This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.
Lung cancer is the most commonly diagnosed cancer and the leading cause of cancer-related death in China. The effectiveness of screening for lung cancer has been reported to reduce lung cancer–specific and overall mortality, although the cost-effectiveness, optimal start age, and screening interval remain unclear.
This study aimed to assess the cost-effectiveness of lung cancer screening among heavy smokers in China by incorporating start age and screening interval.
A Markov state-transition model was used to assess the cost-effectiveness of a lung cancer screening program in China. The evaluated screening strategies were based on a screening start age of 50-74 years and a screening interval of once or annually. Transition probabilities were obtained from the literature and validated, while cost parameters were derived from databases of local medical insurance bureaus. A societal perspective was adopted. The outputs of the model included costs, quality-adjusted life years (QALYs), and lung cancer–specific mortality, with future costs and outcomes discounted by 5%. A currency exchange rate of 1 CNY=0.1557 USD is applicable. The incremental cost-effectiveness ratio (ICER) was calculated for different screening strategies relative to nonscreening.
The proposed model suggested that screening led to a gain of 0.001-0.042 QALYs per person as compared with the findings in the nonscreening cohort. Meanwhile, one-time and annual screenings were associated with reductions in lung cancer–related mortality of 0.004%-1.171% and 6.189%-15.819%, respectively. The ICER ranged from 119,974.08 to 614,167.75 CNY per QALY gained relative to nonscreening. Using the World Health Organization threshold of 212,676 CNY per QALY gained, annual screening from a start age of 55 years and one-time screening from the age of 65 years can be considered as cost-effective in China. Deterministic and probabilistic sensitivity analyses were conducted.
This economic evaluation revealed that a population-based lung cancer screening program in China for heavy smokers using low-dose computed tomography was cost-effective for annual screening of smokers aged 55-74 years and one-time screening of those aged 65-74 years. Moreover, annual lung cancer screening should be promoted in China to realize the benefits of a guideline-recommended screening program.
Lung cancer is a leading cause of death in China and globally. The incidence of lung cancer has recently increased dramatically, both in urban and rural areas, and it is currently the most common form of cancer in China. According to the National Central Cancer Registry of China, in 2015, the incidence of lung cancer was 57.26 cases/100,000 persons and the associated mortality rate was 45.87 deaths/100,000 persons, accounting for 20% and 27% of the values for all cancers, respectively [
The effectiveness of low-dose computed tomography (LDCT) for the screening of lung cancer has been confirmed by the National Lung Screening Trial conducted at 33 medical centers in the United States [
This study was conducted in 2 steps. In the first step, a Markov state-transition model with a lifetime horizon was used to mimic the natural progression of lung cancer and assess the potential impact of LDCT screening compared with a lack of screening in a Chinese cohort aged 50 to 74 years. In the second step, the Markov state-transition model combined with real-world data was used to estimate the ICER of each specific screening strategy as compared with nonscreening. A discount rate of 5% was applied to the costs of both strategies. Important assumptions in this study are summarized in
A simulated cohort of heavy smokers at a start age of 50-74 years was assumed to be followed up until the age of 79 years (mean life expectancy in China) or death.
A heavy smoker in this study was defined as a current smoker who smokes at least 20 pack-years.
Individuals in the screened cohort were assumed to undergo screening by low-dose computed tomography once or annually, and those with positive screening results were assumed to have undergone diagnostic biopsies.
While in the maintenance cancerous stages, the maintenance cost by stage was assumed to be 10% of the treatment cost.
All costs were expressed in CNY (2021; 1 CNY=0.1557 USD).
Future costs and effectiveness were discounted by 5%.
Adherence to screening and follow-up was assumed to be 100%.
The model simulated a cohort of 100,000 heavy smokers in China aged 50 to 74 years until the age of 79 years or death. A heavy smoker was defined as a current smoker who smokes at least 20 pack-years according to the China National Lung Cancer Screening Guidelines with LDCT (2018 version) [
Lung cancer is assumed to progress sequentially from less advanced to more advanced preclinical stages, as depicted in
Schematic diagram of natural history for lung cancer screening. CIS: carcinoma in situ.
Individuals in the nonscreened cohort were diagnosed based on symptoms. The probability of progression to a more advanced stage of lung cancer or a clinical diagnosis, as described by Ten Haaf et al [
It was assumed that patients in the screened cohort underwent screening by LDCT at least once or annually and those with positive results underwent additional testing, including biopsy. The positive result rate and proportion of lung cancer by stage were derived from the Wenling lung cancer screening program, which was initiated in 2018 to conduct annual LDCT screening of local high-risk populations over a 3-year period. Of 10,175 asymptomatic individuals who were screened in 2018, 65 (0.64%) were diagnosed with lung cancer (
Input parameters of the Markov model for lung cancer screening.
