18F-FDG PET/MRI for restaging esophageal cancer after neoadjuvant chemoradiotherapy

Purpose The purpose of this study was to investigate whether 18F-fluorodeoxyglucose (18F-FDG) PET/MRI may potentially improve tumor detection after neoadjuvant chemoradiotherapy (nCRT) for esophageal cancer. Methods This was a prospective, single-center feasibility study. At 6–12 weeks after nCRT, patients underwent standard 18F-FDG PET/computed tomography (CT) followed by PET/MRI, and completed a questionnaire to evaluate burden. Two teams of readers either assessed the 18F-FDG PET/CT or the 18F-FDG PET/MRI first; the other scan was assessed 1 month later. Maximum standardized uptake value corrected for lean body mass (SULmax) and mean apparent diffusion coefficient (ADCmean) were measured at the primary tumor location. Histopathology of the surgical resection specimen served as the reference standard for diagnostic accuracy calculations. When patients had a clinically complete response and continued active surveillance, response evaluations until 9 months after nCRT served as a proxy for ypT and ypN (i.e. ‘ycT’ and ‘ycN’). Results In the 21 included patients [median age 70 (IQR 62–75), 16 males], disease recurrence was found in the primary tumor in 14 (67%) patients (of whom one ypM+, detected on both scans) and in locoregional lymph nodes in six patients (29%). Accuracy (team 1/team 2) to detect yp/ycT+ with 18F-FDG PET/MRI vs. 18F-FDG PET/CT was 38/57% vs. 76/61%. For ypN+, accuracy was 63/53% vs. 63/42%, resp. Neither SULmax (both scans) nor ADCmean were discriminatory for yp/ycT+ . Fourteen of 21 (67%) patients were willing to undergo a similar 18F-FDG PET/MRI examination in the future. Conclusion 18F-FDG PET/MRI currently performs comparably to 18F-FDG PET/CT. Improvements in the scanning protocol, increasing reader experience and performing serial scans might contribute to enhancing the accuracy of tumor detection after nCRT using 18F-FDG PET/MRI. Trial registration Netherlands Trial Register NL9352.


Introduction
Neoadjuvant chemoradiotherapy (nCRT) followed by esophagectomy is a standard treatment for locally advanced esophageal cancer.After nCRT, one-third of patients have a pathologically complete response, opening the way for active surveillance [1].The safety and efficacy of active surveillance are currently investigated in two clinical trials [2,3].Patients in active surveillance undergo clinical response evaluations (CREs) [2].Surgery is performed only when locoregional residual disease is detected in the absence of distant metastases.18F-fluorodeoxyglucose ( 18 F-FDG) PET/computed tomography (CT) has been shown to detect distant metastases before esophagectomy in approximately 10% of patients [4].Within CREs during active surveillance, 18 F-FDG PET/CT also guides the detection of suspected lymph nodes using endoscopic ultrasound with targeted fine-needle aspiration.For the detection of local residual tumors in the esophagus, however, 18 F-FDG PET/ CT has been shown inaccurate [4].A high rate of false Supplemental Digital Content is available for this article.Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's website, www.nuclearmedicinecomm.com.
This is an open access article distributed under the Creative Commons Attribution License 4.0 (CCBY), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
positives was observed because 18 F-FDG cannot reliably discriminate residual tumors from postradiotherapy esophagitis.
MRI is hypothesized to enhance primary tumor detection after nCRT.This is primarily due to higher soft tissue contrast compared to CT and also to the potential to differentiate postradiation inflammation from tumor using the apparent diffusion coefficient (ADC) on diffusion-weighted imaging (DWI) [5,6].Integrated PET/ MRI is a relatively new imaging technique that has the advantage of perfect alignment of PET and MR images.Earlier studies have shown no significant difference between 18 F-FDG PET/MRI and 18 F-FDG PET/CT in the pretreatment staging of esophageal cancer [7][8][9].The feasibility of 18 F-FDG PET/MRI after nCRT has not yet been studied.We hypothesize that in the setting after nCRT, 18 F-FDG PET/MRI might be helpful to distinguish residual tumor from inflammation as well as to detect and characterize new (small) metastatic lesions [8].In the current study, the aim was to evaluate whether 18 F-FDG PET/MRI is feasible to detect residual tumor after nCRT.

