Surgically resected T1‐ and T2‐stage esophageal squamous cell carcinoma: T and N staging performance of EUS and PET/CT

Abstract This study aimed to evaluate the frequency of nodal metastases and to disclose the diagnostic performance of endoscopic ultrasonography (EUS) and PET/CT in T and N staging in surgically resected early‐stage esophageal squamous cell carcinomas (eSCCs). Institutional review board approved this retrospective study with waiver of informed consent for reviewing medical record. We included 435 patients with an early T‐stage (Tis or T1a [≤T1a], T1b and T2) eSCC. The rates of metastatic lymphadenopathy were calculated. Then, the performance of EUS and PET/CT in subdividing T and N stages was assessed. 131 ≤ T1a, 234 T1b, and 70 T2 eSCCs were identified. In discriminating ≤T1a from other cancers, the sensitivity, specificity, and accuracy of EUS were 60.3% (79/131), 80.3% (244/304), and 74.3% (323/435) respectively. With ROC curve analysis, cut‐off value of SUVmax 3.05 at PET provided sensitivity 74.8% (98/131), specificity 70.1% (213/304), and accuracy 71.5% (311/435) for differentiating ≤T1a eSCCs from others. Ten (7.6%) of 131 ≤ T1a cancers had nodal metastasis. In discriminating N0 from node‐positive disease, sensitivity, specificity, and accuracy of EUS were 89.6% (267/298), 41.6% (57/137), and 74.5% (324/435), respectively, whereas those of PET/CT were 88.9% (265/298), 38.7% (53/137), and 73.1% (318/435) respectively. In >70% of patients with ≤T1a eSCCs, the tumor stage can be discriminated from higher stage cancers by using EUS or PET/CT. Substantial percentage (7.6%) of ≤T1a eSCC patients have nodal metastases, which are missed in more than half of the patients in clinical staging.


| INTRODUCTION
Esophageal cancer (ECA) is one of the 10 most prevalent cancers, accounting for 4% of incidence rates among all cancers. The ECA is also one of the top 10 causes of cancer-related deaths. The most common histopathologic subtype of ECA is squamous cell carcinoma (eSCC; 85% of all esophageal cancers). 1 Surgery is the treatment of choice in localized early eSCC (T1b-T2 N0-1 and M0) and endoscopic submucosal dissection (ESD) is a treatment option for selected patients (TIS [high-grade dysplasia] or T1a, collectively called ≤T1a eSCC). 2 Thus, to differentiate T1a eSCC from T1b or higherstage eSCCs would help decide management method for T1 eSCCs. On the other hand, in locally advanced eSCCs, preoperative chemotherapy or chemoradiation therapy may help increase the rates of complete resection, thus decreasing local tumor recurrence and improving survival.
Lymph node metastases are more widely distributed in eSCCs, whereas those in esophageal adenocarcinomas (ADCs) are limited to loco-regional nodes. The rates of lymph node metastases in eSCC are relatively high even in early-stage eSCCs than those in ADCs. 3 According to a meta-analysis study, overall endoscopic ultrasonography (EUS) has good accuracy with pooled sensitivity and specificity for T1a staging of 0.85 (95% confidence interval, 0.82-0.88) and 0.87 (95% confidence interval, 0.84-0.90). Heterogeneous results from various studies may derive from multiple factors including the location and type of lesion, method and frequency of EUS probe, and the experience of the endo-sonographer. 4 In N staging, regardless of T stages, EUS has low sensitivity (usually less than 50% sensitivity), for detecting metastatic nodes. 5 To the best of our knowledge, the role of 18 F-fluorodeoxy glucose (FDG) PET/CT in differentiating T1a from higher stage eSCCs has not been published. Furthermore, PET/CT was falsely negative in early stage esophageal cancer where nodal metastasis is occasional. 5,6 Thus, little datum is available on the role of FDG PET/CT in T and N staging in earlystage esophageal cancers, particularly in eSCCs. Thus, the purpose of this study was to demonstrate the frequency of nodal metastases and to disclose the diagnostic performance of EUS and PET/CT in T and N staging in surgically resected early stage (T1-and T2-stage) eSCCs.

| MATERIALS AND METHODS
Institutional review board (IRB) approved this retrospective study; informed consent for reviewing patients' electronic medical record was not obtained from each patient.

