Cost-Effectiveness Analysis of Helicobacter pylori Diagnostic Methods in Patients with Atrophic Gastritis

Background. There are several diagnostic methods for Helicobacter pylori (H. pylori) infection. A cost-effective analysis is needed to decide on the optimal diagnostic method. The aim of this study was to determine a cost-effective diagnostic method in patients with atrophic gastritis (AG). Methods. A decision-analysis model including seven diagnostic methods was constructed for patients with AG diagnosed by esophagogastroduodenoscopy. Expected values of cost and effectiveness were calculated for each test. Results. If the prevalence of H. pylori in the patients with AG is 85% and CAM-resistant H. pylori is 30%, histology, stool H. pylori antigen (SHPAg), bacterial culture (BC), and urine H. pylori antibody (UHPAb) were dominated by serum H. pylori IgG antibody (SHPAb), rapid urease test (RUT), and urea breath test (UBT). Among three undominated methods, the incremental cost-effective ratios (ICER) of RUT versus SHPAb and UBT versus RUT were $214 and $1914, respectively. If the prevalence of CAM-sensitive H. pylori was less than 55%, BC was not dominated, but its H. pylori eradication success rate was 0.86. Conclusions. RUT was the most cost-effective at the current prevalence of CAM-resistant H. pylori. BC could not be selected due to its poor effectiveness even if CAM-resistant H. pylori was more than 45%.


Introduction
While the prevalence of Helicobacter pylori (H. pylori) has been decreasing [1,2], it remains a critical public health issue. Recently, increasing prevalence of CAM-resistant H. pylori is an emerging problem of public health all over the world as CAM is included in most first-line empiric H. pylori eradication regimens [3,4].
Among these H. pylori-related diseases, AG is more common than PUD or early gastric cancer; in Japan, its prevalence is reported to be approximately 27.9% even in healthy individuals [10]. 85% of AG patients were reported to have H. pylori infection [11]. It is a common situation that, during either diagnostic or screening esophagogastroduodenoscopy (EGD), physicians must choose between one of several H. pylori diagnostic methods.
The diagnostic performance of these tests differs. Using BC for diagnosing H. pylori infection allows us to perform antibiotic-(typically macrolide-) sensitivity testing. The results of the sensitivity testing are useful to make appropriate decisions when choosing the correct first regimen for treatment, a strategy called antimicrobial susceptibilityguided therapy (AMSGT). AMSGT is assumed to be more cost-effective when the prevalence of CAM-resistant H. pylori has been increasing. However, there have been no prior reports mainly focusing on the impact of the prevalence of CAM-resistant H. pylori infection. The aim of this study was to determine a cost-effective diagnostic method for H. pylori infection in patients with AG.

Methods
This study was conducted from a social perspective. A decision-analysis model was constructed for patients in Japan diagnosed with AG suggesting H. pylori infection, using screening or diagnostic EGD. Time horizon was until successful H. pylori eradication or the end of the third regimen. We assumed that this time horizon would fall within 1 year and did not discount either effectiveness or cost.
Undergoing one of seven diagnostic tests (RUT, histology, BC, SHPAb, UBT, SHPAg, and UHPAb), patients' H. pylori infection status was unknown. Excluding BC which can be applied for AMSGT, if one of six tests (RUT, histology, SHPAb, UBT, SHPAg, and UHPAb) was selected and was positive, the patient underwent empiric antibiotic treatment, as none of these six tests provided any information on CAM-sensitivity. If the first standard regimen failed, patients followed the second and third regimens without additional CAM-sensitivity testing. If BC was initially selected as the diagnostic test and was positive, subsequent antibiotic susceptibility testing results were used to decide on the treatment regimen. If detected H. pylori was sensitive to CAM, these patients were treated with CAM-included regimen. If not, these patients were treated with metronidazoleincluded regimen. In the decision tree, all eradication failure was measured by UBT after the previous diagnostic step. We did not include the strategy of initial six diagnostic tests followed by AMSGT as BC required repeat EGD and we considered it unaffordable to perform repeat EGD only for the purpose of BC.
