Implications from Under-reporting at Lifetime, Death Certificate Notifications and Trace-back on the Recorded Incidence of a "Newly" Established Population-based Cancer Registry.

Abstract

Population-based cancer registries (CRs) are powerful tools for measuring cancer burden and progress against cancer. The study's objective was to investigate the effects of under-reporting at lifetime, death certificate notifications, and trace-back on the incidence a newly established population-based CR may record during its initial 15 years of operation. Using cancer data of nine CRs of the SEER programme we performed model calculations to investigate temporal trends of the recorded incidence that might be expected if registration started in 1995 with gradually increasing proportions of cancers reported to the CR. It was assumed that the CR obtains all death certificates providing cancer as the underlying cause of death. Different scenarios with regard to the development of the proportions of cancers reported to the CR and the use of trace-back were evaluated. Our model calculations demonstrated that the inclusion of cancers notified from death certificates which were diagnosed prior to the start of registration and which attribute to the incidence estimates of the year of death ('prevalent' cases) may compensate under-reporting typically observed right after the start of a CR. The recorded incidence may even be overestimated during the first years of registration, if large amounts of prevalent cancers are notified from death certificates (e.g. overestimation of lung cancer incidence by 8% and 21% in the first year of registration, if the proportions of cases reported were 50% and 70% in that year, overestimation of myeloma incidence still exceeding 5% after eight years of registration, if the proportion of cases reported to the CR had reached 97.5% after six years). Trace-back may effectively reduce the recorded surplus cancer cases. During the initial years of registration, the inclusion of prevalent cancers from death certificates may compensate the higher amount of underreporting right after the start of a CR. Accurate incidence estimates may nevertheless be observed for highly fatal cancers if substantial proportions of cancer cases are missed at lifetime, however apparent incidence trends observed in the initial years of registration need to be interpreted with caution.