A Comparison of Detection Rates Between Ultrasound and Post‐Therapeutic I‐131 SPECT/CT for Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma

Abstract

Objective

This study compares the diagnostic performance of Ultrasound and post‐therapeutic I‐131 SPECT/CT for detecting cervical lymph node metastases in postoperative patients. We also characterize the sonographic features of metastatic nodes and map their distribution across cervical levels.

Methods

A retrospective, reference‐standard‐led cohort study was conducted. The cohort comprised 172 post‐thyroidectomy PTC patients, from which 266 individual lymph nodes with pathologically proven metastasis were analyzed. The detection rates of US and I‐131 SPECT/CT were compared using McNemar's test. Multivariate logistic regression was used to identify independent sonographic predictors of a node being classified as suspicious on ultrasound.

Results

US demonstrated a significantly higher detection rate for proven metastatic nodes than SPECT/CT (75.0% versus 55.8%, P < .001). Microcalcifications [adjusted odds ratio (aOR) = 1.64, P = .03] and hypoechogenicity (aOR = 1.57, P = .02) were the strongest independent predictors of a node being classified as suspicious on ultrasound. Level VI was the most frequently involved nodal station (55%) and the only independent predictor of metastatic involvement (OR = 2.64, P = .028).

Conclusion

Ultrasound demonstrated superior detection of metastatic lymph nodes, identifying a substantial subset (32.7%) of RAI‐refractory disease missed by SPECT/CT. SPECT/CT remains valuable for confirming RAI avidity and detecting occult functional disease. These findings support a combined imaging strategy according to the metastatic characteristics. The presence of microcalcifications and hypoechogenicity was an important sonographic predictor of malignancy, with Level VI being the most common and predictable site of metastatic involvement.

Abbreviations

ACR

American College of Radiology

aOR

adjusted odds ratio

ATA

American Thyroid Association

CEUS

contrast‐enhanced ultrasound

CI

confidence interval

CLNM

cervical lymph node metastasis

CNB

core needle biopsy

ETA

European Thyroid Association

FDG

fluorodeoxyglucose

FNA

fine‐needle aspiration

FNAC

fine‐needle aspiration cytology

IRB

Institutional Review Board

kV

kilovolt

LID

low‐iodine diet

LN

lymph node

LT4

levothyroxine

mA

milliampere

mCi

millicurie

NCCN

National Comprehensive Cancer Network

OR

odds ratio

PET/CT

positron emission tomography/computed tomography

PTC

papillary thyroid carcinoma

RAI

radioactive iodine

rhTSH

recombinant human thyroid‐stimulating hormone

S/L ratio

short‐axis to long‐axis ratio

SPECT/CT

single photon emission computed tomography/computed tomography

sTg

stimulated thyroglobulin

Tg

thyroglobulin

TgAb

thyroglobulin antibody

THW

thyroid hormone withdrawal

TI‐RADS

Thyroid Imaging Reporting and Data System

TSH

thyroid‐stimulating hormone

US

ultrasound

VIF

variance inflation factor

WBS

whole body scan

Papillary thyroid carcinoma accounts for 80% of thyroid malignancies, with cervical lymph node metastasis (CLNM) occurring in 20–50% of patients, significantly affecting recurrence and survival. ¹ , ² , ³ Accurate detection of metastatic lymph nodes is crucial for guiding surgical and adjuvant treatment decisions, including radioactive iodine ablation. ⁴ , ⁵

Ultrasound is the primary diagnostic tool due to its accessibility and noninvasiveness, offering a detailed evaluation of lymph node morphology. ⁶ However, its accuracy is operator‐dependent and limited for small or deep‐seated nodes. ⁷ , ⁸ Existing studies report inconsistent diagnostic performance for US and PET/CT in CLNMs detection, highlighting the need for further evaluation. ⁹ , ¹⁰ In contrast, I‐131 SPECT/CT combines functional and anatomical imaging, thereby improving the detection of iodide‐avid metastases, particularly after thyroidectomy. ¹¹ , ¹² Post‐therapeutic SPECT/CT, performed concurrently with ablation, provides functional insights. ¹³

While numerous studies have evaluated the diagnostic accuracy of ultrasound and SPECT/CT separately, a critical limitation persists in the literature: the reliance on cohorts with mixed pathology (both benign and malignant nodes), which introduces verification bias and complicates the interpretation of true test performance. ⁶ , ¹¹ , ¹² , ¹³ This is especially relevant for evaluating metastatic disease, where the key clinical question is not just distinguishing benign from malignant, but understanding the detection capabilities for known metastases. Specifically, it remains unclear what proportion of pathologically proven metastatic lymph nodes are detectable by ultrasound versus SPECT/CT. This knowledge gap has direct clinical implications, as understanding which metastases are detected by one modality but missed by the other is crucial for identifying patients with RAI‐refractory disease, which fundamentally alters therapeutic strategy.

While histopathology remains the gold standard, its invasiveness underscores the need for reliable preoperative imaging. ¹⁴ Ultrasound‐guided FNA cytology and Tg washout remain the clinical gold standard for postoperative lymph node assessment in PTC, offering high diagnostic accuracy (sensitivity: 89–94%; specificity: 91–95%) with minimal invasiveness. ⁷ Specific US features (eg, loss of fatty hilum, microcalcifications, hypoechogenicity) may also predict malignancy, but their clinical utility remains debated. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ A study design that starts with a cohort of proven metastases is therefore necessary to definitively establish the detection rates and complementary roles of these imaging modalities.

Therefore, this study employed a reference standard‐led cohort design comprising exclusively patients with cytologically or histologically confirmed lymph node metastases. The exact proportion of metastases found by ultrasound but missed by SPECT/CT, indicating RAI‐refractory disease, remains unquantified in pathology‐confirmed studies. This design eliminates verification bias and aims to directly compare the detection rates of ultrasound and post‐therapeutic I‐131 SPECT/CT, specifically in pathologically proven metastatic lymph nodes, to clarify the complementary roles of these modalities in postoperative surveillance and identify imaging profiles predictive of RAI refractoriness.

