Should years lost always be equated with life expectancy?

The 'years lost' by a person dying prematurely from some cause is usually equated with life expectancy at the age of death derived from a life table for either the general population or a population in which the cause does not operate. It is suggested that this procedure may not always be valid.

The calculation of years lost by individuals dying prematurely from smoking-related deaths is taken as an example using data from the American Cancer Society Cancer Prevention Study (ACS CPS II) and from Peto et al. An alternative hypothesis, whereby smoking advances the age of death by an amount considerably less than the life expectancy, is examined.

It is shown that when smoking-related deaths are removed from the ACS CPS II data, the life expectancy of the smokers is still less than that of the non-smokers. Secondly, it is demonstrated that, if the alternative hypothesis is used to predict a survival curve in the absence of smoking, it would be incorrect to equate years lost with life expectancy calculated from that curve.

Years lost cannot automatically be equated with life expectancy. In the case of smoking, estimates of years lost must still be subject to considerable uncertainty. Further research is needed to see if smokers dying at a given age have comparable physical and social characteristics to all smokers living at that age.