Helicobacter Pylori Related Gastric Cancer Screening and Cost-Effectiveness Analysis: A Hospital-Based Cross-Sectional Study (SIGES)

Abstract

To evaluate the effectiveness of Helicobacter pylori (Hp) related gastric cancer screening (Hp-GCS) and cost-effectiveness of protocol candidates in a hospital-based cross-sectional study. A total of 163 gastric cancer patients in West China Hospital were retrospectively collected according to ICD-10 code and histologic proof between April 1, 2013 and March 31, 2014, and 15,599 cancer-free controls were simultaneously collected from the health checkup registry. Hp infection was examined by urea breath test (UBT). The prevalence of Hp infection was compared between patients and controls. The diagnostic performance of UBT-based predictive index was tested in both training and validation settings. Cost-effectiveness analysis was conducted to assess candidates of Hp-GCS protocols. The prevalence of Hp infection was 55.8% and 41.2% in gastric cancers and controls, respectively (p < 0.001). UBT-based model showed moderate diagnostic strength in Hp-GCS (AUC = 0.78, 95% CI 0.74–0.82), better than UBT alone (p < 0.001). The sensitivity and specificity of UBT-based index were 80.2% and 61.9% at optimal cutoff in training setting, comparable in validation setting, which sensitivity and specificity were 76.9% and 59.6%. Number needed to screen was decreased along with older age, as well as stronger positivity of UBT. The optimal cost-effective Hp-GCS protocol with detection rate (DR = 77.9%) was endoscopic screening in age 40–59 years and positive UBT, or age ≥60 years without UBT. Incremental analysis suggested a preferable protocol as endoscopic screening in age ≥40 years without UBT (DR = 93.3%). UBT had moderate diagnostic strength in massive gastric cancer screening, and might be cost-effective in middle-aged population (40–59 years). More robust Hp-GCS protocol needs further investigate in test methods and individual biologic features.

Introduction

Gastric cancer is a common malignant tumor with high morbidity and mortality in China (1, 2). In recent years, gastric cancer has become the third largest burden of cancer in China and an important part of disease burden in China (3). China is one of the countries with the most bloated burden of gastric cancer around the globe, with 421,540 people dying from the disease in 2019, contributes the biggest number of incident cases, deaths, and disability-adjusted life-years (DALYs) from gastric cancer on earth (4, 5). Moreover, due to the aging of the population and other factors, the number of new cases and deaths of gastric cancer in China will further grow in the next 25 years, the disease burden of gastric cancer in China is still grievous (4). Although the 5-year survival rate of early gastric cancer can reach more than 90% (6), the rate for advanced gastric cancer is quite low, only 15.8% (7). Therefore, early detection of tumor would be helpful to improve survival. Massive screening can increase the proportion of detected patients with early gastric cancer. Currently, Japan and South Korea have relatively complete screening strategies for gastric cancer (8). Smoking and high-salt diet are momentous risk factors for gastric cancer (9–11). China has formulated a series of tobacco-control strategies (12, 13), and advocated citizens’ low-salt diet (14). Since 2005, China has plotted and implemented a range of screening and early detection projects, which have been expanded from high-risk areas to the whole country in 2019 (15). Presently, the target population of a screening program is people aged 40–69 (16, 17). However, China is a country with a huge population, as the most effective measures of gastric cancer screening, the cost of gastroscopy is also high, it is generally assumed that the entire population endoscopic screening is impracticable (18, 19). In addition, due to limited medical resources and lower population participation, there is no programmatic massive gastric cancer screening in China (20). It is indispensable to develop an economical and effective screening program for gastric cancer to improve detection rate based on the high selected candidates for endoscopy (21, 22).

