BMJ No 7119 Volume 315
Letters Saturday 22 November 1997
New method for expressing survival in cancer
Editor,| I was interested to read of the fraction of normal remaining life span (FNRL) methodology(1) proposed as a new way of expressing survival in cancer. The simplicity of the method was extolled and contrasted with the "cumbersome calculations" of other well established methodologies for incorporating information on background survival rates. I believe this simplicity is beguiling and have concerns about the interpretation of the resulting curves at the level of the individual. If absolute survival is dominated by a specific cause (eg cancer) then it may not be particularly age related. Analysis based on FNRL, however, will be induced to be strongly dependent on age due to the division by the normal remaining life expectancy term - a decreasing function of age. Thus FNRL curves unadjusted for age should not be interpreted at the level of the individual as was done, mistakenly in my opinion, in the section headed Comparison between conventional and real life expectancy method.(1)
An example should clarify the forgoing discussion - the data on 574 women with pathologically staged early breast cancer of ductal histology who underwent primary surgery in the 1980's has been extracted from a database held by Dr A Howell at the Christie Hospital, Manchester. Some simple analyses (for illustration only ignoring known prognostic features) using absolute survival and FNRL are given in figure 1 - normal remaining life expectancies were obtained for England and Wales.(2) Figure 1C reveals no marked difference in absolute survival between three arbitrarily chosen age groups (p = 0.55, logrank test). In contrast Figure 1D reveals the induced differencies in the FNRL curves between the same three age groups (p < 0.0001, logrank test) with the estimated probabilities of surviving at least 1/5 of normal remaining life, for example, being 59% in the under 50's, 71% in the 50-65 year olds and 87% in the over 65's - estimates which deviate markedly from the non age-adjusted estimate of 72%. In addition the age dependency of follow-up in the FNRL scale becomes apparent. A further problem with the proposed FNRL methodology is the static nature of the normal remaining life term - in reality this is an ever changing feature, especially in the context of long term follow-up. My preference is to use additive models for excess risk(3) as these may be formulated in ways that permit scientific interpretation in terms of competing risks of death and also allow the dynamic nature of the background rates to be fully catered for. Hopefully the message is clear - FNRL curves unadjusted for age can be misleading. W D J Ryder Statistician
Department of Medical Statistics, | Fig 1: Survival (A,C) and Fraction of normal remaining life (B,D) curves for 574 cases of early breast cancer - overall (A,B) and stratified by age (C,D) with age < 50 (solid n=166), 50 - 65 (short dash n=254) and > 65 (long dash n=154) | |
 Fig 1A | ||
 Fig 1B | ||
 Fig 1C | ||
 Fig 1D |
1 Vaidya J S, Mittra I. Fraction of normal remaining life span: a new method for expressing survival in cancer. BMJ 1997; 314:1682-4. (7 June.)
2 English Life Tables 14. HMSO. London 1987.
3 Andersen P K, Vth M. Simple parametric and nonparametric models for excess and relative mortality. Biometrics 1989;45:523-35.
Home | Current issue | Past issues | Classified ads | Career Focus | Feedback
Collections | About this site | About the BMJ | BMA | Medline