A Method to Adjust for Ascertainment Bias in the Evaluation of Cancer Registry Data

Abstract

Using the trend of age-standardized incidence rate of cancers (ASR) is inaccurate for registration withincomplete reporting, especially in developing registries. The relative age-standardized ratio (RASR) is a newmeasure that takes ascertainment bias of registration into account. RASR is calculated from the ASR for eachcancer divided by the ASR for leukemia. Leukemia was chosen as the reference because its ASR is rather constantover time in valid registries. The adjusted relative age-standardized rate (ARASR with same unit as ASR) iscalculated by multiplying the RASR for a specific cancer in a particular year by the sum of ASRs of that cancerover the years for which a trend is being determined and then dividing result by the sum of RASRs of the cancerfor those years. Two likely assumptions are behind use of ARASR, first, constant ASR of leukemia over time,second, if under/over-registration occurs, it happens for all cancers to the same extent (random under/overreporting).Using the ARASR with empirical data of valid Finnish and SEER cancer registries proved that trendof ASRs for each cancer is almost equal to its ARASR. Using trends of ARASRs instead of ASRs in a registry withincomplete data collection in first years of registration demonstrated more realistic results. In conclusion, theARASR is more accurate than the ASR for studying cancer incidence trends in registries with incomplete reporting.ARASRs in different countries or different times are comparable since they are age-standardized. Moreover,comparison between trends of ASRs and ARASRs can be used as a test for validity of registration.

Keywords