Abstract
Purpose
We aimed to determine the role of integrated 2-[18F]-fluoro-2-deoxy-glucose (18F-FDG) PET/MRI in preoperative T and N staging and prognosis of esophageal squamous cell carcinoma (ESCC).
Materials and methods
The analysis was conducted on 66 ESCC patients who accepted 18F-FDG-PET/MRI examinations per-operatively. We select the primary lesion as the region of interest to evaluate the diagnostic efficiency of T and N staging. Univariate and stepwise multivariate logistic regression models were performed to determine the T and N staging factors, and we established the optimal prediction model by using the Akaike information criterion. Finally, Cox regression was used for the early recurrence analyses.
Results
The total lesion glycolysis (TLG), metabolic tumor volume (MTV), minimum apparent diffusion coefficient (ADCmin), and mean apparent diffusion coefficient (ADCmean) were significant in univariate analysis with T staging. None of the parameters were significant in the univariate analysis with N staging. Cox regression validated that this model-combining TLG (>47.5), MTV (>9.4 cm3), ADCmin (<1.49 × 10−3 mm2/s), and ADCmean (<1.68 × 10−3 mm2/s)—was the sole predictor significant associated with early tumor recurrence. Notably, no association was found between any single variable and early tumor recurrence.
Conclusion
Integrated 18F-FDG PET/MRI is feasible for preoperative T and N staging of ESCC. The optimal model is composed of TLG, MTV, ADCmin, and ADCmean—all parameters derived from 18F-FDG-PET/MRI—which not only provide valuable information for T staging but also predict early recurrence within the first postoperative year.
Keywords: 2-[18F]-fluoro-2-deoxy-glucose, esophageal squamous cell carcinoma, PET/MRI, prognosis assessment, staging
Introduction
Esophageal cancer (EC) has a high recurrence and low survival rate, and esophageal squamous cell carcinoma (ESCC) accounts for more than 90% of EC cases, particularly in China [1]. Despite several diagnostic modalities and multimodal therapies developed in recent years, the recurrence after curative resection varies between 20 and 50% [2], whereas the 5-year overall survival rate is about 20% [3,4]. To reduce the recurrence and improve survival, a tailored treatment regimen must be provided to individual patients. Therefore, a precise preoperative staging and prognosis assessment is crucial [5]. Preoperative staging is focused on T, which directly guides the oncological and surgical strategy [6]. The N stage is the primary determinant for choosing the appropriate treatment option and may affect the ESCC prognosis [7].
The current practice guidelines for the T/N staging of ESCC included endoscopic ultrasonography (EUS), computed tomography (CT), MRI, and 18F-FDG-PET/CT [8]. EUS is a good tool for assessing the loco-regional stage. However, it is operator-dependent and limited by the risk of esophageal perforation and failure to pass through stenotic esophagi. CT scans have relatively poor diagnostic performance because enlarged lymph nodes may be benign and normal-sized lymph nodes may be malignant [9,10]. Although MRI is highly accurate in determining the T/N stage, it has demonstrated poor agreement between assessing metastatic lymph nodes and pathologically confirmed metastatic lymph nodes [11]. Moreover, there is a lack of reports on the prognostic assessment. Although minimum apparent diffusion coefficient (ADCmin) and mean apparent diffusion coefficient (ADCmean) are often used as MRI quantitative parameters to evaluate the prognosis of EC, the prognosis evaluation is unsatisfactory due to the low resolution of diffusion-weighted imaging (DWI). Giganti et al. [12] studied 18 patients with EC and the results showed that the accuracy of lymph node status by 1.5T MR imaging combined with DWI was only 66%. Ultra-high-field (7.0T) MRI has an advantage in evaluating the depth of mural invasion [13], but it has few clinical applications. 18F-FDG uptake is driven by the increased glucose metabolism of malignant cells and can be measured semiquantitatively using standardized uptake value (SUV) measurements [14]. Moreover, the maximum SUV (SUVmax) of PET/CT helps to identify patients with T-stage and metastatic lymph nodes, although the diagnostic accuracy still remains unsatisfactory [10,15,16].
Recently, 18F-FDG-PET/MRI imaging with multiple MRI sequences and metabolic parameters has been used to diagnose and evaluate tumors [17,18], combining metabolism and anatomy. Previous studies have demonstrated the multiple-parametric advantages of MRI. However, the value of 18F-FDG-PET/MRI-derived metabolic parameters has not been fully exploited in the quantification of T/N staging or been used for prognostic evaluation [19].In the present retrospective study, we aimed to evaluate the value of baseline PET-MRI in preoperative T and N staging in patients with ESCC and whether multiquantitative parameters of PET/MRI can predict early recurrence.
