Strategy Selection and Comparison in the Cost-Effectiveness Analysis of Cancer Screening

Abstract

Cost-effectiveness analysis (CEA) is well-established in application to cancer screening programmes. CEAs of screening differ from analyses of therapeutic interventions as the choice of strategies for comparison is not limited simply by the different interventions to compare but also the intensity at which they are provided. This very much depends on modelling choices made by the analysts conducting the simulation study. Existing checklists and guidelines for the application of CEA fail to fully address important issues regarding strategy choice which are relevant for adequate incremental cost-effectiveness ratio (ICER) estimation. Applied CEA models are an important tool for assessing health and economic effects of healthcare interventions but are not best suited for illustrating methodological issues. The objectives of this thesis are to review the issue of strategy omission in the CEAs of cancer screening, to provide a simple, open-source model for the simulation of disease screening cost-effectiveness for teaching and research purposes, and to demonstrate how the omission of strategy can arise by observing the impacts of characteristics of screening schedules and parameter values on the shape and composition of the efficient frontier. This thesis comprises three studies. The first identified the issue of strategy omission existing in the CEAs of colorectal cancer (CRC) screening. The quality assessment tool developed by this thesis identified methodological concerns around the ICER generation due to inappropriate comparisons and inadequate comparators. This study revealed that many CEAs did not consider different types, i.e., stool-based and image-based, of screening modalities available for CRC screening. Many CEAs did not consider consistent screening schedules for the same type of screening modalities. In addition, this study found around half of the CEAs reviewed reported inappropriate ICERs for decision-making. In most cases, ICERs are confused with average cost-effectiveness ratios (ACERs), which are based on the comparison to the no-screening scenario. Another common inappropriate way of reporting is a cross-tabulated table that listed all the cost-effectiveness ratios (CERs), comparing all strategies to every other strategy. The second study described a microsimulation model expressly designed for methods demonstration in CEAs of cancer screening. This simplified model has rapid simulation time, allowing the simulation of a large range of screening strategies. This framework is highly flexible in adjusting the intensities of screening strategies. This model can be used to demonstrate the complexity of screening strategies and their cost-effectiveness. The scenario analysis of this study showed how to conduct a simple face validation via the observation of the impact of changes in parameter values on the cost and effectiveness estimates. This study also showed the relevance of considering both gross cost and effects estimates and relative outcomes net to a no screening scenario. This distinction between gross and relative outcomes is helpful in providing an intuitive understanding of how cost-effectiveness estimates vary with changes in parameter values. The third study used the pedagogical model proposed by the second study to demonstrate examples of strategy omission. The third study presented the trajectory of strategies with common characteristics on the cost-effectiveness plane with a variation in screening start ages, stop ages, and screening intervals. This study explained the policy-relevant portion of the efficient frontier which is comprised of strategies with ICERs close to the cost-effectiveness threshold and therefore should be of greatest interest to policy makers. Moreover, this study examined the convexity of the efficient frontier and whether this varies with changes in parameter values. In contrast to the position of efficient frontiers which gives a general understanding of the cost-effectiveness of screening relative to the no-screening, the change in the convexity implies the inconsistent change in ICERs of screening strategies with different intensities. This study provides guidance on how to avoid the omission of relevant strategies when conducting CEAs of cancer screening and suggests using an iterative method when choosing strategies to simulate. This approach uses the findings from the initial simulation to inform what other strategies might be worth adding to the simulation in order to include the most relevant strategies. The findings of this study give insights into how to expand the choice of strategies closer to the efficient frontier, improving the chance of strategies being cost-effective. Take as a whole, this thesis describes the issue of strategy omission in cancer screening CEAs in detail. It examines the issue within the published literature and within an abstracted model. Using insights from both the review and a simplified model it provides suggestions on how best to avoid the omission of relevant screening strategies. Future research can extend this thesis to establish more detailed and realistic guidance for CEAs of screening.