Improving Diagnostic Accuracy of Indeterminate Thyroid Fine-Needle Aspiration Using Results of Ultrasound Examination, Cyclin D1 Immunostaining, and Molecular Testing

Abstract

Introduction

Fine-needle aspiration (FNA) is the standard procedure and routine method for diagnosing thyroid nodules. However, indeterminate cytological results diagnoses present considerable challenges in clinical management. This study aimed to evaluate cyclin D1 immunocytochemistry and the Chinese thyroid imaging reporting and data system (C-TIRADS) score, individually and in combination with a simple gene test for cytological diagnosis of Bethesda category III-Ⅴ.

Methods

A consecutive cohort of 177 thyroid FNA specimens with indeterminate diagnosis and available histopathologic follow-up data was collected. The samples were evaluated by cyclin D1 immunocytochemistry and molecular testing for the BRAFV600E mutation or a small panel of markers (BRAF, NRAS, HRAS, KRAS, and TERT). Two experienced sonographers independently reviewed original sonographic images of each nodule and used C-TIRADS to classify the images, obtaining a consensus C-TIRADS score. Identification of the optimal cut-off points of cyclin D1 and C-TIRADS score for the diagnosis of malignancy was evaluated using the receiver operating characteristic (ROC) curves and the assessment of the area under the ROC curve (AUC). The specificity, sensitivity, positive predictive value (PPV) and negative predictive value (NPV) of all tests were evaluated with crosstabs. Logistic regression analysis and ROC curve analysis were used to evaluate the diagnostic accuracy of the three tests individually and in combination.

Results

We enrolled 169 patients (177 thyroid nodules), including 140 malignancies, 15 low-risk neoplasms, and 22 benign lesions. All 177 specimens from the nodules were tested for BRAF V600E, while only 21 specimens adopted the 5-gene detection protocol. With cut-off values set at 10% for cyclin D1 immunocytochemistry and at a C-TIRADS score of 3, along with defining a positive diagnosis as the presence of a mutation in genetic testing, the PPVs for diagnosing thyroid malignancy using the three tests were 97.8%, 90.3%, and 98%, respectively; however, the NPVs were 50%, 24.2%, and 26.3%, respectively. The sensitivities were 87.7%, 83.9%, and 63.9%, and the specificities were 86.4%, 36.4% and 90.9%, respectively. Regarding AUC, cyclin D1 alone demonstrated greater diagnostic accuracy (AUC = 0.921, 95% CI = 0.857–0.985, p = 0.000) compared with the other two methods, which had AUC values of 0.775 (95% CI = 0.703–0.846, p = 0.000) and 0.587 (95% CI = 0.443–0.731, p = 0.235). The combination of two or three tests yielded higher accuracy (AUC = 0.929, 95% CI = 0.873–0.985, p = 0.000; AUC = 0.925, 95% CI = 0.860–0.989, p = 0.000; AUC = 0.937, 95% CI = 0.889–0.985, p = 0.000) compared with cyclin D1 immunocytochemistry alone (AUC = 0.921, 95% CI = 0.857–0.985, p = 0.000).

Conclusion

The addition of cyclin D1 immunocytochemistry or a simple gene panel test, either alone or in combination, significantly improves the limited diagnostic utility of merely repeating the C-TIRADS score in thyroid nodules with indeterminate cytology. Laboratories should integrate these tests to effectively manage uncertain thyroid nodules and reduce diagnostic ambiguity.

Introduction

During ultrasound examinations, thyroid nodules can be detected in 19%–68% of the population, with most being benign and clinically insignificant [1]. Ultrasound-guided fine-needle aspiration (FNA) is the most common and cost-effective method for diagnosing high-risk nodules before surgery. The American National Cancer Institute established the Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) to standardize the diagnostic terminology and criteria for thyroid FNA cytology [2, 3], ensuring effective communication of results to referring physicians. However, the three ‘‘indeterminate’’ cytopathologic interpretations – atypia of undetermined significance (AUS, TBSRTC III), follicular neoplasm (FN, TBSRTC IV), and suspicious for malignancy (TBSRTC V) – account for 20%–30% of all thyroid FNAs [3, 4]. Despite many patients undergoing repeated FNAs or diagnostic surgeries, Bongiovanni et al. [5] found that only 14.4%–32.1% of TBSRTC III and IV nodules were malignant, while 25.1% of TBSRTC-V nodules were benign.

