Value of serum tumor markers in early differential diagnosis of spinal tumors and spinal infections.

Abstract

Objectives: The early imaging features of spinal tumors and spinal infections are similar, and the lack of specific early diagnostic indicators can lead to misdiagnosis or missed diagnosis. Although serum tumor markers have been widely used in early cancer screening, spinal tumors are mostly metastatic, lacking specific markers, and some patients with spinal tumors may even test negative for tumor markers. Therefore, the role of tumor markers in the early differential diagnosis of spinal tumors remains unclear. This study aims to investigate the value of tumor markers in the early differential diagnosis of spinal tumors and spinal infection.

Methods: We retrospectively analyzed the clinical data of 221 patients with spinal bone destruction admitted to Xiangya Hospital of Central South University between April 2017 and October 2022. Peripheral blood levels of 10 tumor markers were measured upon admission, including carcinoembryonic antigen (CEA), alpha fetoprotein (AFP), neuron-specific enolase (NSE), cytokeratin 19 fragment antigen 21-1 (cyfra21-1), carbohydrate antigen 199 (CA199), carbohydrate antigen 125 (CA125), carbohydrate antigen 72-4 (CA72-4), pepsinogen I (PGI), pepsinogen II (PGII), and the PGI/PGII ratio (PGR). Univariate Logistic analysis was used to screen relevant variables, and the correlation between tumor markers was analyzed. Multivariate Logistic regression was then employed to identify risk and protective factors. The optimal cut-off values were calculated using the receiver operating characteristic (ROC) curve and Youden index, and a differential diagnosis model for early spinal tumors and spinal infections was constructed based on the selected indicators. Diagnostic performance was then evaluated.

Results: According to the pathological diagnosis, 91 patients had spinal tumors, and 130 patients had spinal infections. The levels of CEA, AFP, NSE, cyfra21-1, CA199, and CA72-4 were higher in the spinal tumor group than those in the spinal infection group, while PGR was lower, with statistically significant differences (all P<0.05). Univariate Logistic analysis showed significant differences in CEA, AFP, NSE, cyfra21-1, CA199, CA125, CA72-4, and PGR (all P<0.05). Correlation analysis indicated a strong positive correlation between CEA and cyfra21-1, CA199, and CA125, as well as a strong negative correlation of PGII with PGI and PGR. Multivariate Logistic analysis identified AFP and cyfra21-1 as risk factors (P<0.01) and PGR as a protective factor (P<0.05). When the standard cut-off values (AFP<20 ng/mL, cyfra21-1< 3.3 ng/mL, and PGR>3) were applied, AFP had a sensitivity of 3.3% and accuracy of 60.2%, cyfra21-1 had a sensitivity of 20.9% and accuracy of 66.1%, and PGR had a sensitivity of 2.2% and accuracy of 59.3%. The AUC of the early spinal tumor diagnosis model was 0.623. Using the optimal cut-off values (AFP<1.625 ng/mL, cyfra21-1<1.175 ng/mL, and PGR>11.05) from the ROC curve, AFP had a sensitivity of 69.2% and accuracy of 64.7%, cyfra21-1 had a sensitivity of 68.1% and accuracy of 70.1%, and PGR had a sensitivity of 62.6% and accuracy of 57.5%, with the model's AUC at 0.772.

Conclusions: A comparative analysis of 10 serum tumor markers reveals that AFP, cyfra21-1, and PGR are significantly associated with spinal tumors. The diagnostic model constructed based on these 3 markers shows promising predictive performance and may be helpful to the differentiation for the early clinical diagnosis of spinal tumors and spinal infection.