Materials and Methods

Study Design and Population

Study Design

This was a retrospective cohort study designed to compare the detection rates of ultrasound (US) and post‐therapeutic I‐131 SPECT/CT for lymph node metastasis in postoperative PTC patients.

The cohort was assembled by first retrospectively identifying 310 patients who had undergone both a post‐therapeutic I‐131 SPECT/CT and a postoperative neck ultrasound within 2 months between July 2023 and December 2024. From this initial imaging database, we then selected only those patients who had at least one cervical lymph node with cytologically or histologically proven metastasis.

This design ensures a direct “head‐to‐head” comparison of both imaging modalities for the same patient, assessing their performance against a 100% definitive reference standard.

Participant Selection

Consecutive PTC patients who met the following criteria were retrospectively included:

Inclusion criteria:

1. Total thyroidectomy for PTC.

2. Presence of one or more cervical lymph nodes with metastasis confirmed by:

   • ○
     Ultrasound‐guided fine‐needle aspiration cytology and/or thyroglobulin (Tg) washout, and/or
   • ○
     Core needle biopsy, and/or
   • ○
     Surgical histopathology.

3. Undergone post‐therapeutic I‐131 SPECT/CT as per standard clinical management.

4. Comprised postoperative ultrasound and I‐131 SPECT/CT within 2 months of the interval.

Exclusion criteria:

1. Incomplete clinical, imaging, or pathological data for the identified lymph nodes (n = 93).

2. Prior radiation therapy to the neck (n = 11).

3. Patients in whom the final pathological confirmation of the lymph node in question was indeterminate (Bethesda Category III/IV/V) or benign, despite initial suspicion (n = 34).

The initial imaging database search identified 310 patients who underwent both required imaging studies. After applying the above exclusion criteria, 138 patients were excluded (93 + 11 + 34), resulting in the final cohort of 172 patients with definitively proven metastatic lymph nodes. From this cohort, 266 individual pathologically confirmed metastatic lymph nodes were identified and formed the unit of analysis.

Initial surgical management for all patients included total thyroidectomy. The extent of lymph node dissection was guided by preoperative imaging and intraoperative assessment. Most patients (160/172, 93.0%) underwent concomitant central compartment neck dissection (Level VI) only. A minority (12/172, 7.0%) underwent additional therapeutic lateral neck dissection (Levels II–V) due to preoperatively or intraoperatively confirmed lateral nodal disease. All metastatic lymph nodes analyzed in this study were identified during postoperative surveillance and confirmed via cytological/histopathological methods as detailed in the Inclusion Criteria.

Imaging and Correlation

For each patient, the following data were collected and analyzed:

• SPECT/CT Results: Classified as demonstrating iodide avidity (“SPECT/CT‐positive”) or not (“SPECT/CT‐negative”) in the proven metastatic lymph nodes.

• Ultrasound Features: The sonographic features and original radiology reports for the proven metastatic nodes were retrospectively reviewed and classified as “suspicious” or “not suspicious” based on predefined criteria to determine the ultrasound detection rate (Table 1).

Table 1.

Ultrasound Diagnostic Criteria for Lymph Node Metastasis in PTC

Feature	Suspicious of Metastasis	Benign/Indeterminate	Manuscript References
Size	Short‐axis diameter ≥5 mm	<5 mm	21
Shape	Spherical (S/L ratio ≥0.7)	Oval (S/L ratio <0.7)	15 , 21
Echogenicity	Hypoechoic (darker than the strap muscle)	Iso‐/hyperechoic or mild hypoechogenicity	27
Microcalcifications	Punctate hyperechoic foci (≤1 mm) with or without shadowing	Macrocalcifications (>2 mm) or absent	24 , 25 , 26
Cystic Changes	Irregular cystic areas ≥3 mm	Smooth cysts or <3 mm	28 , 29 , 30 , 31
Hilum	Complete loss	Preserved or partial loss	19 , 26
Vascularity	Peripheral/disorganized flow	Central hilar flow	15 , 34

S/L ratio, short axis to long axis ratio.

Multidisciplinary Correlation: This structured correlation protocol was applied retrospectively during the multidisciplinary consensus meetings to ensure accurate matching of lymph nodes between the previously acquired US and SPECT/CT images.

The post‐therapeutic I‐131 SPECT/CT examinations analyzed in this study were performed as part of standard clinical management, not as a research intervention, in accordance with established guidelines. ¹⁹ Ethical approval for the retrospective analysis of this data was granted by the Institutional Review Board (Approval No: bc2020190) of Tianjin Medical University Cancer Institute and Hospital, which waived the requirement for informed consent. The study workflow is detailed in Figure 1.

Figure 1.

Flowchart showing the workflow of the study.

Pre‐Therapy Preparation Protocol

All patients underwent:

1. Thyroid Hormone Withdrawal (THW): Patients were instructed to discontinue levothyroxine (LT4) for 2–4 weeks before I‐131 administration. While recombinant human TSH (rhTSH) is an alternative standard of care to avoid hypothyroid symptoms, THW was employed in this cohort as it represents a common clinical practice in our institution and in many regions where rhTSH may be less accessible or for patients with specific clinical indications (eg, higher risk disease where maximal stimulation is traditionally sought). Serum TSH >30 mIU/mL (median: 98.95 mIU/mL; range: 0.01–249.00) confirmed adequate stimulation.

2. Low‐Iodine Diet (LID): Two to four weeks pre‐SPECT/CT to enhance radioiodine uptake. We acknowledge that the use of THW rather than rhTSH may influence the absolute radiotracer uptake in metastatic foci. However, our internal validation showed no significant difference in TSH levels between SPECT/CT‐positive and SPECT/CT‐negative patients (P = .797), suggesting that within this THW‐prepared cohort, the variation in TSH did not systematically bias the detection rate comparison with ultrasound.

