The screening of optimal primary tumor resection candidates in patients with small cell lung cancer: a population-based predictive model

Introduction

Lung cancer is one of the most common malignant cancers, with almost 2.5 million new cases and over 1.8 million deaths worldwide (1). Among lung cancers, small cell lung cancer (SCLC) poses an intractable clinical problem due to its aggressive pathology, accounting for about 14–16% of total morbidity and only 6% of 5-year survival rate (2-4). In the early treatment of SCLC, radiotherapy and chemotherapy treatments are sensitive. Therefore, chemoradiotherapy, targeted therapy, and immunotherapy are the best treatment options for these patients. However, there is a typically only a brief period of response in SCLC (5). Local and even distant metastases are common phenomena for SCLC. In 70% of cases, metastasis has already occurred at diagnosis (6,7). In these cases, the main goal of SCLC management is to control and alleviate tumor progression.

Surgical treatment of SCLC has always posed a difficult choice for clinicians. In clinical guidelines, stage I–IIA SCLC patients are regarded as a surgical evaluation population. A retrospective study showed that stage I–IIA SCLC patients could benefit from surgical resection (8). Moreover, some large capacity medical centers performed surgery for stage III SCLC patients and found that a node-spreading pattern is a vital prognostic factor (9,10). Other studies reported that surgery significantly improved long-term survival of stage III SCLC patients (11-13). Therefore, surgical intervention is limited to patients with early-stage SCLC, but more recent studies suggest that surgery might be more widely used in SCLC (14-16).

To facilitate screening of SCLC patients suitable for surgery, we have built a predictive model using SCLC patients from the Surveillance, Epidemiology, and End Results (SEER) database based on reliable clinicopathological data related to prognosis. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1419/rc).

Methods

Study population

Patients with the primary SCLC were acquired from SEER Research Plus Data 18 population-based registries (https://seer.cancer.gov/). In total, 73,683 patients with the first primary SCLC were included in the SEER database between 2014 and 2018. The matched clinicopathological data were acquired from the database, including patient diagnose information, sex, age, race, primary site, histology type, grade, TNM stage (T, size of the tumor; N, extent of regional lymph node involvement; M, presence of metastasis), T stage, tumor size, N stage, M stage, surgical methods, chemotherapy, radiotherapy, survival status, cause of death, and survival month. The TNM stage of patients was reclassified based on the 8^th edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual (17).

Eligible patients were included according to the following criteria. The inclusion criteria: (I) patients diagnosed with the first primary SCLC (code: 8041/3–8045/3) in 2004–2018; (II) surgery (code: 0) and non-surgery (code: 21–22, 30–70); (III) age over 18 years. The exclusion criteria: (I) TNM stage is unclear to reclassified or unknown; (II) survival time is 0 or unknown; (III) age, race, and stage are unknown; (IV) tumor position is unknown. The detailed screening process is shown in Figure 1.

Figure 1 The flowchart of screening out patients with small cell lung cancer in this study. SEER, Surveillance, Epidemiology, and End Results; TNM, T, size of the tumor; N, extent of regional lymph node involvement; M, presence of metastasis; CSS, cancer-special survival.

Statistical analyses

Pearson’s Chi-squared test and Student’s t-test were respectively used to compare the basic information and clinicopathologic characteristics of patients in the non-surgery with surgery groups. To balance these characteristics between two groups, the “MatchIt” R package was used to conduct propensity score matching (PSM). The propensity scores were calculated based on variables possibly related to outcomes of treatment. These variables included patient sex, age, race, primary site, histology type, grade, T stage, N stage, M stage, surgery to metastasis site, radiotherapy, and chemotherapy. Cases were 1:1 matched based on the propensity score within a caliper of 0.05. Then, the Kaplan-Meier survival curves were drawn to estimate the overall survival (OS) and cancer-specific survival (CSS) time, and the log-rank test calculated differences between the two groups. Univariate and multivariate Cox proportional hazard regression was used to distinguish independent prognostic factors. Statistical analyses were performed using R software v.4.1.2 and tests were two-sided; P<0.05 was set significant level.

Construction and validation of nomogram

The analyses showed that SCLC patients undergoing surgical resection had longer CSS time than matched SCLC patients who did not undergo surgical resection. The median CSS time was 37.00 vs. 16.00 months, respectively. In the surgery group, patients surviving longer than 16 months were assigned to a benefit group; the remaining patients were assignment to the non-benefit group. The surgical patients were randomly divided into two sets at a 7:3 ratio: 70% were allocated to the training set and 30% to the testing set.

