When a population harbors a high disease burden, asymptomatic members of the population are likely to be aware of the disease and thus may participate in screening. In populations with a high disease burden, the positive predictive value (PPV) of the test will also be high. On a global scale, the IARC has predicted that the annual number of new cases of gastric cancer is expected to increase or remain at a constant level by 2030 (Table 1).⁷ This prediction is consistent with the experience in Taiwan, where gastric cancer is the seventh most common cancer and the sixth most deadly cancer. Every year in Taiwan, there are around 3,800 new cases with an incidence of approximate 17 per 100,000 person-years; approximately 2,300 persons die from gastric cancer each year.¹³ Although the age-standardized incidence of gastric cancer is declining, the absolute numbers of incident cases are stationary or even slightly increased; by contrast, the incidence of colorectal cancer is rapidly increasing. Following the nationwide screening program launched in 2004, its case-fatality rate (ratio of mortality/incidence) is 0.35, which is much lower than that of gastric cancer (0.59), reflecting the urgent need to prevent gastric cancer in Taiwan.

Understanding the natural history of a particular type of cancer is crucial in the design of an effective intervention. As shown in Fig. 1A, the natural course of a cancer can be separated into three phases: (1) the carcinogenic phase (i.e., increased cancer risk due to carcinogen exposure); (2) the PCDP (i.e., early-stage, presymptomatic cancer); and (3) the clinical phase (i.e., advanced-stage, symptomatic cancer). Cancers at the PCDP are the targets of screening, although not every type of cancer is amenable to detection by screening. The longer the PCDP (i.e., the time between cancer onset and clinical symptoms), the longer the lead time (i.e., the time between cancer detection and clinical symptoms); when a diagnosis is made by screening, prolonged survival (or cure) may be possible. The PCDPs of various cancers have been estimated in the work of Bray et al.¹⁴ using the prevalence/incidence ratios. In cancers with very short PCDPs, a screening test may diagnose the cancer earlier but with no effect on the outcome. The apparent effect is the so-called lead time bias. By contrast, in cancers with very long PCDPs (e.g., prostate cancer screened by prostate specific antigen¹⁵), the screening test may detect indolent cancers which may never cause clinical symptoms in the remaining lifetime of the patient, thus raising the issue of overdiagnosis. Obviously, the length of PCDP is not fixed, which is related to the detectability of a screening tool. If a screening tool can identify cancerous foci earlier, the PCDP can be prolonged;¹⁶ the example is the use of low-dose computed tomography, instead of chest X-ray, to screen lung cancer.¹⁷

The mean PCDP of gastric cancer is relatively short, about 1.4 years,¹⁴ which may explain why a 1- to 2-year screening interval is generally recommended for endoscopic surveillance in high risk groups. In fact, upper endoscopy is not an ideal mass-screening tool for gastric cancer as it is expensive, labor-intensive, and not without risk. In addition, the results are subject to interoperator variation so interval cancers¹⁸ (i.e., symptomatic cancer diagnosed during the interscreening interval) are not infrequent. By contrast, colorectal cancer, which has a PCDP of approximate 3 years, is suitable for biennial screening with the noninvasive fecal immunochemical testing, which also has the advantage of a high yield with positive tests.¹⁹

A number of cancers (Fig. 1B) have well-established risk factors such that elimination of, or protection from, these factors can prevent cancer development. Population-based vaccination against viruses that causes cervical cancer²⁰ or liver cancer²¹ has been shown to be effective in reducing the subsequent cancer risk. In contrast, H. pylori infection has not received this kind of attention, despite its well-recognized causative role in gastric cancer. If H. pylori can be removed from the population, it has been estimated that approximately 75% of gastric cancer would be eliminated.²² This is likely an underestimate as early estimates of causation were based on serology, which underestimated the proportion of gastric cancers due to H. pylori; in addition, such a gastric cancer prevention strategy is supported by studies in a real-world setting (Fig. 2),^10,23–30 with favorable results for either premalignant lesions^{10,24–26,28,29} or actual gastric cancer.^10,23,27,30 Furthermore, by pooling six RCTs, a recent meta-analysis consistently showed that the eradication of H. pylori can provide a significant risk reduction of 34% for gastric cancer in asymptomatic H. pylori infected individuals.³¹ For patients diagnosed with early gastric cancer who underwent endoscopic resection, another meta-analysis based on 13 studies, including two RCTs and 11 cohort studies, also showed a significant risk reduction of 58%.³² Because premalignant gastric lesions are expected to be prevalent in the latter group, these observations suggest that it is never too late to eradicate H. pylori.

