Estimated Cost-effectiveness of Endoscopic Screening for Upper Gastrointestinal Tract Cancer in High-Risk Areas in China.

Importance: Upper gastrointestinal tract cancer, including esophageal and gastric cancers, in China accounts for 50% of the global burden. Endoscopic screening may be associated with a decreased incidence of and mortality from upper gastrointestinal tract cancer. Objective: To evaluate the cost-effectiveness of endoscopic screening for esophageal and gastric cancers among people aged 40 to 69 years in areas of China where the risk of these cancers is high. Design, Setting, and Participants: For this economic evaluation, a Markov model was constructed for initial screening at different ages from a health care system perspective, and 5 endoscopic screening strategies with different frequencies (once per lifetime and every 10 years, 5 years, 3 years, and 2 years) were evaluated. The study was conducted between January 1, 2019, and October 31, 2020. Model parameters were estimated based on this project, government documents, and published literature. For each initial screening age (40-44, 45-49, 50-54, 55-59, 60-64, and 65-69 years), a closed cohort of 100 000 participants was assumed to enter the model and follow the alternative strategies. Main Outcomes and Measures: Cost-effectiveness was measured by calculating the incremental cost-effectiveness ratio (ICER), and the willingness-to-pay threshold was assumed to be 3 times the per capita gross domestic product in China (US $10 276). Univariate and probabilistic sensitivity analyses were conducted to assess the robustness of model findings.
Results: The study included a hypothetical cohort of 100 000 individuals aged 40 to 69 years. All 5 screening strategies were associated with improved effectiveness by 1087 to 10 362 quality-adjusted life-years (QALYs) and increased costs by US $3 299 000 to $22 826 000 compared with no screening over a lifetime, leading to ICERs of US $1343 to $3035 per QALY. Screening at a higher frequency was associated with an increase in QALYs and costs; ICERs for higher frequency screening compared with the next-lower frequency screening were between US $1087 and $4511 per QALY. Screening every 2 years would be the most cost-effective strategy, with probabilities of 90% to 98% at 3 times the per capita gross domestic product of China. The model was the most sensitive to utility scores of esophageal cancer- or gastric cancer-related health states and compliance with screening. Conclusions and Relevance: The findings suggest that combined endoscopic screening for esophageal and gastric cancers may be cost-effective in areas of China where the risk of these cancers is high; screening every 2 years would be the optimal strategy. These data may be useful for development of policies targeting the prevention and control of upper gastrointestinal tract cancer in China.

Introduction

The incidence of upper gastrointestinal tract cancer (UGIC), including esophageal cancer (EC) and gastric cancer (GC), is high in China, with an estimated 802 930 new cases (324 422 cases of EC and 478 508 cases of GC) and 674 924 deaths (301 135 due to EC and 373 789 due to GC) in 2020, accounting for approximately 50% of the global burden.[1] However, the geographic distribution of UGIC is uneven in China, mortality rates in some areas being 2- to 3-fold higher than the national average.[2,3,4]

The prognosis of UGIC depends mainly on the disease stage at the time of diagnosis. In China, the overall 5-year survival rates for patients with EC and GC are 30.3% and 35.1%, respectively.[5] However, the rates would be 86% and 90%, respectively, if disease were detected at an early stage.[6,7] This potential for increased survival has provided justification for early detection programs. A series of nationwide screening programs has been established in several East Asian countries where the incidence rates are high. In Japan, radiographic screening for GC was developed in the 1960s, and a national screening program was established in 1983; currently, endoscopic screening every 2 to 3 years is usually recommended.[8] South Korea introduced both radiographic screening and endoscopic screening for GC into national screening programs in 2000 and currently recommends endoscopic screening every 2 years.[9] Over approximately the past 20 years, endoscopic examination has become a major screening method for UGIC because of its high accuracy.[10] Several economic evaluation studies[11,12,13,14,15,16] from South Korea, Singapore, Portugal, the US, and China showed that endoscopic screening for EC or GC was cost-effective compared with no screening. Another study[17] conducted in the US suggested that the cost-effectiveness of combined endoscopic screening for EC and GC was comparable to that of funded screening programs for other cancers when it was integrated into the current colonoscopy screening program.

