Estimation of population-based cancer-specific potential years of life lost in Belgium.

Abstract

The potential years of life lost (PYLL) observed in a cohort of cancer patients cannot be fully assigned to the cancer as both cancer-related and non-cancer-related deaths contribute. A method is presented to decompose the observed all-cause PYLL into cancer mortality and population background mortality fractions when cause of death information is not available. Furthermore, the association of cancer-specific PYLL with cancer-specific mortality and mean age at diagnosis is explored and the impact of the follow-up window length is examined. The framework of the actuarial relative survival and the Ederer II method is applied to estimate the population background mortality contribution, PYLL*. The fraction (PYLL-PYLL*)/PYLL is then attributed to the cancer. The method is applied to cancer incidence in Belgium 2004-2014, about 631 300 cancer patients. The cancer-specific PYLL divided by the number of cancer patients, mean PYLL, in the Belgian cancer population ranges from 0.4 years for prostate cancer to 15 years for tumours of the central nervous system. The cancer-specific mean PYLL increases with both increasing cancer-specific mortality and decreasing age at diagnosis. Longer follow-up periods yield larger cancer-specific mean PYLL until statistical cure of cancer is achieved. The mean PYLL results, obtained by dividing the PYLL by the number of cancer patients, are visualized in combination with cancer incidence and mean age and mean life expectancy at diagnosis, providing an elegant summary to rank and compare cancer sites in terms of incidence, relative survival and PYLL.