Assessment of malignant risk in thyroid nodules classified as category 4 and above against a background of Hashimoto’s thyroiditis

Introduction

Background

Thyroid nodules represent one of the most prevalent clinical thyroid disorders. High-resolution thyroid ultrasonography (US), valued for its non-invasiveness and accessibility, serves as the cornerstone for nodule evaluation (1). To standardize risk stratification, the Thyroid Imaging, Reporting and Data System (TI-RADS) has been widely adopted. Nodules categorized as TI-RADS 4 and above are considered suspicious for malignancy, necessitating further diagnostic workup (2).

Hashimoto’s thyroiditis (HT), or chronic lymphocytic thyroiditis, is the most common autoimmune thyroid disease. Its incidence, reported between 0.03% and 0.15% and rising, shows a marked female predominance (3-6). Notably, thyroid nodules co-exist with HT in approximately 53% of patients (7). The pathogenesis of HT involves autoimmune-mediated destruction of thyroid follicular cells, characterized histologically by lymphoplasmacytic infiltration, lymphoid follicle formation, and parenchymal atrophy (3). This chronic inflammatory process not only leads to clinical manifestations such as hypothyroidism and goiter but also induces diffuse, heterogeneous alterations in the thyroid parenchyma (3,8,9). These changes can promote the development of benign hyperplastic nodules (10). On ultrasound, HT often presents with diffuse hypoechogenicity, indistinct margins, and pseudonodular changes (10)—features that significantly overlap with the classic sonographic appearance of malignant thyroid nodules (11-13). Consequently, differentiating truly suspicious or malignant nodules within an HT background poses a substantial diagnostic challenge, potentially compromising the accuracy of preoperative ultrasound assessment.

Rationale and knowledge gap

The potential link between HT and thyroid cancer has been debated since Dailey et al.’s initial report in 1955 (14). Contemporary research on thyroid nodules in HT predominantly focuses on comparing overall malignancy rates between HT and non-HT cohorts, yielding conflicting conclusions regarding its diagnostic impact (15-17). A critical and underexplored question persists: does HT influence the specific malignancy risk and the predictive value of sonographic features within the high-risk subset of nodules already classified as TI-RADS 4 and above? Current evidence lacks a detailed analysis of how HT modulates the diagnostic weight of individual ultrasound characteristics in this clinically crucial group.

Objective

To address this gap, we conducted a large-scale retrospective cohort study. This study aimed to: (I) compare the histopathologically confirmed malignancy rate of TI-RADS 4+ nodules between patients with and without HT; (II) identify and compare the independent predictive value of specific sonographic features for malignancy in these two distinct populations; and (III) elucidate the interaction between HT status and key ultrasound characteristics. Our findings seek to provide a refined, evidence-based framework for risk assessment and to inform personalized management strategies for patients with HT presenting with sonographically suspicious thyroid nodules. We present this article in accordance with the STROBE reporting checklist (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0024/rc).

Methods

Study design and participants

This retrospective cohort study was conducted in the Department of Vascular and Thyroid Surgery, The First Affiliated Hospital of Xinjiang Medical University. We screened patients who underwent preoperative thyroid ultrasound assessment between January 2018 and December 2024. Inclusion criteria were: (I) a thyroid nodule classified as TI-RADS category 4 or higher; (II) availability of preoperative thyroid function and autoantibody test results; and (III) availability of definitive postoperative histopathological diagnosis. HT was diagnosed if the patient met at least one of the following criteria: (I) preoperative seropositivity for at least one thyroid autoantibody [thyroglobulin antibody (TgAb) or thyroid peroxidase antibody (TPOAb)]; or (II) histopathological confirmation of HT from the surgical specimen. Eligible patients were assigned to the HT group (TI-RADS 4+ nodules with HT background). Patients who did not meet the HT criteria constituted the non-HT group (TI-RADS 4+ nodules without HT).

Exclusion criteria were: (I) incomplete ultrasound data or absence of a TI-RADS classification; (II) ultrasound findings indicating a TI-RADS category below 4; (III) lack of preoperative thyroid serology results; (IV) a previous history of thyroid, parathyroid, or other neck surgery; (V) incomplete medical records; or (VI) a concurrent diagnosis of other malignancies or rheumatic autoimmune diseases. A total of 2,340 patients were finally included in the analysis. The patient selection flowchart is presented in Figure 1. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University (No. K202601-27) and informed consent was taken from all the patients.

Figure 1 Flowchart of patient selection. TI-RADS, Thyroid Imaging, Reporting and Data System.