In China, several endoscopic screening programs for UGIC have been attempted in some areas with high incidence rates since 2005. Some observational studies[18,19,20,21,22] have shown that endoscopic screening can reduce the incidence of and mortality associated with UGIC. To assess the feasibility and efficacy of endoscopic screening for EC and GC, a multicenter randomized controlled trial was launched in 2015, in which more than 140 000 persons aged 40 to 69 years were enrolled.[23,24] A large amount of basic data has been obtained from this trial. The present study evaluated the cost-effectiveness of combined endoscopic screening for EC and GC in people aged 40 to 69 years in areas of China where the risk of these cancers is high. Furthermore, we evaluated the optimal initial age and frequency for screening.

Methods

This model-based economic evaluation was performed from the health care system perspective using TreeAge Pro (Healthcare Version) 2020 (TreeAge Software) and was conducted between January 1, 2019, and October 31, 2020. The project was registered with the Protocol Registration System in the Chinese Clinical Trial Registry and approved by the independent Ethics Committee of the National Cancer Center of China/Cancer Hospital, Chinese Academy of Medical Sciences. Because patient data were deidentified in the analysis, the requirement for informed consent was waived by the Ethics Committee of the National Cancer Center of China/Cancer Hospital, Chinese Academy of Medical Sciences. This evaluation followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guideline.[25]

Markov Model

A Markov model was constructed for different initial screening ages (40-44, 45-49, 50-54, 55-59, 60-64, and 65-69 years) to simulate EC and GC progression and calculate related health and economic outcomes in a lifetime horizon. Five endoscopic screening strategies with different frequencies were considered, including once per lifetime and every 10 years, 5 years, 3 years, and 2 years. No screening was applied as a reference strategy. For each initial screening age, a closed cohort of 100 000 participants with a mean age of 42, 47, 52, 57, 62, or 67 years was assumed to enter the model and follow the alternative strategies.

The Markov model consisted of 26 health states. In addition to the normal and death states, series of EC and GC progression states were considered, and each progression state was divided into undetected and detected states. A posttreatment state was also considered separately from the states requiring medical treatment after diagnosis. Endoscopic screening, reexamination, and related treatment procedures followed the recommendations of the Chinese Expert Consensus on Screening and Endoscopic Diagnosis and Treatment of Early Esophageal Cancer/Gastric Cancer (2014 version).[26,27] These recommendations were also followed in a randomized controlled trial by Chen et al.[23] Details of the trial are given in the eMethods and eFigures 1 to 3 in the Supplement, details of the Markov model are given in eFigure 4 in the Supplement, and validation of the model is shown in eFigures 5 and 6 in the Supplement. Compliance with screening, reexamination, and treatment and complications associated with endoscopic screening were also considered in the model. The model was run with a 1-year cycle length and terminated when the mean age of the cohort reached 90 years. A half-cycle correction was applied.

Model Parameters

Compliance with endoscopic screening was determined to be 49% (range, 30%-80%) according to a pilot project concerning endoscopic screening for EC in several areas of China where the risk of EC is high (Table 1).[18,28,29,30,31,32,33,34,35,36,37,38,39,40,41] Compliance with regular endoscopic reexamination among individuals who had mild esophageal dysplasia, moderate esophageal dysplasia, or low-grade gastric intraepithelial neoplasia detected on screening was assumed to be higher than compliance with endoscopic screening, which was 67% (range, 40%-90%) according to a previous report (Table 1).[15] The model assumed that false-positive results would result in loss of quality of life but could be corrected with an endoscopic reexamination at an additional cost. The sensitivity and specificity of endoscopic examination for EC were 96% (range, 88%-99%) and 90% (range, 59%-100%), respectively, according to previous studies in areas in China where risk of EC is high.[11,30,31] The sensitivity and specificity of endoscopic examination for GC were assumed to be 89% (range, 70%-98%) and 100% (range, 90%-100%), respectively, based on the published literature from other countries where risk for GC is high and other economic evaluation studies of GC[8,12,32,33,34] because of the lack of available data in China. Clinically significant complications (eg, bleeding or perforation) are rare among individuals undergoing diagnostic endoscopic examination.[42,43] The complication rate in the model was estimated to be 0.009% (range, 0%-0.2%) based on the trial by Chen et al[23,24] and other published studies (Table 2).[32,33,35]

Table 1.