Diagnostic performance, including sensitivity and specificity of invasive and noninvasive diagnostic tests, was obtained from past English literatures, searched manually through MEDLINE and EMBASE. If there was a literature of meta-analysis, we adopted pooled values of sensitivity and specificity. Otherwise, we conducted a meta-analysis (bivariate random effects model) to calculate pooled values of sensitivity and specificity. Effectiveness was measured by rate of successful H. pylori eradication.
We used the success rate of H. pylori eradication by the first regimen (lansoprazole 30 mg bid, amoxicillin 750 mg or 1000 mg bid, and CAM 200 mg or 500 mg bid for one week) including CAM in the patients with CAM-sensitive or CAM-resistant H. pylori [12][13][14][15]. We also used success rate of metronidazole included triple therapy (omeprazole 20 mg bid or lansoprazole 30 mg bid, amoxicillin 500 or 750 mg bid, metronidazole 500 mg in the morning and 250 mg in the evening or 250 mg tid for one week) in the patients with CAM-sensitive or CAM-resistant H. pylori [16]. In case of AMSGT, the H. pylori eradication rate of the 2nd regimen for the patients with CAM-resistant H. pylori was used [15]. The success rate of the third regimen (lansoprazole 30 mg bid, amoxicillin 750 mg bid, and sitafloxacin 100 mg bid for one week) for the patients who failed metronidazole-based triple therapy was also used [17].
Costs of each diagnostic procedure and H. pylori eradication regimens were derived from reimbursement of the Japanese governmental health insurance [18], and costs of gastric cancer treatments were derived from diagnosis procedure combination (DPC) by the Japanese government [19].
Our main outcome was a success rate of H. pylori eradication. Cost-effective thresholds, in other words willingness to pay (WTP), were estimated by treatment costs of preventable gastric cancer divided by the number needed to eradicate H. pylori infection.
Ford et al. [20] reported a pooled relative risk of 0.66 (95% confidence interval 0.46 to 0.95), and a number needed to eradicate H. pylori to prevent one patient of gastric cancer was as low as 15 for Chinese men, compared to 245 for US women.
Early gastric cancer is treated by endoscopic mucosal resection (EMR) or endoscopic submucosal dissection (ESD), and advanced gastric cancer is treated by laparoscopic or open gastrectomy. These DPC costs ranged from $2500 to $16000 [19].
Expected values of cost and effectiveness were calculated for BC potentially for AMSGT and other six diagnostic strategies (RUT, histology, SHPAb, UBT, SHPAg, and UHPAb). Costs of each diagnostic method and each H. pylori eradication regimen were estimated from National Health Insurance data in Japan and expressed in US dollars at the exchange rate of 100 yen/US dollar (Table 1) [18,19,21]. We did not include cost of EGD as all patients in our model underwent EGD.
We first performed cost-effective analysis of base-case and calculated incremental cost-effective ratio (ICER) for comparing pairs of undominated diagnostic methods. Then, we conducted a one-way sensitivity analysis, focusing on the prevalence of CAM-resistant H. pylori, prevalence of H. pylori in the patients with AG, and the success rate of the 1st regimen for H. pylori to determine its threshold (Table 1).
We also performed a Monte Carlo simulation using range of uncertain probability in two scenarios of 0.4 or 0.45 of CAM-resistant H. pylori prevalence. All variables were assumed to follow a triangular distribution (Table 1). Ten thousand trials were conducted for simulation. We reported acceptability curve by a simulation of 10000 trials.

Results and Discussion
A decision-analysis model starting at the point of diagnosing H. pylori infection during or just after EGD was constructed ( Figure 1). Costs and probabilities used in the decision model are presented in Table 1.
With the current prevalence of CAM-resistant H. pylori of 30% [37], the most effective test for H. pylori diagnosis was UBT or histology, while the least effective test was SHPAb. Additionally, the most expensive test was histology, while the least expensive test was SHPAb. Histology, SHPAg, and BC were absolutely and UHPAb was weakly dominated by SHPAb, RUT, and UBT. Among the three undominated methods, the ICER of RUT versus SHPAb and UBT versus RUT was $214 and $1914, respectively. The H. pylori eradication success rate of SHPAb, RUT, and UBT was 0.87, 0.94, and 0.96, respectively ( Figure 2).