Diagnostic Modalities

Ultrasound: Performed using a Philips EPIQ5 system (7–15 MHz transducer). Both images and original reports were reviewed by two board‐certified radiologists (>10 years' thyroid imaging experience). Discrepancies between initial reports and re‐evaluations were resolved by consensus. Evaluated features included size (with a short‐axis diameter ≥5 mm used as one criterion for suspicion), shape (S/L ratio), echogenicity, microcalcifications, cystic changes, and loss of hilum. Ultrasound features were classified using predefined criteria, following the American Thyroid Association (ATA) guidelines. ¹⁹

Standardization of Ultrasound Assessment

It is important to note that the American College of Radiology (ACR) TI‐RADS guidelines are designed for risk stratification of thyroid nodules, not for the assessment of cervical lymph nodes. ²⁰ Therefore, our ultrasound analysis utilized the well‐established feature‐based criteria for lymph node malignancy endorsed by the ATA guidelines, which are the standard for postoperative PTC surveillance. ¹⁹ , ²¹ This approach ensured our methods were aligned with the specific clinical context of metastatic lymph node detection.

I‐131 SPECT/CT

Thyroglobulin and Anti‐Tg Antibodies (TgAb): Stimulated Tg (sTg) and TgAb were measured during pre‐ablation assessment. Thirty percent of patients (53/172) had anti‐thyroglobulin antibodies (TgAb ≥20 IU/mL), which can artificially lower measured thyroglobulin levels. These cases were included but interpreted with caution. At our institution, post‐therapeutic I‐131 SPECT/CT is not performed routinely for all thyroidectomy patients. Its use follows the ATA guidelines, and it is indicated for patients with ATA intermediate‐to‐high risk disease who undergo adjuvant radioactive iodine (RAI) therapy following total thyroidectomy. The SPECT/CT was acquired 2–5 days after therapeutic I‐131 ablation (100–200 mCi), as part of the post therapeutic whole body scan protocol, to maximize lesion‐to‐background contrast, using a Symbia Intevo 16 system (Siemens Healthineers) with low‐dose CT (120 kV, 200–220 mA) for attenuation correction and anatomic localization, with iterative reconstruction to minimize noise. SPECT images were obtained at 20 seconds per bed position (128 × 128 matrix). This integrated approach provides functional‐metabolic mapping and anatomical localization without administering an additional diagnostic radioiodine dose, aligning with ATA recommendations for selective use in higher‐risk patients. No intravenous contrast was administered to prevent interference with iodide avidity assessment. Raw images and reconstructed datasets were independently interpreted by two nuclear medicine physicians (≥8 years' experience). Original clinical reports were referred only for anatomic correlation during multidisciplinary reviews. Figure 2 shows a patient's SPECT (A), CT (B), and SPECT/CT fusion images (C), demonstrating how multimodal imaging provides a comprehensive evaluation of both functional and anatomical characteristics of lesions.

Figure 2.

Multimodal imaging of a patient's SPECT (A), CT (B), and SPECT/CT fusion images (C) demonstrates a comprehensive evaluation of both functional and anatomical characteristics of lesions. Focal iodide avidity is seen on the SPECT image (A), which localizes to a lymph node on the CT (B), as confirmed on the fused SPECT/CT image (C).

Correlation of Lymph Nodes Across Imaging Modalities and Multidisciplinary Consensus Protocol

To ensure the accurate and reproducible correlation of individual lymph nodes (LNs) between US and I‐131 SPECT/CT, a structured, multidisciplinary consensus protocol was rigorously followed.

Initial Independent Review

Two board‐certified radiologists, blinded to the SPECT/CT results, independently reviewed the US images and reports. They documented the specific anatomical location and sonographic features of each lymph node.

Two nuclear medicine physicians, blinded to the US results, independently reviewed the SPECT/CT studies and documented the location of any foci of abnormal radiotracer uptake corresponding to lymph nodes.

Structured Correlation Criteria

Lymph nodes were matched across modalities based on a hierarchical assessment of the following fixed anatomical landmarks and standardized neck levels:

Level IA (Submental Group): Lymph nodes situated within the triangle formed by the anterior bellies of the digastric muscles and the hyoid bone.

Level IB (Submandibular Group): This group of lymph nodes is located in an area bounded by the mandible and the digastric and stylohyoid muscles. For radiographic distinction, the posterior border of this region (Level IB) is often defined by a vertical line traced along the back of the submandibular gland, which separates it from Level IIA.

Levels IIA & IIB (Upper Jugular Group): These lymph nodes lie along the upper internal jugular vein and spinal accessory nerve from the skull base to the hyoid bone. The compartment's anterior limit is the lateral border of the sternohyoid/stylohyoid muscles (posterior submandibular gland radiographically), and its posterior limit is the back of the sternocleidomastoid. The spinal accessory nerve itself divides the area into an anteromedial (IIA) and a posterolateral (IIB) sublevel.

Level III (Middle Jugular Group): These nodes lie alongside the middle internal jugular vein from the hyoid bone to the cricoid cartilage. Their compartment is bordered medially by the sternohyoid muscle and laterally by the posterior edge of the sternocleidomastoid muscle.

Level IV (Lower Jugular Group): This nodal compartment contains lymph nodes adjacent to the lower third of the internal jugular vein. Its superior and inferior limits are the cricoid cartilage and the clavicle, respectively. The medial boundary is the lateral border of the sternohyoid muscle, and the lateral boundary is the posterior edge of the sternocleidomastoid muscle.

