Evaluation Of Opportunistic Lung Cancer Screening In The High Incidence Areas Of Lung Cancer And Study Of The Corresponding Risk Prediction Model

BackgroundLung cancer remains to be one of the malignant tumors with the highest morbidity and mortality in China.Weihai is a city located in the northeastern shore of Shandong Province with a higher lung cancer incidence.Signs and symptoms of lung cancer typically appear only when the disease become severe and advanced,and treatments for late-stage lung cancer usually lead to poor prognosis.As a consequence,the 5-year survival rate of advanced lung cancer is less than 10%.Therefore,timely diagnosis and treatment of early lung cancer is an important secondary prevention measure to improve patient survival.Results from the randomized controlled trial(RCT)conducted by the U.S.-based National Lung Cancer Screening Trial(NLST)show that screening for lung cancer using low-dose computed tomography(LDCT)reduced mortality from lung cancer in high-risk individuals by 20%.Many health organizations are calling on governments and international institutions to start offering lung cancer screening through LDCT to high-risk population,and many lung cancer screening guidelines are issued.High-risk individuals are generally referred to heavy smokers,a few guidelines consider other risk factor of lung cancer(such as chronic obstructive pulmonary disease(COPD),etc.).The epidemiological characteristics of lung cancer,however,are constantly changing and vary among different populations.In China,there is no unified standard for the identification of highrisk individuals.Therefore,economic and effective cancer screening should be carried out to accurately identify high-risk groups of cancer,so as to improve the detection rate of lung cancer and optimize the allocation of resources.There are two types of screening,"policy-or program-driven organized population-based screening"(such as the Urban Cancer Early Diagnosis and Early Treatment Project launched in 2012)and "clinical opportunistic screening jointly determined by doctors and participants".However,population-level screening coverage and participation rate of high-risk people received LDCT screening are rather limited,and the criteria of high-risk groups are rough,resulting in lower screening efficiency.The "Healthy China Initiative(2019-2030)" clearly points out that for cancers with high incidence,relatively mature screening measures and technical solutions such as lung cancer,local governments are urged to promote universal opportunistic cancer screening according to the cancer epidemic situation in the region.Like population-based screening,large-scale opportunistic screening for lung cancer requires evidence-based medical evidence of the safety,effectiveness,efficiency,and economy.(1)Effectiveness—Whether it is effective to increase detection rate of early lung cancer,reduce lung cancer mortality in the screened population and improve the prognosis of lung cancer population?At present,whether opportunistic screening for lung cancer is effective lacks systematic research evidence.(2)Efficiency—Having accurate,reliable,simple and easy-touse risk assessment tools for assessing high-risk groups.Most of the lung cancer risk prediction models are based on smokers from Western population.The study design is mainly based on case-control studies with low-quality evidence,and many cohort-based models lack external validation.One of the Guidelines for the Screening and Early Detection of Lung Cancer in China recommends the establishment of risk prediction model and the risk score system of lung cancer based on Chinese population to improve the accuracy of lung cancer screening.(3)Economy-Getting the maximized health benefits at the lowest cost.The research on the health economic evaluation of lung cancer screening in China is still in its primary stage,and there are limited number of studies on the economics evaluation of opportunistic screening for lung cancer with different conclusions.Objective1.Evaluation of the "effectiveness" of opportunistic screening for lung cancer.To evaluate whether opportunistic screening effectively increases the detection rate of early lung cancer,reduces lung cancer mortality and all-cause mortality in the screened population based on the regional lung cancer screening cohort in areas with high incidence of lung cancer.To evaluate whether opportunistic screening improves the prognosis of lung cancer patients based on the regional lung cancer cohort in areas with high incidence of lung cancer.2.Evaluation of the "economy" of opportunistic screening for lung cancer.Based on the regional lung cancer screening cohort in areas with high incidence of lung cancer,economic evaluations of opportunistic screening for lung cancer were preliminarily evaluated.3.Construct an " efficient" risk prediction model of lung cancer.To develop and validate a lung cancer risk prediction model to identify high-risk population for the efficient implementation of opportunistic screening and early diagnosis of lung cancer in large-scale population based on the regional lung cancer screening cohort in areas with high incidence of lung cancer.Methods1.Data collection and collation.