Materials and method
General information
Patients enrollment
We retrospectively screened all patients with a new diagnosis of biopsy-proven EC between October 2018 and October 2020 in Tangdu Hospital (affiliated to Air Force Medical University) for inclusion into this study. The inclusion criteria were as follows: (a) patients underwent radical surgery for ESCC; (b) patients had preoperative 18F-FDG-PET/MR imaging. The exclusion criteria included the following: (a) no surgical intervention (comorbidity, tumor invades adjacent structures, such as the aorta, vertebral body, or trachea, distant metastases, contraindications or lack of consent); (b) diabetes with and poor glycemic control; (c) implanted metallic or electronic devices, allergy history, claustrophobia or inability to lay supine for a longer period of time; (d) a lack of complete clinical data; (e) other pathological tumor types such as adenocarcinoma; or (f) a history of previous malignancy. Ultimaltely, 66 patients were enrolled in the study.
Treatment and follow-up
Patients staged cT1aN0 only received endoscopic resection, cT1b-cT2N0 only received esophagectomy, and cT1b-cT2 N+ or cT3-cT4 any N received preoperative concurrent neoadjuvant chemoradiation therapy + esophagectomy. Patients who underwent esophagectomy and gastric pull-ups underwent two-field or three-field lymphadenectomy.
We evaluated recurrence through regular clinical follow-up of relevant items. Clinical follow-up was performed at 3- to 6-month intervals for the first 2 years after surgery, and subsequent 6- to 12-month intervals. The follow-up content include: (a) carcinoembryonic antigen, squamous cell carcinoma antigen (SCC); (b) Enhanced chest and abdominal CT, PET/CT (for suspected recurrence with negative CT imaging); and (c) Gastroscopy/EUS.
Imaging acquisition
Each patient was required to fast for more than 6h, with no restrictions on drinking water, and blood glucose levels were verified to be less than 11.0 mmol/l before 18F-FDG injection. The scan for treatment position was performed using a 24-channel spine radiofrequency coil integrated 18F-FDG-PET/MRI scanner (Biography mMR, 3.0T, Siemens AG, Germany) 60 min after the intravenous 18F-FDG tracer at 3.7 MBq/kg. The preinjection 18F-FDG radioactivity, measurement time of measurement of preinjection 18F-FDG radioactivity, residual activity after injection, and the measurement time of the residual radioactivity after injection were entered into the scanner during the PET/MRI acquisition [20].
The PET/MRI scan ranges from the upper neck to the mid-thigh by five or six bed positions. For the Dixon-based attenuation correction [21,22], the coronal 3-D volume-interpolated gradient echo (VIBE) sequence was acquired. Diagnostic MR imaging comprised the coronal T2-weighted turbo spin-echo sequence, axial T2-weighted half-Fourier acquisition single-shot turbo spin-echo fat suppression by spectrally adiabatic inversion recovery (SPAIR) technique, and axial 3-D VIBE sequence. In addition, DWI with SPAIR fat suppression was obtained to calculate the ADC map. PET acquisitions were obtained in a 3-D setting, 5 min per bed position, full width at half maximum of 2.0, relative scatter correction, and 2 mm slice thickness. The PET images were reconstructed on a matrix size of 172 × 172 by an iterative algorithm according to the European Association of Nuclear Medicine guidelines [23]. The mean duration of PET/MRI image acquisition was approximately 60 min.
Image analysis
An abdominal radiologist and a nuclear medicine specialist, each with 8 years of experience, completed the PET and MRI images analysis on a PET/MRI workstation (Syngo.via Siemens Healthcare, Germany) and MIM workstation (MIMvista Corp, Cleveland, Ohio, USA).
A volume of interest was manually marked on axial PET images in a MIM workstation to evaluate PET-related image parameters including maximal standardized uptake value (SUVmax), mean standardized uptake value (SUVmean), metabolic tumor volume (MTV), and total lesion glycolysis (TLG). The maximal and mean values of SUV (SUVmax and SUVmean) reflected the maximal SUV (adjusted for body weight) and average SUV within a tumor respectively. MTV was expressed as the tumor volume with FDG uptake, which was segmented using a fixed percentage threshold method at 40% of the SUVmax [24] (Fig. 1), and TLG, which was calculated as MTV multiplied by the average SUV of the included voxels. In the Syngo.via PET/MRI workstation, the slice with the largest tumor area was selected to draw the regions of interest manually. Finally, the SUVmax value of this area exceeded 2.5.
Fig. 1.
18F-FDG PET/MR multiple sequence of ESCC. 18F-FDG PET/MR in a 50-year-old patient with the ESCC (T2N1M0, stage IIIA). (a) The Axial T2W-fs shows that the tumor (white arrow) and lymph node (red arrow) is hyperintense. (c) The b800 DWI-MRI scan and its corresponding ADC-map in (b). Metastatic lymph node adjacent to the lesion is also hypointense (red arrow). (d) The PET/MR fusion image. Coronal PET (e) and axial PET (f) show metabolic tumor volume (light red region, threshold = 40%). ESCC, esophageal squamous cell carcinoma.
