Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2024 Sep 16;102(1):79–90. doi: 10.1111/cen.15130

Diagnostic Nomogram Model for ACR TI‐RADS 4 Nodules Based on Clinical, Biochemical Data and Sonographic Patterns

Yongheng Wang 1,2, Yao Tang 3, Ziyu Luo 1,2, Jianhui Li 1,2, Wenhan Li 1,2,✉
PMCID: PMC11612534  PMID: 39279486

ABSTRACT

Objectives

The objective of this study was to develop and validate a nomogram model integrating clinical, biochemical and ultrasound features to predict the malignancy rates of Thyroid Imaging Reporting and Data System 4 (TR4) thyroid nodules.

Methods

A total of 1557 cases with confirmed pathological diagnoses via fine‐needle aspiration (FNA) were retrospectively included. Univariate and multivariate logistic regression analyses were conducted to identify independent predictors of malignancy. These predictors were incorporated into the nomogram model, and its predictive performance was evaluated using receiver‐operating characteristic curve (AUC), calibration plots, net reclassification improvement (NRI), integrated discrimination improvement (IDI) and decision curve analysis (DCA).

Results

Eight out of 22 variables—age, margin, extrathyroidal extension, halo, calcification, suspicious lymph node metastasis, aspect ratio and thyroid peroxidase antibody—were identified as independent predictors of malignancy. The calibration curve demonstrated excellent performance, and DCA indicated favourable clinical utility. Additionally, our nomogram exhibited superior predictive ability compared to the current American College of Radiology (ACR) score model, as indicated by higher AUC, NRI, IDI, negative likelihood ratio (NLR) and positive likelihood ratio (PLR) values.

Conclusions

The developed nomogram model effectively predicts the malignancy rate of TR4 thyroid nodules, demonstrating promising clinical applicability.

Keywords: ACR TI‐RADS 4, fine‐needle aspiration, nomogram, thyroid nodules

1. Introduction

The incidence of thyroid nodules has significantly increased in recent decades due to the widespread use of high‐resolution ultrasound and chest computed tomography (CT) scans. Studies indicate that 50%–65% of healthy individuals may have thyroid nodules, with more than 95% being asymptomatic and not requiring treatment [1, 2]. However, thyroid cancer has emerged as the most prevalent endocrine malignancy, with annual increases in morbidity rates ranging from 7% to 15% [1, 3]. Ultrasound, the most important diagnostic tool for thyroid nodules, not only presents nodule features but also guides fine‐needle aspiration (FNA) procedures [4]. Various malignancy risk stratification systems based on ultrasound characteristics have been implemented in clinical practice. Tessler et al. introduced the Thyroid Imaging Reporting and Data System (TI‐RADS), modelled after the American College of Radiology (ACR) Breast Imaging Reporting and Data System (BI‐RADS) [5]. TI‐RADS categorizes malignancy risk from TR1 (benign) to TR5 (high suspicion of malignancy) based on ultrasound characteristics such as composition, echogenicity, shape, margin and echogenic foci. TI‐RADS 4 (TR4), which indicates a risk of thyroid carcinoma ranging from 2% to 90%, is more common than TR5 in clinical practice. Despite several studies proposing new risk stratification models for thyroid nodules [6, 7, 8, 9, 10, 11], few large‐scale, high‐quality reports specifically focus on TR4 nodules.

FNA remains the most reliable diagnostic method for suspicious nodules. However, this invasive procedure may lead to complications such as pain and bleeding and can cause emotional distress, including depression and anxiety among patients [12]. Therefore, FNA is typically recommended only for selected patients. Current nodule risk stratification systems predominantly rely on radiologic features alone. However, other factors, such as age, gender, nodule size and thyroid function, have been identified as independent predictors of malignancy [6, 7, 13]. In addition, some benign thyroid conditions, such as asymptomatic subacute thyroiditis, may mimic the ultrasound manifestations of thyroid malignancies, complicating decisions based solely on the TI‐RADS criteria. Currently, artificial intelligence (AI) methods (S‐Detect 2, AmCAD‐UT, Koios DS, Medo Thyroid) based on ultrasound radiomic features have been approved by the Food and Drug Administration (FDA) and demonstrate diagnostic performance comparable to or exceeding that of radiologists [14, 15, 16]. However, the widespread adoption of AI tools is limited by the costs associated with initial investments and potential subscription fees, creating a gap between innovation and practical application.

Therefore, this study aimed to develop and validate a user‐friendly nomogram model that integrated clinical, biochemical and ultrasound features using data from 1557 TR4 cases in our medical centre, thereby further improving the prediction accuracy for malignancy in TR4 thyroid nodules.

2. Methods

2.1. Patient Selection

The study was approved by the Ethical Committee of Shaanxi Provincial People's Hospital. Medical data, including comprehensive demographics, radiological evaluations, thyroid function tests and pathology results, were extracted from patients diagnosed with thyroid nodules at our medical centre between January 2018 and October 2022. Inclusion criteria were as follows: (1) Patients diagnosed with American College of Radiology Thyroid Imaging Reporting and Data System 4 (ACR‐TIRADS 4) thyroid nodules via ultrasound at our institution; (2) all suspicious nodules underwent confirmation of benign or malignant status through FNA. Exclusion criteria were as follows: (1) Patients with insufficient medical records available for analysis; (2) patients with a history of thyroid surgery; (3) patients with inadequate ultrasound images for re‐evaluating the features of thyroid nodules; and (4) patients with undetermined pathology results for thyroid neoplasms.

