Measures
Model structure and screening strategies In this study, different screening strategies were formed by different screening inclusion criteria and screening diseases (LC only or LC, CVD, and COPD screening simultaneously). The screening inclusion criteria were set based on diverse screening starting ages (45, 50, 55, 60, 65, and 70 years), smoking duration (20, 25, 30 packyears) and years since quit smoking. The threshold of years since quit smoking was set to be 15 years in this study since15 years and shorter was recommended by most guidelines in China and other countries.5, 13–15 For brevity, we denoted a screening strategy as “disease screened”“age starting screening”“smoking packyear”“years since quit”. For example, the screening strategy which screened “big3 diseases” simultaneously and included adults aged 45 years and above, with a smoking history of at least 25 packyear, current smoker or smoking quitter for less than 15 years was denoted as 3D452515.
Microsimulation model was used to simulate clinical outcomes, cost, and effectiveness of adults over 45 years under different screening strategies. In this study, the state transition models of LC, CVD and COPD were constructed respectively, and the LC model was detailed in our previous study7. The CVD model was modified according to the research of Huang et al16, which included four states: healthy, nonfatal CVD event, post CVD event, and fatal CVD event. The model assumed that after LDCT examination and scoring of coronary artery calcification (CAC), population with a positive result would be treated with statins, reducing the probability of developing atherosclerotic cardiovascular diseases (ASCVD)17. The COPD model was established according to Behr et al., whose model included 6 states: healthy, GOLD stages (I, II, III, and IV), and death.1 The model hypothesized that COPD stage distribution would change in the presence of screening,1 and treatment of stage I COPD detected by screening with inhaled medication, rather than chronic bronchitis treatment, would slow the rate of disease progression (eFigure 1).18 At the end of each cycle of the model, participants would move to another state or remain in the original state according to the corresponding probability. All screening strategies were annual screening and the age of termination of screening was set as 74 years.15 The model consisted of 45 cycles, each lasting for 1 year, and would stop when all participants reach 90 years of age or die.
Parameters Smoking (starting and quitting probability), LC (incidence and death probability), and LDCT (sensitivity, specificity, screening compliance, overdiagnosis rate, excess relative risk of LC per screening) related transition probabilities are detailed in previous research.7
Incidence probability of ASCVD. The validated ChinaPAR model was adopted in this study to predict ASCVD risk in China.19 Independent variables in this model included gender, age, untreated and treated systolic blood pressure (SBP), total cholesterol, highdensity lipoprotein cholesterol (HDLC), waist circumference, smoking status, diabetes, geographic region, and family history of ASCVD. This model was able to predict 10year ASCVD Risk for a participant, which would be converted into annual incidence probability. The baseline data of 2011 and the third wave data of 2015 were used to establish the prediction model of each independent variable (except age and gender), so as to update the values according to each participant’s characteristics in each cycle. In addition, the first year and longterm recurrent probability during and after nonfatal CVD event were obtained from Li et al.20
Incidence probability of fatal CVD event. For healthy (asymptomatic) population, the probability of fatal CVD events related to age, gender, and smoking exposure were calculated respectively based on allcause and CVD mortality of Chinese population in 2011 obtained from China Health statistics Yearbook 2012,21 and the hazard ratio (HR) values in Asian populations from literature.22 Calculation method adopted was the calculation process of nonLC death probability in previous studies.7 For the population with nonfatal CVD events, the probability of fatalCVD events occurring in the first year was obtained by subtracting the nonCVD mortality (obtained from Yearbook 201221) from the death probability (obtained from Li et al20). The standardized mortality ratio (SMR) on CVD background mortality 1 year after the occurrence of nonfatal CVD events was obtained from Chen et al23, and then the fatalCVD probability was calculated.
Transition probability of COPD model. In this study, in order to calculate the probability of developing COPD, data of participants without COPD were obtained from the 2011 baseline data of CHALS and were matched with corresponding data in 2015. Then, logistic regression was used to construct a COPD risk prediction model. The independent variables of the model included age, gender, smoking status, smoking packyears and duration of smoking cessation, and the 4year COPD risk of an participant can be calculated, which can later be converted into annual transition probability. The GOLD grade stage distribution of COPD without screening was obtained from the study of Wang et al,24 and the stage distribution under screening was derived from the study of Behr1 and Mohamed et al25. The average time for COPD progression at different stages was obtained from literature, which assumed 50% participants would progress into the next stage during this average time. Annual COPD progression probability was calculated using this equation: \(P=1{\left({P}_{t}\right)}^{\frac{1}{t}}\)(P is the annual exacerbation rate, Pt refers to the nonexacerbation rate in t years)1. The mortality of COPD in different stages was obtained from relevant literature26.
Other parameters. The probability of CAC positive results was calculated by the prediction model proposed by Zhang et al.27 The ASCVD occurrence probability for people with CAC would decrease when receiving statin treatment, and relative risk was acquired from literature.28 Similarly, the probability of progressing from given stage to the next stage for COPD patients would decrease with inhaled treatment, and relative risk was calculated based on mean changing value through FEV1 per year.18
Cost. This study evaluated the cost and effectiveness from the perspective of healthcare system. There are four parts of the cost: screeningrelated, and LC, CVD, and COPD treatmentrelated cost. The screening and LC treatment related cost have been detailed in previous research.7
According to relevant guidelines,17 moderateintensity statins, as the initial treatment for lipid reduction, can achieve good clinical outcomes under longterm treatment29. The cost of statins in this study was calculated based on atorvastatin treatment for one year (10mg/d), and the drug price was taken as the average of published bidding prices of 31 provinces in China in 2022. Nonfatal and fatal CVD treatment costs were obtained from 2022 China Health Statistical Yearbook,30 and treatment costs for patients in postCVD status were obtained from literature.23 COPD treatment costs included inhalation therapy (mono bronchodilator) and maintenance therapy (e.g. oxygen inhalation, expectorant, etc.), which took reference from Qu et al.18 Patients of stage I COPD, when not detected, are normally treated as having chronic bronchitis. Treatment costs of COPD acute exacerbations included outpatient fees and stagespecific hospitalization fees.18 All costs were discounted in the year of 2022 and then converted into US dollars according to the average exchange rate (CNY 6.7261 to USD 1)31in 2022 (Table 1).
Table 1 Model Parameters
Parameters

