| OBJECTIVESThis study aimed to evaluate the risk stratification potential of the current pre-screening strategy used in the Early Detection and Early Treatment for Esophageal Cancer in Huai River Regions;to provide optimal strategies for the identification of high-risk individuals for mass esophageal cancer screening by developing a model for esophageal squamous cell carcinoma(ESCC)in the Chinese population.Materials and Methods1.Evaluation of the pre-screening strategy in esophageal cancer screening:Participants without prior cancer history enrolled in the Esophageal Cancer Screening Program in Huai River Regions from 2007 to 2010 and those who had assessment results of the high-risk population were included in this study and were followed until December 31,2015.Screening areas included Xiping County,Henan Province,as well as Jinhu County,Jiangsu Province,and Tengzhou County,Shandong Province.Primary ESCCs diagnosed after a baseline assessment was the main outcome of interest.Secondary outcomes included all primary esophageal cancers and upper gastrointestinal cancers,diagnosed after baseline.A comprehensive evaluation of the association of the assessment results of the high-risk population with ESCC risk was performed by the Log-rank test,propensity score matching,and Cox regression analysis.2.Development and validation of a prediction model for ESCC:In this part,we collected epidemiologic and clinical data based on individuals aged 40-69 years from a prospective esophageal cancer screening cohort and a community-based randomized controlled trial for upper gastrointestinal cancer in Wuwei,Gansu Province to develop and validate a prediction model for ESCC.Existed potential risk factors for ESCC in terms of demographics,behavioral and dietary factors,disease history,and cancer family history were firstly reviewed and then fitted into an unconditional multiple logistic regression model for the development of ESCC within 3 years.A point-based risk prediction system was developed to select high-risk individuals for esophageal cancer screening.The discrimination of the score-based model was evaluated by the area under the receiver operating characteristic curve(AUROC)and its 95%confidence interval(CI),and goodness of fit was assessed by the Hosmer-Lemeshow(H-L)test.The point-based model was internally validated by leave-one-out cross-validation and externally validated in the validation population.To provide optimal cut-off values as a pre-screening strategy for mass esophageal cancer screening,we comprehensively estimated the proportions recommended to be screened for each total score,as well as sensitivity,specificity,Youden’s index,et.al.Results1.A total of 43,875 participants were included for analysis,among which 18,341 individuals were assessed as high-risk population,with the proportion of 41.80%.During approximately 240672.67 person-years of follow-up(median follow-up time of 5.5 years),235 newly diagnosed ESCC cases were identified,with 66 patients in the group of non-high-risk population(incidence density of 46.00/105 person-years)and 169 patients in the group of the high-risk population(incidence density of 173.90/105 person-years).Univariate analysis showed that participants in the group of the high-risk populations had higher incidence rates of ESCC,esophageal cancer,and UGI cancer than those in the non-high-risk group.Besides,the high-risk population had a higher risk of ESCC than the non-high-risk population,with a hazard ratio(HR)of 3.11(95%confidence interval(CI):2.33-4.14)after adjustment for sex,age,education level,income level,and body mass index.Additionally,the assessment results of the high-risk population were significantly associated with the risk of all esophageal cancers(HR=3.30,95%CI:2.51-4.33)and upper gastrointestinal cancers(HR=3.03,95%CI:2.43-3.76).2.A total of 86745 and 9626 participants aged 40-69 years were included to develop and validate the prediction model for ESCC,respectively.In the derivation cohort,a total of 298 individuals were diagnosed with ESCC within 3 years after baseline,during a median follow-up of 4.66 years.In the validation dataset,there were 18 individuals diagnosed with ESCC through endoscopy screening.The reference equation of the risk score model for ESCC was:Y=5.5×age(50 to 59 years old)/8.0×age(60-69 years old)+1.5 X sex(men)+2.5 X family history of UGI cancer(yes)+1.5 × smoking(<30 pack-years)/2.5 X smoking(≥30 pack-years)+1.5 X consumption of pickled and salted food≥once/week(yes)+1.0 X consumption of fresh fruits≥ once/week(no)+2.5 × esophageal clinical symptoms(yes)+1.5×family history of UGI disease(yes),with the total risk scores for each individual ranging from 0 to 21.This score-based risk prediction model for ESCC had a good discrimination ability,with an AUROC of 0.801(95%CI:0.779-0.823)and a good calibration(χ2=6.1918,P=0.6258 in H-L test).AUROCs in the one-leave-out cross-validation and external validation were 0.782(95%CI:0.759-0.805)and 0.855(95%CI:0.794-0.916),respectively.The cut-off threshold of 9 points with the highest Youden’s index had a 13.76%higher sensitivity and a 2.57%higher specificity compared with the strategy of high-risk population assessment applied in the esophageal cancer screening program in Huai River Regions.Conclusions1.The assessment strategy of the high-risk population applied in the esophageal cancer screening program in the Huai River Regions could be helpful for the selection of asymptomatic individuals for priority ESCC screening.More attention should be paid on the composition of assessment items,corresponding risk scores for each item,and the threshold for the high-risk population.2.A new prediction model based on eight routine epidemiological information was developed and validated in the Chinese population,which offered robust evidence for optimizing the current pre-screening strategy for mass esophageal cancer screening in China. |
