Prognostic value of pretherapy CT-based Node-RADS for overall survival in extensive-stage small cell lung cancer treated with chemoimmunotherapy: a comparison with clinical N staging

Introduction

Small cell lung cancer (SCLC) is a highly aggressive subtype, accounting for approximately 14% of all lung cancers, with about 70% of patients diagnosed at the extensive stage (1). Traditionally, platinum-based chemotherapy has been the conventional treatment for extensive-stage SCLC, yet the 5-year survival rate remains around 7% (1). Recently, the integration of immune checkpoint inhibitors (ICIs) targeting programmed death-ligand 1 (PD-L1) with chemotherapy has emerged as a first-line treatment, offering improved overall survival (OS) for extensive-stage SCLC (2-4). However, only a subset of SCLC patients benefit from chemoimmunotherapy and the majority experience limited responses (5). Moreover, PD-L1 expression is absent in most extensive-stage SCLC cases, and its levels do not reliably predict response to chemoimmunotherapy (6). Therefore, there is an urgent clinical need for a biomarker or method that can accurately predict the efficacy of chemoimmunotherapy at an early stage.

The American Joint Committee on Cancer (AJCC) tumor-node-metastasis (TNM) staging system is extensively utilized in clinical practice for selecting treatments and predicting outcomes in SCLC. The clinical N (cN) stage has been linked to survival outcomes in SCLC, according to previous reports (7-9). Currently, the cN stage in SCLC is determined by the extent of suspicious involvement of regional lymph node metastasis (LNM). The existing CT criteria for identifying LNM in SCLC focus primarily on lymph node size, thus failing to distinguish metastatic from reactive enlargement. Furthermore, these criteria overlook the prognostic morphological features of lymph nodes, such as spherical shape, increased size, irregular borders, and necrosis, which are consistently associated with less favorable prognosis in lung and other cancers (10-13). Furthermore, the count of positive LNM stations has been identified as an additional factor for stratifying prognosis in lung cancer (14,15). Nonetheless, the prognostic importance of these factors related to lymph nodes has not been thoroughly investigated in SCLC patients.

Computed tomography (CT) is widely used for characterizing primary tumors and metastatic lesions, determining TNM stage, and monitoring treatment response, offering high spatial resolution at a low cost. For staging lung cancer prior to initiating anti-tumor therapy, contrast-enhanced CT serves as the preferred baseline imaging modality (16). The Node Reporting and Data System (Node-RADS) was introduced to standardize lymph node evaluation across cancer types using CT or magnetic resonance imaging (MRI) (17). This system assigns a grade from 1 to 5 to indicate the probability of LNM, corresponding to a spectrum from very low to very high malignancy risk. Evidence suggests that Node-RADS improves diagnostic accuracy in lymph node assessment across multiple cancer types, including nasopharyngeal, breast, lung, stomach, and rectal cancer (18-22). Recently, studies have preliminarily demonstrated an association between Node-RADS and prognosis in gastric and renal cancers (23,24). However, the relationship between Node-RADS and prognosis in SCLC remains insufficiently explored.

Emerging artificial intelligence and radiomics approaches offer valuable quantitative insights for nodal assessment and prognosis prediction (25,26). However, the higher technical complexity of these methods may limit their immediate clinical adoption. There remains a need for a more straightforward, accessible tool based on routine imaging to facilitate risk stratification in practice. This study aimed to explore the prognostic significance of the pretherapeutic CT-based Node-RADS score in extensive-stage SCLC, compared to the cN stage. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1835/rc).

Methods

Participants

A total of 375 consecutive patients with biopsy pathologically confirmed SCLC who underwent chemoimmunotherapy at Hunan Cancer Hospital between July 2019 and December 2023 were initially included in this study. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Hunan Cancer Hospital (No. KY2024532). The requirement for informed consent was waived in this retrospective study. The inclusion criteria were as follows: (I) pathologically confirmed SCLC and classified as extensive-stage based on National Comprehensive Cancer Network (NCCN) guidelines; (II) underwent first-line standard chemoimmunotherapy; and (III) availability of a baseline contrast-enhanced chest CT scan performed within one week prior to initiating chemoimmunotherapy. The exclusion criteria were as follows: (I) history of anti-tumor treatment before baseline CT; (II) with other concomitant malignancies; (III) treatment regimen was not etoposide and carboplatin (EC)/etoposide and cisplatin (EP) plus atezolizumab/durvalumab; (IV) insufficient CT image quality; or (V) incomplete clinical or follow-up data. Figure 1 shows the study flowchart.

Figure 1 Study flowchart. CT, computed tomography; EC, etoposide and carboplatin; EC-SCLC, extensive-stage small cell lung cancer; EP, etoposide and cisplatin.

Patients’ pretherapeutic clinical data and treatment regimens were collected from the electronic medical records (EMR). Survival outcomes were ascertained through a dual-source verification process to ensure completeness. The primary source was the hospital’s integrated EMR system, which documents vital status and dates of death from all inpatient and outpatient encounters. For patients without recent EMR activity, up to three contact attempts were made on different days/times by a dedicated research coordinator. If direct contact with the patient failed, we sought information from designated emergency contacts or primary care physicians with prior consent. In case of death, the date and cause (if available) were recorded. Vital status and date of death (when applicable) obtained by telephone were cross referenced with EMR data. Patients were censored at their last known clinical contact if they were alive at the study cut-off or were lost to follow-up. The survival endpoint was OS, which was defined as the time from initiating therapy to death from any cause.

