How to Appropriately Extrapolate Costs and Utilities in Cost-Effectiveness Analysis

Abstract

Costs and utilities are key inputs into any cost-effectiveness analysis (CEA). Their estimates are typically derived from individual patient level data collected as part of clinical studies whose follow up duration is often too short to allow a robust quantification of the likely costs and benefits a technology will yield over the entire patient's lifetime.In the absence of long-term data, some form of temporal extrapolation -to project short-term evidence over a longer time horizon -will be required. Temporal extrapolation inevitably involves assumptions regarding the behaviour of the quantities of interest beyond the time horizon supported by the clinical evidence.Unfortunately, the implications for decisions made on the basis of evidence derived following this practice and the degree of uncertainty surrounding the validity of any assumptions made are often not fully appreciated.The issue is compounded by the absence of methodological guidance concerning the extrapolation of nontime to event outcomes such as costs and utilities. This paper considers current approaches to predict long-term costs and utilities, highlights some of the challenges with the existing methods, and provides recommendations for future applications. It finds that, typically, economic evaluation models employ a simplistic approach to temporal extrapolation of costs and utilities. For instance, their parameters (e.g. mean) are typically assumed to be homogeneous with respect to both time and patients' characteristics. Furthermore, costs and utilities have often been modelled to follow the dynamics of the associated time-to-event outcomes. However, cost and utility estimates may be more nuanced, and it is important to ensure extrapolation is carried out appropriately for these parameters. Key points1. Need to consider how costs and utilities are extrapolated in a cost-effectiveness model 2. There is no guidance on methods for extrapolating costs and utilities 3. Existing applications use various approaches with little justification

Longitudinal assessment of utilities in patients with migraine: an analysis of erenumab randomized controlled trials

“…However, the collection of EQ-5D utilities is not always appropriate or possible in every disease state, so other methods may be used [7, 8]. Limited guidance exists on approaches to extrapolating outcomes such as utilities [10]. Applying existing algorithms is one of the options to derive utilities for health-state estimates when they are not available from the original data set [10, 11].…”

“…Limited guidance exists on approaches to extrapolating outcomes such as utilities [10]. Applying existing algorithms is one of the options to derive utilities for health-state estimates when they are not available from the original data set [10, 11]. Extrapolation methods should, however, consider processes that influence utilities that may not be linked to clinical events (e.g.…”

Costs of screening for prostate cancer: Evidence from the Finnish Randomised Study of Screening for Prostate Cancer after 20-year follow-up using register data

“…Perhaps most importantly, we have attempted to present costs in an appropriate manner for a policy-oriented readership. Unfortunately, no robust method seems to exist to extrapolate the observed cost estimates from those men who have already died, to those who may die from PCa in the coming years [34]. A related limitation is that although potentially declining during the study period, overdiagnosis or overtreatment may still be one contributing factor in our results [35].…”

Lifetime Costs of Surviving Cancer—A Queensland Study (COS-Q): Protocol of a Large Healthcare Data Linkage Study

“…Model input parameters & perspective: Applying cohort simulation techniques, the model will capture health and cost outcomes of patients moving through relevant model health states during their life (disease onset until death). If necessary, results beyond the scope of the administrative data will be extrapolated to the end of the model using appropriate parameter estimation methods [44]. Results from Aim 1 will provide estimates of health service utilization and the proportion of patients' co-payments.…”