Relationship between tumor markers and overall survival in small cell lung cancer: a retrospective cohort study

Introduction

Precise prognostic stratification in small cell lung cancer (SCLC) is paramount for optimizing therapeutic interventions, particularly given the recent integration of immune checkpoint inhibitors into first-line protocols (1). Currently, staging (limited vs. extensive) remains the strongest prognostic factor and the basis for treatment decisions; however, it is still insufficient for highly accurate prognosis evaluation (2). Even among patients initially diagnosed with limited-stage disease, prognosis can vary significantly (3). Thus, it remains necessary to explore novel prognostic factors to better evaluate SCLC prognosis.

Serum tumor markers are widely used for tumor diagnosis and prognosis assessment (4). For SCLC, the most widely used diagnostic markers are neuron-specific enolase (NSE) (5), carcinoembryonic antigen (CEA) (6), a fragment of cytokeratin 19 (CYFRA 21-1) (7), and pro-gastrin-releasing peptide (proGRP) (5,8). Several studies have demonstrated that these tumor markers possess prognostic value (9-11). However, the value of tumor markers in assessing prognosis in SCLC remains controversial. For example, a systematic review found no relation between NSE and prognosis in one-third of the studies that included (5). Reasons for heterogeneity across studies include sample size, length of follow-up, data handling methods (such as log transformation, original data, or grouping), and adjustment for confounding factors. Therefore, further research into tumor markers and SCLC prognosis is necessary to identify markers that influence their prognostic value.

It is worth noting that when analyzing the relationship between tumor markers and the prognosis of SCLC patients, previous studies consistently divided patients into two or more groups based on tumor marker levels (12,13). Survival curves and Cox proportional hazards models were used for data analysis (12,13). This data analysis approach has several limitations. First, nearly all previous studies converted continuous tumor markers into categorical variables, leading to a loss of information. Second, existing literature frequently evaluates biomarkers in isolation, often neglecting potential multicollinearity or the synergistic prognostic value of a multi-marker panel. Third, the predictive value of tumor markers for patient prognosis remains unclarified. Additionally, some past studies were published before 2020, when SCLC treatment regimens differed significantly from those used today. Notably, since 2019, immune, anti-angiogenesis, and targeted therapies have been used in SCLC treatment, which has greatly improved patient prognosis (14,15). Therefore, previous conclusions may not be applicable to current clinical practice. Here, we conducted a retrospective study analyzing baseline NSE, CEA, CYFRA 21-1, and proGRP, and their relationship to the prognosis of SCLC patients. We report our findings following the STROBE reporting checklist (16) (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1007/rc).

Methods

Participants

We retrospectively analyzed medical records of patients who visited Tianjin Medical University General Hospital from June 2019 to October 2022. Inclusion criteria were: (I) pathologically confirmed SCLC; (II) age over 18 years; (III) treatment naive. Exclusion criteria were: (I) patients complicated with tumors other than SCLC; (II) patients with key data missing >50%; (III) unavailable prognostic information; (IV) end-stage liver or renal failure. This study was approved by the Ethics Committee of Tianjin Medical University General Hospital (IRB No. 2025-KY-706). Because it was a retrospective study, informed consent was waived. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. We identified patients using our institution’s (Tianjin Medical University General Hospital) electronic medical records system.

Data extraction

We extracted patients’ demographics, laboratory test results (including tumor markers), pathological characteristics, and imaging findings from medical records at the time of their first visit. If multiple laboratory, imaging, or pathological tests were performed at the initial visit, only the first test data were included. Our institution used the Roche e601 (Roche Diagnostics, Mannheim, Germany) to determine NSE, and used the Abbott ARCHITECT i2000 system (Abbott Diagnostics, Wiesbaden, Germany) to determine CEA, CYFRA 21-1, and proGRP.

In October 2023, patients’ overall survival (OS) was confirmed through medical records or telephone follow-up.

Statistical analysis

The Kolmogorov-Smirnov test was used to assess normal distribution (significance level set at 0.10). Data were expressed as mean and standard deviation for normally distributed continuous variables. An independent t-test was used to compare two groups; analysis of variance (ANOVA) was used for more than two groups. For variables with less than 20% missing data, missing values were imputed using the MICE package in R (17). Non-normally distributed data were expressed as median and interquartile range. The Mann-Whitney U test compared two groups, the Kruskal-Wallis test compared three or more groups, and the Spearman correlation analyzed relationships. Kaplan-Meier curves described prognosis by groups, with log-rank tests comparing survival differences. Multivariate Cox proportional hazards models were employed to investigate the relationship between tumor markers and OS. Continuous variables were log-transformed as appropriate. Time-dependent receiver operating characteristic (ROC) curves were used to evaluate the predictive value of tumor markers for OS in SCLC (18). All analyses were conducted in R; P<0.05 was considered statistically significant.

Results

Basic characteristics of the participants

During June 2019 and October 2022, 177 treatment-naive SCLC patients were admitted to our institution. Among them, 17 were excluded due to excessive missing data or end-stage liver or renal disease. Finally, 160 patients were included. Their clinical characteristics are shown in Table 1. Among participants, 53% received immune therapy and 26% received target therapy.