Cytopathologists face challenges in managing three types of indeterminate cytological diagnoses and in reducing unnecessary thyroidectomies. To address the limitations of cytological diagnosis, the 2017 and 2023 versions of TBSRTC recommend molecular testing for managing Grade III–IV cases [2, 3]. However, apart from the common BRAF gene mutation, which has clear diagnostic significance, mutations in genes such as RAS, RET/PTC, and PAX8/PPARγ are less specific for tumor diagnosis due to their low incidence and potential occurrence in benign proliferative follicular nodules or adenomas [6, 7]. Therefore, molecular detection alone is insufficient for triaging patients with indeterminate cytology. Additionally, some commercially available multigene panels, such as ThyroSeq Genomic Classifier and Afirma Gene Sequencing Classifier, are expensive. Their high cost significantly limits their use in clinical practice.

Ultrasound assessment is easier to perform and does not add extra medical costs compared with molecular detection. In a meta-analysis, Gao et al. showed that ultrasound was helpful in differentiating thyroid nodules with indeterminate cytology [8]. Similarly, immunostaining is simple and economical compared with molecular detection. Certain immunohistochemical markers, including galectin-3, Hector Battifora mesothelial cell-1 (HBME-1), and cytokeratin 19, are used to identify papillary thyroid carcinoma (PTC) and differentiate it from benign mimics [9]. PTC showed overexpression of CD44 and cyclin D1 compared with benign thyroid lesions [10]. None of these markers can provide a definitive diagnosis when used alone. This cohort study aimed to evaluate the efficacy of ultrasound classification, immunocytochemical staining, and molecular analysis (BRAFV600E or a five-gene panel) both individually and in combination for assessing TBSRTC III–V nodules.

Materials and Methods

Patients and Specimens

We retrospectively identified all patients who underwent ultrasound-guided thyroid FNA with indeterminate cytological diagnoses, as per the 2017 TBSRTC, using electronic database search methods from December 2018 to December 2022. Two experienced pathologists reviewed the records of all these histologically confirmed patients. Using histopathological analysis after surgery as the reference standard, we identified malignant and low-risk neoplasms as malignancies, while categorizing the remaining cases as benign. This study received approval from the Institutional Review Board of Beijing Hospital, which also granted a waiver for patient informed consent regarding the use of cytology material for ancillary testing.

Ultrasound and Color-Flow Doppler

All grayscale and Doppler sonographic examinations were performed with a 10–13-MHz linear probe (LOGIQ E9/E8 XD-Clear; GE Healthcare, USA) by two radiologists with more than 10 years of experience in thyroid sonography. Before and at the time of FNA, all nodules were routinely submitted to ultrasound risk assessment according to the Chinese guidelines for ultrasound malignancy risk stratification of thyroid nodules (Chinese thyroid imaging reporting and data system [C-TIRADS]) [11].

Cytological Examination

FNA was performed by radiologists using a 25-gauge needle attached to a 10-mL syringe. Each lesion was aspirated with four to five consecutive passes, and conventional smears were prepared; the material remaining after smearing was stored in PreservCyt solution and processed using the ThinPrep 5000 liquid-based cytology (LBC) method (Hologic Co., Marlborough, MA, USA). Each smear or LBC slide was fixed in 95% ethanol. Conventional smears were stained with hematoxylin and eosin, whereas LBC slides were stained with Papanicolaou stain. Two cytopathologists made the cytological diagnosis based on the TBSRTC 2017 and jointly consulted to obtain consensus when there were disagreements.

Manual Macrodissection of Slides

Prior to the surgical procedure, the molecular tests were conducted as described [12]. Suspicious cells were identified and selected by cytopathologists at the microscope to ensure adequate material and to confirm the cytological diagnosis. The lower limit of adequacy for each slide was established according to TBSRTC with 6 groups of thyrocytes with at least 10 well-visualized epithelial cells per group. After choosing the most representative slide or after marking the regions of the slide containing adequate atypical thyrocytes by pathologists, the slides were kept in xylene until the coverslips detached. The slides were then hydrated in a graded series of ethanol baths, followed by a wash in dH₂O for 1 min, and finally air dried. Then DNA extraction was performed using a commercially available kit (AmoyDx, Xiamen, China). Fifty percent of the lysis solution without proteinase K was initially poured on the slide to scrape off the cytological sample using a single-edged razor blade. Scraped cells were collected into 1.5 mL Eppendorf tubes containing the other half of the lysis solution with proteinase K. The extracted DNA was kept at –20°C until used.