Staging of tumor and lymph node
On the SPAIR sequence, the normal esophageal wall showed three signal layers: the mucosa and submucosa with a high signal in the inner layer, the muscularis with a low to moderate signal in the middle, and the outer membrane with a high signal. T-staging was primarily based on the depth through the esophageal wall and involvement of the periesophageal fat or surrounding organs in SPAIR. Based on the 8th edition staging of cancers of the esophagus and esophagogastric junction by the American Joint Committee on Cancer/International Union published in 2017, the diagnostic criteria [19] were: T1-2, muscularis surrounding the lesion had a complete linear low-to-medium signal. T3, low-to-medium signal was interrupted or disappeared in the surrounding muscularis; and T4a: tumor invaded adjacent structures (pleura, pericardium, azygos vein, diaphragm, or peritoneum) and the fat space vanished. It is difficult to distinguish between mucosal invasion and submucosa invasion (T1 and T2) as both the mucosa and submucosa have high signals in SPAIR, so they were pooled.
In terms of lymph staging, we defined nodes with eccentric cortical thickening or obliteration of the fatty hilum or high signal in the DWI sequence. With the SUVmax more than the liver background was considered abnormal, regardless of the size.
Statistical analysis
Statistical analysis was performed with R 3.6.1, an open-source software using the glm package and is available from the Comprehensive R Archive Network [25,26]. Receiver operating characteristic curve (ROC) [27] analysis was used to assess the diagnostic efficacy of 18F-FDG-PET/MRI for T and N staging. Univariate and stepwise multivariate logistic regression models [28] were used to determine the factors associated with T and N staging. These models were constructed by introducing multiparameters from 18F-FDG-PET/MRI and clinical parameters with age as a control variable. In addition, we conducted separate analyses to correlate between T/N staging and multiparameters of PET/MR by Spearman correlation analysis.
Furthermore, Cox proportional hazards regression was used to assess the relationship between early tumor recurrence and the variables. Finally, two-sided P values of less than 0.05 were considered statistically significant.
Establish the optimal prediction model
To further explore the relationship between hazard factors and T/N staging, statistically significant variables in the univariate analysis (P < 0.05) were entered into the multivariate logistic regression analysis. A backward stepwise logistic regression was used to identify possible predictors. This approach begins with a complete model and at each step gradually eliminates variables from the regression model to find a reduced model that best explains the data. At each step, variables were eliminated based on P-values, and the Akaike information criterion (AIC) [29] was used to limit on the total number of variables included in the final optimized model. Finally, we utilized ROC analysis to verify the efficiency of the optimal prediction model (Fig. 2).
Fig. 2.
Data analysis process.
Results
Clinical features
Sixty-six subjects were enrolled in this study. Among them, 55 were males and 11 were females, aged from 42 to 77 years old with an average of 62.9 ± 8.0 years. The time span from the onset of symptoms to surgery ranges from 0.5 to 12 months (2.5 ± 2.3). All patients had ESCC, with well-differentiated (n = 31) and moderately differentiated (n = 29) tumors being the predominant histologic type, and the remainder being poorly differentiated (n = 6). All lesions were located in the thoracic esophagus, and the specific location was as follows: 37 middle, 26 lower, and three upper (Table 1). The distribution of T stages in the 66 patients confirmed via pathology was T1 in 20 patients, T2 in 30 patients, T3 in nine patients, and T4a in seven patients. Concerning the nodal stages, these were N0 in 35 patients, N1 in 13 patients, N2 in 15, and N3 in three patients, classified through their histopathologic findings (Table 1). T1-2, T3, and T4a lesions were correctly staged in 44 (88%, 44/50), six (66.7%, 6/9), and six (85.7%, 6/7) patients by PET/MR, respectively, with a total accuracy of 84.8% (56/66) and, seven overstaged and three understaged. In addition, 27 (77.1%, 27/35) patients with N0, nine (69.2%, 9/13) with N1, 13 (86.7%, 13/15) with N2, and three (100%, 3/3) with N3 were correctly diagnosed with PET/MR imaging. The total accuracy of the N staging was 78.7% (52/66), with nine cases overstaged and five understaged.
Table 1.
Patient characteristics
| Characteristics | Data |
|---|---|
| Age, years | |
| Mean ± SD (range) | 62.9 ± 8.0 (42–77) |
| Gender, n (%) | |
| Male | 55 (83.3) |
| Female | 11 (16.7) |
| Course of the disease, months | |
| Mean ± SD (range) | 2.5 ± 2.3 (0.5–12) |
| SCC (μg/l) | |
| Mean ± SD (range) | 0.60 ± 0.51 (0.29–3.11) |
| Location, n (%) | |
| Upper thoracic esophagus | 3 (4.5) |
| Middle thoracic esophagus | 37 (56.0) |
| Lower thoracic esophagus | 26 (39.3) |
| Histologic grade, n (%) | |
| Well-differentiated | 31 (46.9) |
| Moderately differentiated | 29 (43.9) |
| Poorly differentiated | 6 (9.2) |
| T-stage, n (%) | |
| T1 | 20 (30.4) |
| T2 | 30 (45.4) |
| T3 | 9 (13.6) |
| T4a | 7 (10.6) |
| Regional lymph node, n (%) | |
| N0 | 35 (53.0) |
| N1 | 13 (19.6) |
| N2 | 15 (22.7) |
| N3 | 3 (4.5) |
| Ki67 (%) | |
| Mean ± SD (range) | 35.3 ± 15.6 (10–70) |
18F-FDG-PET/MRI derived parameters and T and N staging
Maximum standardized uptake value and mean standardized uptake value
For all tumors, the median baseline SUVmax and SUVmean were 10.8 (range 2.51–25.0) and 5.46 (range 2.14–8.65), respectively. Advanced T and N stages as well as SCC and Ki67 were associated with higher SUVmax and SUVmean values on a univariate level. However, neither SUVmax and SUVmean were correlated with the T and N stages in univariate analysis, and thus no multivariate analysis was possible for these two parameters (Table 2).
