Cost-effectiveness of screening with polymerase chain reaction for Helicobacter pylori to prevent gastric cancer and peptic ulcers

Background Helicobacter pylori (H. pylori) is a major risk factor for gastric cancer. Screening and treatment of H. pylori may reduce the risk of gastric cancer and peptic ulcer disease (PUD). Polymerase chain reaction (PCR) of gastric biopsies provides superior sensitivity and specificity for the detection of H. pylori. This study explores whether population-based H. pylori screening with PCR is cost-effective in the US. Methods A Markov cohort state-transition model was developed to compare three strategies: no screening with opportunistic eradication, 13C-UBT population screening and treating of H. pylori, and PCR population screening and treating of H. pylori. Estimates of risks and costs were obtained from published literature. Since the efficacy of H. pylori therapy in gastric cancer prevention is not certain, we broadly varied the benefit 30–100% in sensitivity analysis. Results PCR screening was cost-effective and had an incremental-cost effectiveness ratio per quality adjusted life-year (QALY) of $38,591.89 when compared to 13C-UBT strategy with an ICER of $2,373.43 per QALY. When compared to no screening, PCR population screening reduced cumulative gastric cancer incidence from 0.84% to 0.74% and reduced PUD risk from 14.8% to 6.0%. The cost-effectiveness of PCR screening was robust to most parameters in the model. Conclusions Our modeling study finds PCR screening and treating of H. pylori to be cost-effective in the prevention of gastric cancer and PUD. However, the potential negative consequences of H. pylori eradication such as antibiotic resistance could change the balance of benefits of population screening.


Background
Helicobacter pylori is a known risk factor for gastric cancer, which is the third leading cause of cancer death worldwide [1]. Though more than half of the global population is infected with H. pylori, rates of colonization vary greatly, with higher rates in lower income nations generally related to socioeconomic status and hygiene levels [2]. While most people infected with H. pylori are asymptomatic, approximately 10% develop peptic ulcer disease (PUD) and 1 to 3% develop gastric adenocarcinoma [3]. H. pylori promotes gastric carcinogenesis through chronic gastric in ammation which progresses through the stages of atrophic gastritis, intestinal metaplasia, dysplasia, to gastric adenocarcinoma [4]. Eradication of H. pylori infection has been shown to reduce gastric cancer incidence [5]. With more than 60% of gastric cancers attributable to H. pylori, early detection and eradication of infection is important in reducing the risk of cancer [6].
H. pylori infection ful lls many of the criteria for population screening [7]. It can be detected through both noninvasive and invasive methods, each with its own advantages and limitations [8]. There are a variety of diagnostic tests available yet no single gold standard has been established in clinical practice. However, diagnostic tests with sensitivity and speci city exceeding 90% are necessary for accurate diagnosis of H. pylori infection. American College of Gastroenterology guidelines recommend noninvasive 13 C-UBT in populations with low probability of H. pylori infection due to its inexpensive costs and quick results [9]. Compared to this, polymerase chain reaction (PCR) of gastric biopsies has superior sensitivity and speci city for the detection of H. pylori [10]. However, use of PCR testing of gastric biopsies is limited by accessibility, its inherent invasive nature, and expertise level of laboratories. Though screening and treating the population for H. pylori infection may reduce gastric cancer morbidity and mortality, such screening programs can be costly and di cult to implement. In this study, we aimed to evaluate the cost-effectiveness of 13 C-UBT and PCR population screening strategies of H. Pylori for the prevention of peptic ulcer disease and gastric cancer in the United States.

Model Design
A Markov state-transition cohort model was constructed in TreeAge Pro (TreeAge 2020, Williamstown, Massachusetts). The model compared three strategies: (1) no screening with opportunistic eradication, (2) single population screening for H. pylori using 13 C-UBT and treating those who tested positive with eradication therapy, and (3) single population screening for H. pylori with upper endoscopy and PCR of gastric biopsies and treating those who tested positive with eradication therapy. The hypothetical patient for this analysis is a 40-year-old individual from the U.S. general population. A 40-year-old individual was chosen as the target population of our model due to the sharp increase in gastric cancer incidence after 40 years of age [11]. Additionally, in countries such as Japan and Korea where screening programs exist, the recommended start age is also 40 years old [12]. The model follows patients for 60 years or until death and has a cycle length of one year.

Management Strategies
The management strategies in our analysis consisted of no screening, population based 13 C-UBT with eradication therapy, and population-based H. pylori PCR screening with eradication therapy. The health states in our model included H. pylori positive, H. pylori negative, gastric cancer, and death ( Fig. 1). In addition, both H. pylori positive and negative individuals had ongoing risks of peptic ulcer disease throughout the model. Patients could move from any health state to death due to all-cause mortality, bleeding from peptic ulcers, or gastric cancer.
In the no screening strategy, the distribution of the cohort at the beginning of the simulation (i.e. cycle 0) was based on U.S. H. pylori prevalence rates. Though none of the patients initially received screening, those who developed peptic ulcer disease were tested with UBT and given eradication therapy.
For the screening strategies, the entire cohort began with screening and eradication therapy was given to those who tested positive (i.e. true-positive and false-positive). Both rst-line triple therapy (proton pump inhibitors (PPI), clarithromycin, and amoxicillin) and second-line quadruple therapy (PPI, bismuth, tetracyline, and metronidazole) were modeled. Individuals who failed both lines of eradication therapy were considered permanently H. pylori positive in the model. Risk of reinfection was restricted to the rst ve years after successful eradication.