Immunocytochemical Staining

Immunocytochemistry with cyclin D1 antibody (Beijing Zhongshan Golden Bridge Biotechnology Co. Ltd.) was performed on destained slides on an automated stainer (Ventana Medical Systems, Tucson, AZ, USA) according to the manufacturer’s protocol. The staining of cyclin D1 was observable in both the nucleus and cytoplasm. Only thyrocytes with positive nuclei within the designated area were counted, and the percent was calculated. The counting interval was set at 5% for the convenience. The immunocytochemistry results were evaluated by two pathologists who were blinded to the histological diagnosis and molecular status.

Molecular Detection

The detection of the BRAFV600E mutation or 5-gene mutation panel (BRAF, NRAS, HRAS, KRAS, and TERT) was performed using a validated fluorescent PCR detection kit (ADx-ARMS, Amoy Diagnostics Co., Ltd.) approved by the China Food and Drug Administration and based on the Amplification Refractory Mutation System (ARMS). For each sample, an external control assay and a mutation assay were performed in parallel. Each run consisted of a negative control and a positive control. Thermocycling conditions were as follows: stage 1, 5 min at 95°C; stage 2, 25 s at 95°C, 20 s at 64°C, and 20 s at 72°C, repeated for 15 cycles; and stage 3, 25 s at 93°C, 35 s at 60°C, and 20 s at 72°C, repeated for 31 cycles. Data were collected at 60°C in stage 3 using the Life Technologies Prism 7500 series real-time PCR instrument. Run files were analyzed and interpreted according to the manufacturer's instructions.

Statistical Analysis

Statistical analyses were performed using SPSS version 26.0 (IBM Corporation, Armonk, NY, USA). Receiver operating characteristic (ROC) curves were used to assess optimal cut-off points for both the C-TIRADS score and the percent of cyclin D1-positive thyrocytes when diagnosing malignant or low-risk neoplasms, while also evaluating the area under the ROC curve (AUC). The specificity, sensitivity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of C-TIRADS and cyclin D1 were evaluated using crosstabs based on the selected cut-off points. Logistic regression analysis along with ROC curve analysis was performed to compare the diagnostic accuracy of C-TIRADS, cyclin D1 immunostaining, and molecular detection, both individually and in combination. A p < 0.05 was considered statistically significant.

Results

Clinical Presentation

Over the study period, we collected 5,640 thyroid ultrasound-guided FNA specimens. According to the 2017 TBSRTC classification, 919 (16.3%) specimens were categorized as TBSRTC III–V, including 376 TBSRTC III (6.7%), 127 TBSRTC IV (2.2%), and 416 TBSRTC V (7.4%). Of the 919 specimens, histological confirmation was obtained for 177 nodules, which included 138 malignant, 17 low-risk, and 22 benign nodules. This cohort consisted of 169 patients with an average age at diagnosis of 48.0 ± 11.7 years (range: 23–83 years), comprising 39 men (23%) and 130 women (77%). The average longest diameter of the nodules was 9.3 ± 9.1 mm (range: 1–45 mm). Most nodules were not clinically apparent, primarily being microcarcinomas, and were detected incidentally. The diagnoses included 135 PTCs, 3 medullary thyroid carcinomas (MTCs), and 2 follicular thyroid carcinomas (FTCs) as malignant neoplasms. The low-risk neoplasms included 9 follicular tumors with uncertain malignant potential (FT-UMP), 2 well-differentiated tumors with uncertain malignant potential (WDT-UMP), 3 noninvasive follicular thyroid neoplasms with papillary-like nuclear features (NIFTP), and 1 hyalinizing trabecular tumor (HTT). Additionally, the diagnoses encompassed benign lesions such as nodular goiter (n = 10), Hashimoto’s thyroiditis (n = 8), and adenoma (n = 4) (Fig. 1–3). The corresponding categorization of the 177 thyroid FNA specimens is presented in Table 1.

Fig. 1.

