Preoperative Prediction of Lymph Node Metastasis in T1N0M0 Papillary Thyroid Carcinoma by Using Contrast-enhanced Ultrasound

Papillary thyroid carcinoma (PTC) is the most prevalent subtype of thyroid malignancy, accounting for approximately 80%-90% of all thyroid cancers globally. With the widespread application of high-resolution ultrasound in routine physical examinations, an increasing number of early-stage PTC cases, particularly those classified as T1N0M0 (tumor size ≤2 cm, no clinical evidence of lymph node metastasis, and no distant metastasis), are being detected. Despite the relatively favorable prognosis of T1N0M0 PTC, the presence and burden of central lymph node metastasis (CLNM) remain critical factors that directly guide individualized treatment strategies and long-term follow-up plans for patients.

Similarly, the burden of CLNM exerts a pivotal impact on the individualized management of T1N0M0 PTC patients, as distinct metastasis burdens correspond to significantly different therapeutic approaches. Specifically, lymph node metastasis of any burden serves as an absolute contraindication to thermal ablation, a minimally invasive treatment option increasingly used for low-risk PTC; patients with low-burden CLNM (defined as ≤3 positive central lymph nodes with no extracapsular extension) are typically suitable for thyroid lobectomy combined with central lymph node dissection, which balances tumor control and preservation of thyroid function; in contrast, high-burden CLNM (defined as >3 positive central lymph nodes or presence of extracapsular extension) recommends total thyroidectomy with comprehensive central lymph node dissection, followed by radioactive iodine therapy when necessary to reduce the risk of recurrence. However, conventional preoperative imaging examinations, including conventional ultrasound, computed tomography (CT), and magnetic resonance imaging (MRI), have inherent limitations in diagnosing CLNM and evaluating its burden. Conventional ultrasound relies primarily on morphological features (e.g., lymph node size, shape, cortical thickness) which lack sufficient specificity, while CT and MRI are associated with radiation exposure (for CT) or high cost (for MRI) and still demonstrate low sensitivity and specificity for detecting small or micrometastatic lymph nodes in the central compartment, leading to frequent misdiagnosis or underdiagnosis.

This limitation poses a clinical dilemma: inaccurate preoperative assessment of CLNM burden may result in overtreatment (e.g., unnecessary total thyroidectomy in patients with no or low-burden metastasis, leading to lifelong thyroid hormone replacement therapy) or undertreatment (e.g., lobectomy in patients with high-burden metastasis, increasing the risk of local recurrence and the need for reoperation). Therefore, there is an urgent clinical need for a non-invasive, accurate, and cost-effective imaging method to preoperatively evaluate the burden of CLNM in T1N0M0 PTC patients.

Dynamic contrast-enhanced ultrasound (CEUS) has emerged as a promising imaging technique in the field of thyroid and lymph node diagnostics. By intravenously injecting a microbubble contrast agent, CEUS can dynamically visualize the microvascular perfusion of target tissues in real time, providing qualitative (e.g., enhancement pattern, homogeneity, presence of perfusion defects) and quantitative (e.g., time to peak, peak intensity, wash-in rate, wash-out rate) parameters that reflect the pathological microenvironment of lesions. Compared with conventional imaging modalities, CEUS offers several advantages, including non-radiation exposure, real-time imaging, high spatial and temporal resolution, and the ability to assess microcirculation, which may enable more accurate identification of metastatic lymph nodes and evaluation of their burden.

To address the aforementioned clinical gap, this multicenter retrospective cross-sectional study was designed to explore the value of CEUS in preoperatively predicting the burden of CLNM in T1N0M0 PTC patients. The study enrolled 600 consecutive patients with T1N0M0 PTC who were admitted to three tertiary hospitals (the lead center and two collaborating centers) from June 2018 to June 2025 and confirmed by postoperative pathology. All patients met strict inclusion criteria: (1) histopathologically confirmed PTC after surgery; (2) preoperative tumor staging consistent with T1N0M0 based on clinical examination and conventional imaging; (3) no history of thyroid surgery, radioactive iodine therapy, or other malignant tumors; (4) availability of complete preoperative CEUS data of the thyroid and central compartment lymph nodes; and (5) no contraindications to contrast agent injection. Patients with incomplete clinical, imaging, or pathological data were excluded from the study.

