This study was approved by Our Hospital Institutional Research Committee, and written informed consent for medical research was obtained from all patients prior to initial treatment. All methods were performed in accordance with relevant guidelines and regulations.

Our study was a single-center, retrospective cohort analysis of Chinese patients with primary high-risk HNcSCC who underwent surgical treatment between January 2010 and December 2023. Data were extracted from electronic medical records after institutional review board approval. A consecutive sampling approach was employed for all eligible patients meeting the predefined criteria: Histologically confirmed primary HNcSCC with ≥1 high-risk feature; completion of neck staging via ultrasound, contrast-enhanced CT, or SLNB; minimum follow-up of 12 months. Patients with incomplete records, or insufficient follow-up were excluded to minimize selection bias.

Tumor stage was determined according to the 8^th edition of the AJCC staging system and the Brigham and Women’s Hospital (BWH) classification (7). High-risk factors included a maximum lesion diameter of ≥2.0 cm, location on the temple, ear, or lip, immunosuppression, thickness greater than 6.0 mm or invasion beyond subcutaneous fat, poor differentiation, PNI, and bone erosion (8–10). PNI was deemed positive if tumor cells were present within the nerve, measuring 0.1 mm or greater (11).

Primary outcome variables included the five-year regional control (RC) and five-year overall survival (OS). The RC time was calculated from the date of surgery to the date of first neck recurrence or last follow-up, whereas OS time was calculated from the date of surgery to the date of death or last follow-up. Secondary outcome variables encompassed the sensitivity and specificity of radiological imaging and SLNB in detecting LN metastasis, as well as management changes prompted by these imaging modalities. In sensitivity analyses, nodal metastasis was defined as a pathologically confirmed metastasis detected by biopsy or neck dissection during the initial investigation or within six months of follow-up for each patient. The additional six-month period served as validation for the usage of fine needle aspiration biopsy as a reference standard, capturing potential false-negative results at baseline with the assumption that any metastasis detected within this timeframe would have been present during the initial biopsy. If sample analyses yielded insufficient material for robust cytological investigation, the procedure was repeated until a conclusive determination could be made (12).

During management change analysis, a comprehensive chart review of all patients was conducted to collect the following information: the date, type, and outcomes of imaging modalities; the rationale behind imaging; and the impact of imaging on patient management, which was assessed through clinical notes and included alterations to surgical approaches (13–15). Ultrasound was the initial modality for assessing neck status, followed by CT and SLNB when indicated. The management plan was adjusted if surgical approaches were modified based on CT or SLNB findings.

Patients with HNcSCC exhibiting high-risk features should undergo imaging analysis to assess potential lymph node metastasis. Criteria for suspecting LNs on ultrasound included a short axis measuring larger than 5 to 6 mm, a round rather than oval shape, absence of fatty hilum, or presence of extranodal extension (12). Suspicious lymph nodes on contrast-enhanced CT were classified based on appearance into accelerated, enlarged, and necrotic. The short axial diameter was utilized for measuring LN size. Nodes were categorized as accentuated if they exhibited enlargement, especially in side-by-side comparisons, while remaining under 10 mm. Necrosis within melted lymph nodes was defined as a central area of low attenuation surrounded by an irregular rim of enhancing tissue (16).

The SLNB procedure adhered to established protocols as previously described in the literature (17). On the day prior to surgery, 99mTc-nanocolloid was administered via submucosal injection around the perimeter of the primary tumor site. Dynamic lymphoscintigraphy was performed in both anterior and lateral views, followed by single-photon emission CT imaging. Utilizing a γ-probe and methylene blue staining, the sentinel LN was accurately localized on the skin surface, allowing for real-time assessment of sentinel LN status through frozen section analysis. During this procedure, two consecutive slices of fixed LNs were obtained at their maximum cross-section and subjected to hematoxylin and eosin staining, with or without adjunct immunohistochemistry. The samples were subsequently forwarded for postoperative pathological evaluation, and following dehydration and fixation, were sectioned at maximum cross-section to provide one slice for hematoxylin and eosin staining, with or without immunohistochemistry.

For the sample size calculation of 5-year RC, the estimated 5-year RC rates were set at 80% (U group), 90% (UC group), and 99% (UCS group). With a significance level of 0.05 and a power of 90%, the minimum required total sample size was approximately 270 patients. For 5-year OS, the assumed 5-year OS rates were 85% (U group), 95% (UC group), and 99% (UCS group). Using the same significance level and power (90%), the minimum total sample size was estimated at 240 patients.

A multiple imputation approach was employed to address missing data patterns for differentiation, PNI, and LVI. Missing rates among the variables were 11.4% for differentiation, 10.5% for PNI, and 10.0% for LVI (18, 19).