Helicobacter pylori (Hp) is closely related to the development of gastric cancer, which was listed as a class I carcinogen by the World Health Organization (WHO) in 1994 (23, 24). China was a country with a high infection rate of Helicobacter pylori, and the prevalence of Helicobacter pylori infection for the general population was 55.8% (25). It was reported that the infection rate of Helicobacter pylori in Chengdu, Sichuan Province was about 53.1% (26). Long-term infection of Hp can cause atrophic gastritis, a precancerous lesion, and further generate gastric cancer (27, 28). The role of Hp-related gastric cancer screening (Hp-GCS) for endoscopy candidates has been reported in a large number of literature studies (29, 30). Iris, et al. pointed out that due to the low cost of screening test, Hp-GCS had certain cost-effectiveness value (31). For the diagnosis of Hp, urea breath test (UBT) has relatively high sensitivity and specificity, with low cost (32). Therefore, this study was aimed to identify the diagnostic strength of UBT for gastric cancer and the cost-effectiveness of Hp-GCS.

Methods

Study Design

This was a retrospective, high-volume, hospital-based, cross-sectional study within the Sichuan Gastric Cancer Early Detection and Screening (SIGES) project. The SIGES research group was conducted in West China Hospital, Sichuan University, a central hospital in Sichuan province. The period of the study was between April 2013 and March 2014 (33). The subjects covered healthy controls and gastric cancer patients, who were managed or treated in West China Hospital. The West China Hospital is a renowned national medical center in China, located in Chengdu, Sichuan Province of western China. Compared with other hospitals in Sichuan, even in China, it has the most advanced medical equipment all around the world and is one of the hospitals with the highest medical level in China. It is estimated that the West China Hospital receives about 1000 patients with gastric cancer for surgery annually. A total of 15,762 observations were included in this SIGES study.

Ethics

This study retrospectively collected the observations’ basic information and results of 14-C UBT. The SIGES study was approved by the Biomedical Ethical Committee of West China Hospital, Sichuan University (id: 2015-151-V2; 2018-215-V1). The informed consent was waived by the approval of the Biomedical Ethical Committee due to the retrospective nature, while the personal information was anonymized during the whole procedure of analyzing and reporting data.

Eligibility

The eligibility was assessed based on the electronic medical records reported by general practitioners, gastroenterologists, or gastrointestinal surgical oncologists. The healthy controls were collected from the health checkup in the Center of Health Checkup. The general practitioners recorded them as the asymptomatic status and cancer-free status. The gastric cancer patients were collected from the Department of Gastrointestinal Surgery. The diagnosis was proved by the ICD-10 code C16 and histology, regardless of TNM stage. Other kinds of malignancies were excluded, such as lymphoma and gastrointestinal stromal tumor. In addition, all the included observations should be tested by the 14-C UBT. For the gastric cancer patients, the UBT should be performed preoperatively, while the postoperative tests would not be considered.

General Information

The results of gender (male, female), age (<20, 20–39, 40–59, ≥60), ethnicity (Han, Tibetan, Yi, others), 14-C urea breath test (–, 1+, 2+, 3+, 4+) and direct medical expenses of patients with gastric cancer and healthy control group were collected. All the data were collected from the information management system of West China Hospital. When patients visit the hospital, their relevant information will be registered.

14-C Urea Breath Test

Fasting (at least for 3 h) tests took in a 14-C urea capsule (1μCi) with warm water, and then their breath was collected 15 mins, later by blowing to the breath card for 5 mins. The sample was tested by YH04 Helicobacter Pylori Analyzer (Anhui Young Hearty Medical Appliance & Equipment Co.; Ltd., China), and disintegrations per minute (dpm) was measured. The definitions of positivity and severity were according to manufacturer’s instructions. The outcomes were automatically classified as negative (−), mild positive (1+), moderate positive (2+), strong positive (3+), and very strong positive (4+). The cases were diagnosed with Hp infection for any 1+ to 4+ UBT results, and the cases with 3+ or 4+ UBT results were considered severe infections.

