18FDG PET/CT-derived parameters successfully predict clinical stage and prognosis of esophageal cancer.

Although 18FDG PET/CT is validated in baseline workup of esophageal cancer to detect distant metastases, it remains underused in assessing local staging and biology of the primary tumor. This study aimed to evaluate the association between 18FDG PET/CT-derived parameters of esophageal cancer, and its clinico-pathological features and prognosis. All patients with esophageal adenocarcinoma or squamous cell cancer of operated between 2005-2014 were analyzed. Linear regression was used to identify clinicopathologic features of esophageal cancer associated with the tumor’s maximal Standardized Uptake Value (SUVmax), Total Lesion Glycolysis (TLG) and Metabolic Tumor Volume (MTV). ROC curve analysis was performed to precise the optimal cutoff of each variable associated with a locally advanced (cT3/4) status, long-term survival and recurrence. Kaplan Meier curves and Cox regression were used for survival analyses.

accuracy. A SUVmax > 12.7 g/mL may herald early tumor recurrence and poor diseasefree survival.

Background
Esophageal cancer is associated with aggressive lymphatic spread, resulting often in locally advanced or metastatic disease upon diagnosis [1]. Metabolic imaging with 18-Fluorodeoxyglucose Positron Emission Tomography/Computerized Tomography ( 18 FDG-PET/CT) has been integrated into the preoperative workup of esophageal cancer for the detection of distant suspicions lesions [1][2][3], interval metastases [4] or assessment of response to neoadjuvant treatment [5]. Esophageal cancer workup should use threemodality staging with Computerized Tomography (CT), endoscopic ultrasound (EUS) and 18 FDG PET/CT [2,3], as failure to identify locally advanced tumors (cT3/4 or N+ ) may lead to omission of neoadjuvant treatment before surgery, compromising patient survival [2,6].
To this day, 18 [7]. Obtaining accurate cTN staging information through 18 FDG PET/CT may be of prime importance particularly where EUS is unavailable, either by lack of an expert operator, or if the tumor is obstructive (up to 19% of patients) [7].
The aim of our study was to assess the association of several clinico-pathological characteristics of esophageal cancer and the 18 FDG-PET/CT derived parameters SUV m ax , TLG and MTV, as well as to evaluate the predictive value of these parameters in survival and tumor recurrence.

Methods
All patients operated for esophageal adenocarcinoma or squamous cell cancer, with curative intent, from 2005-2014 in our tertiary referral center and a baseline 18 FDG-PET/CT in the preoperative workup were included in this study. Demographic, clinical and histological data were retrieved from our prospectively maintained database.
In all patients, routine preoperative staging was performed by esophagogastroduodenoscopy, EUS and thoraco-abdominal CT scan. Since 2005 18 FDG-PET/CT was integrated in the baseline preoperative workup, according to current recommendations [2]. TNM stage was defined according to the 7 th TNM classification [8].
Neoadjuvant treatment was administered for locally advanced lesions (cT3/4 and/or N+), with 5FU-platin or carboplatin-paclitaxel based chemotherapy and external beam radiation of 41-54 Gy. R0 resection was defined as the presence of tumor within 1mm of resection margins. Postoperative follow-up included a thoraco-abdominal CT scan every 4 months for the first two postoperative years and further workup in cases of suspected recurrence [9]. Early recurrence was defined as any documented recurrence in the first 12 postoperative months. Follow-up data were last updated in November 2018, to assure a minimum follow-up of 4 years for all patients.