Levels VA and VB (Posterior Triangle Group): Primarily containing lymph nodes of the lower spinal accessory nerve and transverse cervical artery—plus the supraclavicular nodes—this group occupies the posterior triangle. Its superior, inferior, anterior/medial, and posterior/lateral boundaries are the sternocleidomastoid‐trapezius apex, the clavicle, the posterior sternocleidomastoid border, and the anterior trapezius border, respectively. Sublevels VA and VB are divided by a horizontal line at the inferior cricoid cartilage arch.

Level VI [Anterior (Central) Compartment Group]: Its anatomical limits are defined by the hyoid bone superiorly, the suprasternal notch inferiorly, and the common carotid arteries laterally.

Multidisciplinary Consensus Meeting

A final consensus meeting was held with the radiologists and nuclear medicine physicians. All initially identified lymph nodes from both modalities were reviewed side‐by‐side. A node was considered a definitive match only if there was unanimous agreement on its identity across US and SPECT/CT based on the structured criteria above. Any discrepancies in initial localization were discussed and resolved by consensus, ensuring the highest confidence in the one‐to‐one correlation used for the subsequent diagnostic performance analysis. Post‐surgery imaging and FNAC/Tg washout, core needle biopsy, and histopathology results (where available) helped confirm metastasis. Since post‐op imaging directly located any new or remaining nodes, initial surgery pathology reports were not used for level assignment.

Statistical Analysis

The detection rate of US and SPECT/CT was evaluated and compared using McNemar's test, as it is designed for paired nominal data (the same lymph nodes assessed by two modalities). Univariate and multivariate logistic regression analyses were performed to identify associations between specific ultrasound features and the outcome of a node being US‐positive. The multivariate model was adjusted for nodal size and level. The assumptions of the logistic regression model were checked, including the absence of multicollinearity, which was confirmed by all Variance Inflation Factor (VIF) values being below 5. Nonparametric tests (Spearman's rho for correlation, Mann–Whitney U for group comparisons) were used to assess relationships between TSH, Tg, and imaging results, as the data were non‐normally distributed (Kolmogorov–Smirnov tests, all P < .001). All data were managed in a secure electronic database. Statistical analysis was performed with SPSS version 27.0 (SPSS Inc., Chicago, IL, USA), and a P‐value <.05 was considered statistically significant.

Results

Study Population and Baseline Characteristics

The study cohort was derived from 310 PTC patients screened for imaging availability, from which 172 patients who had undergone both postoperative US and I‐131 SPECT/CT within 2 months were retrospectively included. The cohort had a mean age of 44.8 ± 14.8 years and consisted of 104 (60.5%) females and 68 (39.5%) males. From these 172 patients, a total of 266 individual cervical lymph nodes with pathologically proven metastasis were identified and formed the unit of analysis for this study.

The pathological confirmation of metastasis was obtained via the following methods:

Ultrasound‐guided fine‐needle aspiration cytology and/or thyroglobulin (Tg) washout: 192 nodes (72.2%), core needle biopsy: 20 nodes (7.5%), and surgical histopathology (from subsequent neck dissection): 54 nodes (20.3%). The mean number of metastatic lymph nodes per patient was 1.55 (range: 1–6).

Diagnostic Detection Rates and Imaging Profiles

The detection rates for each imaging modality are summarized in Table 2. Ultrasound identified 75.0% (199/266) of the proven metastatic nodes, significantly outperforming I‐131 SPECT/CT, which showed iodide avidity in 55.8% (148/266) of nodes (P < .001).

Table 2.

Detection Rates of Ultrasound and I‐131 SPECT/CT in 266 Pathologically Proven Metastatic Lymph Nodes

Modality	Detection Rate (Positivity Rate)	95% Confidence Interval
Ultrasound	75.0% (199/266)	69.4–80.0
I‐131 SPECT/CT	55.8% (148/266)	49.6–61.8
P‐value	<.001 (McNemar's test)

The interplay between US and SPECT/CT findings is detailed in Table 3. The largest subgroup of metastases (32.7%, 87/266) was detected by ultrasound but showed no iodide avidity on SPECT/CT. In contrast, a smaller but substantial subgroup (13.5%, 36/266) was SPECT/CT‐positive but did not exhibit suspicious features on ultrasound. The remaining 31 nodes (11.7%) were not detected by either modality.

Table 3.

Imaging Profiles of 266 Metastatic Lymph Nodes

Imaging Profile	Number of LNs	Percentage of Cohort
US+/SPECT/CT+	112	42.1%
US+/SPECT/CT−	87	32.7%
US−/SPECT/CT+	36	13.5%
US−/SPECT/CT−	31	11.7%
Total	266	100%

A representative case of a metastatic lymph node positive on both modalities is shown in Figure 3, which depicts a 45‐year‐old female with PTC who underwent a postoperative ultrasound and SPECT/CT. Fused I‐131 SPECT/CT images showed focal radiotracer uptake in a right cervical lymph node, indicating iodide‐avid metastatic disease. Targeted ultrasound images at the corresponding level reveal the node to be round, hypoechoic, and containing microcalcifications, confirming its morphologically suspicious appearance. Ultrasound‐guided fine‐needle aspiration confirmed metastatic papillary thyroid carcinoma.

Figure 3.

Representative case of a metastatic lymph node detected by both ultrasound and I‐131 SPECT/CT. Ultrasound (A) revealed a hypoechoic (calipers), round lymph node with microcalcifications, highly suspicious for metastasis. SPECT/CT (B) showed focal iodide uptake (intersection point) in the same lymph node.

Ultrasound Features and Metastasis

Analysis was performed to identify which sonographic features were associated with a metastatic node being successfully identified as “suspicious” on ultrasound.