Ultrasound examination and feature assessment

All patients underwent a standardized high-frequency ultrasound examination. Examinations were performed by two sonographers, each with over ten years of experience in thyroid ultrasound, using high-resolution ultrasound systems equipped with a linear array probe (frequency range: 10–14 MHz). The systems included models from Canon Medical Systems, Samsung Medison, GE Healthcare, Mindray, and Fornia Medical. The following sonographic features of the index nodule (the nodule with the highest TI-RADS score) were systematically recorded: maximum diameter, location, composition (cystic, predominantly cystic, predominantly solid, or solid), echogenicity (hyperechoic, isoechoic, hypoechoic, or markedly hypoechoic), margin (smooth, ill-defined, irregular, or extrathyroidal extension), shape (wider-than-tall or taller-than-wide), presence of hyperechoic foci (with or without acoustic shadowing), type of calcification (microcalcification, macrocalcification, or rim calcification), and internal echotexture (homogeneous or heterogeneous). All nodules were assessed and stratified according to the 2020 Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) guidelines (18).

Laboratory assays

Preoperative serum levels of TgAb and TPOAb were measured by electrochemiluminescence immunoassay on a Cobas E601 automated analyzer (Roche Diagnostics, Mannheim, Germany) using the manufacturer’s reagents, calibrators, and quality control materials. The reference ranges were defined as 0–115 IU/mL for TgAb and 0–34 IU/mL for TPOAb.

Reference standard

The definitive diagnosis (gold standard) was established by either ultrasound-guided fine-needle aspiration biopsy (FNAB) cytology, reported using the Bethesda System for Reporting Thyroid Cytopathology, or by postoperative histopathological examination of the surgical specimen. Malignancy was confirmed for all pathological subtypes, including but not limited to papillary thyroid carcinoma, follicular thyroid carcinoma, medullary thyroid carcinoma, oncocytic carcinoma, poorly differentiated carcinoma, anaplastic carcinoma, as well as carcinomas with extrathyroidal extension or lymph node metastasis.

Statistical analysis

Statistical analyses were performed using SPSS software (version 27.0.1; IBM Corp., Armonk, NY, USA) and R software (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables are presented as median with interquartile range (IQR), as the data exhibited non-normal distribution (assessed by the Shapiro-Wilk test, P<0.05). Comparisons between the HT and non-HT groups were performed using the Mann-Whitney U test. Categorical variables are presented as number (percentage) and were compared using the chi-square test or Fisher’s exact test, as appropriate.

To identify independent predictors of malignancy, univariate analyses were first performed on candidate variables (including clinical baseline data and all sonographic features). Variables with a P value <0.05 in the univariate analysis were subsequently entered into a multivariate logistic regression model. Separate models were constructed for the HT and non-HT groups. The discriminative performance of each group-specific prediction model was evaluated by plotting receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC). The pROC package (version 1.19.0.1) in R was used for this analysis. The difference in AUC between the two models was compared using DeLong’s test.

To further assess the clinical utility of the models, we determined the optimal probability cutoff for each model by maximizing Youden’s index (sensitivity + specificity − 1). Based on these optimal cutoffs, we calculated the corresponding sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). All AUC estimates were validated using R software. A two-sided P value <0.05 was considered statistically significant throughout the analysis.

Results

Patient characteristics

The study cohort comprised 2,340 patients with thyroid nodules classified as TI-RADS category 4 or higher. Based on the presence of HT, patients were stratified into an HT group (n=733) and a non-HT group (n=1,607). The distribution of age was non-normal in both groups (Shapiro-Wilk test, P<0.05). The median age was 48 years (IQR: 38–55 years) in the HT group and 48 years (IQR: 38–56 years) in the non-HT group. Comparative analysis using the Mann-Whitney U test indicated no significant difference between the two groups (U=575666.5, P=0.38). The proportion of female patients was significantly higher in the HT group (85.4%, 626/733) compared to the non-HT group (69.6%, 1,119/1,607; χ²=66.0, P<0.001). The distribution of TI-RADS categories (4, 5, etc.) was comparable between the two groups (P>0.05).

Postoperative histopathological examination confirmed malignancy in 1,972 of the 2,340 nodules, yielding an overall malignancy rate of 84.3% [95% confidence interval (CI): 82.8–85.8%]. The malignancy rate was significantly higher in the HT group (86.6%, 635/733; 95% CI: 84.3–89.0%) than in the non-HT group (83.2%, 1,337/1,607; 95% CI: 81.3–85.0%; χ²=5.12, P=0.02). Detailed baseline characteristics are presented in Table 1.

Univariate and multivariate analysis of sonographic features

Univariate analysis

Within both the HT group (Table 2) and the non-HT group (Table 3), malignant nodules demonstrated a significantly higher prevalence of several suspicious sonographic features compared to benign nodules. These features included solid composition, hypoechogenicity, indistinct margins, irregular shape (taller-than-wide), hyperechoic foci without acoustic shadowing, the presence of any calcification, and heterogeneous internal echotexture (all P<0.001).