CT acquisition and Node-RADS evaluation

The baseline CT scan was performed within 1 week before chemoimmunotherapy utilizing either a 64-slice scanner [United Imaging uCT760 (United Imaging Healthcare, Shanghai, China) or Siemens Definition AS+ (Siemens, Erlangen, Germany)] or a 256-slice scanner [GE Revolution Xtream (GE Healthcare, Chicago, IL, USA)]. The scan parameters for the chest were standardized based on the manufacturer’s guidelines. An iodinated contrast medium (ioversol) was used at an injection rate of 2.5 mL/s, with a dosage of 1.5 mL/kg of body weight. Multiplanar reconstructions (axial, coronal, and sagittal) were created at a slice thickness of 1 mm for detailed analysis.

Lymph node stations were defined and assessed according to the International Association for the Study of Lung Cancer (IASLC) lymph node map. Two experienced radiologists (Reader A with 10 years and Reader B with 28 years of expertise in chest radiology) independently assessed regional lymph nodes on pretherapeutic CT images. In cases of initial disagreement in Node-RADS scoring, the two radiologists jointly reviewed the images and reached a final consensus score through direct discussion. Evaluators were kept unaware of survival outcome data. Lymph nodes from 13 stations (bilateral supraclavicular, 2R, 2L, 3a, 3p, 4R, 4L, 5, 6, 7, 8, contralateral hilar) were assessed individually, among which the highest score was selected to represent the patient’s Node-RADS score, defined as maximum Node-RADS score. Given that the majority of cases in this cohort involved central lung cancer, and the primary tumors often merged with ipsilateral hilar lymph nodes, Node-RADS scoring was not performed on the ipsilateral hilar lymph nodes.

The Node-RADS system was used to assess lymph nodes, guided by a three-tiered flowchart (Figure 2). In summary, this system evaluates lymph nodes based on size and structural characteristics, assigning an overall score from 1 to 5 to indicate the level of suspicion for LNM. Figure 3 provides examples of the Node-RADS assessment. Lymph nodes with a score of 4 or 5 were considered positive, based on the Node-RADS reporting guidelines. The categorization of score 3 varies based on the type of primary malignancy involved. So far, no studies have clearly determined whether lymph nodes with a score of 3 should be considered positive or negative LNM in lung cancer. Therefore, both Node-RADS scores of 3 and 4 were used as criteria for identifying positive lymph nodes in this study. Additionally, we defined a metric called LNM-Station to reflect the number of stations with positive lymph nodes. Specifically, LNM-Station3 was defined as the number of lymph node stations in which at least one lymph node had a Node-RADS score ≥3. In contrast, LNM-Station4 represented the count of lymph node stations, among which each station contained at least one lymph node with a Node-RADS score ≥4.

Figure 2 Node-RADS scoring flowchart and assessment principle for lymph node. Node-RADS, Node Reporting and Data System.

Figure 3 Examples of Node-RADS scoring. White boxes highlight the magnified regions and arrows point to the selected lymph nodes. (A,B) A 57-year-old male with TNM 4a stage SCLC. The selected lymph node in station 4R measures 9 mm × 6 mm with homogeneous texture, smooth border, and kidney-bean-like shape. (C,D) A 72-year-old male with TNM 4a stage SCLC. The selected lymph node in station 2R measures 9 mm × 9 mm with homogeneous texture, smooth border and spherical shape. (E,F) A 58-year-old male with TNM 4b stage SCLC. The selected lymph node in station 5 measures 20 mm × 13 mm with heterogeneous texture, smooth border, and oval shape. (G,H) A 66-year-old male with TNM 4b stage SCLC. The selected lymph node in station 7 measures 31 mm × 22 mm with a heterogeneous texture and irregular border. (I,J) A 45-year-old male with TNM 4b stage SCLC. The selected lymph node in station 4R measures 28 mm × 25 mm with gross necrosis and irregular border. Node-RADS, Node Reporting and Data System; SCLC, small cell lung cancer; TNM, tumor node metastasis.

Sensitivity analyses

To evaluate potential confounding and immortal-time bias related to treatment variables, we performed the following sensitivity analyses: (I) restriction to patients who received at least 2 treatment cycles; (II) landmark analysis at 6 months, including only patients who survived beyond this time; and (III) categorization of treatment cycles (1–2, 3–4, ≥5 cycles). All prognostic models were refitted under each scenario.

Statistical analysis

Statistical analyses were performed using R (version 4.3.0, R Foundation for Statistical Computing, Vienna, Austria) and SPSS (version 25.0, IBM Corp., Armonk, NY, USA), with P<0.05 as the significance threshold. The reliability of the Node-RADS scores within and between observers was assessed through the intraclass correlation coefficient (ICC).

