OBJECTIVES : Although patient preference studies are gaining interest from drug developers, regulatory agencies and reimbursement decision makers, there is lack of standards for study design and reporting. The aim of this literature review is to systematically assess reporting standards and quality of patient preference studies in the cardiovascular domain.

METHODS : Studies were identified from PubMed and Embase databases, screened by title and abstract, and selected on pre-defined inclusion/exclusion criteria, followed by data extraction. Eligible studies were scored on a 5-point scale with a critical appraisal method assessing purpose of study, respondent sampling, explanation of methods, findings reported, and significance testing (PREFS). Studies with a PREFS score ≥3 were reviewed for appropriateness for quantitative synthesis, where preferences on safety and efficacy endpoints were normalized to measure maximum acceptable risk (MAR).

RESULTS : 423 studies were screened and 22 studies met all inclusion/exclusion criteria. Attribute definitions varied between the studies. Seven studies reported cognitive interviews to assess patient understanding of the preference elicitation task. Fourteen studies reported internal consistency testing. The average PREFS score in the 22 studies was 2.5. Nine studies with average PREFS score of 3.25 were selected for MAR synthesis. Seven of these studies included major stroke/major bleeding attributes with varying MAR (2.20-7.62). The MAR also varied across three studies with major stroke/MI attributes, with the highest MAR (7.62; 11.00) in a best worst scaling (BWS) study and the lowest MAR (2.20; 2.52) in two discrete choice experiments (DCE). MAR values were more similar for non-major stroke/major bleeding (0.23-1.07; n=2 studies) and for moderate stroke/major bleeding (1.04-3.17; n=5 studies).

CONCLUSIONS : Patient preference studies in the cardiovascular domain vary in reporting, endpoint (attribute) definitions, and methods used for preference elicitation. Differences in study design and attribute definition may explain some differences in the observed MAR values.