Estimates of Parameters Used in the Model

Parameter	Base-case value	Range	Distribution	Source
Compliance with screening	0.49	0.30-0.80	Triangular (0.30, 0.49, 0.80)	Wei et al,[18] 2015; Feng et al,[28] 2015; Wang et al,[29] 2015
Compliance with reexamination	0.67	0.40-0.90	Triangular (0.40, 0.67, 0.90)	Yang et al,[15] 2015
Endoscopic examination characteristics
Sensitivity for EC	0.96	0.88-0.99	Triangular (0.88, 0.96, 0.99)	Chang et al,[11] 2012; Dawsey et al,[30] 1998; Nagami et al,[31] 2014
Specificity for EC	0.90	0.59-1.00	Triangular (0.59, 0.90, 1.00)	Chang et al,[11] 2012; Dawsey et al,[30] 1998; Nagami et al,[31] 2014
Sensitivity for GC	0.89	0.70-0.98	Triangular (0.70, 0.89, 0.98)	Zhou et al,[12] 2013; Hamashima et al,[8] 2018; Yeh et al,[32] 2016; Lee et al,[33] 2007; Hamashima et al,[34] 2013
Specificity for GC	1.00	0.90-1.00	Triangular (0.90, 1.00, 1.00)	Zhou et al,[12] 2013; Hamashima et al,[8] 2018; Yeh et al,[32] 2016; Lee et al,[33] 2007; Hamashima et al,[34] 2013
Endoscopic examination complications	0.00009	0-0.002	Triangular (0, 0.00009, 0.002)	Zeng et al,[24] 2020; Yeh et al,[32] 2016; Lee et al,[33] 2007; Espino et al,[35] 2012
Annual self-initiated examination
Severe esophageal dysplasia and CIS or HGIN and CIS	0.01	0.005-0.02	Triangular (0.005, 0.01, 0.02)	Chang et al,[11] 2012
Early EC or GC	0.20	0.10-0.40	Triangular (0.10, 0.20, 0.40)	Chang et al,[11] 2012
Advanced EC or GC	0.70	0.56-0.90	Triangular (0.56, 0.70, 0.90)	Chang et al,[11] 2012
Compliance with treatment
Severe esophageal dysplasia or CIS	0.7458	0.5625-0.9654	β (31.72, 10.81)	Chen et al,[23] 2017
Early EC	0.9405	0.7149-1.0000	β (8.17, 0.52)	Chen et al,[23] 2017
Advanced EC	0.9643	0.8393-1.0000	β (16.99, 0.63)	Chen et al,[23] 2017
HGIN or CIS	0.5455	0.4425-0.7746	β (92.43, 77.01)	Chen et al,[23] 2017
Early GC	0.9000	0.6792-1.0000	β (12.41, 1.38)	Chen et al,[23] 2017
Advanced GC	0.9643	0.8393-1.0000	β (16.99, 0.63)	Chen et al,[23] 2017
Costs, $
Screening mobilization and administration per capita	1.05	±50%	γ (0.16, 0.15)	Chen et al,[23] 2017
Endoscopic examination	47.87	±50%	γ (46.57, 0.97)	Chen et al,[23] 2017
Treatment for endoscopic complications	113.68	±50%	γ (5.82, 0.05)	Chen et al,[23] 2017
Initial treatment
Severe esophageal dysplasia or CIS	1604	±50%	γ (3.33, 0.002)	Yang et al,[36] 2018
Early EC	7732	±50%	γ (2.33, 3.01)	Yang et al,[36] 2018
Advanced EC	7320	±50%	γ (3.33, 4.55)	Yang et al,[36] 2018
HGIN or CIS	1423	±50%	γ (1.41, 9.88)	Yang et al,[36] 2018
Early GC	7548	±50%	γ (4.61, 6.11)	Yang et al,[36] 2018
Advanced GC	7086	±50%	γ (5.96, 8.41)	Yang et al,[36] 2018
Annual health care
Severe esophageal dysplasia or CIS	216	±50%	γ (1.22, 0.006)	Yang et al,[36] 2018
Early EC	367	±50%	γ (1.23, 0.003)	Yang et al,[36] 2018
Advanced EC	342	±50%	γ (2.05, 0.006)	Yang et al,[36] 2018
HGIN or CIS	243	±50%	γ (1.24, 0.005)	Yang et al,[36] 2018
Early GC	409	±50%	γ (1.18, 0.003)	Yang et al,[36] 2018
Advanced GC	435	±50%	γ (1.19, 0.003)	Yang et al,[36] 2018
Utility scores
Mild esophageal dysplasia	1.00	0.98-1.00	Triangular (0.98, 1.00, 1.00)	Sharaiha et al,[37] 2014; Inadomi et al,[38] 2009
Moderate esophageal dysplasia	1.00	0.98-1.00	Triangular (0.98, 1.00, 1.00)	Sharaiha et al,[37] 2014; Inadomi et al,[38] 2009
Severe esophageal dysplasia or CIS	0.84	0.79-0.89	β (3.57, 0.68)	Liu et al,[39] 2018
Early EC	0.70	0.66-0.74	β (2.63, 1.13)	Liu et al,[39] 2018
Advanced EC	0.61	0.56-0.66	β (1.12, 0.71)	Liu et al,[39] 2018
LGIN	1.00	0.98-1.00	Triangular (0.98, 1.00, 1.00)	Sharaiha et al,[37] 2014; Inadomi et al,[38] 2009
HGIN or CIS	0.92	0.86-0.99	β (2.53, 0.22)	Xia et al,[40] 2020
Early GC	0.75	0.71-0.78	β (3.15, 1.05)	Xia et al,[40] 2020
Advanced GC	0.57	0.53-0.62	β (1.35, 1.02)	Xia et al,[40] 2020
Discount rate	0.05	0-0.08	NA	Weinstein et al,[41] 1996