Univariate and multivariable logistic analyses were performed to distinguish independent factors of surgical benefit. The multivariable logistic regression model was built based on preoperative variables related to surgical benefit. These variables included N stage, T stage, primary site of tumor, age, M stage, and sex. According to this logistic regression model, a nomogram was built to distinguish optimal primary tumor resection candidates from SCLC patients using the “rms” package (18).

To test the discrimination and stability of this nomogram, the receiver operating characteristic (ROC) curves, decision curve analysis (DCA) plots, and calibration plots were performed in the training and testing cohorts. Meanwhile, after PSM, we redivided the SEER database into three groups according to this predictive model: the surgery & benefit group, the surgery & non-benefit group, and the non-surgery group. Lastly, Kaplan-Meier survival curve was constructed to compare the CSS time. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

Results

Patient characteristics

After screening, 35,214 eligible SCLC patients were included in final study population from the SEER database, with a median follow-up time of 9 months (Figure 1). Among them, only 1,364 (3.9%) patients underwent primary tumor resection. The obvious difference was observed in the basic information and clinicopathologic characteristics of patients between the two groups (Table 1). Surgery was commonly performed in patients who were White and female. Tumors in these patients were highly differentiated, had early TNM stage, and had combined small cell lung cancer (cSCLC) pathological features. However, patients not undergoing surgery had more exposure to radiation and chemotherapy. A 1:1 PSM was performed based on patient age, race, sex, primary site, histology type, grade, T stage, N stage, M stage, surgery to metastasis site, radiotherapy, and chemotherapy. After PSM, significant differences were not found in the basic information and clinicopathologic characteristics of patients between the two groups (Table 1 and Figure S1). The standard mean difference (SMD) for the matched groups was presented in the Table S1. Additionally, the “love plot” showed the SMD before and after matching, visually illustrating the matching effect (Figure S2).

Impact of surgery on survival outcome in SCLC patients

Kaplan-Meier curves suggested that patients who underwent surgery had longer survival time than those who did not undergo surgery (Figure 2A,2B). After PSM, the results of survival analyses were consistent with those before PSM (Figure 2C,2D). Meanwhile, the 1-, 3-, 5-year OS rates respectively were 53.25%, 21.48%, and 15.80% in the non-surgery group, compared with 77.77%, 45.79%, and 35.70% in the surgery group. And, the 1-, 3-, 5-year CSS rates in the surgery group were more than those in the non-surgery group (1-year: 80.96% vs. 58.65%, 3-year: 50.33% vs. 25.39%, 5-year: 42.53% vs. 20.85%, respectively; Table S2). The median CSS time was 37.00 months [95% confidence interval (CI): 30.56–43.44 months] for surgical patients but only 16.00 months (95% CI: 14.73–17.27 months) for nonsurgical patients after PSM.

Figure 2 Kaplan-Meier curves of patients with small cell lung cancer between surgery and non-surgery groups before and after PSM. (A,B) Kaplan-Meier curve comparing OS (A) and CSS (B) before PSM. (C,D) Kaplan-Meier curve comparing OS (C) and CSS (D) after PSM. PSM, propensity score matching; OS, overall survival; CSS, cancer-special survival.

Identification of independent prognostic factors

Univariate and multivariate Cox proportional hazard regression was performed to distinguish independent prognostic factors in patients with SCLC population after PSM. The results indicated that receiving surgery was a protective factor of OS [multivariate analyses hazard ratio (HR) =0.44; 95% CI: 0.40–0.49; P<0.001] and CSS (multivariate analyses HR =0.43; 95% CI: 0.38–0.48; P<0.001). Moreover, patient age, sex, T stage, N stage, primary site of tumor, M stage, radiotherapy, and chemotherapy were also independent prognostic factors (Table 2).

A nomogram to identify optimal candidates for primary tumor resection

When surgical patients showed a CSS time of longer than 16 months, we wondered whether these patients might benefit from surgery. From a total of 1,364 cases of receiving surgery, we identified 874 cases of surgery & benefit along with 317 cases of surgery & non-benefit. Patients who died of non-lung cancer or lived but who did not follow-up for more than 16 months were excluded. Univariate and multivariable logistic analyses were used to distinguish independent factors of surgical benefit based on the benefit and non-benefit groups. The multivariable logistic analyses showed N stage, T stage, primary site of tumor, age, M stage, and sex as independent factors of surgery benefit. According to this logistic model, a nomogram was established to predict optimal primary tumor resection candidates in SCLC patients (Figure 3). Collecting patient age, sex, TNM stage, and primary tumor site, clinicians can calculate the likelihood of survival of each patient by adding score values corresponding to the variable in the nomogram.

