Gilsenan A, Hollis K, Zografos L, Calingaert B, Andrews E. Examination of respondent bias and patient characteristics in multiple REMS assessment surveys. Poster presented at the 27th ICPE International Conference on Pharmacoepidemiology & Therapeutic Risk Management; August 18, 2011. Chicago, IL.

BACKGROUND: Draft FDA guidance requires Risk Evaluation and Mitigation Strategies (REMS) that include a medication guide to be assessed through a survey of patients’ understanding of the serious risks of the drug. Such surveys should be designed to minimize selection and response bias, and allow analysis of characteristics associated with response/nonresponse. We conducted three surveys using data from a large nationwide pharmacy database to identify eligible participants. Invitations were mailed from the pharmacy headquarters, thereby minimizing potential for an intervention effect. Participants could complete the brief survey by phone or Web.

OBJECTIVES: Determine whether the survey approach for patient accrual, including recruitment through a national pharmacy chain and bimodal survey, for three REMS patient surveys, was successful at enrolling a diverse population of respondents who were similar to the targeted populations.

METHODS: Characteristics of patients invited to participate in a surveys for three different drugs treating three types of conditions (acute: n¼150: intermittent: n¼208: or chronic: n¼200) were examined for response bias. The age/gender distribution of respondents was qualitatively compared with the expected age/gender distribution of patients using each of the three drugs.

RESULTS: Approximately 50% to 66% of respondents opted to complete the survey by phone with older respondents more likely to select this mode. Respondents were geographically distributed across the US and were educationally diverse (27%–38% had high school education). For each survey, characteristics of respondents and nonrespondents were generally similar—% female ( 8% difference), % new prescription (<5% difference), and payer type distribution ( 5% difference)—although respondents were slightly younger than nonrespondents. Age and sex distributions were consistent with what would be expected for the target population for each type of drug.

CONCLUSIONS: Identifying and recruiting participants through pharmacies and offering both Web and phone survey options yielded samples for each survey that closely resembled the expected target populations with no evidence of respondent bias.