The sheet-arranged follicular epithelial cells have various sizes of nuclei, with irregular nuclear membranes and some longitudinal nuclear grooves (AUS, TBSRTC III). The nodule was confirmed as an adenomatous hyperplasia nodule by histopathological examination.

Fig. 3.

The sheet-arranged follicular epithelial cells have various sizes of nuclei, with irregular nuclear membranes and some longitudinal nuclear grooves (AUS, TBSRTC III).

Table 1.

TBSRTC categorization corresponding to the results of BRAF testing, 5-gene testing, and cyclin D1 immunostaining

TBSRTC	BRAF	Histological diagnosis
		malignant/low risk (n = 155)	benign (n = 22)	total (n = 177)
III	Mutant	40	0
	Non-mutant	30	11	81
IV	Mutant	1	0
	Non-mutant	11	6	18
V	Mutant	50	0
	Non-mutant	23	5	78

	5-gene
III	Mutant	6	0
	Non-mutant	1	0	7
IV	Mutant	5	1
	Non-mutant	4	1	9
V	Mutant	1	0
	Non-mutant	2	0	3

	Cyclin D1
III	≥10%	61	2
	<10%	9	9	81
IV	≥10%	9	1
	<10%	3	5	18
V	≥10%	66	0
	<10%	7	5	78

Fig. 2.

The follicular epithelial cells partially overlap and are arranged in a microfollicular pattern, exhibiting slightly enlarged nuclei (suspicious for follicular neoplasm, TBSRTC IV).

Ultrasound C-TIRADS Classifications

The benign group comprised 2 specimens classified as C-TIRADS 3, 6 specimens as C-TIRADS 4a, 3 specimens as C-TIRADS 4b, 7 specimens as C-TIRADS 4c, and 4 specimens as C-TIRADS 5. The malignant or low-risk group comprised 1 specimen classified as C-TIRADS 3, 24 specimens as C-TIRADS 4a, 25 specimens as C-TIRADS 4b, 83 specimens as C-TIRADS 4c, and 22 specimens as C-TIRADS 5. The nodules classified as C-TIRADS 3 were assigned a score of 1, and those categorized as C-TIRADS 4a received a score of 2. Nodules falling under the category of C-TIRADS 4b were allocated a score of 3, and those belonging to the category of C-TIRADS 4c were given a score of 4. The highest score of 5 was assigned to nodules classified as C-TIRADS 5 (Fig. 4–6). The ROC curve analysis indicated an AUC of 0.587 (95% CI = 0.440–0.734, p = 0.187). We determined the optimal cut-off value to be 2.5, where a score of ≥3 indicated malignancy or low-risk neoplasm, and a score of <3 suggested a benign lesion. In terms of reclassified performance for the nodules with indeterminate cytology, the C-TIRADS score had an 83.9% sensitivity, a specificity of 36.4%, a PPV of 90.3%, a NPV of 24.2%, and a diagnostic accuracy of 78% in diagnosing malignant/low-risk neoplasms, with histopathological analysis after surgery serving as the reference standard.

Fig. 4.

The same specimen depicted in Figure 1. A hypoechoic nodule is observed in the superior aspect of the isthmus of the left thyroid lobe, measuring 5.6 mm × 4.1 mm. It exhibits well-defined margins, an irregular shape, and heterogeneous internal echogenicity (C-TIRADS 4a).

Fig. 6.

The same specimen depicted in Figure 3. The upper right lobe of the thyroid gland shows a small, hypoechoic nodule measuring 2.8 mm × 4.1 mm. The boundary is indistinct, the shape is irregular, and the aspect ratio exceeds 1; no punctate echogenic foci are observed within the nodule (C-TIRADS 4c).

Fig. 5.

The same specimen depicted in Figure 2. The middle and upper part of the left lobe exhibited a well-defined, regular-shaped nodule measuring 9.4 mm × 8.1 mm with markedly hypoechoic characteristics (C-TIRADS 4a).