Abbreviations: CIS, carcinoma in situ; EC, esophageal cancer; GC, gastric cancer; HGIN, high-grade intraepithelial neoplasia; LGIN, low-grade intraepithelial neoplasia; NA, not applicable.

Table 2.

Base-Case Cost-effectiveness Results Compared Among Different Strategies by Initial Screening Age Among 100 000 Cohort Members

Initial screening age, strategy	QALYs	Incremental QALYs		Cost ($, thousand)	Incremental cost ($, thousand)		ICER ($/QALY)
		Vs no screening	Vs the next most effective strategyb		Vs no screening	Vs the next most effective strategyb	Vs no screening	Vs the next most effective strategyb
40-44 y
No screening	1 659 260	NA	NA	25 035	NA	NA	NA	NA
Screening once per lifetime	1 660 347	1087	1087	28 334	3299	3299	3035	3035
Screening every 10 y	1 663 677	4417	3330	31 954	6919	3620	1566	1087
Screening every 5 y	1 666 345	7085	2668	36 707	11 672	4753	1647	1781
Screening every 3 y	1 668 371	9111	2026	42 218	17 183	5511	1886	2720
Screening every 2 y	1 669 622	10 362	1251	47 861	22 826	5643	2203	4511
45-49 y
No screening	1 572 532	NA	NA	26 817	NA	NA	NA	NA
Screening once per lifetime	1 574 161	1629	1629	30 347	3530	3530	2167	2167
Screening every 10 y	1 576 941	4409	2780	33 961	7144	3614	1620	1300
Screening every 5 y	1 579 145	6613	2204	37 745	10 928	3784	1653	1717
Screening every 3 y	1 580 974	8442	1829	42 531	15 714	4786	1861	2617
Screening every 2 y	1 582 334	9802	1360	48 190	21 373	5659	2180	4161
50-54 y
No screening	1 468 506	NA	NA	30 229	NA	NA	NA	NA
Screening once per lifetime	1 470 890	2384	2384	34 142	3913	3913	1641	1641
Screening every 10 y	1 472 583	4077	1693	36 373	6144	2231	1507	1318
Screening every 5 y	1 474 773	6267	2190	40 356	10 127	3983	1616	1819
Screening every 3 y	1 476 325	7819	1552	44 120	13 891	3764	1777	2425
Screening every 2 y	1 477 651	9145	1326	49 010	18 781	4890	2054	3688
55-59 y
No screening	1 342 830	NA	NA	36 095	NA	NA	NA	NA
Screening once per lifetime	1 346 031	3201	3201	40 491	4396	4396	1373	1373
Screening every 10 y	1 347 264	4434	1233	42 811	6716	2320	1515	1882
Screening every 5 y	1 348 747	5917	1483	45 240	9145	2429	1546	1638
Screening every 3 y	1 350 365	7535	1618	49 171	13 076	3931	1735	2430
Screening every 2 y	1 351 491	8661	1126	52 843	16 748	3672	1934	3261
60-64 y
No screening	1 195 742	NA	NA	45 876	NA	NA	NA	NA
Screening once per lifetime	1 199 618	3876	3876	51 081	5205	5205	1343	1343
Screening every 5 y	1 201 091	5349	1473	53 970	8094	2889	1513	1961
Screening every 3 y	1 202 293	6551	1202	56 498	10 622	2528	1621	2103
Screening every 2 y	1 203 210	7468	917	58 779	12 903	2281	1728	2487
65-69 y
No screening	1 025 119	NA	NA	60 442	NA	NA	NA	NA
Screening once per lifetime	1 029 058	3939	3939	67 233	6791	6791	1724	1724
Screening every 2 y	1 030 594	5475	1536	70 810	10 368	3577	1894	2329

Abbreviations: ICER, incremental cost-effectiveness ratio; NA, not applicable; QALY, quality-adjusted life-year.

