Objectives To apply correction in life tables in mortality analyses to address selection effect with respect to the reference population in a cohort of sewer workers.

Methods We used excess hazard model to assess the excess mortality from all-cause and from all-cancer in an historical cohort of 1898 sewer workers between 1960 and 2011. National and regional life tables were available to assess the all-cause and the all-cancer background mortality. The corrections of these life tables were modeled by spline functions in the logit of survival scale. The excess hazard was modeled using splines functions with the time since hiring as the time scale. The parameters of the model were estimated by maximizing the likelihood. The expected excess number of cases were estimated using both regional and national model-based corrected tables and compared to those obtained without correcting the life tables. In a simulation study estimates were obtained using the corrected life table with a known model of correction (i), using the uncorrected life table but applying the model of correction (ii), and without model of correction (iii).

Results The simulation study showed that applying the model of correction reduces the estimation bias in the excess rate model. In the cohort study, for all-cause mortality, the difference between the excess numbers of cases estimated reduced from 28.0 using the original life tables to 1.5 when a model of correction was applied. For all-cancer mortality, the difference reduced from 24.3 to 11.8. However, the standard error was doubled.

Conclusions The differences between estimates obtained using two reference life tables decreased when the model of correction was applied at a cost of larger confidence intervals. Correction in life tables can be applied in mortality analyses when the life tables available are not fully suitable to the cohort studied.