Univariate Analysis

The association between specific ultrasound features and pathology‐positive metastasis is shown in Table 4. Microcalcifications, defined as punctate hyperechoic foci <1 mm, were significantly associated with ultrasound detection (OR = 1.67, 95% CI: 1.05–2.65, P = .031). They increased the odds of a metastatic node being detected by ultrasound by 66.7% (95% CI: 1.05–2.65), followed by hypoechogenicity (OR = 1.52, 95% CI: 1.13–2.12) and loss of hilum (OR = 1.51, 95% CI: 1.12–2.11). Cystic changes were significant but had wider confidence intervals (OR = 1.30, 95% CI: 1.09–2.10), indicating variable diagnostic utility. Round shape (OR = 0.30, P > .05) was not associated with metastasis.

Table 4.

Univariate Analysis of Ultrasound Features Associated with Metastasis

Feature	OR	95% CI	P‐Value
Microcalcifications	1.67	1.05–2.65	.031
Echogenicity	1.52	1.13–2.12	.024
Loss of hilum	1.51	1.12–2.11	.041
Cystic Changes	1.30	1.09–2.10	.012
Round Shape	0.30	0.28–10.75	>.05

Analysis performed on 266 pathologically confirmed metastatic lymph nodes. All models adjusted for nodal size.

Multivariate Analysis

Multivariate logistic regression analysis was performed to identify independent predictors of a node being classified as suspicious on ultrasound (Table 5). After adjustment for size and confounders, microcalcifications remained the strongest independent predictor (aOR = 1.64, 95% CI: 1.02–2.63), closely followed by hypoechogenicity (aOR = 1.57) and loss of hilum (aOR = 1.45) (all P < .05). Cystic changes, while significant (aOR = 1.43, P = .01), may reflect necrosis in advanced metastases.

Table 5.

Multivariate Logistic Regression of Independent Predictors of Metastasis

Feature	aOR	95% CI	P‐Value
Microcalcifications	1.64	1.02–2.63	.03
Hypoechogenicity	1.57	1.21–2.52	.02
Loss of hilum	1.45	1.37–2.69	.04
Cystic changes	1.43	1.04–1.96	.01
Round Shape	0.42	0.32–6.44	>.05

Analysis performed on 266 pathologically confirmed metastatic lymph nodes. All models adjusted for nodal size.

The model explained only 3.7% of the variance (Nagelkerke R ²) and showed marginal improvement over the null model (P = .059). The multicollinearity test revealed no issues with multicollinearity.

Subgroup Analysis: Metastatic Lymph Nodes <5 mm

Among the 266 metastatic lymph nodes, 38 (14.3%) measured <5 mm in short‐axis diameter. Of these, 21 (55.3%) exhibited suspicious ultrasound features (predominantly microcalcifications or loss of fatty hilum) and were correctly classified as malignant. The remaining 17 small nodes (44.7%) lacked definitive suspicious features and were not sonographically flagged. This highlights that while size is a common criterion, microcalcifications and other features in subcentimeter nodes can be critical indicators of malignancy.

Subgroup Analysis: Metastatic Distribution

A multidisciplinary review of ultrasound and SPECT/CT imaging identified Level VI as the most frequently involved nodal station (55%), followed by Level IV (42%), Level III (40%), and Level II (31%). Levels I and V were rarely involved (0% and 5%, respectively). This distribution was mapped during the multidisciplinary consensus review of postoperative imaging, providing level‐specific assessment of metastatic involvement in the study cohort. The number of metastatic lymph nodes per patient averaged 1.55 (range: 1–6 nodes), based on 266 pathologically confirmed nodes identified across 172 patients. Chi‐square tests for association (Table 6) showed that Levels II, III, IV, VI, and “Other LN” were strongly associated with neck lymph node metastasis from papillary thyroid carcinoma (P < .05).

Table 6.

Chi‐Square Tests for Association

LN Level	P‐Value	Interpretation
II	.013	Significant association with metastasis
III	<.001	Highly significant
IV	<.001	Highly significant
VI	<.001	Highly significant
Other LN	.002	Significant

Logistic Regression Results of LN Levels

After adjusting for other LN levels, only Level VI was independently associated with metastatic involvement (OR = 2.64, 95% CI: 1.11–6.28, P = .028). Levels II/III/IV and other lymph nodes showed significance in unadjusted analyses by chi‐square but not in regression, suggesting their effects may overlap with Level VI. Level VI was the only nodal level that remained statistically significant in multivariate analysis (OR = 2.64, P = .028), while Level III showed a nonsignificant trend (OR = 2.73, P = .098). This suggests that Level VI metastasis is the primary driver of nodal spread, with lateral node involvement (eg, Level III) often co‐occurring but not independently predictive (Table 7).

Table 7.

Logistic Regression Results

Predictor	(OR)	P‐Value	95% CI	Interpretation
Level VI LN	2.639	.028	[1.11, 6.28]	The significant predictor with 2.6 times higher odds of involvement
Level III LN	2.730	.098	[0.83, 8.98]	Trend (not significant)
Level II LN	0.853	.786	[0.27, 2.69]	Not significant
Level IV LN	0.396	.215	[0.09, 1.71]	Not significant
Level V LN	1.05	.961	[0.09, 11.26]	Not significant
Other LN	1.829	.242	[0.67, 4.99]	Not significant

Level‐Specific Ultrasound Features

The level‐specific analysis was based on anatomical localization from the retrospective image review consensus. Metastatic Level VI LNs predominantly exhibited hypoechogenicity (78%) and hilum loss (65%), whereas lateral LNs (Levels II‐IV) more frequently showed microcalcifications (60.7%) and cystic changes (42%). Level VI's independent predictive value (OR = 2.64) persisted despite fewer microcalcifications (28.6%), aligning with its role as the primary drainage site for early metastasis. ²²

Among metastatic nodes with microcalcifications (n = 112), 42% (47/112) were positive on both US and SPECT/CT, while the remainder showed discordant imaging profiles.

These findings support level‐specific diagnostic criteria, with Level VI assessment prioritizing hypoechogenicity and hilum loss, while lateral level evaluation emphasizes microcalcifications.