Statistics

Demographic characteristics and UBT results of the observations were compared by Chi-square test for categorical variables, and Wilcoxon rank-sum test for ranked variables or continuous variables without normality distribution. Shapiro-Wilk test for normal data was used to check the normality of variables. Multivariable analyses were performed by Logistic regression, and adjusted odds ratios (aORs) with 95% confidence intervals (CIs) were estimated. Non-parametric ROC analysis was used to assess the capability of predicting gastric cancer cases, while sensitivity (SEN) and specificity (SPE) were estimated. The area under the curve (AUC) with standard error (SE) was calculated. The diagnostic strength of UBT alone for gastric cancer was not high. Age, sex and ethnicity was related to gastric cancer in Logistic regression analysis. Therefore, UBT-based index was generated from the model of Logistic regression. The maximal Youden index (=SEN + SPE) was calculated to determine the optimal cutoff of UBT-based index. The difference of AUCs was compared by the Z test (34, 35). Number needed to screen (NNS) was calculated in diverse subpopulation, stratified by sex, age group, and positivity strength of UBT, and relative ratio (RR) was estimated. The cost-effectiveness ratio (CER) was estimated throughout nine candidate protocols of Hp-GCS. The direct medical cost contained UBT and endoscopy according to different protocols. The effect was number of detected cases by endoscopy. The detection rate (DR) of each protocol of Hp-GCS was estimated. Additionally, incremental cost-effectiveness ratio (ICER) was estimated to identify the protocol of higher detection rate regarding willing-to-pay manner. The strength of AUC, SEN, SPE or DR was classified as weak (0.5–0.7), moderate (0.7–0.9), and strong (0.9–1). All the tests were two-sided and statistical significance was defined as p < 0.05. The STATA/SE 12.0 software was used for statistical analysis.

Results

In this study, a total of 15,599 healthy controls and 163 gastric cancer patients were analyzed. The basic characteristics of gastric cancer patients and cancer-free controls are shown in Table 1. There were statistically significant differences, older age and more males in gastric cancer patients, comparing to controls (p < 0.001), but not in terms of ethnicity (p = 0.119). The prevalence of Hp infection was 55.8% and 41.2% in gastric cancers and controls, respectively (aOR = 1.79, 95% CI 1.31–2.46, p < 0.001) (Table 2). Age ≥ 40 years and males were associated with higher prevalence of Hp infection.

Table 1. Baselines between cancer-free controls and gastric cancer cases.

Baselines	Controls, n (%)	Gastric cancer, n (%)	p value
Total	15,599	163
Age (yrs)			<0.001^a
<20	367 (2.3)	0
20–39	4,410 (28.3)	11 (6.7)
40–59	7,969 (51.1)	66 (40.5)
≥60	2,853 (18.3)	86 (52.8)
Sex			<0.001^b
Male	7,237 (46.4)	100 (61.4)
Female	8,362 (53.6)	63 (38.6)
Ethnicity			0.119^b
Han	15,358 (98.5)	158 (96.9)
Tibetan, Yi, others	241 (1.5)	5 (3.1)
Setting	(4:1)	(2:1)	0.002^b
Training	12,204 (78.2)	111 (68.1)
Validation	3,395 (21.8)	52 (31.9)

a, Wilcoxon test; b, Chi-square test.

Table 2. Hp prevalence between cancer-free controls and gastric cancer cases.

	Controls Hp +, n (%)	Gastric cancers Hp +, n (%)	aOR	p value
Total	6,419 (41.2)	91 (55.8)	1.79 (1.31–2.46)	<0.001
Subsets
Age
20–39 yrs	1,773 (40.2)	5 (45.5)	1.27 (0.39–4.15)	0.699
40–59 yrs	3,386 (42.5)	41 (62.1)	2.21 (1.34–3.64)	0.002
≥60yrs	1,141 (40.0)	45 (52.3)	1.60 (1.04–2.47)	0.033
Sex
Male	2,984 (41.2)	58 (58.0)	1.98 (1.32–2.96)	0.001
Female	3,435 (41.1)	33 (52.4)	1.55 (0.94–2.55)	0.086
Ethnicity
Han	6,277 (40.9)	87 (55.1)	1.78 (1.30–2.45)	<0.001
Tibetan, Yi, others	142 (58.9)	4 (80.0)*	1.89 (0.19–18.97)	0.588

* A trend of higher Hp prevalence in ethnic minority, but without significance.

