Risk factors of early mortality in patients with small cell lung cancer: a retrospective study in the SEER database

Small cell lung cancer (SCLC) is a highly aggressive neuroendocrine cancer with a high risk of early mortality (i.e., survival time less than 1 month). This study aimed to identify relevant risk factors and predict early mortality in SCLC patients. A total of 27,163 SCLC cases registered between 2010 and 2019 were extracted from the Surveillance, Epidemiology, and End Results (SEER) data. Significant independent risk factors were identified by univariate and multivariate logistic regression analyses. Nomograms for all-causes and cancer-specific early death were constructed and evaluated. Age, sex, clinical stage, presence of metastasis (liver and lung), and absence of treatment (surgery, radiotherapy and chemotherapy) were identified for significant association with all-causes and cancer-specific early death. Nomograms based on these predictors exhibited high accuracy (area under ROC curve > 0.850) and potential clinical practicality in the prediction of early mortality. We identified a set of factors associated with early mortality from SCLC and developed a clinically useful nomogram to predict high-risk patients. This nomogram could aid oncologists in the administration of individualized treatment regimens, potentially improving clinical outcomes of SCLC patients.


Introduction
Lung cancer is the most commonly diagnosed cancer and the greatest causes of cancer-related deaths worldwide (Dyba et al. 2021). Small cell lung cancer (SCLC) represents approximately 15% of all lung cancers (Rudin et al. 2021). SCLC is a highly aggressive neuroendocrine cancer characterized by rapid proliferation, early dissemination and poor prognosis (Rudin et al. 2021). Cigarette smoking is the dominant risk factor accounting for 95% of all SCLC cases (Bennett et al. 2017). Therefore, the incidence and mortality of males are much higher than that of females (Wang et al. 2018). Nowadays, although the development of targeted therapy and immunotherapy have changed the treatment paradigm for SCLC, chemotherapy and radiotherapy are still the mainstream therapy and merely achieve minor survival benefits. The overall survival (OS) for SCLC patients is roughly 1 to 2 years, with a 5-year OS rate of 2.8 to 7.2% (Iams et al. 2019;Wu et al. 2020).
Tumor heterogeneity and distinct therapeutic regimens could lead to divergence of survival time between patients (Hsu et al. 2019), such as early death of some SCLC patients after disease diagnosis. Exploring relevant factors can help clinicians identify high-risk patients and develop targeted treatment plans to improve the survival rate and quality of life of patients (Song et al. 2020a, b). Existing evidence has found that risk factors for early death could be identified in gastric cancer (Zhu et al. 2020), endometrial cancer  and nasopharyngeal cancer (Chen et al. 2021), but the results on SCLC cases were rarely reported. Therefore, it is practical and necessary to identify risk factors for early death in SCLC. Based on the above-mentioned research, we defined early death as mortality within 1 month after diagnosis of SCLC (Jones et al. 2018). The objective of this study was to investigate potential risk factors of early death based on publicly available data from the Surveillance, Epidemiology, and End Results (SEER) database, in order to develop prognostic prediction tools.

Raw data source
Cancer incidence and survival data for approximately 47.9% of the American cancer registry population have been recorded in the SEER database, which is largest publicly available data source for cancer incidence and survival in the United States. We accessed the SEER Research Data 1975-2019 (www. seer. cancer. gov), which was released in April 2022. Clinical information of SCLC between 2010 and 2019 was obtained using SEER * Stat software (version 8.4.0).
Selected patients were randomly assigned in a 7:3 ratios to training and validation cohorts to construct and validate nomogram. The demographic and clinical characteristics collected for analysis were as follows: age at diagnosis, sex, race, primary site, grade, laterality, clinical stage and TNM stage, bone metastasis, brain metastasis, liver metastasis, lung metastasis, surgery, radiotherapy, chemotherapy, causes of death, and survival months. The clinical outcome variables of this study were all-causes and cancer-specific early death.

Statistical analysis
Univariate and multivariate logistic regressions were used to identify individual risk factors, whereby a nomogram was constructed to help predict high-risk patients for early death. The receiving operator characteristic (ROC) curve was plotted to calculate the area under the curve (AUC), which is positively correlated with the predictive accuracy (Janssens et al. 2020). By bootstrapping with 1000 resamples, the calibration plots were produced to evaluate the consistency between the actual and predicted probabilities of early death (Correa et al. 2019). Additionally, decision curve analysis (DCA) was performed to evaluate the clinical applicability of the nomogram (Chen et al. 2019;Gao et al. 2022). All analyses were performed using the R software (version 4.2.0). A two-tailed P-value less than 0.05 was considered statistically significant.

Ethics approval
The data in the SEER database do not require informed patient consent and are deemed exempt from a review by the ethics committee of the Third Affiliated Hospital of Shandong First Medical University. The present study was complied with the ethical standards in 1964 Helsinki Declaration.