Candidate variables were first screened by univariate Cox regression (P<0.05). Variables retained then underwent multicollinearity assessment [generalized variance inflation factor (GVIF)]; variables with GVIF >5 were considered to exhibit substantial collinearity and were excluded. The remaining variables were subsequently included in the final multivariable Cox regression model. Pearson correlation analysis was performed between key variables of interest. The discriminative ability of each final model was quantified using Harrell’s concordance index (C-index). To correct for potential overfitting, an optimism-adjusted C-index was obtained via 1,000 bootstrap resamples. Model calibration was evaluated using calibration plots at 1-, 2-, and 3-year time points. Time-dependent receiver operating characteristic (ROC) curves were generated to assess predictive accuracy across follow-up, and decision curve analysis (DCA) was performed to quantify the clinical net benefit of each model. Continuous net reclassification improvement (cNRI) was calculated to quantify the incremental prognostic value of Node-RADS metrics.

Results

Patient characteristics

The study cohort ultimately included 297 patients, with a gender distribution of 271 men and 26 women. The mean age was 61.0±7.8 years. According to the NCCN guidelines, the patients underwent 1–6 cycles of chemoimmunotherapy, including EC + atezolizumab (n=142), EC + durvalumab (n=66), EP + atezolizumab (n=32), and EP + durvalumab (n=57). The other clinical characteristics of all included patients are detailed in Table 1. Of the 297 patients, definitive vital status (including date of death) was ascertained for 262 (88.2%), and 35 (11.8%) patients were lost to follow-up. The median follow-up time was 25.7 months [95% confidence interval (CI): 21.9–29.5], during which 194 patients (65.32%) died. The median survival was 14.7 months (95% CI: 13.7–16.5).

Interobserver and intraobserver reliability

The inter-observer ICC of Maximum Node-RADS scores at the patient level was 0.796 (95% CI: 0.774–0.813). For intraobserver agreement, Reader A rescored all cases after 3 months, yielding an ICC of 0.825 (95% CI: 0.802–0.884), indicating good reliability.

Distribution of pretherapeutic Node-RADS scores

To characterize the lymph nodal burden in our cohort, we described the frequency and distribution of Node-RADS scores. At the patient level, 9.4% (28/297), 18.9% (56/297), and 56.2% (167/297) of patients had lymph nodes with a maximum Node-RADS score of 3, 4, and 5, respectively. The station-level distribution is shown in Table S1. The lymph node stations most frequently involved with scores ≥3 were 4R (51.8%), 7 (50.5%), and 2R (42.8%).

Pretherapeutic maximum Node-RADS score and LNM-Station associated with OS

The cut-off value of cN stage for predicting OS was 1, which divided the enrolled patients into the low (0–1) and high (≥2) cN groups. Similarly, the cut-off value of maximum Node-RADS score was 4, classifying patients into the low [0–4] and high [5] Node-RADS groups. The cut-off value of LNM-Station3 was 3, which classified patients into the limited (0–3) and extensive (≥4) LNM station groups. Similarly, the cut-off value of LNM-Station4 was 2, which also classified patients into the limited (0–2) and extensive (≥3) LNM station groups.

The limited LNM-Station3 group showed a higher 3-year OS rate than the extensive group (36.09% vs. 1.86%, P<0.001) (Figure 4). Similarly, the limited LNM-Station4 group showed a higher 3-year OS rate than the extensive group (38.77% vs. 0.00%, P<0.001; Figure 4). The low maximum Node-RADS score (47.98% vs. 5.77%, P<0.001) and low cN stage (52.95% vs. 16.23%, P<0.001) were also associated with a higher 3-year OS rate (Figure 4).

Figure 4 Kaplan-Meier survival curves of overall survival. Kaplan-Meier survival curves of overall survival stratified by clinical N stage (A), maximum Node-RADS score (B), LNM-Station3 (C), and LNM-Station4 (D). CI, confidence interval; HR, hazard ratio; Node-RADS, Node Reporting and Data System; LNM-Station, the number of positive lymph node stations involved.

Performance of prognostic models for OS

To predict OS, the Clinical-NR3 model was developed using the pretherapeutic clinical variables and Node-RADS features (the maximum Node-RADS score and LNM-Station3). Additionally, the Clinical-NR4 prognostic model was constructed based on the clinical variables, maximum Node-RADS score, and LNM-Station4. The Clinical-cN prognostic model was also developed as a comparison, using clinical variables and cN stage. Univariate Cox regression analysis indicated that performance status, Maximum Node-RADS score, LNM-Station3, LNM-Station4, liver metastasis, bone metastasis, number of metastatic organs, thoracic radiotherapy, and the number of chemoimmunotherapy cycles were significantly correlated with OS (Tables 2-4). Node-RADS showed a positive correlation with LNM-Station3 (Pearson r=0.631) and LNM-Station4 (Pearson r=0.627, Figure S1). Multicollinearity was assessed for all models, with adjusted variance inflation factors (VIFs) ranging from 1.028 to 1.516 (Clinical-NR4), 1.025 to 1.572 (Clinical-NR3), and 1.010 to 1.512 (Clinical-cN). All GVIF and adjusted VIF values were well below the conventional threshold of 5, indicating that multicollinearity among the predictor variables was negligible (Table S2). Multivariate Cox regression analysis demonstrated that performance status, liver metastasis, thoracic radiotherapy, and the number of chemoimmunotherapy cycles were independent prognostic factors in all three predicting models for OS (Tables 2-4). Simultaneously, LNM-Station3 and LNM-Station4 acted as independent predictors for the Clinical-NR3 and Clinical-NR4 models, respectively (Tables 2,3). The cN stage was not significantly associated with OS in the Clinical-cN model (P>0.05, Table 4).

