With the application of high-resolution ultrasound, thyroid nodules are found in 19%–67% of ultrasound exams (1). A portion of them (15%-53.8%) are mixed echoic nodules, which have both cystic and solid components (2, 3). Partially cystic thyroid nodules (PCTNs) are considered to result from cystic degeneration due to hemorrhage, ischemia, necrosis and liquefaction. Nodules with cystic degeneration in the thyroid gland, such as nodular goiters and thyroid adenoma, are mostly benign lesions. Previous studies have shown that the proportion of cystic components in nodules is inversely proportional to their malignant potential (3–5).

However, 13%-26% of thyroid carcinomas may have cystic degeneration (6). For PCTNs, the percentage of malignancy varies from approximately 2% to 18% (1, 2, 7). PCTNs are often underestimated. Currently, the ultrasound-guided fine needle aspiration (US-FNA) of thyroid nodules is recognized as the most reliable method for preoperative differential diagnosis. This method, however, may be limited by several factors, such as its invasiveness, insufficient sampling and/or indeterminate cytology, and operator dependence (8–10). Moreover, the presence of cystic components in PCTNs may further reduce the diagnostic accuracy of FNA (11). There are few studies on the US features related to malignant PCTN, and these features differ from those of solid nodules (12–15). Therefore, the purpose of this study was to determine the ultrasound features that predict the malignancy of PCTNs and to analyze the correlations to construct a scoring system and present it as a forest plot. Such a scoring system is expected to improve the preoperative diagnostic efficacy and to facilitate patient management in clinical practice.

As a subtype of thyroid nodules, PCTNs are fairly common and can be differentiated easily according to conventional US features. Most PCTNs are regarded as benign, as degenerative processes arise in underlying benign lesions, such as nodular goiters and adenomatous nodules (18), and true epithelial cysts are rare. However, previous studies have revealed cystic changes in many malignant thyroid nodules, and the incidence of PCTN malignancy is up to 18% (1). PCTNs are easy to misdiagnose, especially for less experienced radiologists, due to their typical US features being different from those of solid nodules. The clinical management of PCTNs is still a challenge.

Ultrasound plays an essential role in evaluating the status of PCTNs because of its economical, noninvasive, and convenient features. Many studies have summarized the malignant features of solid thyroid nodules, such as their taller-than-wide shape, hypoechogenicity, and microcalcifications, while no clear diagnostic criteria have been established for malignant PCTNs. The sizes, compression symptoms, rates of recurrence after repeated aspirations, and local invasion of PCTNs are considered to be indicators of surgery (3, 19, 20). However, no general clinical features completely distinguish benign PCTNs from malignant PCTNs (3, 20, 21). Among 119 patients with PCTNs who underwent thyroidectomy after FNA, Bellantone (3) reported that 21 were malignant, yielding a 17.6% malignancy rate, and he therefore recommended FNA for all PCTNs; however, the malignancy rate among PCTNs may have been overestimated due to selection bias for the enrolled patients. Nevertheless, the presence of cystic components in a thyroid nodule may reduce the diagnostic efficiency of FNA. In this case, US-FNA is no longer recommended as a routine diagnostic method for PCTNs (22). Frates MC et al. (2) and LI W et al. (11) found that US features can help prioritize PCTNs for FNA. The main purpose of this study was to elucidate the US features that can predict PCTN malignancies and to construct a scoring system based on these US features to guide clinical practice.

Many studies have revealed the US features of malignant solid thyroid nodules, mainly including their echogenicity, margin, shape, microcalcification, and vascularity, which were all included in this study. In addition, PCTNs have unique nodule compositions, solid portion positions and free margins, and these characteristics were therefore also included in this study. The eccentric configuration of PCTNs was reported to indicate malignancy in a previous study (19); however, the sensitivity (44.4%) and positive predictive value (13.1%) were low. We further subdivided the eccentric configuration based on the angle to the adjacent cyst wall and found that only an acute angle was associated with malignancy (p <0.001), which was confirmed in the reports (12, 23). This phenomenon may be attributed to malignant PCTNs being more likely to protrude from the cystic wall, while diffuse lesions are circumferentially located within the cystic wall, forming a concentric configuration (24). Significantly, some researchers (13, 25) reported that when patients were regrouped into the ≥50% solid portion and <50% solid portion groups, an eccentric position of the solid component was a significantly malignant feature in only the solid portion ≥50% group. In our study, however, we did not find that a solid portion ≥50% was significantly associated with malignancy. This indicates that we should pay special attention to nodules that have both a solid portion ≥50% and an eccentric configuration.

A taller-than-wide shape was revealed to be an independent predictor of PCTN malignancy in our study. Previous studies found that the sensitivity and specificity of a taller-than-wide shape for the diagnosis of malignancy were 43.33% and 91.36%, respectively (26, 27). In contrast, some studies (12, 28) reported no significant correlation between a taller-than-wide shape and the risk of PCTN malignancy, which they suggested was potentially attributed to the inter- and intraobserver variability in taller-than-wide shapes.

