Non–small-cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancers, with up to 40% of patients with NSCLC exhibiting metastatic disease upon diagnosis.1 Accurate mediastinal staging is imperative for devising optimal treatment strategies and effectively managing patients. While invasive procedures serve as the gold standard for nodal staging, their utilisation is hampered by inherent risks of complications, cost concerns and diagnostic delays.2 Non-invasive mediastinal lymph node (LN) staging presents a promising alternative to identify patients requiring invasive interventions. Positron emission tomography/computed tomography (PET/CT) with 18F-fluorodeoxyglucose (18F-FDG) emerges as a non-invasive modality for cancer staging, restaging, radiotherapy planning and treatment response prediction.3,4

However, reliance solely on the maximum standardised uptake value (SUVmax) for quantifying metabolic activity in 18F-FDG PET/CT poses limitations. SUVmax provides a single voxel value that is insufficient for capturing the entirety of tumour volume or offering comprehensive insights into overall tumour metabolism. Consequently, it falls short as a determinant of the heterogeneous nature of lung tumours characterised by distinct biological activity regions.5 Moreover, guidelines caution against considering SUVmax of the primary tumour as a reliable biomarker for determining disease prognosis.6,7 Recent research underscores the potential of alternative parameters such as metabolic tumour volume (MTV), total lesion glycolysis (TLG) and heterogeneity factor (HF), derived from SUV calculations on PET/CT scans, as robust prognostic indicators.5,8–12 MTV reflects metabolically active tumour volume, while TLG signifies both volumetric and metabolic tumour burden, offering more nuanced measures of tumour aggressiveness and burden.13–16 Notably, intra-tumoural heterogeneity, quantified by the FDG uptake heterogeneity factor, emerges as a salient predictor of lymph node metastasis (LNM) across various malignancies.17–19 Leveraging semi-quantitative parameters of the primary tumour and LNs in F-18 FDG PET/CT, our study aims to discern their efficacy in predicting LNM during NSCLC clinical staging. Additionally, we analyse the utility of a novel scoring system incorporating these parameters alongside visual LN assessment for enhanced mediastinal staging accuracy.

This study delved into the efficacy of various semi-quantitative parameters, such as LN size, SUVmax, MTV and TLG, in predicting mediastinal LN metastasis. Additionally, we explored how these parameters, combined with primary tumour characteristics, including size, SUVmax, MTV, TLG, and HF, measured via 18F-FDG PET/CT, contribute to predictive accuracy. Our research suggests that a scoring system that integrates lymph node size, SUVmax, MTV, TLG and visual comparison is more effective in non-invasively staging lymph nodes for detecting positivity than relying on these parameters individually. Additionally, our study boasts the largest sample size to date for evaluating lymph node parameters.

Approximately 60%–70% of lung cancer diagnoses occur in individuals aged 65 or older, with men exhibiting a higher incidence. Notably, Lu et al. comprehensive analysis spanning 1973 to 2015 revealed a declining incidence of lung cancer in men, decreasing with a rising trend in women. Furthermore, their study highlighted a higher 5-year survival rate among women. Consistent with prevailing trends, our study predominantly comprised male participants (91.5%, n=292), with an average age at diagnosis of 61.62±8.23 years.

Age and tumour size serve as well-established risk factors for LNM in NSCLC.24 For instance, investigation of Kong et al. on patients with resected adenocarcinoma in situ (adenosquamous cell carcinoma) revealed a heightened likelihood of LNM among individuals younger than 65 years with tumours larger than 3cm in size.25 Intriguingly, neither sex nor histological subtype significantly influenced LNM likelihood, with optimal cut-off points identified at 64.5 years and 2.85cm for age and tumour size, respectively (AUC: 0.567, 95% CI: 0.544-0.591; AUC: 0.677, 95% CI: 0.656-0.698, respectively). Similarly, in a study focusing on peripheral early-stage lung ADC, tumour size emerged as a robust predictor of LNM, while age and sex exhibited negligible impact.26 However, our study did not unveil any significant associations between age, sex, tumour size and LNM (P=0.4, P=0.607, P=0.62). Notably, Miao et al. reported analogous findings, underscoring the non-predictive nature of age, sex and tumour size in patients operated for lung ADC.27 It is crucial to acknowledge the inherent heterogeneity within NSCLC subtypes, which may contribute to divergent prognostic outcomes across studies. The discrepancies observed in study outcomes may be attributed to heterogeneity.

LNM stands out as a key marker of malignancy, often indicating highly invasive tumours with poor prognoses.28 While invasive methods are typically preferred for mediastinal LN staging, there are situations, such as in patients with significant comorbidities, where invasive procedures may not be viable. In such cases, treatment decisions can be guided by a PET/CT-based clinical staging approach.29,30 In our study, the development of a scoring system based on predefined PET/CT parameters revealed a substantial 14-fold increase in the risk of metastasis in LNs scoring 3 and above.