2.2. Acquisition of Clinical Data

Clinical data, including age, gender and body mass index (BMI), were retrieved from patients' medical records. Thyroid function tests were performed at the clinical laboratory of Shaanxi Provincial People's Hospital using commercially available chemiluminescence immunoassays. These tests were conducted within 1 week before FNA, with cutoff values set according to reagent protocols: thyroid‐stimulating hormone (TSH): 0.35–5.5 uIU/mL; thyroid peroxidase antibodies (TPO Abs): < 30 IU/mL; thyroglobulin antibodies (TG Abs): < 75 IU/mL; free thyroxine (fT4): 10–22 pmol/L; free triiodothyronine (fT3): 3.5–7 pmol/L; thyroglobulin (TG): < 50 ng/mL.

2.3. Ultrasound Examination and Image Analysis

All ultrasound examinations were carried out by a sonographer with 9 years of specialized experience in thyroid ultrasound. Images were archived in the image database of our medical centre. Subsequently, two independent radiologists, blinded to pathology results, classified nodule image features, including nodule size, position, margin, extrathyroidal extension, halo, composition, echogenicity, calcification, suspicious lymph node metastasis (LNM) and aspect ratio. Any discrepancies were resolved through collaborative discussion.

2.4. Ultrasound‐Guide FNA

FNA was performed on nodules exhibiting suspicious ultrasound features. A radiologist used a disposable needle (21 G) and a disposable syringe (10 mL) to puncture the target nodule under ultrasound guidance. Rapid and multidirectional movements with negative pressure were applied to obtain the specimen. The collected material was then fixed in 95% ethanol and transferred to the Department of Pathology at our medical centre for cytological analysis. Cytopathological results were categorized into six groups according to the Bethesda System of reporting [17]. For this study, only nodules with definitive pathological diagnoses—specifically Bethesda II for benign cases and Bethesda VI for malignant cases—were included in the analysis.

2.5. Statistical Analysis

Statistical analysis was performed using SPSS (Version 26.0) and R software (Version 4.3.2). Univariate and multivariate binary logistic regression models were used to identify independent predictors for the diagnosis of thyroid nodule pathology. Following predictor selection, both categorical and continuous variables were integrated into a novel nomogram model. The performance of our nomogram was compared with the traditional ACR score model using the area under the time‐dependent receiver‐operating characteristic curve (AUC) and calibration plots in separate training and validation cohorts. Additionally, we assessed predictive accuracy, specificity, sensitivity, positive likelihood ratio (PLR) and negative likelihood ratio (NLR) for both models. To further elucidate the predictive superiority of our model, we conducted net reclassification improvement (NRI) and integrated discrimination improvement (IDI) analyses. NRI evaluated the model's ability to correctly reclassify individuals compared to the traditional model, while IDI measured the overall improvement in discriminatory accuracy [18, 19]. Finally, the clinical utility of our nomogram was evaluated through decision curve analysis (DCA). Two‐sided p < 0.05 were considered statistically significant.

3. Results

3.1. Demographics and Thyroid Nodule Characteristics

A total of 1557 TR4 nodules were included in this study, consisting of 431 males and 1126 females, with a mean age of 47.02 ± 12.76 years. Of these nodules, 1086 were malignant (Figures S1, S2 and 3A,C) and 471 were benign (Figures S1, S2 and 3B,D). The nodules were categorized as TR4a (677 cases, 43.5%), TR4b (308 cases, 19.8%) and TR4c (572 cases, 36.7%). The average nodule diameter was 1.34 ± 1.22 cm. The nodules were randomly divided into a training group (1092 cases) and a validation group (465 cases) in a 7:3 ratio. There were no significant differences between these two groups in terms of baseline clinical data, tumour ultrasound features and blood biochemical indexes (Table 1).

Table 1.

Characteristics of patients with thyroid nodules in the training and validation sets.

Variable Total (n = 1557) Training set (n = 1092) Validation set (n = 465) p value
Age 47.02 ± 12.76 46.96 ± 12.72 47.17 ± 12.86 0.76b
Gender
Male 431 301 130
Female 1126 791 335 0.874a
Nodule size 1.34 ± 1.22 1.34 ± 1.17 1.36 ± 1.34 0.759b
BMI 24.09 ± 3.82 24.16 ± 3.72 23.94 ± 4.03 0.297b
Nodule Position 1
Left lobe 701 485 216
Right lobe 604 570 234
Isthmus 52 37 15 0.76a
Nodule Position 2
Upper region 317 223 94
Middle region 711 501 210
Lower region 460 320 140
Isthmus 69 48 21 0.989a
Histology
Benign 471 329 142
Malignant 1086 763 323 0.872a
Margin
Regular 244 176 68
Irregular 1313 916 397 0.458a
Extrathyroidal extension
Yes 344 258 86
No 1213 834 379 0.25a
Halo
Absent 1359 948 411
Complete 82 56 26
Incomplete 116 88 28 0.36a
Composition
Cystic or spongiform 9 7 2
Cystic and solid (cystic ≥ 50%) 38 25 13
Cystic and solid (solid ≥ 50%) 95 70 25
Solid 1415 990 425 0.757a
Calcification
No calcification 489 343 146
Macrocalcification 168 116 52
Macro‐ and microcalcification 105 82 23
Microcalcification 795 551 244 0.317a
Vascular distribution pattern
Avascularity 627 425 202
Peripheral vascularity 312 215 97
Mainly central vascularity 429 310 119
Mixed vascularity 189 142 47 0.168a
Suspicious LNM
Yes 288 208 80
No 1269 884 385 0.391a
Aspect ratio > 1
Yes 948 660 288
No 609 432 177 0.58a
TSH
< 0.35 33 20 13
0.35–5.5 1375 971 404
≥ 5.5 149 101 48 0.372a
TPO Ab
< 30 1254 871 383
≥ 30 303 221 82 0.235a
TG Ab
< 75 1277 906 371
≥ 75 280 186 94 0.135a
fT4
< 10 39 26 13
10–22 1503 1056 447
≥ 22 15 10 5 0.852a
fT3
< 3.5 40 25 15
3.5–7 1502 1055 447
≥ 7 15 12 3 0.403a
Tg
< 50 1248 884 364
≥ 50 349 248 101 0.934a
NLR 3.46 ± 3.60 3.45 ± 3.64 3.48 ± 3.51 0.902b
Open in a new tab

