Background: Reports of dramatic increases in gastric cardia cancer incidence warrant concern. However, the recent introduction of a separate diagnostic code, the lack of a consensus definition of the cardia area, and the accelerating interest in cardia cancer may affect classification practices. Little is known about the magnitude of cardia cancer misclassification in large cancer registries.
Methods: In a well-defined Swedish population (1.3 million), we uniformly classified all patients with newly diagnosed gastric adenocarcinoma (from 1989 through 1994) with respect to gastric subsite, and we used this patient group as our gold standard. We then evaluated the completeness of the Swedish Cancer Registry in registering gastric adenocarcinomas against this gold standard and, further, assessed the completeness of cardia cancer registration and the rate of falsely included cases to estimate the potential impact on observed incidence trends.
Results: Our gold standard contained 1337 case subjects with gastric adenocarcinoma. Overall, the Swedish Cancer Registry was 98% complete with regard to gastric adenocarcinomas and had a 4% rate of falsely included cases. The completeness of coding cardia cancer was only 69%, and the positive predictive value for cardia cancer was 82%, with no improvement over time.
Conclusions: Although overall completeness of gastric cancer registration by the Swedish Cancer Registry was excellent, accuracy in registering cardia tumors was surprisingly low. Our estimates suggest that true cardia cancer incidence could be up to 45% higher or 15% lower than that reported in the Cancer Registry. This margin of error could accommodate the observed increase in cardia cancer in Sweden. Therefore, secular trends in cardia cancer incidence should be interpreted cautiously.