The decision to eradicate H. pylori to prevent gastric cancer is a public health decision where the problem must be weighed against the other pressing issues and the budget. The risks of a persistent infection include progressive exponential increase in gastric cancer risk, risk of developing a peptic ulcer (approximately 1 in 6), and among those an ulcer complication (approximately 1 in 4); this is also the reservoir of infection responsible for transmission (i.e., H. pylori is a transmissible infectious disease). In addition to the elimination of gastric cancer and peptic ulcer disease, there are many potential advantages of eradication (Table 2), including the mucosa associated lymphoid tissue lymphoma, functional dyspepsia, atrophic gastritis, vitamin B₁₂ deficiency, iron deficiency anemia, idiopathic thrombocytopenic purpura, and so forth.³³ Ideally, population-wide eradication would be done as it is once-off that would eliminate these diseases. However, new cases would continue the population as most immigrants are likely to be infected. Case finding and follow-up of individuals subsequently discovered to be infected would be needed as is done for many transmissible infectious diseases.

Reliable tests are available to diagnose H. pylori infections including histologic evaluation of the gastric mucosa, rapid urease test, culture, urea breath test (UBT), stool antigen test (SAT), IgG anti-H. pylori serology, and so forth. Noninvasive tests are preferred for population screening.^33,34 Both the UBT and SAT are >90% sensitive and specific; however, both are relatively expensive and require patient cooperation.^35,36 IgG anti-Helicobacter serology is cheap and if validated for that population, is relatively accurate. The main disadvantage of serology is that the test may remain positive after the infection is cured.³⁷ One way to utilize serology is to confirm positivity by using a second noninvasive test (e.g., UBT or SAT) on those with positive results.

During screening, the PPV of a screening test is the key process indicator to support the effectiveness of a program.³⁸ In a screen-and-treat strategy, the PPV is dependent on both the performance of H. pylori testing and the prevalence rate of H. pylori in a specific population. The equation can be expressed as follows:¹¹

In a population with a H. pylori prevalence of 80% (e.g., Latin American),^39,40 the PPVs for serological testing and SAT are estimated at 94% and 99%, respectively, which shows a minimal difference. Thus, the choice between two tests may depend on the infrastructure of a program and the available budget. When the prevalence is 40% (e.g., East Asia),⁴¹ the PPVs for serological testing and SAT are estimated to be 73% and 95%, respectively, showing a significant difference in performance between the two tests. Taking the screen-and-treat programs in Taiwan as examples, the prevalence of H. pylori was 60% and 35% for the populations residing in Matsu Island and Changhua County, respectively, before mass screening. The ¹³C-UBT has been adopted in the Matsu population, given a small high-risk population with adequate funding support and manpower,^10,42 while SAT was selected for the Changhua population for a large population with limited manpower and infrastructure in delivery of the screening test.¹¹