Multivariate analysis

Variables significant in the univariate analysis (P<0.05) were entered into separate multivariate logistic regression models for the HT and non-HT groups. For the HT group, independent predictors of malignancy were: hypoechogenicity [odds ratio (OR) =2.54, 95% CI: 1.27–5.09; P=0.009], irregular shape (OR =1.85, 95% CI: 1.10–3.11; P=0.02), and the presence of calcification (OR =2.09, 95% CI: 1.23–3.53; P=0.006). For the non-HT group, independent predictors were: hypoechogenicity (OR=3.18, 95% CI: 2.15–4.69; P<0.001), irregular margins (OR =2.02, 95% CI: 1.48–2.75; P<0.001), irregular shape (OR =1.69, 95% CI: 1.24–2.30; P=0.001), and heterogeneous internal echotexture (OR =3.65, 95% CI: 1.90–6.99; P<0.001). Notably, irregular margins and heterogeneous echotexture were not retained as independent predictors in the HT model. Full results are detailed in Table 4.

Predictive performance of sonographic features across groups

Comparison of feature-specific PPV

We compared the PPV (malignancy rate when the feature is present) for each suspicious sonographic feature between the HT and non-HT groups. The PPV for “hyperechoic foci without acoustic shadowing” and for “calcification” was significantly higher in the HT group than in the non-HT group (P<0.05 for both). For the other features (solid composition, hypoechogenicity, indistinct margins, irregular shape, heterogeneous echotexture), there was no statistically significant difference in PPV between the two groups (all P>0.05). These results are summarized in Figure 2 and Table 5.

Figure 2 Comparison of malignancy prediction rates for different ultrasonic malignant features between HT and non-HT groups. *, P<0.05; **, P<0.01; ns, P>0.05. HT, Hashimoto’s thyroiditis.

Synergistic effect of calcification and HT

To further investigate the interaction, we analyzed malignancy rates based on the combination of “calcification” and “hyperechoic foci without shadowing”. In the non-HT group, the malignancy rates were: 79.4% for nodules with neither feature, 85.8% for those with only hyperechoic foci, 81.3% for those with only calcification, and 86.4% for those with both. In the HT group, a distinct pattern emerged: 81.2% (neither feature), 84.1% (only hyperechoic foci), 87.7.3% (only calcification), and 92.3% (both features). Risk ratio (RR) analysis indicated that HT significantly amplified the risk associated with calcification: RR =1.08 (95% CI: 1.02–1.14) for “calcification only” and RR =1.07 (95% CI: 1.01–1.13) for “both features”. Conversely, HT did not amplify the risk associated with “hyperechoic foci only” (RR =0.98; 95% CI: 0.91–1.06). This demonstrates a specific synergistic effect between HT and the presence of calcification (Figure 3A,3B).

Figure 3 Synergistic effect characteristics of ultrasonic features like hyperechoic foci without acoustic shadowing and calcification in HT and non-HT groups: (A) synergistic effect of calcification and hyperechoic foci without acoustic shadowing in an HT background. (B) Malignancy RR of HT group relative to non-HT group (HT group malignancy rate/non-HT group malignancy rate). Δ, absolute risk difference. HT, Hashimoto’s thyroiditis; RR, risk ratio.

Positive likelihood ratio (LR+) analysis of sonographic features

To evaluate the diagnostic value of individual sonographic features independently of disease prevalence, we calculated the LR+ for each feature in both groups. As shown in Table 6, all LR+ values ranged from 1.0 to 1.6, with no single feature exceeding 2. The LR+ for calcification was 1.54 (95% CI: 1.45–1.63) in the HT group and 1.14 (95% CI: 1.09–1.19) in the non-HT group, with non‑overlapping CIs. The LR+ for tumor margin was 1.32 (95% CI: 1.22–1.42) in the HT group and 1.62 (95% CI: 1.52–1.72) in the non-HT group, also with non-overlapping CIs.

Performance of group-specific prediction models

Logistic regression models were constructed using the respective sets of independent predictors identified for the HT and non-HT groups. The model for the HT group yielded an AUC of 0.680 (95% CI: 0.626–0.734). The model for the non-HT group yielded an AUC of 0.686 (95% CI: 0.649–0.723). The difference between the two AUCs was not statistically significant (DeLong’s test, P=0.84). This indicates that while the constitutive predictors differed between the groups, the overall discriminative ability of the prediction models remained comparable (Table 7, Figure 4).

Figure 4 ROC curves comparing ultrasonic malignant feature prediction models for the HT group and the non-HT group. The AUC is shown for each model; P value is derived from DeLong’s test. AUC, area under the curve; CI, confidence interval; HT, Hashimoto’s thyroiditis; ROC, receiver operating characteristic.

To further assess the clinical utility of these models, we determined the optimal probability cutoff for each model by maximizing Youden’s index. Based on these optimal cutoffs, we calculated the corresponding sensitivity, specificity, PPV, and NPV for both groups (Table 7). In the HT group, the optimal cutoff was 0.909, yielding a sensitivity of 39.9% and a specificity of 86.6%, with a PPV of 95.1% and an NPV of 18.0%. In the non-HT group, the optimal cutoff was 0.795, yielding a sensitivity of 81.5% and a specificity of 48.9%, with a PPV of 88.8% and an NPV of 34.8%. The HT model demonstrated higher specificity (86.6%) than sensitivity (39.9%), whereas the non‑HT model demonstrated higher sensitivity (81.5%) than specificity (48.9%; Table 7).