(1)Lung cancer screening cohort from Weihai Municipal Hospital Healthcare Group(2013-2021,59633 participants).The data were collected including demographic characteristics,baseline disease characteristics,blood indexes and outcomes of lung cancer.In addition,we linked to the database of tumor registration data and cause-of-death registration data of Shandong Provincial Center for Disease Control and Prevention.The short-term outcome indicators were the incidence of early lung cancer(stage 0-I)and lung cancer,and the long-term outcome indicators were lung cancer-specific death and all-cause death.The study was followed until December 30,2021.The data was used to evaluate the effectiveness and economy of opportunistic screening for lung cancer,and construct the risk prediction model of lung cancer as the training cohort.(2)Lung cancer cohort from Weihai Municipal Hospital Healthcare Group(2016-2021,5246 cases).Data were collected including demographic indicators,tumor characteristics,baseline disease characteristics,blood indexes and treatment information.The outcomes were lung cancer-specific mortality and all-cause mortality.The study was followed until December 30,2021.This data was used to evaluate the effectiveness of opportunistic screening for lung cancer.(3)The health examination data of residents’ electronic health records(EHR)in the Shandong Multi-Center Healthcare Big Data Platform(SMCHBDP)in Shandong Province(2013-2019,35824 people)was used for external validation of the risk model of lung cancer.Information regarding the predictors in the model was collected and the outcome was the occurrence of lung cancer by October 30,2020.2.Evaluation of the "effectiveness" of opportunistic screening for lung cancer.Propensity score matching(PSM)was applied to evaluate of the effectiveness of opportunistic screening for lung cancer.(1)Generating a propensity score using L1-regularized logistic regression(Least Absolute Shrinkage and Selection Operator,LASSO)based on the selected covariates to get the probability of LDCT screening.(2)The matched samples were obtained by performing nearest neighbor matching,with a caliper set at 0.2 of the standard deviations of the logit of the propensity score in a 1:1 ratio between the two groups.Balance diagnosis was carried out on the matched data evaluated by the absolute value of the standardized mean difference(SMD)of each covariate.SMD less than 0.10 was considered to be accepted.(3)The correlation(the population hazard ratio(HR)and 95%confidence interval(CI))between LDCT screening and the interested outcomes was analyzed by the univariate Cox regression model based on the matched data.(4)Sensitivity analysis.Using the propensity score regression adjustment and inverse probability treatment weighting(IPTW)method to estimate whether the effects were consistent with the effect of the PSM method.3.Evaluation of the "economy" of opportunistic screening for lung cancer.The health economic evaluation was conducted based on data from the LDCT opportunistic screened group(22059 participants)in the lung cancer screening cohort.The health economics indexes included the detection rate of lung cancer,early detection rate,cost,cost-effectiveness ratio(CER)and early detection cost index(EDCI).4.Construction and validation of the lung cancer risk prediction model.(1)Univariable Cox regression was first performed in the training cohort(the lung cancer screening cohort)to select the candidate factors,and LASSO-Cox algorithm was constructed to identify the final predictors of the model by including the variables with a P<0.05 in the univariable Cox model.The risk prediction model was built using the Cox regression model.(2)Internal validation was performed using 1000 Bootstrap resamples and 10-fold crossvalidation.Discrimination ability(concordance index,C-index),calibration ability(calibration curve),utility in clinical application evaluation(decision curve)were used to assess the performance of the established model in both training cohort and validation cohort(EHR-based physical examination cohort).(3)The model was converted to a score model to facilitate clinical application.The coefficients in the model were divided by the minimum coefficient in the model,multiplied by 2,and rounded to an integer,that is,the corresponding score.The sum of total scores was the individual lung cancer risk score.The score model was evaluated according to the receiver operating characteristic curve(ROC),area under curve(AUC),sensitivity,specificity,Youden index etc.Model reporting follows the "Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis"(TRIPOD)and"Strengthening the Reporting of Observational Studies in Epidemiology"(STROBE).Results1.Cohort description.(1)A total of 59633 participants were enrolled in the lung cancer screening cohort,with a median age of 51 years and 30963(51.92%)males.The median follow-up time was 4.20 year.There were 607 cases of lung cancer,with an incidence density of 230.95/100000 person-years.Screening-detected lung cancers were substantially more often diagnosed in early stage(70.30%).Dealing with death as an outcome,the median follow-up time was 4.22 years.There were 755 deaths with the death density 286.19/100000 person-years.There were 99 deaths from lung cancer with the death density 37.53/100000 person-years.A total of 22059 participants(36.99%)received LDCT screening.As the follow-up time increased,though the incidence of lung cancer in the LDCT screened group was higher than that in the non-screened group,the mortality of lung cancer in the non-screened group showed an increasing trend compared with that in the screened group.