Predictive modeling of high-grade lung neuroendocrine neoplasms Overall survival and Cancer-specific survival based on a machine learning approach

Abstract

Abstract Objective: We wished to construct a prognostic prediction model regarding high-grade lung neuroendocrine neoplasms(L-NENs) by using demographic characteristics and clinical information. Method: The demographic characteristics and clinical data of 5268 patients with pathologically diagnosed high-grade lung neuroendocrine tumors between 2010 and 2019 were retrospectively analyzed using the Surveillance, Epidemiology and End Results (SEER) database, and a Nomogram of overall survival(OS) and cancer-specific survival(CSS) at 1, 3, and 5 years was constructed using LASSO regression and COX regression analysis. Nomogram of OS and CSS at 1, 3, and 5 years were constructed, and the performance of the predictive models was evaluated using the consistency index (C-index), calibration curves, Receiver Operating Characteristic(ROC) curves, and decision curve analysis (DCA), and internal validation of our constructed models was performed by validation sets. Results: We divided the included patients into training and validation sets in a ratio of 7:3, and analysis using the chi-square test revealed no statistically significant difference between the baseline information of the two data sets (p > 0.05); The training set was analyzed using COX univariate analysis and found that gender, age, AJCC stage, whether treated or not, and distant metastasis were the influencing factors of OS, and these influencing factors were found to be independent prognostic influences of OS after further screening by including these influencing factors in LASSO regression, and we constructed a Nomogram plot of OS by including these influencing factors in COX multivariate;We used the same method to screen the independent prognostic influences affecting CSS were gender, age, race, AJCC stage, whether treated or not, bone metastasis, brain metastasis, and liver metastasis, etc., which we included in the COX multifactorial to construct a Nomogram of CSS;Validation of the OS and CSS models using ROC curves, C-indexes, calibration curves and DCA curves after construction proved the accuracy and reliability of our models. Conclusion: This prediction model can more accurately predict the prognosis of patients with high-grade L-NENs.