Discussion

Thyroid nodules represent a common clinical entity, with the primary diagnostic challenge lying in the accurate discrimination of the minority that are malignant (19-21). High-resolution US, complemented by standardized risk stratification systems such as the TI-RADS, serves as the cornerstone for this evaluation. Nodules categorized as TI-RADS 4 and above warrant heightened suspicion and further diagnostic investigation (1,2). Concurrently, HT, the most prevalent autoimmune thyroid disorder, is frequently associated with thyroid nodules, with co-occurrence rates reported between 10% and 58% (22). A critical and unresolved question is whether and how the distinct immune-inflammatory milieu of HT modulates the pre-operative risk assessment of these already suspicious (TI-RADS 4+) nodules. Specifically, does HT alter the predictive value of individual sonographic features? To address this, our large-scale retrospective cohort study systematically analyzed the malignancy risk and the sonographic predictors in patients with TI-RADS 4+ nodules, stratified by HT status.

Key finds

HT as an independent risk factor in high-risk nodules

Our study revealed a strikingly high overall malignancy rate of 84.3% among 2,340 patients with TI-RADS 4+ nodules. More importantly, within this high-risk cohort, the presence of HT conferred an additional, independent risk. The malignancy rate was significantly higher in the HT group (86.8%) compared to the non-HT group (83.2%), despite comparable distributions of TI-RADS categories. This finding robustly establishes HT as an independent risk factor for High-Risk Nodules even among nodules deemed suspicious by conventional imaging criteria, aligning with and extending previous reports (11,23-25). This suggests that the elevated risk is intrinsic to the HT disease state rather than being merely a reflection of nodules with more aggressive sonographic phenotypes at presentation.

The underlying mechanism likely resides in the chronic immune-inflammatory microenvironment characteristic of HT. Persistent lymphocytic infiltration, release of pro-inflammatory cytokines [e.g., interleukin-6, tumor necrosis factor (TNF)-α], and cycles of tissue damage and repair (3,9) are believed not only to foster nodulogenesis (26) but also to create a permissive landscape for the malignant transformation of thyroid follicular cells (27-30). Therefore, in clinical practice, the assessment of a TI-RADS 4+ nodule should integrate the patient’s HT status as a key component of a comprehensive risk stratification model, moving beyond reliance on imaging features alone to guide personalized management.

Recalibration of sonographic predictors by the HT background

Our analysis revealed a fundamental recalibration of the predictive hierarchy of sonographic features in the context of HT. While the classic feature set of hypoechogenicity, indistinct margins, irregular shape, and heterogeneous echotexture formed the core predictive model in the non-HT group—closely mirroring the emphasis of standard TI-RADS (31)—a distinct model emerged for the HT group. Here, independent predictors were simplified to hypoechogenicity, irregular shape, and the presence of calcification.

This shift is mechanistically insightful. The loss of independent predictive value for “indistinct margins” in the HT group is a pivotal finding, contradicting some earlier studies (32-34). We postulate that this is a direct consequence of the diffuse parenchymal pathology of HT. Widespread lymphocytic infiltration and associated fibrosis disrupt the normal tissue architecture and acoustic interfaces (26,35), likely causing both benign and malignant nodules to frequently display indistinct margins, thereby diminishing this feature’s specificity for malignancy.

Conversely, “calcification” emerged with enhanced predictive power in the HT group. This “model reshaping” underscores that applying a universal TI-RADS algorithm, derived primarily from a non-HT population, to nodules in an HT background may introduce significant assessment bias. Clinically, this mandates caution in interpreting “indistinct margins” in HT patients and suggests that “calcification” should be accorded greater diagnostic weight in this specific population.

Feature-selective synergistic effect of HT

Delving deeper into feature-specific performance, we discovered that the risk-amplifying effect of HT is not uniform across all suspicious features. While the PPV of most features was similar between groups, the PPV for “hyperechoic foci without shadowing” and particularly for “calcification” was significantly higher in the HT group.

Most notably, our analysis of feature combinations uncovered a specific synergistic interaction. HT did not augment the malignancy risk associated with “hyperechoic foci without shadowing” alone. However, a clear synergistic effect was observed when HT coexisted with calcification. Nodules in the HT group featuring “calcification only” or “calcification with hyperechoic foci” exhibited malignancy rates exceeding 87%, with significantly elevated relative risks. This indicates that HT selectively potentiates the risk signaled by calcification.

The pathological nexus between HT and calcification may involve shared pathways of chronic inflammation promoting tissue fibrosis and dystrophic calcification (26). This novel observation of a selective synergy highlights that in HT patients, the identification of calcification within a suspicious nodule should trigger a substantially higher index of suspicion for malignancy.

