Influence of site classification on cancer incidence rates: an analysis of gastric cardia carcinomas.

BACKGROUND Recent reports suggest that the incidences of cardia and gastroesophageal junction carcinomas have increased markedly. The influence of improvements in cancer site classification (i.e., from no specific site to a specific site) on these incidence rates is unknown. METHODS We analyzed data for all gastric cancers reported to the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) cancer registries from 1974 through 1998. We compared incidence figures adjusted for improvements in site classification with the standard unadjusted incidence rates traditionally reported from SEER data. All analyses used two-sided statistical tests. RESULTS Among white males, the proportion of gastric cancers with an unspecified location decreased from 38% from 1974 to 1976 to 14% in 1996 to 1998. Between 1974-1976 and 1996-1998, the adjusted cardia cancer incidence rate for white males was unchanged (5.3% increase, from 3.6 to 3.8 per 100,000 population/year, respectively; P =.59), whereas the unadjusted cardia cancer incidence rate underwent a statistically significant increase (77% increase, from 1.9 to 3.4 per 100 000 population/year, respectively; P<.001). During the same period, the adjusted noncardia gastric cancer incidence rate in white males decreased from 6.8 to 3.8 per 100,000 population/year (P<.001), an absolute decrease more than twice as large as that seen using standard unadjusted SEER data (from 4.5 to 3.2 per 100,000 population/year; P<.001). Similar findings were observed for black males. CONCLUSIONS Improved specification of gastric cancer sites may largely account for the purported increase in cardia cancer incidence in recent decades. Noncardia gastric cancer incidence may be decreasing much more rapidly than previously appreciated. These results illustrate the potentially large influence of changes in site classification on some cancer incidence rates.