Highly effective antibiotic regimens are available for most populations. The success of antimicrobial therapy for a bacterial infection depends largely on the presence of susceptibility to the antibiotic or antibiotics and the adherence of the patients to the therapy. With a multiple-drug therapy, one only needs to know the antibiotic-resistance pattern and the eradication rates for each subgroup in relation to the antibiotics (e.g., susceptible to all, resistant to one antibiotic, etc).⁴³ With these data, one can create a reliable model to estimate the treatment outcome with combination of a proton pump inhibitor, amoxicillin, clarithromycin, and metronidazole. The on-line calculator available at https://hp-therapy.biomed.org.tw/.⁴⁴ In brief, for a regimen using amoxicillin, clarithromycin, and metronidazole, the treatment outcome can be predicted based on the summarized eradication rates according to the percentages of strains dual sensitive, single resistant, and dual resistant to clarithromycin and metronidazole, assuming that the amoxicillin resistance is a rare event, which can be summarized as:⁴⁵ (% success with all-susceptible strains)×(proportion with all-susceptible infections)+(% success with clarithromycin-susceptible strains)×(proportion with clarithromycin-susceptible infections)+(% success with metronidazole-susceptible strains)×(proportion with metronidazole-susceptible infections)+(% success with dual resistant strains)×(proportion with dual resistant strains).

Based on this model, the reported outcomes of the two-antibiotic and three-antibiotic regimens can be predicted by the prevalence of antibiotic resistance in a given population (Table 3);^44–58 the predicted values are indeed comparable to the observed ones. The results showed that triple therapy for 14 days may provide an incremental efficacy of about 8% as compared with 7 days’ regimen, while four-drug regimens for 14 days can provide an incremental efficacy of about 7% as compared with 14-day triple therapy.

An effective program will require effective therapies. For most populations, it is possible to identify two alternatives that are highly effective when given empirically. Patients can then be assigned to therapy using a checklist to identify those likely to fail (e.g., use of both clarithromycin and metronidazole in the past). As a general rule, the 14-day therapy provides the highest treatment success.⁴⁵ The local patterns of antibiotic resistance in the general population should be collected before the implementation of a mass eradication program. In addition, special treatment centers would be needed to offer the susceptibility-based therapy for difficult-to-cure cases. In Taiwan, the regimens to be considered for most patients are triple therapy, concomitant therapy, and sequential therapy all for 14 days. Treatment failures and penicillin allergic patients would receive levofloxacin-based therapy, bismuth quadruple therapy, or susceptibility-based therapy. The considerations for primary therapy include cost, effectiveness, and side effects. Concomitant therapy will always be equal or superior to sequential therapy but has higher cost. Sequential therapy is also more complicated but this can be eliminated by prepackaged drugs.

When the decision has been made to start eradication, the considerations may include whether to originally target the entire population, to target high risk populations, or to prevent new cases. H. pylori infections are typically acquired in childhood such that H. pylori eradication of young individuals before marriage would largely prevent transmission to their children and within several generations the disease would largely disappear. Also, this population has the highest proportion with nonatrophic gastritis and after the eradication therapy, the follow-up endoscopic surveillance is rarely needed. One would need to also ask whether to treat children and if so at what age. Screening would identify infected parents and thus identify families at greatest risk of having children with the infection. Reduction in the population burden of infection would also markedly reduce the chance of reinfection as the infection is almost exclusively restricted to humans.

However, in the real-world setting, an ideal design may be constrained by limitations in the framework for the delivery of screening tests or the system for providing treatment. For example, in Matsu Island,¹⁰ the residents had a high prevalence of H. pylori infection. The incidence of gastric cancer was about 50 cases per 100,000 person-years, which was approximately 3 to 5 times the incidence on Taiwan Main Island. H. pylori eradication would be expected to nearly eliminate the risk of gastric cancer in those with nonatrophic gastritis (most of whom would be younger persons) and stabilizing or reducing the risk among those with established atrophic damage (concentrated among older persons). The eligible age-range was set at 30 years and older (n=approximately 5,000) with the aim of accelerating the elimination of gastric cancer in this population. By contrast, in the Changhua County Study on the main Island of Taiwan, the incidence of gastric cancer was about 17 per 100,000 subjects per year.¹¹ Because the carcinogenic process was believed to be slower than that on Matsu Island, the decision was not to initially target young adults; the eligible age range was set at 50 to 69 years. In addition, because the number of eligible subjects was large (n=approximately 260,000), it was important to perform the screening test as economically as possible. Therefore, the SAT was selected and added to the established protocol for colorectal cancer screening with the fecal immunochemical test. This program demonstrated an easy-to-implement approach for an intermediate-risk population for gastric cancer while the efficacy of this approach requires further observation.

