| PART ⅠDeveloping a lung cancer risk prediction model based on LASSO and estimating its performance in screeningObjective We explored effectrive predictors in low-dose computed tomography(LDCT)screening cohort by least absolute shrinkage and selection operator(LASSO),and developed a lung cancer risk prediction model based on data from Chinese population groups.Then we estimated the prediction ability of the model,investigated high-risk groups selection,and evaluated screening technology effectiveness.Materials and Methods From October 28th 2008 to December 31th 2019,participants of the derivation cohort for LDCT lung cancer screening were recruited in Cancer Hospital,Chinese Academy of Medical Sciences.From July 1st 2017 to September 2nd 2020,participants of the validation cohort for LDCT lung cancer screening were recruited in Cancer Hospital,Chinese Academy of Medical Sciences,Yunnan Cancer Hospital,and Changzheng Hospital,Shanghai.Data on risk factors,LDCT and tumor markers were collected.Effective predictors selected by LASSO were applied to develop a logistic regression model with R software.Then a combined model was developed based on effective predictors and tumor markers.Receiver operating characteristic(ROC)curve analysis was used to evaluate the predictive ability of the model.Internal validation was performed with a bootstrap method.The discrimination of the model was determined by C-index.Calibration curve was used to assess the calibration of the model.Independent verification was performed in validation cohort data.Results Of 22 976 participants in the derivation cohort,131 lung cancer cases were identified;while of 10 000 participants in the validation cohort,72 lung cancer cases were confirmed.There were 32 candidate variables documented.Age,smoking history,and education level were different between lung cancer and non-lung cancer individuals(P<0.05).Diameter of the largest part solid nodule,number of non-solid nodules,diameter of the largest non-solid nodule,nodule classification arising from I-ELACAP follow-up strategy,and nodule classification based on NCC guideline were screened as effective predictors.The area under curve(AUC)of the lung cancer risk prediction model was 0.955 and the C-index was 0.970.The calibration curve showed that the predictive results were in good consistency with the actual results.The AUC of independent verification was 0.810{95%CI(0.791,0.862)}.The integrated discrimination improvement of combined model was 0.061(Z=5.071,P=0.002),which means the performance of the combined model improved by 6.1%.The net reclassification index of combined model was 0.193(Z=2.405,P=0.015).Conclusions For the selection of high-risk population,age,smoking history,and education level can be considered reference risk factors.The lung cancer risk prediction model based on effective predictors shows high accuracy and discrimination,and performs well in generalization.For the selection of screening technology,tumor makers added to LDCT screening can provide limited additional benefits.PART ⅡDiagnostic value of conventional tumor markers and combined with chest CT in patients with stage IA lung cancerObjective To investigate the diagnostic efficiency of conventional serum tumor marker and tumor markers combined with chest CT in stage IA lung cancer.Materials and Methods From January 2016 to October 2020,1155 patients with stage IA lung cancer and 200 patients with benign lung lesions all confirmed by surgery were retrospectively enrolled.Six conventional serum tumor markers[carcinoembryonic antigen(CEA),carbohydrate antigen 125(CA125),squamous cell carcinoma associated antigen(SCCA),cytokeratin 19 fragment(CYFRA21-1),neuron-specific enolase(NSE),gastrin-releasing peptide precursor(ProGRP)]and chest thin-slice CT were performed on all patients 1 month before surgery.Pathology was taken as the gold standard to analyze the difference of positivity rate tumor markers between lung cancer group and benign group,moderate/poor differentiation group and well differentiation group,adenocarcinoma group and squamous cell carcinoma group,lepidic and non-lepidic predominant adenocarcinoma,solid nodules group and subsolid nodules group based on thin-slice CT,and subgroups of IA1～IA3 lung cancer.Then the diagnostic performance of tumor markers and tumor markers combined with chest CT was analyzed using the receiver operating characteristic(ROC)curve.Result ①Positivity rate of tumor markers:lung cancer group vs.benign group was 2.32%-20.08%vs.0-13.64%in single tumor marker detection,and 40.93%vs.18.18%in combined detection of tumor markers.The single and combined detection of lung cancer group was higher than that of benign group(χ2=8.20,P=0.004).Moderate/poor differentiation of lung cancer group was higher than well differentiation group(χ2=7.41,P=0.025).Squamous cell carcinoma group was higher than adenocarcinoma group(χ2=3.66,P=0.045).Non-lepidic adenocarcinoma was higher than lepidic adenocarcinoma(χ2=18.43,P=0.001).Solid nodules were higher than subsolid nodules in lung cancer group(χ2=8.10,P=0.028).②Positivity rate of tumor markers in different tumor staging(tumor size):stage IA1,IA2,IA3 was 33.33%,48.96%,69.23%,respectively.③Diagnostic efficacy of tumor markers and tumor markers combined with CT:combined detection was slightly better than marker detection alone,with sensitivity of 83.00%vs.74.00%,specificity of 78.30%vs,56.30%,and area under ROC curve of 0.721 vs.0.692.Conclusions For stage IA lung cancer,positivity rate of clinical tumor markers commonly used are generally low.Combined detection of 6 markers can increase the positivity rate.The positivity rate of markers tends to be higher in poorly differentiated lung cancer,squamous cell carcinoma,or solid nodules.Tumor markers combined with thin-slice CT show limited improvement of diagnostic efficiency for early lung cancer. |