Variable | Base case value | Distribution | Source | |||||
|
Male | Female | Overall |
|
|
|||
|
|
|
|
|
|
|||
|
50-54 | 81.0559 | 89.6626 | N/Aa | Beta | [ |
||
|
55-59 | 162.0833 | 112.4574 | N/A | Beta | [ |
||
|
60-64 | 256.0943 | 154.6871 | N/A | Beta | [ |
||
|
65-69 | 373.6808 | 190.2521 | N/A | Beta | [ |
||
|
70-74 | 498.0681 | 242.6310 | N/A | Beta | [ |
||
|
|
|
|
|
|
|||
|
50-64 | 0.60 | 0.04 | N/A | Beta | [ |
||
|
65-74 | 0.45 | 0.07 | N/A | Beta | [ |
||
RRb (>20 pack-years) | N/A | N/A | 3.87 | Beta | [ |
|||
|
|
|
|
|
|
|||
|
CISc | N/A | N/A | 0.000 | Beta | [ |
||
|
I | N/A | N/A | 0.190 | Beta | [ |
||
|
II | N/A | N/A | 0.165 | Beta | [ |
||
|
III | N/A | N/A | 0.346 | Beta | [ |
||
|
IV | N/A | N/A | 0.299 | Beta | [ |
||
|
|
|
|
|
Wenling lung cancer screening program | |||
|
CIS | N/A | N/A | 0.0370 | Beta | N/A | ||
|
I | N/A | N/A | 0.6852 | Beta | N/A | ||
|
II | N/A | N/A | 0.0370 | Beta | N/A | ||
|
III | N/A | N/A | 0.1852 | Beta | N/A | ||
|
IV | N/A | N/A | 0.0556 | Beta | N/A | ||
Sensitivity of LDCT (%) | N/A | N/A | 79 | Beta | [ |
|||
Specificity of LDCT (%) | N/A | N/A | 81 | Beta | [ |
|||
|
|
|
|
|
|
|||
|
50-54 | N/A | N/A | 3.59 | Beta | [ |
||
|
55-59 | N/A | N/A | 4.73 | Beta | [ |
||
|
60-64 | N/A | N/A | 8.19 | Beta | [ |
||
|
65-69 | N/A | N/A | 12.99 | Beta | [ |
||
|
70-74 | N/A | N/A | 21.08 | Beta | [ |
||
|
|
|
|
|
|
|||
|
50-54 | N/A | N/A | 28.81 | Beta | [ |
||
|
55-59 | N/A | N/A | 52.86 | Beta | [ |
||
|
60-64 | N/A | N/A | 101.93 | Beta | [ |
||
|
65-69 | N/A | N/A | 153.34 | Beta | [ |
||
|
70-74 | N/A | N/A | 248.57 | Beta | [ |
||
|
|
|
|
|
|
|||
|
Lung cancer stage CIS to I | N/A | N/A | 0.0980 | Beta | [ |
||
|
Lung cancer stage I to II | N/A | N/A | 0.3682 | Beta | [ |
||
|
Lung cancer stage I to III | N/A | N/A | 0.0328 | Beta | [ |
||
|
Lung cancer stage I to IV | N/A | N/A | 0.0745 | Beta | [ |
||
|
Lung cancer stage II to III | N/A | N/A | 0.2260 | Beta | [ |
||
|
Lung cancer stage II to IV | N/A | N/A | 0.1510 | Beta | [ |
||
|
Lung cancer stage III to IV | N/A | N/A | 0.1455 | Beta | [ |
||
|
Lung cancer stage CIS to death | N/A | N/A | 0.00 | Beta | [ |
||
|
Lung cancer stage I to death | N/A | N/A | 0.04 | Beta | [ |
||
|
Lung cancer stage II to death | N/A | N/A | 0.07 | Beta | [ |
||
|
Lung cancer stage III to death | N/A | N/A | 0.13 | Beta | [ |
||
|
Lung cancer stage IV to death | N/A | N/A | 0.18 | Beta | [ |
||
|
|
|
|
|
|
|||
|
CIS | N/A | N/A | 0.87 | Beta | [ |
||
|
I | N/A | N/A | 0.84 | Beta | [ |
||
|
II | N/A | N/A | 0.84 | Beta | [ |
||
|
III | N/A | N/A | 0.87 | Beta | [ |
||
|
IV | N/A | N/A | 0.75 | Beta | [ |
||
|
|
|
|
|
Survey data | |||
|
Direct screening cost | N/A | N/A | 245.86 | Gamma | N/A | ||
|
Indirect screening cost | N/A | N/A | 23.07 | Gamma | N/A | ||
|
Prediagnosis cost | N/A | N/A | 628.36 | Gamma | N/A | ||
|
Biopsy diagnosis cost | N/A | N/A | 1232.44 | Gamma | N/A | ||
|
|
|
|
|
|
|||
|
CIS | N/A | N/A | 47,341.85 | Gamma | N/A | ||
|
I | N/A | N/A | 53,344.51 | Gamma | N/A | ||
|
II | N/A | N/A | 83,365.95 | Gamma | N/A | ||
|
III | N/A | N/A | 90,643.18 | Gamma | N/A | ||
|
IV | N/A | N/A | 116,471.34 | Gamma | N/A |
aN/A: not applicable.