Abbreviations: BMI, body mass index; LNM, lymph node metastasis; NLR, neutrophil‐to‐lymphocyte ratio; TG Ab, thyroglobulin antibodies; TSH, thyroid‐stimulating hormone; TPO Ab, thyroid peroxidase antibody.

a

Using the χ 2 test for this statistic.

b

Using two‐sample t‐test for this statistic.

3.2. Univariate and Multivariate Logistic Regression Analyses

In the training set, univariate logistic regression analysis was conducted to identify factors that differentiate benign from malignant nodules. As shown in Table 2, 14 variables (age, gender, nodule size, Nodule Position 2, margin, extrathyroidal extension, halo, composition, calcification, suspicious LNM, aspect ratio, TPO Ab, TG and NLR showed significant differences between benign and malignant nodules (all p < 0.05). In contrast, BMI, marital status, Nodule Position 1, vascular distribution pattern, TSH, TG Ab, fT34 and fT33 did not show significant differences between the two groups (all p > 0.05). Multivariate logistic regression analysis further investigated the risk factors associated with malignancy. Eight out of the 14 variables (age, margin, extrathyroidal extension, halo, calcification, suspicious LNM, aspect ratio and TPO Ab) were identified as independent predictors of malignancy in TR4 nodules (all p < 0.05).

Table 2.

Univariate and multivariate analysis of risk factors for thyroid cancer in the training group.

Variables Histology Univariate analysis Multivariate analysis
Benign Malignant OR (95%CI) p value OR (95%CI) p value
Age (mean ± SD) 52.31 ± 12.58 44.65 ± 12.07 0.95 (0.94–0.96) < 0.01* 0.95 (0.94–0.97) < 0.01*
Gender
Male 70 231 Reference Reference
Female 259 532 0.62 (0.46–0.84) < 0.01* 0.80 (0.52–1.23) 0.32
BMI 24.01 ± 3.89 24.22 ± 3.64 1.02 (0.98–1.05) 0.4
Marital status
Married 315 727 Reference
Other 14 36 1.11 (0.61–2.16) 0.74
Nodule size (mean ± SD) 1.64 ± 1.39 1.21 ± 1.04 0.74 (0.66–0.83) < 0.01* 0.95 (0.80–1.13) 0.56
Nodule Position 1
Left lobe 136 349 Reference
Right lobe 184 386 0.82 (0.63–1.06) 0.14
Isthmus 9 28 1.21 (0.58–2.78) 0.63
Nodule Position 2
Upper region 46 177 Reference Reference
Middle region 175 326 0.48 (0.33–0.7) < 0.01* 0.63 (0.38–1.04) 0.08
Lower region 97 223 0.6 (0.4–0.89) 0.01* 0.59 (0.34–1.01) 0.06
Isthmus 11 37 0.87 (0.42–1.92) 0.72 0.73 (0.27–2.10) 0.54
Margin
Regular 140 36 Reference Reference
Irregular 189 727 14.96 (10.14–22.59) < 0.01* 5.10 (3.06–8.69) < 0.01*
Extrathyroidal extension
Yes 8 250 Reference Reference
No 321 513 0.05 (0.02–0.1) < 0.01* 0.08 (0.03–0.17) < 0.01*
Halo
Absent 268 680 Reference Reference
Complete 52 4 0.03 (0.01–0.07) < 0.01* 0.06 (0.02–0.19) < 0.01*
Incomplete 9 79 3.46 (1.8–7.5) < 0.01* 4.83 (2.08–12.62) < 0.01*
Composition
Cystic or spongiform 5 2 Reference Reference
Cystic and solid (cystic ≥ 50%) 20 5 0.63 (0.1–5.25) 0.63 0.85 (0.02–38.88) 0.93
Cystic and solid (solid ≥ 50%) 42 28 1.67 (0.33–12.21) 0.56 0.69 (0.02–28.41) 0.85
Solid 262 728 6.95 (1.49–48.71) 0.02* 0.55 (0.07‐6.34) 0.59
Calcification
No calcification 137 206 Reference Reference
Macrocalcification 63 53 0.56 (0.37–0.85) < 0.01* 0.69 (0.37–1.28) 0.24
Macro‐ and microcalcification 21 61 1.93 (1.14–3.38) 0.02* 1.12 (0.53–2.42) 0.77
Microcalcification 108 443 2.73 (2.02–3.69) < 0.01* 1.53 (1.02–2.29) 0.04*
Vascular distribution pattern
Avascularity 130 295 Reference
Peripheral vascularity 66 149 0.99 (0.7–1.42) 0.98
Mainly central vascularity 87 223 1.13 (0.82–1.56) 0.46
Mixed vascularity 46 96 0.92 (0.61–1.39) 0.69
Suspicious LNM
Yes 14 194 Reference Reference
No 315 569 0.13 (0.07–0.22) < 0.01* 0.24 (0.11–0.48) < 0.01*
Aspect ratio > 1
Yes 124 536 Reference Reference
No 205 227 0.26 (0.19–0.34) < 0.01* 0.41 (0.27–0.61) < 0.01*
TSH
< 0.35 7 13 Reference
0.35–5.5 290 681 1.26 (0.47–3.12) 0.62
≥ 5.5 32 69 1.16 (0.4–3.12) 0.77
TPO Ab
< 30 233 638 Reference Reference
≥ 30 96 125 0.48 (0.35–0.65) < 0.01* 0.42 (0.27–0.65) < 0.01*
TG Ab
< 75 276 630 Reference
≥ 75 53 133 1.1 (0.78–1.57) 0.6
fT4
< 10 8 18 Reference
10–22 320 736 1.02 (0.42–2.3) 0.96
≥ 22 1 9 4 (0.59–80.17) 0.22
fT3
< 3.5 8 17 Reference
3.5‐7 317 738 1.1 (0.44–2.49) 0.83
≥ 7 4 8 0.94 (0.22–4.38) 0.94
Tg
< 50 235 609 Reference Reference
≥ 50 94 154 0.63 (0.47–0.85) < 0.01* 0.75 (0.49–1.16) 0.20
NLR (mean ± SD) 2.73 ± 2.45 3.77 ± 4.01 1.12 (1.06–1.18) < 0.01* 1.05 (0.98–1.12) 0.17
Open in a new tab

