Trend analysis of cancer incidence in Japan using data from selected population-based cancer registries.

Population-based cancer registries are operated by over 80% of prefectures in Japan. However, only a limited proportion of the registries can provide long-term incidence data. Here, we aimed to establish a method for monitoring cancer incidence trends in Japan using data from selected prefectures. Based on the availability of long-term (≥ 20 years) high-quality data, we collected incidence data from five prefectures (Miyagi, Yamagata, Fukui, Osaka, and Nagasaki), which included an annual average of 54,539 primary cancer cases diagnosed between 1985 and 2004. Cancer mortality data for 1995-2004 were obtained from the vital statistics. Representativeness and homogeneity of the trends were examined by funnel plot analysis of log-linear regression coefficients calculated for the most recent 10 years of data (1995-2004) of age-standardized rates (ASR). The ASR of incidence for five prefectures in total (5-pref total) showed a significant decrease, with an annual percent change (APC) of -1.0 (95% confidence interval [CI] -1.4: -0.6) for males and -0.4 (95% CI -0.8: -0.1) for females. Excluding data from Osaka (4-pref total) reversed the decreasing trend; the corresponding APC was +0.4 (95% CI -0.2: +1.0) for males and +0.7 (95% CI +0.5: +0.9) for females. The APCs for the ASR of mortality for the 4-pref total (males, -1.5; females, -1.3) were more representative of nationwide data (males, -1.4 [95% CI -1.7: -1.2]; females, -1.1 [95% CI -1.4: -0.9]) than those for the 5-pref total (males, -1.7; females, -1.4). We conclude that using data from Miyagi, Yamagata, Fukui, and Nagasaki prefectures, with continuous monitoring of the representativeness of the data, is a provisionally relevant way to evaluate cancer incidence trends in Japan.