Cyclin D1 Immunocytochemical Staining

Eighteen nodules, 14 benign and 4 malignant/low-risk, exhibited a cyclin D1 positivity of 0%. Twenty nodules, including 5 benign and 15 malignant/low-risk, had cyclin D1 positivity levels between 0% and 10%. There were 139 nodules with cyclin D1 positivity of 10% or higher. Among these, only three were benign, while the remaining 136 were classified as malignant or low-risk (Fig. 7–9). The ROC curve analysis revealed an AUC of 0.921 (95% CI: 0.857–0.985, p = 0.000). The optimal cut-off value for the percent of cyclin D1-positive thyrocytes was 7.5%; a percentage ≥10% indicated malignancy or low-risk neoplasm, and a percentage of <10% suggested a benign lesion. The cyclin D1 immunostaining showed a diagnostic accuracy of 87.6% for malignant/low-risk neoplasms, with a sensitivity at 87.7% and specificity at 86.4%. The PPV was 97.8%, and the NPV was 50.0% (Tables 1, 2).

Fig. 7.

The same specimen depicted in Figure 1. The cyclin D1 immunostaining of the follicular cells was negative. The nodule was confirmed as an adenomatous hyperplasia nodule by histopathological examination.

Fig. 9.

The same specimen depicted in Figure 3. The cyclin D1 immunostaining of the follicular cells was positive. The nodule was confirmed as PTC by histopathological examination.

Table 2.

Diagnostic characteristics of cyclin D1 immunohistochemistry, gene test, C-TIRADS, and combination of two or three tests for confirmation of malignancy of indeterminate cytology

	Accuracy, %	Sensitivity, %	Specificity, %	PPV, %	NPV, %	AUC (95% CI)	p value
Cyclin D1	87.6	87.7	86.4	97.8	50	0.921±0.033 (0.857–0.985)	0
Gene test	67.2	63.9	90.9	98	26.3	0.775±0.036 (0.703–0.846)	0
C-TIRADS	78	83.9	36.4	90.3	24.2	0.587±0.073 (0.443–0.731)	0.235
Cyclin D1+gene	91	98.6	81.8	97.3	60	0.929±0.029 (0.873–0.985)	0
Cyclin D1+C-TI-RADS	86.4	85.8	90.9	98.5	47.6	0.925±0.033 (0.860–0.989)	0
C-TIRADS+gene	67.2	63.9	90.9	98	26.3	0.811±0.042 (0.729–0.893)	0
3-test combination	91	91.6	86.4	98	59.4	0.937±0.024 (0.889–0.985)	0

Fig. 8.

The same specimen depicted in Figure 2. The cyclin D1 immunostaining of the follicular cells was positive. The nodule was confirmed as a noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) by histopathological examination.

Molecular Detection

We conducted the BRAF V600E mutation test on 177 specimens, with 91 specimens (51.4%) identified as mutant and histologically confirmed to be PTC. The remaining 86 specimens (48.6%) were classified as wild-type. Only 21 specimens were tested with the 5-gene panel, and 11 (52.4%) showed mutations. Among the 11 specimens, 5 had mutations in the NRAS gene. One specimen was diagnosed as follicular cell adenoma, while 4 specimens were classified as FT-UMPs. One specimen demonstrated a mutation in the HRAS gene and was histologically confirmed as follicular thyroid adenoma. One specimen displayed a mutation in the KRAS gene and was diagnosed with an FT-UMP. One specimen exhibited a mutation in the TERT gene and was histologically confirmed as FTC. The presence of both HRAS and TERT mutations was observed in 2 specimens, with histological confirmation of one specimen as FTC and the other specimen as a FT-UMP. Additionally, there was 1 specimen with both BRAF and TERT mutations, which was histologically confirmed as PTC. By combining the results of BRAF V600E and the 5-gene tests, we considered any identified mutation to be a positive genetic test. The ROC curve had an AUC of 0.775 (95% CI: 0.703–0.846, p = 0.000), indicating an accuracy of 67.2% for malignant/low-risk neoplasms, with a sensitivity of 63.9%, specificity of 90.9%, PPV of 98.0%, and NPV of 26.3% (Tables 1, 2).