The Clinical-NR4 model exhibited an increase in C-index compared to the Clinical-cN model for OS [0.759 (95% CI: 0.724–0.793) vs. 0.726 (95% CI: 0.690–0.762), P=0.009]. The C-index [0.748 (95% CI: 0.712–0.784)] of the Clinical-NR3 model for OS did not show a statistically significant difference from that of the Clinical-NR4 (P=0.093) and Clinical-cN (P=0.059) models. After optimism correction, the C-indices were 0.690 (95% CI: 0.611–0.769) for the Clinical-NR3 model, 0.636 (95% CI: 0.511–0.761) for the Clinical-cN model, and 0.703 (95% CI: 0.627–0.779) for the Clinical-NR4 model, as detailed in Table S3.

Furthermore, the Clinical-NR4 prognostic model demonstrated superior predictive performance compared with the Clinical-cN model across all time points (Figure 5A-5C): areas under the curve (AUCs) were 0.801 versus 0.785 (P=0.384) at 1 year, 0.858 versus 0.772 (P=0.003) at 2 years, and 0.867 versus 0.687 (P=0.001) at 3 years. Moreover, the time-dependent ROC AUC values exhibited a progressive increase over time, indicating that the Clinical-NR4 model offers enhanced predictive accuracy for long-term survival in SCLC (Figure 5D).

Figure 5 ROC curves. ROC curves of the Clinical-NR4 and Clinical-cN models for 1-, 2-, 3-year OS (A-C). Time-dependent ROCs of Clinical-NR4 and Clinical-cN models for OS (D). AUC, area under the curve; C-index, concordance index; CI, confidence interval; OS, overall survival; ROC, receiver operating characteristic.

To further evaluate the added prognostic value of the Node-RADS-based model, we calculated the cNRI for the Clinical-NR4 model compared with the Clinical-cN model. The cNRI values for 1-, 2-, and 3-year OS were 0.500, 0.502, and 0.503, respectively, indicating an improvement in individual risk ranking. Similarly, the Clinical-NR3 model showed a cNRI of 0.480–0.483 across the same time points (Table S4).

Based on the predicted risk scores from the Clinical-NR4 model, we established preliminary clinical thresholds as defined by the tertiles (33rd and 66th percentiles). Patients were categorized as low-risk (score ≤0.763), intermediate-risk (score >0.763 and ≤1.878), and high-risk (score >1.878). These groups exhibited significantly different median OS (48.5, 15.2, and 8.6 months, respectively, log-rank P<0.001), demonstrating potential risk stratification (Figure S2).

The nomograms of the Clinical-NR4 for OS are depicted in Figure 6A. The calibration curves indicated excellent agreement between the predictions of the Clinical-NR4 model and actual observations at 1, 2, and 3 years (Figure 6B) and the DCA showed that the Clinical-NR4 model provided a greater net benefit for predicting OS compared to the Clinical-cN model (Figure 6C).

Figure 6 Nomogram, calibration curves, and DCA curves of the Clinical-NR4 model. Nomogram of the Clinical-NR4 model for overall survival (A), calibration curve of the Clinical-NR4 model at 1-, 2-, and 3-years overall survival (B), and DCA curves of the Clinical-NR4 and Clinical-cN models (C). DCA, decision curve analysis; LNM-Station, the number of positive lymph node stations involved; Node-RADS, Node Reporting and Data System.

Sensitivity analyses addressing treatment-related biases

We applied the three analytical strategies described above, which respectively included 269 patients (≥2 cycles), 296 patients (landmark ≥6 months), and the full cohort with categorized cycle numbers. As shown in Table S5, the Clinical-NR4 model consistently demonstrated superior discrimination compared to the Clinical-cN model across all scenarios (original, restricted cycles, landmark, and categorized cycles), with C-indexes between 0.725 and 0.759 and all corresponding P values <0.05. Although the Clinical-NR3 model also tended to outperform the Clinical-cN model in most settings, this difference was not statistically significant in the landmark analysis (P=0.278).

Discussion

SCLC is associated with high mortality and recurrence rates, despite the survival advantages offered by chemoimmunotherapy. LNM plays a crucial role in the unfavorable prognosis of SCLC. In this research, we retrospectively examined the correlations between pretherapeutic Node-RADS-related CT features and OS outcome of SCLC. Clinical factors were separately combined with the cN stage and Node-RADS-related features to construct two types of predictive models. Our results revealed that the model incorporating Node-RADS features offered better predictive performance for SCLC prognosis. We calculated the number of regional node stations with positive LNM based on different Node-RADS score thresholds, and subsequently constructed two Node-RADS-related predictive models. When using a Node-RADS score of 4 as the threshold, the constructed model demonstrated superior predictive performance.