Despite its widespread use, SUVmax, as a single parameter, has been found to be an unreliable indicator and inadequate for predicting prognosis, as highlighted in several studies.6,7,31,32 Relying solely on SUVmax for PET/CT interpretation can lead to the overlooking of LNM, with reported incidence rates varying between 8% and 58%.33 In another study, the false positive rate of PET-CT was found to be 25% (57 of 234 patients). Upon further analysis, the researchers determined that the histological subtype (squamous cell carcinoma) was the primary reason for this discrepancy.34 Recognising these limitations, recent research has focused on alternative methods for predicting metastasis. Volumetric parameters such as MTV, TLG and HF in PET/CT imaging have emerged as promising predictors and prognostic markers for LNM and staging across various cancer types. Clear and concise imaging metrics could significantly enhance cancer diagnosis and management.8,11,12,14,15,18,35 Our study revealed no association between the semi-quantitative parameters of the primary tumour and LNM. Although increased values for tumour size, MTV, TLG and HF were linked to advanced stage, they did not correlate with primary tumour SUVmax. These findings underscore the limited reliability of SUVmax as a parameter. However, they do suggest the potential suitability of other parameters as prognostic indicators, a notion supported by the study of Ouyang et al.17 In their analysis, tumour SUVmax and SUVmean values for ADC and SCC were not significant factors for LNM. Nevertheless, primary tumour MTV emerged as a crucial predictive factor, with an optimal cut-off value of 29.50 (AUC: 0.659, sensitivity: 70.4%, specificity: 59.2%) for occult LNM in patients with stage I ADC (P=0.044; OR: 1.671, 95% CI: 1.013-2.756). Similarly, an optimal cut-off value of 0.198 (AUC: 0.759, sensitivity: 88.9%, specificity: 61.1%) for primary tumour HF in patients with SCC was identified as an independent risk factor (P=0.023; OR: 8.799, 95% CI: 1.351–57.317).

In the mediastinal region, LNs with a short axis diameter greater than 10mm are commonly diagnosed as malignant on contrast-enhanced CT36; however, their sensitivity is relatively low at 59%, with a specificity of 78%.37 On the other hand, PET/CT imaging demonstrates a higher sensitivity (85%) in LNs larger than 1cm in diameter, dropping to 32% in smaller nodes.38 Kuzucuoglu et al. found that LNs with high FDG uptake, a size greater than 16mm and an SUVmax greater than 2.9 have a higher risk of metastasis.39 However, assessing LN metabolic and volumetric parameters for metastasis prediction is relatively uncommon compared with tumour evaluations. In a separate study, the SUVmax values of metastatic LNs were notably higher than in non-metastatic ones when considering SUVmax of ≥4 as PET-positive for metastasis.40 In this study, a correlation was found between tumour SUVmax, histological subtype and tumour size, with no such correlation found for LN SUVmax. Studies have established an SUVmax threshold of ≥2.5 as appropriate for detecting pathological LNs.41 A meta-analysis suggested that for the mediastinal staging of NSCLC, PET/CT based on the SUVmax ≥2.5 positivity criterion has a sensitivity of 81% and specificity of 79.4%, while based on LN FDG uptake higher than the background activity criterion, the sensitivity and specificity estimates were 77% and 90%, respectively.4

In a 2016 study, Rogasch et al. demonstrated that LNs with visual FDG uptake higher than that of the liver (LN>L) were 3.3 times more effective in detecting malignancy than LN with lower uptake than the liver but with more MBPS (LN>MBP).42 This study involved 86 patients with NSCLC, with three experts of varying experience levels analysing the PET images. Their evaluations produced comparable results, indicating the reproducibility of the suggested classification. In a related recent study with a larger patient sample, it was noted that LNs with higher uptake from the liver (LN>LBC) necessitate less invasive staging than those with higher MBPS (LN>MBPS).43 However, while a lower false-positive rate would be achieved, there is a risk of missing more N2 diseases, including multiple diseases. Therefore, to mitigate excessive invasiveness in staging, the authors proposed limiting invasive staging to patients with all three risk factors (cN1, tumours larger than 3cm and central tumours), in line with existing guideline recommendations.20,21

In our study, we observed a 1,933-fold visual increase in the risk of LNM with high FDG uptake from the liver. Our results showed that 44.3% (n=424) of LNs were oversampled because high FDG uptake corresponding to the mediastinal blood pool (LN>MBPS) was considered positive for PET. If LNs with a higher FDG uptake than the liver (LN>L) had been considered positive, this figure would have been 22.3%. Notably, our scoring system achieved the lowest number of unnecessary invasive procedures, with a rate of 15.1%.

