A guide to interpreting estimated median age of survival in cystic fibrosis patient registry reports

Abstract

Survival statistics, estimated using data collected by national cystic fibrosis (CF) patient registries, are used to inform the CF community and monitor survival of CF populations. Annual registry reports typically give the median age of survival, though different registries use different estimation approaches and terminology, which has created confusion for the community. In this article we explain how median age of survival is estimated, what its interpretation is, and what assumptions and limitations are involved. Information on survival from birth is less useful for individuals who have already reached a certain age and we propose use of conditional survivor curves to address this. We provide recommendations for CF registries with the aim of facilitating clear and consistent reporting of survival statistics. Our recommendations are illustrated using data from the UK Cystic Fibrosis Registry.

Keywords: Annual report; Birth cohort; Conditional survival; Median survival; Patient registry; Period estimates; UK Cystic Fibrosis Registry.

Fig. 1

Illustrations contrasting the period approach and the birth cohort approach. Lines represent 12 individuals followed-up during 1st January 1985–31st December 2015. Individuals 1 and 2 were born prior to 1985 and individuals 1, 4, 8 and 11 were known to be alive at the end of follow-up. Individual 7 was lost-to-follow-up. Individuals 2 and 10 do not contribute to either analysis.

Fig. 2

Estimated survivor curves using data from the UK Cystic Fibrosis Registry. The shaded area shows the 95% confidence intervals and the down-arrow indicates the estimated median survival age. Estimates are based on the period approach using data on 10,946 individuals and 654 deaths observed in the 5-year period 1st January 2011–31st December 2015.

Fig. 3

The upper plot shows estimates of median age of survival, conditional on survival to a given age. The dots show the estimated conditional median age of survival from each age and the bars show the 95% CIs. Estimates are based on the period approach using data on individuals observed in the 5-year period 1st January 2011–31st December 2015. The lower plots show the estimated conditional survival curves from ages 30 and 40 indicating the conditional median age of survival. Supplementary Fig. 1 shows estimates for males and females separately.