage-matched analysis (6, 7). Recent studies demonstrated that serial determination of CA 15-3 and CEA levels during the postoperative follow-up of breast cancer patients may be useful for early detection of preclinical recurrence or metastatic disease (8, 9). In addition, the serum levels of CA 15-3 and CEA may be used to predict response to chemotherapy in patients with metastatic breast cancer (MBC) (10). Similarly, a model for predicting disease progression using CA 15-3 and CEA in MBC patients without target lesions in first-line chemotherapy was suggested by Hashimoto et al (11), which delineated that evaluations of CA 15-3 and CEA at baseline contributed more significantly to the prediction of disease progression compared to other variables.

Despite continuous interest in this topic, only a limited number of studies have described the correlation of increased CA 15-3 and CEA levels at recurrence with clinicopathological parameters and specific sites of recurrence. A better understanding of the correlation between serum CA 15-3 and CEA levels and the intrinsic biological characteristics of breast cancer may affect the selection of adjuvant therapy and surveillance decisions and determine which investigations and therapies are appropriate upon diagnosis of distant disease.

In the present retrospective study, we analysed the association of serum CA 15-3 and CEA levels with clinicopathological parameters in patients with MBC and determined whether the elevation of these two tumor markers is correlated with metastatic site(s).

Breast cancer is a heterogeneous group of diseases that differ in their pathological and molecular characteristics and clinical behavior. Despite advances in treatment, 20–30% of patients with early breast cancer eventually develop relapse with distant metastatic disease (2). The risk of recurrence is affected by stage at initial presentation and the underlying biology of the tumor. To improve early detection of recurrence and survival, CA 15-3 and CEA are widely used for the early diagnosis of metastatic disease in the clinical setting. During follow-up, CA 15-3 and CEA levels have been shown to detect 40–60% of recurrences prior to clinical or radiological evidence of disease, with a lead time of 2–18 months. The combined measurement of both serum markers allows early diagnosis of metastases in up to 60–80% of breast cancer patients (12, 13). It was previously demonstrated that CA 15-3 and CEA levels are elevated in ∼80 and 40%, respectively, of patients diagnosed with MBC (14, 15). In the present study, elevated CA 15-3 and CEA levels at initial diagnosis of recurrence were identified in 163 (57.4%) and 97 (34.2%) of the 284 patients, respectively, in accordance with previous findings.

Previous studies reported an association between elevated CA 15-3 levels and ER positivity, namely that the incidence of increased CA 15-3 levels was higher in patients with ER-positive primary tumors compared to those with ER-negative tumors (9, 16). Our results indicated that CA 15-3 and CEA elevation at initial diagnosis of MBC were associated with breast cancer molecular subtypes (P<0.001 and P=0.032, respectively). Elevated CA 15-3 and CEA levels were most commonly observed in luminal subtypes, whereas the incidence of CA 15-3 and CEA elevation was lower in the HER2-enriched and TN subtypes. The reason for this difference has not been fully elucidated. According to a previous study (17), the expression patterns of luminal, HER2-enriched and basal-like tumors are closely associated with mature luminal, luminal progenitor and the basal stem/progenitor cells of normal breast tissue, respectively. Based on that finding, it is hypothesized that less differentiated subtypes lack certain circulating antigens. Luminal subtypes are characterized by a high expression of hormone receptor (HR)-related genes, whereas HER2-enriched or basal-like subtypes exhibit a low expression of HR-related genes (18), indicating an association between CA 15-3 and CEA elevation and HR expression. In addition, a greater proportion of patients with preoperatively elevated tumor marker levels exhibited a more prominent elevation of tumor markers at recurrence compared to those without preoperatively elevated marker levels (19). The measurement of serum tumor markers is relatively easy and cost-effective. The incidence of the increase in CA 15-3 and CEA levels according to breast cancer molecular subtype may be useful in clinical practice.

As regards the correlation between marker levels and metastatic site(s), the results thus far have been conflicting: Tampellini et al (16) reported that increased CA 15-3 levels were found more frequently in patients with liver metastases and in those with pleural effusion. Yerushalmi et al (20) did not identify significant differences in CA 15-3 and CEA levels between different sites of metastasis, whereas al-Jarallah et al (21) reported that the highest proportions of elevated CA 15-3 levels were found in patients with bone metastases and the highest proportions of elevated CEA levels in patients with lung, bone or liver metastases. Our study demonstrated that the elevation CA 15-3 at recurrence was correlated with bone metastasis (P=0.017), whereas an elevated CEA level was observed regardless of metastatic site (P>0.05). Further studies, investigating the effect of metastatic site on CA 15-3 and CEA elevation, are required. Increased CA 15-3 was significantly more common in patients with ≥2 metastatic sites compared to MBC with a single metastasis, in agreement with previously reported findings (19, 22). The significantly elevated CA 15-3 levels observed in cases with multiple metastases may reflect tumor burden.

In conclusion, CA 15-3 and CEA level elevation at initial diagnosis of recurrence were found to be associated with breast cancer molecular subtype. These serum tumor markers are frequently increased in the luminal subtypes of breast cancer. In addition, an elevated CA 15-3 level was correlated with bone metastasis, whereas elevated CEA was observed regardless of metastatic site. The elevation in CA 15-3 levels also differed according to the number of metastatic sites.