Establishment And Evaluation Of Risk-stratified Screening Strategies For Colorectal Cancer In Chinese Population

Background:Traditional screening strategies,including colonoscopy-based or fecal immunochemical test(FIT)-based screening strategies are effective for decreasing the incidence and mortality of colorectal cancer.In 2011,a risk-stratified screening strategy was established by the Asia-Pacific Working Group on Colorectal Cancer Screening.In this strategy,screening subjects are assessed by the Asian-Pacific Colorectal Screening(APCS)score,based on which ones who are assessed as high-risk ones were recommended for colonoscopy-based screening,and the other are recommended for FIT-based screening.This strategy is expected to further improve the effectiveness and cost-benefits of screening when compared with the traditional screening strategies and has been applied in parts of the Asia-Pacific region.However,the accuracy of the risk stratification method used in this strategy has not been fully verified,and there is a lack of a risk-stratified screening strategy suitable for Chinese population,and its long-term effectiveness and cost-effectiveness are unclear.Objective:The present study aims to establish scientifically effective colorectal cancer risk-stratified screening strategies suitable for Chinese population and to evaluate their long-term screening effectiveness and cost-effectiveness.This will provide scientific evidence for formulating the national policies on colorectal cancer prevention and control.Methods:This study is based on a Chinese multicenter population colorectal cancer screening randomized controlled trial called the Target-C study.The present study included 3,144 screening subjects who underwent colonoscopy examination.of these subjects,270 cases of colorectal cancer and advanced adenoma were included as the case group,while 2,874 cases of non-advanced adenoma and normal subjects were included as the control group.Firstly,differences in predictive efficacy between the APCS score and other risk assessment models were compared.Differences in screening accuracy between different categories of FIT(quantitative FIT vs.qualitative FIT)were also compared.Then,this study established risk-stratified screening strategies based on the APCS score and quantitative FIT with different positivity thresholds.Screening accuracy and the singleround screening effectiveness of these strategies were evaluated.Finally,a Markov microsimulation model was used to evaluate the long-term effectiveness and costeffectiveness of these strategies and to compare them with colonoscopy strategy and FIT strategy.1.For risk assessment models,the published risk assessment models for colorectal cancer based on environmental factors were collected through a systematic review.These models were evaluated separately in our study,and were compared with APCS score with respect to model discrimination and model calibration;For FIT,three qualitative FITs with different positivity thresholds(8.0,14.4,20.8 μg Hb/g,respectively)and a quantitative FIT were included.Comparisons of screening accuracy were conducted between each of the three qualitative FITs and the quantitative FIT with the positivity threshold recommended by the manufacturer(20 μg Hb/g).Furthermore,the comparisons were conducted after adjusting thresholds of the quantitative FIT to yield the same positivity thresholds or specificities(90%,92%,and 97%)as each of the three qualitative FITs respectively.Accuracy rate was used to evaluate the screening accuracy,and statistical differences were tested using paired χ2 test.2.The results of the first part did not observe significant differences in screening accuracy between different types of risk assessment models,but showed that the quantitative FIT had better screening accuracy than the qualitative FITs.Therefore,a framework of riskstratified screening strategy was constructed based on the APCS score and quantitative FIT as it was at present.Based on the framework,nine scenarios were constructed by a pairwise combination of the positivity thresholds of APCS score(3,4,and 5 points)and FIT(10,20,and 30 μg Hb/g).Screening accuracy of these scenarios was compared with indicators used in screening test and by comparing the likelihood ratio in the receiver operator characteristic curve(ROC)space,respectively.In the ROC space,the proportional relationship between incremental false-positive number and incremental true-positive number ratio was plotted for these nine scenarios compared with the FIT strategy.A decision tree model was used to compare the single-round screening effectiveness,among these nine scenarios,colonoscopy strategy,and FIT strategy to select the optimal combination of positivity thresholds for APCS score and for FIT in the risk-adapted screening strategy.3.A Markov microsimulation model was established to simulate colorectal cancer screening in Chinese population based on the parameters with respect to cost of screening and diagnosis and treatment,as well as different utility values