Abbreviations: BMI, body mass index; CI, confidence interval; LNM, lymph node metastasis; NLR, neutrophil‐to‐lymphocyte ratio; SD, standard deviation; TG Ab, thyroglobulin antibodies; TPO Ab, thyroid peroxidase antibody; TSH, thyroid‐stimulating hormone.

*

p value indicates a significant difference.

3.3. Development and Validation of a Nomogram for Predicting Thyroid Cancer

Based on the results of the univariate and multivariate analyses, we developed a nomogram to quantitatively assess the risk of malignancy in thyroid nodules. As shown in Figure 1, clinicians can calculate the malignancy probability of thyroid nodules by summing the scores assigned to selected variables. Notably, halo status contributed most significantly to the model, followed by age and extrathyroidal extension. We compared the predictive performance of our nomogram with that of the traditional ACR score model. The AUC values for the nomogram were 0.884 and 0.882 in the training and validation cohorts, respectively, whereas the ACR score model achieved AUC values of 0.790 and 0.831, respectively (Figure 2). Additionally, the predictive accuracy, specificity, sensitivity, PLR and NLR for both models are detailed in Table 3. NRI and IDI analyses demonstrated the superior performance of our nomogram compared to the ACR score model (IDI: 0.1437, 95% CI: 0.1213–0.1660, p < 0.001; categorical NRI: 0.1162, 95% CI: 0.0613–0.1711, p < 0.001; continuous NRI: 0.861, 95% CI: 0.7469–0.9751, p < 0.001). These findings underscored the superior predictive capability of our nomogram in comparison with the ACR score model. For calibration assessment, calibration plots in both training and validation groups (Figure 3A,B) showed strong agreement between predicted probabilities and actual outcomes, with deviations consistently within a 10% margin of error. Finally, DCA demonstrated the satisfactory clinical utility of our model (Figure 3C).

Figure 1.

Figure 1

Open in a new tab

Nomogram for predicting the risk of malignancy for thyroid nodules in TR4. All the points assigned on the top point scale for each factor are summed together to generate a total point score. The total point score is projected on the bottom scales to determine the overall survival rate in an individual.

Figure 2.

Figure 2

Open in a new tab

The receiver‐operating characteristics (ROC) curve and area under the ROC curve (AUC) of the nomogram in the training (A) and validation (B) cohorts. The ROC curve and AUC of the ACR score model in the training (C) and validation (D) cohorts.

Table 3.

Diagnostic performances of the nomogram and ACR score model.

Performance parameter Accuracy Specificity Sensitivity PLR NLR
Nomogram training cohort 0.81 0.757 0.934 3.428 0.22
Nomogram validation cohort 0.815 0.761 0.839 3.504 0.212
ACR model training cohort 0.791 0.55 0.895 1.99 0.191
ACR model validation cohort 0.809 0.57 0.913 2.216 0.152
Open in a new tab

Abbreviations: NLR, negative likelihood ratio; PLR, positive likelihood ratio.

Figure 3.

Figure 3

Open in a new tab

Nomogram validation. Calibration curve showing nomogram‐predicted malignant nodules probabilities compared with the actual malignant nodules in the training (A) and validation (B) cohort. Decision curve analyses (DCAs) of the nomogram model for predicting malignant nodules in TR4 (C).

4. Discussion

The prevalence of thyroid nodules has increased with the widespread use of imaging technologies [1, 5, 20]. Currently, FNA remains the most accurate diagnostic method for determining the nature of thyroid nodules. However, this invasive procedure should be used cautiously due to potential complications. Although several ultrasound‐based risk stratification systems have been proposed to reduce unnecessary FNA procedures, improvements are still needed. Malignant thyroid nodules typically present with distinct ultrasound features compared to benign nodules; however, benign conditions like subacute thyroiditis can still present diagnostic challenges. Existing systems, like the ACR TI‐RADS, offer broad estimates of malignancy risk (e.g., TR4 indicates a risk range from 2% to 90%). However, studies indicate that TR4 nodules have variable benign probabilities, ranging from 29% to 70% [6, 9, 10, 21] due to differing inclusion criteria and sample sizes. For example, malignancy rates are usually higher in studies confirmed by surgical procedures compared to those based on FNA. Additionally, high‐level medical centres often report a higher proportion of malignant TR4 nodules than primary medical centres. The exclusion of low‐risk thyroid cancers in some studies due to incomplete pathological results further complicates risk assessment. In our study, 30% (471/1557) of cases were benign, highlighting diagnostic uncertainty that can cause patient anxiety despite recommendations for active surveillance. Moreover, ultrasonography is a relatively subjective examination heavily dependent on the radiologists' experience. It can introduce variability, potentially leading to misdiagnosis, especially in less‐experienced or primary care settings where interpretation relies heavily on imaging reports alone. Besides sonographic characteristics, clinicopathologic features and thyroid function also contribute to the primary diagnosis [6, 7, 13]. Therefore, we developed a diagnostic nomogram integrating clinical and ultrasonographic patterns specific to TR4 nodules. Our model demonstrated superior discrimination and calibration capability compared to the ACR TI‐RADS system in both the training and validation cohorts (AUC: 0.884 vs. 0.790; IDI: 0.1437; categorical NRI: 0.1162; continuous NRI: 0.861).