Study design
This was a single-center, prospective observational feasibility study.The study was registered on the International Clinical Trials Registry Platform (NL9352) and approved by the Medical Ethical Committee of the Erasmus MC (MEC-2020-0784).All patients provided written informed consent.

Patients
Eligible patients were diagnosed with an adenocarcinoma or squamous cell carcinoma of the esophagus or esophagogastric junction, located at or below the carina.This region was chosen to obtain a homogenous cohort for which the scanning protocol was optimized.All patients underwent five weekly cycles of carboplatin/paclitaxel with concurrent 41.1 Gy of radiotherapy [1].Patients were consecutively identified between February 2021 and May 2022 from the Surgery As Needed for Oesophageal Cancer (SANO)-2 study, an extension study of the SANO trial [10].Exclusion criteria were contra-indications for MRI and an 18 F-FDG nonavid tumor at diagnosis.

Active surveillance
As part of the SANO-2 study, patients underwent the first CRE-1 at 4-6 weeks after nCRT [2].When a residual tumor was detected or highly suspected using endoscopy with bite-on-bite biopsies, patients underwent 18 F-FDG PET/CT to exclude distant metastases.If residual tumor was not identified at CRE-1, CRE-2 was scheduled after 4-6 weeks.CRE-2 included 18 F-FDG PET/CT, endoscopy with bite-on-bite biopsies, and endoscopic ultrasound with fine-needle aspiration.In patients with clinically complete response, subsequent CREs were scheduled every 3 months in the first year, with intervals becoming longer until 5 years after nCRT [2].

Study procedures
The 18 F-FDG PET/CT at CRE-1/-2 was complemented with a contrast-enhanced CT scan (to have the highest-quality 18 F-FDG PET/CT available for study assessments) and a PET/MRI acquisition (Fig. 1).Before PET/ MRI was performed, patients had a short break to eat and drink.During this break and directly after the 18 F-FDG PET/MRI, patients completed a self-constructed questionnaire to evaluate the burden of undergoing these scans, based on a similar study in esophageal cancer patients [11].

Scanning protocols
Patients underwent a whole-body PET/low dose CT scan 60 ± 5 min after injection of 18 F-FDG [median 2.7 MBq/kg; interquartile range (IQR) 2.2-2.9MBq/kg], according to the guidelines of the European Association of Nuclear Medicine v.2.0 [12], as implemented in the institutional protocols.Scans were performed on a 40-or 128-slice Siemens BioGraph PET/CT system (Siemens Medical Systems, Erlangen, Germany).After PET/low dose CT acquisition at 3 min/bed, a contrast-enhanced CT of the neck, chest and abdomen was acquired on the same scanner without repositioning the patient, according to standard clinical protocol.
Directly following the 18 F-FDG PET/CT scan, a nonenhanced PET/MRI was performed on an integrated 3.0 Tesla PET/MRI whole-body system (Signa PET/ MR, GE Healthcare, Waukesha, Wisconsin, USA).MRI sequences were acquired simultaneously with the PET bed positions and comprised sequences for wholebody, dedicated esophagus and dedicated liver imaging.Full protocol details are listed in Supplementary Table 1, Supplemental digital content 1, http://links.lww.com/NMC/A270.Briefly, the whole-body MRI sequences comprised axial T1-weighted (T1w) liver acquisition with volume acceleration Flex, axial T2w fast recovery fast spin echo Flex, and axial DWI single-shot echo planar imaging (b-values: 50, 800 s/mm 2 ) sequence.For the primary tumor location, additionally, an axial T2w periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) and an axial field-of-view optimized and constrained undistorted single-shot (FOCUS) DWI (b-values: 50, 200, 800 s/mm 2 ) sequences were obtained.For each of the 6-7 bed positions, a default zero echo time-(headonly) and Dixon-based sequence was performed to calculate attenuation correction maps for PET image reconstruction.ADC maps were calculated on the scanner console using the acquired b-values and a mono-exponential fit.