Subgroup Analysis: Microcalcifications and Iodide Avidity

Of the 112 metastatic LNs with microcalcifications, 60.7% (68/112) exhibited focal iodide avidity on SPECT/CT, while 39.3% (44/112) were SPECT/CT‐negative. This indicates that while microcalcifications are a strong morphological marker, they do not universally predict functional iodide avidity.

Thyroglobulin and Antibody Findings

Stimulated thyroglobulin levels varied widely (median: 10.9 ng/mL, range: 0.04–194.0) without correlation to TSH (Spearman's ρ = −0.024, P = .751). Anti‐thyroglobulin antibodies (≥20 IU/mL) were detected in 30.8% of patients (n = 53/172), potentially affecting Tg assay accuracy in these cases.

Discussion

Principal Findings and Interpretation

This study, employing a reference standard‐led cohort of 266 pathologically proven metastatic lymph nodes, provides a clear quantification of the detection capabilities of ultrasound and I‐131 SPECT/CT in postoperative PTC. Our key finding is that ultrasound demonstrated a significantly higher overall detection rate than SPECT/CT (75.0% versus 55.8%). This underscores the primary role of ultrasound in morphological surveillance. The superior detection rate of ultrasound aligns with its ability to identify structural abnormalities irrespective of their functional status, making it indispensable for detecting RAI‐refractory disease. ⁶ , ¹⁹

The detailed imaging profiles we established (Table 3) offer critical clinical insights. The largest subgroup of metastases (32.7%, US+/SPECT/CT−) represents RAI‐refractory disease that would be missed by functional imaging alone. Conversely, the 13.5% of metastases that were US−/SPECT/CT+ represent functionally active, morphologically occult disease, highlighting SPECT/CT's unique value in specific contexts. This clear delineation of discordant groups validates the complementary roles of both modalities. ²² , ²³

Ultrasound Features in a Proven Metastasis Cohort

In this cohort of known metastases, microcalcifications and hypoechogenicity emerged as the strongest independent predictors of a node being sonographically suspicious. This refines the conventional understanding of ultrasound criteria, suggesting that in a post‐operative setting where all nodes are malignant, these features are most reliably associated with a positive ultrasound interpretation.

The strong association of microcalcifications (aOR = 1.64) with ultrasound detection is pathologically grounded, as they represent psammoma bodies, a hallmark of PTC. ²⁴ Our finding that microcalcifications were more frequent in lateral neck nodes (60.7%) than in Level VI (28.6%) suggests they may be a marker of more advanced or established metastatic deposits. ⁹ , ²¹ This is consistent with studies by Rosario et al ²⁵ and Kravchenko et al, ²⁶ which reported similar associations between microcalcifications and metastatic involvement. Hypoechoic echogenicity is a well‐established feature of metastatic lymph nodes, reflecting the replacement of normal lymphoid tissue by tumor cells. ²⁷ Also, loss of fatty hilum (aOR = 1.45, 95% CI: 1.37–2.69) is consistent with studies by Kravchenko et al, ²⁶ supporting its continued relevance in post‐thyroidectomy surveillance. For cystic changes, our results (aOR = 1.43, 95% CI: 1.84–2.43) were consistent with their known high specificity (85–100%) and lower sensitivity (70%). ²⁸ , ²⁹ , ³⁰ This aligns with pathological evidence that cystic degeneration typically occurs in larger metastatic deposits, ²⁷ , ³¹ which may explain why this feature maintained significance despite our size adjustment.

The loss of predictive value for a round shape (aOR = 0.42, P > .05) in our postoperative cohort is a novel finding, contrasting with the results of preoperative studies. ⁵ While shape remains a valuable criterion in preoperative nodal assessment, ¹⁵ , ²¹ our findings imply that postoperative ultrasound evaluation might benefit from increased attention to echotextural features, particularly in a previously dissected neck where postsurgical anatomical changes can alter nodal morphology.

Anatomical Distribution and Clinical Implications

The predominance of Level VI involvement (55%) and its status as the only independent predictor of metastatic involvement (OR = 2.64) reinforce its role as the primary drainage basin and a critical sentinel station. ³² , ³³ , ³⁴ These findings mandate that Level VI receive meticulous attention in any postoperative ultrasound examination. The loss of independent significance for lateral levels after adjusting for Level VI involvement suggests that lateral disease often occurs in the context of central compartment spread, supporting the “gateway” hypothesis for metastatic progression in PTC. ³⁵ , ³⁶

These findings suggest that postoperative ultrasound criteria may weigh microcalcifications, hypoechogenicity, and loss of the fatty hilum more heavily than shape, in contrast to preoperative TIRADS systems. ²⁰ , ³⁷ A modified scoring system for surveillance settings may improve accuracy. While microcalcifications were strongly associated with metastasis (aOR = 1.64), their distribution varied by nodal level. Lateral lymph nodes (Levels II–IV) had a higher rate of microcalcifications (60.7%) than Level VI nodes (28.6%). This aligns with pathological findings that psammoma bodies are more common in larger, advanced metastatic deposits. ⁹ , ²⁰ Level VI's predictive dominance (OR = 2.64) likely reflects its anatomical role as the initial metastatic site, where architectural distortion (eg, loss of fatty hilum, hypoechogenicity) precedes calcification. Our findings suggest that during postoperative ultrasound surveillance, characteristic sonographic patterns may vary by nodal level: metastatic nodes in Level VI frequently exhibited hypoechogenicity and loss of fatty hilum, whereas lateral compartment nodes (Levels II–IV) more commonly displayed microcalcifications. This level‐specific distribution of sonographic features (Table 8) may help inform a more nuanced, anatomy‐aware approach to nodal evaluation in patients with PTC.

Table 8.

