Diagnostic utility of metabolic parameters on FDG PET/CT for lymph node metastasis in patients with cN2 non-small cell lung cancer

Accurate lymph node (LN) staging of primary non-small cell lung cancer (NSCLC), especially in patients with clinically suspected N2, is crucial in making therapeutic strategy decisions and in determining clinical outcomes [1‐3]. In recent years, after receiving the results of the Pacific trial [4], the treatment strategy for stage III unresectable locally advanced NSCLC has changed dramatically, and the importance of assessing mediastinal LN metastasis accurately has increased. In clinical practice, computed tomography (CT) and/or fluorine-18-fluoro-2-deoxy-D-glucose-positron emission tomography (FDG-PET/CT) are usually performed for clinical LN staging. However, these modalities do not play a complete role in LN staging and causes diagnostic ambiguity (providing false-positive or false-negative results) in the clinical practice. Although LNs with a short-axis diameter of more than 10 mm on a contrast-enhanced CT are generally diagnosed as malignant, a previous meta-analysis showed a relatively low sensitivity (59%) and specificity (78%) [5]. In previous studies, using FDG-PET/CT in diagnosing mediastinal LNs also showed a relatively low sensitivity of 50–79% and specificity of 72–94% [6‐10]. Inflammatory changes in LN due to tuberculosis or granulomatous disease can increase FDG uptake and false-positive results [6, 11]. On the contrary, a maximum standardized uptake value (SUVmax) threshold is associated with a greater number of false-negative findings [12, 13]. Mediastinal LN staging by endobronchial ultrasound guided-transbronchial needle aspiration (EBUS-TBNA)/mediastinoscopy was reported to be useful and was recommended when LN metastasis is suspected on imaging [14]. However, there are LNs, including those in stations #5 or #6, which are difficult to evaluate with EBUS-TBNA. Furthermore, there are possible complications, such as bleeding and mediastinitis [15]. Hence, a more accurate and less invasive diagnosis of LN involvement is required in patients with cN2 NSCLC.

In recent years, the usefulness of metabolic parameters, including metabolic tumor volume (MTV) and total lesion glycolysis (TLG) has been investigated. Several studies reported that MTV and TLG in the primary tumor of lung cancer, which provides information on tumor activity and volume, could predict the pathological tumor invasive size [16], LN status [17, 18], and prognosis [19‐22] more accurately than SUVmax. Furthermore, the LN-to-primary tumor ratio of SUVmax (LPR) was also assessed to determine whether this parameter can be a good predictor for LN metastasis in NSCLC patients [13, 23]. However, only a few studies examined the diagnostic utility of these metabolic parameters and focused not on the primary tumor but on LNs for LN staging of primary lung cancer. Therefore, we aimed to investigate the role of metabolic parameters on FDG-PET/CT for LN metastasis in patients with cN2 NSCLC.

Forty-five patients with 84 clinically diagnosed positive hilar or mediastinal LNs were enrolled in the present study. The clinicopathological characteristics are summarized in Table 1. All cohorts comprised 39 men (87%) and six women, with a median age of 70 years (40–79). The median tumor size was 35 mm (9–106 mm), and the clinical T factor was mainly T1–T2 (73%). The primary histology was adenocarcinoma in 18 patients (40%), squamous cell carcinoma in 18 patients (40%), and others including adenosquamous cell carcinoma or large cell carcinoma and neuroendocrine tumors in nine patients (20%). In the present study, lobectomy was mainly performed (76%), and systemic LN dissection of more than ND2a-2 was performed on 36 patients (80%).