Base case(range)

Distribution

Source

Discount rate

0.03(0, 0.08)

triangle(0, 0.03, 0.0.08)

42

Overdiagnosis rate when screening

0.11

beta(120,969)

43

RR of incident ASCVD after statin treatment

0.79(0.77, 0.81)

beta(1257.9, 334.4)

28

RR of exacerbation after treatment

0.88(0.85, 0.93)

beta(222.24, 30.1)

18

Specificity of LDCT

0.765(0.70, 0.93)

beta(56936, 17497)

4

Sensitivity of LDCT

0.937(0.89, 1)

beta(649, 44)

Screening compliance

0.3532(0.305, 1)

beta(197251,558480)

44

LC stagespecific annual probability of death


Estimated45

Stage I

0.047(±50%)

beta(414.708, 8380.292)

Stage II

0.109(±50%)

beta(200.565, 1631.435)

Stage III

0.22(±50%)

beta(1310.297, 4657.703)

Stage IV

0.43(±50%)

beta(379.543, 502.457)

Stage distribution with screening

Estimated 4, 4652

LC Stage I

0.623(0.563, 0.717)

beta(804, 487)

LC Stage II

0.091(0.067, 0.13)

beta(118, 1173)

LC Stage III

0.170(0.168, 0.177)

beta(220, 1071)

LC Stage IV

0.115(0.048, 0.130)

beta(149, 1142)

COPD Stage I

0.636(0.61, 0.662)

beta(807, 463)

1, 25

COPD Stage II

0.315(0.289, 0.341)

beta(400, 870)

COPD Stage III

0.046(0.034, 0.058)

beta(59, 1211)

COPD Stage IV

0.003(0, 0.006)

beta(4, 1266)

Stage distribution without screening

53

LC Stage I

0.19(0.152, 0.228)

beta(1331, 5682)

LC Stage II

0.164(0.131, 0.197)

beta(1161, 5852)

LC Stage III

0.347(0.278, 0.416)

beta(2432, 4581)

LC Stage IV

0.299(0.239, 0.359)

beta(2089, 4924)

COPD Stage I

0.537(0.523, 0.551)

beta(2719, 2189)

24

COPD Stage II

0.38(0.366, 0.394)

beta(1798, 3110)

COPD Stage III

0.074(0.067, 0.081)

beta(349, 4559)

COPD Stage IV

0.009(0.006, 0.012)

beta(42, 4866)

Excess relative risk of LC per screening

0.001(0.0003, 0.0019)

beta(6,5995)

8, 54

Health utility value of different stages of LC



LC Stage I

0.85(0.78, 0.89)

beta(136.78,24.14)

3234

LC Stage II

0.75(0.68, 0.8)

beta(149.31,49.77)

LC Stage III

0.69(0.56, 0.79)

beta(42.18,18.95)