| Patients
From January 2010 to December 2016 when surgery and its staging was recorded in consideration of 7th version of esophageal staging system, 1498 patients received esophagectomy and lymph node dissection. Of them, 732 patients proved to have eSCCs of ≤T1a, T1b, or T2 stage. Among them, 297 patients were excluded with the following reasons; 24 patients were excluded owing to neoadjuvant chemoradiation therapy (n = 7) or concurrent chemoradiation therapy (n = 17), 10 patients did not undergo either enhanced chest CT or PET/CT, 165 patients underwent PET/CT in outside hospital (difficulty in measuring maximum standardized uptake value [SUVmax]), and 98 patients underwent chest CT in outside hospital (with incomplete or different CT parameters for the evaluation of esophageal cancer and its staging). Thus, the remaining 435 patients who underwent both PET/ CT and chest CT at our institution were finally included in this study ( Figure 1).

| Endoscopic ultrasonography (EUS)
Endoscopic ultrasonography was performed by one of two physicians (with endoscopic experience of 10 and nine years, respectively) using water-filling method and adopting a radial scanning catheter probe (12/20-MHz, UM-2R/3R, Olympus, Tokyo, Japan). The probe was passed through one instrument channel of a 2-channel endoscope (GIF-2T240/GIF-2TQ260M, Olympus). The second channel was used for the instillation of 100-200 mL water by using a pump system. The depth of tumor invasion and the lymph-node status was assessed by cross-sectional images of the esophageal wall and peri-esophageal area. Lymph nodes were considered malignant, when it was larger than 10 mm in any axis, hypoechoic, and with rounded shape and clearly defined margin.

| CT and PET/CT scanning
CT scanning was performed with various helical CT scanners of four vendor companies (Data S1). Helical CT scans were obtained from the 3 cm above sternal notch to the renal hila with 2.5-3-mm detector collimation and a pitch of 0.984-1. 5 granules or air insufflation was not performed. Scanning parameters were 120 kVp and 114-275 mA under automatic exposure control; beam width, 10-20 mm; rotation time was 0.3-0.4 seconds. Image data were reconstructed by both transverse and coronal planes with a soft tissue algorithm and with 2.5-mm-thickness and at the same-thickness intervals. The reconstructed image data were directly interfaced and sent to picture archiving and communication system (Centricity 3.0; GE Healthcare, Mount Prospect, IL, USA) which displayed all image data on two monitors (1536 × 2048 matrix, 8-bit viewable gray scale, and 60-ft-lambert luminescence). Both mediastinal (width, 400 HU; level, 20 HU) and lung (width, 1500 HU; level, −700 HU) window images were viewed on these monitors.
All patients fasted for at least 6 hours before PET examination. Blood glucose levels were measured before injection of FDG and were required to be <150 mg/dL in all patients. Whole-body PET and unenhanced CT images were acquired using two kinds of PET/CT scanners (Discovery LS, GE Healthcare, Milwaukee, WI, USA; Discovery STe, GE Healthcare, Milwaukee, WI, USA), 60 minutes after the injection of FDG (5.5 MBq/kg). When the Discovery LS scanner was used, whole-body CT was performed with a continuous spiral technique with an 8-slice helical CT (140 keV; 40-120 mA; section width, 5 mm). After the CT scan, an emission scan was obtained from head to middle thigh for 4 min per frame in 2-dimensional mode. Attenuation-corrected PET images (4.3 × 4.3 × 3.9 mm) were reconstructed from the CT data using an ordered-subset expectation maximization (OSEM) algorithm (28 subsets, 2 iterations). When the Discovery STe scanner was used, whole-body CT was performed with a continuous spiral technique with 16-slice helical CT (140 keV; 30-170 mA; section width, 3.75 mm). After the CT scan, an emission scan was obtained from the head to middle thigh for 2.5 minutes per frame in 3-dimensional mode. Attenuation-corrected PET images (3.9 × 3.9 × 3.3 mm) were reconstructed from the CT data using a 3-dimensional OSEM algorithm (20 subsets, 2 iterations). Standardized uptake value (SUV) was derived from the injected dose of FDG and the patient's body weight.