Of the 84 hilar or mediastinal LNs clinically diagnosed as positive, 63 were pathologically proven as positive (75%), whereas 21 were pathologically proven as negative (25%). The FDG-PET/CT parameters for negative and positive LNs are presented in Table 2. The SUVmax, MTV, TLG, and LPR of metastatic LNs were higher than those of benign nodes. No significant difference was observed in the primary tumor SUVmax. In the ROC analysis, the AUC value of LPR [AUC, 0.776; 95% confidence interval (CI), 0.640–0.913] was higher than that of LN SUVmax (AUC, 0.753; 95% CI, 0.626–0.880) or LN TLG3.5 (AUC, 0.746; 95% CI, 0.607–0.885) (Fig. 3a). MTV had the lowest AUC value compared with the above parameters (not shown in Fig. 3a). Using the optimal LPR cutoff value of 0.47, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 84.1, 66.7, 88.3, 58.3, and 79.8%, respectively. On the contrary, the LN SUVmax (cutoff value, 4.15) and LN TLG3.5 (cutoff value, 1.26) showed sensitivity values of 79.4 and 81.0%, specificity values of 76.2 and 76.2%, positive predictive values of 90.9 and 91.1%, negative predictive values of 55.2 and 57.1%, and accuracies of 78.6 and 79.8%, respectively. The results of the multivariate analysis of predictors associated with LN metastasis by logistic regression are shown in Table 3. LPR was the independent predictor for LN metastasis (odds ratio, 6.45; 95% CI, 1.785–23.301; P = 0.004).

A subgroup analysis based on the histology of primary tumor was performed. In the subgroup analysis of patients with adenocarcinoma (n = 18; 32 LNs), the SUVmax, MTV, TLG, and LPR of metastatic LNs were higher than those of benign nodes, but there was no significant difference in primary tumor SUVmax (Supplementary Table S1). In the subgroup analysis of patients with squamous cell carcinoma (n = 18; 34 LNs), the SUVmax and LPR of metastatic LNs were higher than those of benign nodes, but there were no significant differences in primary tumor SUVmax, MTV and TLG of LNs (Supplementary Table S2). In the ROC analysis based on adenocarcinoma, TLG3.5 was a better predictor (AUC, 0.816; 95% CI, 0.639–0.985) than LPR (AUC, 0.792; 95% CI, 0.599–0.986) or LN SUVmax (AUC, 0.792; 95% CI, 0.625–0.959) (Fig. 3b). Using the optimal TLG3.5 cutoff value of 1.26, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 71.4, 90.9, 93.8, 62.5, and 78.1%, respectively. In the ROC analysis based on squamous cell carcinoma, LPR was a better predictor (AUC, 0.831; 95% CI, 0.642–1.000) than LN SUVmax (AUC, 0.804; 95% CI, 0.593–1.000) or TLG3.5 (AUC, 0.701; 95% CI, 0.436–0.966). On the other hand, there was no predictor in other histological group. A subgroup analysis based on the LN location was performed. All LNs were assigned to the following two groups: N2 nodes (n = 58) and N1 nodes (n = 26); PET parameters were compared between these groups. The LN SUVmax, MTV, and TLG of N1 nodes were significantly higher than those of N2 nodes, but there was no significant difference in the LPR between the two groups (Supplementary Table S3). In the ROC analysis, the AUC value of LPR was higher than that of the LN SUVmax or TLG3.5 in the N2 nodes as well as in the entire cohort (AUC, 0.908; 95% CI, 0.821–0.995) (Supplementary Fig. S1). On the other hand, there was no significant predictor in the N1 nodes.

The present study suggested that LPR on FDG-PET before surgery, and not SUVmax or TLG, is the best predictor of LN metastasis in patients with cN2 NSCLC. TLG can be a good predictor for LN metastasis in patients with adenocarcinoma. To the best of our knowledge, this is the first study to investigate metabolic parameters such as MTV or TLG for the diagnosis of LN metastasis. This study evaluated the presence of LN metastasis through surgical pathology, including #5 or #6 LN station, which was difficult to evaluate with EBUS-TBNA.