One-way sensitivity analysis with change of prevalence of CAM-resistant H. pylori was showed in Figures 3 and 4 Figure 4).
One-way sensitivity analyses using other two variables the prevalence of H. pylori in the patients with AG and H. pylori eradication success rate by the 1st regimen suggested that our results were insensitive for these two variables ( Table 2).
In acceptability curves using Monte Carlo simulation with current (0.3) and increased (0.45) prevalence of CAMresistant H. pylori, the optimal strategy was either SHPAb, RUT, or UBT (Figures 5(a) and 5(b)). This is the first cost-effective analysis of H. pylori diagnostic methods mainly taking into account increasing prevalence of CAM-resistant H. pylori. First, our study showed that SHPAb, RUT, and UBT were undominated and RUT was the most cost-effective at the current prevalence of CAM-resistant H. pylori considering both their effectiveness and WTP. Second, although BC for AMSGT can be a suitable option if the proportion of CAM-resistant H. pylori increases to more than 45%, RUT was the most cost-effective as the effectiveness of BC was remarkably poorer than RUT and UBT. Third, although SHPAg was dominated in the base-case analysis, Monte-Carlo analyses showed that SHPAg was cost-effective in about 20% of trials if WTP was more than $1000. This would be caused by uncertainty of diagnostic performance of UBT and SHPAg.
Elwyn et al. [38] performed cost-effective analysis including three methods (SHPAb, SHPAg, and UBT) and concluded that UBT was dominated by SHPAg and the ICER of SHPAg versus SHPAb was €10. This study disregarded the three invasive tests, as well as UHPAb, all of which were included in our decision model. The cost of SHPAg ($33) is relatively more expensive than other diagnostic tests in Japan, and the above study used higher sensitivity and specificity data for SHPAg than those used in our model. The outcome of this study by Elwyn et al. was not the H. pylori eradication rate but the number of true outcomes. Additionally, authors did not discuss about WTP. These might be some of the reasons why SHPAg was not found to be cost-effective in our study. They followed "test and treat" policy without considering a referral to EGD, common in general practitioners' practice outside of Japan. In our decision model, we assumed that EGD would be performed prior to H. pylori testing in patients with or without dyspepsia.
According to the most recent guidelines for gastric cancer screening in Japan, EGD can be used not only as an opportunistic screening but also as a population-based screening tool [39]. It is anticipated that the number of asymptomatic individuals with the diagnosis of AG will increase and a costeffective diagnostic tool for H. pylori infection is therefore needed. As such, the results of our study can be applied to choosing a diagnostic method for H. pylori infection mainly in the context of a screening population undergoing EGD.
Our analysis has some limitations and strengths. First, we did not take into account possible adverse events from taking antibiotics or taking biopsy specimen for invasive tests, which we anticipate are very rare and not severe. Second, as treatment completion was assumed to be within 1 year, we did not consider the time needed until H. pylori eradication. Third, we did not take into account costs of several H. pylori-associated diseases except gastric cancer in estimating WTP. Fourth, our results can apply only for medical practice in Japan as our model assumes AG prevalence and standard H. pylori eradication regimen in Japan, both of which are different from western countries.
However, this is the first study to investigate the impact of increasing prevalence of CAM-resistant H. pylori infection from a cost-effectiveness perspective. In addition, we used the results of meta-analyses for all diagnostic methods' performance, which should be valid.
In conclusion, RUT was the most cost-effective diagnostic procedure given the present prevalence of CAM-resistant H. pylori. Although BC can be a costeffective diagnostic method if the proportion of CAMresistant H. pylori continues to increase to ≥45%, BC potentially for AMSGT will not be cost-effective due to its poor effectiveness.  Helicobacter pylori (H. pylori). Acceptability curve showed that serum H. pylori antibody or rapid urease test or urea breath test was an optimal diagnostic method depending on willingness to pay (WTP). Even if the prevalence of CAM-resistant H. pylori increased to 0.45, the probability that bacterial culture becomes an optimal method was low regardless of WTP.