Oral Presentation COSA 2015 ASM

Estimating rare cancer net survival in New South Wales: Comparison of Ederer II and Pohar-Perme estimators (#56)

Hui You 1 , Richard Walton 1 , Deborah Baker 1
  1. Cancer Institute NSW, PO Box 41, Alexandria, NSW, Australia

Aims
Cancer net survival is a major epidemiological indicator to compare cancer survival between areas or countries. Recent findings have shown that the traditional methods including Ederer II provided biased estimates for net survival. A new less biased estimator, the Pohar-Perme estimator, was proposed in 2012. It takes into account the fact that the competing risks of death from other causes increase with age. We compared two estimates for rare cancers by time of follow-up, cancer site and age.


Methods
Using cancer registry data for New South Wales during 1995-2009 we estimated unstandardized and age-standardized net survival Ederer II and Pohar-Perme estimators for 15 rare cancers, adjusted by background mortality. The period method was used for those who were at risk during the period 2005-2009.


Results
A total of 29,574 cases were eligible for inclusion. Differences between the estimators increased with time of follow-up but were, in general, small during the first 10 years of follow-up. The good-prognosis cancers generally had greater bias than poorer-prognosis cancers. For unstandardized survival, at 10 years the absolute difference was 0.1% for mesothelioma (Ederer II 2.4%(95%CI 1.4%–4.0%) vs. Pohar 2.3%(95%CI 1.2%–4.1%)), 2.0% for lip cancer (Ederer II 84.4% (95%CI 80.9% - 87.7%) vs. Pohar 82.4% (95%CI 76.5% – 87.0% )) and 0.2% for bone cancer(Ederer II 57.5% (95%CI 50.4% - 64.1% ) vs. Pohar 57.7% (95%CI 49.7% – 64.9% )). For age-standardized survival, at 10 years the absolute difference was 0.4% for mesothelioma, 0.1% for lip cancer and 0.1% for bone cancer.


Conclusions
Both estimators can be used to estimate rare cancer net survival for population based cancer studies. The Pohar-Perme estimator provided the less biased estimates for net survival but with higher variance than Ederer II estimates.