Lymph Node Level‐Specific Features

Level	Key Features	Frequency in Metastatic LNs	Literature Support
VI	Hypoechogenicity, hilum loss	55%	26 , 27
II–IV	Microcalcifications, cystic changes	42–60%	24 , 25 , 26 , 28 , 29 , 30 , 31
V	Extranodal extension	5%	30

The detection of these features also differed by modality (Table 9). SPECT/CT's low‐dose CT component (120 kV, 200–220 mA) demonstrated lower sensitivity for small calcifications (<1–2 mm) compared to ultrasound, primarily due to spatial resolution limitations. However, when coarse calcifications coexisted with focal iodide avidity (observed in 60.7% of calcified nodes), specificity approached 100%. This suggests the calcification's primary diagnostic value lies in risk stratification when detected by ultrasound and in multimodal confirmation when paired with SPECT/CT avidity.

Table 9.

Comparative Table of Modality Performance

Feature	Ultrasound	SPECT/CT	Clinical Implication
Microcalcifications	High sensitivity	Low sensitivity (size‐dependent)	US for screening, SPECT/CT for confirmation
Iodide avidity	NA	High specificity	Gold standard for metabolic activity
Anatomic localization	Moderate	Excellent	SPECT/CT is superior for deep nodes

While microcalcifications, hypoechogenicity, and hilum loss were significant, our multivariate model explained only 3.7% of variance (Nagelkerke R ²), suggesting the need to incorporate additional parameters like vascular patterns and molecular markers in future studies to improve predictive accuracy. ³⁸ , ³⁹

The anatomical distribution of metastatic nodes revealed clinically important patterns (Table 10). Level VI emerged as both the most frequently involved station (55%) and the only independent predictor of metastasis (OR = 2.64, P = .028). ⁴⁰ The predominance of Level VI involvement, as identified by multidisciplinary imaging review, aligns with its anatomical role as the primary drainage site for metastasis. This finding was independently derived from postoperative ultrasound and SPECT/CT, reinforcing the reliability of imaging‐based level assignments in postoperative settings. While Level III, though significant in univariate testing (P < .001), lost statistical independence when adjusted for Level VI (OR = 2.73, P = .098). This aligns with the gateway hypothesis proposed by Zheng et al ³⁰ wherein central compartment (Level VI) involvement precedes lateral spread. Some studies, for example, Nie et al, ⁴¹ note lateral skip metastases in 5–15% of cases, though our cohort showed less. This discrepancy may reflect differences in surgical extent or tumor biology. Clinically, this underscores the importance of prioritizing Level VI assessment in imaging and surgical planning, as isolated lateral node enlargement (eg, Level III) without Level VI metastasis is rare in our cohort (4%). The trend toward significance for Level III (P = .098) suggests that lateral nodes may still warrant scrutiny when Level VI is positive, but their predictive value is dependent on central compartment status. This strong association underscores the critical role of central compartment evaluation in diagnostic protocols, ³³ while suggesting that routine evaluation of Levels I and V may be of limited value in similar patient populations. The loss of significance for Levels III and IV in multivariate analysis suggests their involvement may often reflect concurrent Level VI disease rather than independent metastatic spread. ³³ , ³⁴ , ³⁵ , ³⁶ These findings raise important questions about the biological mechanisms underlying the preferential involvement of Level VI nodes. We hypothesize that anatomical factors such as primary lymphatic drainage patterns, as well as potential molecular tropisms, may contribute to this observed distribution. Future studies incorporating detailed molecular characterization of metastatic nodes across different levels could provide valuable insights into these mechanisms.

Table 10.

Comparison of Univariate and Multivariate Associations Across Nodal Levels

LN Level	Univariate (Chi‐Square, P)	Multivariate Regression (OR, P)	Interpretation
VI	<.001	OR = 2.64, P = .028	Only an independent predictor
III	<.001	OR = 2.73, P = .098	Trend, not independent
II	.013	OR = 0.85, P = .786	Confounded by Level VI

Clinical Integration and Modality Selection

The high rate of US+/SPECT/CT− metastases (32.7%) has direct therapeutic implications. These nodes represent diseases that are unlikely to respond to RAI, guiding clinicians toward alternative management options, such as surgery or thermal ablation. ⁴² , ⁴³ Our data strongly support the ATA guidelines, which position ultrasound as the cornerstone of postoperative surveillance. ¹⁹ They also clarify the use of SPECT/CT: confirming RAI avidity when treatment is planned and detecting occult disease in high‐risk patients with negative ultrasound but rising Tg levels.

The subgroup of metastatic lymph nodes that were undetected by both ultrasound and SPECT/CT (11.7%, 31/266) represents a clinically challenging entity. While these nodes lack both suspicious sonographic features and iodide avidity, they are not definitively classifiable as radioiodine‐refractory in the traditional sense, because the absence of I‐131 uptake on SPECT/CT does not necessarily imply biochemical refractoriness or dedifferentiation. Instead, these may represent:

1. Very small or early metastatic deposits below the spatial resolution threshold of both modalities, as spatial resolution limits for ultrasound and SPECT/CT are well‐documented in detecting sub‐centimeter pathology. ¹⁶ , ¹⁹ , ²⁰

2. Nodes with diffuse infiltration that preserve nodal architecture, lacking typical malignant ultrasound features such as round shape, hypoechogenicity, or loss of fatty hilum. ²¹ , ²²

3. Metastases with low‐level iodide uptake below the functional detection limit of SPECT/CT, especially in the setting of competing physiological uptake or suboptimal imaging timing, are a known limitation of post‐therapeutic scintigraphy. ¹³ , ²³

In clinical practice, these nodes were identified only through histopathological confirmation prompted by other indicators (eg, elevated stimulated thyroglobulin, clinical suspicion). Thus, while they are imaging‐occult, their biological behavior regarding RAI responsiveness remains uncertain. This underscores the complementary value of biochemical markers and the potential role of more sensitive functional imaging (eg, PET/CT) or molecular testing in such cases. ³⁸

The SPECT/CT, while less sensitive, remains a vital adjunct for two specific purposes: (1) to confirm RAI avidity in suspicious nodes when RAI therapy is being considered, and (2) to detect occult, functionally active disease in high‐risk patients with negative or equivocal ultrasounds but rising Tg levels. This integrated approach optimizes metastatic lymph node detection and characterization.