Two senior ultrasound physicians with more than 10 years of experience in thyroid and lymph node imaging, who were completely blinded to the postoperative pathological results (including the presence and burden of CLNM), independently analyzed the stored CEUS images. They extracted a comprehensive set of qualitative and quantitative CEUS features. Qualitative features included: (1) enhancement pattern of the thyroid lesion (homogeneous vs. heterogeneous, hyperenhancement vs. hypoenhancement vs. isenhancement compared with surrounding normal thyroid tissue, presence of peripheral rim enhancement); (2) enhancement characteristics of suspicious lymph nodes (homogeneous vs. heterogeneous enhancement, presence of perfusion defects, cortical enhancement, medullary hilum disappearance); and (3) time sequence of enhancement (whether the lesion/lymph node enhances earlier or later than normal tissue). Quantitative features were measured using dedicated offline analysis software, including time to peak (TTP), peak intensity (PI), wash-in slope (WIS), wash-out rate (WOR), and area under the time-intensity curve (AUC-TIC) for both the thyroid lesion and suspicious lymph nodes.

Taking postoperative pathological results as the gold standard (the gold standard for CLNM diagnosis and burden assessment), patients were divided into three groups: (1) non-metastasis group: no positive central lymph nodes confirmed by pathology; (2) low-burden metastasis group: ≤3 positive central lymph nodes with no extracapsular extension; (3) high-burden metastasis group: >3 positive central lymph nodes or presence of extracapsular extension (extracapsular extension was defined as the infiltration of metastatic lymph node tissue beyond the lymph node capsule into adjacent soft tissues). Pathological examination was performed by two senior pathologists who were blinded to the CEUS results, and any discrepancies were resolved through consensus discussion.

Statistical analysis was performed using SPSS 26.0 and R 4.2.0 software. First, inter-observer agreement for qualitative and quantitative CEUS features was evaluated using the kappa coefficient (for qualitative features) and intraclass correlation coefficient (ICC) (for quantitative features), with kappa/ICC >0.75 indicating excellent agreement. Features with excellent inter-observer agreement were included in subsequent analyses. Then, univariate Logistic regression analysis was used to screen potential predictors of CLNM burden (i.e., comparing non-metastasis vs. metastasis, and low-burden vs. high-burden metastasis). Variables with a P value <0.1 in univariate analysis were included in multivariate Logistic regression analysis to identify independent predictors of CLNM burden. Based on the independent predictors, a combined predictive model was constructed, and the regression coefficient of each predictor was used to assign a weight to generate a risk score for CLNM burden.

To evaluate the diagnostic efficacy of the combined model, receiver operating characteristic (ROC) curves were drawn, and the area under the curve (AUC) was calculated as the primary outcome measure. The optimal cut-off value of the risk score was determined using the Youden index (sensitivity + specificity - 1). Secondary outcome measures included sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy, which were calculated based on the optimal cut-off value. Additionally, decision curve analysis (DCA) was performed to evaluate the clinical practicality of the model by quantifying the net benefit of using the model to guide clinical decision-making compared with the "treat all" or "treat none" strategies across different threshold probabilities.

This study is expected to achieve several key objectives: first, to identify the most valuable qualitative and quantitative CEUS features for predicting the burden of CLNM in T1N0M0 PTC; second, to construct a reliable combined predictive model based on these CEUS features that can accurately distinguish between non-metastasis, low-burden metastasis, and high-burden metastasis preoperatively; third, to validate the diagnostic efficacy and clinical applicability of the model through ROC and DCA analyses. Ultimately, this project aims to provide a non-invasive, accurate, and practical tool for preoperative assessment of CLNM burden in T1N0M0 PTC patients, thereby guiding clinicians to formulate more precise and individualized treatment plans, reducing the risk of overtreatment or undertreatment, improving patient prognosis, and optimizing the use of medical resources.