Relationships among CEA, NSE, proGRP, and CYFRA21-1

Figure 1 is a heat map showing correlations among tumor markers. CEA correlated positively with CYFRA21-1; proGRP correlated positively with NSE and CYFRA21-1.

Figure 1 Correlations among tumor markers. CEA, carcinoembryonic antigen; CYFRA21-1, a fragment of cytokeratin 19; NSE, neuron-specific enolase; proGRP, pro-gastrin-releasing peptide.

Relationship between tumor markers and patients OS

Patients were divided into two groups based on the median concentrations of each tumor marker. Survival curves illustrated the relationship between tumor markers and OS, showing that only NSE and CYFRA21-1 were significantly associated with OS (Figure 2).

Figure 2 Kaplan-Meier curves showing the relationship between tumor markers and overall survival. CEA, carcinoembryonic antigen; CYFRA21-1, a fragment of cytokeratin 19; NSE, neuron-specific enolase; proGRP, pro-gastrin-releasing peptide.

We used Cox proportional hazards models to analyze the prognostic value of tumor markers. Before using the model, we assessed collinearity among tumor markers using variance inflation factors. The variance inflation factors for all tumor markers were <5, and all tumor markers, along with stage, age, and sex, were included in the Cox model.

As shown in Table 2, although univariate analysis showed all tumor markers (log-transformed) were associated with OS, in a multivariate model including stage and tumor markers, only CYFRA21-1 was statistically significant (P<0.05). Besides, we also analyzed the prognostic value of tumor markers in the dataset without imputation. CYFRA21-1 and stage were also independently associated with OS in this analysis. The hazard ratios (HRs) for stage and CYFRA21-1 (log-transformed) were 1.99 [95% confidence interval (CI): 1.14–3.46] and 1.73 (95% CI: 1.18–2.55), respectively.

Restricted cubic spline (RCS) analysis revealed a linear relationship between tumor markers and OS (Figure 3).

Figure 3 Restricted cubic spline depicting tumor markers and OS of SCLC. CI, confidence interval; CEA, carcinoembryonic antigen; CYFRA21-1, a fragment of cytokeratin 19; HR, hazard ratio; NSE, neuron-specific enolase; OS, overall survival; proGRP, pro-gastrin-releasing peptide; SCLC, small cell lung cancer.

Finally, time-dependent ROC curve analysis revealed moderate tumor marker areas under the curve (AUCs), generally ranging from 0.60 to 0.70 (Figure 4).

Figure 4 Time-dependent ROC curves depicting tumor markers and overall survival of SCLC. AUC, area under the curve; CEA, carcinoembryonic antigen; CYFRA21-1, a fragment of cytokeratin 19; NSE, neuron-specific enolase; proGRP, pro-gastrin-releasing peptide; ROC, receiver operating characteristic; SCLC, small cell lung cancer.

Discussion

This study used Kaplan-Meier curves, the Cox proportional hazards model, time-dependent ROC curves, and RCS to analyze the relationship between four serum tumor markers and OS in SCLC patients. We found that individual tumor markers were associated with OS in univariate analyses; however, after adjusting for staging, only CYFRA21-1 was associated with patient OS. There is a linear relationship between tumor markers and patient OS. For patients with the highest concentration of tumor markers, the risk of death was 1.5 times that of patients with low concentration tumor markers. Time-dependent ROC curves also indicate that the AUC for tumor markers predicting patient OS was approximately 0.6. These results suggest that serum tumor markers are linearly correlated with the prognosis of SCLC patients, and CYFRA 21-1 shows the strongest association.

Although many previous studies have investigated the relationship between tumor markers and the prognosis of SCLC patients (7-10), this study has several advantages over previous studies. The first advantage of our study is that we used multiple advanced statistical methods to analyze the relationship between tumor markers and OS of SCLC. For example, we applied a logarithmic transformation to tumor markers to overcome the influence of outlier values. We used RCS analysis to assess the relationship between tumor markers and patient OS. The linear association between tumor markers and OS supports the use of the logarithmic transformation of tumor markers in the Cox proportional hazards model. In contrast, previous studies often categorized patients into two groups based on the median tumor marker level before performing survival curve analysis and Cox model analysis. This grouping method is relatively crude because of information loss. For example, in a categorical analysis of CEA with a threshold of 4 ng/mL, a patient with CEA of 5 ng/mL and a patient with CEA of 50 ng/mL are both categorized as high CEA. We also used time-dependent ROC curves to analyze the relationship between tumor markers and patient OS. Traditional methods for analyzing the prognostic value of tumor markers include Kaplan-Meier curves and Cox proportional hazards models, but these methods only indicate whether tumor markers independently associate with patient OS and cannot assess their predictive value for OS. Our study used time-dependent ROC curves to evaluate the predictive value of tumor markers. We found that the overall performance of tumor markers in predicting patient prognosis is moderate, with an AUC of around 0.70. These findings suggest that tumor markers, when used alone, may not be suitable as the basis for selecting patient treatment regimens. This result also supports our conclusion from the Cox proportional hazards model that, among the four tumor markers studied, only CYFRA21-1 was independently associated with patients’ prognosis. In addition, more than half of the participants in our study received immune, anti-angiogenesis, and targeted therapies, while many previous studies only studied the prognostic value of tumor markers in patients who received chemotherapy. Maybe because of the small sample size, we failed to find significant prognostic value for NSE, CEA, and proGRP. However, we found that CYFRA21-1 was significantly associated with OS, suggesting that tumor markers can be used to estimate SCLC prognosis in the era of immunotherapy and anti-angiogenesis therapies.

