Surgically Resected Esophageal Squamous Cell Carcinoma: Patient Survival and Clinicopathological Prognostic Factors

We aimed to report patients’ survival after surgical resection of eSCC and to ascertain the clinical, imaging, and pathological factors related to patient prognosis. This retrospective study included 435 patients with eSCC of <stage T2 (median follow-up period, 49.3 months). A total of 103 (23.7%) patients died, and 89 (20.5%) experienced recurrence during follow-up. The maximum standardized uptake value (SUVmax) on positron emission tomography (PET)/computed tomography (CT) of the primary tumor was significantly correlated with tumor length, nodal metastasis, and pathologic T stage in a positive linear fashion. In the multivariate analysis, higher SUVmax on PET/CT was a negative prognostic factor for both disease-free survival (DFS) and overall survival (OS). Contrarily, the presence of nodal metastasis was a prognostic factor only for DFS, and pathologic T stage only for OS. By applying SUVmax cut-off, both DFS and OS were significantly different among three groups when divided by cut-off values (A: SUVmax ≤ 3.05, B: SUVmax 3.06 - 5.64, C: SUVmax ≥ 5.65). In patients with a surgically resectable eSCC, measuring the SUVmax of the primary tumor during PET/CT can help predict patient survival. Additionally, PET/CT renders triage criterion for endoscopic submucosal dissection (ESD; T1a cancer and SUVmax, ≤3.05).

Relationships among SUV max and tumor characteristics, t, or n stage. SUV max was significantly correlated with both pathologic T stage and the presence of nodal metastasis in a linearly positive fashion (r = 0.536, p < 0.001; r = 0.282, p < 0.001; respectively). However, there was no significant correlation between SUV max and tumor differentiation (p = 0.520). SUV max was also significantly correlated with pathologic N stage in a linearly positive fashion (r = 0.313, p < 0.001). Tumor size, defined as the maximum diameter of the tumor in a pathological specimen, was also significantly linearly correlated with SUV max (r = 0.342, p < 0.001).
Validation of the diagnostic performance of preoperative T staging with PET/CT. There were no significant differences in terms of the demographics or tumor characteristics of patients in the validation group and the original experimental group. Details of patients' characteristics are described in Table 2.
Prognostic significance of SUV max with survival and recurrence. In the receiver operating characteristic (ROC) curve analysis, a cut-off value of SUV max 3.05 and 5.65, respectively, was the most useful for differentiating <T1a eSCCs from other cancers and for differentiating T1 (<T1b) eSCCs from T2 eSCCs 9 .
By applying these SUV max cut-off values, we statistically proved that both DFS and OS were significantly different among the three groups when divided by the cut-off values (A: SUV max ≤ 3.05, B: SUV max 3.06-5.64, C: SUV max ≥ 5.65), except the OSs between groups A and B (DFS, p < 0.001; A vs. B, p = 0.005; A vs. C, p < 0.001; B vs. C, p = 0.010; OS, p < 0.001; A vs. B, p = 0.167; A vs. C, p < 0.001; B vs. C, p = 0.009; Figs. 2 and 3). The five-year DFS was 86.5% in group A, 78.4% in group B, and 59.5% in group C ( Table 3). The five-year OS was 80.6% in group A, 78.7% in group B, and 59.5% in group C (Table 4). However, clinical T stage, determined by EUS, was not statistically different between the groups in terms of DFS or OS (DFS: p = 0.324, OS: p = 0.753) (Figs. 2 and 3).
In the univariate analysis, sex, adjuvant therapy, nodal metastasis, SUV max , and pathological T stages were significant prognostic factors for tumor recurrence. SUV max , the presence of nodal metastasis, and pathological T stage were significant prognostic factors for survival. As the cut-off values of SUV max were extracted by pathological T stage groups, we assumed two different models for the multivariate analysis using pathological T stage and hypothetical stages using SUV max (PET stage groups, Tables 3 and 4). This was despite the severity of multicollinearity not being significant between those with pathological T stage and hypothetical stage as determined by the SUV max (variance inflation factor [VIF] = 1.594). In the multivariate analysis, SUV max and the presence of nodal metastasis were significant factors for disease recurrence. SUV max and pathological T stage were significant factors for patient survival (Tables 3 and 4 www.nature.com/scientificreports www.nature.com/scientificreports/ at all time points (pathological T staging, 0.598; clinical T staging using EUS, 0.533). For predicting the OS, the iAUC value of hypothetical staging using SUV max was 0.602. This was also greater than those of other staging systems at all time points (pathological T staging, 0.580; clinical T staging using EUS, 0.525).

