Advancing regulatory science and assessment of FDA REMS programs: A mixed-methods evaluation examining physician survey response

Abstract

Purpose: Food and Drug Administration’s (FDA) Draft Guidance for Industry on pharmaceutical REMS (Risk Evaluation and Mitigation Strategies) assessment and survey methodology highlights physician knowledge–attitudes–behaviors (KAB) surveys as regulatory science tools. This mixed-methods evaluation advances regulatory science and the assessment of FDA REMS programs when using physician surveys. We: (1) reviewed published physician survey response rates; and (2) assessed response bias in a simulation study of secondary survey data using different accrual cut-off strategies. Methods: A systematic literature review was conducted of US physician surveys (2000–2014) on pharmaceutical use (n = 75). Kruskal–Wallis tests were used to examine the relationships between response rates and survey design characteristics. The simulation was conducted using secondary data from a population-based physician KAB survey on diabetes risk management with antipsychotic use in Missouri Medicaid (n = 973 accrued over 30 weeks). Survey item responses were compared using Pearson’s chi-square tests for two faster completion simulations: Fixed Sample (n = 300) and Fixed Time (8 weeks). Results: Survey response rates ranged from 7% to 100% (median = 48%, IQR = 34%–68%). Surveys of targeted populations and surveys using member lists were associated with higher response rates (p = 0.02). In the simulation, 9 of 20 (45%) KAB items, including diabetes screening advocacy, differed significantly using the smaller Fixed Sample strategy (achieved in 12 days) versus full accrual. Fewer response differences were found using the Fixed Time strategy (2 of 20 [10%] items). Conclusions: Published data on physician surveys report low response rates with most associated with the sample source selected. FDA REMS assessments should include formal evaluation of survey accrual and response bias.