Applications of Mathematical Modeling in Implementing Cancer-Control Strategies

Abstract

Mathematical modeling (MM) is a valuable tool in planning and evaluating health interventions. Modeling and simulating change prior to implementing interventions or policies is a recognized strategy in the field of implementation science. However, in spite of its potential, it is under-utilized in this field, since the majority of studies applying MM simulate health outcomes rather than implementation outcomes, such as equity, reach and cost. The overall objective of this research is to demonstrate, through case three studies, applications of MM in implementing cancer-control interventions.The first application involves exploring whether race-informed, prostate-specific antigen (PSA)-based screening strategies could mitigate observed racial disparities in fatal prostate cancer (PCa) in the US. This aim is motivated by other studies concluding that Black men have greater disease burden at an earlier onset. I used a microsimulation model of PCa natural history, which was calibrated to incidence data from population-based registry data and randomized trials. The model approximates historical incidence and mortality rates for the general US population and for Black men under historical PSA-based screening. I projected and compared age-specific incidence of fatal PCa (fPCa) for the general US population and for Black men under both historical and hypothetical intensified PSA-based screening strategies, shortening the inter-screening interval from biennial to annual and lowering the starting age for Black men from 50 to 40 years. I concluded that Targeted PSA-based screening can potentially mitigate racial disparities in fPCa incidence particularly among young men, but yet does not eliminate those disparities. Outputs from this aim could inform designing race-specific prostate-cancer screening guidelines. In the second application, I evaluate the cost-effectiveness implications of disseminating blood-based tests for colorectal cancer (CRC) screening among currently-screened populations in US. This aim was motivated by other studies concluding that blood tests are less cost-effective but more convenient compared to other screening tests. I used two microsimulation models which simulate individual life histories of adenoma development, growth, and progression to CRC. Both models were calibrated to incidence data from population-based registry data. The main parameters that defined my simulation scenarios were the percentage of switching to blood tests among those who have been participating in screening, and the percentage of new uptake of blood tests among those who never screened before. Under different combinations of these two parameters, I projected health outcomes, costs, and cost-effectiveness of introducing blood tests, calculated and reported model-specific thresholds levels of how much new uptake is needed to offset the losses resulting from switching. By factoring in trade-offs involved in improving adherence, reducing effectiveness and increasing costs, this aim will inform policy decision making with regards to screening modalities. In the third application, I projected the health outcomes and implementation costs for Uganda adopting a selected list of comprehensive cancer-control strategies targeting the five highest-brden cancer types (cervical, prostate, esophagus, breast, and liver). This was a proof-of-concept model-building exercise and was meant to help identify financing strategies to deliver cancer care, since efforts in cancer modeling lack comprehensive, integrated policy-focused models that cover multiple standardized sets of interventions. I built a hybrid model that is composed of: (i) an open-population average, state-transition (Markov) model for health state projection; and (ii) a Cohort Component Model for population projections in future years. I simulated a scenario where an intervention package with coverage scale up (80%) and out-of-pocket expenditure scale down (0%) by 2050. I projected deaths averted, costs, and cases of inducted poverty under different financing strategies. The work also allowed quantifying financial risk protection and assessing Universal Health Coverage. Building on this work, I also plan on developing a decision-support tool to aid policymakers in prioritizing cancer prevention and early detection strategies. Collectively, my work will facilitate adoption and dissemination of evidence-based interventions and promote evidence-informed decision-making in the field of cancer control.