There is a noticeable gap in the current literature regarding LN metabolic parameters, leading to uncertainty about their clinical utility. While LN SUVmax, MTV and TLG have been linked to recurrence,44 volumetric parameters (MTV and TLG) have been reported to carry a higher prognostic risk than metabolic parameters (SUVmax and SUVmean). A recent study by Nakanishi et al. explored LN SUVmax, MTV, TLG and LPR (LN/primary tumour SUVmax) in 84 pN+ patients, revealing significantly higher values in metastatic LNs than in benign ones.45 The authors established cut-off values for LN SUVmax and TLG as 4.15 (sensitivity: 79.4%, specificity: 76.2%) and 1.26 (sensitivity: 81%, specificity: 76.2%), respectively. However, MTV, TLG and SUVmax did not emerge as independent risk factors for predicting metastasis. To the best of our knowledge, this investigation was the primary one to examine metabolic parameters in the LN, wherein the authors used an SUVmax cut-off value of 3.5 for calculating MTV and TLG, highlighting potential limitations due to sample size inadequacy.

In a study by Hua et al., which analysed 206 LNs, parameters such as LN size, SUVmax, SUVmean and TLG were found to be higher in metastatic LNs than in non-metastatic ones (P<0.001). However, MTV was higher in non-metastatic LNs, although not statistically significant.46 The study identified optimal diagnostic cut-off values as 5.47 for SUVmax, 10.30 for TLG and 9.90mm for LN size. Both studies suggest that LN TLG is elevated in metastatic LNs, but differences in results between studies underscore the lack of standardisation in calculating MTV. In our investigation, TLG emerged as a distinct risk factor for LN metastasis, particularly when the SUV cut-off was set at 2.5. If semi-quantitative parameters continue to offer reliable predictions of LN metastasis, the development of an automated system utilising a validated method could enhance clinicians’ ability to assess metastatic potential based on routine PET values.

In our study, we identified semi-quantitative measures of LNs and non-squamous histology as independent risk factors for LNM. Our novel scoring system represents a simple and promising approach for treatment planning and monitoring in patients with early and locally advanced NSCLC who are ineligible for invasive staging due to contraindications, refuse surgery, or when traditional methods fail to provide a definitive diagnosis.

It is important to note that this scoring system not only aids in avoiding unnecessary invasive procedures but also ensures prompt initiation of treatment, thus potentially reducing the economic burden on healthcare systems by avoiding unnecessary resource utilisation. Additionally, reducing the time between diagnosis and treatment of lung cancer can lead to better patient outcomes. Overall, we believe that implementing this scoring system could significantly reduce the cost of managing lung cancer while improving patient care.

Nonetheless, our study has several limitations. First, its single-centre design constrained the sample size. Second, the retrospective nature of the study meant that not all LNs with heightened uptake on PET/CT were sampled. Moreover, the inclusion criteria focused solely on patients undergoing EBUS for staging, potentially overlooking other cases. Furthermore, LNs evaluated with EBUS exhibiting sonographically benign criteria were not histopathologically evaluated, potentially introducing bias. A prospective study design, especially categorising cancer subtypes separately, could yield more robust results. Additionally, the lack of standardisation in the 18F-FDG PET/CT imaging protocol for measuring MTV and TLG poses a challenge. The segmentation method for determining MTV is still evolving, and there is currently no validated technique. Variability in studies arises from calculations using different SUV threshold values. Moreover, the fixed threshold SUV of 2.5 used to calculate MTV and TLG may lead to overestimation in tumours with high FDG uptake and underestimation in those with low uptake due to the exclusion of areas below the threshold value.

The developed scoring system aims to predict the likelihood of LNM in malignancies and to reduce unnecessary invasive staging. Our research identifies high FDG uptake in LNs from the liver, along with LN size, SUVmax, MTV and TLG, as independent risk factors for LNM. These parameters offer supplementary advantages to current clinical staging methods, potentially decreasing the need for invasive procedures. For individuals with a total score below 2.5, invasive staging may not be warranted. LNs scoring 3 or above exhibit a high likelihood of metastasis, indicating their suitability for invasive pathological staging. However, multi-centre studies with larger patient cohorts are necessary to validate this scoring system before routine implementation. Furthermore, this scoring system could be applied not only for LNM but also for the differential diagnosis of various LN conditions, such as granulomatous, reactive, or anthracotic LNs. Future studies on a larger scale are imperative to comprehensively validate this approach.

This study has not been published or presented any where before.

In this study, national and international ethical rules are observed.

Ethics Committee of the University of Health Sciences, Keçiören Training and Research Hospital (14.09.2021/2012-KAEK-15/2365).

The authors declared that this study has received no financial support.

Ayperi Öztürk and Özlem Özmen: conceptualization, methodology, software. Özlem Özdağ: data curation, writing- original draft preparation. Ebru Tatcı and Figen Öztürk Ergür: visualization, investigation. Özlem Özmen: supervision, validation. Figen Öztürk Ergür: writing. Aydın Yılmaz and Ayperi Öztürk: reviewing and editing.

No conflict of interest was declared by the authors.