UBT-based multivariate model showed moderate diagnostic strength in predicting gastric cancer (AUC = 0.78, 95% CI 0.74–0.82), better than UBT alone (AUC = 0.60, 95% CI 0.55–0.65) (p < 0.001) (Figure 1). The diagnostic strength of UBT-based model in each subset of age group or sex was not a significant difference (p > 0.05) (Table 3). The SEN and SPE of UBT-based index were moderate 80.2% (95% CI 71.5–87.1%) and weak 61.9% (61.1–62.8%) at optimal cutoff in training setting (Table 4), comparable in validation setting moderate 76.9% (63.2–87.5%) and weak 59.6% (57.9–61.3%), respectively. Providing SPE at 80%, SEN of UBT-based index was estimated as weak 59.5% (49.7–68.7%).

Figure 1. ROCs of UBT alone and UBT-based model.

Table 3. AUC values of UBT predicting gastric cancer in training setting.

Category	AUC	95% CI	SE	p value
UBT alone	0.60	0.55–0.65	0.027	Ref
Model^a	0.78	0.74–0.82	0.021	<0.001
Subsets^b
Age (years)
20–39	0.65	0.43–0.87	0.112	Ref
40–59	0.67	0.59–0.75	0.040	0.861
≥60	0.64	0.56–0.71	0.040	0.925
Sex
Male	0.79	0.75–0.84	0.023	Ref
Female	0.75	0.67–0.83	0.039	0.362

Abbreviations: AUC: area under the curve; CI: confidence interval; Ref: reference; SE: standard error; UBT: urea breath test.

^aThe model included four variables, i.e., dpm value of UBT, age (years), gender, and ethnicity.

^bThe AUCs among the subsets were calculated using the model above.

Table 4. Diagnostic performance of the UBT-based index ^a generated from UBT for predicting gastric cancer.

Setting and cutoff	TP	FP	FN	TN	SEN	SPE
	N	N	N	N	% (95% CI)	% (95% CI)
Training setting
Optimal cutoff ^a	89	4,646	22	7,558	80.2 (71.5–87.1)	61.9 (61.1–62.8)
SPE at 80%	66	2,438	45	9,766	59.5 (49.7–68.7)	–
SPE at 90%	49	1,219	62	10,985	44.1 (34.7–53.9)	–
SEN at 80%	89	4,646	22	7,558	–	61.9 (61.1–62.8)
SEN at 90%	100	7,304	11	4,900	–	40.2 (39.3–41.0)
Validation setting
Optimal cutoff ^b	40	1,371	12	2,024	76.9 (63.2–87.5)	59.6 (57.9–61.3)

Abbreviations: CI: confidence interval; FN: false negative; FP: false positive; SEN: sensitivity; SPE: specificity; TN: true negative; TP: true positive; UBT: urea breath test.

^aIndex = 0.000432× dpm (value) + 0.0702× age (yrs) – 0.5756× sex (female =1, male = 0) + 1.5833× ethnicity (Tibetan, others =1, Han =0).

^bThe optimal cutoff was >2.864 for positivity, which was settled by the maximal Youden index in training setting.

NNS was significantly lower along with older age, as well as stronger positivity of UBT (Table 5). NNSs were 109 males (RR = 145.23, 95% CI 17.43–1,228.73) and 268 females (RR = 58.73, 95% CI 7.87–534.55) in age group ≥60 years with UBT positivity 3+∼4+, compared to reference male <40 years with negative UBT (NNS = 15,762). The optimal CER of Hp-GCS protocol was endoscopic screening in age 40–59 years and positive UBT, or age ≥60 years without UBT (DR = 77.9%) (Table 6, Figure 2). ICER suggested a preferable protocol as endoscopic screening in age ≥40 years without UBT (DR = 93.3%). The protocols containing age group <40 years might not obviously improve DRs (Table 6).

Figure 2. Cost-effectiveness analysis on protocol candidates of Hp-GCS.

Table 5. Subpopulation strategy for endoscopic screening.