Patients were categorized into three groups based on the method of neck evaluation, and their clinicopathological variables were compared using the Chi-square test. Potential factors influencing RC and OS were assessed via univariate analysis firstly, and those significant variables were further evaluated via the Cox proportional hazards model with the presentation by hazard ratios (HR) and 95% confidence intervals (CI). Sensitivity and specificity of different neck evaluations in predicting lymph node metastasis was assessed by 2×2 confusion matrix. Sensitivity represented the proportion of true positives correctly identified by the test, calculated as the ratio of true positives to the sum of true positives and false negatives. Specificity reflected the proportion of true negatives accurately detected, defined as the ratio of true negatives to the sum of true negatives and false positives. The positive predictive value indicated the probability that a positive test result truly represented a positive condition, computed as the ratio of true positives to the sum of true positives and false positives. Conversely, the negative predictive value measured the probability that a negative test result genuinely corresponded to a negative condition, derived as the ratio of true negatives to the sum of true negatives and false negatives.

All statistical analyses were conducted using R version 3.4.3, with a p-value of < 0.05 considered statistically significant.

A total of 270 patients were analyzed, with a mean age of 65 ± 8 years, comprising 213 males and 57 females. Seven patients had undergone organ transplants. Clinical LN metastasis was identified in 32 individuals. Among the 283 tumors, 178 were located in the temple, ear, or lip regions. Tumor size was less than 2.0 cm in 148 patients, while 78 patients exhibited invasion of subcutaneous fat. Poor differentiation was noted in 70 cases, and PNI and LVI occurred in 42 and 34 patients, respectively. Thirteen tumors presented with bone erosion. According to the AJCC staging system, 217 tumors were classified as T1/2 and 66 as T3/4. Based on the BWH stage, 113 cases were determined as T1/2a and 170 as T2b/3. All tumors underwent surgical excision, with no case demonstrating positive margins.

Neck status was evaluated by ultrasound alone in 91 patients (U group), by simultaneous ultrasound and CT in 102 patients (UC group), and by ultrasound, CT, and SLNB in 77 patients (UCS group). A statistically significant difference was observed among the three groups regarding clinical neck stage distribution (p=0.001), with no patients classified as cN+ in the UCS group. Other clinicopathological variables exhibited similar distributions across the three groups ( Table 1 ).

During a median follow-up period of 4.5 years, regional recurrences were recorded in 8 patients from the U group and 10 from the UC group, whereas no neck failures were noted in the UCS group. Death occurred in 11 patients in the U group, 16 in the UC group, and 3 in the UCS group.

In univariate analysis, factors such as tumor size, invasion depth, differentiation, PNI, AJCC tumor stage, BWH tumor stage, neck evaluation method, and LN metastasis were found to be associated with both RC and OS (all p<0.05, Figure 1 , Table 2 , and Table 3 ). These variables were further evaluated in multivariable analysis. Immunosuppression was significantly related to OS (p=0.008) but not to RC (p=0.118) and was included in the Cox model for OS calculation.

In the multivariable analysis for RC, compared to the UCS group, patients in the UC group exhibited a HR of 1.48 [95% CI: 1.06-2.77], while those in the U group demonstrated an HR of 1.43 [95% CI: 1.10-3.00]; both findings were statistically significant (p=0.013 and p=0.009, respectively). Other independent factors included tumor size, invasion depth, PNI, differentiation, AJCC tumor stage, BWH tumor stage, and LN metastasis ( Table 2 ).

For OS in multivariable analysis, comparisons to the UCS group revealed that patients in the UC group had an HR of 1.67 [95% CI: 1.11-2.89], while those in the U group also had an HR of 1.69 [95% CI: 1.21-3.12]; both differences were significant (p=0.008 and p=0.005, respectively). Other independent factors included immunosuppression, tumor size, invasion depth, PNI, differentiation, AJCC tumor stage, BWH tumor stage, and LN metastasis ( Table 3 ).

A subgroup analysis was performed to assess whether the impact of diagnostic modalities varied by treatment method. All patients underwent curative resection, but only 21 received adjuvant chemoradiation. The patients were divided into two groups: those who did not receive adjuvant therapy (n=146) and those who underwent adjuvant therapy (n=124). In both cohorts, UCS demonstrated the best RC and OS ( Figures 2 and 3 ).

Within the UC group, contrast-enhanced CT identified seven occult metastases that ultrasound failed to detect; these patients subsequently underwent aspiration (n=5) and neck dissection (n=2), yielding a management change rate of 6.8%. In the UCS group, a cN0 status was documented via ultrasound and CT, though SLNB revealed LN metastasis in six patients, all of whom underwent neck dissection, resulting in a management change rate of 7.8% ( Table 4 ).

In the overall population, LN metastasis occurred in 22 patients at the initial treatment. Among the three modalities evaluated, SLNB demonstrated the highest diagnostic accuracy, with a sensitivity of 85.7% and a specificity of 100%. Ultrasound and CT exhibited comparable sensitivity and specificity; their negative predictive value aligned closely with that of SLNB. However, both ultrasound and CT had significantly lower positive predictive values compared to SLNB ( Table 4 ).