Thyroid cancer is the most prevalent endocrine malignancy, with its incidence steadily increasing over the past two decades due to factors such as improved healthcare accessibility and widespread thyroid imaging utilization (1, 2). Papillary thyroid carcinoma (PTC) accounts for 75–85% of cases and, together with follicular thyroid carcinoma (FTC), is classified as differentiated thyroid carcinoma (DTC) (3). The standard treatment for thyroid cancer includes near-total or total thyroidectomy, with neck dissection performed in cases of nodal metastases detected via preoperative imaging. Patients with distant metastases, commonly in the lungs or bones, undergo total thyroidectomy followed by radioactive iodine (RAI) uptake scanning and I-131 ablation therapy for remnant ablation or targeted metastatic treatment (4).

Most patients who receive appropriate treatment exhibit favourable survival outcomes, with cancer-related mortality rates below 10%. However, recurrence occurs in approximately 20–30% of cases, predominantly within the neck, either in the thyroid bed or lymph nodes, typically within the first decade post-treatment (5, 6). Regular follow-up involves physical examinations, ultrasound (US) assessments, and periodic serum thyroglobulin (Tg) measurements. Additional imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), 18-fluorodeoxyglucose positron emission tomography (¹⁸FDG-PET), or I-131 scanning may be employed based on clinical indications to detect residual disease or recurrence at an early stage. High-resolution neck US is considered the primary surveillance tool for detecting persistent or recurrent disease. The American Thyroid Association (ATA) recommends serum Tg measurement with anti-Tg antibodies (TgAb) at 6- to 12-month intervals post-surgery and/or RAI therapy, alongside periodic neck US evaluations (7).

Serum thyroglobulin (Tg) remains a cornerstone biochemical marker for post-thyroidectomy monitoring, yet its utility is limited in certain scenarios. Approximately one-fourth of DTC patients develop anti-thyroglobulin antibodies that interfere with assay results, and Tg levels may remain low despite residual or recurrent disease, particularly in non–radioiodine-avid or non-ablated patients (8–10). Although ultrasensitive Tg assays (functional sensitivity <0.1 ng/mL) have improved detection thresholds and reduced need for TSH stimulation, their standalone specificity remains modest, warranting integration with structural imaging modalities such as US for accurate disease surveillance (11, 12). Consequently, neck US, being non-invasive, widely available, radiation-free, and highly sensitive, remains the preferred modality for detecting abnormal lymph nodes or masses in the thyroid bed. The integration of colour Doppler enhances the assessment of thyroid bed nodules for malignant characteristics, though confirmation via fine-needle aspiration cytology is warranted (10, 13).

Despite its utility, ultrasonography has limitations, including operator dependency and the potential for false-positive findings leading to unnecessary additional testing. Emerging applications of artificial intelligence (AI) and machine learning (ML) are increasingly being explored to address these limitations by reducing operator dependency and enhancing diagnostic specificity. Deep learning models trained on large annotated ultrasound datasets can automatically detect and characterize key sonographic features, thus showing promise in improving reproducibility, streamlining workflow, and mitigating reader variability in thyroid imaging (14).

The ATA acknowledges that some residual disease often remains post-thyroidectomy and emphasizes the role of neck US in identifying or excluding neoplastic disease, even in patients with undetectable serum Tg levels (15). However, studies evaluating US efficacy in patients with undetectable Tg levels report variable findings, with one recent study identifying recurrence in only 1 out of 170 such patients (16). Furthermore, certain studies highlight a high rate of false-positive abnormalities on neck US, leading to excessive testing without detecting clinically significant disease (16). An inherent limitation of local US is also its inability to adequately assess hematogenous dissemination to distant sites, particularly in certain subtypes of thyroid carcinoma such as follicular carcinoma.

The primary outcomes of this study were to determine the incidence of abnormal findings on neck US in post-total thyroidectomy patients with DTC and to evaluate their correlation with persistent or recurrent locoregional disease. Secondary outcomes included establishing the relationship between Tg levels and locoregional disease persistence or recurrence in DTC patients.

This single-centre retrospective analytical study included patients who underwent neck US examinations at our institution over a 10-year period (1^st January 2009 to 31^st December 2019). The study population comprised patients who met specific inclusion and exclusion criteria based on clinical history, imaging records, and surgical and biochemical data.

