The aim of this study was to compare stomach cancer incidence and survival rates between four very distinct areas: Campinas (Brasil), Latin America, Iowa (USA), Northern America, Varese (Italy), Europe and Osaka (Japan) in Asia, and determine which of the differences are due to variations in the case mix and which are due to the care received. A proportional hazards regression method was applied to the relative survival rates to obtain geographical differences that were adjusted for age, gender, period of diagnosis, sub-site and stage. Age, gender, period and stage explained most of the variability between the areas (50-100% excess risk of death with respect to Osaka) in the survival rates for stomach cancer patients. In Iowa and Varese, information on the sub-site fully explained the remaining variability. The large survival differences between the four areas were almost totally due to the different case mixes of the stomach cancer patients. The importance of stage indicates that diagnostic delay may be a major clinical factor affecting survival.