Discussion
It is known that with surgically resected eSCCs, the pathological stage is an independent risk factor for recurrence within the first year after surgery and that the presence of lymph node metastasis is the most common relapse pattern after an esophagectomy 10 .
In our study, the SUV max on PET/CT was also observed as an independent factor for predicting both future recurrences and patient survival, along with the pathologic T stage (for predicting survival) and nodal metastasis presence (for disease recurrence). Additionally, the SUVmax was positively correlated with both pathological T and N stages and tumor length.
There have been a few studies published recently regarding the value of measuring SUV max on PET/CT as a prognostic factor in eSCCs. These studies showed similar results to those seen in this study. According to Jeon et al. 11 , venous invasion and high SUV max could be important prognostic factors for disease recurrence in T1N0M0 eSCCs. Song et al. also reported similar results from their cohort in which all stages of eSCCs were included 12 . Our results were derived from a patient cohort, included a larger study population. And our study had a longer follow-up period than other studies. Moreover, we performed a more comprehensive analysis to look at predicting pathological T stage and also survival using the SUV max of primary tumors.
In the Kaplan-Meier curve analyses for both DFS and OS based on the SUV max cut-off, two survival curves of groups B and C crossed over each other at the time point of 65 months after surgery (Figs. 2b and 3b). The same phenomenon was also seen in the Kaplan-Meier curve for OS based on the pathologic T staging (Fig. 3a). This may have resulted from a small number of samples whose follow-up periods were greater than 70 months. The 95% CIs of group B for both DFS and OS widened rapidly 70 months after surgery (Supplemental Fig. 2).
The usefulness of FDG PET/CT for tumor staging or prognostication in esophageal cancer has been unclear, particularly in patients with early-stage (T1) cancers and in esophageal adenocarcinomas. Cuellar et al. 13 asserted that FDG PET/CT is not useful in evaluating adenocarcinoma of the esophagus when endoscopic biopsy discloses Tis and T1 in tumor stage. They believed that because regional nodal metastases are uncommon and distant www.nature.com/scientificreports www.nature.com/scientificreports/ metastases are rare in patients with T1-stage esophageal cancers as well as because FDG PET/CT can lead to inappropriate management, FDG PET/CT should not be used in evaluating patients with clinical Tis and T1 esophageal adenocarcinomas. Contrastingly, in our study, in which eSCCs of T1a, T1b, and T2 were included, measuring SUV max appeared to help differentiate the T stages and in predicting oncoming recurrent disease (DFS) and patient survival (OS) in patients who had a surgically resected eSCC.
Reduced spatial and contrast resolutions of PET/CT were considered one of the interpretative downsides of using PET/CT for T-descriptor 14 . However, by using SUV max of the primary tumor rather than the visualized gross tumor volume, PET/CT can help differentiate pathologic T stages. In our previous study, measuring the SUV max of primary cancer was shown to assist in differentiating <T1a from T1b or T2 cancers and between <T1 and T2 cancers 9 .
In our study, the cut-off value of SUV max 3.05 was effective in discriminating T1a-stage from T1b-or T2-stage eSCCs. Similarly, Furukawa et al. 15 showed that FDG-PET helped to diagnose tumors in 40 consecutive cT1N0M0 eSCC patients involving the submucosa but not beyond the middle one-third of the submucosa (SM2) and beyond, or tumors having occult lymph node metastasis. Their proposed optimal cut-off SUV max value of 2.7  www.nature.com/scientificreports www.nature.com/scientificreports/ can be used distinguishing ESD candidate patients from advanced stage eSCC patients with an SM2 involvement and beyond (21, 52.5%) or lymph node metastasis (6, 15%). Our optimized cut-off value SUV max 3.05 was slightly higher, partly because we tried to discriminate eSCCs of T1a or less from higher-stage cancers. It could also be because we used a larger study population (435 eSCC patients). The pathologic T1b stage includes any tumors involving the submucosa (SM1, SM2, and SM3).
In a prospective validation study using different patient cohorts, PET/CT, and the SUVmax of the tumor showed its efficaciousness in triaging patients with various T stages. Furthermore, in the current study, www.nature.com/scientificreports www.nature.com/scientificreports/ hypothetical T stages determined with the measured SUV max regarding the primary tumor appear to be more practical and accurate in predicting patients' prognoses than those determined with EUS. When a time-dependent AUC metric was used, the prediction model adopting the hypothetical T sage on disease recurrence and survival excelled the other models adopting pathologic T stage and clinical T stage by using EUS. The SUV max of the tumor positively correlated with the depth of tumor invasion (pathologic T stage) and also with the tumor length and nodal metastasis (pathologic N stage). The pathologic T stage or clinical T stage (determined by the use of EUS) might not have reflected the relationship with nodal metastasis. www.nature.com/scientificreports www.nature.com/scientificreports/ It is known that the length of eSCC has a positive correlation with the SUV max of the primary tumor 16,17 . In a study by Xu et al. 18 , the tumor length demonstrated a positive correlation with tumor recurrence. However, in the present study, the tumor length was not an independent prognostic factor in surgically resectable eSCCs.
