The study population was obtained from a historical cohort of consecutive patients who were newly diagnosed with OPSCC without evidence of distant metastases at Asan Medical Center between July 2006 and November 2016. The inclusion criteria used to select patients were 1) patients who underwent p16 immunohistochemistry analysis or HPV deoxyribonucleic acid (DNA) detection, and 2) patients who underwent contrast-enhanced neck CT or MRI before treatment. Patients were excluded from the study population if they 1) showed a negative result in p16 immunohistochemistry analysis or HPV DNA detection, 2) had a prior cancer history or synchronous cancers, or 3) showed no metastatic lymph node on pretreatment CT or MRI.

Therefore, 313 consecutive OPSCC patients underwent p16 immunohistochemistry analysis or HPV DNA detection, and 237 were identified as having HPV-positive OPSCC. After excluding 103 patients, 134 patients were included in this study (Fig. 1), which included 118 men (mean age, 59.9 years; age range, 58.2–61.5 years) and 16 women (mean age, 59.9 years; age range, 55.5–64.2 years) with a mean age of 59.9 years (age range, 58.3–61.4 years).

Patients were staged according to the 8th edition of the AJCC-TNM staging system by using physical and endoscopic exams, CT, MRI, and 18F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET)/CT imaging. Information regarding patient age, sex, smoking status, treatment methods, and survival status at last follow-up were obtained from the medical records.

Two methods are mainly used to evaluate HPV infection status in our institution, with one being tumor p16 immunohistochemical analysis using CINtec p16 histology (anti-p16^INK4a mouse monoclonal antibody and immunohistochemical detection kit; Roche MTM Laboratories, Heidelberg, Germany). A positive p16 result was defined as diffuse strong staining in the cytoplasm and nucleus. Focal or faint reactivity was considered negative for p16. The other method used was tumor HPV DNA detection by polymerase chain reaction (PCR)/DNA chip scanning, which can detect 43 subtypes of HPV (high-risk subtypes 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, and 82 and other lower or undetermined risk subtypes). HPV-related OPSCC was diagnosed when the results of either p16 or HPV DNA PCR were positive (18).

After completing treatment, patients were evaluated by physical examination, endoscopy, and imaging (e.g., CT, MRI, or ¹⁸F-FDG PET/CT) every 2–3 months for the first year, according to the protocols of our hospital, which were based on the guidelines of the National Comprehensive Cancer Network. Follow-up intervals were increased to 3–6 months in the second year and 6–12 months thereafter. To avoid extended follow-up and ensure efficient assessment, progression-free survival (PFS) was chosen as the primary outcome (19). PFS was calculated from the first day of treatment to the date of disease progression (locoregional recurrence or distant metastases), death from any cause, or the date of the last follow-up visit (censored). The minimum follow-up period for ascertaining PFS was 12 months. All locoregional recurrences were diagnosed by physical examination, flexible endoscopy and biopsy, and/or CT or MRI of the neck showing progressive bone erosion and/or a soft tissue mass. Distant metastases were diagnosed using imaging methods that included whole-body bone scan, CT, MRI, or ¹⁸F-FDG PET/CT. A secondary outcome was the diagnostic performance of CT or MRI for the diagnosis of ENE among the patients who underwent neck dissection for lymph nodes. The reference standard was surgical pathologic results.

A total of 122 patients underwent contrast-enhanced CT scans with one of several different multidetector CT systems with 64–128 channels (Siemens Healthineers, Erlangen, Germany; GE Healthcare, Milwaukee, WI, USA). The typical imaging parameters were as follows: 120-kV tube voltage; 200-mAs effective tube current; 22-cm display field-of-view (FOV); 50-cm large-body scan FOV; and axial and coronal sections with a 3-mm thickness reconstructed using a soft tissue algorithm. The scan range extended from the upper margins of the frontal sinus to the top of the aortic arch. All scans were acquired at 70 seconds after intravenous administration of 140 mL of non-ionic iodinated contrast media (iopamidol, Isovue-370; Bracco Diagnostics, Princeton, NJ, USA) at a rate of 2.5 mL/s.

MRI was performed on 117 patients by using a 3T MR scanner (Achieva; Philips Healthcare, Best, The Netherlands) with a 16-channel neurovascular coil (SENSE NV coil; Philips Healthcare). The neck imaging protocol contained the following sequences: axial T1-weighted imaging, axial T2-weighted imaging, axial fat-suppressed contrast-enhanced T1-weighted imaging (CE-T1WI), coronal T1-weighted imaging, coronal fat-suppressed T2-weighted imaging, and coronal fat-suppressed CE-T1WI. CT or MR images were analyzed on Picture Archiving and Communication System (PACS). The mean interval between staging CT or MRI and the start of treatment was 49 days (range, 7–90 days).

