Purpose: Helicobacter pylori (Hp) infection is responsible for ~40% of all cases of gastric cancer (GC), the second leading cause of cancer-related deaths worldwide. Potentially promising interventions include antibiotic treatment, screening for precancerous lesions, and an eventual Hp vaccine. Our objective was to develop a natural history model of GC that would be capable of evaluating cost-effectiveness of primary and secondary prevention efforts.

Methods: Our state-transition model simulates the natural history of intestinal-type GC in China and Columbia, two high-risk regions with an adult Hp seroprevalence of ~70%. Health states include gastritis (with and without Hp), atrophy, metaplasia, dysplasia, and invasive cancer. Initial transition probabilities were derived from country-specific prospective cohort studies and the literature. A series of calibration exercises were conducted to elucidate the unobserved parameters for the model while ensuring model output was consistent with age-specific prevalence of precancerous lesions and GC incidence.

Results: In both China and Columbia lifetime GC risk was ~3% although the shape of the age-specific cancer incidence curve differed. Unobserved natural history parameters reflecting progression between precancerous health states, and derived from our calibration exercises, differed between countries. For example, although the majority of transition probabilities were within the plausible ranges, the probability of progression to metaplasia was ½-fold higher in Columbia while those for progression to atrophy and dysplasia were 5- and 2-fold higher in China, respectively. Use of a more stringent definition of atrophy in Columbia may explain the differences in progression probabilities to atrophy and metaplasia. Higher prevalence of advanced precancerous lesions in initial distribution and probability of progression to dysplasia in China may reflect country-specific levels of Hp infection and other factors, such as smoking, diet, and genetics. A refined set of natural history parameters will be imputed through a second series of calibration exercises that leverages country-specific data on these risk factors.

Conclusions: Natural history models calibrated to different epidemiologic settings offer an opportunity to leverage region-specific data in order to understand geographical variation of risk factors and their influence on disease progression.