Combined Tests

Logistic regression and ROC curve analysis revealed that combining the cyclin D1 immunocytochemical test with the genetic test achieved an AUC of 0.929 (95% CI: 0.873–0.985, p = 0.000) for the diagnosis of thyroid malignant/low-risk neoplasms, with an accuracy of 91.0%, sensitivity of 98.6%, specificity of 81.8%, PPV of 97.3%, and NPV of 60.0%. By combining the cyclin D1 immunocytochemical test with the C-TIRADS score, we found that the AUC was 0.925 (95% CI: 0.860–0.989, p = 0.000), with an accuracy of 86.4%, sensitivity of 85.8%, specificity of 90.9%, PPV of 98.5%, and NPV of 47.6%. By integrating the genetic test with the C-TIRADS score, we achieved an AUC of 0.811 (95% CI: 0.729–0.893, p = 0.000), with a sensitivity of 63.9%, specificity of 90.9%, PPV of 98.0%, and NPV of 26.3%. Additionally, combining cyclin D1 immunocytochemical testing, the C-TIRADS score, and simple molecular testing enabled us to accurately diagnose 91% of specimens, with sensitivity and specificity recorded at 91.6% and 86.4%, respectively. Furthermore, the PPV reached 98.0%, and NPV stood at 59.4%. The calculated AUC was 0.937 (95% CI: 0.889–0.985; p = 0.000) (Table 2).

Discussion

The incidence of thyroid cancer, the most common malignant tumor of the endocrine system, is increasing annually by 14.51% in China. This rise is accompanied by a corresponding increase in the mortality rate of 1.42% [13, 14]. There are no specific limits on the maximum diameter of nodules for ultrasound-guided FNA in China. It is emphasized that for nodules <5 mm, particularly those located in challenging anatomical sites or difficult-to-access areas, FNA should be performed by skilled physicians or under their expert supervision [15]. Therefore, most specimens in our study had nodules smaller than 1.0 cm, with a median maximum diameter of 9.3 mm (range: 1–45 mm).

Ultrasound-guided FNA is the most commonly used and cost-effective method for the preoperative diagnosis of thyroid malignancies [4]. Implementation of TBSRTC has improved the quality of FNA reporting and prevented the overtreatment of thyroid nodules [16]. According to a retrospective analysis of comprehensive data published by Chen et al. [17], the prevalences of class III, IV, and V diagnoses based on the TBSRTC system were 6.46%, 3.16%, and 6.72%, respectively. Our findings align with these statistics: TBSRTC Class III had a prevalence of 6.7%, Class IV was 2.2%, and Class V was 7.4%. This suggests that specimens with indeterminate cytology represent roughly 16% of all thyroid FNAs. These data differ from the published data of studies outside of China [5, 16]. The main distinction lies in the slightly higher proportion of Class III and V observed in domestic data, whereas the incidence of class IV, which corresponds to follicular tumors, is comparatively lower. This discrepancy may result from differences in the criteria that Chinese and foreign cytopathologists use to identify atypical cytomorphology. This explanation is supported by evidence from a multicenter study by Padmanabhan et al. [18], who found that AUS to have the highest interobserver variability among cytopathologists. Furthermore, due to the use of diverse reporting systems, the ratio of indeterminate classification will likewise vary [19–21]. Additionally, variations in race, geographic region, and iodine intake may also contribute to differences in the incidence of follicular tumors. The efficacy of FNA is well established; however, it is not without limitations. Relying solely on this technique for preoperative diagnosis may result in overdiagnosis and unnecessary treatment of some patients. Therefore, enhancing the sensitivity and specificity of preoperative diagnosis is the optimal approach to reducing unwarranted thyroidectomies.

Ultrasound is the preferred imaging method for thyroid nodules and plays a crucial role in diagnostic and treatment decision-making. Various versions of the Thyroid Imaging Reporting and Data System (TIRADS) have been developed with reference to the American College of Radiology’s Breast Imaging Reporting and Data System (BI-RADS) in certain countries [22–24]. The C-TIRADS, introduced in 2020 by the Chinese Society of Ultrasound in Medicine, is widely used in China [11]. At our center, we typically use C-TIRADS along with relevant clinical information and nodule size to decide whether ultrasound-guided FNA is necessary. Some studies suggest that repeating ultrasound risk stratification, whether alone or in combination with cytological score and/or other factors such as molecular detection and clinical information, can help predict malignancy in indeterminate FNA nodules [25–28]. However, our study’s statistical results showed that the ultrasound scores for the malignant or low-risk groups were slightly higher than those for the benign group; however, this difference was not statistically significant. Thus, in indeterminate thyroid nodules at high risk of malignancy, ultrasound C-TIRADS scoring may have limited utility.