The reporting and data systems have been effective in staging and prognostic predictions for various cancers (27,28). Node-RADS was recently introduced to standardize lymph node evaluations, enhancing diagnostic accuracy across different cancer types (18-24). To our knowledge, there have been no studies on using Node-RADS for predicting lung cancer prognosis, making this study potentially the first. Pretreatment Node-RADS-related CT metrics (maximum Node-RADS score and LNM-Station) were associated with OS in SCLC, as demonstrated by Kaplan-Meier and Cox proportional hazards analyses. The lower maximum Node-RADS scores and LNM-Station values are linked to improved survival outcomes. This could offer new perspectives on using Node-RADS in cancer prognosis studies, including SCLC.

In this study, a maximum Node-RADS score of 5 at the patient level was linked to a worse prognosis in SCLC. The Node-RADS system combines information on lymph node size and morphological abnormalities, including texture, border, and shape. Its scoring criteria are cumulative, meaning that larger lymph nodes with more abnormalities receive higher scores, suggesting a higher chance of malignancy. A Node-RADS score of 5 was the highest possible score, which indicates the most morphological abnormalities and/or the largest size. The size of lymph nodes is widely acknowledged as an important prognostic factor in cancers (29,30). Furthermore, necrosis, spherical shape, and irregular borders of lymph nodes frequently imply structural disruption and extranodal extension, which generally indicate more advanced tumor progression and worse survival outcomes (31-33). In the setting of SCLC, these factors might partly indicate increased aggressiveness or a more advanced stage of the disease (34), leading to reduced efficacy of chemoimmunotherapy and a consequently poorer prognosis.

Similar to the maximum Node-RADS score, the presence of multiple positive lymph node stations, defined as either LNM-Station4 ≥3 or LNM-Station3 ≥4, was also associated with poorer survival outcomes in SCLC patients. The occurrence of multiple positive stations indicates a wider spread of metastatic lymph nodes, often indicating advanced tumor progression. Earlier research has demonstrated that the number of involved lymph node stations correlates with prognosis in non-small cell lung cancer (15,35), consistent with this research. Additionally, the recent AJCC 9th edition TNM staging system for lung cancer now subdivides stage N2 into N2a (single-station) and N2b (multi-station) metastasis (36).

Interestingly, among the two types of Node-RADS-related CT metrics investigated in this research, only LNM-Station entered the final multivariable OS model as an independent predictor, implying greater prognostic utility than the maximum Node-RADS score.

Although previous research on Node-RADS score indicates that a threshold of 2 can effectively distinguish pathological negative from positive of mediastinal lymph nodes in lung cancer patients at the patient level (18), there remains a lack of studies investigating suitable diagnostic LNM thresholds at the lymph node level. Our study shows that the Clinical-NR4 model demonstrated incremental predictive value for OS in SCLC, compared with the Clinical-cN model (P=0.009). After optimism correction, the Clinical-NR4 model maintained a corrected C-index of 0.703, demonstrating better discrimination and greater stability compared with the model based solely on cN stage. The cNRI values provide additional supportive evidence for the incremental prognostic value of the Node-RADS-based model beyond conventional cN staging. Moreover, the proposed risk stratification tertiles (thresholds: ≤0.763, 0.763–1.878, >1.878) effectively identify patients with poor prognosis, who may benefit from more intensive monitoring or adjunctive therapies.

The Clinical-NR4 model also performed better than the Clinical-NR3 model, with this difference nearing statistical significance (P=0.093). Given that LNM-Station3 and LNM-Station4 denote the number of positive LNM stations at different Node-RADS score thresholds, our findings suggest that using a score of 4 as the threshold for positive LNM could be more beneficial for guiding clinical treatment decisions. Our findings require further investigation to validate their clinical value.

After conducting sensitivity analyses to account for potential treatment-related biases (including restriction by cycle number, landmark analysis, and cycle categorization), the Clinical-NR4 model maintained stable discriminative performance (C-index ≥0.725). These findings support the robustness of the association between the Node-RADS-based model (threshold =4) and OS, although residual time-dependent bias cannot be fully excluded.

In the Clinical-cN model of this study, the cN stage was not identified as an independent predictor of OS in SCLC, which differs from previous reports (7-9) and is similar to prior observations (37). This discrepancy may be attributable to differences in cohort composition. Performance status, liver metastasis, bone metastasis, and number of metastatic organs were negatively related to the prognosis of SCLC in the present study, similar to previous reports (38,39). A higher performance status score, more liver and bone metastases, and a greater number of metastatic organs indicate poorer physical function and a higher tumor burden, which undoubtedly suggests a worse prognosis. In this study, both thoracic radiotherapy and the number of chemoimmunotherapy cycles showed a positive correlation with the OS of SCLC, aligning with findings from previous reports (40,41).