Baseline 18 FDG-PET/CT and derived parameters
Since the beginning of this study, we introduced our own PET/quality control program used in several national and international PET studies [10] until our center participated to the quality control program by EANM EARL as PET/CT Center of Excellence in October 2011, for which we have been accredited each year so far. For 18 FDG-PET/CT, patients fasted for at least 6 hours before and blood glucose was measured before administration of the radiotracer (<8.5 mmol/L). Each patient received 3.5 MBq/kg of 18 F-FDG intravenously and remained in a calm and warm area for 1 hour. Thereafter, the patient was asked to void and subsequently was placed in the scanner. Images were acquired on PET/CT scanner (Discovery LS before 09/2011 and then Discovery D690 TOF; GE Healthcare, Waukesha, WI) with scatter and point-spread function recovery corrections. The CT scan (140 kV, 80-200 AutomA/SmartmA) was used for attenuation correction. The CT scan was followed by a PET over the same body region (2min/bed position). Two nuclear medicine physicians closely reviewed the images using for analysis an Advantage Workstation (version 4.6, GE Healthcare, Waukesha, WI) using PET VCAR to compute SUV m ean , SUV m ax , TLG and MTV.
SUV m ax was defined as the point of maximal radiotracer uptake value within the delineated tumor volume (g/mL). MTV represents the metabolically active tumor volume (cm 3 ), whereas TLG was computed as the product of MTV multiplied by the tumor's SUV m ean . In order to define the contouring margin of primary tumor, a volume of interest around the tumor was drawn carefully to incorporate the target lesion in transaxial, sagittal and coronal planes. For tumor delineation we used a 42% threshold, as it is one of the most commonly used in the literature [7,11].

Statistical analysis
Linear regression was performed to assess correlation between several clinicopathological variables and baseline SUV m ax , TLG and MTV. For each PET-CT derived parameter, a ROC curve analysis was performed to assess whether an optimal cutoff could be associated with locally advanced lesions (cT3/4), overall survival and early tumor recurrence. Overall and disease-free survival were analyzed with the Kaplan-Meier method and log-rank test as well as and a Cox regression analysis. Exploratory survival analyses were carried out for each histological type (adenocarcinoma and squamous cell). Co-variates with a p-value<0.2 on a univariate level were entered to a backward elimination process, allowing to build the final multivariate model with the lowest AIC value. Significance level was set at p<0.05 and all tests were two-sided. Statistical analysis was performed with RStudio

Results
From the 141 patients operated in the study period, 89 had a baseline 18 FDG-PET/CT in their workup (63%). Three of them were excluded from analysis because of histology other than adenocarcinoma or squamous cell; thus, the current series consists of 86 patients.

Baseline 18 FDG-PET/CT -derived parameters and initial tumor staging a. SUV m ax ( maximal Standardized Uptake Value)
Median baseline SUV m ax was 12.1 g/mL (range 2.8-48.0) for all tumors. Middle third tumor location, advanced cT and cN stage as well as squamous cell histology were associated with higher SUV m ax values on a univariate level, however only tumor location and cT stage remained significant on multiple regression ( Table 2). cT3/4 tumors had an expected SUV m ax 6.61 higher than a cT1-2 lesion (β coefficient 6.61, 95%CI 2.40, 10.81, p=0.002), and middle third tumors an expected SUV m ax 7.01 higher than GEJ lesions (β coefficient 7.01, 95%CI 0.71-13.32, p=0.029). The multivariable model presented a good fit to the data (R 2 = 0.2804, F-statistic 4.676 on 6 and 72 DF, p<0.0001).
Baseline SUV m ax presented a good prognostic value of a cT3/4 status in ROC curve analysis (Figure 1a). A SUV m ax of 8.25g/mL predicted a cT3/4 lesion with a sensitivity of 83.9% and a specificity of 68.4%. Overall accuracy as indicated by the area under the curve was 81.6% (AUC=0.816, 95%CI=0.704-0.928, p<0.001).

c) MTV (Metabolic Tumor Volume)
Median MTV for all FDG-avid tumors was 22.7 cm 3 (range 1-519). Univariate analysis identified only active smoking being associated with higher baseline MTV (β coefficient 32.81, 95%CI 4.99-70.62, p=0.093) and thus, no multivariable analysis was possible for this parameter.