Table 2.
Value of multiparameters
| n = 66 | Normal appearing esophageal | Esophageal lesion | P-value |
|---|---|---|---|
| Mean ± SD (range) | Mean ± SD (range) | ||
| SUVmax | 1.94 ± 0.19 (1.61–2.30) | 11.51 ± 5.63 (2.51–25.0) | 0.000 |
| SUVmean | 1.68 ± 0.24 (0.83–1.98) | 5.41 ± 1.59 (2.14–8.65) | 0.000 |
| TLG | — | 85.87 ± 84.16 (1.18–386.45) | — |
| MTV (cm3) | — | 14.01 ± 11.07 (0.55–46.04) | — |
| ADCmin (10−3 mm2/s) | 1.55 ± 0.36 (0.62–2.31) | 1.22 ± 0.31 (0.85–2.17) | 0.000 |
| ADCmean (10−3 mm2/s) | 1.86 ± 0.40 (0.70–2.59) | 1.35 ± 0.29 (0.98–2.23) | 0.000 |
ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; SUVmax, maximum standardized uptake value; SUVmean, mean standardized uptake value; TLG, total lesion glycolysis.
Total lesion glycolysis
The median baseline TLG was 59.1(range 1.2–386.4) for all tumors. Simple linear regression revealed that TLG remained significant in the univariate analysis with T staging (Table 2, Fig. 3), and the expected TLG for the T3/4a tumors was 88.1 higher than T1/2 tumors (29.9). In addition, ROC curve analysis identified a TLG = 47.5 as the optimal cutoff to detect a T3/4a lesion, with a sensitivity of 86%, specificity of 74%, and overall accuracy of 85% (AUC 0.863, P < 0.001, Fig. 4). However, multivariate regression analysis showed that TLG was not included in the optimal model for predicting T staging. Finally, TLG was not correlated with the N stage in the univariate regression analysis (Table 3).
Fig. 3.
Univariate analysis results. In the univariate analysis, TLG, MTV, ADCmin, and ADCmean were found to be significantly associated with T staging. ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; TLG, total lesion glycolysis.
Fig. 4.
ROC curve analysis results. ROC curve analyses for the predictive value of the 18F-FDG PET/MR parameters in relation to a T3/4a status. The above four parameters predicted significantly the T3/4a status of the primary tumor. TLG>47.5 (sensitivity 86%, specificity 74%, P < 0.001). MTV > 9.4 cm3 (sensitivity 86%, specificity 78%, P < 0.001). ADCmin<1.49 (sensitivity 57.1%, specificity 100%, P < 0.001). ADCmean<1.68 (sensitivity 57.1%, specificity 100%, P < 0.001). ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; ROC, receiver operating characteristic curve; TLG, total lesion glycolysis.
Table 3.
Univariate and multivariate analyses for multiparameters that predict T and N staging in esophageal squamous cell carcinoma
| Variables | T stage | N stage | ||||||
|---|---|---|---|---|---|---|---|---|
| Univariate | Multivariate | Univariate | Multivariate | |||||
| Rs | P value | Rs | P value | Rs | P value | Rs | P value | |
| SCC | 0.220 | 0.251 | N/A | N/A | 0.219 | 0.252 | N/A | N/A |
| Ki67% | 0.225 | 0.238 | N/A | N/A | 0.194 | 0.311 | N/A | N/A |
| SUVmax | 0.299 | 0.115 | N/A | N/A | 0.188 | 0.328 | N/A | N/A |
| SUVmean | 0.312 | 0.099 | N/A | N/A | 0.155 | 0.423 | N/A | N/A |
| MTV | 0.517 | 0.004 | 0.211 | 0.139 | 0.136 | 0.481 | N/A | N/A |
| TLG | 0.492 | 0.007 | N/A | N/A | 0.117 | 0.545 | N/A | N/A |
| ADCmin | −0.700 | 0.000 | −5.044 | 0.106 | −0.270 | 0.150 | N/A | N/A |
| ADCmean | −0.751 | 0.000 | N/A | N/A | −0.270 | 0.160 | N/A | N/A |
ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; SUVmax, maximum standardized uptake value; SUVmean, mean standardized uptake value; TLG, total lesion glycolysis.
Metabolic tumor volume
The median MTV for all FDG-avid tumors was 12.1cm3 (range 0.6–46.0). Simple linear regression revealed that MTV remained significant in univariate analysis with T staging (Table 3, Fig. 3), and the expected MTV for T3/4a tumors was 14.4 cm3 higher than for the T1/2 tumors (3.9 cm3). ROC curve analysis identified a MTV = 9.4 cm3 as the optimal cutoff to detect a T3/4a lesion, with a sensitivity of 86%, specificity of 78% and overall accuracy of 89% (AUC 0.894, P < 0.001, Fig. 4). Meanwhile, multivariate regression analysis revealed that MTV was included in the optimal model for predicting T staging and the model presented a good fit to the data (Rs = 0.211, P = 0.139). Finally, consistent with TLG, MTV was not correlation with the N stage in the univariate regression analysis (Table 3).