Inclusion criteria required patients to have undergone neck US within the designated study period, with corresponding reports accessible in the radiology information system. Eligible patients had also undergone total thyroidectomy, with or without neck dissection or adjuvant radioiodine therapy, or had undergone revision surgery for suspected or confirmed recurrence, with or without additional radioiodine therapy. Exclusion criteria included a diagnosis of thyroid neoplasms other than DTC, treatment limited to lobectomy alone, or an indeterminate status regarding disease persistence or recurrence.

Surveillance for recurrent or persistent disease in post-thyroidectomy patients is critical, and US and serum Tg estimation remain cornerstone modalities in the follow-up of patients with DTC. US is especially valuable in monitoring residual thyroid tissue, detecting locoregional recurrent disease, and evaluating neck nodal metastases (17, 18). In this study, we examined the diagnostic performance of US and serum Tg, assessed the role of small nodal metastases, and evaluated recurrence timing to optimize surveillance strategies.

Locoregional recurrence rates in thyroid cancer range from 9-30%, depending on initial tumour burden, surgical completeness, and adjuvant therapy. In our cohort, 261 of 941 patients (27.7%) developed recurrence, aligning with the findings of Luo et al., who reported a 12% recurrence rate (19). The vast majority of thyroid bed recurrences (97.4%) were diagnosed within the first decade post-surgery, with a median interval of 11 months, emphasizing the need for early and sustained surveillance. Although 2.6% of recurrences occurred beyond 10 years post-surgery (mean interval: 12.1 years), these late recurrences suggest the possibility of slow-growing clones or dedifferentiation over time.

Patient age at surgery did not significantly impact recurrence rates, consistent with studies by Luo et al. and Ryoo et al. (2, 19). Furthermore, although neck dissection at the time of thyroidectomy was associated with a lower recurrence rate, this finding was not statistically significant, mirroring prior reports from Luo et al. (19). Postoperative RAI therapy significantly reduced recurrence rates, consistent with Toniato et al. (20), reinforcing its role in reducing microscopic residual disease and improving long-term disease-free survival.

US demonstrated high sensitivity (98.9%) and NPV (99.3%) in detecting locoregional disease, confirming its role as a first-line imaging modality. However, its specificity (63.1%) was relatively low, likely due to false positives in patients with post-surgical changes or reactive nodes. These limitations are consistent with a prior study by Simeone et al. (21), who also reported high US sensitivity but reduced specificity in the post-operative setting.

Serum Tg was less sensitive than US but had higher specificity (91.8%). A Tg cut-off of 1.8 ng/ml was associated with an increased risk of recurrence (OR: 13.86). The cut-off of 1.8 ng/mL was derived from ROC analysis within our cohort to maximize sensitivity and specificity; however, it should be acknowledged that this relatively high threshold may inevitably miss low-level recurrences. Despite its strong predictive value, Tg alone was insufficient for ruling out disease, as 22 of 149 patients with Tg <1.8 ng/ml still had proven recurrence. Additionally, 11.6% of patients with undetectable Tg levels (<0.1 ng/ml) developed recurrence, consistent with prior studies by Pacini et al., who reported that even in patients with undetectable Tg, 1.2% had true recurrence (22). This highlights the limitations of Tg in detecting low-volume or structurally occult disease and reinforces the need for concurrent US evaluation.

In our study, Tg demonstrated higher overall diagnostic accuracy than US (88.3% vs. 73.0%), reflecting its strength as a biochemical marker of recurrence. Nonetheless, Tg alone may be insufficient, as 11.6% of patients with undetectable Tg (<0.1 ng/mL) still harboured proven disease. US remains indispensable as the primary structural detector, and when combined with basal Tg, both sensitivity and NPV can reach 100%, while also reducing the need for stimulation testing in nearly 80% of patients (12). This complementary role underscores that Tg provides greater overall accuracy, but US ensures detection of structurally occult locoregional disease.

Our results demonstrated that US findings, such as hypoechoic or heteroechoic thyroid bed lesions with microcalcifications and vascularity, were significantly associated with recurrent thyroid-bed disease. Similarly, heteroechoic nodes with abnormal vascularity and loss of fatty hilum were predictive of nodal metastases. The presence of nodal involvement in both central and lateral neck compartments conferred a higher risk of recurrence, underscoring the importance of systematic nodal evaluation during US surveillance. These findings align with those of Leboulleux et al., highlighting the prognostic value of nodal architecture and vascularity in identifying malignant nodes (23).