In our analysis, a total of 21 variables were considered, including baseline information (age, gender, BMI, nodule size and nodule location), sonographic features (margin, extrathyroidal extension, halo, composition, calcification, vascular distribution pattern, suspicious LNM and aspect ratio > 1), as well as blood biochemical parameters (TSH, TPO Ab, TG Ab, fT4, fT3, TG and NLR) from a cohort of 1557 individuals. Of these, eight variables (age, margin, extrathyroidal extension, halo, calcification, suspicious LNM, aspect ratio and TPO Ab) were selected for inclusion in our nomogram model. Our findings were consistent with those of Yang, Gao, and Yang [6], who developed a similar nomogram model for diagnosing C‐TR4 nodules. However, our study benefited from a significantly larger sample size, three times that of Yang, Gao, and Yang, and included a broader range of factors potentially influencing diagnosis. Consequently, our study may provide more statistically robust findings and serve as both a complement and refinement of their research.

In this study, we focused on parameters readily accessible in clinical practice. Gene biomarkers, such as BRAF V600E [6], may be involved in future practice when widely applied. Our nomogram underscored the continued significance of nodule sonographic features in diagnosis. Although individual ultrasound features alone may not have a high predictive value for malignancy, their combined patterns—such as spiculated margins, microcalcifications, taller‐than‐wide shape and marked hypoechogenicity—remained integral to various ultrasound‐based risk stratification systems [22, 23, 24, 25]. Specifically, our predictive model included margin characteristics, extrathyroidal extension, halo sign presence, calcification pattern and aspect ratio. Hypoechogenicity has been identified as highly specific for malignancy with the specificity ranging between 92% and 94% [26]. However, this discipline may be less discriminative for TR4 nodules, given that the majority (1510/1557) exhibited hypoechoic features.

A thyroid nodule may present a pseudocapsule, often seen as a halo or hypoechoic rim, composed of fibrous connective tissue, compressed thyroid tissue and chronic inflammatory elements [27]. In our study, we observed that 82% (386 out of 471) of benign nodules did not have a halo, consistent with findings by Trimboli et al. [28]. Conversely, halos were generally absent in the majority of thyroid carcinomas (973 out of 1086), and when present, they were often incomplete (104 out of 113). Thus, in line with previous research [4], our study underscores the significance of halo integrity in distinguishing between benign and malignant tumours.

The borders of a tumour, often characterized by unclear or uneven edges, present challenges in delineating the boundary between the nodule and surrounding normal tissue. Speculated or microlobulated margins are strong indicators of malignancy, with a specificity rate of 92% and a positive predictive value of 81%. However, some studies define ill‐defined nodule margins as those with more than 50% unclear demarcation from the surrounding tissue [26]. We recommend refining the definition of nodule irregularity to mitigate biases across studies. Additionally, nodules may extend beyond the thyroid capsule, invading adjacent soft tissue and/or vascular structures—a condition known as extrathyroidal extension. Therefore, irregular margins and extrathyroidal extension were independent predictors of malignancy in our study.

Calcification is a significant sonographic feature in diagnosing thyroid nodules. Psammoma bodies, characterized by round, laminar and crystalline deposits ranging from 10 to 100 μm, appear as microcalcifications on ultrasound. These bodies are strongly associated with papillary thyroid carcinoma, with a specificity of 86%–95% and a positive predictive value of 42%–94% [29, 30, 31, 32]. In contrast, dystrophic calcifications are larger and irregularly shaped, resulting from tissue necrosis, and have limited diagnostic value [33, 34]. Additionally, a taller‐than‐wide shape (aspect ratio > 1) is another crucial indicator of malignancy, with a specificity of 89% and a positive predictive value of 86% [26]. This shape difference arises from distinct growth patterns: benign nodules typically grow parallel to the tissue plane, while malignant nodules tend to expand centrifugally through normal tissue [4, 35]. Our study confirmed the diagnostic significance of both calcification patterns and aspect ratio in differentiating thyroid nodule pathology.

The identification of highly suspicious neck lymph nodes during ultrasound is crucial for diagnosing thyroid cancer. Normally, a lymph node appears hypoechoic on ultrasound, with an oval shape and a central hyperechoic streak corresponding to the hilum. Abnormal lymph nodes, however, may exhibit characteristics such as cystic or solid composition, isoechoic or hyperechoic appearance, round or irregular shape and the absence of a visible hilum [36]. It has been reported that the pooled sensitivity, specificity and AUC for ultrasound in diagnosing central LNM are 33%, 93% and 0.69, respectively. For lateral LNM, these values are 70%, 84% and 0.88, respectively [37]. In our study, we included lymph node status as an independent risk factor in our nomogram for predicting the risk of thyroid cancer.

