Use of Individual Participant Data to Evaluate Cost-Effectiveness Across Heterogeneous Population
Posted: 14 Jun 2007
Date Written: June2007
Abstract
Cost-effectiveness analyses based on clinical trials often report a single cost-effectiveness estimate and ignore the potential for variation across the study participants. Cost-effectiveness decision analytic models, on the other hand, frequently explicitly model heterogeneity, but often based solely on published summary data and structured modelling assumptions. The objectives of this presentation are to illustrate the potential value of individual participant data (IPD) when dealing with heterogeneity in cost-effectiveness analysis, and to identify appropriate methods of evaluating cost-effectiveness across heterogeneous population.
IPD from the 20536-large Heart Protection Study are employed to explore heterogeneity in cost-effectiveness of statins for cardiovascular disease prevention. The IPD were used to develop and validate a decision analytic lifetime model with particular focus on heterogeneity and uncertainty of cost-effectiveness. A series of risk and cost equations estimate disease progression and healthcare costs with and without the intervention for different patient profiles. Cost-effectiveness is evaluated for each trial participant and summarized across subcategories defined through the main determinants of heterogeneity. Parameter uncertainty is evaluated through bootstrapping the IPD. We show that using IPD permits development and validation of the model in subcategories of participant and improves the efficiency of cost-effectiveness estimates.
Baseline vascular disease risk is the main determinant of statins' cost-effectiveness, with age and gender having an additional impact when lifetime cost-effectiveness is studied. The event risk equations, estimated using IPD, showed the importance of capturing the interdependence between non-fatal disease events and subsequent survival. For example, the HPS data showed that experiencing a heart attack or stroke increases the hazard of a subsequent such event more than 3-fold in the first year (hazard ratio 3.2 (95% CI 2.9 to 3.5)), about 2-fold in the second year (hazard ratio 2.0 (1.7 to 2.3)) and by about 70% during each subsequent year (hazard ratio 1.7 (1.4 to 2.0). Using IPD to model cost-effectiveness parameters also improves the efficiency of the estimates, narrowing the confidence intervals around incremental costs, major vascular events avoided and life years gained at five years by about 30% to 40% compared to censoring adjusted estimates of trial data. Significant improvements in efficiency are observed also across the five patient subgroups studied. Appropriate evaluation of parameter uncertainty by preserving the interdependence between the different elements of the model and impact of heterogeneity are also illustrated.
We argue that IPD has advantages in informing the sources and structure of heterogeneity of cost-effectiveness parameters across this and other populations. IPD also offers substantial advantages in validating cost-effectiveness models, and in evaluating cost-effectiveness uncertainty for subcategories of participant. In summary, exploring the structure of heterogeneity at the IPD level and using the results to inform subsequent cost-effectiveness analysis seems to be the most promising approach towards subgroup analysis.
Keywords: cost-effectiveness, decision models, heterogeneity, individual participant data, clinical trials
Suggested Citation: Suggested Citation