(2)A total of 5246 lung cancer cases were enrolled with a median age of 63 years and 2725 were males(51.94%).The median follow-up time was 1.69 years,and the longest followup time was 5.99 years.There were 1666 deaths and 1539 deaths from lung cancer.2251 patients(42.91%)were diagnosed with,lung cancer through opportunistic screening.Lung cancer survival was higher in the opportunistic screening group than in the non-opportunistic screening group(symptomatic group)with P<0.001.(3)A total of 35824 participants were enrolled in the EHR-based physical examination cohort.The median age was 58 years and 17250(48.15%)were males.The median follow-up time was 3.45 years.There were 651 cases of lung cancer,with an incidence density of 494.21/100000 person-years.2.Evaluation of the"effectiveness" of opportunistic screening for lung cancer.(1)In the lung cancer screening cohort,we used the 10-fold cross-validation LASSOLogistic algorithm to select 57 optimal predictors such as age,gender,smoking,and family history of lung cancer.After 1:1 matching according to propensity-score matching,32206 individuals(16103 in each group)were finally included.The baseline characteristics of the matched individuals were balanced between the two groups.Early-stage lung cancer incidence density was 114%higher(HR=2.14,95%CI,1.58-2.89),lung cancer-specific mortality was 75%lower(HR=0.25,95%CI,0.12-0.52)and all-cause mortality was 55%lower(HR=0.45,93%CI,0.35-0.57)for participants in the screened group compared with those in the nonscreened group.The results obtained in propensity score regression adjustment and IPTW were similar with those obtained by PSM.(2)In the lung cancer cohort,we used the 10-fold cross-validation LASSO-Logistic algorithm to select 46 optimal predictors such as age,smoking,and COPD.After 1:1 matching according to propensity-score matching,2754 individuals(1377 in each group)were finally included.The baseline characteristics of the matched individuals were balanced between the two groups.Lung cancer-specific mortality was 50%lower(HR=0.50,95%CI,0.42-0.61)and all-cause mortality was 48%lower(HR=0.52,95%CI,0.44-0.63)for participants in the screened group compared with those in the non-screened group.The results obtained by propensity score regression adjustment and IPTW were consistent with those obtained by PSM.3.Evaluation of the "economy" of opportunistic screening for lung cancer.Among 22059 participants who underwent LDCT opportunistic screening in the lung cancer screening cohort,265 patients were diagnosed with lung cancer.The detection rate of lung cancer was 1.20%,the early detection rate was 86.49%,the total cost was 18630638 Chinese yuan,the CER was 70304.29 Chinese yuan per lung cancer cases,and the EDCI was 1.29.4.Lung cancer risk prediction model used in identifying high-risk individuals.The variables included in the model after variable selection were age,family history of lung cancer,COPD,pneumoconiosis and interstitial lung disease(ILD).The C-index after 1000 Bootstrap resamplings was 0.75(95%Cl,0.73-0.78)in internal validation and 0.73(95%Cl,0.70-0.76)in the validation cohort.The calibration curve showed good agreement both in training cohort and validation cohort.The model was transformed into a score model:age(4549:2 points,50-54:4 points,55-59:5 points,60-64:6 points,65+:7 points);family history of lung cancer(Yes,6 points);COPD(Yes,2 points);pneumoconiosis(Yes,6 points);ILD(Yes,2 points).The AUC of 10-fold cross-validation was 0.75(95%CI,0.72-0.78)in the training cohort and 0.74(95%CI,0.72-0.77)in the validation cohort.Taking 5 or 6 points as the threshold,the Youden index was relatively high.About 45-56 people needed to be screened to detect 1 case of lung cancer.Conclusion1.Based on the regional lung cancer screening cohort study in areas with high incidence of lung cancer,the study demonstrated that the opportunistic screening with LDCT effectively increases the detection rate of early lung cancer,reduces lung cancer mortality and all-cause mortality in the screened population.Based on the regional lung cancer cohort study in areas with high incidence of lung cancer,opportunistic screening was further verified to improve the prognosis of lung cancer patients.2.The EDCI was less than 5,indicating that LDCT opportunistic screening was economic and it can be more beneficial if the screening was carried out in high-risk population.3.The study established a lung cancer risk prediction model and validated it for identifying high-risk individuals.The prediction model contained risk factors including age,family history of lung cancer,COPD,pneumoconiosis,and ILD.The model had good discriminative ability,calibration ability when applied to the different population who went for health examination in basic public health services.The prediction model had the potential to be applied in clinical opportunistic screening efficiently.