Sex	Age	Hp infection	Controls	Cancer	RR (95% CI)	NNS
Male	<40 yrs	UBT (–)	1,404	1	Ref	15,762
		UBT (1+∼2+)	862	1	1.63 (0.10–26.00)	9,682
		UBT (3+∼4+)	81	0*	1.73 (0.003–1,148.55)	9,087
	40–59 yrs	UBT (–)	2,047	15	9.81 (1.35–77.29)	1,607
		UBT (1+∼2+)	1,369	26	25.13 (3.56–192.71)	627
		UBT (3+∼4+)	133	3	29.74 (3.25–295.93)	530
	≥60 yrs	UBT (–)	802	26	43.78 (6.00–324.53)	360
		UBT (1+∼2+)	496	23	61.79 (8.43–459.90)	255
		UBT (3+∼4+)	43	5	145.23 (17.43–1,228.73)	109
Female	<40 yrs	UBT (–)	1,481	5	4.35 (0.55–40.42)	3,620
		UBT (1+∼2+)	684	3	5.65 (0.64–58.88)	2,789
		UBT (3+∼4+)	265	1	4.87 (0.33–84.19)	3,240
	40–59 yrs	UBT (–)	2,536	10	4.78 (0.71–43.07)	3,296
		UBT (1+∼2+)	1,466	8	6.61 (0.95–60.89)	2,385
		UBT (3+∼4+)	418	4	11.54 (1.49–118.83)	1,366
	≥60yrs	UBT (–)	910	15	20.63 (3.01–172.19)	764
		UBT (1+∼2+)	478	11	28.62 (4.09–244.18)	551
		UBT (3+∼4+)	124	6	58.73 (7.87–534.55)	268

Abbreviations: CI: confidence interval; Hp: Helicobacter pylori; NNS: number needed to screen; Ref: reference; RR: relative risk; UBT: urea breath test.

* A value 0.5 was assigned for null event.

Table 6. Cost-effectiveness analysis of protocol candidates for Hp-GCS.

Protocol	Detail	Cost ($)	Cases	DR	CER	Δcost ($)	ΔEffect	ICER
P1	Any age without UBT	1,300,990	163	100%	7,981.54	604,836	36	16,801.00
P2	Any age, but UBT (1 ∼ 4+)	782,520	91	55.8%	8,599.12
P3	Age ≥40 yrs and UBT (1 ∼ 4+)	551,462	86	52.8%	6,412.35
P4	Age ≥40 yrs and UBT (3 ∼ 4+)	231,456	18	11.0%	12,858.66
P5	Age 40-59 yrs and UBT (1 ∼ 4+), or age ≥60 yrs without UBT	696,154	127	77.9%	5,481.53	Ref	Ref
P6	Any age and UBT (3 ∼ 4+)	334,577	19	11.7%	17,609.32
P7	Age ≥40 yrs without UBT	905,790	152	93.3%	5,959.15	209,636	25	8,385.50
P8	Age <40 yrs and UBT (1 ∼ 4+), or age ≥40 yrs without UBT	1,136,848	157	96.3%	7,241.07	440,694	30	14,689.80
P9	Age <40 yrs and UBT (3 ∼ 4+), or age ≥40 yrs without UBT	1,008,912	153	93.9%	6,594.19	312,757	26	12,029.10

Abbreviations: CER: cost-effectiveness ratio; DR: detection rate; Hp-GCS: Helicobacter pylori related gastric cancer screening; ICER: incremental cost-effectiveness ratio; Ref: reference; UBT: urea breath test; yrs: years.

Discussion

In this hospital-based cross-sectional study, the higher prevalence of Hp infection was observed in gastric cancer patients, especially in age group ≥40 years or males. UBT-based model containing demographic covariates showed just moderate diagnostic strength for predicting gastric cancer. Strong and very strong positivity of UBT might additionally decrease the NNSs in any age groups of either males or females. The cost-effectiveness analysis demonstrated that endoscopic screening might not bring obvious benefit of detection rate in age group <40 years, but the selection of positive UBT individuals for Hp-GCS obtained optimal cost-effectiveness with a moderate detection rate 77.9%.