The cyclin D1 oncogene encodes the regulatory subunit of a holoenzyme that phosphorylates and inactivates the Rb protein, facilitating the transition from the G1 phase to the S phase of the cell cycle [29]. Several studies in which one or more immunohistochemical antibodies were used to identify the optimal diagnostic markers for thyroid tumors have shown that cyclin D1 was one of the most reliable markers [30–33]. The histology study also suggests that cyclin D1 is a valuable biomarker for distinguishing between benign thyroid hyperplasia and thyroid tumors [34]. Our study found that the malignant group had a significantly higher prevalence of positive cyclin D1 specimens than the benign group. Moreover, the percent values of cyclin D1 expression in the malignant group were larger than those in the benign group. The ROC curve analysis revealed an AUC >0.9, indicating that cyclin D1 has a substantial diagnostic utility for malignant thyroid tumors. Ancillary cyclin D1 immunohistochemistry can serve as a valuable adjunct in the diagnosis and differentiation of thyroid FNAs with indeterminate cytology, as validated in our prior research [35]. However, since the NPV was only 50%, relying solely on this technique for cytological diagnosis could lead to many benign specimens being misidentified.

The integration of molecular detection with pathological diagnosis is becoming more widespread, thanks to advancements in molecular Detection technology and the ongoing discovery of specific molecular events related to various tumor types. The 2017 version of the TBSRTC reporting system included molecular detection in the clinical management of diagnostic categories III and IV [2]. Molecular testing has gradually emerged as a complementary tool for conventional cytology and is increasingly recommended for differentiating between malignant and benign lesions in nodules with indeterminate cytology [36].

The BRAF V600E is a common mutation in PTC. Its prevalence ranges from 80% to 90% in East Asian countries such as China and Korea, which is significantly higher than the 30%–50% prevalence observed in Western countries [17, 37]. The tumor cells with the BRAF V600E mutation often exhibit distinct morphological characteristics. These include an increase in nuclear perimeter and area, plump cells, large, polygonal tumor cells with moderate to abundant homogeneous acidophilic squamous metaplastic cytoplasm, and small off-center nuclei with a unique sickle shape [38, 39]. These identifiable features facilitate the cytological diagnosis of malignant tumors. However, the effectiveness of detecting the BRAF V600E mutation is limited in cases of uncertain cytology categories.

The RAS gene mutation, which includes three highly homologous isoforms (HRAS, KRAS, and NRAS), is the second most prevalent mutation found in thyroid malignant tumors. It can be observed in approximately 40%–50% of FTCs and 10%–20% of PTCs [6]. Additionally, HRAS and KRAS mutations can be detected in MTC [40]. However, RAS gene mutations are found in 20%–40% of benign follicular lesions, leading to a lower PPV compared with BRAF mutations (74%–87%) [41]. Additionally, Liu et al. [42] recently identified a molecular marker known as the telomerase reverse transcriptase promoter mutation, which is frequently observed in aggressive thyroid cancers. This mutation helps diagnose malignant nodules and also acts as a prognostic indicator for tumor progression.

High cost is a major drawback and limitation to the use of the available molecular diagnostic tools recommended in TBSRTC. To address this limitation, Amoy Diagnostics Co., Ltd., China, has developed a customized 5-gene rule-in panel to analyze the most frequent genetic alterations in thyroid cancer. BRAF mutation testing is performed on cytology specimens that are suspicious for PTC and AUS, and 5-gene molecular testing is performed on specimens classified as TBSRTC IV by FNA. Although the BRAFV600E and 5-gene tests have low accuracy and sensitivity for diagnosing thyroid malignant tumors (67.2% and 63.9%, respectively), their specificity (90.9%) exceeds that of cyclin D1 (86.4%) and C-TIRAD (36.4%). This discrepancy may arise from the low mutation rate of genes other than BRAF in thyroid tumors. Most specimens collected were PTC, with only a small proportion being other thyroid tumors. Additionally, this disparity could be influenced by the limited number of samples tested for the 5-gene panel in this particular dataset.