A cancer-screening program can easily fail without adequate participation. The relationship of adherence and effectiveness of a program can be expressed as follows:

Effectiveness of a program=efficacy of antibiotic treatment×adherence

Lower adherence will result in lower effectiveness of the screening program in preventing gastric cancer even when the intervention is highly efficacious. In both programs in Taiwan, potential participants were contacted either by telephone or with a pamphlet sent by mail, inviting them to participate in screening. Furthermore, as is the case with most cancer prevention programs, advertising and educational materials are helpful (an example available at http://epaper.ntuh.gov.tw/health/201208/special_1_1.html). Such materials may improve participation by illustrating the purpose of screening and describing the follow-up diagnostics when screening tests show positive results.

To ensure the quality of an organized screening program, it is mandatory to check whether the guidelines are being followed and that the results of the screening program are being regularly reported and evaluated. To reach this goal, a series of consensus meetings should be held to develop the treatment guidelines. Educational programs should be provided to both primary care physicians and first-line healthcare workers before implementation. After implementation, standardized quality indicators should be audited periodically to ensure the quality of screening.³⁸ In the first round in the Chunghua program,¹¹ among 3,621 tested subjects, the return rate of fecal samples was 95.4%, the positivity rate of SAT was 36.2%, and the referral rate for antibiotic treatment of those who tested positive was approximate 70%. The eradication rate of first-line therapy was approximate 88%. Among the 643 subjects also undergoing upper endoscopy, the PPV of SAT positivity (i.e., number with lesions/total number of diagnostic endoscopies) for gastrointestinal tract lesions was 31.9% and the detection rate was 5.9% (i.e., number with lesions/tested population). These indicator variables were periodically checked for outliers among the 26 townships.

The optimal regimen for treating the asymptomatic population within the community remains unclear. Previous studies evaluating the efficacy of H. pylori eradication focused mainly on symptomatic subjects who were recruited from hospitals. Researchers tended to increase the duration of treatment and the number of antibiotics in order to increase the eradication rate; however, this practice should be weighed against decreased adherence in asymptomatic carriers because they are less likely to tolerate the length and complexity of the antimicrobial regimen. Adherence to the antibiotic treatment is crucial for the success of this program.

In the two prevention programs in Taiwan, first-line triple therapies consistently showed an eradication rate of approximate 88%, given the antibiotic-resistance rates of approximately 1%, 8.2%, 3.8%, 21.6%, and 0.3% for amoxicillin, clarithromycin, levofloxacin, metronidazole, and tetracycline, respectively.⁷ In many areas that results might be considered sufficient for treatment of the population. However, in Taiwan a retest-and-retreat practice was utilized to improve the overall eradication and minimize the possible spread of antibiotic-resistant strains.^7,10,42 The second-line treatment, utilized a levofloxacin-containing triple therapy for 10 days that provides an approximately 80% eradication rate resulting in an overall eradication rate of approximately 98% after two courses of treatments. For those who did not respond to the two courses of treatment (2/100), a tailored rescue regimen was designed on the basis of drug susceptibility testing.⁶⁰ For a country-wide test-and-treat strategy, it would be important to examine the cost effectiveness of regimens with a higher initial yield such as concomitant therapy for initial therapy and 14-day levofloxacin triple therapy for treatment failures provided that 100% eradication was shown to be a cost effective goal.