In addition to ultrasound patterns, our nomogram model integrates age and TPO Ab levels as crucial parameters. Currently, their diagnostic roles in thyroid nodules have remained controversial. Consistent with previous research [10, 11], our study revealed that younger age was associated with a higher likelihood of malignancy in thyroid nodules. We also observed an inverse relationship between TPO Ab levels and the prevalence of thyroid cancer in both univariate and multivariate logistic regression analyses. Paparodis et al. reported that elevated TPO Ab levels might offer protection against thyroid cancer in patients with Hashimoto's thyroiditis in a large multicenter cohort of 8461 participants [38]. However, other studies have presented conflicting or even contradictory results [39, 40, 41, 42, 43]. These inconsistencies may arise from differences in study populations, sample sizes and selection biases. Further research is needed to clarify the true relationship between age, TPO Ab levels and thyroid cancer risk, given the complexities and potential confounding factors involved in thyroid disease and its diagnostic markers.

Although our model has shown promising diagnostic capability for TR4 nodules, several limitations should be acknowledged in clinical practice. First, the majority of our study subjects were recruited from the northwestern region of China, highlighting the need for multicenter studies to assess the generalizability of our findings across different populations. Second, there may be inherent selection bias, as nodules without a pathological diagnosis were excluded from the study. Finally, certain cancer‐related features, such as BRAF mutation status and ultrasonic elastography, were not included due to insufficient data.

5. Conclusion

We developed a diagnostic nomogram incorporating eight independent factors to assist in evaluating the malignancy risk of TR4 nodules. This model shows potential for reducing unnecessary FNA procedures and provides new insights into managing TR4 nodules.

Author Contributions

W.L. and Y.W. conceived this study and wrote the manuscript. Y.W. and Y.T. conducted data curation and data analysis. Z.L. and J.L. assisted in revising the manuscript. All authors read the manuscript and approved it for publication.

Ethical Statements

This retrospective study was approved by the Institutional Review Board of the Shaanxi Provincial People's Hospital.

Consent

Informed consent for this study was waived.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

FIGURE 1 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4a thyroid nodules.

CEN-102-79-s002.tif (1.9MB, tif)

FIGURE 2 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4b thyroid nodules.

CEN-102-79-s003.tif (1.6MB, tif)

FIGURE 3 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4c thyroid nodules.

CEN-102-79-s001.tif (1.7MB, tif)

Acknowledgements

The authors have nothing to report. This work was supported by grants from the Shaanxi Provincial People's Hospital (2021JY‐45) and the key R & D Programme of Shaanxi Provinces (2022SF‐291 and 2023‐YBSF‐621).

Yongheng Wang and Yao Tang authors contributed equally to this work.

Development and validation of a novel diagnostic nomogram model for predicting malignancy rates of ACR TI‐RADS 4 thyroid nodules based on clinical, biochemical and sonographic characteristics.

Data Availability Statement

The data that support the findings of this study are openly available in the Science Data Bank at https://www.scidb.cn/en/s/QRFrQz.