Figure 3 A nomogram to distinguish optimal surgical candidates in patients with small cell lung cancer; TNM, T, size of the tumor; N, extent of regional lymph node involvement; M, presence of metastasis.

Validation of the predictive model

Internal and external validation was performed to test the discrimination and stability of this nomogram. The ROC curves show the area under the curve (AUC) index for predicting CSS was 0.737 (95% CI: 0.670–0.775) and 0.681 (95% CI: 0.619–0.745) in training and testing sets respectively (Figure 4A,4B). The fit test demonstrated that the model fit well in the training and testing sets (Hosmer-Lemeshow goodness: P=0.70 and P=0.29). Moreover, the calibration plots showed that the nomogram predicted probability was closely correlated with the actual probability (Figure 4C,4D). Additionally, DCA curves also indicated that this predictive model appeared good predictive ability (Figure 4E,4F).

Figure 4 The internal and external validation of this nomogram. The ROC curve in training set (A) and testing set (B). The calibration plots in training set (C) and testing set (D). The decision curve analysis curve of nomogram in training set (E) and testing set (F). AUC, area under the curve; ROC, receiver operating characteristic.

In the PSM population, we reclassified the surgical patients using a cut-off 0.5 to validate the discrimination of the model. As shown in Figure 5, patients in the surgery & benefit group had a longer CSS time than those in the surgery & non-benefit group (HR =2.89; 95% CI: 2.33–3.59; P<0.001) and those in the non-surgery group (HR =2.38; 95% CI: 2.11–2.69; P<0.001). However, significant difference was not found in CSS time for SCLC patients between the surgery & non-benefit group and non-surgery group (HR =0.82; 95% CI: 0.67–1.01; P=0.07).

Figure 5 Kaplan-Meier curve to compare differential beneficial groups in the population after PSM according to this nomogram. PSM, propensity score matching.

Clinical application

To apply the nomogram, we drew a vertical line above each variable to obtain the corresponding score and then calculated total score. And then, we again drew a vertical line below the total point to locate probability of benefit (Figure 3). For example, facing a 66-year-old male patient with upper-lobe SCLC diagnosed as T3N2M0 stage, we calculated total score 134 points. The final probability of benefit is 36%, which suggests that this patient would not benefit from primary tumor resection. Therefore, this nomogram provides a novel visual evaluation model that provides a convenient basis for clinicians to make decisions about whether to operate in SCLC patients.

Discussion

This is the first predictive model to distinguish optimal primary tumor resection candidates based on preoperative population and tumor characteristics in SCLC patients. Firstly, we demonstrated that surgery could improve the long-term survival of SCLC patients, which is consistent with results of previous studies (12,19-21). Moreover, detailed TNM stage contributed to the evaluation of prognosis of SCLC patients.

In this model, the N stage, primary site, and T stage were the three most important predictive factors for distinguishing surgical benefit. Previous studies have confirmed that lymph metastasis status and numbers are associated with prognosis in SCLC patients (11,22-24). Moreover, ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) was also used to stratify the prognosis of SCLC patients through evaluating metabolic parameters of lymph nodes (17). Additionally, patient age, sex, and M stage are also important for screening surgical patients. Taken together, these results show that self-conditions and TNM stage (based on AJCC 8^th edition) are crucial for making surgical decisions. Earlier staging and younger age create better conditions for surgery. Clinically, TNM stage should be applied more thoroughly and evaluated more accurately for SCLC patients.

In Cox proportional hazard regression, radiotherapy and chemotherapy are also important prognostic factors of SCLC patients, in addition to age, sex, and TNM stage. Although surgery has been shown to improve long-term survival time of SCLC patients, surgery should be only important part of comprehensive treatment because of inferior results receiving surgery alone (25,26). Surgery should be more valued and performed more than ignored for patients with SCLC. With the development of rapid recovery concepts and minimally invasive surgery, surgery will be used more to deal quickly with tumor burden, and with less trauma either with or without distant and lymph node metastases. Meanwhile, preoperative chemoradiotherapy may be used to downgrade TNM stage and help SCLC patients benefit from surgery based on this predictive model (27).