Clinical implications and model performance

Importantly, formal comparison of the two group‑specific models using DeLong’s test revealed no significant difference in their discriminative performance (AUC 0.680 vs. 0.686, P=0.84). This finding is central to our thesis: although HT fundamentally alters which sonographic features independently predict malignancy, the overall diagnostic accuracy achievable with HT-appropriate feature weighting is equivalent to that in the non-HT population. Thus, a stratified, HT‑aware assessment strategy can achieve the same level of risk discrimination as conventional approaches, without compromising diagnostic efficacy.

This statistical equivalence argues compellingly against a “one-size-fits-all” approach and lays the groundwork for developing refined, context-specific risk assessment tools. For clinical practice, our results advocate for a dual adjustment: first, incorporating HT status as a binary risk modifier; and second, mentally recalibrating the diagnostic weight of sonographic features when evaluating nodules in patients with known HT.

Strengths and limitations

This study has several strengths. First, it is based on a large, well-characterized surgical cohort, providing robust histopathological confirmation for all included nodules. Second, we employed pragmatic and clinically relevant diagnostic criteria for HT (seropositivity and/or histopathological confirmation), enhancing the real-world applicability and generalizability of our findings. Third, the comprehensive analysis of sonographic features within a standardized TI-RADS framework allows for direct comparison with established guidelines.

Nevertheless, certain limitations must be acknowledged. The retrospective, single-center design may introduce selection bias and limit generalizability; our findings require validation in prospective, multi-center studies. We acknowledge that external validation of our group-specific prediction models in independent, diverse populations is an essential next step before these findings can be broadly integrated into clinical practice. Such validation efforts are currently being planned. Furthermore, while we focused on sonographic anatomy, we did not integrate potentially synergistic biochemical data [e.g., thyroid-stimulating hormone (TSH) levels, antibody titers] into the predictive models. Future research should aim to construct multi-modal risk assessment tools that combine imaging, serological markers, and perhaps artificial intelligence to achieve more nuanced, quantitative risk stratification for nodules in an HT background.

Comparison with similar research

Our finding that HT is an independent risk factor for malignancy in TI-RADS 4+ nodules aligns with a growing body of literature suggesting an association between chronic thyroiditis and thyroid cancer (11,23-25). However, our study advances this field by specifically focusing on the already suspicious nodule population, demonstrating that HT adds risk beyond that captured by imaging alone.

Regarding sonographic predictors, our observation that the predictive hierarchy changes in HT provides a potential explanation for conflicting results in prior studies. While some reports found features like “indistinct margins” retained value in HT (32-34), our data suggest its diagnostic specificity is significantly diminished, likely masked by the diffuse parenchymal changes of HT itself. Conversely, our finding of a strengthened association between calcification and malignancy in HT patients resonates with prior work highlighting calcification as a particularly worrisome feature in this context (36). The novel demonstration of a specific synergistic effect between HT and calcification, to our knowledge, has not been previously detailed in the literature.

Explanations of findings

The modest ORs observed in our multivariate analysis, with lower confidence limits approaching 1.0, reflect the inherent challenge of risk stratification in a high‑prevalence surgical population. To confirm this interpretation, we calculated LR+—a prevalence-independent measure of diagnostic value. All features demonstrated LR+ values below 2, indicating that no single sonographic characteristic possesses strong standalone discriminatory power in TI-RADS 4+ nodules.

Notably, the LR+ for calcification was significantly higher in the HT group than in the non-HT group, with non‑overlapping CIs confirming a synergistic effect between HT and calcification. Conversely, the LR+ for tumor margin was higher in the non-HT group than in the HT group, consistent with the loss of independent predictive value of margins in the HT group observed in multivariate analysis. These patterns align with the differential sets of independent predictors identified in Table 4 and support the conclusion that HT fundamentally reshapes the diagnostic hierarchy of sonographic features.

A case in point is the feature “heterogeneous internal echotexture”. In the non-HT group, this feature yielded an OR of 3.65, but its 95% CI was notably wide (1.90–6.99), signaling imprecision. This instability was corroborated by its LR+ (1.13, 95% CI: 0.82–1.58), a CI that includes 1.0, indicating that even the direction of association is uncertain. The underlying cause is sparse data: very few benign nodules in the non‑HT group exhibited this feature, leading to unstable estimates in both the logistic regression model and the raw 2×2 table. Clinically, this means that while heterogeneous echotexture may contribute to a multi-feature risk assessment, it should not be overinterpreted as a standalone predictor.