QALYs, costs, and ICERs are expressed as the values in 2019.

Compared with the next most effective strategy at the same initial screening age.

The prevalence rates of EC- and GC-related health states were estimated based on baseline screening reports from the trial by Chen et al[23] in areas where the risk of these cancers is high (eTable 1 in the Supplement); these reports were used to determine initial distributions of cohort members across health states in the model. A wide range was set for each rate to cover the values reported in these areas by referring to previous studies from China.[29,44,45,46] The annual transition probabilities were derived from published observational studies concerning the natural history of EC and GC and economic evaluation studies of EC and GC (eTable 2 in the Supplement).[15,47,48,49]

The model assumed that in the absence of active screening, individuals with severe esophageal dysplasia and carcinoma in situ, high-grade gastric intraepithelial neoplasia and carcinoma in situ, or EC or GC would receive a diagnosis on the basis of self-initiated examinations according to state-specific probabilities (Table 1).[11] Individuals with these diagnoses would receive state-specific treatments, whereas individuals who did not receive these diagnoses would remain untreated. The rates of state-specific compliance with treatment were calculated for the proportion of screened patients who actually completed the entire treatment procedure in areas where the risk of UGIC is high in the trial by Chen et al[23] (eTable 3 in the Supplement). The state-specific probabilities of recurrence after treatment and cancer-related mortality rates for advanced EC and GC were estimated based on the survival rates among patients with EC and GC (eTable 2 in the Supplement),[15] whereas age-specific natural background death rates were obtained from the China Population & Employment Statistics Yearbook, 2019.[50]

Costs were converted from Chinese renminbi to 2019 US dollars (US $1 = ¥6.8968 in 2019). The cost of screening included screening mobilization and administration costs, endoscopic examination costs, and costs of treatment for endoscopic complications, all of which were obtained from the 7 study centers that participated in the trial by Chen et al[23] (eTable 4 in the Supplement). The cost of EC- and GC-related treatment included the initial treatment cost after detection of the cancer and the subsequent annual health care cost after treatment, both of which were obtained from the survey included in the trial by Chen et al[23] that was administered to assess the economic burden of UGIC in China.[36] Calculation details are shown in eTable 5 in the Supplement.

The health outcome was utility-weighted life expectancy expressed as quality-adjusted life-years (QALYs). Utility scores of EC- and GC-related health states were obtained from the survey included in the trial by Chen et al[23] that was administered to assess the quality of life of patients with UGIC in China (eTable 6 in the Supplement).[39,40] Considering that patients diagnosed with mild esophageal dysplasia, moderate esophageal dysplasia, and low-grade gastric intraepithelial neoplasia do not have symptoms, we used a utility score of 1 in the base-case analysis and set a range of 0.98 to 1 in the sensitivity analyses.[37,38] We assumed a discount rate of 5% (range, 0%-8%) for both QALYs and costs.[41] The choice of distribution for all parameters was based on consideration of the properties of the parameters and the data informing the parameters.

Statistical Analysis

Effectiveness and Cost-effectiveness Evaluation

Expected QALYs and costs of each strategy were obtained from the model. We evaluated the cost-effectiveness of screening strategies at different initial screening ages with 2 approaches. First, we calculated the incremental cost-effectiveness ratio (ICER), defined as incremental costs per QALY gained of each screening strategy compared with no screening. Second, we calculated the ICER of each screening strategy compared with the next most effective screening strategy to identify the optimal strategy at each initial screening age. Because of the lack of data regarding the willingness to pay in the population of China, we adopted the cost-effectiveness definition used by the World Health Organization. Highly cost-effective was defined as an ICER less than 1 time the per capita gross domestic product (GDP) in China; cost-effective, an ICER of 1 to 3 times the per capita GDP; and not cost-effective, and an ICER greater than 3 times the per capita GDP.[51] The per capita GDP in China in 2019 was US $10 276.