In previous studies, more focus has been placed on I–II stage SCLC patients. These studies demonstrated that primary tumor resection should be extended to stage II SCLC patients (13,20,25). Although some studies investigated the role of surgery in stage I–II SCLC patients, such models for predicting surgical benefits have not been established. Lobectomy was observed to be advantageous over other surgical approaches. The same conclusions were reached for stage III SCLC patients (12). Zeng et al. developed a nomogram to forecast the long-term survival of patients with resected limited-stage SCLC (28). Similarly, Wang et al. also established a nomogram and risk stratification system for resectable SCLC based on clinicopathological characteristics and surgical procedures (29). In these two predictive models, surgery methods, age, sex, chemotherapy, T stage, and lymph status were included. However, detailed lymph status information, such as lymph node metastasis ratio, was hard to verify before surgery. Therefore, lymph node metastasis was not included in our predictive model. Similarly, lymph status was indicated to be the most important factor for deciding whether a patient would benefit from surgery in our model. Moreover, tumors of the main bronchus urgently required surgical treatment, and patients with M1 stage SCLC may benefit from surgery in certain patients. Nowadays, there is a lack of prospective clinical studies to validate the prognostic value of surgery in stage II–IV SCLC patients. Some clinicians do not consider surgery as a treatment option for SCLC. Therefore, it is urgent to perform prospective clinical studies evaluating surgery in II–IV SCLC, to determine whether more clinicians should use surgical treatment approaches.

In this model, the ROC curves, DCA plots, and calibration plots show the good discrimination and stability of this nomogram. We hope that more clinical database will be used to verify the value of the predictive model and that this model will help guide further personalization of clinical management. Our study is not without limitation. First, some important clinical data were unknown, such as nutritional status and comorbidities, which may lead to a selection bias prior to making decisions about surgery. Second, neoadjuvant therapy was not clearly listed in the SEER database; immunotherapy and target therapy were not included in SEER database. Therefore, we could only speculate that primary tumor resection would bring more survival benefit after systemic treatment. Third, data on tumor markers were not improved in the SEER database, such as neuron-specific enolase (NSE). Lastly, an external data set from a single center or multicenter sample population is needed to valid this predictive model.

Conclusions

We have built a novel visual predictive model to distinguish optimal SCLC candidates who would benefit from primary tumor resection based on preoperative population and tumor characteristics. This model confirmed that surgery improved the long-term survival of SCLC patients, and showed good specificity and stability. Given accurate clinical and tumor TNM stage information, this nomogram allows clinicians to better assess the extent to which each patient would benefit from surgery. We hope another external validation with a multicenter database will be used to valid our model, which may encourage clinicians to perform surgery in SCLC patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1419/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1419/prf

Funding: This work was supported by the Young Talent Development Plan of Changzhou Health Commission (Nos. CZQM2020034 and CZQM2020004); Social Development Projects of Changzhou Science and Technology Bureau (No. CE20205039); Top Talent of Changzhou “14th Five-Year Plan” High-Level Health Personnel Training Project (No. KY20221388); and Major Projects of the Changzhou Health Commission (No. ZD202205).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-1419/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 (as revised in 2013).