This study had some limitations. Firstly, the use of post-baseline treatment intensity variables may introduce the potential risk of treatment confounding and immortal time bias. To address this concern, we conducted multiple sensitivity analyses (including landmark analysis, restriction to patients receiving ≥2 cycles, and categorization of cycle numbers). These analyses yielded results consistent with the primary findings, although residual time-dependent bias cannot be entirely ruled out. Secondly, the single-center, retrospective design is inherently subject to selection bias and limits the generalizability of our findings. Thirdly, the lack of external validation and possible variability in CT scanners across institutions may affect the reproducibility of Node-RADS assessments. Fourthly, the cut-off values for the maximum Node-RADS score and LNM-Station metrics were derived within the same dataset. Although internal validation via bootstrapping supports the robustness of model performance, these data‑driven thresholds may be cohort-specific and could exhibit instability in external populations. Therefore, external validation in independent, multi-center cohorts is warranted to confirm the generalizability of both the proposed thresholds and the overall prognostic model. Finally, for patients with central lung tumors, the primary lesion is often anatomically indistinct from adjacent hilar nodes on CT imaging, making reliable delineation and separate Node-RADS scoring technically unfeasible. Consequently, ipsilateral hilar lymph nodes were excluded from our LNM-Station evaluation. Although this ensures the specificity of scoring for non-contiguous nodes, it potentially leads to an underestimation of the true nodal burden in cases of central tumors. Future studies incorporating more advanced imaging segmentation techniques or revised criteria for central tumor lymph node assessment are warranted to address this challenge.

Conclusions

The baseline LNM-Station based on Node-RADS is an effective prognostic indicator for extensive-stage SCLC. When predicting the prognosis of SCLC, the baseline Clinical-NR models provide superior predictive capabilities compared to the Clinical-cN model. These models may help to identify SCLC patients at a higher risk of adverse outcomes, which may facilitate timely interventions and more targeted monitoring in clinical settings.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1835/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1835/dss

