18F-FDG PET/CT metabolic parameter to predict pathological complete response and survival after neoadjuvant immunotherapy plus chemoradiotherapy in resectable esophageal squamous cell carcinoma Authors: Jichen Yang, Li Ma, Xiaoming Zhao, Peng Jin, Xue Meng, Wenxing Du, Zhe Wu, Wenjie Jiao, Wenfeng Yang Published: 2026 ABSTRACT AbstractBackgroundNeoadjuvant immunotherapy plus chemoradiotherapy (nICRT) for esophageal squamous cell carcinoma (ESCC) received increasing attention because of high pathological complete response (pCR) rate. Accurately assessing residual disease after nICRT is beneficial for dynamic monitoring of patients. This exploratory, single-arm, phase II clinical study aimed to evaluate the diagnostic efficacy of 18F-FDG PET/CT for achieving pCR after nICRT for ESCC, and its prognostic value for survival.MethodsFrom January 12, 2021 to February 21, 2023, a total of 21 resectable ESCC patients receiving nICRT were enrolled and underwent baseline PET/CT scans (scan-1) and surgery. 19 patients underwent scans (scan-2) before surgery. The recorded PET/CT parameters included maximum, mean, and peak standardized uptake values (SUVmax, SUVmean, and SUVpeak), metabolic tumor volume (MTV), and total lesion glycolysis (TLG). The diagnostic performance of PET/CT parameters was analyzed between the patients with pCR of the primary tumor (pCR-tu) and non-pCR-tu. The Cox regression and Kaplan-Meier method were performed for multivariable analysis and survival analyses, respectively.Results13(61.9%) achieved pCR-tu after nICRT among the 21 participants. None of the single metabolic parameters of primary tumor were significant predictors of pCR-tu (all p > 0.05). SUVmax2 of lymph nodes (Ln) has diagnostic value for lymph nodes metastasis (p < 0.01), the area under the curve (AUC), sensitivity and specificity were 0.812, 77.8% and 76.9%, respectively. The 3-year overall survival (OS) and disease-free survival (DFS) rates were 76.2% and 66.7%, respectively. Univariate Cox regression indicated that OS and DFS were correlated with the postoperative ypN stage, ypTNM stage, pCR, pCR-tu, SUVmax2, and LnSUVmax2. Long rank analysis showed that patients with a cutoff value of ≥12.4 for SUVmax2 and of ≥7.9 for LnSUVmax2 had poorer OS and DFS. However, multivariate Cox regression showed that the parameters mentioned above were not independent prognostic factors for OS and DFS.ConclusionsNone of the single metabolic parameters of the primary tumor were significant predictors of pCR-tu. LnSUVmax2 has diagnostic value for lymph nodes metastasis. The patients with SUVmax2 ≥ 12.4 and LnSUVmax2 ≥ 7.9 had poorer OS and DFS, but they are not independent prognostic factors, which needs to be confirmed by larger scale clinical studies.Trial registrationNCT05323890. Registered on April 5, 2022. Retrospectively registered.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s40644-026-01118-5. Background Although neoadjuvant chemoradiotherapy (nCRT) can improve the survival benefits of patients with locally advanced esophageal squamous cell carcinoma (ESCC) [1–3], high recurrence and metastasis rates remain the main reasons for treatment failure [4–7]. Programmed cell death-1 (PD-1) immune checkpoint inhibitors (ICIs), such as tislelizumab, has shown good therapeutic effects in the treatment of esophageal cancer (EC) [8–11]. Studies combining ICIs with CRT, such as the PALACE-1 [8], reported an improved pathological complete response (pCR) rate of 55.6%. Multiple studies have shown that pCR is associated with longer overall survival (OS) and disease-free survival (DFS) [12, 13]. Neoadjuvant immunotherapy plus chemoradiotherapy (nICRT) for resectable EC has attracted increasing attention due to its high pCR rate [9, 10]. However, this treatment strategy shows significant heterogeneity in terms of effectiveness, some patients forgo surgical treatment due to the favorable therapeutic effects of immunotherapy [14]. Some patients did not respond to nICRT or had poor efficacy in postoperative pathological analysis [15]. A small number of patients even experienced disease progression during immunotherapy, including those with hyperprogression on computed tomography (CT) imaging [16–18]. In addition, some patients exhibited pseudoprogression during the treatment [19]. This indicates that immunotherapy can lead to different treatment responses from nCRT or neoadjuvant chemotherapy (nCT), including changes in CT morphology and tissue metabolism. Given the heterogeneity of the efficacy of immunotherapy, it is essential to find effective imaging methods [20] to accurately monitor changes in lesions and predict prognosis, as well as screen EC patients who are most likely to benefit from nICRT. The CT radiological model was used to predict the pCR of ESCC after nICRT [20], due to the irregular morphology of EC lesions, the traditional manual segmentation method is time-consuming and laborious, and is not suitable for large-scale data processing. Moreover, these studies [20, 21] focused on CT radiomics and its machine learning models, mainly concentrating on morphological data, overlooking the potential ramifications of the metabolic differences in tumor tissues on efficacy evaluation. Among many imaging methods, 18F-FDG PET/CT is the most commonly used molecular imaging technique, which can reflect the metabolic changes in tumor tissues. This study aims to apply the tumor metabolomics of 18F-FDG PET/CT to predict the pathological response and survival to nICRT for resectable ESCC. Patients and methods Study design and participants This prospective, single-arm, single-center, exploratory, phase II clinical study was approved by the