Uncertainty and Sensitivity Analyses

Univariate sensitivity analyses of all parameters within their respective ranges were performed to identify the main sensitivity parameters. Probabilistic sensitivity analyses were further conducted to determine the probability of each screening strategy being cost-effective compared with no screening and the probability of each strategy being optimal compared with all other strategies.

Results

Base-Case Results

The study included a hypothetical sample of 100 000 individuals aged 40 to 69 years. All 5 screening strategies increased QALYs and costs compared with no screening at the same initial screening age (Table 2) by 1087 to 10 362 QALYs and US $3 299 000 to $22 826 000 for a cohort of 100 000 participants over a lifetime, and the corresponding ICERs were US $1343 to $3035 per QALY, which were lower than the per capita GDP ($10 276). Therefore, all screening strategies would be more cost-effective than no screening.

Further comparisons of the screening strategies were performed, and more frequent screening would be associated with a gain of more QALYs but with higher costs (Table 2). When a screening strategy was compared with the next most effective screening strategy in the cohort with the same initial screening age, 917 to 3939 QALYs would be gained and an additional cost of US $2 231 000 to $6 791 000 gained for a cohort of 100 000 participants over a lifetime; the corresponding ICERs were US $1087 to $4511 per QALY, which was lower than the per capita GDP. Therefore, the high-frequency screening strategy would be more cost-effective than the low-frequency screening strategy, and screening every 2 years would be the optimal strategy at each initial screening age. Screening every 2 years and screening once per lifetime had the most incremental QALYs at the initial screening ages of 40 to 44 years and 65 to 69 years, respectively.

Sensitivity Analysis Results

Univariate sensitivity analyses revealed that the results remained largely unchanged over the plausible range of each parameter. The ICER upper limits of the most sensitive parameters of each screening strategy at different initial screening ages are shown in Table 3. When a screening strategy was compared with no screening, only varying utility scores of EC- and GC-related health states resulted in ICERs exceeding the per capita GDP at initial screening ages of 40 to 44 years and 65 to 69 years. When a screening strategy was compared with the next most effective screening strategy, varying utility scores of EC- and GC-related health states resulted in ICERs at some initial screening ages exceeding 3 times the per capita GDP, and varying compliance with screening resulted in ICERs of screening every 2 years vs screening every 3 years exceeding the per capita GDP. Univariate sensitivity analyses demonstrated the stability of the cost-effectiveness and ranking of the cost-effectiveness of the screening strategies.

Table 3.

Upper Limits of Incremental Cost-effectiveness Ratio for Each Screening Strategy in Univariate Sensitivity Analyses by Initial Screening Age