The altered predictive landscape in HT can be mechanistically linked to its underlying pathology. The diffuse lymphocytic infiltration and fibrosis characteristic of HT (3,26,35) obscure tissue planes, rendering “indistinct margins” a non-specific finding common to both benign and malignant processes. This fundamentally reduces its discriminatory power. The enhanced risk associated with calcification in HT may stem from shared pathological pathways. Chronic inflammation can promote tissue remodeling, fibrosis, and dystrophic calcification (26). Therefore, calcification in an HT nodule may not merely be a passive marker but could reflect a more aggressive, inflammation-driven pathological process, thereby carrying greater malignant potential. The selective synergy we observed—where HT amplifies the risk of calcification but not other features like hyperechoic foci—supports the notion of a specific biological interaction rather than a non-specific increase in background risk.

We acknowledge that the AUC of approximately 0.68 achieved by our prediction models indicates only moderate discriminatory capacity. This limitation reflects several factors inherent to our study design: the high baseline malignancy rate (84.3%) in this surgical cohort, the restriction to conventional sonographic features without incorporation of clinical or molecular data, the relative homogeneity of TI-RADS 4+ nodules, and the confounding effect of HT-induced parenchymal changes. Despite these constraints, the equivalent AUCs achieved in the HT and non-HT groups—using different sets of predictors—support our central thesis that HT fundamentally recalibrates the predictive hierarchy of sonographic features. Thus, the primary contribution of this study lies not in providing a high-performance prediction tool, but in generating mechanistic insights that can inform more nuanced, HT-aware risk stratification.

The clinical implications of the low NPV observed in our models merit further discussion. As shown in Table 7, the NPV was only 18.0% in the HT group and 34.8% in the non-HT group. This means that when our models predict a nodule to be benign, the prediction is incorrect in approximately 82% of HT patients and 65% of non‑HT patients. As shown in Table 7, the HT model demonstrated higher specificity (86.6%) than sensitivity (39.9%), whereas the non‑HT model demonstrated higher sensitivity (81.5%) than specificity (48.9%). This differential performance reflects the distinct impact of HT on the predictive landscape: in HT patients, the diffuse parenchymal changes increase the risk of false positives for malignancy, shifting the model toward specificity; in non‑HT patients, the sonographic features retain their classic discriminatory patterns, favoring sensitivity.

This limitation is not a failure of our specific models but a mathematical consequence of working in a high‑risk selected population with a pre‑test malignancy probability of 84.3%. In this context, no ultrasound‑based model can achieve a high NPV without incorporating additional data. Therefore, the clinical utility of our models lies not in ruling out malignancy—which is inherently unreliable—but in providing risk stratification among patients already destined for intervention. For the clinician, this means that a “benign” prediction from any ultrasound‑based model in a TI-RADS 4+ nodule should not delay or alter management decisions. The low NPV also highlights the urgent need for multimodal risk stratification tools that integrate sonographic features with clinical, serological, and molecular markers to achieve higher predictive accuracy in this challenging population. Future studies should focus on developing such integrated models to improve the preoperative identification of truly benign nodules among those classified as TI-RADS 4+.

Implications and actions needed

The relative risk increase associated with the HT-calcification synergy is modest (8%). However, in the context of a high baseline malignancy risk (exceeding 87%), even this modest relative increase translates into a clinically meaningful absolute risk difference of 3.6 percentage points. With a number needed to harm of 28, a high-volume center evaluating 1,000 TI-RADS 4+ patients annually would detect approximately 6 additional malignancies attributable to this synergy. For the clinician, this means that a calcified nodule in an HT patient warrants a higher index of suspicion, which may appropriately influence decisions regarding the threshold for intervention.

Our results have direct implications for clinical practice and research:

• Risk stratification: HT status should be formally incorporated into the preoperative risk assessment of TI-RADS 4+ nodules, potentially as a binary risk modifier in existing models or guidelines.
• Ultrasound interpretation: sonographers and clinicians should adopt a differentiated interpretative approach for patients with HT. This involves down-weighting the significance of “indistinct margins” while up-weighting the importance of “hypoechogenicity”, “irregular shape”, and, most critically, “calcification”.
• Personalized management: recognition of the HT-calcification synergy should inform patient counseling and surgical decision-making. A calcified TI-RADS 4 nodule in an HT patient may carry a risk profile approaching that of a higher-category nodule.
• Future research direction: the development and validation of HT-specific risk prediction models, integrating imaging and molecular data, is warranted to move beyond a “one-size-fits-all” diagnostic algorithm and towards truly personalized medicine in thyroid nodule evaluation.

Conclusions

In conclusion, this study establishes HT as an independent risk factor that significantly elevates the malignancy risk of thyroid nodules already deemed suspicious by conventional imaging (TI-RADS 4+). We demonstrate that the presence of HT fundamentally recalibrates the sonographic risk-prediction model: it diminishes the diagnostic value of classic features like “indistinct margins” while augmenting the predictive importance of “hypoechogenicity”, “irregular shape”, and, most critically, “calcification”. A key novel finding is the feature-selective nature of this risk modulation, with HT exhibiting a specific synergistic effect with the “calcification” feature rather than universally amplifying all suspicious sonographic findings. These insights, derived using pragmatic clinical diagnostic criteria for HT, underscore the necessity of incorporating HT status into preoperative risk stratification. Moving forward, adopting a differentiated, HT-aware approach to ultrasound interpretation and risk factor weighting is essential to advance the management of thyroid nodules towards greater precision and personalization.