Clinical Research Ethics Committee of the institution and was registered at ClinicalTrials.gov (No. NCT05323890). Written informed consent was obtained from all patients. Inclusion criteria for patients in the study: age 18–75 years; ESCC histologically confirmed by gastroscopy biopsy within 2 weeks before treatment and potentially surgically resectable thoracic ESCC; clinical stage T1-2N+M0/T3-4a nay N M0 [22]; normal hematologic, hepatic, and renal function; and Karnofsky performance score of 90 or greater. Exclusion criteria for patients in the study: treatment with radiotherapy, chemotherapy, or other treatment before enrollment; presence of any other primary malignancy; inability to undergo stomach conduit reconstruction due to previous gastrectomy; unsuitable for surgery due to significant malnutrition or comorbidities such as cardiovascular and pulmonary disease; or inability to provide informed written consent due to social, family, or other reasons. Exit criteria: If participants or their legal representatives withdraw their informed consent to participate in the study, they must withdraw from the study. 18F-FDG PET/CT scanning and analysis All patients fasted for at least 6 hours before examination, and blood glucose levels in the peripheral blood should be less than 11.1 mmol/L. After intravenous injection of 3.7–5.55 MBq/kg of 18F-FDG, the patient rested for approximately 60 min before undergoing PET/CT scanning (Biograph Horizon; Siemens Healthineers). The CT scan was first performed with an X-ray tube voltage of 130 kV, pitch of 0.8, slice thickness of 3 mm, and a rotation speed of 0.6 s/rotation. PET scanning was performed in whole-body mode at 1.2 min per bed position with an axial slice thickness of 3.375 mm/slice from the head to thigh. Data were reconstructed using the Iterative + TOF method at a postprocessing workstation (Syngo.via; Siemens Healthcare). The patient had normal shallow respiration during image acquisition. Two experienced nuclear medicine physicians analyzed and evaluated PET/CT images using an image-processing workstation (Beijing Mozi Medical Co., Ltd.). In cases of disagreement, a third senior nuclear medicine physician was consulted to make the final judgment after discussion. In order to reduce bias, nuclear medicine physicians were blinded when analyzing PET/CT metabolic parameters after treatment. All nuclear medicine physicians are unaware of the neoadjuvant therapy regimens. Tumor efficacy was evaluated according to PERCIST 1.1 [23]. Delineation of the region of interest (ROI) for each target lesion was performed on axial sections. Five parameters, maximum, mean, and peak standardized uptake values (SUVmax, SUVmean, and SUVpeak), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), were extracted from each lesion for semiquantitative analysis, with an SUVmax value of 30% serving as the ROI threshold. The post-treatment percentage change (Δ%) values for each metabolic parameter were calculated and analyzed as potential diagnostic indicators based on prior knowledge. For example, ΔSUVmax% = (SUVmax of scan-1 – SUVmax of scan-2)/SUVmax of scan-1 × 100. PET/CT before treatment was labeled as scan-1, and PET/CT after treatment was labeled as scan-2. Additionally, SUVmax and short-axis diameter (SAD) values for the regional lymph nodes were recorded. Neoadjuvant treatment regimens The patients received albumin-bound paclitaxel (100 mg/m2 on days 1 and 8), cisplatin 75 mg/m2 (day 1) or 25 mg/m2 (days 1–3), and tislelizumab (Beigene (Beijing) Biotechnology Co., Ltd) 200 mg (day 1) every three weeks for two cycles. For radiotherapy, a total dose of 41.4 Gy was administered in 23 fractions of 1.8 Gy per fraction, with 5 fractions per week, starting on the first day of the first cycle of chemotherapy. All patients were treated with intensity-modulated radiation therapy technique. Radiation therapy was delivered using megavoltage equipment with photon energies of 6–8 MV. Treatment was interrupted or delayed if severe adverse events occurred. Complete tumor resection was performed 6–8 weeks after completion of nICRT. Total tumor volume (GTV) and clinical target volume (CTV) respectively encompass primary tumor (GTVp, CTVp) and metastatic regional lymph nodes (GTVn, CTVn). CTVp provided a proximal and distal margin of 3 cm and a 0.5 cm radial margin around the GTVp. CTVn was formed by uniformly expanding each direction of GTVn by 5 mm. If anatomical barriers such as vertebral bodies, large blood vessels, and other structures are present, the CTV edge needs to be retracted. The planning target volume (PTV) was defined as a 5–8 mm margin of the CTV for tumor motion and setup variations. Treatment was interrupted or delayed if severe adverse events occurred. Complete tumor resection was performed 6–8 weeks after completion of nICRT. Pathologic analysis Postoperative pathology reports included the tumor type and pathological stage [22]. Tumor regression grade (TRG) was graded according to the CAP/AJCC criteria for pathological assessment after neoadjuvant therapy [3]: 0 (complete response (pCR), no viable cancer cells, including lymph nodes); 1 (near complete response; single cells or rare small groups of cancer cells); 2 (partial response; residual cancer cells with evident tumor regression, but more than single cells or rare small groups of cancer cells); and 3 (poor or no response; extensive residual cancer with no evident tumor regression). R0 resection meant the tumor was completely removed with all microscopic margins negative. pCR is defined as the absence of evidence of tumor cells in the surgical specimen, including lymph nodes. Follow-up Follow-up were