In addition to antibiotic resistance of H. pylori, incomplete adherence to treatment is an issue that warrants attention. To improve adherence, a call center has been set up in Changhua County.^7,11 The personnel at the call center are equipped to evaluate the eradication rate of a specific regimen using the intention to treat (ITT) and per protocol (PP) analyses. For example, the eradication rates of the triple therapy for 7 days for the ITT and PP analyses were 88.1% and 89.0%, respectively. For sequential therapy for 10 days, the eradication rates were both 2% higher, reaching a level that can be regarded as a good eradication rate (i.e., >90%).⁶¹ Studies of different strategies in terms of eradication, adherence and cost effectiveness could easily be done as part of the initial implementation of a population-wide program.

The reinfection rate of H. pylori, which has the potential to overturn the effectiveness of this strategy, should be monitored by periodically testing those who had undergone successful eradication. In Matsu Island, the reinfection rate has been estimated to be about 1% per person-year during 2004 to 2008.¹⁰ This rate declined to 0.67% per person-year during 2008 to 2012,⁷ a decline that appeared in parallel with improvements in sanitation and hygiene in this population. However, a proportion of those with apparent reinfection are actually recrudescence as in a small proportion of subjects treatment results in prolonged suppression and false-negative test of cure results. This can be reduced by employing more effective regimens initially.

In Matsu Island, the screening program applied a quasi-experimental, before-and-after study design, designating the whole population of Taiwan as an external comparator group.¹⁰ The main outcome measure was the impact of mass eradication on the changes of premalignant gastric lesions and gastric cancer, which was obtained by comparing data from the pre- and posteradication program eras in the same population. The results showed that the effectiveness of reducing the incidence of gastric atrophy and peptic ulcer was significant at 77.2% and 67.4%, respectively; the reduction in incidence of gastric cancer was 25% although it was not statistically significant. To date, there are four rounds of mass eradication implemented in this population in 2004, 2008, 2012, and 2014, and the prevalence of H. pylori declined steadily from 60% to approximate 10%.⁷ We predict that a significant risk reduction of 47% will be observed in 2015.

In Changhua County, following the design of Matsu Gastric Cancer Prevention program, a pilot study was implemented in 2012. Among 7,463 participants during a 2-year period, the positivity rate of SAT was 34.4% with a referral rate of approximate 75%. Among the referrals, antibiotic treatment was prescribed to 99%, and five gastric cancers were found.⁷ Using the pilot study as a model, a randomized controlled study was launched in 2014 with inclusion of 10,000 subjects per year in each arm, and primary end-points of incident gastric cancer and death.

Not every proposed strategy can be realized at the population level, even though there is incontrovertible evidence to support the effectiveness of the strategy. The greatest impediment to widespread implementation is the fact that the proposed strategy may be costly, thus reducing its competitiveness with other health priorities. To address this problem, it requires cost-effectiveness analyses (CEA).⁶²

In the CEA, the first step is to create a model to simulate the natural history of the disease in the absence of screening, in which the parameters are generated from empirical studies of the specific population. The second step is to compare the natural course model (i.e., null hypothesis) with active screening (i.e., alternative hypothesis); taken together, these models constitute a decision model. Third, by inputting both the effectiveness and cost data into the decision model, the ratio of change in cost to the change in effect can be calculated (i.e., the incremental cost-effectiveness ratio [ICER]) between different strategies. We can also make a projection of the long-term outcome (i.e., Markov model) without the requirements of a large sample size and long-term follow-up, and without ethical issues that may hinder the performance of an RCT. Fourth, by specifying distributions for all relevant parameters, the uncertainty in this model can be presented as an acceptability curve, which can provide the probability of each intervention being cost-effective according to different levels of the maximal willingness to pay (WTP) for a specific outcome. The final step involves varying the input of a parameter in the model by a given amount and examining whether the model’s results will change (i.e., the sensitivity analysis). From the public health perspective, a one-size-fits-all intervention may not be optimal, and the purpose of a sensitivity analysis is to inspect whether a fine-adjustment of strategy is needed to make the screening program work well on the population level, which is, in fact, composed of diverse subpopulations.

