Development and internal validation of dynamic nomograms for predicting recurrence and survival in patients with resected combined small cell lung cancer.

Background: The clinical features and prognosis of combined small cell lung cancer (C-SCLC) are not well understood. Given the unique histological heterogeneity of C-SCLC, the 8th tumor-node-metastasis (TNM) staging system has limited prognostic accuracy in this population, necessitating a dedicated prediction model. This study aimed to develop and internally validate dynamic nomograms incorporating combined subtypes and serum tumor markers to predict recurrence and survival in patients with resected C-SCLC. Methods: A total of 223 patients with resected C-SCLC were enrolled in this study between 2008 and 2021. Eligibility criteria were as follows: surgically resected and pathologically confirmed C-SCLC with complete clinical data. Due to the limited sample size, no training/validation split was performed; internal validation was conducted using 1,000 bootstrap resamples. All serum tumor markers including neuron-specific enolase (NSE) were measured within one week before surgery. Visceral pleural invasion (VPI) and combined histological components were independently reviewed by two professional pathologists. The independent prognostic factors were identified and integrated to build the nomograms for predicting 3- and 5-year disease-free survival (DFS) and overall survival (OS) based on stepwise Cox regression. The discrimination and predictive accuracy of the models were evaluated using the concordance index (C-index) and calibration curves. Decision curve analyses (DCAs) were performed to verify the clinical utility of the model compared with that of the 8th edition of the International Association for the Study of Lung Cancer (IASLC) TNM staging system. Based on the nomogram scores, the C-SCLC patients were divided into high- and low-risk subgroups. An online webserver was applied to facilitate the convenient use of the model. Results: Ultimately, six independent prognostic factors, including tumor location, combined components, VPI, adjuvant chemotherapy, lymph node metastasis, and serum NSE level, were identified and incorporated into the nomograms. The median follow-up duration was 42.6 months. During follow-up, 101 patients developed recurrence or metastasis, and 121 died. The median age of the cohort was 64 years, with a predominance of male patients (89.7%). Among the patients, 27.8% were classified as stage I, and large cell neuroendocrine carcinoma (LCNEC) was the most common combined component (67.3%). The C-index values of the nomograms for predicting DFS [0.730, 95% confidence interval (CI): 0662-0.720] and OS (0.748, 95% CI: 0.682-0.734) were significantly higher than those of the TNM staging system (0.601, 95% CI: 0.574-0.628 and 0.615, 95% CI: 0.586-0.644 for DFS and OS, respectively; P<0.001). The calibration plots indicated good agreement between the model-predicted and observed survival. The DCA results showed that the developed nomograms demonstrated better predictive performance than the TNM staging system. Conclusions: Our internal validation suggests that the nomograms outperform the TNM staging system; however, these findings should be considered as hypothesis-generating. External validation with multicenter cohorts is essential before clinical implementation. The online tool is primarily intended to facilitate further research.