conducted regularly through outpatient clinics or phone calls. All patients were followed up with every three months in the initial two years, thereafter every six months until the last follow-up. Barium meal, EUS and contrast-enhanced CT were used to evaluate treatment response. OS and DFS were selected as the index to evaluate the prognosis. OS was defined as the time from the date of pathological biopsy diagnosis to the date of death due to any reason or until the last end of follow-up. DFS was defined as the duration time from the initiation of treatment until the date of disease recurrence or death. The follow-up deadline was April 30, 2026. Statistical analysis Categorical variables are presented as frequencies and proportions. All data were verified for normal distribution with Shapiro-Wilk test and for homogeneity of variance with Levene’s test. Normally distributed data are expressed as mean±standard deviation (SD) or median with interquartile range (IQR). Non-normally distributed data are expressed as median with IQR. Differences in categorical variables were analyzed using chi-square test. Paired sample t-test and independent sample t-test was used to compare the metabolic parameters between two groups. Pearson’s test was used for correlation analysis between metabolic parameters and pCR at baseline and after treatment. Receiver operating characteristic (ROC) curve analysis was used to evaluate the diagnostic performance, and to determine the optimal cut-off value of each metabolic parameter using Youden index to identify pCR patients. The comparison between ROC curves was conducted using DeLong’s test. Variance inflation factors (VIFs) were calculated to detect the multicollinearity of variables. A binary logistic regression analysis was established for metabolic parameters that had no collinearity with other variables to analyze their diagnostic performance. Univariate and multivariate analyses of clinical-pathological-metabolic variables were carried out using Cox proportional hazards regression model. Collinearity analysis was used to eliminate interference factors. Survival was analyzed using the Kaplan-Meier method, and the differences between the variables were evaluated using the log-rank test. All p values were two-sided and p values less than 0.05 was considered statistically significant. All data analyses were performed using the IBM SPSS Statistics for Windows (version 25.0; IBM Corp., Armonk, NY, USA). Results Patient characteristics A total of 23 participants newly diagnosed with ESCC were enrolled between January 12, 2021and February 21, 2023.All patients underwent 18F-FDG PET/CT examination at baseline (scan-1) (Fig. 1). One patient refused surgery after nICRT and the other died during nICRT. A total of 21 patients received surgery. Among them, 19 patients underwent PET/CT scanning after nICRT and before surgery (scan-2). Thirteen patients (61.9%, 13/21) with primary tumors achieving pCR were included in the pCR-tu group, and the other eight patients were included in the non-pCR-tu group. No significant differences in age, sex, tumor location, differentiation, or clinical and pathologic stage were observed between the pCR-tu and non-pCR-tu groups (Table 1). Flowchart of the study. pCR-tu, pathological complete response of the primary tumor; PET/CT, positron emission tomography/computed tomography Table 1Clinical and pathological characteristics of enrolled ESCC patientsGroupAll (n = 21)No. (%)pCR-tu (n = 13), No. (%)non-pCR-tu (n = 8), No. (%)χ2pAge (year)≥609 (42.9)5 (38.5)4 (50.0)0.0340.854<6012 (57.1)8 (61.5)4 (50.0)SexMale18 (85.7)11(84.6)7(87.5)0.0340.854Female3 (14.3)2 (15.4)1(12.5)Locationupper6 (28.6)2 (15.4)4 (50.0)0.3320.189Middle6 (28.6)5 (38.5)1 (12.5)Lower9 (42.8)6 (46.1)3 (37.5)Clinical T stageT21 (4.8)1(7.7)0 (0)0.6460.421T320 (95.2)12 (92.3)8 (100.0)Clinical N stageN07 (33.3)4 (30.8)3 (37.5)4.3670.113N18 (38.1)7 (53.8)1 (12.5)N26 (28.6)2 (15.4)4 (50.0)AJCC StageII7 (33.3)5 (38.5)2 (25.0)0.4040.525III14 (66.7)8 (61.5)6 (75.0)DifferentiationWell3 (14.3)3(23.1)0 (0)2.3420.310Moderate8 (38.1)4 (30.8)4 (50.0)Poor10 (47.6)6 (46.1)4 (50.0)Pathologic T stageypT013 (61.9)13(100.0)0 (0)21.0000.000ypT12 (9.5)0 (0)2 (25.0)ypT24 (19.1)0 (0)4 (50.0)ypT32 (9.5)0 (0)2 (25.0)Pathologic N stageypN013 (61.9)10 (76.9)3 (37.5)3.4400.179ypN16 (28.6)2 (15.4)4 (50.0)ypN22 (9.5)1 (7.7)1 (12.5)ypTNM stage*I12 (57.1)10 (76.9)2 (25.0)6.0170.111II1 (4.8)0(0)1 (12.5)IIIa5 (23.8)2 (15.4)3 (37.5)IIIb3 (14.3)1 (7.7)2 (25.0)TRG013 (61.9)13 (100.0)0 (0)21.0000.00013 (14.3)0 (0)3 (37.5)23 (14.3)0 (0)3 (37.5)32 (9.5)0 (0)2 (25.0)pCR-tu, pathological complete response of primary tumor* Pathological stage: stage I (10 in ypT0N0, 2 in ypT2N0), II (1 cases in ypT3N0M0), and IIIa (2 in ypT0N1M0; ypT1aN1M0, ypT1bN1M0, ypT2N1M0 each with 1), IIIb (ypT0N2M0, ypT3N1M0, ypT2N2M0 each with 1)Values in bold indicate statistically significant results (p < 0.05) Clinical and pathological characteristics of enrolled ESCC patients pCR-tu, pathological complete response of primary tumor * Pathological stage: stage I (10 in ypT0N0, 2 in ypT2N0), II (1 cases in ypT3N0M0), and IIIa (2 in ypT0N1M0; ypT1aN1M0, ypT1bN1M0, ypT2N1M0 each with 1), IIIb (ypT0N2M0, ypT3N1M0, ypT2N2M0 each with 1) Values in bold indicate statistically significant results (p < 0.05) Comparison of tumor metabolic parameters Except for the MTV and TLG, the metabolic parameters on scan-2 for the primary tumors were significantly lower than those on scan-1 among all included patients (p < 0.01) (Table 2). However, the metabolic parameters of both scan-1 and scan-2, as well as the post-treatment percentage changes (Δ%), did not differ significantly between the pCR-tu and non-pCR-tu groups (p > 0.01). Moreover, no significant correlation was observed between pCR-tu and the metabolic parameters of scan-1 or scan-2, or with percentage changes (all p > 0.05; Table 3, Fig. 2). ROC curve analysis showed that none of the single metabolic parameters on scan-2 or post-treatment percentage changes (Δ%) in metabolic parameters were significant predictors of pCR-tu (all p > 0.05; Table 4). Table 2Comparison of PET/CT metabolic parameters at baseline and after nICRTParameterScan-1(n = 19)Scan-2(n = 19)paired samples t-testMedian and IQRMedian and IQRtp95% CIEffect size (Cohen d)SUVmax14.0 (11.7, 17.0)6.6 (5.8, 9.7)5.7450.0004.42, 9.515.28SUVmean8.8 (7.5, 11.1)4.0 (2.7, 5.3)5.6800.0003.61, 6.391.30SUVpeak12.0 (8.9, 13.2)4.8 (3.8, 6.9)5.3610.0003.61, 8.251.23MTV8.2 (4.7, 18.9) a9.3 (7.3, 13.7)−1.1560.248bTLG54.9 (30.8, 147.6) a38.9 (25.9, 55.0) a−1.8170.069bLnSUVmax3.6 (2.5, 6.1)2.2(1.7, 3.6)2.4380.0250.33, 4.54.33NTR0.3(0.2, 0.5)0.3 (0.2 0.5)−0.5390.596−0.11, 0.070.18IQR: interquartile range, CI: confidence interval, Ln:lymph node, NTR:nodal-to-primary tumor SUVmax ratio. Bold font indicates statistical significance after Bonferroni correction (i.e., p < 0.01). aShapiro-Wilk test, p < 0.05, The normality assumptions were violated. b Mann-Whitney test, Z value and p value (2-tailed) Comparison of PET/CT metabolic parameters at baseline and after nICRT IQR: interquartile range, CI: confidence interval, Ln:lymph node, NTR:nodal-to-primary tumor SUVmax ratio. Bold font indicates statistical significance after Bonferroni correction (i.e., p < 0.01). aShapiro-Wilk test, p < 0.05, The normality assumptions were violated. b Mann-Whitney test, Z value and p value (2-tailed) Table 3Comparison of PET/CT metabolic parameters of primary tumorsMetabolic parameterTumor pCRTumor non-pCRindependent sample t-testMedian and IQRMedian and IQRtp95% CIEffect size (Cohen d)Scan-1n = 13n = 8 SUVmax1b13.3 (9.6, 16.9)15.7 (13.9, 20.0)−1.4030.177−8.34, 1.655.22 SUVmean1b8.5 (5.1, 10.7)9.8 (8.7, 12.1)−1.4100.175−5.19, 1.01−0.63 SUVpeak1b9.2 (6.1, 12.7)12.4 (11.8, 16.4)−2.1430.045−8.06, −0.10−0.96 MTV18.2 (5.0, 26.7)a13.8 (5.2, 39.6)a−0.4350.664c TLG139.9 (26.0, 64.0)a129.6 (52.4, 368.1)a0.0360.037c NTR10.25 (0.2, 0.5)0.3 (0.2, 0.3)0.5770.570−0.16, 0.290.24Scan-2n = 11n = 8 SUVmax2b6.1(4.1, 9.0)6.9 (5.8, 11.9)−1.2350.234−4.42, 1.16−0.58 SUVmean2b3.4 (2.5, 5.1)5.17 (3.5, 6.8)−2.0810.053−2.94, 0.02−0.97 SUVpeak2b4.8 (3.4, 6.0)5.9 (4.7, 8.5)−1.8880.076−3.31, 0.18−0.88 MTV2b11.5 (7.2, 14.1)8.7 (4.8, 13.6)0.8260.420−4.13, 9.450.38 TLG242.1(24.9, 63.9)36.5 (25.7, 49.2)a−0.3300.741c NTR20.4 (0.2, 0.5)0.34 (0.2, 0.5)0.2290.821−0.20, 0.250.23Percentage changes (Δ%)n = 11n = 8 ΔSUVmax%b47.1(14.3, 59.0)51.1 (32.1, 67.1)−0.7620.457−32.28, 15.16−0.35 ΔSUVmean%b51.1(8.4, 68.8)49.4 (23.1, 65.9)−0.3290.746−30.25, 22.09−0.15 ΔSUVpeak%b43.7 (9.2, 55.8)52.6 (28.4, 72.5)−1.2170.240−38.47, 10.32−0.57 ΔMTV%−52.5 (−118.3, 57.0)44.8 (24.9, 54.0)a−1.4860.137c ΔTLG%8.2 (−67.1, 82.7)64.2 (55.3, 83.6)a−1.5690.117cTumor pCR: Pathological complete response of the primary tumor, IQR: interquartile range, CI: confidence interval, NTR: nodal-to-primary tumor SUVmax ratioaShapiro-Wilk test, p < 0.05, The normality assumptions were violatedbLevene’s test, p > 0.05, The homoscedasticity assumptions were not violatedcMann-Whitney test, Z value and pvalue (2-tailed). Bold font indicates statistical significance after Bonferroni correction (i.e., p < 0.01) Comparison of PET/CT metabolic parameters of primary tumors Tumor pCR: Pathological complete response of the primary tumor, IQR: interquartile range, CI: confidence interval, NTR: nodal-to-primary tumor SUVmax ratio aShapiro-Wilk test, p < 0.05, The normality assumptions were violated bLevene’s test, p > 0.05, The homoscedasticity assumptions were not violated cMann-Whitney test, Z value and pvalue (2-tailed). Bold font indicates statistical significance after Bonferroni correction (i.e., p < 0.01) Correlation between 18F-FDG PET/CT parameters and pCR of the primary tumor (Pearson correlation) at baseline (n = 13) and after treatment (n = 11). Red represents a positive correlation between the two variables and blue represents a negative correlation. *p < 0.05. pCR: pathological complete response of the primary tumor Table 4Diagnostic value of 18F-FDG PET/CT metabolic parameters for tumor pCR and lymph node involvement, and exploratory multivariable logistic regression analysisMetabolic parameterAUCPCut-off ValueSensitivity%Specificity%Accuracy %PPV%NPV%95% CIPrimary tumorScan-1(n = 21) SUVmax10.7020.12811.80100.046.262.046.7100.00.478, 0.925 SUVmean10.6630.2188.63587.561.571.458.388.90.424–0.903 SUVpeak10.7500.06010.6387.561.566.7100.065.00.540, 0.960 MTV10.5580.66412.2662.561.561.950.072.70.296, 0.819 TLG10.7790.03647.5587.561.571.450.0100.00.572, 0.985 NTR10.5190.8850.25546.287.561.985.750.00.220–0.742Scan-2 (n = 19) SUVmax20.6480.28311.5537.510073.710068.80.385–0.910 SUVmean20.7500.0695.77100.050.079.0100.073.30.521, 0.979 SUVpeak20.7100.1276.6650.090.973.780.071.40.466, 0.955 MTV20.4200.5637.46075.036.452.662.572.70.152, 0.688 TLG20.5110.934109.912.510057.954.6750.240–0.783 NTR20.3980.4570.2954.57563.27554.50.128–0.667Percentage change (Δ%) (n = 19) ΔSUVmax%0.5570.68022.9087.536.457.950.080.00.289, 0.824 ΔSUVmean%0.5230.86914.10100.027.357.950.0100.00.249, 0.797 ΔSUVpeak%0.6250.36457.5050.081.268.458.385.70.359, 0.891 ΔMTV%0.7050.13721.2087.572.720.543.866.70.488, 0.961 ΔTLG%0.7160.11753.0087.563.620.542.960.00.476, 0.956Exploratory multivariable logistic regression analysis* SUVmax2+ΔSUVmax%0.8300.0170.633a62.590.979.083.376.90.642, 1.000 