PUBLISHED ECONOMIC EVIDENCE TO SUPPORT PERSONALIZED MEDICINE DECISION-MAKING: THE EXAMPLE OF HER2-POSITIVE EARLY BREAST CANCER

Purpose: We evaluated how decision-analytic models of trastuzumab and human epidermal growth factor receptor-2 (HER2) , an established example of personalized medicine, provided evidence to  decision-makers  by considering relevance to the local setting and representation of decision uncertainty in the context of international economic evaluation guidelines. 

Method: A systematic review of published economic evaluations considering trastuzumab treatment or HER2 testing in early-stage breast cancer (ESBC) was conducted.  Electronic citations indexed in Biosis, Cochrane, CRD, EconLit, EMBASE, HEED, MEDLINE and PubMed published up to October 2008 in English were considered.  Included articles were reviewed for [1] relevance to the local setting (evidence of effectiveness, local data source use, model calibration), [2] sensitivity analysis methods (scenario analyses, univariate, multivariate and probabilistic sensitivity analyses [SA]) and [3] representation of decision uncertainty (e.g., cost-effectiveness acceptability curve [CEAC]).

Result: 958 citations were identified; nine treatment evaluations, two testing evaluations and one test-treat evaluation were included.  All studies used real-world or local clinical practice data to estimate costs.  Drug efficacy was not estimated from local or real-world data (n=0). Real-world HER2 test parameters and utilities were each included in one study. One model reported calibration. Univariate (n=9) was more common than probabilistic SA (n=7).  Decision uncertainty was quantified with CEACs in less than half of all probabilistic analyses (n=3).

Conclusion: Guidelines for economic evaluation in Australia, Canada, Denmark, France, the United Kingdom and the United States require deterministic univariate SA at a minimum and request probabilistic SA to assess joint parameter uncertainty.  Our review reveals that published economic evaluations of trastuzumab and HER2 testing in ESBC meet minimum requirements, but fall short in assessing stochastic parameter uncertainty.  Moreover, few studies provide CEACs to illustrate decision uncertainty.    Although the availability of real-world treatment effectiveness of trastuzumab in ESBC is limited, other variables such as test accuracy and health-related quality-of-life rarely reflected the local setting.  Local relevance is greatly influenced by these parameters, as well as model calibration.  This review suggests that the evidence base to support personalized medicine decision-making can be improved by including real-world factors and improving representation of stochastic parameter and decision uncertainty.