References

  • 1. Haugen B. R., Alexander E. K., Bible K. C., et al., “2015 American Thyroid Association Management Guidelines for Adult Patients With Thyroid Nodules and Differentiated Thyroid Cancer: The American Thyroid Association Guidelines Task Force on Thyroid Nodules and Differentiated Thyroid Cancer,” Thyroid 26, no. 1 (January 2016): 1–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Gharib H., Papini E., Garber J. R., et al., “American Association of Clinical Endocrinologists, American College of Endocrinology, and Associazione Medici Endocrinologi Medical Guidelines for Clinical Practice for the Diagnosis and Management of Thyroid Nodules—2016 Update,” Endocrine Practice: Official Journal of the American College of Endocrinology and the American Association of Clinical Endocrinologists 22, no. 5 (May 2016): 622–639. [DOI] [PubMed] [Google Scholar]
  • 3. Chen W., Zheng R., Baade P. D., et al., “Cancer Statistics in China, 2015,” CA: A Cancer Journal for Clinicians 66, no. 2 (March/April 2016): 115–132. [DOI] [PubMed] [Google Scholar]
  • 4. Rago T., and Vitti P., “Risk Stratification of Thyroid Nodules: From Ultrasound Features to Tirads,” Cancers 14, no. 3 (January 2022): 717. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Tessler F. N., Middleton W. D., Grant E. G., et al., “ACR Thyroid Imaging, Reporting and Data System (TI‐RADS): White Paper of the ACR TI‐RADS Committee,” Journal of the American College of Radiology 14, no. 5 (May 2017): 587–595. [DOI] [PubMed] [Google Scholar]
  • 6. Yang Z., Gao X., and Yang L., “Predictors and a Prediction Model for Positive Fine Needle Aspiration Biopsy in C‐TIRADS 4 Thyroid Nodules,” Frontiers in Endocrinology 14 (2023): 1154984. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. Hu T., Li Z., Peng C., et al., “Nomogram to Differentiate Benign and Malignant Thyroid Nodules in the American College of Radiology Thyroid Imaging Reporting and Data System Level 5,” Clinical Endocrinology 98, no. 2 (February 2023): 249–258. [DOI] [PubMed] [Google Scholar]
  • 8. Strieder D. L., Cristo A. P., Zanella A. B., et al., “Using an Ultrasonography Risk Stratification System to Enhance the Thyroid Fine Needle Aspiration Performance,” European Journal of Radiology 150 (May 2022): 110244. [DOI] [PubMed] [Google Scholar]
  • 9. Tao Y., Yu Y., Wu T., et al., “Deep Learning for the Diagnosis of Suspicious Thyroid Nodules Based on Multimodal Ultrasound Images,” Frontiers in Oncology 12 (2022): 1012724. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Chen L., Zhang J., Meng L., Lai Y., and Huang W., “A New Ultrasound Nomogram for Differentiating Benign and Malignant Thyroid Nodules,” Clinical Endocrinology 90, no. 2 (February 2019): 351–359. [DOI] [PubMed] [Google Scholar]
  • 11. Xu S., Ni X., Zhou W., Zhan W., and Zhang H., “Development and Validation of a Novel Diagnostic Tool for Predicting the Malignancy Probability of Thyroid Nodules: a Retrospective Study Based on Clinical, B‐Mode, Color Doppler and Elastographic Ultrasonographic Characteristics,” Frontiers in Endocrinology 13 (2022): 966572. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Polyzos S. A. and Anastasilakis A. D., “Clinical Complications Following Thyroid Fine‐Needle Biopsy: A Systematic Review,” Clinical Endocrinology 71, no. 2 (August 2009): 157–165. [DOI] [PubMed] [Google Scholar]
  • 13. Kwong N., Medici M., Angell T. E., et al., “The Influence of Patient Age on Thyroid Nodule Formation, Multinodularity, and Thyroid Cancer Risk,” The Journal of Clinical Endocrinology & Metabolism 100, no. 12 (December 2015): 4434–4440. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Wildman‐Tobriner B., Taghi‐Zadeh E., and Mazurowski M. A., “Artificial Intelligence (AI) Tools for Thyroid Nodules on Ultrasound, From the AJR Special Series on AI Applications,” American Journal of Roentgenology 219, no. 4 (October 2022): 1–8. [DOI] [PubMed] [Google Scholar]
  • 15. Reverter J. L., Vázquez F., and Puig‐Domingo M., “Diagnostic Performance Evaluation of a Computer‐Assisted Imaging Analysis System for Ultrasound Risk Stratification of Thyroid Nodules,” American Journal of Roentgenology 213, no. 1 (July 2019): 169–174. [DOI] [PubMed] [Google Scholar]
  • 16. Han M., Ha E. J., and Park J. H., “Computer‐Aided Diagnostic System for Thyroid Nodules on Ultrasonography: Diagnostic Performance Based on the Thyroid Imaging Reporting and Data System Classification and Dichotomous Outcomes,” American Journal of Neuroradiology 42, no. 3 (March 2021): 559–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Cibas E. S. and Ali S. Z., “The Bethesda System for Reporting Thyroid Cytopathology,” Thyroid 19, no. 11 (November 2009): 1159–1165. [DOI] [PubMed] [Google Scholar]
  • 18. Parikh C. R., Coca S. G., Thiessen‐Philbrook H., et al., “Postoperative Biomarkers Predict Acute Kidney Injury and Poor Outcomes After Adult Cardiac Surgery,” Journal of the American Society of Nephrology 22, no. 9 (September 2011): 1748–1757. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19. Steyerberg E. W., Vickers A. J., Cook N. R., et al., “Assessing the Performance of Prediction Models: A Framework for Traditional and Novel Measures,” Epidemiology 21, no. 1 (January 2010): 128–138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Remonti L. R., Kramer C. K., Leitão C. B., Pinto L. C. F., and Gross J. L., “Thyroid Ultrasound Features and Risk of Carcinoma: A Systematic Review and Meta‐Analysis of Observational Studies,” Thyroid 25, no. 5 (May 2015): 538–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Nixon I. J., Ganly I., Hann L. E., et al., “Nomogram for Selecting Thyroid Nodules for Ultrasound‐Guided Fine‐Needle Aspiration Biopsy Based on a Quantification of Risk of Malignancy,” Head & Neck 35, no. 7 (July 2013): 1022–1025. [DOI] [PubMed] [Google Scholar]
  • 22. Horvath E., Majlis S., Rossi R., et al., “An Ultrasonogram Reporting System for Thyroid Nodules Stratifying Cancer Risk for Clinical Management,” The Journal of Clinical Endocrinology & Metabolism 94, no. 5 (May 2009): 1748–1751. [DOI] [PubMed] [Google Scholar]
  • 23. Kwak J. Y., Han K. H., Yoon J. H., et al., “Thyroid Imaging Reporting and Data System for US Features of Nodules: A Step in Establishing Better Stratification of Cancer Risk,” Radiology 260, no. 3 (September 2011): 892–899. [DOI] [PubMed] [Google Scholar]