Our study had several limitations. First, it was conducted in a single tertiary referral hospital. Second, our study population was retrospectively recruited from our surgical registry, and the patients had a resectable early-stage (only T1 and T2-stage) eSCC. This may have contributed to selection bias. Third, we included patients who underwent both CT and PET/CT at our institution. Thus the measurement methods of SUV max and CT and PET/CT study protocols were uniformly standardized. Therefore, the results of our study may not be generalized for patients with eSCC worldwide. However, we tried to include a large number of early-stage eSCC patients. Further studies with a multi-centered prospective design and with a larger number of patients may be needed to validate our study results. Fourth, we excluded patients who underwent neoadjuvant chemotherapy. Future studies including patients undergoing neoadjuvant chemo-or radiation therapy may be needed to assess the effectiveness of measuring SUV max on PET/CT in eSCC patients. Lastly, we did not consider the possible effect of the angiolymphatic or perineural invasion of the primary tumor in the survival analysis. This was because there were only a few identified angiolymphatic or perineural invasion cases in our patient cohort.  www.nature.com/scientificreports www.nature.com/scientificreports/ In conclusion, of the 435 patients undergoing surgical resection with a T1-or T2-stage eSCC, 103 (23.7%) patients died, and 89 (20.5%) patients experienced tumor recurrence during the median follow-up period of 49.3 months (range; 0.8 to 105 months). In the multivariate analysis, the SUV max of the primary tumor (eSCC) on PET/CT was an independent factor for predicting future recurrences and patient survival. The pathologic T stage was a predictor for survival, and nodal metastasis was a predictor for recurrence. Furthermore, the SUV max was significantly correlated with both pathological T and N stages in a linearly positive manner. Therefore, in patients with a surgically resectable eSCC, measuring the SUV max of the primary tumor could help predict patient survival.

Materials and Methods
Study population and data collection. Using the Esophageal Cancer Surgery Registry at the Samsung Medical Center (a 1,979-bed tertiary referral hospital in Seoul, South Korea), 1498 patients who received esophagectomy and lymph node dissection were identified between January 2010 and December 2016. Of them, 732 patients had <T1a, T1b, or T2 stage eSCCs. Among them, 297 patients were excluded for the following reasons: 24 patients were excluded due to having undergone neoadjuvant chemoradiation therapy (n = 7) or concurrent  Table 4. Univariate and Multivariate Analyses Evaluating Overall Survival. Note __ * Calculated using the log-rank test of the differences between the two survival curves generated using the Kaplan-Meier curve; **Calculated using the Multivariate Cox proportional hazard model; ***Groups by tumor length were divided using a cut-off 2.5 cm close to the mean value of the tumor length. EUS: Endoscopic ultrasonography; PET: positron emission tomography; SUV max : maximum standardized uptake value; CI: confidence interval.
Scientific RepoRtS | (2020) 10:5077 | https://doi.org/10.1038/s41598-020-62028-5 www.nature.com/scientificreports www.nature.com/scientificreports/ chemoradiation therapy (n = 17), 10 patients did not undergo either enhanced chest CT or PET/CT, 165 patients had their PET/CT in outside hospital (difficulty in measuring SUV max ), and 98 patients had their chest CT in an outside hospital (with incomplete or different CT parameters for evaluating ECA and its staging). The remaining 435 patients who underwent both PET/CT and chest CT at our institution were included in this study 9 (Fig. 4).
For the validation study of FDG-PET/CT as a possible T descriptor, we also found 224 patients who received an esophagectomy and lymph node dissection between January 2017 and December 2018. Of these, 110 patients were excluded for the following reasons: 50 were more than T3 stage, 38 had undergone their PET/CT in an outside hospital, and 22 had undergone surgery for recurrent ECA. Thus, 114 patients were included in the validation study, where we prospectively performed T staging using the SUV max criteria acquired with an experimental study. The diagnostic performance of FDG-PET/CT for a T descriptor was compared with that of the experimental study.
Patient-related (age, sex, adjuvant chemotherapy, and survival), surgery-related (type of surgery, and surgical resection margin)-, and tumor-related (length, location, histology, and pathologic stage) factors were collected from the database. Details of patients' surgeries and their pathologic specimen analyses are described in our previously published article 9 .