Prognostic value of 18 FDG-PET/CT -derived parameters for recurrence and patient survival
Among the three parameters studied, SUV m ax at baseline was the only one with a significant predictive value for early tumor recurrence (Figure 2). A SUV m ax ≥12.7g/mL predicted early recurrence with 70.4% sensitivity and 64.6% specificity (AUC 0.660, 95%  (Table 4).
When the two histological subtypes were analyzed separately, there was no significant association of SUV m ax with DFS for adenocarcinoma. For squamous cell carcinoma, a baseline SUV m ax ≥12.7g/mL along with pT and pN stage independently predicted worse DFS (Table 4).

Discussion
In this study, higher baseline SUV m ax of esophageal cancer was significantly related to a middle-third tumor location and a cT3/4 stage, whereas higher TLG was related to cT3/4 and cN+ stage. Baseline SUV m ax >8.25 g/mL, TLG> 41.7 and MTV>10.70 cm 3 were associated with cT/4 stage, whereas SUV m ax ≥12.7 g/mL predicted early recurrence and poor disease-free survival.
The value of 18 FDG-PET/CT in preoperative workup of esophageal cancer has been extensively studied since its first report in 1995 [12]. It is mostly used for the detection of distant metastases as it can identify suspicious lesions as small as 1cm [2,13,14]. Walker et al reported 18 FDG-PET/CT -detected distant lesions precluding curative treatment in 21% of patients [14], even though the specificity of an FDG 'hot spot' remains low and may reflect an inflammatory process or even a synchronous neoplastic lesion in up to 9.3% of patients [15]. Limited spatial resolution of PET for esophageal wall layers and adjacent structures had restrained this modality as a detector of distant metastases [14], whereas metabolic information on the primary tumor are largely overlooked. Recent data, however, reinforce the role of 18 FDG-PET/CT in better defining cTNM stage and the tumor's biology, the latter being FDG-avid in 84-92% of cases especially if it infiltrates the submucosa [7,14,16].
Although it is generally admitted that all 18 FDG-PET/CT-derived parameters are higher in advanced and aggressive tumors, no conclusive data exist as to their optimal cutoff, limiting their practical utility on preoperative cT/N staging. Our study used the most methodologically robust method, linear regression, to define whether and which associations exist between the tumor's clinicopathologic characteristics and the 18 [7], obtaining a much lower accuracy for these cutoffs compared to our study.
One might argue that EUS is sufficient to identify locally advanced lesions (cT3/4 or N+) and thus to direct the patient to neoadjuvant treatment before surgery. However, previous data from our institution suggest a rather low rate of accurate usT (51%) and usN (72%) staging, with the highest rates of understaging among active smokers [17]. Indeed, a three-modality workup strategy ( 18 FDG-PET/CT, CT and EUS) offers the highest probability (84%) to correctly select patients for surgery, a fortiori when 18 FDG-PET/CT is the first exam performed [3]. Rather than advocating the superiority of a diagnostic modality over the others they should be used as complementary to improve staging accuracy, directing patients with locally advanced lesions to neoadjuvant treatment and avoiding its unnecessary toxicity for early-stage tumors.
Several studies have reported poor long-term prognosis associated with high baseline SUV m ax [15,[18][19][20][21], although there is great variability in the suggested cutoff, ranging from 3 to 9 g/mL. A universally accepted SUV m ax associated to overall survival cannot be suggested so far, as cutoffs are chosen either arbitrarily, or using the median of each individual series. In the present study none of the 18 FDG-PET/CT derived parameters demonstrated significant association with overall survival. However, ROC curve analysis identified a significant association between baseline SUV m ax >12.7g/mL and early tumor recurrence. This was confirmed in a multivariate Cox regression, whereby SUV m ax >12.7g/mL, pT3/4 status and active smoking were independent predictors of poor DFS.
The independent predictive value of SUV m ax for DFS was separately confirmed in squamous cell cancer, whereas it did not reach significance in adenocarcinoma patients. This study has some limitations that need to be addressed. Retrospective analysis has an inherent drawback in data completeness, even though our institutional database is maintained prospectively, with a stringent follow-up of all patients. Although there was practically no heterogeneity in 18