H. pylori is a formidable public health adversary. More than 40 percent of adults worldwide carry it, and infection, typically acquired in childhood, generally persists for life unless deliberately eradicated. Most carriers never develop symptoms, so few seek testing or treatment, allowing a vast silent reservoir of chronic infection to accumulate. The International Agency for Research on Cancer estimates that roughly 850,000 cancer cases worldwide in 2020 were attributable to the bacterium, more than from any other cancer-causing infection, including human papillomavirus and the hepatitis B and C viruses. China, which accounts for about 35 percent of global gastric cancer cases, sits at the epicentre of this burden.

Earlier meta-analyses had reported that H. pylori prevalence in China fell from roughly 55 to 60 percent in 1990 to around 40 to 45 percent by 2020, fuelling optimism that the epidemic was receding on its own. The new study, led by Dianqin Sun of Erasmus MC in collaboration with researchers in the United States, Mexico, and the International Agency for Research on Cancer, argues that this optimism rested on shaky statistical ground. Previous pooled estimates ignored the age structure of the populations being surveyed, conflated distinct measures of infection, and failed to correct for the imperfect accuracy of diagnostic tests, biases that can make prevalence appear to fall faster than it truly does.

The team assembled an unusually deep evidence base. From six published meta-analyses and an updated search of PubMed, Embase, and Chinese databases, 741 original studies were initially identified; after rigorous screening for general-population relevance, adequate reporting, and acceptable detection methods, 180 unique studies covering 1983 to 2023 and all 31 provinces of mainland China qualified for meta-analysis, contributing 805 records of current infection or seroprevalence. One additional study measuring ever-infection was reserved for the transmission model, and a large nationwide survey was held back entirely as an independent validation target.

Crucially, the researchers recognised that the three main ways of measuring H. pylori are not interchangeable. Current-infection tests detect active bacteria, serology detects antibodies that can linger long after the infection has cleared or been treated, and ever-infection measures capture both. Pooling them together, as earlier reviews did, distorts trends. The team also applied the Rogan-Gladen correction to adjust each study’s observed prevalence for the sensitivity and specificity of the diagnostic test used, then pooled estimates separately by age, calendar period, and birth cohort using random-effects models on the logit scale.

The centrepiece of the analysis is a Bayesian catalytic model, a class of transmission model that infers the rate at which susceptible individuals acquire infection, known as the force of infection, from age-specific prevalence data. Individuals in the model begin uninfected, may become infected, and after treatment can clear the bacterium while remaining seropositive before eventually seroreverting. The model incorporated cohort-specific limits on lifetime infection risk, period-varying treatment rates, a constant rate of antibody waning, and diagnostic misclassification, and was fitted to the record-level data with Hamiltonian Monte Carlo. Four competing model structures were compared using leave-one-out cross-validation, and the selected model reproduced observed prevalence closely and passed external validation against the nationwide survey.

The model’s estimates tell a striking generational story. For the 1920 birth cohort, the annual force of infection was 3.21 percent at age 10; for the 2000 cohort it was just 1.72 percent at the same age, and childhood infection rates in recent cohorts now resemble those seen among young adults born in the 1910s. A clear inflection appears after the 1980 birth cohort, which the authors link to the sweeping improvements in water supply, sanitation, housing, and household hygiene that followed China’s reform and opening-up policy from 1978. Modernisation, in other words, reshaped transmission dynamics well before any medical intervention reached scale.

Yet the population-level picture is far less encouraging. Overall prevalence declined only modestly, from 43.1 percent in 1990 to 36.3 percent in 2023, and much of that apparent decline disappears once the ageing of the population is accounted for; age-standardised prevalence fell further than crude prevalence, indicating that demographic ageing has propped up the raw numbers. Because gastric cancer and peptic ulcer disease strike predominantly at older ages, the children who escaped infection have not yet reached the ages where disease risk peaks, while large, heavily infected cohorts continue to age into the danger zone. The population attributable fractions in 2023 remained 70.1 percent for non-cardia gastric cancer and 49.4 percent for cardia gastric cancer, using China-specific relative risks from the China Kadoorie Biobank and a ten-year latency window for cancer and a five-year window for ulcers.

The authors are careful about the limits of their estimates. Subtype-specific incidence data were unavailable, so they modelled bounding scenarios for the mix of cardia and non-cardia tumours and for duodenal versus gastric ulcers; the gastric ulcer attributable burden remained statistically compatible with zero. They also note that the synergy between H. pylori and nonsteroidal anti-inflammatory drugs was not modelled, which may underestimate the ulcer burden, and that substantial regional variation across China is masked by the national averages. Sensitivity analyses using alternative model structures produced similar attributable burden estimates, reinforcing the central conclusion.

The policy implications are pointed. The eradication benefits of treating H. pylori are well established, and guidelines from the World Health Organization, the Taipei Global Consensus, and Chinese expert bodies increasingly endorse organised screen-and-treat programmes and family-based approaches, which China’s expert consensus recommends but no national policy yet implements. Comparable efforts are advancing elsewhere: South Korea is trialling population screening within its national cancer programme, Japan reimburses eradication therapy for H. pylori gastritis, and the European Council has urged high-burden member states to consider screening. The study’s age- and cohort-specific estimates provide the quantitative foundation such programmes need, identifying which cohorts carry the heaviest infection load, how many people would be eligible for treatment, and what costs and antibiotic-use concerns each strategy would entail. The modelling framework, the authors add, is transferable to other high-burden countries undergoing similar epidemiological transitions, particularly in East and Southeast Asia, where declining childhood infection coexists with large infected adult cohorts moving into the gastric cancer risk window. Without active intervention, the analysis warns, China’s H. pylori-attributable disease burden is unlikely to decline rapidly for decades.

Subject of Research: Trends in Helicobacter pylori prevalence, force of infection, and attributable gastric cancer and peptic ulcer disease burden in China from 1990 to 2023

Article Title: Trends in Helicobacter pylori prevalence, force of infection, and attributable disease burden in China: a systematic review and Bayesian modelling study

Article References: Sun, D., Han, W., Alarid-Escudero, F., Roa-Contreras, J. A., Wang, S., Mülder, D. T., Huang, R. J., Ladabaum, U., Spaander, M. C., Camargo, M. C., Park, J. Y., Chen, W., Lansdorp-Vogelaar, I., & Hontelez, J. A. (2026). Trends in Helicobacter pylori prevalence, force of infection, and attributable disease burden in China: a systematic review and Bayesian modelling study. The Lancet Regional Health – Western Pacific, 74, Article 101979. https://doi.org/10.1016/j.lanwpc.2026.101979

Image Credits: AI Generated

DOI: 10.1016/j.lanwpc.2026.101979

Keywords: Helicobacter pylori, gastric cancer, peptic ulcer disease, Bayesian modelling, force of infection, systematic review, China, epidemiology, screen-and-treat, birth cohorts, population attributable fraction, public health