Among 134 patients, 105 patients underwent both CT and MR examinations. Seventeen patients underwent only CT scans and 12 patients underwent only MR scans.

All CT or MRI studies were interpreted in an independent manner by two board-certified radiologists with 5 and 12 years of clinical experience in head and neck imaging. Before evaluation, the two neuroradiologists completed a training session with five patients to help them reach a consensus on evaluation of the imaging findings. Lymph nodes were considered as positive for radiologic ENE when at least one of the following criteria was met (92021222324): 1) enhancement, thickening, and irregularity of the nodal rim and 2) infiltration of the adjacent fat or other soft tissue planes (Fig. 2). If there was a discrepancy over the presence of radiologic ENE between CT and MRI (i.e., one of the two yielded a positive result), radiologic ENE was considered to be present. If patients underwent lymph node surgery, the images depicting the surgical dissection sites were reviewed. The imaging results, including the presence of radiologic ENE in any cervical lymph node of the patients, were compared with the nodal histopathology results.

The patient's clinicopathological features were analyzed according to their radiologic ENE status by using a chi-squared test or Fisher's exact test for categorical variables, and a t test for continuous variables. Univariate Cox regression analysis was performed to identify predictors of PFS among the following variables: age, sex, smoking status, T-stage, N-stage, overall stage, radiologic ENE, primary tumor site, treatment modality, and interval between imaging and treatment. A multivariate Cox regression analysis was performed using backward elimination. The distribution of PFS was estimated by the Kaplan-Meier method and compared using a log-rank test.

The sensitivity, specificity, accuracy, positive and negative predictive values, and area under the receiver operating characteristic curve of CT and MRI for depicting the presence of radiologic ENE were calculated relative to the histopathologic results.

Interobserver agreement in the image analysis of neck CT or MRI for radiologic ENE was calculated by using Cohen's kappa index. Kappa values were interpreted as indicating poor (κ < 0.1), slight (0.1 ≤ κ ≤ 0.2), fair (0.2 < κ ≤ 0.4), moderate (0.4 < κ ≤ 0.6), substantial (0.6 < κ ≤ 0.8), and nearly perfect (0.8 < κ ≤ 1.0) agreement (252627). Statistical analysis was performed with SPSS version 23 (IBM Corp., Armonk, NY, USA).

The median follow-up period was 38 months (range, 12–107 months). The mean lymph node size (the minimum axial diameter) was 1.9 cm (range, 0.5–4.0 cm). Radiologic ENE showed a statistically significant association with high N-stage and overall stage (p = 0.012 and p = 0.041, respectively), as shown in Table 1.

In the univariate Cox regression analysis, the presence of radiologic ENE, increased N-stage (N1 vs. N2, N3), increased T-stage (T0, T1, T2 vs. T3, T4), and overall stage (I, II vs. III) were significant adverse variables for PFS (p = 0.035, p = 0.004, p = 0.026, and p = 0.006, respectively) (Table 2). In the multivariate Cox regression analysis, N-stage and overall stage were statistically associated with PFS (p = 0.041 and p = 0.022, respectively), but the presence of radiologic ENE was not (p = 0.141; hazard ratio, 2.68; 95% confidence interval [CI], 0.72–9.97). The Kaplan-Meier disease-free survival curves show that PFS was significantly lower in patients with radiologic ENE-positive disease, with a 3-year PFS of 95.3% versus 83.7%, respectively (log-rank p = 0.023) (Fig. 3).

Of the 57 patients who underwent neck dissection and received a pathologic evaluation for ENE, 21 (36.8%) had pathologically confirmed ENE. Among these patients, the area under the receiver operating characteristic curve for diagnosis of radiologic ENE on CT or MRI was 0.70 (range, 0.57–0.83), and the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 62.0%, 77.8%, 61.9%, 77.8%, and 71.9%, respectively (p = 0.002).

The interobserver agreement between the two neuroradiologists for interpreting radiologic ENE was substantial (kappa = 0.79; 95% CI, 0.67–0.89). Specifically, for enhancement, thickening, and irregularity of the nodal rim, it was nearly perfect (kappa = 0.83) while it was substantial (kappa = 0.73) for infiltration of the adjacent fat or other soft tissue planes. The interobserver agreement between the two neuroradiologists for radiologic ENE was substantial on both MRI (kappa = 0.76; 95% CI, 0.64–0.88) and CT (kappa = 0.77; 95% CI, 0.66–0.88).