bRR: relative risk.
cCIS: carcinoma in situ.
dLDCT: low-dose computed tomography.
eA currency exchange rate of 1 CNY=0.1557 USD is applicable.
The total cost of the screening program included direct expenses (ie, public advertising, management of screening invitations, salaries of staff members, and depreciation of screening equipment) and indirect expenses (ie, transportation and wages for missed work). In addition, the cost of diagnostic biopsies for participants with positive results after initial LDCT was considered. Screening-related costs were retrieved from data provided by the Wenling lung cancer screening program. Costs of treatment of lung cancer by stage were derived from a database of local medical insurance bureaus, which included 4947 patients and 107,248 relevant records. The cost of maintenance by stage accounted for 10% of the total treatment cost. All costs in this study are expressed in Chinese yuan (CNY) at a discount of 5% of rates in 2018. A currency exchange rate of 1 CNY=0.1557 USD is applicable.
The putative benefit of cancer screening for early diagnosis was assumed to be a difference in life expectancy and QALY after treatment. As the severity and responsiveness to treatment vary according to stage, the specified utility score for each stage was used for calculation [
As the scheduled screening program included several key characteristics, different combinations of screening intervals and start ages, as well as a nonscreening cohort, were evaluated (
Evaluation strategies.
Scenario | Screening tool | Screening interval | Start age (years) |
LDCTa#1 | LDCT | Annual | 50, 55, 60, 65, and 70 |
LDCT#2 | LDCT | One time | 50, 55, 60, 65, and 70 |
Nonscreening | N/Ab | N/A | 50, 55, 60, 65, and 70 |
aLDCT: low-dose computed tomography.
bN/A: not applicable.
The main outcomes of the cost-effectiveness analysis for each strategy were QALYs and total costs. The ICER was calculated by dividing the incremental costs by the incremental QALYs gained for each screening strategy as compared to nonscreening. In China, there is no regulated or published cost-effectiveness threshold. Hence, the threshold recommended by the World Health Organization (WHO) is commonly used. Given that 3 times the gross domestic product per capita was used as a reference point, a tentative threshold value of 212,676 CNY was adopted in this study.
The Markov state-transition model was developed using TreeAge Pro 2021 software (TreeAge Software, Inc). The parameters of direct screening cost, maintenance cost, discount rate, consumer price index (CPI) rate, incidence rate of heavy smokers, and specificity and sensitivity of LDCT uncertainty were investigated by 1-way deterministic sensitivity analyses. The costs of direct screening, as well as maintenance costs, CPI rate, and incidence rate of heavy smokers, were set to vary by 30% as compared to base values. The discount rate was set to range from 0% to 8%, and the sensitivity and specificity of LDCT were set to range from 0.63 to 0.95 and 0.65 to 0.97, respectively. Input parameters were randomly drawn from beta or gamma distributions (
The results of the model suggested that the QALYs of the screening cohort increased by 0.001 to 0.042 as compared to that of the nonscreening cohort. The reduction in lung cancer–associated mortality ranged from 0.004% to 1.171% for one-time screening and from 6.189% to 15.819% for annual screening (
The sensitivity of the model for the above-mentioned parameters is shown in
Base case results with different screening settings (per 100,000 persons).