Hp infection is a common cause of gastric cancer (36). The proportion of Hp infection in Chinese residents was estimated to be 55.8% (25). And a 14-year follow-up study in rural China found that the incidence of gastric cancer in patients with Hp infection was about 23.1‰, which was significantly higher than that of uninfected patients (5.93‰) (37). The present study confirmed the association between gastric cancer and higher Hp prevalence. However, Hp-GCS has not been widely carried out around the world, with the exception of Japan and South Korea (20). A western study in the United States showed that the screening and treatment of Hp for high-risk groups was also potentially cost-effective in preventing gastric cancer (38). The Taiwan region also launched a population-based screening and eradication project of Hp in 2004 (39). A prospective study in Japan found that Hp infection could predict gastric cancer well, and as the aggravation of infection, the incidence of gastric cancer increased (40).

Regarding the Hp-GCS, either serologic antibody test (41, 42) or UBT (43–45) was a convenient and noninvasive approach. Besides UBT, serological screening of Hp might be helpful to identify high-risk individuals of gastric cancer development (46). Although a Japanese study found that large-scale endoscopic screening was the best mean to detect and identify gastric cancer patients (47), actually, endoscopic screening was more costly and less receptive by the testee of health checkup. Similarly, those reasons were why it was difficult to organize large-scale endoscopic screening in China. Additionally, in recent years, the Hp-GCS has been considered in the scheme of health checkup in many Chinese institutes, but Hp-GCS is still not mandatory in the governmental program, due to the insufficiency of robust evidence. In addition, it was reported in the United States that serological examination of Hp was low in sensitivity and cannot detect precancerous lesions (48). The 14-C UBT was used to diagnose Hp infection in this study, but some studies suggested that the cost of enzyme-linked immunosorbent assay was lower (49). Along with our findings, either UBT or serology might not be suitable as an independent screening test.

Hp-GCS may potentially consider several serologic factors of virulent Hp strains, while Hp-GCS can combine certain tumor biomarkers or precancerous lesion-associated biomarkers (50). Cytotoxin-associated gene A (CagA), vacuolating cytotoxin A (VacA), helicobacter cysteine-rich protein C (HcpC), and chaperonin (GroEL), etc, which might be informative factors of virulent Hp strains in predicting gastric cancer (28, 51). Pepsinogens and gastrin-17 were associated with precancerous lesions of gastric cancer (16). Studies found the serologic pepsinogen test could be used as a supplementary screening test for high-risk individuals, but might not be ideal for gastric cancer itself (52, 53). Gastrin-17 was a serologic marker, which could reflect the function of gastric mucosa and predict gastric cancer by increasing level along with decreasing pepsinogen levels (54, 55). There were many reports on gastric cancer screening related to tumor biomarkers (56, 57). Commonly, higher levels of serologic CEA, CA724, CA125, and CA199, etc, were associated with gastric cancer risk, which could be considered as supplementary tests in the gastric cancer screening (34, 58). Besides Hp, some other pathogens might be associated with gastric cancer risk, such as Epstein-Barr virus (EBV), hepatitis B virus (HBV), human cytomegalovirus (CMV), human papillomavirus (HPV), and John Cunningham virus (JCV) (59–61). All these factors can be investigated on the effectiveness and cost-effectiveness in gastric cancer screening.

There were certain limitations to be considered with caution. First, the study design was a single-hospital-based cross-sectional study, so the representativeness might be impaired to some extent in the simulative manner and sampling errors could not be waived out. Second, the detected cases were not classified by the TNM stage, and the role of Hp-GCS in early detection was kept unclear. Third, this study has an important limitation, that is, extrapolating a cross-sectional study to make conclusions on population-based cost-effectiveness, a major methodological limitation which has to be acknowledged. Cost-effectiveness analysis on screening cohort or whole-life health checkup might be more practicable for organized massive screening. Finally, the cost only contained direct medical cost during single round screening, while the cost of Hp eradication and endoscopic surveillance would better be considered.

In conclusion, UBT had moderate diagnostic strength in massive gastric cancer screening, and might be cost-effective in middle-aged population (40–59 years). More robust Hp-GCS protocol needs further investigate in test methods and individual biologic features.

Acknowledgments

The authors thank the substantial work of Volunteer Team of Gastric Cancer Surgery (VOLTGA), West China Hospital, Sichuan University, China.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by the Wu Jieping Medical Foundation (320.2710.1815), and the 1•3•5 Project for Disciplines of Excellence, West China Hospital, Sichuan University (ZY2017304). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.