| CT and PET/CT study interpretation for T and N staging
One of two radiologists (with two-year-experience of thoracic CT interpretation and with 4-year-experience of both thoracic and abdominal CT interpretation, when they first read the CT studies) who were blinded to the results of PET/ CT and pathologic results but aware of having an esophageal cancer, prospectively interpreted the CT images. Each radiologist analyzed CT images and recorded the presence or absence of abnormalities in the esophagus; no identifiable tumor along the entire course of the esophagus (≤T1a stage), or definitely visible wall thickening or tumor with >5 mm in thickness or tumor diameter but less than 10 mm (T1b stage), with 10-15 mm in thickness or tumor diameter (T2), and with >15 mm in thickness or tumor diameter (T3). When identifiable esophageal lesions were present, the location and length of the lesions were recorded in consideration of four anatomic landmarks: thoracic inlet, azygos arch, inferior pulmonary veins, and the esophago-gastric junction.
Lymph nodes were classified into the following 11-group nodal stations according to a modified version of the lymph node mapping system for esophageal cancer proposed by Korst et al 7 group Cx, cervical; group Pt, paratracheal; group 5, aortopulmonary; group 7, subcarinal; group 8, paraesophageal; group 9, inferior pulmonary ligament; group 10, hilar; group 15, diaphragmatic; group 17, left gastric; group 18, common hepatic; and group 20, celiac. In addition, lymph nodes at gastric cardia were classified into para-cardial group. Intrathoracic and abdominal lymph nodes >10 mm in shortaxis diameter, supraclavicular nodes >5 mm in diameter, and retrocrural nodes more than 6 mm in diameter were regarded as abnormal lymph nodes. 2 One of two nuclear medicine physicians (16 years and 10 years of experience in PET/CT interpretation, respectively) and one chest radiologist (26 years of chest CT interpretation and 10 years experience of PET/CT interpretation), who were blinded to clinical and pathologic results, prospectively evaluated PET/CT in consideration of chest CT scan results. As for tumors, SUVmax was measured at tumor sites. When the primary cancer was not visualized or could not be distinguished from the background (n = 70), SUVmax was assigned an assumed default value of 1.0 similar to background uptake. Regarding nodal staging, the area of FDG uptake at PET greater than mediastinal blood pool uptake where identifiable lymph nodes were present at CT, was considered to be positive for malignancy. Any lymph nodes, when they were higher in attenuation (>70 HU) than mediastinal structures or containing calcifications on unenhanced scans, were considered negative for metastasis at PET/CT, even though they showed higher FDG uptake at PET than mediastinal blood pool or SUVmax greater than 3.5. 8

| Surgery and pathologic specimen analyses
Patients with the primary tumor involving the cervical or upper thoracic esophagus (azygos arch or above, n = 49) underwent transthoracic esophagectomy (involving laparotomy, right thoracotomy, and cervical anastomosis) with LN dissection in three fields (thoracic, abdominal, and cervical [including supraclavicular nodes]). Patients with the primary tumor located in the middle portion of the esophagus (below the azygos arch down to the right inferior pulmonary vein, n = 169), the lower thoracic esophagus or gastroesophageal junction (n = 189), and middle to lower thoracic esophagus (n = 28) underwent transthoracic esophagectomy (involving laparotomy, right thoracotomy, and high thoracic anastomosis) with two-field (thoracic and abdominal) lymph node dissection (Table 1). During surgery, one of three thoracic surgeons (with 31, 28, and 17 years of esophageal cancer surgery experience, respectively), dissected all visible and palpable LNs in the surgical field, taking into consideration all results from the preoperative imaging examinations, including EUS, enhanced CT and FDG PET/CT.
The esophagectomy specimen was opened longitudinally in the fresh state. After periesophageal fat was dissected, LNs were sought. Thereafter, the specimen was fixed overnight in 10% neutral buffered formalin. Descriptions of the tumor (i.e., appearance, depth of invasion, length from both lines of resection and from cardia), the mucosal appearance, the wall thickness, and the lymph nodes (the location, size, and number of nodes seen) were recorded. Sections were obtained for the histopathologic evaluation of the tumor, the non-neoplastic mucosa, the proximal and distal lines of resection, and the lymph nodes. The specimens were stained with a standard hematoxylin-eosin technique and examined with light microscopy.
In each patient, the pathologic stage of primary esophageal cancer was recorded. A total 15 432 lymph nodes were dissected and classified as nodal stations.

| Statistical analysis
Endoscopic intervention (endoscopic submucosal dissection, ESD) can be indicated for eSCCs confined to the mucosa or lamina propria (T1a or less in stage) without submucosal invasion. 9,10 Thus, it was very important to know depth of tumor invasion and the presence of lymph node metastasis. In T staging, the distinction capability (sensitivity, specificity and accuracy) of EUS between ≤T1a and T1b or T2 was compared with that of CT or PET. For the SUVmax of primary tumor, ANOVA test was carried out among three groups (≤T1a, T1b, and T2). Then, the Pearson product-moment correlation coefficient was calculated for analyzing relationship between SUVmax and pathologic T stages and the receiver operating characteristic (ROC) curves were constructed and depicted in order to obtain the most appropriate cut-off values in terms of differentiating ≤T1a from T1b or T2 cancers and differentiating ≤T1 and T2.
The sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) of EUS and PET/CT in differentiating N0 disease from N1, N2 or N3 disease were calculated on a per-person basis.
Intermodality agreements on both T and N staging of each modality were calculated by weighted kappa for comparing its performance, because the data range of both T and N staging was not dichotomous. 11 Both bootstrap method and GEE (generalized estimating equations) were used for calculating confidence interval and comparing correlated kappa coefficients.
All statistical analyses were performed with SPSS software (SPSS for Windows, version 22.0; SPSS, Chicago, IL) and the statistical computing language R (version 3.4.3, R Foundation). Statistically significant differences were defined as having P values <.05.