SUVmax2+ΔSUVpeak%0.7730.0480.487a75.072.752.644.460.00.559, 0.986 ΔSUVmean%+ΔSUVpeak%0.7730.0480.344a100.054.552.646.266.70.558, 0.988Regional lymph nodesScan-1(n = 21) SUVmax10.6570.0521.7057.175.077.031.394.40.504, 0.809 SAD10.6570.0510.8542.990.097.959.192.10.499, 0.815Scan-2(n = 19) SUVmax20.8120.0021.6977.876.975.629.873.70.669, 0.955 SAD20.7360.0180.5577.868.776.931.984.20.528, 0.943Percentage change (Δ%) (n = 19) ΔSUVmax%0.6020.3057.1044.471.486.522.797.010.157, 0.639 ΔSAD%0.6860.0622.6366.779.278.916.797.50.495, 0.877Tumor pCR: Pathological complete response of the primary tumor, SAD:Short-axis diameter of lymph node, AUC: Area under the ROC curve, CI: confidence interval. NTR: nodal-to-primary tumor SUVmax ratio. * Multivariate: SUVmax2, ΔSUVmax%, ΔSUVpeak%, ΔSUVmean%aMulticollinearity test: VIF < 10 and tolerance > 0.1. Values in bold indicate statistically significant results (p < 0.05) Diagnostic value of 18F-FDG PET/CT metabolic parameters for tumor pCR and lymph node involvement, and exploratory multivariable logistic regression analysis Tumor pCR: Pathological complete response of the primary tumor, SAD:Short-axis diameter of lymph node, AUC: Area under the ROC curve, CI: confidence interval. NTR: nodal-to-primary tumor SUVmax ratio. * Multivariate: SUVmax2, ΔSUVmax%, ΔSUVpeak%, ΔSUVmean% aMulticollinearity test: VIF < 10 and tolerance > 0.1. Values in bold indicate statistically significant results (p < 0.05) Because the tumor metabolic parameters in some pCR-tu patients decreased by less than 30%, the overall metabolic parameters in pCR-tu group patients on scan-2 remained relatively highs (Tables 2 and 3; Fig. 3), this is an important reason why there was no significant difference or correlation in metabolic references between the pCR-tu and non-pCR-tu groups. 31.6% (6/21) of the patients showed a decrease in ΔSUVmax% of less than 30%. Among them, 4 cases (4/21, 21.1%) were pCR-tu patients, their postoperative pathology revealed extensive inflammatory cell infiltration, multinucleated multinuclear giant cell reaction, and interstitial fibrotic tissue hyperplasia, suggesting a significant immune-inflammatory response. The other two non-pCR-tu cases (TRG3) had a large amount of residual tumor in the primary tumor and fewer tumor-infiltrating lymphocytes, with only slight immune-inflammatory response, indicating insensitivity to nICRT (Fig. 4). Among patients with a decrease of less than 30% in ΔSUVmean%, ΔMTV%, or ΔTLG%, there were a considerable number of pCR-tu and TRG1 patients. Postoperative pathology revealed a large number of infiltrating inflammatory cells and a significant immune-inflammatory response in the primary tumor (Fig. 4). Violin distribution plots of changes in 18F-FDG PET/CT metabolic parameters in 19 patients. pCR-tu, pathological complete response of the primary tumor. Patients with a Δ% decrease of less than 30% were as follows: Δsuvmax%: six patients, including four with pCR-tu and 2 cases of TRG3. Δsuvmean%: five patients, including three with pCR-tu, and two cases of non-pCR-tu (1TRG3 and 1TRG2, respectively). ΔMTV%: 11 patients, including 7 cases with pCR-tu, and 4 cases of non-pCR-tu (3 case of TRG1and 1 TRG2); eight of these had an increase in MTV2 and a negative value of ΔMTV%, including 6 patients with pCR-tu. ΔTLG%: seven patients, including 5 cases with pCR-tu and 2 cases of non-pCR-tu (one case of TRG1and TRG2, respectively) 18F-FDG PET/CT and histopathological images of representative cases of pCR-tu and non-pCR-tu. pCR-tu:pathological complete response of the primary tumor. a, b, c: a 66-year-old man with middle EC who achieved pCR-tu after nICRT (scan-1: SUVmax 9.2, SUVmean 5.08, TLG 30.78, MTV 6.06; scan-2: SUVmax 9.7, SUVmean 5.1, TLG 42.13, MTV 9.24). Postoperative pathological examination revealed a pCR of the primary tumor. a: scan-1; b: scan-2; c: postoperative pathology (hematoxylin and eosin staining; original magnification × 100). d, e, f: a 63-year-old man with lower EC who achieved pathological partial response after nICRT (scan-1: SUVmax16.9, SUVmean 9.94, TLG 47.61, MTV4.85; scan-2: SUVmax 11.9, SUVmean 6.74, TLG 52.17, MTV7.74). Postoperative pathology revealed a large amount of residual tumor tissue (TRG3). d: scan-1; e: scan-2; and f: postoperative pathology (hematoxylin and eosin staining; original magnification × 100) The prediction of therapeutic response with model of tumor metabolic parameters To distinguish hypermetabolism caused by immune inflammation or tumor residue, binary logistic regression analysis models of metabolic parameters were established, to analyze their performance to predict pCR-tu after nICRT. Among the three analysis models, the independent variables in the regression models showed no significant collinearity (VIF < 10 and tolerance > 0.1). The area under the curve (AUC) of the model SUVmax2+ΔSUVmax% was the highest among the three models (AUC = 0.830, p = 0.017). but DeLong’s test showed that the ROC curve of this model was not superior to the other two models (p = 0.203, 0.254). Although its sensitivity (62.5%) was lower than that of the model ΔSUVmean%+ΔSUVpeak%, its specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV) were superior to the other two models (Table 4, Fig. 5, Supplementary table S1). ROC curves for the diagnostic value of 18F-FDG PET/CT metabolic parameters for pCR of the primary tumor and lymph node involvement after nICRT in patients. The metabolic parameters of the primary lesion have no predictive effect on the pCR of the primary tumor. The LnSUMmax2 and LnSAD2 have predictive effects on the metastasis involvement of the lymph node PET/CT detection of regional lymph nodes involvement