The ATA risk stratification system is commonly used to estimate the risk of disease recurrence [5], and DTC is managed postoperatively based on its recurrence risk pattern, categorized into low, intermediate, and high categories. Individual features are considered to accurately stratify the recurrence risk profile. Minimally invasive interventions are viable options for the lowest risk disease requiring interventional treatment. However, for high nodules, a more aggressive approach involving surgery combined with lymph node dissection is typically chosen [22]. Therefore, risk stratification of patients with thyroid cancer is crucial for accurate adjustment of treatment plan. In this study, we collected preoperative ultrasonic and cytologic features from 396 patients and performed univariate and multivariate analyses to make a systematic analysis of the relationship between ultrasonic and cytologic features and risk stratification of PTC. Our findings showed that Papillary arrangement, Escape like arrangement, Nucleolus, Size, Echo, Margin, and ECE were independent predictors for determining the risk stratification of patients with PTC. In addition, we devised a Nomogram model leveraging the identified metrics, as it allows for the integration of all these variables to individually estimate the risk stratification of each PTC patient. The Nomogram model, utilizing ultrasonic and cytologic features, exhibited strong discriminatory power, yielding an AUC of 0.799 for the ROC curve within the training cohort. Furthermore, in an external testing cohort comprising 104 PTC patients, the ROC curve achieved an AUC of 0.778. This model enables swift and straightforward estimation of the risk stratification for PTC patients. Notably, to our knowledge, this is the first study to amalgamate preoperative FNA and ultrasonic features for predicting the preoperative risk stratification of PTC.
High-resolution ultrasound is commonly regarded as the preferred method for preoperative assessment of thyroid tumors [23]. However, its effectiveness can be limited in visualizing deeper anatomical structures or those obscured by bone or airborne acoustics [5], resulting in lower diagnostic efficacy for ATA risk stratification. Numerous studies have utilized ultrasonic features to diagnose key risk stratification factors such as lymph node metastases and extrathyroidal extensions[24-26],These metrics rely on ultrasonic features to determine tumor aggressiveness. In our research, features including Size, Echo, Margin, and ECE were identified as independent predictors for PTC risk stratification. While ultrasonic features often reflect the macroscopic aggressiveness of a tumor, they may overlook crucial microscopic biological information. Cytologic images, on the other hand, contain subtle and complex details [20]. Recent studies have leveraged cytological features from FNA to assess tumor aggressiveness, with findings suggesting that different cytological findings correlate with distinct clinical tumor phenotypes. Notably, TIR3A/TIR3B cytology predicts a more inert behavior compared to TIR4/TIR5 cytology. Laura Croce posits that varying cytology examinations correspond to diverse clinical tumor phenotypes and that tumor cytology responds to tumor aggressiveness[27]. Our analysis of tumor cytological features through both unifactorial and multivariate approaches revealed a Papillary arrangement, Escape like arrangement, and Nucleolus as independent predictors of PTC risk stratification at the microscopic level.
Nomograms are widely employed predictive models in oncology, integrating key factors to visualize the probability of clinical outcomes [28]. Currently, there are no published Nomograms for predicting preoperative risk stratification in patients with PTC. We developed a Nomogram by amalgamating the crucial clinical features outlined earlier. The Nomogram illustrates the relative contribution of each factor to risk stratification, with Size emerging as the strongest predictor, followed by Nucleolus, Echo, ECE, escape like arrangement, Margin, and papillary like arrangement. The Nomogram enables the integration of all features, allowing for individualized estimation of ATA risk for each patient. Our Nomogram achieved a C-index of 0.799, signifying accurate prediction of PTC risk stratification; the calibration curve indicated good alignment with actual conditions. To validate the model's stability, we collected data from 104 PTC patients across two hospitals for testing; the Nomogram model also performed well in the testing cohort (AUC=0.778), confirming the robust diagnostic efficiency of the model. Leveraging the quantitative risk stratification provided by our Nomogram, clinicians can precisely identify PTC risk. Patients with a high Nomogram score may necessitate prophylactic cervical lymph node dissection and rigorous postoperative evaluation to enhance prognosis, whereas those with a low risk Nomogram score may consider active surveillance or ablative therapy.
Our study presents several limitations that warrant consideration. Firstly, due to its retrospective nature, the results are contingent upon the available data, and some inherent biases may have influenced the findings. Secondly, our study exclusively focused on patients diagnosed with PTC, thus potentially limiting the generalizability of our conclusions to other thyroid cancer subtypes.
In this investigation, we identified several key morphological features including Size, Nucleolus, Echo, ECE, Margin, Escape like arrangement, and Papillary like arrangement as pivotal factors for risk stratification among patients with PTC. Leveraging these factors, we developed a predictive model for PTC risk stratification, which exhibited robust discrimination, calibration, and clinical utility. Our model offers clinicians a reliable tool for accurately assessing PTC risk, thereby facilitating more precise treatment planning.