LC Stage IV

0.69(0.38, 0.7)

beta(21.46,9.64)

Nonfatal CVD event

0.76(0.54, 0.96)

beta(2.1, 0.67)

1

Post CVD event

0.773(0.6, 0.9)

beta(15.31, 4.5)

28

COPD Stage I

0.897(0.65, 0.97)

triangle(0.65, 0.897, 0.97)

1, 18

COPD Stage II

0.755(0.58, 0.86)

triangle（0.58, 0.76, 0.86)

COPD Stage III

0.748(0.54, 0.80)

triangle(0.54, 0.748, 0.8)

COPD Stage IV

0.549(0.54, 0.80)

triangle(0.54, 549, 0.7)

Cost ($)


Screening related


Cost of LDCT test

53.49(44.62, 73.94)

gamma(4,13.3725)

55

Cost of publicity in screening

1.54(±50%)

gamma(4,0.385)

56

Cost of management in screening

1.85(±50%)

gamma(4,0.4625)

Cost of human resources in screening

3.08(±50%)

gamma(4,0.77)

57

Cost of diagnostic test

291.34(±50%)

gamma(4,72.835)

55

LC treatment related




Treatment for LC stage I

8704.68(±50%)

gamma(4,2176.17)

55

Treatment for LC stage II

13603.55(±50%)

gamma(4,3400.8875)

Treatment for LC stage III

14791.04(±50%)

gamma(4,3697.76)

Treatment for LC stage IV

19005.64(±50%)

gamma(4,4751.41)

CVD treatment related




Annual cost of statin

13.95(±50%)

gamma(4,3.4875)

Public bidding announcement

Treatment for nonfatal CVD event

2211.79(±50%)

gamma(4,552.9475)

30

Treatment for fatal CVD event

2915.08(±50%)

gamma(4,728.77)

30

Treatment for post CVD event

957.18(±50%)

gamma(4,239.295)

23

COPD treatment related




Treatment for COPD stage I

269.52(±50%)

gamma(4,67.38)

18

Treatment for COPD stage II

796.37(±50%)

gamma(4,199.09)

Treatment for COPD stage III

1016.51(±50%)

gamma(4,254.13)

Treatment for COPD stage IV

1016.51(±50%)

gamma(4,254.13)

Cost of exacerbation stage II

1383.42(±50%)

gamma(4,345.85)

Cost of exacerbation stage III

2706.2(±50%)

gamma(4,676.55)

Cost of exacerbation stage IV

4028.98(±50%)

gamma(4,1007.24)

Cost of chronic bronchitis treatment

529.23(±50%)

gamma(4,132.31)

LDCT, Lowdose computed tomography; LC, lung cancer; ASCVD, atherosclerotic cardiovascular diseases; CVD, cardiovascular diseases; COPD, chronic obstructive pulmonary disease; RR, relative risk.
Health utility value and discount rate. The health utility value of different status of models in this study was obtained from published studies.1, 18, 28, 32–34 If a participant had multiple diseases (LC, CVD, and COPD) at the same time, the health utility would be set as the one with the lowest utility value among coexisting diseases. An annual discount rate of 3% was adopted to discount the cost and utility value into 2022 equivalents.
Statistical Analysis
This study bootstrapped 100,000 real participants from included cases to maintain the correlations between the attributes of participants (e.g., age, sex, blood biomarkers). R (Version 4.1.2) was used to construct the model. Under different screening strategies, the cumulative personyears of different states over the entire model cycle were calculated. Healthcare costs and qualityadjusted lifeyears (QALYs) for all participants in different scenarios were calculated, and then the average cost and effectiveness of each strategy can be obtained. China’s GDP per capita in 2022 was 12,741 US dollars 31 . The program would be regarded costeffective, as recommended by WHO, if the incremental costeffectiveness ratio (ICER) is less than three times as much as China’s GDP per capita ($38,223 in this study). The ICER was calculated as incremental cost divided by incremental effectiveness compared with the baseline strategy. In addition, net health benefit (NHB), calculated based on different willingness to pay (WTP), was adopted to compare different strategies and to help select the optimal one. Oneway sensitivity analysis was conducted after changing each parameter over a plausible range to examine the effect of the uncertainty of each model parameter. In addition, probabilistic sensitivity analyses were conducted by running the model for 1,000 times, with each run sampled from the prespecified distributions of each parameter (Table 1).
The model was verified in the following aspects. First, the lung cancer incidence rate of each age group over 45 years old simulated by the model was compared with data in Yearbook 201235. Second, the nonLC, CVD, and COPD mortality rate of each age group over 45 years simulated by the model was compared with data in Yearbook 2012.36