A total of 174 nodal stations from 21 participants were evaluated in the study (mean number of nodal stations sampled per patient: 8.3). The comparison between the lymph node positive and negative groups is shown in table 5. From ROC analysis, the optimal cut-off value for SUVmax1, SAD (short-axis diameter of lymph node)1, SUVmax2, and SAD2 distinguishing metastatic lymph nodes from benign ones were 1.7, 8.50 mm, 1.05, and 5.5 mm, respectively (Table 5). The AUCs of SUVmax2 and SAD2 were 0.812 and 0.736 (p = 0.002, 0.018), respectively. The DeLong’s test indicated that the diagnostic performance of the ROC curve for SUVmax2 was not superior to that of SAD2 (p = 0.059). Although the sensitivity did not increase compared to SAD2, its specificity was higher than that of the latter (Table 5), and it can be used as a predictive indicator for lymph node metastasis (Table 4, Fig. 5). Table 5Characteristics of lymph node (ln) stations according to pathological involvementCharacteristicsLn involvement (+)Ln involvement (-)Mann-Whitney testMedian and IQRMedian and IQRZ valuep (2-tailed)Scan-1 (21 patients)n = 27an = 147 SUVmax13.8 (0.7, 6.8) d0.7 (0.50, 1.0) d−4.8580.000 SAD1(cm)0.9 (0.4, 1.5) d0.4 (0.30, 0.6) d−4.4690.001Scan-2 (19 patients)n = 9bn = 147 SUVmax22.6 (1.2, 7.4)0.7 (0.50, 1.1) d−3.1910.001 SAD2(cm)0.8 (0.5, 1.6)0.4 (0.30, 0.6) d−2.4530.014Percentage changes (Δ%) (19 patients)cn = 21b,n = 135 ΔSUVmax%55.3 (3.2, 68.5) d0.0 (0.0, 5.0) d−3.6900.000 ΔSAD%6.7 (0.0, 31.7)0.0 (0.0, 0.0) d−3.0810.002IQR, interquartile range; SAD, short-axis diameter. Bold font indicates statistical significance after Bonferroni correction (p < 0.008)a Including 14 Ln stations (14/174, 8.0%) in 7 patients (7/21, 33.3%) proved positive after surgery, and 13 Ln stations from 10 patients showed treatment response and postoperative pCRb Two patients without scan-2 were excluded. c Including 12 Ln stations that showed treatment response and postoperative pCRdp < 0.05, Shapiro-Wilk test, the normality assumptions were violated Characteristics of lymph node (ln) stations according to pathological involvement IQR, interquartile range; SAD, short-axis diameter. Bold font indicates statistical significance after Bonferroni correction (p < 0.008) a Including 14 Ln stations (14/174, 8.0%) in 7 patients (7/21, 33.3%) proved positive after surgery, and 13 Ln stations from 10 patients showed treatment response and postoperative pCR b Two patients without scan-2 were excluded. c Including 12 Ln stations that showed treatment response and postoperative pCR dp < 0.05, Shapiro-Wilk test, the normality assumptions were violated Prognostic value of 18F- FDG PET/CT parameters for the survival The median follow-up of the 21 patients, calculated with the reverse Kaplan-Meier method, was 57.0 months (IQR: 35.0 - 59.0). Among the 21 patients who underwent surgery, 8 (38.1%) experienced disease progression during the follow-up period. As of the follow-up date, all 8 patients died. Among them, 4 patients developed distant metastasis (lung, liver, bone, retroperitoneum, and supraclavicular lymph node, respectively), and 4 patients developed regional recurrence. The median DFS and OS of the entire cohort were not achieved. The 2-year OS and DFS rates were 90.5% (95% CI: 71.1–97.4) and 85.7% (95% CI: 65.4–95.0), respectively. Furthermore, the 3-year OS and DFS rates were 76.2% (95% CI: 54.9–89.4) (Fig. 6A) and 66.7% (95% CI: 45.4–82.8) (Fig. 7B), respectively. The univariate Cox regression indicated that OS and DFS were related to the postoperative N stage, TNM stage, pCR, and pCR-tu of the patients (Supplementary table S2). The OS and DFS of patients in stage I and II were longer than those in stage IIIa and IIIb (Fig. 6C, D). The OS and DFS of patients with pCR and pCR-tu were longer than those of patients with non-pCR and non-pCR-tu, respectively (Fig. 7A–D). Kaplan-Meier survival analysis of OS and DFS in different postoperative stages. (A, B) Overall survival (OS) and Disease-free survival (DFS) in 21 patients who received nICRT. Median OS and DFS were not reached. (C, D) The patients with pathological stage I and II had better OS and DFS than those with pathological stage IIIa and IIIb Kaplan−Meier survival analysis of overall survival (OS) and disease-free survival (DFS) for patients (n = 21) with pCR or tumor pCR. A, B: OS and DFS according to pCR, C, D: OS and DFS according to tumor pCR. The patients with pCR or primary tumor pCR have better OS and DFS Patients were grouped based on the median of metabolic parameters as the cutoff value, with values greater than or equal to the median being the high metabolic group and values less than the median being the low metabolic group. Univariate Cox regression was performed between the metabolic parameters and survival. No significant correlation was found between OS and DFS and the metabolic parameters in the low and high metabolic groups (Supplementary table S3). Then, using ROC analysis, the optimal cutoff value of metabolic parameters for patients with recurrence and metastasis was determined. Based on this cutoff value, patients are divided into groups greater than or equal to the cutoff value and groups below the cutoff value. A univariate Cox regression was conducted between the metabolic parameters and survival. The univariate Cox regression indicated that, OS and DFS were associated with SUVmax2 and LnSUVmax2, but not with other metabolic parameters such as NTR, (Supplementary table S4). Long rank analysis showed that patients with a cutoff value of ≥12.4 for SUVmax2 and a cutoff value of ≥7.9 for LnSUVmax2 had poor OS and DFS (Fig. 8). In the multivariate analysis, we included the