Study endpoints
The following criteria were defined to determine the feasibility of 18 F-FDG PET/MRI in the restaging of esophageal cancer after nCRT: 1. the concordance between 18 F-FDG PET/MRI and the reference standard (see below); 2. the possibility to perform quantitative measurements, including the interobserver variability and the concordance with the reference standard; 3. the burden for the patient of undergoing 18 F-FDG PET/MRI.

Reference standard
Histology of the resection specimen served as the reference standard in patients who underwent surgery.Resection specimens were assessed using the tumor regression grade (TRG): TRG 1, 0% residual tumor cells; TRG 2, 1-10%; TRG 3, 11-50%; TRG 4 > 50% [13].In patients without surgery, subsequent CREs during active surveillance until 9 months after nCRT served as a proxy for ypT and ypN.For example, when patients had a persistent clinically complete response until 9 months after nCRT, the reference standard at the time of scanning was considered 'ycT0N0' (i.e.postneoadjuvant clinical staging) [14].The cutoff of 9 month was chosen since we expected that in patients with a residual tumor at the time of the study scan, while undetected at that moment, this would become apparent after two subsequent CREs (i.e.timed at 6 and 9 months after nCRT).

Sample size
A formal sample size calculation was not performed because this was a feasibility study.A sample of at least 20 patients was considered sufficient for an indication of parameters for diagnostic accuracy.

Qualitative assessments
Two teams of readers assessed the scans qualitatively and quantitatively using VUE Carestream (Carestream Health, Rochester, New York, USA).Each team included two members: one radiologist with expertise in MRI and one nuclear medicine physician with expertise in PET (all with >10 years of experience).A random sequence was generated to determine whether the 18 F-FDG PET/CT or the 18 F-FDG PET/MRI was assessed first.The other scan was assessed 1 month later, to prevent recall bias as much as possible.The evaluation of pretreatment imaging was allowed during scoring, but readers were blinded from all other clinical and outcome data.Qualitative assessments were performed using European Association of Nuclear Medicine Research Ltd (EARL)-1 reconstruction of the PET/CT and the Q.Clear 300 or 150 of the PET/ MRI.
First, team members independently assessed the scans, allowing them to study the scans before the subsequent consensus meeting per team.Independent assessments included confidence scores (CS) for the presence of residual tumor, tumor-involved lymph nodes and distant metastases [4,6]: 1 = benign; 2 = probably benign; 3 = equivocal; 4 = probably malignant; 5 = malignant.During team consensus, the two members assessed the scan together.They reconsidered and discussed their independent scores to generate an integrated conclusion regarding the presence of residual tumor in the esophagus, lymph nodes, and distant metastases: 'benign' (CS1-2); 'equivocal' (CS3); 'malignant' (CS4-5).Furthermore, the quality of the CT of the 18 F-FDG PET/CT and MRI of the 18 F-FDG PET/MRI was scored per team as either 'good', 'artifacts, but sufficient', or 'poor'.

Quantitative measurements
EARL-1 PET reconstructions were used for both scans to measure the maximum standardized uptake value corrected for lean body mass (SUL max ) at the primary tumor bed [4].During team consensus meetings, the mean apparent diffusion coefficient (ADC mean ) was measured on the ADC map corresponding to the DWI FOCUS b = 800 s/mm 2 .An oval or free-form shape was delineated in the axial plane, covering at least all hyperintense parts.In the absence of such a signal, an area representative of the primary tumor location was delineated, using the pretreatment 18 F-FDG PET/CT scan for reference.
Five-point Likert-scale items of the questionnaire were described with mean and SD, using the Wilcoxon signedranks test for pair-wise comparisons.Other items were described with numbers and percentages.The burden was predefined in the study protocol as acceptable when ≥60% of patients were neutral or willing to undergo another 18 F-FDG PET/MRI scan.
The interobserver agreement between teams for qualitative assessments was reported using the percentage exact  Quantitative measurements were compared using a Student's t-test for normally distributed data and otherwise a Mann-Whitney U test.Bland-Altman analysis was performed to compare measurements of the same patients on 18 F-FDG PET/CT vs. 18 F-FDG PET/MRI.
Statistical analysis was performed using R version 4.0.4(www.r-project.org).The code can be accessed via github.com/mjvalkema/PRIMERO.

