SCLC is a subtype of lung cancer with poor prognosis, mainly occurring in elderly patients. Studies have shown that aging predicts a worse outcome, with elderly patients likely to have poorer performance status and more treatment-related toxicities [6, 8]. Therefore,age should be considered as an important variable in selecting therapeutic methods. However, current TNM staging criterion used to predict prognosis of SCLC patients ignore significant risk factors that can improve individualized survival predictions, such as age, gender, histologic grade, and treatment-related factors. In this study, we developed a more accurate nomogram based on a combination of independent prognostic factors to predict the probability of survival in elderly SCLC patients. Our novel nomogram incorporated ten variables: sex, age, AJCC stage, surgery, chemotherapy, radiation, bone metastasis, brain metastasis, liver metastasis and tumor size, which was able to provide more accurate assessment and prediction of elderly SCLC patients compared with the TNM staging criteria.
In this study, most elderly SCLC patients were white people aged 60–69 years. The tumor was mainly located in the upper lobe of the lung. Most patients had advanced AJCC staging with histological grade III and IV. A large proportion of elderly patients received chemotherapy, but less than 7% of cases underwent surgery. These were some unique disease features for elderly SCLC patients. In addition, we identified ten independent prognostic factors for OS and CSS, which were consistent with the previous findings [7, 13,14,15,16, 23,24,25]. Our model indicated that chemotherapy made the greatest contribution to the prognostic score. Multiple studies have proved that chemotherapy, as the most important treatment for SCLC, can prolong the survival time of patients [7, 13, 15, 23, 26]. Our analysis also indicated that AJCC stage and surgery had greater impacts on patient survival. Wang et al. established a prognostic nomogram for SCLC patients, in which AJCC stage made the greatest contributions to the final risk score [16]. A recent review also indicated that surgery has the greatest impact on the prognosis of patients with SCLC and should be recommended for certain patients, especially in the early stages [24].
Multiple studies have shown that risk factors such as increased age, male sex and larger tumor are negatively correlated with long-term survival [7, 13,14,15, 23, 25]. Zhong et al. constructed a novel predictive nomogram for extensive-stage SCLC patients by screening out independent prognostic factors such as gender, age, TNM staging and treatment methods [15]. Another study conducted by shan et al. identified seven prognostic factors and developed a predictive model for SCLC patients with brain metastasis [25]. These results were consistent with our study. In addition, the survival time of SCLC patients varies depending on the number and site of metastasis [27]. Nakazawa et al. found that extensive-stage SCLC most often metastasized to the liver, lung, brain, bone and adrenal gland [28]. They also demonstrated that patients with liver and multiple organ metastases had the worst survival outcomes. In accordance with these findings, our study showed that liver, bone and brain metastasis had a significant impact on the prognosis of elderly SCLC patients.
Validation of predictive models is critical to determine generalization and avoid overfitting [29]. In the current study, calibration plots showed a good agreement between the nomogram predicted and actual observed 1- and 2-year OS and CSS, which verified the repeatability and reliability of the established nomograms [21, 30]. The higher C-indexes and AUC values of the nomogram compared with the AJCC staging system indicated better discrimination ability of the nomograms. Besides, the C-index and AUC value in the validation cohort of our model were also higher than those of the previously SCLC nomogram published by Wang et al. (C-index: 0.745 vs. 0.722, AUC value: 0.826 vs. 0.789) [16]. Further DCA analyses also testified its obvious clinical application benefit versus traditional AJCC staging model. In addition, according to risk stratification model of this nomogram, patients in the external validation cohort can be effectively divided into high risk and low risk groups with distinguished OS and CSS. To our knowledge, this is the first nomogram survival prediction using SEER database and external validation cohort to predict survival in elderly SCLC patients. It also can be inferred from our study that the characteristics of a high-risk SCLC patient are: elderly male, late stage, large tumor, no surgery or radiotherapy or chemotherapy, with bone, liver or brain metastases. More importantly, our nomogram shows better ability and value than the TNM staging system. We believe that a well-designed nomogram can accurately predict each patient’s prognosis, thus benefiting both clinicians and patients.
Our study has some limitations. First of all, the data in this study was collected retrospectively, which may lead to unavoidable bias. Second, we did not include several treatment-related factors, such as chemotherapy regimens, numbers of cycles, doses and methods of radiation and targeted therapy, which could also influence the prognosis. Third, the external validation cohorts were all from the Asian population, and the sample size was relatively small. Future prospective clinical trials with larger sample size and different ethnic populations are necessary to validate our findings. Finally, it would be interesting to validate the already existing models from Wang et al. [16] on the SEER and our external datasets. However, the SEER database did not contain Charlson/Deyo score information, which prevented direct comparison of performance between our model and the published nomogram from Wang et al. Besides, we do not have the access to the NCBD database. Therefore, we compared the C-index and AUC value of our model with Wang’s nomogram and found that our model performed better. In spite of these shortcomings, our nomogram is established based on large population data collection from the SEER database, which provides a good opportunity to predict OS and CSS for elderly SCLC patients.