Demographic and clinical characteristics
As shown in Fig. 1, 27,163 patients were included in the study based on the aforementioned inclusion and exclusion criteria (see Materials and Methods), including 7435 all-causes and 6816 cancer-specific early death cases. The majority of early deaths occurred in patients who were male (51.8%), between the ages of 65 and 85 years (66.3%), and of white race (88.9%). The most common primary site among early death cases were upper lobe (43.8%). Moreover, most of the patients who died early were not treated with surgery (99.5%), radiotherapy (83.3%), or chemotherapy (71.3%). The characteristics of the patients are summarized in Table 1.  ACED all-causes early death, CSED Cancer-specific early death

Risk factors for early death
Univariate and multivariate logistic regressions were performed with the training dataset (see Materials and Methods). The results demonstrated that age, sex, clinical stage, and liver or lung metastases, and patients who did not receive surgery, radiotherapy or chemotherapy had a higher risk of all-causes early death (Table 2). And age, clinical stage, and liver or lung metastases, and patients who did not receive surgery, radiotherapy or chemotherapy were subject to a higher risk of cancer-specific early death (Table 3).

Nomogram construction
The multivariable logistic regression model showed that specific variables were statistically significant correlated with early death. We therefore constructed nomograms of all-causes and cancer-specific early death with these significant predictors, including age, sex (only in all-causes early death), clinical stage, presence of metastasis (liver and lung), and treatment (surgery, radiotherapy and chemotherapy) (Fig. 2). The nomogram enabled the calculation of a point value for each patient (Supplementary Table 1), which corresponded to the odds of early death (Supplementary Table 2).

Performance evaluation
The efficiency of nomogram was further verified in the validation dataset (see Materials and Methods). As measured by the area under ROC curve, the nomograms had reliable prediction capability for all-causes and cancer-specific early death (Fig. 3). The results of decision curve analysis (DCA) (Fig. 4) also showed a better clinical benefit compared to treat-all and treat-no strategies. In addition, the calibration curves were observed to be close to the ideal 45-degree line, suggesting good consistency between the actual and predicted values (Fig. 5).

Discussion
SCLC is a type of lung cancer with the worst prognosis (Sabari et al. 2017). While most clinicians are committed to prolonging the survival time and improving the quality of life of patients, few studies have focused on early death (Bertagnolli et al. 2020). Although the definition of early death in cancer varies, the threshold was set as 3 months in many studies (Song et al. 2020a, b;Zhang et al. 2021). However, SCLC is a very aggressive neuroendocrine carcinoma with extremely high risk of early death. Therefore, following the criteria adopted in previous research (Cortinovis et al. 2021), we defined early death as mortality within 1 month after diagnosis. Understanding the risk factors for early death is crucial to improve the survival of SCLC cancer patients. This study was the first one to evaluate the risk factors for early deaths in SCLC. Among the included patients, 27.4% the suffered all-causes early death and 25.1% suffered cancer-specific early death, suggesting an evidently higher risk than many other types of cancer. A nomogram for predicting early death in SCLC patients was constructed and validated in this study (Iasonos et al. 2008). Age, sex, clinical stage, presence of In this study, the P value was two-sided, and P < 0.05 was considered statistically significant (in bold) Reference Reference metastasis (liver and lung), and absence of treatment (surgery, radiotherapy and chemotherapy) were found to be closely related to early death in SCLC. First of all, our model demonstrated that older SCLC patients are more likely to experience early death, which is consistent with previous findings (Miranda-Filho et al. 2017;Shen et al. 2018;Hristov et al. 2019). And clinical stage III/ IV is significantly associated with high risk for early death in SCLC. It indicates that early diagnosis of SCLC can significantly reduce the early mortality. Admittedly, liver and lung metastases were more likely to cause early death in SCLC. Therefore, clinicians should take into consideration this difference when developing strategies to prevent early mortality (Li et al. 2019).
Using nomogram to calculate the probability of early death in SCLC patients could provide a reference for individualized treatment of subsequent patients (Balachandran et al. 2015). Our nomograms constructed with a training dataset were further evaluated by ROC curves and the calibration curves with a validation set. In addition, the results of DCA also showed a better clinical benefit compared to treat-all and treat-none strategies. (Van Calster et al. 2018).
All this evidence corroborated the good discriminative ability and potential applicability of our nomograms.
Despite the potential utility of the current results, there are still some limitations that should be acknowledged. First, the data from SEER database was retrospective, so the generalization ability of our findings should be validated in prospective research and multiple cohorts. Second, the data regarding chemotherapy and radiation therapy were incomplete in SEER database, without details on drug name, dose or duration. Finally, while the nomogram assumes the prognosis as a linear combination of covariates, the clinical factors could be complicated and non-linear. Therefore, more sophisticated models such as machine learning are warranted in the future to detect the complex relationships in data.
In conclusion, we identified a set of factors associated with early mortality from SCLC and developed a clinically useful nomogram to predict high-risk patients. This nomogram could aid oncologists in the administration of individualized treatment regimens, potentially improving the prognosis of patients with SCLC. In this study, the P value was two-sided, and P < 0.05 was considered statistically significant (in bold)