Patients
Twenty-one patients were included (Fig. 2).Clinicopathological characteristics are presented in Table 1.Fourteen of 21 patients (67%) had yp/ycT+.Twelve of 14 underwent surgery.The other two patients had highly suspected disease (i.e.ycT+), but did not undergo surgery; one had an interval bone metastasis, and one refused surgery.In seven of 21 patients (33%), no residual cancer was detected during CREs until 9 months after nCRT (ycT0).One of these seven patients nevertheless underwent surgery, because a solitary lymph node metastasis (Fig. 3) and high-grade  F-FDG PET/T2w FrFSE Flex.The 18 F-FDG uptake in the lymph node at level 2R is more pronounced on the 18 F-FDG PET/MRI than on the 18 F-FDG PET/CT since it is acquired at a prolonged interval after 18 F-FDG injection.Subtle characterization of the lymph node on the MRI was however, hindered by lung motion artifacts.Three months later, a further increase in 18 F-FDG uptake was seen in this node and the patient underwent surgery, which confirmed this lymph node metastasis (ypT0N1).FrFSE, fast recovery fast spin echo; PROPELLER, periodically rotated overlapping parallel lines with enhanced reconstruction.
dysplasia were detected at 6 months after nCRT.The resection specimen confirmed the lymph node metastasis, but high-grade dysplasia or residual tumor was absent (TRG 1, ypT0N1).
Six of 21 (29%) patients had ypN+.Thirteen of 21 (62%) had ypN0 (n = 7) or ycN0 (n = 6).For the other two of 21 patients, the ypN stage remained unknown since these patients did not undergo surgery because of resp.distant metastasis and refusal.These two patients were left out of the analysis for yp/ycN+ detection.

Scanning parameters
Twenty patients completed both 18 F-FDG PET/CT and 18 F-FDG PET/MRI.One patient decided to quit during the 18 F-FDG PET/MRI scan, and a DWI was missing from the esophagus region.Because a sufficient-quality T2-weighted PROPELLER was acquired, the patient was included in the analysis.Median glucose levels were 5.7 (IQR 5.4-6.4).Scans were performed at a median of 11.9 weeks after nCRT (IQR 11.6-12.1).Median scan duration was 32 min (IQR 30-35) for 18 F-FDG PET/CT and 61 min (IQR 56-64) for 18 F-FDG PET/MRI.The interval between 18 F-FDG injection and scanning was 61 min (IQR 56-63) for 18 F-FDG PET/CT and 124 min (IQR 121-130) for 18 F-FDG PET/MRI.No adverse events occurred.

Qualitative assessments
The image quality of the scans was scored at least sufficient for all 18 F-FDG PET/CT scans and for 18 of 21 (86%) 18 F-FDG PET/MRI scans (Supplementary Table 2, Supplemental digital content 1, http://links.lww.com/NMC/A270).Some examples of the image quality of the two techniques are shown in Fig. 3.
For teams 1 and 2 resp., a sensitivity for yp/ycT+ detection with 18 F-FDG PET/MRI was achieved of 36% and 78%, and the specificity of 43% and 14%, respectively.The sensitivity for ypN+ was 17% and 33%, and the specificity was 85% and 62%, respectively.The one-interval bone metastasis was detected on both scans.All diagnostic accuracy parameters of 18 F-FDG PET/MRI and 18 F-FDG PET/CT are presented in Tables 2 and 3. Examples of patients correctly and incorrectly assessed by both teams are shown in Figs.4-7.
Quantitative measurements SUL max measurements on both 18 F-FDG PET/CT and 18 F-FDG PET/MRI were not discriminative for locoregional tumor; neither was ADC mean (Table 4).Bland-Altman analysis showed good agreement between 18 F-FDG PET/CT and 18 F-FDG PET/MRI regarding SUL max (Supplementary Figure 1, Supplemental digital content 1, http://links.lww.com/NMC/A270).The intraclass coefficient for ADC mean between the two teams was 0.27, indicating poor agreement.