An important challenge identified in our study was the evaluation of small cervical nodes. Among 112 patients with nodal lesions ≤6 mm, 30.4% had metastatic disease, demonstrating that small node size does not preclude malignancy, particularly when additional suspicious features such as loss of fatty hilum, abnormal vascularity, or microcalcifications are present. While prior studies (such as one by Kwak et al.) suggested that small nodes with preserved fatty hilum are more likely reactive (24, 25), our findings emphasize that short-axis diameter alone should not be the sole determinant in ruling out malignancy, While FNAB is technically challenging in very small nodes, it may be warranted in nodes ≤6 mm that demonstrate such suspicious sonographic characteristics.

Regarding recurrence timing, 38.5% of recurrences were diagnosed within 6 months post-surgery. Furthermore, 70.6% of recurrences occurred within the first 2 years, aligning with Ryoo et al., who suggested that early recurrences reflect incomplete initial treatment or residual microscopic disease (2). Our findings support US surveillance every 6 months for the first 2 years post-surgery for patients with any of the following: high-risk US features; Tg ≥1.8 ng/mL or rising Tg/TgAb; suspicious/indeterminate US findings; residual thyroid or no RAI; or multi-compartment nodal disease (central + lateral). For low-risk patients with undetectable basal Tg (<0.1 ng/mL) and a negative US at 6–12 months, de-escalate to annual US, then every 12–24 months thereafter. In TgAb-positive patients with negative imaging, perform US every 6–12 months until antibodies decline.

While our study reinforces the central role of US in post-thyroidectomy surveillance, it also highlights challenges of false-positive results, which may provoke patient anxiety, lead to unnecessary biopsies, and contribute to increased healthcare costs (25, 26). These limitations highlight the need for balanced surveillance strategies that minimize overtreatment without compromising early detection. Emerging AI and ML-based tools hold promise to reduce operator dependency and improve the specificity of US interpretation (14). Future prospective, multicentre trials incorporating AI-assisted US workflows (along with liquid biopsy assays), could provide an integrated, non-invasive framework for personalized recurrence monitoring.

Recent literature strongly emphasizes the complementary role of neck US and serum Tg for DTC surveillance, as each modality offsets the other’s limitations. Tg testing serves as an efficient high-sensitivity screen: if Tg remains undetectable (and anti-Tg antibodies are absent), the likelihood of active structural disease is exceedingly low. After total thyroidectomy and RAI, surveillance US can be safely deferred in patients with low or undetectable Tg levels, since virtually no clinically significant recurrences occurred in Tg-negative, antibody-negative patients. In those patients, neck US is more likely to yield false-positive nodules than true cancer. On the other hand, when Tg is elevated or rising, ultrasound plays a crucial role in localizing the recurrence (27–29).

The strengths of this study include its large cohort and detailed ultrasonographic analysis of recurrent disease patterns. However, several limitations should be acknowledged. The retrospective, single-centre design may introduce inherent selection bias and limit generalizability, as only patients with available follow-up and complete biochemical and imaging data were included. US hardware evolved over the 10-year study period, introducing variability in doppler-based vascularity assessment. Despite protocol standardization, neither inter-device calibration nor inter-observer agreement testing was performed which could affect reproducibility. Follow-up intervals were inconsistent, reflecting real-world practice but limiting uniform comparisons. ATA recurrence risk stratification (low, intermediate, high) was not available for all patients due to the retrospective and imaging-focused design of this study, which may limit correlation between baseline risk and recurrence outcomes. Another limitation is that while near-total thyroidectomy was included in the cohort, the average volume of residual thyroid tissue in these cases was not quantified, which could potentially influence Tg interpretation. TSH levels were also not consistently measured at the time of Tg assessment, which may influence Tg interpretation. Tg measurements were derived from a heterogeneous cohort including both RAI-treated and non-RAI patients, which may confound the analysis of Tg’s diagnostic accuracy. We also specifically evaluated the role of US in detecting locoregional recurrence; its utility does not extend to assessing hematogenous dissemination to distant sites.

Despite these limitations, our findings reinforce the pivotal role of US and serum Tg in the surveillance of post-thyroidectomy patients, emphasizing the need for a risk-adapted, multimodal follow-up strategy. Future studies incorporating advanced imaging techniques and molecular markers may further refine recurrence detection and risk stratification, ultimately improving long-term patient outcomes.

Neck US and serum Tg together provide a robust, generalizable framework for detecting recurrence after thyroidectomy. Broader implementation of this combined approach could standardize follow-up care across centres. Future research should focus on refining risk-adapted surveillance algorithms incorporating molecular markers and AI-assisted US to enhance early recurrence detection and optimize long-term management of thyroid cancer patients.