(120) Distribution of Samples of a New Preventive Migraine Drug and Potential Implications for Validity of Pharmacoepidemiology Studies Using Administrative Data

Publication Number: 794

Background: Pharmacoepidemiology studies often utilize data from administrative claims. As claims are financial records, free medication samples are not captured. We previously identified new users of galcanezumab, a migraine drug with a recommended initial (“loading”) dose of two injections, but this dosing pattern was not observed in the claims for 72% of patients. We speculated that this could be due to drug samples not recorded in claims.

Objectives: We conducted a provider survey to characterize sample pack dissemination of an injectable preventive migraine treatment in the HealthCore Integrated Research Database (HIRD).

Methods: We identified users of galcanezumab in the HIRD, a US administrative claims database, from September 2018 through August 2020. Using these data, we emailed a four-item Qualtrics survey to galcanezumab prescribers (n=601) for whom email addresses were available. Reminders were sent on a weekly basis. The survey was available from October 26 through December 21, 2021.

Results: After 50 undeliverable emails, we reached an estimated 551 providers and received 70 survey responses (response rate: 13%). Of 65 prescribers who completed the survey, 40 (62%) had received sample packs provided by the manufacturer explicitly for loading doses (the manufacturer does not promote the use of samples for anything beyond the loading dose). Of those, 97% responded that they provided sample packs for the loading dose and 56% responded that they provided the sample packs for doses beyond the loading dose. Most providers (62%) indicated that sample pack administration is routinely recorded in the electronic medical record (EMR).

Conclusions: Medication samples were given to patients, including for doses beyond the loading dose. Drug samples not recorded in claims can lead to exposure misclassification in claims-based pharmacoepidemiology studies. Patients who do not fill a prescription after receiving a sample, possibly due to adverse events, intolerability, lack of effectiveness, long-lasting effectiveness, or cost are not identifiable in claims. Use of samples beyond the initial dose will also appear as non-exposure in claims studies, affecting assessment of safety, adherence, and cost. Although medical records may contain data on samples, not all providers indicated that they record samples in their EMR, indicating that even medical records data may not be sufficient to capture all samples. Claims- or medical record-based studies of medications should explore the impact of sample utilization and consider incorporating quantitative bias and other sensitivity analyses into the study protocol.