[The development and validation of risk prediction model for lung cancer: a systematic review].

To systematically understand the global research progress in the construction and validation of lung cancer risk prediction models. "lung neoplasms" , "lung cancer" , "lung carcinoma" , "lung tumor" , "risk" , "malignancy" , "carcinogenesis" , "prediction" , "assessment" , "model" , "tool" , "score" , "paradigm" , and "algorithm" were used as search keywords. Original articles were systematically searched from Chinese databases (CNKI, and Wanfang) and English databases (PubMed, Embase, Cochrane, and Web of Science) published prior to December 2018. The language of studies was restricted to Chinese and English. The inclusion criteria were human oriented studies with complete information for model development, validation and evaluation. The exclusion criteria were informal publications such as conference abstracts, Chinese dissertation papers, and research materials such as reviews, letters, and news reports. A total of 33 papers involving 27 models were included. The population characteristics of all included studies, study design, predicting factors and the performance of models were analyzed and compared. Among 27 models, the number of American-based, European-based and Asian-based model studies was 12, 6 and 9, respectively. In addition, there were 6 Chinese-based model studies. According to the factors fitted into the models, these studies could be divided into traditional epidemiological models (11 studies), clinical index models (6 studies), and genetic index models (10 studies). 15 models were not validated after construction or were cross-validated only in the internal population, and the extrapolation effect of models was not effectively evaluated; 8 models were validated in single external population; only 4 models were verified in multiple external populations (3-7); the area under the curve (AUC) of models ranged from 0.57 to 0.90. Research on risk prediction models for lung cancer is in development stage. In addition to the lack of external validation of existing models, the exploration of potential clinical indicators was also limited.