This study also has some limitations. First, it is a single-center retrospective study with a relatively small sample size, so the results’ precision needs further improvement. Second, we studied only four tumor markers simultaneously and did not consider whether these markers remain associated with patient prognosis when other markers are present. This is because our institution only performs routine testing for these four tumor markers.

Conclusions

In summary, common tumor markers are associated with OS in SCLC in a linear manner. Among the conventional tumor markers, CYFRA21-1 has the strongest association with OS. The predictive value of a single tumor marker for OS is moderate and should not be used alone as a basis for treatment decision-making.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1007/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1007/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1007/prf

Funding: This study was supported by the National Natural Science Foundation of China (No. 82272672).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1007/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. This study was approved by the Ethics Committee of Tianjin Medical University General Hospital (IRB No. 2025-KY-706). Because it was a retrospective study, informed consent was waived. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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. Kim SY, Park HS, Chiang AC. Small Cell Lung Cancer: A Review. JAMA 2025;333:1906-17. [Crossref] [PubMed]
2. Oronsky B, Abrouk N, Caroen S, et al. A 2022 Update on Extensive Stage Small-Cell Lung Cancer (SCLC). J Cancer 2022;13:2945-53. [Crossref] [PubMed]
3. Yao Z, Li C, Han F, et al. Development and validation of a serum inflammatory biomarker-driven machine-learning model for prognostic stratification in surgical limited-stage small cell lung cancer. Transl Lung Cancer Res 2025;14:2983-95. [Crossref] [PubMed]
4. Zhou Y, Tao L, Qiu J, et al. Tumor biomarkers for diagnosis, prognosis and targeted therapy. Signal Transduct Target Ther 2024;9:132. [Crossref] [PubMed]
5. Tian Z, Liang C, Zhang Z, et al. Prognostic value of neuron-specific enolase for small cell lung cancer: a systematic review and meta-analysis. World J Surg Oncol 2020;18:116. [Crossref] [PubMed]
6. Guo D, Jing W, Zhu H, et al. Clinical value of carcinoembryonic antigen for predicting the incidence of brain metastases and survival in small cell lung cancer patients treated with prophylactic cranial irradiation. Cancer Manag Res 2018;10:3199-205. [Crossref] [PubMed]
7. Li J, Dai CH, Chen P, et al. Survival and prognostic factors in small cell lung cancer. Med Oncol 2010;27:73-81. [Crossref] [PubMed]
8. Huang Z, Xu D, Zhang F, et al. Pro-gastrin-releasing peptide and neuron-specific enolase: useful predictors of response to chemotherapy and survival in patients with small cell lung cancer. Clin Transl Oncol 2016;18:1019-25. [Crossref] [PubMed]
9. Liu X, Zhang W, Yin W, et al. The prognostic value of the serum neuron specific enolase and lactate dehydrogenase in small cell lung cancer patients receiving first-line platinum-based chemotherapy. Medicine (Baltimore) 2017;96:e8258. [Crossref] [PubMed]
10. Petrović M, Bukumirić Z, Zdravković V, et al. The prognostic significance of the circulating neuroendocrine markers chromogranin A, pro-gastrin-releasing peptide, and neuron-specific enolase in patients with small-cell lung cancer. Med Oncol 2014;31:823. [Crossref] [PubMed]
11. Sirohi A, Alsaedi BSO, Ahelali MH, et al. Biased proportional hazard regression estimator in the existence of collinearity. Heliyon 2023;9:e21394. [Crossref] [PubMed]
12. Zhu C, Huang J, Jin X, et al. The predictive value of delta-like3 and serum NSE in evaluating chemotherapy response and prognosis in patients with advanced small cell lung carcinoma: An observational study. Medicine (Baltimore) 2024;103:e38487. [Crossref] [PubMed]
13. Ando S, Suzuki M, Yamamoto N, et al. The prognostic value of both neuron-specific enolase (NSE) and Cyfra21-1 in small cell lung cancer. Anticancer Res 2004;24:1941-6.
14. Horn L, Mansfield AS, Szczęsna A, et al. First-Line Atezolizumab plus Chemotherapy in Extensive-Stage Small-Cell Lung Cancer. N Engl J Med 2018;379:2220-9. [Crossref] [PubMed]
15. Goldman JW, Dvorkin M, Chen Y, 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]
16. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008;61:344-9. [Crossref] [PubMed]
17. Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med 2016;4:30. [Crossref] [PubMed]
18. Kamarudin AN, Cox T, Kolamunnage-Dona R. Time-dependent ROC curve analysis in medical research: current methods and applications. BMC Med Res Methodol 2017;17:53. [Crossref] [PubMed]