Initial screening age and screening strategy	Utility scoresb		Compliance with screening		Discount rate		Prevalences of EC- and GC-related health statesb		Costs of screeningc		Annual transition probability from severe esophageal dysplasia or CIS to early EC
	Vs no screening	Vs the next most effective strategy	Vs no screening	Vs the next most effective strategy	Vs no screening	Vs the next most effective strategy	Vs no screening	Vs the next most effective strategy	Vs no screening	Vs the next most effective strategy	Vs no screening	Vs the next most effective strategy
40-44 y
Screening once per lifetime	11 778d	11 778d	4016	4016	6671	6671	4711	4711	4422	4422	4592	4592
Screening every 10 y	4498	2877	1883	1271	3440	2144	1737	1129	2272	1570	2186	1658
Screening every 5 y	5116	6395	1845	2277	3467	3513	1811	1927	2397	2603	2308	2513
Screening every 3 y	6202	11 273d	2085	5204	3871	5260	2082	3003	2758	4022	2609	3659
Screening every 2 y	7452	19 296d	2619	13 156d	4480	8663	2451	5127	3240	6747	2996	5740
45-49
Screening once per lifetime	6153	6153	2760	2760	4486	4486	3448	3448	3092	3092	2677	2677
Screening every 10 y	5236	4571	1912	1473	3248	2358	1934	1389	2304	1842	2259	1986
Screening every 5 y	5310	5455	1853	2155	3246	3241	1949	1979	2370	2502	2284	2333
Screening every 3 y	6217	10 190	2020	4874	3583	4782	2206	3090	2687	3830	2539	3455
Screening every 2 y	7678	22 118d	2552	10 504d	4128	7457	2605	5002	3163	6125	2936	5387
50-54 y
Screening once per lifetime	3976	3976	1674	1974	3216	3216	2609	2609	2296	2296	2134	2134
Screening every 10 y	4253	4844	1564	1409	2897	2367	2010	1479	2122	1876	2069	1972
Screening every 5 y	5210	7982	1850	2752	2998	3197	2091	2219	2292	2609	2238	2560
Screening every 3 y	5803	8366	2233	6715	3252	4255	2320	3550	2542	3197	2424	3161
Screening every 2 y	7025	17 484d	2821	12 807d	3695	6278	2710	4887	2956	5401	2768	4787
55-59 y
Screening once per lifetime	3271	3271	1394	1394	2601	2601	2127	2127	1882	1882	1920	1920
Screening every 10 y	4841	52 896e	1527	2043	2746	3194	2151	2191	2084	2608	2194	2954
Screening every 5 y	5026	5620	1680	2573	2777	2868	2186	2277	2155	2366	2210	2258
Screening every 3 y	6162	13 002d	2076	5658	3046	4045	2476	3421	2443	3496	2455	3346
Screening every 2 y	6920	12 320d	2523	13 669d	3357	5377	2814	5008	2745	4770	2694	4255
60-64 y
Screening once per lifetime	3635	3635	1360	1360	2476	2476	1995	1995	1794	1794	2054	2054
Screening every 5 y	5310	31 253e	1566	2611	2697	3324	2235	2736	2055	2741	2324	3033
Screening every 3 y	6082	11 374d	1805	4871	2847	3515	2448	3321	2228	3001	2459	3044
Screening every 2 y	6492	9464	2076	11 769d	3011	4159	2683	4334	2399	3621	2586	3459
65-69 y
Screening once per lifetime	8653	8653	1741	1741	3128	3128	2375	2375	2185	2185	3053	3053
Screening every 2 y	14 302d	25 436d	2019	4801	3372	4011	2747	3644	2456	3153	3313	3959

Abbreviations: CIS, carcinoma in situ; EC, esophageal cancer; GC, gastric cancer; ICER, incremental cost-effectiveness ratio.

Only main sensitive parameters are shown; ICERs are expressed as the value in 2019.

As a set of parameters, all the parameters changed simultaneously with a positive correlation in univariate sensitivity analyses.

Including screening mobilization and administration costs, endoscopic examination costs, and treatment costs for endoscopic complications. The 3 parameters changed simultaneously with a positive correlation in univariate sensitivity analyses.

The upper limit of ICER was higher than the per capita GDP but lower than 3 times the per capita GDP.

The upper limit of ICER was higher than 3 times the per capita GDP.

The results of the probabilistic sensitivity analyses of each screening strategy compared with no screening are shown in Figure 1. The probability of each screening strategy being cost-effective increased as the willingness-to-pay threshold increased and reached 91% to 98% even at a willingness-to-pay threshold of the per capita GDP. The results of the probabilistic sensitivity analyses conducted for each strategy compared with all other strategies are shown in Figure 2. Screening every 2 years maintained its dominance from a willingness-to-pay threshold of less than the per capita GDP to a willingness-to-pay threshold equal to 3 times the per capita GDP at each initial screening age, with probabilities of 82% to 93% and 90% to 98% for being optimal at 1 and 3 times per capita GDP, respectively. Therefore, all screening strategies would be more cost-effective than no screening, and screening every 2 years would be the optimal strategy at a willingness-to-pay threshold of 3 times the per capita GDP.

Figure 1.

Cost-effectiveness Acceptability Curves of All Screening Strategies Compared With No Screening by Initial Screening Age

Dashed vertical blue lines represent per capita gross domestic product (GDP); dashed vertical black lines represent 3 times per capita GDP. QALY indicates quality-adjusted life-year.

Figure 2.

Cost-effectiveness Acceptability Curves of All Strategies Competing With Each Other by Initial Screening Age

Dashed vertical blue lines represent per capita gross domestic product (GDP); dashed vertical black lines represent 3 times per capita GDP. QALY indicates quality-adjusted life-year.