The Institutional Review Board (IRB) of the Samsung Medical Center approved this retrospective study (IRB no. 2017-04-093). Informed consent for reviewing patients' electronic medical records was waived by the IRB. Our study design, data collection, and analysis were performed as per the relevant guidelines and regulations.
Preoperative PET/CT scanning and Interpretation. All patients fasted for at least 6 hours before their PET examination. Blood glucose levels were measured before the injection of FDG and were required to be <200 mg/dL in all patients. Whole-body PET and unenhanced CT images were acquired using two types 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 STe scanner was used, a whole-body CT was performed using a continuous spiral technique with a 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 min per frame in a 3-dimensional mode. Attenuation-corrected PET images (3.9 × 3.9 × 3.3 mm) were reconstructed from the CT data using an ordered-subset expectation maximization (OSEM) algorithm (20 subsets, 2 iterations). When the Discovery LS scanner was used, a whole-body CT was performed using 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 the head to middle thigh for 4 min per frame in a 2-dimensional mode. Attenuation-corrected PET images (4.3 × 4.3 × 3.9 mm) were reconstructed from the CT data using an OSEM algorithm (28 subsets, 2 iterations). The standardized uptake value (SUV) was derived from the injected dose of FDG, and the patient's body weight 9 .
One of the 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 of PET/CT www.nature.com/scientificreports www.nature.com/scientificreports/ interpretation), who were blind to the clinical and pathologic results, evaluated the PET/CT in consideration of the chest CT results. When identifiable esophageal lesions were present, the location was recorded with four anatomic landmarks used for categorization: thoracic inlet, azygos arch, inferior pulmonary veins, and the esophagogastric junction.
As for T staging, the SUV max was measured at the tumor sites. When the primary cancer was not visualized or could not be distinguished from the background (n = 70), the SUV max was assigned an assumed default value of 1.0, similar to the background uptake.
In our previous study, the ROC curves were constructed and depicted to obtain the most appropriate cut-off values in terms of differentiating <T1a from T1b or T2 and differentiating <T1 and T2 9 . We divided patients into three hypothetical PET staging groups A, B and C using two cut-off values of SUV max 3.05 and 5.65: A = SUV max < 3.05, SUV max 3.05 < B < SUV max 5.65, and C = SUV max ≥ 5.65, respectively. patient surveillance. Patients in this retrospective cohort were followed up regularly following the specific surveillance protocols at our institution after surgery. Specifically, out-patient based clinic appointments were arranged every 6 months at our institution for 5 years. After 5 years, patients were usually referred to a secondary referral hospital in their hometown. Follow-up imaging studies were also performed at specific intervals: a chest x-ray every month for the first 2 months to check for complications after surgery, a contrast-enhanced chest CT scan every 6 months for 5 years, a PET/CT and esophagogastroduodenoscopy every year, or at any time when symptoms indicated a recurrence had occurred.
Overall survival (OS) was defined as the length of time from either the date of diagnosis or the start of treatment to death. Disease-free survival (DFS) was defined as the diagnosis of a locoregional recurrence, including lymph node metastasis or distant metastasis at any site. The closing date for survival data collection was December 31, 2018, which was 2 years after the surgery of the most recently enrolled patients in our cohort.

Statistical analysis.
Pearson's product-moment correlation coefficient was calculated for analyzing the relationship between the SUV max of the primary eSCC and the pathologic T and N stages, the presence of nodal metastasis, or primary tumor histologic differentiation.
Five-year OS and DFS were calculated and plotted using the Kaplan-Meier method. Differences in survival among the three (T1a, T1b, and T2 stages) groups were assessed by the three staging methods (pathologic, clinical [EUS], and hypothetical PET by using the SUV max ) were compared using the log-rank test. Multivariate Cox proportional hazard models for OS and DFS were built for those prognostic factors with a p-value of <0.1 in the univariate analysis. For detecting multi-co-linearity among the prognostic factors used for the multivariate analysis, VIFs were calculated.
To evaluate the predictive value for survival (prognostication performance) among the three staging systems, including pathologic T staging, clinical staging using EUS, and hypothetical PET staging, we used an integrated time-dependent AUC (iAUC) calculated from time 0 to 60 months after surgery 19 .
For the validation of SUV max as a T descriptor, we calculated the sensitivity, specificity, accuracy, PPV, and NPV, and compared the results with those of the experimental study. To see whether there were demographic differences between the experimental and validation groups, the data were analyzed using the chi-square test.
All statistical analyses were done using SPSS (SPSS for Windows, version 22.0; SPSS, Chicago, IL) and the statistical computing language R (version 3.4.3, R Foundation). A p-value of <0.05 was considered statistically significant.

Data availability
The datasets generated during and/or analyzed in the current study are available from the corresponding author on reasonable request.