Minimum apparent diffusion coefficient and mean apparent diffusion coefficient
The median baseline ADCmin and ADCmean were 1.2 × 10−3 mm2/s (range 0.9–2.1) and 1.3 × 10−3 mm2/s (range 1.0–2.2), respectively. Advanced T and N staging were associated with lower ADCmin and ADCmean values on a univariate level. For the T staging, ADCmin and ADCmean remained significant in the univariate analysis (Fig. 3). ROC curve analysis identified an ADCmin = 1.49 × 10−3 mm2/s as the optimal cutoff to detect a T3/4a lesion, with a sensitivity of 57.1%, a specificity of 100% and an overall accuracy of 82.3% (AUC 0.823, P < 0.001, Fig. 4), and an ADCmean = 1.68 × 10−3 mm2/s with sensitivity of 57.1%, a specificity of 100% and an overall accuracy of 83.9% (AUC 0.839, P < 0.001, Fig. 4). Multivariate regression analysis revealed that ADCmin was also included in the optimal model for predicting T staging (Rs = −5.044, P = 0.106, Table 3), whereas ADCmean was not included. Consistent with the analysis results of the above parameters, ADCmin and ADCmean had no correlation with the N stage in the univariate regression analysis (Table 3).
Optimal prediction model for predicting T staging
Parameters comprising SUVmax, SUVmax, TLG, MTV, ADCmin, ADCmean, and others were defined as variables, and univariate and multivariate analyses were performed with T and N staging. In addition, TLG and MTV were screened based on the statistical significance of T staging (these parameters were not correlated with N staging). Based on the AIC, a backward stepwise variable elimination strategy, with repeated attempt to remove a variable, was utilized to form an optimal prediction model composed of MTV and ADCmin (Fig. 2).
According to the backward stepwise analysis, the optimal prediction model using AIC (AIC = 23.06) demonstrated that MTV and ADCmin for predicting ESCC T staging showed the best predictive performance and the coefficient and P values were 0.211, −5.044 and 0.139, 0.106, respectively (Table 3). In addition, the ROC curve analysis identified 0.873 as the optimal prediction model cutoff with a sensitivity of 73.9% and a specificity of 100% (AUC 0.925, P < 0.001, Fig. 5).
Fig. 5.
ROC curve of the predictive model. ROC curve of a predictive model for T staging. Sensitivity was 73.9%, specificity 100%, and AUC 0.925. ROC, receiver operating characteristic curve.
Prognostic value of 18F-FDG PET/MR-derived parameters for early recurrence
Early recurrence of EC is defined as recurrence occurring within 12 months of curative-intent surgery. The median follow-up in all patients was 20 months (range 12–24 months). Fourteen patients recurrences were noted within 5 months (Fig. 6) whereas others showed no recurrence during follow-up. Cox regression confirmed the model composed of TLG (>47.5), MTV (>9.4 cm3), ADCmin (<1.49 × 10−3 mm2/s) and ADCmean (<1.68 × 10−3 mm2/s) was the only one with a significant predictive value for risk of early tumor recurrence, within the first postoperative year. No association was found between any single variable and early tumor recurrence on Cox analysis (Table 4).
Fig. 6.
Cox regression analysis of early tumor recurrence. Cox regression analysis of early tumor recurrence confirmed the model comprising TLG, MTV, ADCmin, and ADCmean was the only one with a significant predictive value for risk of early tumor recurrence in the first postoperative year. ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; TLG, total lesion glycolysis.
Table 4.
Cox analysis of the relationship between early tumor recurrence and multiple variables
| Univariate model | Multivariate model | ||||
|---|---|---|---|---|---|
| Variable | Coefficient | P value | Variable | Coefficient | P value |
| Gender | 0.747 | 0.425 | MTV | −0.323 | 0.040 |
| Location | 0.500 | 0.803 | TLG | 0.0391 | 0.044 |
| Histologic grade | 1.030 | 0.382 | ADCmin | −8.687 | 0.031 |
| SCC | 1.103 | 0.871 | ADCmean | 9.861 | 0.024 |
| Ki67% | 5.702 | 0.135 | |||
| T stage | 2.270 | 0.101 | |||
| N stage | 3.190 | 0.071 | |||
| MTV | −0.186 | 0.168 | |||
| TLG | 0.018 | 0.294 | |||
| SUVmax | 0.095 | 0.122 | |||
| SUVmean | 0.081 | 0.158 | |||
| ADCmin | 1.352 | 0.107 | |||
| ADCmean | 1.681 | 0.140 |
ADCmean, mean apparent diffusion coefficient; ADCmin, minimum apparent diffusion coefficient; MTV, metabolic tumor volume; SUVmax, maximum standardized uptake value; SUVmean, mean standardized uptake value; TLG, total lesion glycolysis.