| The agreement of pathologic T and N staging with EUS or CT staging
The agreement of EUS and pathologic T staging was moderate, whereas that of CT with pathologic T staging was fair (0.411 vs 0.361), but there was no significant statistical difference in T staging capability between the two modalities ( Table 4).
The agreement of EUS and pathologic N staging was fair and that of PET/CT and pathologic N staging was also fair (0.379 vs 0.321). There was no significant intermodality difference in N staging capability between the two modalities when applying weighted kappa coefficients at the 5% significance level (Table 4).
With ROC curve analysis, cut off value of

T A B L E 2 EUS, PET/CT and pathologic N staging characteristics
In clinical nodal staging of these 14 patients, none of the lymph nodes were enlarged (>10 mm in diameter) on CT. FDG PET/CT was falsely negative in the 13 patients who had pN1 (12 patients) and pN2 (one). As for T1a lesions, lymph node metastases have been variably reported but generally found to be <10% in recent series, even though most reports have dealt with esophageal adenocarcinomas rather than eSCC. 13 In a cumulative-data analysis, Merkow et al 13 found that 5% of patients with T1a lesions (of 5390 patients, 10.6% eSCC) have involved lymph node metastases and 16.6% for T1b tumors. Differently Duan et al 14 reported that lymph node metastasis rates are 17.5% for pT1 eSCC tumors, 16.0% (8 of 50) for pT1a tumors, and 22.6% (21/93) for T1b tumors. Of 25 patients with lymph node metastasis, one had cervical metastasis, 15 had thoracic metastasis, and 17 had abdominal metastasis. In our study, 10 (7.6%) of 131 with ≤T1a cancers, and 88 (37.6%) of 234 with T1b cancers had nodal metastases. The rates were similar to those in a previous report published at our institution, where the rates of lymph node involvement were 6.25% (4 of 64) in T1a cancers and those of T1b cancers were 29.3% (39 of 133) (P < .001). 15 Thus, In ≤T1a eSCCs, lymph node metastases were much lesser than those in the study of Duan et al, 14 whereas those were much higher in T1b esophageal cancer. Given the context that the management could be ESD in T1a esophageal cancers, the lymph node metastasis rates in our study appear to be more applicable and reliable.
Comparison of FDG uptake in the primary tumor and clinicopathologic staging findings showed that there is a significant association between the FDG uptake in the primary tumor and depth of invasion. 16  Endoscopic ultrasonography has some limitations in T staging, because EUS is not an option for evaluating stenotic tumors. Stenotic esophageal cancers account for 30% of the cancers. However, in our case series of T1a-T2 stages, stenotic tumors hampering the passage of endoscopic probe were absent. EUS also has some difficulty in nodal staging when the target nodes are not accessible. 17 In our study, T staging in both EUS and CT showed only less than moderate degree of agreement with pathologic T staging and there was no intermodality difference between EUS and CT or PET/CT for predicting the pathologic T staging. PET/CT showed high sensitivity in particular for differentiating ≤T1a eSCCs from higher-stage cancers by noticing no tumor on CT and by measuring SUVmax of ≤3.05 in patients with an esophageal cancer seen on endoscopy. We think the criteria of no identifiable tumefaction on CT and little FDG uptake of <3.05 at PET help discriminate the ≤T1a cancers, thus, allowing us to select candidates for ESD.
N staging in both EUS and PET/CT demonstrated only fair agreement with pathologic N staging, and there was no significant difference in their N staging performance between the two modalities. Therefore, it would provide interesting prognostication results to compare survival between those showing N0 disease in clinical staging but positive nodal metastasis in pathologic examination and those demonstrating true clinical and pathologic N0 stage.
In conclusion, with SUVmax cut-off value of 3.05, differentiation of T1a from higher T stage cancers can be achieved in >70% of surgically resectable eSCCs and PET/CT and EUS provide comparable performance in differentiating N0 eSCCs from node metastasis-positive cancers. Substantial percentage (7.6%) of ≤T1a eSCC patients have nodal metastases, and nodal metastasis rates increase as T stage increases (T1b [37.6%] and T2 [55.7%]). Moreover, more than half of nodal metastases were missed on PET/CT or EUS. Thus, after endoscopic surgery or even after curative surgical resection of <T1a eSCCs, adjuvant therapy is needed for those having nodal metastasis.