clinical pathological and metabolic variables that were significantly correlated with OS and DFS in the univariate analysis. However, multivariate Cox regression showed that none of these variables were independent prognostic factors for OS and DFS (Supplementary table S5). Kaplan−meier survival analysis of overall survival (OS) and disease-free survival (DFS) for patients (n = 19) with primary tumor SUVmax2 ≥ 12.4 versus primary tumor SUVmax < 12.4, stratified by recurrence and metastasis.metastasis.The patients with SUVmax2 ≥ 12.4 in primary tumors have poor OS and DFS Discussion This study indicates that neither single metabolic parameters of the primary tumor nor the percentage change (Δ%) in metabolic parameters after treatment were significant predictors of pCR-tu (all p > 0.05). A study found that in patients with ESCC undergoing surgery after nCRT, ΔSUV > 7.2 and ΔSUV ratios > 58% [24], or the value of MTV after neoadjuvant therapy and % decrease of SUVavg [25] were predictors of pCR, changes in metabolic parameters such as TLG and MTV of the primary tumor were potentially predictive for clinical and pathological response after nCRT for EC [26]. In patients with esophageal adenocarcinoma undergoing nCT [27], both a larger MTV and TLG after induction chemotherapy were significantly associated with a higher chance of a poor pathologic response. However, these studies were conducted in EC patients receiving nCRT or nCT, The PERCIST 1.1 criteria [23] consider a decrease of ≥30% in metabolic parameters as a response to treatment. In this study, it was found that the decrease in ΔSUVmax%, ΔSUVmean%, ΔMTV%, and ΔTLG% in some pCR and TRG1 patients was less than 30%. The postoperative pathology showed that the tumor tissues of these patients exhibited a strong immune-inflammatory response, resulting in increased glucose metabolism, which led to the metabolic parameters remaining at a relatively high level after the neoadjuvant treatment. This would interfere with the evaluation of treatment response by 18F-FDG PET/CT. Therefore, in this group of patients, there was no significant differences or correlations in the metabolic parameters of scan-1 and scan-2, as well as the post-treatment percentage change (Δ%), between the patients with pCR-tu and non-pCR-tu. Therefore, for EC patients receiving nICRT, the addition of immunotherapy induces immune-inflammatory responses in tumor tissue that differ from nCRT or nCT, leading to changes in tumor tissue metabolic parameters. This immune-inflammatory response also exists in the postoperative tumor tissue of patients with breast cancer [28] and non-small cell lung cancer [29] who receive neoadjuvant immunotherapy, but there is no such immune-inflammatory response in the tumor tissue of patients who receive nCT. Therefore, the differences in tumor metabolism between EC patients receiving nICRT and nCRT or nCT may be due to the unique immune-mediated inflammatory response. It is crucial to distinguish whether hypermetabolism is caused by a large amount of tumor residues or by immune-inflammatory responses for assessing therapeutic efficacy. Although, this study showed that none of the single metabolic parameters were significant predictors of tumor pCR, we have made exploratory attempts to analyze it through a regression model of metabolic parameters. In this study, the binary regression analysis model using ΔSUVmax2+ΔSUVmax%, showed a sensitivity and specificity of 62.5% and 90.9%, and false positive (9.1%), respectively, which was significantly lower than the previously reported false positive rates of EC patients receiving nCRT or nCT (27 and 28%, respectively) [30, 31], indicating that this model can reduce the false-positive predictions caused by increased metabolism due to immune-inflammatory responses. Meta-analysis [30] showed that SUVmax had a lower ability to detect residual disease in EC patients receiving nCRT or nCT, with sensitivity, specificity of only 62 and 73%, respectively, which were lower than the predictions of the binary regression model in this study. Another review [31] reported a relatively high sensitivity (69%), but a lower specificity (72%). The detection of residual disease in patients with ESCC receiving nICRT is very important, as accurate prediction of pCR before surgery could potentially allow patients to forgo surgery and would enable researchers to study the feasibility of organ preservation strategies after nICRT. This analysis model was used to predict pCR of primary tumor after treatment, although it has slightly lower sensitivity, it has high specificity and significantly reduces false positive predictions. However, due to the small number of cases, the model suffers from statistical overfitting and is only an exploratory indicator, it is necessary to expand the number of patients to reduce statistical errors, increase statistical power, and apply sufficient cases as a verification cohort for independent validation. Improving the sensitivity of lymph node detection is beneficial for determining lymph node staging and selecting treatment plans, especially for ESCC patients who give up surgery after nICRT. A meta-analysis [27] evaluated the diagnostic accuracy of regional lymph node involvement in EC patients receiving nCRT and found that SUVmax (<2.5) showed high sensitivity (89.0%) but low specificity (50.0%). In this study, the sensitivity and specificity of SUVmax2 in predicting lymph node metastasis were 77.8% and 76.9%, respectively. The sensitivity was slightly lower than that of the above-mentioned studies, but the specificity was higher [27], which reduces false positive predictions. Although the performance