Standardized reporting criteria, such as those proposed by the ATA and the European Thyroid Association (ETA), could be considered to ensure consistent interpretation of ultrasound findings. ¹⁹ , ⁴⁴ The discordance highlights the modalities' complementary roles: US excels in morphological detection (especially microcalcifications), while SPECT/CT identifies functional avidity in morphologically ambiguous nodes. ⁴⁵ This supports our conclusion that a combined approach optimizes the detection of metastatic LN.

Strengths and Limitations

The methodological approach of this study, a direct comparison of established imaging modalities against a pathological reference standard, is well‐established. Its application in this specific clinical context provides novel and critical insights. The primary strength and innovation of this work lie not in the introduction of new technology, but in the definitive clarification it provides using a rigorous, reference‐standard‐led cohort design. By exclusively analyzing pathologically proven metastatic nodes, this study eliminates the verification bias common in studies with mixed pathology. It offers an unambiguous quantification of the true detection rates and complementary roles of US and SPECT/CT. This foundational evidence is crucial for optimizing postoperative surveillance protocols. Furthermore, while our cohort of 172 patients and 266 lymph nodes was sufficiently powered for our primary analyses and represents a substantial investigation, we acknowledge that expanding the sample size in future multi‐center studies would be valuable. A larger, more diverse population would strengthen the generalizability of our findings, allow for more robust subgroup analyses (eg, based on specific PTC variants or serum Tg levels), and provide greater statistical power to investigate rarer sonographic features or metastatic patterns in less common nodal levels.

A limitation is the retrospective design, which precluded the standardized assessment of all ultrasound features, such as vascularity, in all cases. ¹⁵ , ³⁴ The single‐center nature may affect generalizability, though our standardized protocols enhance reproducibility. Also, our cohort, by design, represents a high‐risk population selected for RAI therapy; therefore, the absolute detection rates may not be generalizable to low‐risk patient groups. Furthermore, while our model for ultrasound features was significant, the low explained variance (Nagelkerke R ² = 3.7%) suggests other unmeasured factors influence sonographic detection, warranting future investigation.

Notably, our study utilized thyroid hormone withdrawal (THW) for TSH stimulation. While recombinant human TSH (rhTSH) is a widely accepted alternative that improves patient quality of life, THW remains a valid and commonly used method, particularly in settings where rhTSH is unavailable or for specific high‐risk patients. The comparable TSH levels achieved across our cohort (median 98.95 mIU/mL, with 95.3% achieving >30 mIU/mL) support adequate and uniform stimulation. Future studies comparing detection rates in rhTSH‐prepared cohorts would be valuable to assess the generalizability of our findings across different preparation strategies. While Tg levels were routinely measured, they showed no significant association with imaging results, likely due to confounding factors like TgAb interference (30.8% of cases) and residual thyroid tissue variability. This aligns with prior studies suggesting Tg's limited standalone value in localized nodal assessment. ⁴⁶ Interestingly, the lack of TSH‐Tg correlation (ρ = −0.024, P = .751) in our uniformly prepared cohort suggests metastatic nodes may exhibit TSH‐independent Tg secretion, a finding requiring molecular investigation. ³⁸

We acknowledge the potential for diagnostic ambiguity when cytology and Tg washout results are discordant, a known challenge in clinical practice. In our center, to mitigate this and ensure the highest confidence in our reference standard, we employed a sequential diagnostic protocol for such cases. This included repeat Tg washout testing and, where clinically indicated, progression to core needle biopsy or surgical excision for definitive histopathological confirmation. While this rigorous approach strengthens the validity of our cohort's classification, it is important to note that the reference standard in thyroid cancer surveillance is a composite clinical benchmark rather than a single perfect test.

Our ultrasound assessment incorporated a short‐axis diameter ≥5 mm as one criterion for suspicion, consistent with ATA guidelines. ¹⁹ However, 38 metastatic nodes (14.3%) in our cohort were <5 mm, of which over half displayed suspicious features such as microcalcifications. This indicates that strict adherence to a size threshold may underestimate ultrasound's true detection capability for small metastatic deposits. In clinical practice, careful evaluation of subtle features like microcalcifications in small lymph nodes is essential to avoid missing early metastatic disease.

While our study evaluated nodal size as a component of the ultrasound assessment criteria, we did not perform a comparative analysis of the mean or median sizes of metastatic lymph nodes across the different imaging profiles (eg, US+/SPECT/CT+ versus US+/SPECT/CT−). Such an analysis could provide further insight into the size thresholds at which each modality detects metastatic disease. Future studies designed to correlate precise nodal dimensions with imaging profile subtypes would be valuable to better understand these detection dynamics.

Conclusion

In conclusion, by using a rigorous reference standard‐led cohort, this study definitively shows that ultrasound has a superior detection rate for metastatic lymph nodes compared to I‐131 SPECT/CT. The imaging profiles we defined clearly quantify the substantial fraction of RAI‐refractory disease (detected by US only) and the smaller fraction of occult disease (detected by SPECT/CT only). Microcalcifications, hypoechogenicity, and loss of hilum are the most reliable sonographic predictors in this setting, and Level VI is the dominant site for metastatic involvement. These findings validate a surveillance paradigm where ultrasound serves as the primary tool, with SPECT/CT providing critical functional stratification to guide targeted therapy, ultimately supporting a precision medicine approach in the management of recurrent PTC.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.