Outcomes
The primary outcomes of interest were total cost, life expectancy, quality-adjusted life years (QALYs), and incremental cost-effectiveness ratios (ICERS). QALYs are a composite measure of the value of health outcomes that can be used to compare medical interventions where a value of 0 represents death and a value of 1.0 represents a year of perfect health. To calculate a QALY for a patient, we multiplied the utility value associated with a given health state multiplied by the cycle length of 1 year. Life expectancy was calculated as the total number of years the hypothetical patient was alive throughout the model duration without applying the health state utility values that re ected disease morbidity. A willingness to pay (WTP) threshold of $100,000/QALY was used to determine cost-effectiveness. Secondary clinical outcomes of interest included total lifetime gastric cancer incidence. In the model, total lifetime gastric cancer incidence was determined by aggregating the proportion of patients who were in the gastric cancer health state with those who died of gastric cancer by the end of the simulation duration.

Parameter Estimates and Model Assumptions
Model parameters were based on estimates from the literature. Base-case values and ranges used in sensitivity analyses are summarized in Table 1. H. pylori prevalence in the U.S. was obtained from a meta-analysis by Hooi et al [2]. Risks of PUD and gastric cancer varied based on H. pylori status. The incidence of PUD in H. pylori positive individuals was 6 to 10-fold higher than for uninfected individuals [13,14]. For patients who were H. pylori negative, they had a 0.66 relative risk of gastric cancer compared to those who were H. pylori positive [15]. Due to scarce data regarding gastric cancer risk for individuals who were never infected compared to those were successfully eradicated of H. pylori, the 0.66 relative risk was applied for both groups. Though the risk of gastric cancer in individuals successfully treated of H. pylori could still be higher than the risk for those never infected due to developed gastric atrophy, we assumed the two risks were the same in the principal analysis.

Utilities
Quality of life utility values relating to healthy, H. pylori positive, and gastric cancer states were incorporated in our model. Utility decrements due to upper endoscopy and PUD were also applied. Quality adjusted life years were discounted at an annual rate of 3%.

Sensitivity Analyses
We performed one-way deterministic sensitivity analyses by altering individual variables across a range of values to investigate the key parameters that most impacted the outcomes of the model. Due to uncertainty in the true clinical e cacy of H. pylori therapy for reducing excess gastric cancer risk, we varied e cacy rates with sensitivity analysis from 30-100%. In addition, a probabilistic sensitivity analysis (PSA) was performed to address parameter uncertainty. β distributions were tted for transition probabilities and utilities, while γ distributions were tted for cost parameters. The PSA was performed using Monte Carlo simulations with 100,000 reiterations.

Results
Our base case analysis demonstrated that both 13 C-UBT and PCR were cost-effective treatment strategies (  The cumulative gastric cancer incidence in the no-screening strategy was 0.84% and was calibrated to re ect published SEER life gastric cancer risks for an average 40-year old in the US. [11] The lifetime gastric cancer incidence for the PCR strategy and 13 C-UBT strategy was 0.74% and 0.75% respectively. In the no-screening strategy, there was a 14.8% lifetime risk of PUD. In the PCR and UBT strategies, the lifetime risk of PUD dropped to 6.0% and 6.4%, respectively. When each H. pylori eradication strategy was compared to the no-screening strategy, the results of the one-way sensitivity analysis showed that the 13 C-UBT and PCR strategies were most sensitive to the risk of gastric cancer if H. pylori positive, the risk of gastric cancer after H. pylori eradication, H. pylori prevalence, and costs of screening (Fig. 2). However, even within the prescribed ranges of the one-way sensitivity analysis, the 13 C-UBT and PCR strategies remained more cost effective than the no-screening strategy. When varying the e cacy of H. pylori eradication from 30%-100%, PCR remained the most costeffective strategy at a WTP of $100,000 per QALY (Fig. 3).
Cost-effectiveness acceptability curves were used to present the results of the PSA and determine the probability of any strategy being the cost effective at a given WTP (Fig. 4). When the WTP was <$5000 per QALY, the no screening strategy was the most cost-effective strategy majority of the time. As the WTP exceeded $60,000 per QALY, PCR became the cost-effective the majority of times.