In our study, we demonstrated that nonsmooth free margins were significantly more common in malignant thyroid nodules than in benign nodules. This may be illustrated by the histological tendencies of malignant PCTNs, including their uneven growth in an infiltrative manner and absence of pseudocapsule formation (14). Regarding the margins of entire nodules, no significant differences were found between benign and malignant PCTNs. Because large amounts of cystic fluid from benign lesions can be absorbed in a short time, the margin of the entire nodule is either smooth or ill-defined. Moreover, some malignant PCTNs can have smooth margins, especially those that are less than 10 mm in diameter.

It is well known that the presence of microcalcifications is highly specific for papillary thyroid carcinoma (14, 19). In this study, we defined microcalcifications as those less than 1 mm in size on US images; while this definition may not represent the actual pathological condition, all calcifications were confirmed by pathology. In this way, we excluded punctuate hyperechoic foci with comet tails, which may be another factor underlying misdiagnosis. Pathologically, nodule microcalcifications are correlated with psammoma bodies, which result from the degeneration of tumor cells and collagen produced by tumor cells (29).

Color Doppler ultrasound is a diagnostic tool for predicting thyroid carcinoma based on the hypothesis that peripheral or intranodular flow is caused by the angiogenesis-dependent growth of malignant tumors, while degenerating cystic lesions mainly present as the absence of internal flow signals or a small amount of peripheral flow. In this study, no significant differences were found among nodules with three different vascular patterns, which was possibly caused by rare blood flow signals detected in nodules during the period of stagnant growth (27).

To our knowledge, the risk of malignancy increases as more suspicious US features are detected in nodules (19). In our study, a logistic regression model was developed to assess the malignancy of PCTNs based on multiple ultrasonographic features. To make the model simple and convenient for clinical application, we proposed and validated a scoring system to predict the malignancy of PCTNs, which was based on the logistic regression model, achieving accuracies of 0.876 in the primary cohort and 0.829 in the validation cohort. A series of different TI-RADS guidelines have been proposed to evaluate solid thyroid nodules, although ACR TI-RADS has indeed taken the PCTN into consideration, but the ultrasound features in this guideline are mainly referred to solid thyroid nodules and the solid component of PCTN. This guideline has not fully considered the ultrasound features specific and important to PCTNs, such as free-margin, position of solid portion, and composition. Therefore, the strength of this scoring system was that it used eight common PCTN US features to construct the model. The scoring system displayed a remarkable ability to predict the malignancy of PCTNs (AUC=0.876), yielding a sensitivity of 71.05% and a specificity of 87.63%. These results indicate the potential value of our model for decreasing the rate of misdiagnosis and the use of US-FNA of PCTNs prior to surgery. Therefore, clinicians can now efficiently manage PCTN patients.

In univariate analysis, we did not find that nodule size was significantly associated with malignancy of PCTNs, and the finding was in accordance with previous studies (11, 19, 30). Hence, nodule size has not been described as a risk factor of malignant PCTNs. Nevertheless, size was considered as an important parameter to predict the malignancy in solid thyroid nodules. At present, many studies on nodule size are available, but no consensus as to its influence on probability of malignancy. Some studies have demonstrated that larger lesions had lower risk of malignancy (31, 32), while the other studies reached opposite conclusions (33, 34). The majority studies (31, 33) indicates that size threshold for malignant risk was approximately 2.0 cm in solid nodules, and cutoff value 2cm have been adopted in the study of PCTNs (12). Therefore, we divided the PCTNs into two groups to evaluate the impact of nodule size on the diagnostic efficacy of the scoring system. It was found that the effect of nodule size on the scoring system was nonsignificant, and the scoring system exhibited a high diagnostic accuracy in 2 groups of nodules with different sizes.

There are several limitations in our study. First, this study included some benign PCTNs for which only cytological results were available; for these lesions, pathological confirmation was absent or only a limited follow-up evaluation was performed. Second, we did not include all benign PCTNs. All nodules included in the study were removed by FNA or surgery due to suspicious ultrasound features or clinical indications. Therefore, we could not estimate the percentage of malignant PCTNs in our study, which may have been affected by selection bias to some extent. Third, this study did not investigate the inherent relationship between suspicious US features, like in the study of Zhang M et al. (22), but it may provide some new insights due to the more detailed classification of US features.

In conclusion, a scoring system based on the US features of PCTNs was developed to predict the risk of malignant PCTN in this study. This model provides a noninvasive, effective, and easy-to-use method that can be clinically implemented to identify malignant PCTNs preoperatively. This method is expected to assist radiologists in diagnosis and clinical decision making.

Due to the privacy of patients, the data related to patients cannot be available for public access but can be obtained from the corresponding author on reasonable request approved by the institutional review board of First Hospital of Shanxi Medical University.

This retrospective study was approved by the Ethics Review Committee of First Hospital of Shanxi Medical University. The requirement for informed consent of patients was waived.

YX, LL, and JZ designed the study. YX, FL, YS, and XY collected the clinical data, ultrasonography images and histopathological/cytopathology results. FL and YX processed the clinical and images data. YS and YX performed the statistical analysis. YX and FL drafted and revised the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by Application Basic Research Project of Science and Technology Department of Shanxi Province (201801D121340), and Key Research and Development Program of Science and Technology Department of Shanxi Province (201903D321190).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.