Discussion
Our primary endpoint was to assess the role of 18F-FDG-PET/MRI in preoperative T/N staging of ESCC. The results demonstrated that 18F-FDG-PET/MRI for T and N staging had a high accuracy of 84.8 and 78.8% respectively. For T staging, the accuracy in this study was consistent with that reported by Giganti et al. [12] which showed the feasibility of MRI in identifying esophageal wall layers comparable to EUS [30]. We showed the accuracy of T1-2 staging was 88%, and the accuracy of T3 was 66.7% and of T4a was 85.7%. PET/MRI performed relatively poorly in patients with T3 tumors, possibly due to local inflammatory exudation and fibrosis of the esophageal wall, resulting in a reduced definition of the boundary between the lesion and the adjacent normal structures.
Another important finding was that in univariate analysis, MTV and TLG were positively correlated with T staging, whereas ADCmin and ADCmean were negatively correlated with T staging. Furthermore, in multivariate analysis, MTV and ADCmin showed no significant correlation with T staging. However, the optimal model for T staging established by multivariate analysis included MTV and ADCmin, and it had high diagnostic efficacy in T staging with an AUC of 0.925. This may have been due to the small number of cases and we could still quantify T staging by MTV and ADCmin and provide support for subsequent treatment in clinical work.
Lymph node metastasis is an independent risk factor for the prognosis of EC [31]. In general, lymph nodes with an SUVmax of more than the liver background or with eccentric cortical thickening or obliteration of the fatty hilum are considered abnormal, regardless of their size. The present study demonstrated that the accuracy of N staging by 18F-FDG-PET/MRI was 78.8%, higher than that by MRI alone [12,32], showing that the combination of MRI multisequence imaging and PET metabolism information could improve the accuracy of N staging. However, we did not find any PET/MR parameters of primary lesions that could indicate regional lymph node metastasis. Given that relatively few patients were included, the study was underpowered to make a definitive conclusion in this regard, and multiparameter imaging omics and metabolomics will be combined to provide more evidence for the N staging of ESCC in our subsequent studies.
Although high baseline SUVmax may suggest tumor aggressiveness and early recurrence [33,34], according to our study, SUVmax was neither an independent predictor nor associated with early tumor recurrence. The semiquantitative indexes SUV reflect the metabolic characteristics of some cells in the tumor [35], as the volume metabolic parameters, MTV, and TLG can reflect the activity and differentiation rate of the tumor as a whole. Furthermore, the ADC value can indirectly reflect the invasiveness of the tumor through the microscopic movement of tumor cells [36]. It is the complementarity of PET/MRI multiparameters that provides us with more valuable information about preoperative staging and prognosis for ESCC. Our study’s added value lies in the identification of a four-parameter model comprising TLG, MTV, ADCmean, and ADCmin that could be used to evaluate ESCC early recurrence within the first postoperative year. This result is of significant clinical importance, as it might allow for the early identification of patients with resectable ESCC who may not benefit from surgical resection as their risk of early recurrence is significantly increased [34]. In addition, the multivariable result may provide valuable prognostic information for the individual patient and guide therapeutic management.
This study had several limitations. First, it was a retrospective, single-center study and included a relatively small cohort of patients, which may have resulted in type 2 errors and the nonsignificance of some results. By increasing the sample number and performing a multicenter study, our group hope that the statistical association between PET/MR and N staging can be found in future studies. Secondly, the study lacked other histological types such as esophageal adenocarcinoma. A single histological type can be considered as exploratory, necessitating validation in large cohorts with multiple histological types.
Conclusions
Our findings demonstrated that integrated 18F-FDG PET/MRI are feasible for preoperative T/N staging of ESCC. The 18F-FDG-PET/MRI-derived parameters TLG, MTV, ADCmin, and ADCmean provide valuable information for T staging and predicting early recurrence, and based on these, the optimal model composed of these four parameters is initially developed and can be used for the prediction of early recurrence within the first postoperative year. However, the results require further investigation in a larger cohort to determine their validity.
Acknowledgements
The authors would like to thank Prof. Wei Wang and Qiang Li (Tangdu Hospital, Fourth Military Medical University, Xi’an, China) for their valuable input in this study. The authors would also like to thank Prof. Peng Liu (Life Science Research Center, School of Life Science and Technology, Xidian University, Xi’an, China) for editing the figures.
All procedures performed in this study were approved by the Medical Ethics Committee of Tangdu Hospital (No. TDLL-202201-04).
All participants included in this study written informed consent.
The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.
J.H. and W.H. collected the clinical and PET/MRI data. W.Z. and Y.W. conducted the PET/MRI data analyses. J.W. and Y.L. wrote the manuscript and edited it. L.J. and P.Y. contributed to the study design. M.Y. contributed to the improvement of the manuscript.
Conflicts of interest
The authors declare that they have no potential conflict of interest with respect to the research, authorship, or publication of this article.
Footnotes
Yunbo Li and Junyan Wang contributed equally to this work.
