Objectives In Japan, prostate cancer constitutes the fourth common cancer and the sixth major cause of cancer death in men. Despite the rising incidence, it remains unclear whether Prostate Specific Antigen (PSA) screening has contributed to early detection and subsequent reduction in late-stage incidence of and mortality from prostate cancer, or it is a manifestation of overdiagnosis. The time-trend analysis of prostate cancer incidence by stage at diagnosis will help understand the impact of PSA screening. However, analyses of trends in cancer incidence have been restricted due to unreported statuses of stage at diagnosis in early days. Using multiple imputation approach, this study aims to assess the long-term trends in the incidence of prostate cancer by stage at diagnosis before and after the introduction of PSA screening in Japanese municipalities.

Method We used data from three population-based cancer registries (Yamagata, Fukui and Nagasaki prefectures) in Japan. These three registries have been established to monitor the long-term trends in cancer incidence ensuring a high degree of completeness and timeliness. A total of 29,458 malignant prostate cancer cases diagnosed between 1993 and 2014 were used for the analysis. Sequential polytomous logistic regression, or alternatively called multiple imputation with chained equations (MICE) was used to impute a specific stage at diagnosis (localized, regional or distant) for the cases with ‘Unknown’ status and cases with missing status. For the imputation, we included variables from the cancer registry data that can predict the stage at diagnosis and its missingness. For the trend analyses, we estimated the age-standardized incidence rates by stage at diagnosis from 1993 to 2014, and Joinpoint linear regression models were used to show the joined trend with 95% CIs.

Results A more pronounced increase in the incidence of localized prostate cancer was seen from 2001 (12.1 per 100,000) to 2003 (34.5 per 100,000) after imputation compared to a complete-case analysis from the original data (7.2 per 100,000 in 2001 to 11.4 per 100,000 in 2003). Joinpoint analyses also showed a significant increase in the imputed dataset in 2003 and after (Annual Percentage Change, APC: 4.1%). For regional cancer, the original data before imputation showed a significantly increasing trend until 2009 (APC: 12.6%), while the imputed data showed that the increasing trend stopped earlier in 2001 (APC: 8.4%). For distant prostate cancer, the original data showed a constant increase in the incidence rates throughout the study period (APC: 2.4%), while the results from the imputed data showed that the increasing trend stopped in 2005 (APC: 2.8%), followed by an insignificantly declining trend since 2005.

Conclusions Our study suggests a stage shifting - a decrease in the number of advanced-stage prostate cancer coupled with an increase in localized cancer - which coincides with the time when PSA screening rapidly spread among Japanese municipalities in early 2000s. However, confirmation of the impact of PSA screening on mortality reduction would require a comprehensive analysis including changes in mortality, survival and treatment practices over time, which warrant further investigation.