Funding: This work was supported by Hunan Provincial Natural Science Foundation of China (No. 2025JJ80823) and High-Level Talent Support Program of Hunan Cancer Hospital (No. 20250806-1002).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1835/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Hunan Cancer Hospital (No. KY2024532). Informed consent was waived in this retrospective study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Park H, Tseng SC, Sholl LM, Hatabu H, Awad MM, Nishino M. Molecular Characterization and Therapeutic Approaches to Small Cell Lung Cancer: Imaging Implications. Radiology 2022;305:512-25. [Crossref] [PubMed]
2. Paz-Ares L, Dvorkin M, Chen Y, Reinmuth N, Hotta K, Trukhin D, et al. Durvalumab plus platinum-etoposide versus platinum-etoposide in first-line treatment of extensive-stage small-cell lung cancer (CASPIAN): a randomised, controlled, open-label, phase 3 trial. Lancet 2019;394:1929-39. [Crossref] [PubMed]
3. Goldman JW, Dvorkin M, Chen Y, Reinmuth N, Hotta K, Trukhin D, et al. Durvalumab, with or without tremelimumab, plus platinum-etoposide versus platinum-etoposide alone in first-line treatment of extensive-stage small-cell lung cancer (CASPIAN): updated results from a randomised, controlled, open-label, phase 3 trial. Lancet Oncol 2021;22:51-65. [Crossref] [PubMed]
4. Horn L, Mansfield AS, Szczęsna A, Havel L, Krzakowski M, Hochmair MJ, et al. First-Line Atezolizumab plus Chemotherapy in Extensive-Stage Small-Cell Lung Cancer. N Engl J Med 2018;379:2220-9. [Crossref] [PubMed]
5. Chen T, Wang M, Chen Y, Cao Y, Liu Y. Advances in predictive biomarkers associated with immunotherapy in extensive-stage small cell lung cancer. Cell Biosci 2024;14:117. [Crossref] [PubMed]
6. Longo V, Catino A, Montrone M, Pizzutilo P, Annese T, Pesola F, Marech I, Cassiano S, Ribatti D, Galetta D. What Are the Biomarkers for Immunotherapy in SCLC? Int J Mol Sci 2021;22:11123. [Crossref] [PubMed]
7. Shan Q, Shi J, Wang X, Guo J, Han X, Wang Z, Wang H. A new nomogram and risk classification system for predicting survival in small cell lung cancer patients diagnosed with brain metastasis: a large population-based study. BMC Cancer 2021;21:640. [Crossref] [PubMed]
8. Li Q, Zhao Y, Xu Z, Ma Y, Wu C, Shi H. Development and validation of prognostic models for small cell lung cancer patients with liver metastasis: a SEER population-based study. BMC Pulm Med 2024;24:13. [Crossref] [PubMed]
9. Deng L, Zhou Z, Xiao Z, Chen D, Feng Q, Liang J, Lv J, Wang X, Bi N, Wang X, Zhang T, Wang W, Wang L. Impact of thoracic radiation therapy after chemotherapy on survival in extensive-stage small cell lung cancer: A propensity score-matched analysis. Thorac Cancer 2019;10:799-806. [Crossref] [PubMed]
10. Nomura K, Aokage K, Nakai T, Sakashita S, Miyoshi T, Tane K, Samejima J, Suzuki K, Tsuboi M, Ishii G. Prognostic impact of extranodal extension in patients with pN1-N2 lung adenocarcinoma. J Cancer Res Clin Oncol 2021;147:3699-707. [Crossref] [PubMed]
11. Luchini C, Veronese N, Nottegar A, Cheng M, Kaneko T, Pilati C, Tabbò F, Stubbs B, Pea A, Bagante F, Demurtas J, Fassan M, Infante M, Cheng L, Scarpa A. Extranodal extension of nodal metastases is a poor prognostic moderator in non-small cell lung cancer: a meta-analysis. Virchows Arch 2018;472:939-47. [Crossref] [PubMed]
12. Li J, Zhao Q, Zhang Y, Li H, Ruan G, Liu L, Yan Y, Cui C. Prognostic value of quantitative cervical nodal necrosis burden on MRI in nasopharyngeal carcinoma and its role as a stratification marker for induction chemotherapy. Eur Radiol 2022;32:7710-21. [Crossref] [PubMed]
13. Rollvén E, Blomqvist L, Öistämö E, Hjern F, Csanaky G, Abraham-Nordling M. Morphological predictors for lymph node metastases on computed tomography in colon cancer. Abdom Radiol (NY) 2019;44:1712-21. [Crossref] [PubMed]
14. Sayar A, Turna A, Kiliçgün A, Solak O, Urer N, Gürses A. Prognostic significance of surgical-pathologic multiple-station N1 disease in non-small cell carcinoma of the lung. Eur J Cardiothorac Surg 2004;25:434-8. [Crossref] [PubMed]
15. Ichinose Y, Kato H, Koike T, Tsuchiya R, Fujisawa T, Shimizu N, Watanabe Y, Mitsudomi T, Yoshimura M, Tsuboi MJapanese Clinical Oncology Group. Completely resected stage IIIA non-small cell lung cancer: the significance of primary tumor location and N2 station. J Thorac Cardiovasc Surg 2001;122:803-8. [Crossref] [PubMed]
16. Ganti AKP, Loo BW, Badiyan S, Chaudhuri AA, Chin EK, Dowlati A, et al. NCCN Clinical Practice Guidelines in Oncology: Small Cell Lung Cancer. Version 4.2025.
17. Elsholtz FHJ, Asbach P, Haas M, Becker M, Beets-Tan RGH, Thoeny HC, Padhani AR, Hamm B. Introducing the Node Reporting and Data System 1.0 (Node-RADS): a concept for standardized assessment of lymph nodes in cancer. Eur Radiol 2021;31:6116-24. [Crossref] [PubMed]
18. Meyer HJ, Schnarkowski B, Pappisch J, Kerkhoff T, Wirtz H, Höhn AK, Krämer S, Denecke T, Leonhardi J, Frille A. CT texture analysis and node-RADS CT score of mediastinal lymph nodes - diagnostic performance in lung cancer patients. Cancer Imaging 2022;22:75. [Crossref] [PubMed]
19. Niu Y, Wen L, Yang Y, Zhang Y, Fu Y, Lu Q, Wang Y, Yu X, Yu X. Diagnostic performance of Node Reporting and Data System (Node-RADS) for assessing mesorectal lymph node in rectal cancer by CT. BMC Cancer 2024;24:716. [Crossref] [PubMed]
20. Yang X, Yang J, Li J, Leng J, Qiu Y, Ma X. Diagnostic Performance of Node Reporting and Data System Magnetic Resonance Imaging Score in Detecting Metastatic Cervical Lymph Nodes of Nasopharyngeal Carcinoma. Clin Med Insights Oncol 2024;18:11795549241231564. [Crossref] [PubMed]
