Developing non-invasively methods to evaluate and differentiate MLN from BLN is clinically important. Our study is the first time to focus on the issue with dynamic 18F-FDG PET/CT in patients with ESCC and the main results indicated that: 1) diagnostic parameters, including Ki and SUVmax in MLN were higher than in BLN, and the difference with statistically significant (p < 0.00001); 2) SUVmax and Ki was best correlated in both lymph nodes and primary tumors (r = 0.858 verse r = 0.952); 3) quantitative dynamic 18F-FDG PET/CT protocol may suggest higher accuracy for distinguishing MLN from BLN in ESCC patients; 4) All parameters of the primary tumors in N0 stage was slightly higher than non-N0 stage, but without significant difference (p > 0.05).
Maximal standard uptake value (SUVmax) is usually used as the parameter for PET semi-quantitative analysis in clinic to evaluate glucose metabolism of static imaging3,16,17. Pharmacokinetic analysis of dynamic PET/CT allows quantitative assessment of FDG influx constant (Ki) using Patlak model. Time-activity curves (TACs) provide kinetic parameters that allow the assessment of physiological processes in space and time. For most malignant lesions the TACs were persistent ascending, whereas the majority of the benign lesions showed a low slope and the FDG uptake were lower 12,19,20. We compared of SUVmax, Ki, in MLN and BLN. In general, we found each parameter had great significant difference between MLN and BLN (p < 0.0001, Fig.1). Further more, to investigate the impact of different locations of primary ESCC, BLN and MLN parameters at different ESCC location were compared in detail (Table 2), upper and middle thoracic locations seemed to show more significant difference (all p < 0.001), but the cervical and the lower thoracic & abdominal showed less significance (p < 0.01), this was possibly associated with the clinical primary ESCC of upper and middle thoracic ESCC accounted for the most (Table S2). We compared the BLN from N0 stage and non-No stage, and no difference was found, which was reasonable in clinic.
We also compared the above mentioned parameters in PT (N0 stage verse non-No stage). Interestingly, we found both SUVmax and Ki in non-N0 group was slightly higher than in N0 groups (Table S1, Fig. 2), however, there was no significant difference in two groups (p > 0.05). This may be caused by limited patient population in the study. It has been reported that SUVmax was associated with the histopathological malignancy grade and differentiation.
The correlation between SUVmax and Ki was great in both lymph nodes (r = 0.858) and primary tumors (r = 0.952) (Fig. 3), but it was clear that the less correlation in lymph notes arributed to the difference between BLN and MLN. ROC curve showed the diagnostic accuracy that Ki (90.61%) was greater than SUVmax (88.16%) (Table 2, Fig. 4). Which may indicate Ki is a more sensitive diagnostic parameter in dynamic imaging than static modality (SUVmax). Yuan S et al. assessed locoregional lymph nodes in 32 ESCC patients, and they reported the sensitivity, specificity and accuracy of PET/CT for malignant lymph nodes diagnosis were 93.90%, 92.06% and 92.44%, respectively.17 And later, Hu Q et al reported that the static FDG PET/CT for differentiating malignancy from benign were 76.06%, 85.16%, 83.33%21. Our results were moderate compared with others.
This study is the first time to evaluate MLN with dynamic 18F-FDG PET/CT in patients with ESCC. Previous dynamic 18F-PET FDG study mainly focus on differentiation of benign from malignant primary lesions12,13,22,23, but rarely was related to MLN diagnosis. Yang M et al discussed dynamic 18F-FDG PET scans in 62 non-small cell lung cancer (NSCLC), and concluded that the dynamic modeling for MLN (Ki MLN) was more sensitive than the SUVmax to detect metastatic lymph nodes14. Lockau H et al underwent dynamic 18F-FDG PET lymphography for identification of lymph node metastases in murine melanoma, and indicated the MLN showed significantly longer retention of the radiotracer than in nonmetastatic lymph nodes20. This is consisting with our findings. Lockau H et al performed multiple time points dynamic PET/CT in 74 patients with oral/head and neck cancer, and their results indicated that the 18F-FDG-PET did not predictably identify metastatic cervical lymph nodes24. Since not all of lymph nodes were operated and pathologically confirmed, and the enrolled patients were limited in this study, further investigations are needed to confirm its potential in MLN differentiation. Studies have demonstrated the value of dynamic PET/CT scan in lesions differential diagnosis12,18-20,22,23,25-27, but it has not been translated to the clinic, mainly because its complexity and only allows the assessment of one field of view (typically 15–25 cm), limited the coverage extent of scanner18,28. Dynamic whole-body PET/CT, with iterative image reconstruction, it is possible to acquire eyes-to-thighs imaging in a shorter time, which may overcome the drawback of routine PET/CT scan29.