Non-progressive micrometastatic malignant lymph nodes associated with the influx rate of [18F]FDG PET/CT in esophageal squamous cell carcinoma patients

Abstract

BACKGROUND: Lymph node metastasis (LNM) is closely related to the prognosis of esophageal squamous cell carcinoma (ESCC). Most primary tumors (PTs) of ESCC can be diagnosed via conventional radiology. Nonetheless, suspicious micrometastases of LNM in ESCC patients with indistinct imaging features are generally misdiagnosed, which may lead to further deterioration or relapse. In this study, [¹⁸F]fluorodeoxyglucose ([¹⁸F]FDG) positron emission tomography/computed tomography (PET/CT) was used to quantify the features of micrometastatic LNM to confirm the stage and location of ESCC. METHODS: ESCC patients (n = 31) were divided into two groups, non-N0 stage (n = 16) and N0 stage (n = 15), according to the TNM classification, and [¹⁸F]FDG PET/CT 0–60 dynamic scans were performed. Lymph nodes (LNs; n = 102), including malignant lymph nodes (MLNs; n = 44) and benign lymph nodes (BLNs; n = 58), were identified by professional radiologists. Thirty-five confirmed malignant lymph nodes (cMLNs) were detected based on pathological biopsy examination, along with nine suspicious MLNs (sMLNs), which were not confirmed by biopsies but appeared enlarged compared with regular BLNs. Furthermore, percentiles of PTs and LNs were sectioned as cervical and upper thoracic (70.69%), middle thoracic (52.27%), and lower thoracic (88.89%) according to their anatomical locations. In this study, p-values from an unpaired, nonparametric Mann‒Whitney test, the area under the curve (AUC), and the sensitivity and specificity of differentiating non-N0 stages from N0 stages or MLNs from BLNs by applying the maximum standard uptake value (SUV_max) or the net influx rate constant (K_i) were statistically compared. RESULTS: Regarding PTs, both the K_i value and the SUV_max were significantly higher in the non-N0 stage than in the N0 stage, with a p-value of 0.0078 (0.0299 SUV_max; all SUV_max values are in parentheses), an AUC of 77.25% (72.55%), a sensitivity of 76.47% (70.59%), and a specificity of 73.33% (73.33%). Both K_i and the SUV_max were significantly higher in total MLNs than in BLNs (p < 0.0001), with an AUC of 98.12% (93.85%), a sensitivity of 100% (100%), and a specificity of 93.18% (86.36%). However, significantly greater K_i values were noted in the sMLNs than in the BLNs (p < 0.0001), whereas no significant differences were found for the SUV_max (p = 0.0561). The AUC of K_i was 93.63% (79.18% SUV_max), and the sensitivity and specificity of K_i (SUV_max) were 93.10% (77.59%) and 92.31% (76.92%), respectively. In different segments from the cervical and upper thoracic regions to the middle thoracic region, the sMLN proportions were 88.89%, 11.11%, and 0%, the MLN proportions were 52.27%, 34.09%, and 13.64%, and the BLN proportions were 70.69%, 20.69%, and 8.62%, respectively. From the cervical and upper thoracic regions to the middle thoracic and lower thoracic regions, the PT proportions were 19.35%, 61.29%, and 19.35%, respectively. CONCLUSIONS: K_i was a more promising parameter than the SUV_max non-progressive micrometastatic MLNs of ESCC. For PTs and progressive MLNs, K_i and the SUV_max were equally effective in differentiating between malignancies and BLNs.