References
- 1.Liang H, Fan JH, Qiao YL. Epidemiology, etiology, and prevention of esophageal squamous cell carcinoma in China. Cancer Biol Med 2017; 14:33–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Tanaka K, Yamasaki M, Makino T, Yamashita K, Saitoh T, Takahashi T, et al. Analysis of prognostic factors in patients with lymph node recurrence after radical esophagectomy: importance of locoregional therapy. Esophagus 2021; 18:195–202. [DOI] [PubMed] [Google Scholar]
- 3.Pennathur A, Gibson MK, Jobe BA, Luketich JD. Oesophageal carcinoma. Lancet 2013; 381:400–412. [DOI] [PubMed] [Google Scholar]
- 4.Rustgi AK, El-Serag HB. Esophageal carcinoma. N Engl J Med 2014; 371:2499–2509. [DOI] [PubMed] [Google Scholar]
- 5.Winiker M, Mantziari S, Figueiredo SG, Demartines N, Allemann P, Schäfer M. Accuracy of preoperative staging for a priori resectable esophageal cancer. Dis Esophagus 2018; 31:1–6. [DOI] [PubMed] [Google Scholar]
- 6.Soeno T, Harada H, Hosoda K, Mieno H, Ema A, Ushiku H, et al. Lymph node progression and optimized node dissection of middle thoracic esophageal squamous cell carcinoma in the latest therapeutic surgical strategy. Ann Surg Oncol 2019; 26:996–1004. [DOI] [PubMed] [Google Scholar]
- 7.Zhang X, Jain D. Updates in staging and pathologic evaluation of esophageal carcinoma following neoadjuvant therapy. Ann N Y Acad Sci 2020; 1482:163–176. [DOI] [PubMed] [Google Scholar]
- 8.Varghese TK, Jr, Hofstetter WL, Rizk NP, Low DE, Darling GE, Watson TJ, et al. The Society of Thoracic Surgeons guidelines on the diagnosis and staging of patients with esophageal cancer. Ann Thorac Surg 2013; 96:346–356. [DOI] [PubMed] [Google Scholar]
- 9.Van Vliet EP, Heijenbrok-Kal MH, Hunink MGM, Kuipers EJ, Siersema PD. Staging investigations for oesophageal cancer: a meta-analysis. Br J Cancer 2008; 98:547–557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Yoon YC, Lee KS, Shim YM, Kim B-T, Kim K, Kim TS. Metastasis to regional lymph nodes in patients with esophageal squamous cell carcinoma: CT versus FDG PET for presurgical detection prospective study. Radiology 2003; 227:764–770. [DOI] [PubMed] [Google Scholar]
- 11.Gao Z, Hua B, Ge X, Liu J, Xue L, Zhen F, Luo J. Comparison between size and stage of preoperative tumor defined by preoperative magnetic resonance imaging and postoperative specimens after radical resection of esophageal cancer. Technol Cancer Res Treat 2019; 18:1533033819876263. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Giganti F, Ambrosi A, Petrone MC, Canevari C, Chiari D, Salerno A, et al. Prospective comparison of MR with diffusion-weighted imaging, endoscopic ultrasound, MDCT and positron emission tomography-CT in the pre-operative staging of oesophageal cancer: results from a pilot study. Br J Radiol 2016; 89:20160087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Yamada I, Miyasaka N, Hikishima K, Tokairin Y, Kawano T, Ito E, et al. Ultra-high-resolution MR imaging of esophageal carcinoma at ultra-high field strength (7.0T) ex vivo: correlation with histopathologic findings. Magn Reson Imaging 2015; 33:413–419. [DOI] [PubMed] [Google Scholar]
- 14.Bose S, Le A. Glucose Metabolism in Cancer. Adv Exp Med Biol 2018; 1063:3–12. [DOI] [PubMed] [Google Scholar]
- 15.Noble F, Bailey D, Tung K, Byrne JP; SWCIS Upper Gastrointestinal Tumour Panel. Impact of integrated PET/CT in the staging of oesophageal cancer: a UK population-based cohort study. Clin Radiol 2009; 64:699–705. [DOI] [PubMed] [Google Scholar]
- 16.Hsu WH, Hsu PK, Wang SJ, Lin KH, Huang C-S, Hsieh C-C, Wu Y-C. Positron emission tomography-computed tomography in predicting locoregional invasion in esophageal squamous cell carcinoma. Ann Thorac Surg 2009; 87:1564–1568. [DOI] [PubMed] [Google Scholar]
- 17.Rakheja R, Demello L, Chandarana H, Glielmi C, Geppert C, Faul D, Friedman KP. Comparison of the accuracy of PET/CT and PET/MRI spatial registration of multiple metastatic lesions. AJR Am J Roentgenol 2013; 201:1120–1123. [DOI] [PubMed] [Google Scholar]
- 18.Kubiessa K, Purz S, Gawlitza M, Kühn A, Fuchs J, Steinhoff KG, et al. Initial clinical results of simultaneous 18F-FDG PET/MRI in comparison to 18F-FDG PET/CT in patients with head and neck cancer. Eur J Nucl Med Mol Imaging 2014; 41:639–648. [DOI] [PubMed] [Google Scholar]