21. Sun Y, Xiao H, Wen L, Xiang W, Luo X, Yang X, Chen L, Yang Y, Zhang Y, Yu S, Yu X. Value of Post-NAC CT-based Node-RADS score for Predicting residual lymph node metastasis and survival outcome of locally advanced gastric cancer. Abdom Radiol (NY) 2025;50:3936-49. [Crossref] [PubMed]
22. Pediconi F, Maroncelli R, Pasculli M, Galati F, Moffa G, Marra A, Polistena A, Rizzo V. Performance of MRI for standardized lymph nodes assessment in breast cancer: are we ready for Node-RADS? Eur Radiol 2024;34:7734-45. [Crossref] [PubMed]
23. Li X, Lin D, Xiong Y, Lin W, Zhang S, Huang S, Zhou J, Dai C. Node-RADS category on preoperative CT predicts prognosis in patients with papillary renal cell carcinoma. Eur Radiol 2025;35:5480-91. [Crossref] [PubMed]
24. Sun Y, Wen L, Xiang W, Luo X, Chen L, Yang X, Yang Y, Zhang Y, Yu S, Xiao H, Yu X. Added value of pretreatment CT-based Node-RADS score for predicting survival outcome of locally advanced gastric cancer: compared with clinical N stage. BMC Cancer 2025;25:598. [Crossref] [PubMed]
25. Li Y, Deng J, Ma X, Li W, Wang Z. Diagnostic accuracy of CT and PET/CT radiomics in predicting lymph node metastasis in non-small cell lung cancer. Eur Radiol 2025;35:1966-79. [Crossref] [PubMed]
26. Khorrami M, Mutha P, Barrera C, Viswanathan VS, Ardeshir-Larijani F, Jain P, Higgins K, Madabhushi A. AI-based radiomic features predict outcomes and the added benefit of chemoimmunotherapy over chemotherapy in extensive stage small cell lung cancer: A multi-institutional study. Cancer Lett 2025;628:217872. [Crossref] [PubMed]
27. Fujimoto A, Sakamoto S, Horikoshi T, Zhao X, Yamada Y, Rii J, Takeuchi N, Imamura Y, Sazuka T, Matsusaka K, Ikeda JI, Ichikawa T. Tumor localization by Prostate Imaging and Reporting and Data System (PI-RADS) version 2.1 predicts prognosis of prostate cancer after radical prostatectomy. Sci Rep 2023;13:10079. [Crossref] [PubMed]
28. Kim BK, Ryu JM, Oh SJ, Han J, Choi JE, Jeong J, Suh YJ, Lee J, Sun WY. Comparison of clinicopathological characteristics and prognosis in breast cancer patients with different Breast Imaging Reporting and Data System categories. Ann Surg Treat Res 2021;101:131-9. [Crossref] [PubMed]
29. Sugimura K, Miyata H, Kanemura T, Takeoka T, Sugase T, Yamamoto M, Shinnno N, Hara H, Omori T, Motoori M, Ohue M, Yano M. Clinical Impact of Metastatic Lymph Node Size on Therapeutic Effect and Prognosis in Patients with Esophageal Squamous Cell Carcinoma Who Underwent Preoperative Chemotherapy Followed by Esophagectomy. Ann Surg Oncol 2023;30:4193-202. [Crossref] [PubMed]
30. Huang CL, Chen Y, Guo R, Mao YP, Xu C, Tian L, Liu LZ, Lin AH, Sun Y, Ma J, Tang LL. Prognostic value of MRI-determined cervical lymph node size in nasopharyngeal carcinoma. Cancer Med 2020;9:7100-6. [Crossref] [PubMed]
31. Bin Y, Meng Z, Huang LL, Hu XY, Song JM, Xie YT, Kang M, Wang RS. Prognostic value of the cervical lymph node necrosis ratio in nasopharyngeal carcinoma. Radiother Oncol 2022;177:185-90. [Crossref] [PubMed]
32. Jang S, Lee S, Chung JH, Lee KW, Lee KH. Radiologic Extranodal Extension of Metastatic Lymph Nodes in Patients With Non-Small Cell Lung Cancer: Prognostic Utility and Diagnostic Performance. AJR Am J Roentgenol 2023;221:471-84. [Crossref] [PubMed]
33. Sugai K, Sekine Y, Kawamura T, Yanagihara T, Saeki Y, Kitazawa S, Kobayashi N, Kikuchi S, Goto Y, Ichimura H, Sato Y. Sphericity of lymph nodes using 3D-CT predicts metastasis in lung cancer patients. Cancer Imaging 2023;23:124. [Crossref] [PubMed]
34. Rice SR, Molitoris JK, Vyfhuis MAL, Edelman MJ, Burrows WM, Feliciano J, Nichols EM, Suntharalingam M, Donahue J, Carr SR, Friedberg J, Badiyan S, Simone CB 2nd, Feigenberg SJ, Mohindra P. Lymph Node Size Predicts for Asymptomatic Brain Metastases in Patients With Non-small-cell Lung Cancer at Diagnosis. Clin Lung Cancer 2019;20:e107-14. [Crossref] [PubMed]
35. Demir A, Turna A, Kocaturk C, Gunluoglu MZ, Aydogmus U, Urer N, Bedirhan MA, Gurses A, Dincer SI. Prognostic significance of surgical-pathologic N1 lymph node involvement in non-small cell lung cancer. Ann Thorac Surg 2009;87:1014-22. [Crossref] [PubMed]
36. Detterbeck FC, Woodard GA, Bader AS, Dacic S, Grant MJ, Park HS, Tanoue LT. The Proposed Ninth Edition TNM Classification of Lung Cancer. Chest 2024;166:882-95.
37. Wu J, Zhou Y, Xu C, Yang C, Liu B, Zhao L, Song J, Wang W, Yang Y, Liu N. Effectiveness of CT radiomic features combined with clinical factors in predicting prognosis in patients with limited-stage small cell lung cancer. BMC Cancer 2024;24:170. [Crossref] [PubMed]
38. Ma M, Wang M, Xu Y, Hu K, Liu H, Li L, Zhong W, Zhang L, Zhao J, Wang H. First-line chemotherapy and its survival analysis of 394 patients with extensive-stage small cell lung cancer in a single institute. Zhongguo Fei Ai Za Zhi 2014;17:8-14. [Crossref] [PubMed]
39. Slotman BJ, Faivre-Finn C, van Tinteren H, Keijser A, Praag J, Knegjens J, Hatton M, van Dam I, van der Leest A, Reymen B, Stigt J, Haslett K, Tripathi D, Smit EF, Senan S. Which patients with ES-SCLC are most likely to benefit from more aggressive radiotherapy: A secondary analysis of the Phase III CREST trial. Lung Cancer 2017;108:150-3. [Crossref] [PubMed]
40. Han J, Fu C, Li B. Clinical outcomes of extensive-stage small cell lung cancer patients treated with thoracic radiotherapy at different times and fractionations. Radiat Oncol 2021;16:47. [Crossref] [PubMed]
41. Hoffmann E, De-Colle C, Potkrajcic V, Baumann D, Spengler W, Gani C, Utz D. Is consolidative thoracic radiotherapy of extensive-stage small cell lung cancer still beneficial in the era of immunotherapy? A retrospective analysis. Strahlenther Onkol 2023;199:668-75. [Crossref] [PubMed]