Acknowledgments

We thank the supervisors and departmental colleagues for their guidance and strong support during this research and manuscript collaboration.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0024/rc

Data Sharing Statement: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0024/dss

Peer Review File: Available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0024/prf

Funding: This work was supported by National Natural Science Foundation of China (No.82560104); and Autonomous Region Scientific Research Plan Project (No. 2024D01D23).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://gs.amegroups.com/article/view/10.21037/gs-2026-1-0024/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University (No. K202601-27) and informed consent was taken from all the patients.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Liu XH, Yin HQ, Shen H, et al. A multivariable model of ultrasound and biochemical parameters for predicting high-volume lymph node metastases of papillary thyroid carcinoma with Hashimoto's thyroiditis. Front Endocrinol (Lausanne) 2024;15:1501142. [Crossref] [PubMed]
2. Vassallo E, Péporté A, McQueen A, et al. ESR Essentials: thyroid imaging-practice recommendations by the European Society of Head and Neck Radiology. Eur Radiol 2026;36:3788-96. [Crossref]
3. Ralli M, Angeletti D, Fiore M, et al. Hashimoto's thyroiditis: An update on pathogenic mechanisms, diagnostic protocols, therapeutic strategies, and potential malignant transformation. Autoimmun Rev 2020;19:102649. [Crossref] [PubMed]
4. Ragusa F, Fallahi P, Elia G, et al. Hashimotos' thyroiditis: Epidemiology, pathogenesis, clinic and therapy. Best Pract Res Clin Endocrinol Metab 2019;33:101367. [Crossref] [PubMed]
5. Resende de Paiva C, Grønhøj C, Feldt-Rasmussen U, et al. Association between Hashimoto's Thyroiditis and Thyroid Cancer in 64,628 Patients. Front Oncol 2017;7:53. [Crossref] [PubMed]
6. Batool S, Das B, Arif M, et al. Frequency of hashimoto thyroiditis in papillary thyroid cancer patients and its impact on their Outcome. J Ayub Med Coll Abbottabad 2022;34:251-5. [Crossref] [PubMed]
7. Zhao C, Xiong H, Zhu L, et al. An investigation of the pattern and mechanism of comorbidity in patients with Hashimoto's thyroiditis. Front Endocrinol (Lausanne) 2025;16:1615095. [Crossref] [PubMed]
8. Gan XX, Li YY, Li SJ, et al. Significance of DMBT1 in Papillary Thyroid Carcinoma Concurrent With Hashimoto's Thyroiditis. Front Oncol 2021;11:680873. [Crossref] [PubMed]
9. Li R, He T, Xing Z, et al. The immune system in Hashimoto's thyroiditis: Updating the current state of knowledge on potential therapies and animal model construction. Autoimmun Rev 2025;24:103783. [Crossref] [PubMed]
10. El Aghoury A, Abd ElHafeez S, El Sabaa B, et al. Hashimoto's thyroiditis in an Egyptian cohort: clinical, functional, and ultrasonographic features with insights into nodule risk. Front Endocrinol (Lausanne) 2025;16:1664047. [Crossref] [PubMed]
11. Zhao T, Xu S, Zhang X, et al. Comparison of Various Ultrasound-Based Malignant Risk Stratification Systems on an Occasion for Assessing Thyroid Nodules in Hashimoto's Thyroiditis. Int J Gen Med 2023;16:599-608. [Crossref] [PubMed]
12. Tan HL, Ge H, Qin ZE, et al. Evaluating the diagnostic efficiency of ultrasound and serum autoantibodies in Hashimoto's thyroiditis: a cross-sectional study. Sci Rep 2025;15:12674. [Crossref] [PubMed]
13. Mao L, Zheng C, Ou S, et al. Influence of Hashimoto thyroiditis on diagnosis and treatment of thyroid nodules. Front Endocrinol (Lausanne) 2022;13:1067390. [Crossref] [PubMed]
14. DAILEY ME. LINDSAY S, SKAHEN R. Relation of thyroid neoplasms to Hashimoto disease of the thyroid gland. AMA Arch Surg 1955;70:291-7. [Crossref] [PubMed]
15. Hu M, Zong S, Xu N, et al. The Value of Thyroid Ultrasound Computer-Aided Diagnosis System in the Evaluation of Thyroid Nodules With Concurrent Hashimoto's Thyroiditis. J Ultrasound Med 2022;41:1117-24. [Crossref] [PubMed]