  • 24. Ha E. J., Chung S. R., Na D. G., et al., “2021 Korean Thyroid Imaging Reporting and Data System and Imaging‐Based Management of Thyroid Nodules: Korean Society of Thyroid Radiology Consensus Statement and Recommendations,” Korean Journal of Radiology 22, no. 12 (December 2021): 2094–2123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Shin J. H., Baek J. H., Chung J., et al., “Ultrasonography Diagnosis and Imaging‐Based Management of Thyroid Nodules: Revised Korean Society of Thyroid Radiology Consensus Statement and Recommendations,” Korean Journal of Radiology 17, no. 3 (May/June 2016): 370–395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Moon W. J., Jung S. L., Lee J. H., et al., “Benign and Malignant Thyroid Nodules: US Differentiation—Multicenter Retrospective Study,” Radiology 247, no. 3 (June 2008): 762–770. [DOI] [PubMed] [Google Scholar]
  • 27. Rago T. and Vitti P., “Diagnostic Role of Ultrasound and Elastosonography in Nodular Goiter,” Best Practice & Research Clinical Endocrinology & Metabolism 28, no. 4 (August 2014): 519–529. [DOI] [PubMed] [Google Scholar]
  • 28. Trimboli P., Castellana M., Piccardo A., et al., “The Ultrasound Risk Stratification Systems for Thyroid Nodule Have Been Evaluated Against Papillary Carcinoma. A Meta‐Analysis,” Reviews in Endocrine & Metabolic Disorders 22, no. 2 (June 2021): 453–460. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Teefey S. A., Middleton W. D., Reading C. C., et al., “Effect of Decreasing the ACR TI‐RADS Point Assignment for Punctate Echogenic Foci When They Occur in Mixed Solid and Cystic Thyroid Nodules,” American Journal of Roentgenology 216, no. 2 (February 2021): 479–485. [DOI] [PubMed] [Google Scholar]
  • 30. Papini E., Guglielmi R., Bianchini A., et al., “Risk of Malignancy in Nonpalpable Thyroid Nodules: Predictive Value of Ultrasound and Color‐Doppler Features,” The Journal of Clinical Endocrinology & Metabolism 87, no. 5 (May 2002): 1941–1946. [DOI] [PubMed] [Google Scholar]
  • 31. Khoo M. L., Asa S. L., Witterick I. J., and Freeman J. L., “Thyroid Calcification and Its Association With Thyroid Carcinoma,” Head & Neck 24, no. 7 (July 2002): 651–655. [DOI] [PubMed] [Google Scholar]
  • 32. Peccin S., de Castro J. A. S., Furlanetto T. W., Furtado A. P. A., Brasil B. A., and Czepielewski M. A., “Ultrasonography: Is It Useful in the Diagnosis of Cancer in Thyroid Nodules?,” Journal of Endocrinological Investigation 25, no. 1 (January 2002): 39–43. [DOI] [PubMed] [Google Scholar]
  • 33. Yoon D. Y., Lee J. W., Chang S. K., et al., “Peripheral Calcification in Thyroid Nodules: Ultrasonographic Features and Prediction of Malignancy,” Journal of Ultrasound in Medicine 26, no. 10 (October 2007): 1349–1355; quiz 56–57. [DOI] [PubMed] [Google Scholar]
  • 34. Kim B. M., Kim M. J., Kim E. K., et al., “Sonographic Differentiation of Thyroid Nodules With Eggshell Calcifications,” Journal of Ultrasound in Medicine 27, no. 10 (October 2008): 1425–1430. [DOI] [PubMed] [Google Scholar]
  • 35. Cappelli C., Castellano M., Pirola I., et al., “Thyroid Nodule Shape Suggests Malignancy,” European Journal of Endocrinology 155, no. 1 (July 2006): 27–31. [DOI] [PubMed] [Google Scholar]
  • 36. Leenhardt L., Erdogan M. F., Hegedus L., et al., “2013 European Thyroid Association Guidelines for Cervical Ultrasound Scan and Ultrasound‐Guided Techniques in the Postoperative Management of Patients With Thyroid Cancer,” European Thyroid Journal 2, no. 3 (September 2013): 147–159. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37. Zhao H. and Li H., “Meta‐Analysis of Ultrasound for Cervical Lymph Nodes in Papillary Thyroid Cancer: Diagnosis of Central and Lateral Compartment Nodal Metastases,” European Journal of Radiology 112 (March 2019): 14–21. [DOI] [PubMed] [Google Scholar]
  • 38. Paparodis R., Livadas S., Karvounis E., et al., “Elevated Preoperative TPO Ab Titers Decrease Risk for DTC in a Linear Fashion: A Retrospective Analysis of 1635 Cases,” The Journal of Clinical Endocrinology & Metabolism 109, no. 1 (December 2023): e347–e355. [DOI] [PubMed] [Google Scholar]
  • 39. Januś D., Kujdowicz M., Wójcik M., et al., “Ultrasound Evolution of Parenchymal Changes in the Thyroid Gland With Autoimmune Thyroiditis in Children Prior to the Development of Papillary Thyroid Carcinoma—A Follow‐Up Study,” Frontiers in Endocrinology 14 (2023): 1172823. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Silva de Morais N., Stuart J., Guan H., et al., “The Impact of Hashimoto Thyroiditis on Thyroid Nodule Cytology and Risk of Thyroid Cancer,” Journal of the Endocrine Society 3, no. 4 (April 2019): 791–800. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Cho Y. A., Kong S. Y., Shin A., et al., “Biomarkers of Thyroid Function and Autoimmunity for Predicting High‐Risk Groups of Thyroid Cancer: A Nested Case–Control Study,” BMC Cancer 14 (November 2014): 873. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Liu Y., Li C., Zhao W., and Wang Y., “Hashimoto's Thyroiditis Is an Important Risk Factor of Papillary Thyroid Microcarcinoma in Younger Adults,” Hormone and Metabolic Research 49, no. 10 (October 2017): 732–738. [DOI] [PubMed] [Google Scholar]
  • 43. Li L., Shan T., Sun X., et al., “Positive Thyroid Peroxidase Antibody and Thyroglobulin Antibody Are Associated With Better Clinicopathologic Features of Papillary Thyroid Cancer,” Endocrine Practice 27, no. 4 (April 2021): 306–311. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

FIGURE 1 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4a thyroid nodules.

CEN-102-79-s002.tif (1.9MB, tif)

FIGURE 2 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4b thyroid nodules.

CEN-102-79-s003.tif (1.6MB, tif)

FIGURE 3 | Representative ultrasound and cytological images of malignant (A, C) and benign (B, D) ACR TI‐RADS 4c thyroid nodules.

CEN-102-79-s001.tif (1.7MB, tif)

Data Availability Statement

The data that support the findings of this study are openly available in the Science Data Bank at https://www.scidb.cn/en/s/QRFrQz.

Articles from Clinical Endocrinology are provided here courtesy of Wiley

RESOURCES