Discussion
In this study, we compared the lifetime cost-effectiveness of different population-based H. pylori screening strategies for the prevention of gastric cancer for average 40-year-old Americans. To our knowledge, our study is the rst to determine the cost-effectiveness of PCR as a potential screening strategy. Based on our results, the PCR strategy was the most cost-effective at the prescribed WTP threshold of $100,000 per QALY as it resulted in the lowest lifetime risks of gastric cancer at 0.74% and peptic ulcer disease at 6.0%.
Our study ndings are similar to the conclusions of prior cost-effectiveness analyses that suggest population-based screening and eradication treatment can be a viable option for the reduction of gastric cancer [17,18]. While previous cost-effectiveness studies focused mostly on gastric cancer reduction, we also considered the additional bene ts of H. pylori screening and eradication on peptic ulcer disease. Relative to gastric cancer, the incidence of peptic ulcer disease is much higher and can confer signi cant physical and nancial burden in patients due to hospitalization costs, bothersome symptoms, and chronic PPI use. Our model determined that screening and eradicating H. pylori decreased the lifetime incidence of peptic ulcer disease from 14.8-6%. The inclusion of peptic ulcer disease allowed us to incorporate opportunistic eradication within the no screening strategy and provide a more clinically accurate representation within our model to mitigate any bias towards the intervention strategies.
Due to its superior sensitivity and speci city, PCR screening and treatment of H. pylori can achieve signi cant bene ts towards long term gastric cancer reduction. However, implementation of population PCR screening is challenged by unequal availability of necessary diagnostic facilities. Furthermore, PCR of gastric biopsies requires endoscopy, an invasive procedure that comes with risk of complications such as bleeding. As determined by our model, compared to no screening, 13 C-UBT is a cost-effective and noninvasive alternative that also reduces gastric cancer and peptic ulcer disease. While the results of our model showed only marginal bene ts towards gastric cancer reduction with PCR compared 13 C-UBT, there can be cases where the most accurate measures are required. High risk patients for gastric cancer, such as in the context of hereditary diffuse gastric cancer or Lynch syndrome, often have earlier onset and worst prognoses [19]. In these higher risk populations, PCR screening and treating of H. pylori could be recommended due to its superior sensitivity and specify to reduce as much excess risk of gastric cancer as possible.
Our study does have several limitations. Data were combined from multiple sources and some values for variables were imprecise. The relative risk of gastric cancer for H. pylori positive patients in our model was more conservative than what other studies had published [20,21]. However, this conservative measure would bias the results of our model against screening and treatment. Adherence to eradication therapy is critical to the effectiveness of screening programs, and within our model we assumed 100% adherence. Compliance to a treatment program is a multifactorial process which can include factors such as complexity of treatment, trust in doctor-patient relations, and side effects of treatment. Incorporating such factors into the model was not possible and therefore a limitation in our study. While 100% adherence is not feasible in a real-world context, we assumed a perfect scenario for model simplicity and function. The exact e cacy of H. pylori therapy toward gastric cancer reduction is unknown. We assumed 100% e cacy in our principal analysis but varied the e cacy from 100%-30% within the sensitivity analysis to account for this limitation. While development of resistance to existing eradication therapies is an important consideration, and signi cant concern to public health, it is di cult to quantify the speci c clinical harm it would confer to an individual patient and therefore was not included in our model. Risk of antibiotic resistance is a signi cant concern to public health and a considerable counterargument to a population wide test and treat strategy. Though implementation of population screening can be challenged by risks of antibiotic resistance, PCR screening can confer an additional advantage by determining which strains of H. pylori could be carrying genes for antibiotic resistance. This information could potentially decrease costs by informing which antibiotic regimens are most likely to succeed, reducing the need for re-tests. PCR could also provide similar bene ts when testing strains for cytotoxin-associated gene A (CagA) status. As CagA positive strains are associated with higher risk of gastric cancer development, patients that are determined CagA negative by PCR could potentially withhold or delay eradication treatment for surveillance. While our study focused on the attributable risk of gastric cancer due to H. pylori infection, we do not account for inherited predispositions to gastric cancers. Though inherited gastric cancers make up only small fraction of all gastric cancers, the subset of the population with these heightened risk factors can bene t from a PCR screening strategy as cancer onset often occurs at younger ages for them. Furthermore, we did not consider the potential adverse effects of curing H. pylori, such as the development of re ux esophagitis, esophageal cancer, and autoimmune diseases [22,23]. Rubenstein [23,24].
Model uncertainty is an inevitable challenge faced in cost-effectiveness analyses. While the true e cacy of H. pylori eradication toward gastric cancer reduction is controversial, our model remained robust despite varying e cacy rates. Although our study considers the bene ts of screening speci cally within the United States, other countries with higher prevalence of H. pylori and gastric cancer, screening and treating could be even more cost-effective, or possibly cost-saving. In the US context, there is a wide ethnic and racial disparity in H. pylori and gastric cancer prevalence and incidence. While our model does not take into account speci c racial and ethnic differences in gastric cancer risk, cost-effectiveness studies targeted for immigrants and people from higher risk regions and people groups could show increased effectiveness of screening programs.

Conclusions
In summary, our modeling analysis nds that H. pylori testing with upper endoscopy and PCR testing of gastric biopsies is a cost-effective option to prevent gastric cancer and peptic ulcer disease in the US population. Additional clinical studies to a rm some of our model assumptions, especially those regarding the differences in risk of gastric cancer between individuals cured of H. pylori and those were never infected with H. pylori, are needed.

Abbreviations
CagA: cytotoxin-associated gene A; 13  Availability of data and materials All data generated or analyzed during this study are included in this article

Competing interests
The authors declare that they have no competing interests