16. Park SK, Ryoo JH, Kim MH, et al. Association Between Eight Autoimmune Diseases and Thyroid Cancer: A Nationwide Cohort Study. Thyroid 2024;34:206-14. [Crossref] [PubMed]
17. Feldt-Rasmussen U. Hashimoto's thyroiditis as a risk factor for thyroid cancer. Curr Opin Endocrinol Diabetes Obes 2020;27:364-71. [Crossref] [PubMed]
18. Superficial Organ and Vascular Ultrasound Group, Society of Ultrasound in Medicine, Chinese Medical Association. 2020 Chinese Guidelines for Ultrasound Malignancy Risk Stratification of Thyroid Nodules: The C-TIRADS. Chinese Journal of Ultrasonography 2021;30:185-200.
19. Roy M, Ghander C, Bigorgne C, et al. An update on management of cytologically indeterminate thyroid nodules. Ann Endocrinol (Paris) 2025;86:101711. [Crossref] [PubMed]
20. Alexander EK, Doherty GM, Barletta JA. Management of thyroid nodules. Lancet Diabetes Endocrinol 2022;10:540-8. [Crossref] [PubMed]
21. Kobaly K, Kim CS, Mandel SJ. Contemporary Management of Thyroid Nodules. Annu Rev Med 2022;73:517-28. [Crossref] [PubMed]
22. Fang M, Lei M, Chen X, et al. Radiomics-based ultrasound models for thyroid nodule differentiation in Hashimoto's thyroiditis. Front Endocrinol (Lausanne) 2023;14:1267886. [Crossref] [PubMed]
23. Vita R, Ieni A, Tuccari G, et al. The increasing prevalence of chronic lymphocytic thyroiditis in papillary microcarcinoma. Rev Endocr Metab Disord 2018;19:301-9. [Crossref] [PubMed]
24. Uhliarova B, Hajtman A. Hashimoto’s thyroiditis—an independent risk factor for papillary carcinoma. Braz J Otorhinolaryngol 2018;84:729-35. [Crossref] [PubMed]
25. Le Y, Geng C, Gao X, et al. The risk of thyroid cancer and sex differences in Hashimoto's thyroiditis, a meta-analysis. BMC Endocr Disord 2024;24:151. [Crossref] [PubMed]
26. Ehlers M, Schott M. Hashimoto's thyroiditis and papillary thyroid cancer: are they immunologically linked? Trends Endocrinol Metab 2014;25:656-64. [Crossref] [PubMed]
27. Xue X, Wu D, Yao H, et al. Mechanisms underlying the promotion of papillary thyroid carcinoma occurrence and progression by Hashimoto's thyroiditis. Front Endocrinol (Lausanne) 2025;16:1551271. [Crossref] [PubMed]
28. Boi F, Pani F, Calò PG, et al. High prevalence of papillary thyroid carcinoma in nodular Hashimoto's thyroiditis at the first diagnosis and during the follow-up. J Endocrinol Invest 2018;41:395-402. [Crossref] [PubMed]
29. Jackson D, Handelsman RS, Farrá JC, et al. Increased Incidental Thyroid Cancer in Patients With Subclinical Chronic Lymphocytic Thyroiditis. J Surg Res 2020;245:115-8. [Crossref] [PubMed]
30. Harmantepe AT, Ozdemir K, Bayhan Z, et al. The Underestimated Impact of Hashimoto Thyroiditis on Thyroid Papillary Carcinoma. Updates Surg 2024;76:1085-9. [Crossref] [PubMed]
31. Tessler FN, Middleton WD, Grant EG, et al. ACR Thyroid Imaging, Reporting and Data System (TI-RADS): White Paper of the ACR TI-RADS Committee. J Am Coll Radiol 2017;14:587-95. [Crossref] [PubMed]
32. Baser H, Ozdemir D, Cuhaci N, et al. Hashimoto's Thyroiditis Does Not Affect Ultrasonographical, Cytological, and Histopathological Features in Patients with Papillary Thyroid Carcinoma. Endocr Pathol 2015;26:356-64. [Crossref] [PubMed]
33. Wang B, Ou X, Yang J, et al. Contrast-enhanced ultrasound and shear wave elastography in the diagnosis of ACR TI-RADS 4 and 5 category thyroid nodules coexisting with Hashimoto's thyroiditis. Front Oncol 2022;12:1022305. [Crossref] [PubMed]
34. Teng C, Pang K, Zhang L, et al. Impact of Hashimoto's thyroiditis on ultrasound diagnosis of papillary thyroid carcinoma: a retrospective study. Front Oncol 2025;15:1551114. [Crossref] [PubMed]
35. Khorrami M, Prasanna P, Gupta A, et al. Changes in CT Radiomic Features Associated with Lymphocyte Distribution Predict Overall Survival and Response to Immunotherapy in Non-Small Cell Lung Cancer. Cancer Immunol Res 2020;8:108-19. [Crossref] [PubMed]
36. Mao L, Zheng C, Ou S, et al. Influence of Hashimoto thyroiditis on diagnosis and treatment of thyroid nodules. Front Endocrinol 2022;13:1067390.