Semi-Markov modeling for disease incidence risk and duration

Over the last decade, the number of years of life lost (YLL) has become a popular metric in biostatistics for assessing mortality and life expectancy discrepancies between patient cohorts and the general population. Using data from the Belgian Cancer Registry (161,007 cases of melanoma, thyroid, and female breast cancer), a three-state (healthy–cancer–death) illness-death model is used to illustrate how it can be applied to cancer registry data to estimate the incidence risk and YLL due to cancer at various ages of diagnosis and survival times post-diagnosis. The probabilities of being diagnosed with cancer over the next 20 years remain low for melanoma and thyroid cancers for both sexes but considerably increase with age for female breast cancer. YLL before age 70 due to cancer is highest for early diagnoses of female breast cancer but peaks at later ages for melanoma and thyroid cancers. Additionally, male patients generally experience higher YLL before age 70 due to cancer than females for melanoma and thyroid cancers. For patients surviving 10 years post-diagnosis, YLL before age 70 due to cancer remains below one year for melanoma and thyroid cancers, suggesting a limited impact on life expectancy compared to the general population.