Childhood cancer survival and avoided deaths in Australia, 1983-2016

Large improvements in childhood cancer survival have been reported over recent decades. Data from cancer registries have the advantage of providing a 'whole of population' approach to gauge the success of cancer control efforts.The aim of this study was to investigate recent survival estimates for children diagnosed with cancer Australia and to examine the extent of changes in survival over the last 35 years. For the first time, we also estimated the number of deaths among Australian children that were potentially avoided due to improvements in survival.A retrospective, population-based cohort study design was used. Case information was extracted from the Australian Childhood Cancer Registry for 1983-2016, with follow-up to 31 December 2017. Eligible children were aged 0-14 with a basis of diagnosis other than autopsy or death certificate only. Five-year relative survival was calculated using the semi-complete cohort method for three diagnosis periods (1983-1994, 1995-2006 and 2007-2016), and changes in survival over time were assessed via flexible parametric models. Avoided deaths within 5 years for those diagnosed between 1995 and 2016 were estimated under the assumption that survival rates remained the same as for 1983-1994.Overall 5-year survival within the study cohort (n = 20,871) increased from 72.8% between 1983 and1994 to 86.1% between 2007 and 2016, equating to an adjusted excess mortality hazard ratio of 1.82 (95% confidence interval 1.67, 1.97). Most cancers showed improvements in survival; other gliomas, hepatoblastoma and osteosarcoma were exceptions. Among children diagnosed between 1995 and 2016, 38.7% of expected deaths within 5 years of diagnosis (n = 1537 of 3970) were avoided due to temporal improvements in survival.Survival for childhood cancer has continued to improve over recent years, thanks mainly to ongoing progress in treatment development combined with improved supportive care. Providing innovative measures of survival, such as avoided deaths, may assist with understanding outcome data produced by cancer registries.