In Taiwan, the CEA has been adopted in the evaluation of costs in curing H. pylori infection⁴⁴ as well as in the prevention of death from gastric cancer.⁶³ With regards to the former, the relative cost-effectiveness has been evaluated between two-antibiotic and three-antibiotic therapies. Results based on the base-case values showed a higher treatment efficacy for sequential therapy for 14 days than for triple therapy for 14 days. Sensitivity analyses according to prevalence of antibiotic-resistant strains showed that sequential therapy for 14 days was the more efficacious regimen (compared to triple therapy) in all areas, except in areas with very low (<5%) clarithromycin resistance and very high (>80%) metronidazole resistance; sequential therapy for 10 days appeared to be more effective than triple therapy for 14 days only in areas where the metronidazole resistance was <40%. Taking cost into consideration (Fig. 3A), the triple therapy for 14 days was found with a low (<5%) possibility of being cost-effective. The sequential therapy for 10 days was the more cost-effective across a wide range of WTPs, while the same regimen for 14 days could be an alternative choice when the WTP is high. It is important to note that these analyses were restricted to two regimens and did not compare regimens such as concomitant or hybrid therapy which are often superior to sequential therapy in specific populations.

With regards to the prevention of gastric cancer, cost-effectiveness of the screen-and-treat strategy has been evaluated on a global scale (Table 4).^63–72 When compared with no screen, this strategy was generally considered to be cost-effective, and there was a trend towards recommending a younger age for initiation of screening. Among the various subpopulations to be considered are those at risk of transmission to children (an elimination of transmission subgroup), those with nonatrophic gastritis (almost complete prevention of gastric cancer subgroup), and those with atrophic gastritis (a reduction in gastric cancer subgroup). Each may require a different strategy.

It deserves mention that all studies may have underestimated the relative cost-effectiveness of mass eradication because the benefit of reducing dyspepsia and peptic ulcers and transmission was not take into consideration.⁷³ In Taiwan, the economic evaluation has been performed for the high-risk population residing on Matsu Island, comparing the screen-and-treat for H. pylori infection approach with traditional endoscopic screening based on the pepsinogen method (Fig. 3B).⁶³ Within the acceptable range of WTP for a life-year gain, once-only H. pylori eradication at the younger age was the most cost-effective. In addition to recommending an initial screening age, sensitivity analyses showed that the relative cost-effectiveness was subject to the risk of H. pylori reinfection and endoscopic quality in detecting early gastric cancer, suggesting that the choice is also dependent on the socioeconomic status of a given subpopulation.

Some studies have suggested that the widespread eradication of H. pylori may lead to unforeseen dangers, such as increases in gastroesophageal reflux disease, childhood asthma and atopy, obesity, inflammatory bowel disease, and the activation of latent tuberculosis.^74–78 However, there are confounders in all studies, and the underlying mechanism may not follow the presumed causal pathway (Fig. 4). This issue has been discussed in detail and the overall conclusion is that there is no current evidence that H. pylori infection provides a personal or population benefits.⁷⁹

In Taiwan, the prevalence of endoscopic esophagitis was found to increase following mass eradication in Matsu Island population, an area where atrophic gastritis is highly prevalent.¹⁰ The presumed mechanism may be related to the regeneration of gastric glands after the elimination of H. pylori, increased intragastric acid output, and more acid reflux into the esophagus; however, obesity may be a confounder.⁸⁰ Furthermore, a reduction in gastric cancer risk can definitely outweigh this potential drawback, and low-grade erosive esophagitis can be managed by lifestyle modification such as weight loss (if indicated), avoidance of nicotine, alcohol, and smoking, and the use of short-course or on-demand acid-suppressing therapy. Also, this observation may be subject to the extent of gastric atrophy in the baseline gastric histology; a previous meta-analysis based mostly on populations at low-risk for gastric cancer has shown that there was no association between H. pylori eradication and development of new cases of gastroesophageal reflux disease in subjects with dyspepsia.⁸¹