Estimating complete prevalence of cancers diagnosed in childhood.

A method of estimating the complete prevalence of cancers diagnosed in childhood called CHILDPREV (CHILDhood PREValence), is presented. It is a semi-parametric method based on cancer registry data and on the completeness index method. It allows estimating prevalence even when no observation is available (typically older patients alive at the prevalence date may have been diagnosed with cancer before the introduction of the registry). The method was validated on Connecticut Tumor Registry data, which has 62 years of follow-up and provides complete prevalence, and compared with the fully parametric PIAMOD method. Results of complete childhood prevalence estimates based on SEER-9 cancer registries data for acute lymphocytic leukemia and all cancer sites combined are presented. Copyright (c) 2007 John Wiley & Sons, Ltd.