Discussion

Upper gastrointestinal tract cancer continues to be a major public health burden in areas in China where incidence and mortality rate for this cancer are higher than those in other areas. The present study targeted local residents aged 40 to 69 years in areas where the risk of UGIC is high who were most likely to benefit from endoscopic screening according to the age characteristics of UGIC incidence and life expectancy in China, and this population accounts for a large proportion of patients with UGIC in China. Our base-case results suggest that combined endoscopic screening for EC and GC would be more cost-effective than no screening regardless of the initial screening age or screening frequency; this finding is consistent with the conclusion reported in a systematic review.[52] Screening every 2 years would be the optimal strategy, and an initial screening age of 40 to 44 years was associated with the most health benefits. A screening frequency of 1 to 5 years has been previously proposed,[53] with the optimal interval being less than 3 years.[54] Most cases of early UGIC that were missed at the first endoscopic examination and subsequently identified in the second screening within 3 years were still amenable to curative surgery,[55] which could theoretically increase the sensitivity of endoscopic examination after 2 consecutive screenings.[13] The optimal strategy in this study was similar to the guidelines recommended in Japan and South Korea,[8,9] where the incidence rates are as high as those in China. Areas in China where the risk of UGIC is high are usually rural and have limited health resources and an underdeveloped economy.[56,57] Policy makers should consider the cost and effectiveness of the screening strategy, local economic level, and disease burden of UGIC when choosing appropriate screening strategies. If screening once per lifetime, which was the least expensive screening strategy, is preferable, the optimal initial screening age suggested by this study is 65 to 69 years.

Univariate sensitivity analyses revealed that utility scores of EC- and GC-related health states and compliance with screening were the 2 main sensitivity parameters. In this study, because most patients diagnosed with mild esophageal dysplasia, moderate esophageal dysplasia, or low-grade gastric intraepithelial neoplasia did not have symptoms, we used a utility score of 1 in the base-case analysis and a range of 0.98 to 1 in the sensitivity analyses according to other reports.[37,38] Whether any decrements in the quality of life could be present in patients diagnosed with mild esophageal dysplasia, moderate esophageal dysplasia, or low-grade intraepithelial neoplasia if they were asymptomatic and received education regarding their true cancer risk is unknown.[17] Although previous studies have found that patients with precancerous lesions have a poorer quality of life than the general population,[58,59] whether the decrement in the quality of life is caused by coexisting symptoms (eg, esophagitis, gastric ulcers) or the patient’s perception of cancer risk remains unknown.[60] Future studies should focus on the effect on the quality of life among those diagnosed with precancerous lesions. Compliance with screening affected whether the optimal strategy was screening every 2 years or screening every 3 years at a willingness-to-pay threshold of the per capita GDP. Endoscopic screening may be associated with reduced incidence and mortality of UGIC, and improving compliance among the target population is critical for achieving prevention effectiveness. The compliance with screening in this study was only 49%, which was derived from the areas with the highest incidence of UGIC in China[18]; these areas have the most longstanding promotion times and the most abundant experience with endoscopic screening and have residents with the highest cognitive understanding of endoscopic screening. Endoscopic screening is relatively expensive and slightly invasive, resulting in discomfort when individuals are examined. In addition, most of the target population, aged 40 to 69 years (especially men), in areas where there is high risk of UGIC are the primary sources of family income and migrate elsewhere for work. Therefore, compliance with endoscopic screening for UGIC is still low in China, and further popularization and promotion of endoscopic screening are needed to improve compliance.

Limitations

This study has limitations. The accuracy of the model depends on the accuracy of parameter estimates. First, base-case initial probabilities of EC- and GC-related health states were derived from the study by Chen et al.[23] Residents of areas where there is high risk of UGIC enrolled in the project may have had a higher incidence of this cancer; thus, the estimated parameters may not represent the status of the entire region. Second, annual progression or regression transition probabilities of health states should increase or decrease with age in the real world. However, those between specific precancerous lesion states were fixed owing to the lack of relevant observational data. As a result, the cost-effectiveness of screening strategies may be overestimated in younger age groups and underestimated in older age groups. Third, compliance with different screening frequencies was assumed to be consistent in the model. In the real world, a higher screening frequency may be associated with a reduction in compliance because of patients’ concerns about pain and sedation and competing life demands.[61,62] Fourth, annual screening was not considered as an alternative strategy in this study because it has not been recommended in any country until now. In addition, the burden may be difficult to address in countries with low GDP.

Conclusions

To our knowledge, this is the first comprehensive cost-effectiveness analysis of endoscopic screening for both EC and GC in China. The findings from the present study suggest that from the perspective of the health care system, combined endoscopic screening for EC and GC would be highly cost-effective for people aged 40 to 69 years in areas of China where there is a high risk of UGIC; screening every 2 years would be the optimal strategy. The findings provide important evidence for policies targeting the prevention and control of UGIC in China.