In patients with NSCLC, accurate preoperative LN evaluation is required, especially in those with clinically suspected mediastinal LN metastasis. This is because the status of mediastinum LNs determines the patient’s treatment strategies or modalities. The present study demonstrated that the sensitivity and specificity for LN metastasis were 84.1 and 66.7% using LPR, which could reduce the false-negative findings due to its high sensitivity. In other words, LPR was useful for the exclusionary diagnosis of LN metastasis. These results might highlight the issue of false-positive test results, which resulted from a small volume of benign LN being present or relatively high metabolic activity or both. Several previous studies reported the usefulness of LPR when assessing LN status [13, 23]. In a series of 44 patients with 92 mediastinal LNs with a SUVmax of 2.5–4.0, Moloney et al. reported that the LPR of 0.3 was the optimal cutoff value for predicting malignancy (sensitivity, 91%; specificity, 71%) [13]. In a series of 172 patients with 504 PET-positive LNs, Mattes and colleagues found that a cutoff value of 0.28 was optimal (sensitivity, 93%; specificity, 87%) [23]. Their optimal cutoff values were somewhat lower than that of the present study, which was 0.47. This was most likely due to study population differences. For example, the former study excluded LNs with a SUVmax of more than 4.0, while the latter study included LNs with a SUV max of less than 2.5. Although there were differences in the inclusion criteria, background, and rate of LN malignancy, these results, according to LPR with a high sensitivity and a relatively low specificity, supported our results.

On the contrary, using LPR reduced the specificity and increased the number of false-positive cases. This result might be associated with the concurrent infectious disease such as obstructive pneumonia and granulomatous inflammation owing to the relatively high prevalence of tuberculosis in our ethnic background. According to predictors for LN metastasis using the SUV threshold, previous studies showed high specificity but relatively low sensitivity [6‐10]. Darling and colleagues reported the accuracy of PET-CT in 149 patients who underwent mediastinoscopy and/or thoracotomy for mediastinal staging [9]. Although their study had PET-CT sensitivity of 70% and specificity of 94%, they highlighted the risk of missing the chance of surgery because of the relatively high number of false-positive findings. There are limitations in comparing the sensitivities, specificities, and other parameters between studies due to the heterogeneity of the prevalence of malignancies, ethnic background, tumor histology, and cutoff values. However, including the LPR into the diagnostic approach could increase the accuracy for predicting LN metastasis. Indeed, in a subgroup analysis, the AUC value of the LPR was higher than that of the LN SUVmax or TLG3.5, especially in N2 nodes. We considered that tumor volume had an effect on hilar LNs, but SUV uptake rather than tumor volume may have had a stronger effect on mediastinal LNs. Therefore, LPR was the best predictor of LN metastasis, especially in N2 nodes.

In recent years, the usefulness of metabolic parameters, including MTV and TLG, has been investigated. Several studies could predict the LN status according to MTV and TLG in the primary tumor of lung cancer [17, 18]. However, no studies have investigated metabolic parameters of LN itself, and whether these parameters were useful remains unclear. We also investigated the usefulness of metabolic parameters, and revealed the association between those parameters and pathological invasive size in patients with adenocarcinoma [16]. Initially, we assumed that these metabolic parameters would have a more diagnostic value in predicting LN, as well as the primary tumor, but the results were contrary to that hypothesis. This might be because, unlike the primary tumor, the metabolic activity in the LN was lower and the difference in LN volume was less apparent. In cases of adenocarcinoma, TLG had a higher diagnostic value than LPR. Although the mechanism was unknown, tumor activity or glucose transporter family expression might be different at the metastatic site depending on the tissue type [27]. Hence, further studies are warranted to clarify this mechanism.

The present study has several limitations. First, our retrospective study revealed a single-center investigation conducted in individuals with the same ethnic background and geographical region. Therefore, the insufficiency of data and external validity are potential problems. Second, there was one limitation to confirm that the LN evaluated on imaging was the same LN diagnosed pathologically in any study of this nature. Third, our study had a small sample size. PET scans performed at other institutions were excluded to ensure uniform FDG-PET/CT imaging accuracy. A further prospective multicenter study is needed to assess not only the diagnostic ability but also the effect of metabolic parameters on making treatment decisions.

Our findings suggest that LPR on FDG-PET is a useful predictor for LN metastasis in patients with cN2 NSCLC. In addition, TLG can be a good predictor for LN metastasis in patients with adenocarcinoma. Further prospective, multi-institutional study is needed to validate the outcomes of the present study and to assess the effect of metabolic parameters on the therapeutic strategy.