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. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Gazdar AF, Bunn PA, Minna JD. Small-cell lung cancer: what we know, what we need to know and the path forward. Nat Rev Cancer 2017; Epub ahead of print. [Crossref]
3. Redin E, Quintanal-Villalonga Á, Rudin CM. Small cell lung cancer profiling: an updated synthesis of subtypes, vulnerabilities, and plasticity. Trends Cancer 2024;10:935-46. [Crossref] [PubMed]
4. Drapkin BJ, Rudin CM. Advances in Small-Cell Lung Cancer (SCLC) Translational Research. Cold Spring Harb Perspect Med 2021;11:a038240. [Crossref] [PubMed]
5. Roskoski R Jr. Targeted and cytotoxic inhibitors used in the treatment of lung cancers. Pharmacol Res 2024;209:107465. [Crossref] [PubMed]
6. Rudin CM, Ismaila N, Hann CL, et al. Treatment of Small-Cell Lung Cancer: American Society of Clinical Oncology Endorsement of the American College of Chest Physicians Guideline. J Clin Oncol 2015;33:4106-11. [Crossref] [PubMed]
7. Shim JS, Kim Y, Yuh T, et al. Real-World Outcomes with Lurbinectedin in Second Line and Beyond for Extensive Stage Small Cell Lung Cancer in Korea. Lung Cancer (Auckl) 2024;15:149-59. [Crossref] [PubMed]
8. Li H, Song L, Zhou Y, et al. The effects of surgical resection in the treatment of limited-stage small cell lung cancer: a multicenter retrospective study. Updates Surg 2024;76:1483-92. [Crossref] [PubMed]
9. Qiao R, Zhong R, Xu J, et al. Prediction of lymph node status in completely resected IIIa/N2 small cell lung cancer: importance of subcarinal station metastases. J Cardiothorac Surg 2019;14:63. [Crossref] [PubMed]
10. Leuzzi G, Lococo F, Alessandrini G, et al. Prognostic Impact of Node-Spreading Pattern in Surgically Treated Small-Cell Lung Cancer: A Multicentric Analysis. Lung 2017;195:107-14. [Crossref] [PubMed]
11. Yang H, Mei T. The prognostic value of lymph node ratio in patients with surgically resected stage I-III small-cell lung cancer: a propensity score matching analysis of the SEER database. Eur J Cardiothorac Surg 2021;60:1212-20. [Crossref] [PubMed]
12. Gao L, Shen L, Wang K, et al. Propensity score matched analysis for the role of surgery in stage III small cell lung cancer based on the eighth edition of the TNM classification: a population study of the US SEER database and a Chinese hospital. Lung Cancer 2021;162:54-60.
13. Chen F, Wang Z, Gu X, et al. Different treatment modalities on the prognosis of patients with stage I-IIIa small cell lung cancer: a population based study. J Thorac Dis 2024;16:2822-34. [Crossref] [PubMed]
14. Casiraghi M, Sedda G, Del Signore E, et al. Surgery for small cell lung cancer: When and how. Lung Cancer 2021;152:71-7. [Crossref] [PubMed]
15. Fong AJ, Reich H, Mirocha J, et al. Disparities and Underutilization of Surgery for Early Stage Small Cell Lung Cancer. Ann Thorac Surg 2024;117:1095-102. [Crossref] [PubMed]
16. Yu L, Xu J, Qiao R, et al. Pathological Stage N1 Limited-Stage Small-Cell Lung Cancer Patients Can Benefit From Surgical Resection. Clin Lung Cancer 2023;24:e1-8. [Crossref] [PubMed]
17. Kalemkerian GP, Loo BW, Akerley W, et al. NCCN Guidelines Insights: Small Cell Lung Cancer, Version 2.2018. J Natl Compr Canc Netw 2018;16:1171-82. [Crossref] [PubMed]
18. Harrell FE. rms: Regression Modeling Strategies. 2021. Available online: https://CRAN.R-project.org/package=rms
19. Zeng C, Li N, Li F, et al. Prognostic factors of patients with small cell lung cancer after surgical treatment. Ann Transl Med 2021;9:1146. [Crossref] [PubMed]
20. Kauffmann-Guerrero D, Walter J, Kovács J, et al. The Role of Thoracic Surgery in Small Cell Lung Cancer - A Large Longitudinal Analysis (2002-2015) Based on Real-World Data. Clin Lung Cancer 2022;23:244-52. [Crossref] [PubMed]
21. Caput B, Peretti L, Lacomme S, et al. Effect of surgery on survival of patients with small-cell lung cancer undiagnosed before resection. Ann Thorac Med 2024;19:258-65. [Crossref] [PubMed]
22. Rucker AJ, Raman V, Jawitz OK, et al. Effect of Lymph Node Assessment on Outcomes in Surgery for Limited Stage Small Cell Lung Cancer. Ann Thorac Surg 2020;110:1854-60. [Crossref] [PubMed]
23. Jiang X, Luo C, Peng X, et al. Incidence rate of occult lymph node metastasis in clinical T(1-2)N(0)M(0) small cell lung cancer patients and radiomic prediction based on contrast-enhanced CT imaging: a multicenter study: Original research. Respir Res 2024;25:226. [Crossref] [PubMed]
24. Gao T, Chang Y, Yue H. Association of log odds of positive lymph nodes with survival in patients with small cell lung cancer: Results from the SEER database. Clinics (Sao Paulo) 2024;79:100369. [Crossref] [PubMed]
25. Zhao X, Kallakury B, Chahine JJ, et al. Surgical Resection of SCLC: Prognostic Factors and the Tumor Microenvironment. J Thorac Oncol 2019;14:914-23. [Crossref] [PubMed]
26. Jia J, Trassl L, Kong F, et al. Improved survival of patients with stage III small-cell lung cancer with primary resection: A SEER-based analysis. Transl Oncol 2024;49:102070. [Crossref] [PubMed]
27. Zhang Y, Yang Z, Chen R, et al. Histopathology images-based deep learning prediction of prognosis and therapeutic response in small cell lung cancer. NPJ Digit Med 2024;7:15. [Crossref] [PubMed]
28. Zeng Q, Li J, Tan F, et al. Development and Validation of a Nomogram Prognostic Model for Resected Limited-Stage Small Cell Lung Cancer Patients. Ann Surg Oncol 2021;28:4893-904. [Crossref] [PubMed]
29. Wang Y, Pang Z, Chen X, et al. Development and validation of a prognostic model of resectable small-cell lung cancer: a large population-based cohort study and external validation. J Transl Med 2020;18:237. [Crossref] [PubMed]