Start age and strategya | Cost (CNY,b millions) | QALYsc (10,000 years) | Lung cancer mortality reduction vs nonscreening (%) | ICERd |
ICER |
|
|
|
|
|
|
|
|
|
Non_scr | 2489.69 | 135.92 | N/Ae | N/A | N/A |
|
Scr_once | 2552.16 | 135.93 | 0.0041 | 614,167.75 | N/A |
|
Scr_annu | 3410.57 | 136.30 | 6.1886 | 245,746.19 | 235,467.06 |
|
|
|
|
|
|
|
|
Non_scr | 2380.25 | 121.21 | N/A | N/A | N/A |
|
Scr_once | 2448.97 | 121.23 | 0.0145 | 365,289.96 | N/A |
|
Scr_annu | 3176.64 | 121.62 | 6.7044 | 192,119.62 | 183,886.78 |
|
|
|
|
|
|
|
|
Non_scr | 2154.69 | 104.08 | N/A | N/A | N/A |
|
Scr_once | 2230.61 | 104.11 | 0.0467 | 263,083.31 | N/A |
|
Scr_annu | 2808.69 | 104.50 | 7.7816 | 154,401.89 | 146,456.38 |
|
|
|
|
|
|
|
|
Non_scr | 1773.45 | 84.40 | N/A | N/A | N/A |
|
Scr_once | 1860.84 | 84.45 | 0.1997 | 192,574.66 | N/A |
|
Scr_annu | 2260.66 | 84.77 | 10.0628 | 131,284.57 | 122,745.38 |
|
|
|
|
|
|
|
|
Non_scr | 1184.22 | 61.56 | N/A | N/A | N/A |
|
Scr_once | 1279.79 | 61.61 | 1.1705 | 180,280.19 | N/A |
|
Scr_annu | 1476.25 | 61.80 | 15.8193 | 119,974.08 | 103,182.45 |
aNon_scr: nonscreening; Scr_once: one-time screening; Scr_annu: annual screening.
bA currency exchange rate of 1 CNY=0.1557 USD is applicable.
cQALY: quality-adjusted life year.
dICER: incremental cost-effectiveness ratio.
eN/A: not applicable.
Tornado diagrams. The tornado diagrams illustrate the change in the incremental cost-effectiveness ratio (ICER). The blue column shows the impact of decreasing the input parameters on the results. Similarly, the red column shows the impact of increasing the input parameters on the results. CPI: consumer price index; EV: expected value; LDCT: low-dose computed tomography.
Probabilistic sensitivity analyses. The screening strategies are labeled as follows: screening or not screening interval_start age. CEA: cost-effectiveness analysis; non_scr: nonscreening; scr_annu: annual screening; scr_once: one-time screening.
This is the first cost-effectiveness analysis of a lung cancer screening program with different start ages and screening intervals using real-world data in China. In summary, using a lifetime societal perspective for one-time or annual LDCT for screening of heavy smokers, the annual screening strategy with a start age of 55-74 years showed strong dominance as compared with the nonscreening strategy. These results were sensitive to the rate of newly developed lung cancer and the specificity of LDCT. As compared with the nonscreening strategy, the one-time screening strategy was cost-effective for patients aged 65-74 years, using a cost-effectiveness threshold of 212,676 CNY per QALY gained. This finding is consistent with that in the UK Lung Screen trial, which demonstrated a long-term benefit from a single screen and provided potentially important data for inclusion in future modeling studies to optimize the screening interval [
Although the analytical approach was somewhat similar to that in a previous study by Yuan et al [
Regarding the implications of policies related to lung cancer screening, the China National Lung Cancer Screening Guidelines with LDCT (2018 version) [
There were several limitations to this study that should be addressed. First, like most mathematical models, the model used in this study to estimate the incidence of lung cancer in heavy smokers was a simplification of the biological complexity of lung carcinogenesis and neglected the influence of various endogenous and exogenous risk factors, such as family history and residential/occupational exposure to radon, which may have led to underestimation of the incidence of lung cancer in the targeted population. Further, as heavy smokers are more likely to die from other diseases (eg, cardiovascular diseases and other cancers), its application to estimate the general probability of all-cause death in this population might have slightly underestimated the mortality rate in this work. Nevertheless, the use of this nomothetic approach has aided the development of prevention and control strategies against lung cancer in the United States [
This economic evaluation revealed that a population-based lung cancer screening program in China for heavy smokers using LDCT could result in more QALYs, although with greater expense than nonscreening. Using the WHO threshold for cost-effectiveness analysis, the annual screening strategy from 55 to 74 years and one-time screening strategy from 65 to 74 years can be considered cost-effective. Moreover, annual screening was the most promising; thus, annual screening should be promoted in China to realize actual benefits.
Validation of the natural history model of lung cancer.
carcinoma in situ
consumer price index
incremental cost-effectiveness ratio
low-dose computed tomography
odds ratio
quality-adjusted life year
relative risk
World Health Organization
Conceptualization: HD and ZZ; methodology: ZZ and LD; software: YL and LW; investigation: YW and YY; data curation: HD; writing-original draft preparation: ZZ; writing-review and editing: ZZ and LD; supervision: HD; funding acquisition: LD. All authors have read and agree to the published version of the manuscript.
None declared.