of this model in detecting small residual tumors in lymph nodes is still insufficient and there are still a certain number of false positive lymph nodes, overall, this model has relatively high effectiveness in predicting lymph node involvement and can be used as an indicator for predicting lymph node metastasis. This study found that patients with metabolic parameters SUVmax2 ≥ 12.4 and LnSUVmax2 ≥ 7.9 had poorer OS and DFS. However, multivariate Cox regression indicated that these two parameters were not independent prognostic factors. Studies [32] showed that for patients with ESCC undergoing definitive chemoradiotherapy (dCRT), MTV was an independent prognostic factor for OS. However, studies [33] have shown that SUVmax > 12.7 g/mL was an independent prognostic factor for early recurrence in patients with EC, especially ESCC, who underwent nCRT followed by surgery, and their DFS was significantly shortened, which suggests that this result has significant clinical implications for early identification of resectable esophageal cancer patients who may not benefit from surgery due to their significantly increased short-term recurrence risk. Although these studies have shown that a high baseline SUVmax may herald the tumor aggressiveness and early recurrence, the underlying mechanism remains poorly understood. Compared with other tumors, the high FDG-uptake tumors did not have clear immunohistochemical characteristics [34]. Studies [35] have reported that NTR was an independent prognostic factor for both distant metastasis-free survival (DMFS) and OS in patients with ESCC after dCRT. In addition, a meta-analysis [36]. showed that in cancer patients including ESCC, lung cancer, breast cancer, nasopharyngeal cancer, head and neck squamous cell carcinoma, elevated NTR was significantly associated with worse survival outcomes. Elevated NTR can serve as a useful prognostic biomarker for cancer patients and may be potentially employed to guide treatment decisions and monitor disease progression. However, this study found no correlation between NTR and the survival of ESCC patients who underwent surgery after nICRT. It is possible that after nICRT, the immune-inflammatory response of the tumor tissue led to an increase in metabolic parameters, thereby causing data bias, or an increase in sampling error and a decrease in statistical efficiency due to a small number of patients. This study has certain limitations. (1) This was a single-center study, most of the enrolled patients were classified as stage III, with a later clinical stage. Therefore, a selection bias cannot be avoided. (2) All included patients with ESCC underwent surgical treatment; thus, it is not clear whether the results are applicable to other patients. (3) Due to the small number of enrolled cases, our regression model was not independently validated, which limits the reliability of the conclusions. Despite these limitations, this real-world study evaluated a relatively homogeneous group of ESCC patients. The current findings should be regarded as exploratory pending further validation. Validation of our findings will require more large-scale prospective studies with a larger number of cases, including phase III clinical studies. Or computer-aided detection methods [37], integrate imaging data, radiotherapy dose, immune drug application to enhance prediction accuracy. Or the use of new PET/CT tracer such as 18F-FAPI-04 to validate these results [38]. Conclusion In summary, none of the single metabolic parameters of the primary tumor on scan-2 were significant predictors of pCR. LnSUVmax2 has diagnostic value for lymph node metastasis. The patients with metabolic parameters SUVmax2 ≥ 12.4 and LnSUVmax2 ≥ 7.9 had poorer OS and DFS. but these two parameters were not independent prognostic factors. The validation of these data in larger, prospective, multicenter studies is essential. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Acknowledgment We are grateful to all patients, their families, and the site investigators who participated in the study. We also thank Beigene (Beijing) Biotechnology Co.,Ltd. for supplying camrelizumab for this study. Author contributions Jichen Yang was involved in study design and manuscript preparation. Li Ma, Xiaoming Zhao, and Peng Jin contributed to the image acquisition and interpretation. Du and Wu collected and processed data. Xue Meng, Wenjie Jiao and Wenfeng Yang edited and reviewed the manuscript. All authors discussed the results and approved the final manuscript. Funding This study was supported by the National Natural Science Foundation of China (82172720) and the Project of the Medical and Health Science and Technology Development Research Center of the National Health Commission (WA2021RW03), Wujieping Science Foundation of China (320.6750.2022-09-36, 320.6750.2022-17-52), Natural Science Foundation of Shandong Province (ZR2020MH234). Data availability No datasets were generated or analysed during the current study. Declarations Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of our institution and with the principles of the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Consent to participate Informed consent was obtained from all patients enrolled in the study. Consent to publish The authors affirms that the human research participants provided informed consent for the publication of the images in Figures. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Wenjie Jiao, Email: jiaowj@qduhospital.cn. Wenfeng Yang, Email: wfyang@sdfmu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Data Availability Statement No datasets were generated or analysed during the current study.