of health states.The model was used to evaluate the long-term effectiveness and cost-effectiveness of the riskstratified screening strategy,and to compare it with the no-screening,colonoscopy strategy,and FIT strategy.For long-term effectiveness analysis,the incidence rate and mortality rate of colorectal cancer were evaluated.For cost-effectiveness analysis,quality-adjusted life year(QALY)was used to evaluate the utility,while RMB(yuan)was used as the monetary unit of all cost data.All the data on the utility and costs were discounted into 2015.Incremental cost-effectiveness ratio(ICER)was calculated,and one time to three times gross domestic product(GDP)per capita per QALY of China(49,200 to 147,600 yuan)in 2015 were used as the cost-benefit thresholds.Deterministic sensitivity analyses and probabilistic sensitivity analyses were used to verify the result stability.Results:1.In this study,a total of eleven sets of risk assessment models including APCS score were included,which showed small discrepancies for colorectal cancer and advanced adenoma with an area under the curve(AUC)of the ROC curve ranging from 0.612(95%CI:0.583-0.640)to 0.640(95%CI:0.607-0.673).The goodness of fit for these models was good,and all P values for the included models were>0.05 when using Hosmer-Lemeshow tests.For FIT,when using the positivity thresholds recommended by the manufacturer,the quantitative FIT had a higher accuracy rate of 91.3%(95%CI:90.0%～92.5%),while the accuracy rates of the three qualitative FITs were 85.7%(95%CI:84.0%～87.3%;P<0.001),89.3%(95%CI:88.0%-90.7%;P<0.001),90.6%(95%CI:89.2%-92.1%;P=0.044),respectively.After adjusting to the same positive thresholds,the accuracy rate of quantitative FIT is still higher than that of each qualitative FIT;however,comparable accuracy rates were observed between each of the three qualitative FIT and the quantitative FIT(all P values>0.05).2.The nine scenarios of risk-stratified screening strategy increased sensitivity from 26.3%(95%CI:19.0%-33.3%)to 74.4%(95%CI:67.6%-81.2%),with a subsequent decrease in specificity from 90.8%(95%CI:89.4%-92.4%)to 46.6%(95%CI:44.3%-49.0%).The comparison of likelihood ratios within the ROC space showed no absolute advantage of incremental false-positive number versus incremental true-positive number ratio between these nine scenarios and the FIT strategy.In the single-round screening,when the positivity threshold of the APCS score was set as the 4 points,the risk stratified screening scenario achieved better efficacy and cost-benefit,missing 12.5%to 15.0%of cases of colorectal cancer and advanced adenoma,while saving 48.7%to 50.1%of colonoscopy resources and 33.5%to 35.0%of total costs compared with the colonoscopy strategy.3.With the participation rate of 40%for colonoscopy and of 90%for FIT,all screening strategies have reduction in incidence rates of colorectal cancer from 42%to 59%,and in mortality rates from 51%to 78%,when compared with no-screening group.The ICERs of these strategies ranged from 7,030 yuan/QALYs to 11,223 yuan/QALYs when compared with no-screening group.All of these strategies were lower than the one-time GDP per capita per QALYs of China in 2015 and were extremely cost effective.Further analysis using cost-effectiveness diagrams showed that FIT strategy and colonoscopy strategy were superior economically than the risk-stratified strategies.When the participation rate of colonoscopy reached 80%and the participation rate of FIT maintained 90%,the colonoscopy strategy(colonoscopy screening every 10 years)and risk-stratified screening strategy II(colonoscopy screening every 10 years for the high-risk ones,and FIT screening every two years for the low-risk ones)became cost effective(ICER=18,038 yuan/QALY,and 11,102 yuan/QALY).At this point,however,the former needs 2.3 times of colonoscopies greater than the risk-stratified screening strategy.The results of sensitivity analyses were stable.Conclusions:1.There was little difference between APCS score and other risk assessment models based on environmental factors in predicting colorectal cancer and advanced adenoma;The accuracy rate of quantitative FIT is better than qualitative FIT,but this discrepancy disappeared when they yielded the similar specificities.2.The risk stratification screening strategy based on APCS score with positivity thresholds of 4 points could ensure the screening effectiveness and save colonoscopy resources and costs,when compared with colonoscopy strategy in single round of screening,which has the potential value of population promotion.3.Currently,with the low participation rate of colonoscopy and the high participation rate of FIT among Chinese population,the risk-stratified screening strategy has not shown better long-term effects and benefits than FIT.With the increase in the participation rates of colonoscopy,the economic effectiveness of the risk-stratified screening strategy will be superior to FIT,and will not only ensure the cost effectiveness,but also save colonoscopy resources,compared with the colonoscopy strategy.