The feasibility of using a multivariate regression model incorporating ultrasound findings and serum markers to predict thyroid cancer metastasis.

OBJECTIVE

This study aimed to assess the viability of a multivariate regression model utilizing ultrasound findings and serum markers for predicting thyroid cancer metastasis.

METHODS

A retrospective analysis of 98 thyroid patients admitted from January 2022 to October 2022 was conducted to categorize them into a metastasis group (n=20) and a non-metastasis group (n=78) based on postoperative pathological results. Both groups underwent ultrasound examination and serum marker testing. Correlative analysis was performed to explore the association between various indicators and thyroid cancer metastasis. A multivariate regression model was developed, and receiver operating characteristic (ROC) curves were used to assess the predictive value of ultrasound findings, serum markers, and their combination for thyroid cancer metastasis.

RESULTS

Statistically significant differences were found in the levels of ultrasound findings and serum markers between the two groups. Nodule boundaries, presence or absence of halos, margins, lobulation, capsular invasion, surface smoothness, nodule aspect ratio, uric acid, total cholesterol, triglyceride, and LDL cholesterol levels were predictors of metastasis in thyroid cancer. The AUC value of 0.950 for the prediction of thyroid cancer metastasis by ultrasound signs combined with serologic indicators was significantly higher than 0.728 and 0.711 predicted by ultrasound signs or serologic indicators alone.

CONCLUSION

The multivariate regression model incorporating ultrasound findings and serum markers enhances the predictive accuracy for thyroid cancer metastasis, offering essential guidance for early prediction and intervention in a clinical setting.