- 19.Linder G, Korsavidou-Hult N, Bjerner T, Ahlström H, Hedberg J. 18F-FDG-PET/MRI in preoperative staging of oesophageal and gastroesophageal junctional cancer. Clin Radiol 2019; 74:718–725. [DOI] [PubMed] [Google Scholar]
- 20.Lopci E, Hicks RJ, Dimitrakopoulou-Strauss A, Dercle L, Iravani A, Seban RD, et al. Joint EANM/SNMMI/ANZSNM practice guidelines/procedure standards on recommended use of [(18)F]FDG PET/CT imaging during immunomodulatory treatments in patients with solid tumors version 1.0. Eur J Nucl Med Mol Imaging 2022; 49:2323–2341. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Schulz V, Torres-Espallardo I, Renisch S, Hu Z, Ojha N, Börnert P, et al. Automatic, three-segment, MR-based attenuation correction for whole-body PET/MR data. Eur J Nucl Med Mol Imaging 2011; 38:138–152. [DOI] [PubMed] [Google Scholar]
- 22.Zaidi H, Ojha N, Morich M, Griesmer J, Hu Z, Maniawski P, et al. Design and performance evaluation of a whole-body Ingenuity TF PET-MRI system. Phys Med Biol 2011; 56:3091–3106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Law I, Albert NL, Arbizu J, Boellaard R, Drzezga A, Galldiks N, et al. Joint EANM/EANO/RANO practice guidelines/SNMMI procedure standards for imaging of gliomas using PET with radiolabelled amino acids and [(18)F]FDG: version 1.0. Eur J Nucl Med Mol Imaging 2019; 46:540–557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Yu CW, Chen XJ, Lin YH, Tseng Y-H, Lu C-C, Chen B-B, et al. Prognostic value of 18F-FDG PET/MR imaging biomarkers in oesophageal squamous cell carcinoma. Eur J Radiol 2019; 120:108671. [DOI] [PubMed] [Google Scholar]
- 25.Marschner CI. Glm2: fitting generalized linear models with convergence problems. R J 2011; 3:12–15. [Google Scholar]
- 26.Gagolewski M. Bibliometric impact assessment with R and the CITAN package. J Informetrics 2011; 5:678–692. [Google Scholar]
- 27.Tu DG, Chen HY, Yao WJ, Hung Y-S, Chang Y-K, Chang C-H, Wang Y-W. Verification of the efficacy of new diagnostic criteria for retropharyngeal nodes in a cohort of nasopharyngeal carcinoma patients. Int J Med Sci 2021; 18:3463–3469. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Verma R, Bhardwaj N, Singh PD, Bhavsar A, Sharma V. Estimation of sex through morphometric landmark indices in facial images with strength of evidence in logistic regression analysis. Forensic Sci Int Rep 2021; 4:100226. [Google Scholar]
- 29.Hong W, Liang L, Gu Y, Qi Z, Qiu H, Yang X, et al. Immune-related lncRNA to construct novel signature and predict the immune landscape of human hepatocellular carcinoma. Mol Ther Nucleic Acids 2020; 22:937–947. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Lee G, I H, Kim SJ, et al. Clinical implication of PET/MR imaging in preoperative esophageal cancer staging: comparison with PET/CT, endoscopic ultrasonography, and CT. J Nucl Med 2019; 55:1242–1247. [DOI] [PubMed] [Google Scholar]
- 31.Kayani B, Zacharakis E, Ahmed K, Hanna GB. Lymph node metastases and prognosis in oesophageal carcinoma--a systematic review. Eur J Surg Oncol 2011; 37:747–753. [DOI] [PubMed] [Google Scholar]
- 32.Van Rossum PSN, Van Lier A, Lips IM, et al. Imaging of oesophageal cancer with FDG-PET/CT and MRI. Clin Radiol 2015; 70:81–95. [DOI] [PubMed] [Google Scholar]
- 33.Goense L, Heethuis SE, Van Rossum PSN, et al. Correlation between functional imaging markers derived from diffusion-weighted MRI and 18F-FDG PET/CT in esophageal cancer. Nucl Med Commun 2018; 39:60–67. [DOI] [PubMed] [Google Scholar]
- 34.Mantziari S, Pomoni A, Prior JO, Winiker M, Allemann P, Demartines N, Schäfer M. 18F- FDG PET/CT-derived parameters predict clinical stage and prognosis of esophageal cancer. BMC Med Imaging 2020; 20:7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bailey DL, Pichler BJ, Gückel B, Barthel H, Beer AJ, Botnar R, et al. Combined PET/MRI: from status quo to status go. Summary report of the fifth international workshop on PET/MR imaging; February 15-19, 2016; Tübingen, Germany. Mol Imaging Biol 2016; 18:637–650. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Baiocco S, Sah BR, Mallia A, Kelly-Morland C, Neji R, Stirling JJ, et al. Exploratory radiomic features from integrated 18F-fluorodeoxyglucose positron emission tomography/magnetic resonance imaging are associated with contemporaneous metastases in oesophageal/gastroesophageal cancer. Eur J Nucl Med Mol Imaging 2019; 46:1478–1484. [DOI] [PMC free article] [PubMed] [Google Scholar]