Complete Cancer Prevalence Predictions via LASSO Regularization and Flexible Parametric Survival Cure Models
Cancer prevalence is a key measure of disease burden, and accurate projections are essential for healthcare planning and resource allocation. The Prevalence Incidence Analysis Model (PIAMOD) is widely used to estimate and project cancer prevalence in population-based studies by combining incidence rates and net survival within a forward calculation framework. We propose an improved approach to projecting cancer prevalence by using regularization techniques to select the optimal incidence model and flexible parametric survival cure models to improve the estimation of net survival. The proposed and conventional approaches were compared through a simulation study and an application to population-based data on patients diagnosed with colon cancer in Sweden between 1958 and 2019. Applying the proposed methodology, colon cancer prevalence in Sweden was projected through 2030. The number of individuals living with a previous diagnosis of colon cancer is projected to continue increasing, reaching approximately 550 per 100\,000 males and 500 per 100\,000 females by 2030. Careful model selection is essential when estimating measures of disease burden such as cancer prevalence. The projected increase in colon cancer prevalence in Sweden reflects continued improvements in survival for both sexes and highlights the growing number of individuals living with colon cancer.