Patient burden
All patients completed questionnaires (Supplementary Table 3, Supplemental digital content 1, http://links.lww.com/NMC/A270).Fourteen of 21 patients (67%) were neutral or willing to undergo a similar 18 F-FDG PET/MRI examination in the future. 18F-FDG PET/MRI was less comfortable, and patients had more anxiety.The most stressful aspects of 18 F-FDG PET/MRI were scan duration (7 of 21, 33%) and the noise of the scanner (6 of 21, 29%).Sixteen of 21 patients (76%) experienced the additional 18 F-FDG PET/MRI as not (so) unpleasant.

Discussion
This feasibility study shows that the diagnostic performance of 18 F-FDG PET/MRI at 12 weeks after nCRT appears comparable to 18 F-FDG PET/CT for the detection of locoregional residual tumor.Therefore, an added value to improve clinical response evaluations was not yet demonstrated in this first exploration of 18 F-FDG PET/MRI in the post-treatment setting.
Table 2 Qualitative assessments vs. reference standard for detecting primary tumor in the esophagus Data are median (95% confidence interval).NPV, negative predictive value; PPV, positive predictive value.
Table 3 Qualitative assessments vs. reference standard for detecting locoregional lymph node metastases Data are median (95% confidence interval).NPV, negative predictive value; PPV, positive predictive value.
To our best knowledge, this is the first study that prospectively investigated the value of 18 F-FDG PET/MRI for esophageal cancer response evaluation after nCRT.The DWI did not appear to provide complementary value regarding the discrimination of residual tumor vs. inflammation in the present study cohort.Although DWI may show diffusion restriction in substantial tumor masses, diffusion restriction may be less clearly observed in small residual tumor volumes or in post-treatment necrotic tumor masses [15].As illustrated in Figs.6-7, a single 18 F-FDG PET/MRI scan will not always provide clear guidance regarding the response after nCRT and, as such, the decision to proceed to surgery or continue active surveillance. 18F-FDG PET/MRI might be more suitable for assessing larger tumor volumes, for example, in the pretreatment staging and delineation of gross tumor volumes [7][8][9].
A striking finding was that the performance of the two teams of readers was relatively similar regarding the detection of ypN+ using 18 F-FDG PET/MRI.The assessment of yp/ycT+ only reached a fair agreement.
In retrospect, team 1 seems to have considered the MRI component as a leading factor in their integrated conclusion (data not shown): when 18 F-FDG uptake was considered probably malignant or equivocal, but diffusion restriction was visually absent, postradiotherapy inflammation was considered more likely than the presence of a residual tumor.On the contrary, team 2 applied the DWI more in support of their integrated conclusion rather than considering it as leading.This phenomenon was not the case for ypN+ detection because diffusion restriction is not indicative of lymph node metastases and was not used as such by both teams.With the application of a new technique in esophageal cancer restaging, a learning curve may play a role.The obvious advantage of 18 F-FDG PET/MRI is the perfect alignment between the PET and MR images, which aids the characterization of 18 F-FDG PET-avid lesions.The methodology for the interpretation of a residual tumor, however, requires further investigation.
Overall, the present study indicates a sensitivity for yp/ycT+ of 36/78% (team 1/team 2) with 18 F-FDG PET/MRI, and a specificity of 43/14%, compared with 86/86% sensitivity and 57/14% specificity with 18 F-FDG PET/CT.In contrast, another study demonstrated better sensitivity and specificity for ypT+ with DWI than with 18 F-FDG PET/CT (sensitivity reader 1/ reader 2 : 92/96% vs. 69/62%, resp.; specificity: 57/43% vs. 43/43%, resp.) after nCRT [6].MRI has thus been shown to achieve high sensitivity at the cost of low specificity [16].Notably, in the present study, an integrated assessment of 18 F-FDG PET/MRI was performed instead of assessing MRI alone.This might partially explain the discrepancy between results because the interpretation of the 18 F-FDG PET might have influenced the integrated conclusion of the scan.In addition, with a different implementation of the DWI of another scanner, a higher-quality image might be achieved in terms of less artifacts or distortions.Additional data on the value of MRI post-treatment is expected, because two studies are currently investigating ( 18 F-FDG PET/) MRI in patients treated with nCRT [17,18].