The performance of both the C-TIRADS score and cyclin D1 ICC is straightforward, with the former already incorporated into standard diagnostic tests, thereby avoiding any additional medical expenses. The C-TIRADS score and cyclin D1 immunocytochemistry demonstrate higher sensitivity (83.9%, 87.7%) in the diagnosis of malignant or low-risk thyroid tumors compared with genetic testing (63.9%); however, genetic testing exhibits significantly higher specificity (90.9%) than the other two methods (36.4%, 86.4%) and also provides prognostic value. Our study data indicate that repeating ultrasound stratification alone does not effectively predict malignancy in indeterminate FNA nodules, Adding an ultrasound score does not improve the diagnostic accuracy, sensitivity, specificity, PPV, or NPV of the genetic testing, which contradicts previous published findings [28, 43]. Genetic testing has high specificity (90.9%) for diagnosing malignant or low-risk tumors but low sensitivity (63.9%). However, when cyclin D1 immunostaining is added, sensitivity can significantly increase to 98.6%.

In terms of AUC, the diagnostic accuracy of cyclin D1 alone (0.921, 95% CI = 0.857–0.985, p = 0.000) was significantly higher than that of the other two methods (0.775 [95% CI = 0.703–0.846, p = 0.000]; 0.587 [95% CI = 0.443–0.731, p = 0.235]). Cyclin D1 immunostaining not only increased the suspicion of malignancy in thyroid nodules with indeterminate cytology but also showed improved diagnostic performance when the cut-off point was optimized and combined with gene testing and/or the C-TIRADS score. The combination of two or three methods yields higher accuracy (0.929 [95% CI = 0.873–0.985]; 0.925 [95% CI = 0.860–0.989]; 0.937 [95% CI = 0.889–0.985]) compared with the use of cyclin D1 immunostaining alone (0.921 [95% CI = 0.857–0.985, p = 0.000]).

Limitation

This study was a 5-year retrospective review conducted at a single institution, and it has notable limitations. The results were based on diagnoses from two experienced cytologists at the same laboratory. Other laboratories may have different grading judgments for TBSRTC, highlighting the need for verification through multicenter large-scale studies. A significant limitation of the 5-gene panel is that it was only used in 21 out of 177 specimens. This limited application affects its ability to improve the accuracy of cytology results. The true value can only be derived from cyclin D1 and BRAFV600E molecular testing. This limitation could potentially explain why the NPV of combined gene detection contradicts the findings of Decaussin-Petrucci et al. [44]. It is essential to increase the sample size of the 5-gene test to confirm its diagnostic value in the FNA cases with indeterminate cytology. Moreover, not all specimens with indeterminate cytology receive histopathological confirmation. As a result, the actual incidence of malignancy in thyroid nodules diagnosed as TBSRTC III–IV remains unknown. Most patients opted for a strategy involving periodic ultrasound follow-up, with some patients choosing to change hospitals, resulting in only 19.5% (177/919) of specimens undergoing histopathological confirmation in our study, among which only 12.4% (22/177) had benign nodules. The uneven distribution of data in both groups will ultimately impact the statistical outcomes. Additionally, most malignant thyroid tumors in our study were follicular cell tumors, particularly PTC. Further validation with larger patient cohorts and multicenter data is necessary to determine the applicability of these detection methods to other rare types of thyroid tumors.

Conclusion

Our data indicate that combining the C-TIRADS score, cyclin D1 immunostaining, and a simple gene test (BRAF V600E or 5-gene panel) can help differentiate between malignant and benign thyroid nodules with indeterminate cytology. Laboratories may integrate these various tests to guide the clinical management of uncertain thyroid nodules, based on their resources. This integration enhances the accuracy of preoperative diagnosis, facilitates optimal treatment decisions, and minimizes patient harm as well as unnecessary waste of medical resources.

Acknowledgment

The authors thank AiMi Academic Services (www.aimieditor.com) for English language editing and review services.

Statement of Ethics

This retrospective study was approved and granted an exemption from requiring written informed consent by the Ethics Committee of the Beijing Hospital, Approval No. BJ-2019-155.

Conflict of Interest Statement

The authors do not have any conflict of interest to declare.

Funding Sources

No funds, grants, or other support was received.

Author Contributions

S.H.: conceptualization, methodology, data curation, investigation, cytopathological evaluation, and original draft writing. J.Z. and M.W.: data curation and investigation. D.L. and M.J.: conceptualization, methodology, project administration, supervision, and editing. F.W. and K.G.: ultrasound evaluation and data curation. R.C.: taking photographs. S.H.: immunostaining. D.J.: molecular detection.

Funding Statement

No funds, grants, or other support was received.

Data Availability Statement

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.