An important strength of our study was the prospective design, allowing similar scanning protocols and a similar follow-up protocol in all patients.Furthermore, between scans, patients were allowed to eat and drink.This contributed to the toleration of another hour of scanning.Furthermore, two teams of readers were involved in Patient with cTxN0M0 adenocarcinoma, who underwent surgery for a positive lymph node (ypT0N1).Both teams correctly classified the primary tumor area (a-d; arrows) as benign on 18 F-FDG PET/MRI.Some hyperintense signal in the esophageal wall was observed on the T2-weighted PROPELLER image (a).This was considered more likely to be reactive than suggestive for residual tumor: linear, 18 F-FDG uptake was observed (b), without clear diffusion restriction (c,d).PROPELLER, periodically rotated overlapping parallel lines with enhanced reconstruction.
reading the scans.This was advantageous for assessing intra-and inter-team comparisons.Some limitations should be taken into consideration.The sample size was limited, which fit the purpose of a feasibility study.As indicated by wide confidence intervals, the parameters for diagnostic accuracy thus only provide a gross estimation of the expected diagnostic performance of 18 F-FDG PET/MRI.Moreover, the reference standard was not available at the same time point for all patients.A proxy until 9-month follow-up had to be defined in patients without surgery.However, such a choice was unavoidable in an active surveillance setting.Moreover, 18 F-FDG PET/CT was performed first in all patients to retain it as the standard of care.This may have introduced bias regarding the perceived burden, which was more favorable with 18 F-FDG PET/CT.The standard longer uptake time for 18 F-FDG PET/MRI also affected distribution.
Different tumor/background ratios might have impacted assessments, although it is unclear to what extent.Finally, baseline 18 F-FDG PET/MRI and dynamic contrast-enhanced (DCE) MRI were not performed in this study.Therefore, it was not possible to exploit delta-ADC and dynamic contrast-enhanced parameters.
This was our first experience with post-treatment 18 F-FDG PET/MRI, and several aspects could be considered in future studies.Scanning the mediastinum with 18 F-FDG PET/MRI remains challenging, and the scanning protocol needs to be optimized to reduce artifacts.The addition of filling liquid in the esophageal lumen might be considered to optimize tissue contrast.The possibility of serial scanning, exploring delta-MRI features, might provide more guidance in assessing response during active surveillance [19].Eventually, the best improvement in specificity might be obtained with another, more tumor-selective tracer.Fibroblast activation protein inhibitor may be a candidate, as well as other radiotracers in the future, which needs to be confirmed in new studies [20].

Conclusion
The current study indicates that the novel 18 F-FDG PET/ MRI and the present standard 18 F-FDG PET/CT show similar performance in terms of accuracy of locoregional esophageal cancer detection after nCRT.Improvements in technique, reader experience and application of serial scanning offer potential sources for improvement of the performance of 18 F-FDG PET/MRI.  to thank S.J. Baart for her statistical advice during the design of the study.
Data generated or analyzed during the study are available from the corresponding author on request.
All procedures performed in the patients described herein were in accordance with the principles of the 1964 Declaration of Helsinki and its later amendments or comparable standards.The study was approved by the Medical Ethical Committee of the Erasmus MC (MEC-2020-0784).Written informed consent for participation and for publication was obtained from all patients in this study.

Table 1
Patient and tumor characteristics [14]inuous data are median (interquartile range).Categorical data are numbers (percentages).Staging was performed according to the AJCC cancer staging manual, 8th edition[14]a Only for patients undergoing resection

Table 4
Quantitative assessments of the primary tumor location *P value calculated with Mann-Whitney U test.