(A-185) Evaluating Sequence Symmetry Analysis Across Six European Healthcare Databases

Background: Sequence Symmetry Analysis (SSA) uses real-world data to assess temporal patterns between drug exposure and potential adverse drug events, which may be recorded as conditions or drug initiations. The validity of SSA can be tested by exploring its ability to correctly identify known positive or negative associations between exposures and outcomes.

Objectives: To apply SSA across 6 healthcare databases across Europe using a set of known positive and negative controls.

Methods: 23 positive and 10 negative control outcomes were identified from the literature (e.g., positive: ACE Inhibitors to cough; negative: amiodarone to allopurinol). We included individuals from CPRD GOLD (UK) and THIN® databases (UK, Romania, Italy, Spain, & Belgium) who experienced events of interest between 2010 and 2021. Patients who received the exposure medicine and had a recorded event of interest within one year were included (i.e., initiation gap). Adjusted sequence ratios (aSRs) with 95% confidence intervals (CIs) were calculated, with values >1 indicating positive (i.e., potential) signals. Sensitivity analyses assessed various initiation gaps and prior data availability settings. Results with less than 50 counts were not reported. We further assessed the consistency of the results for each control by counting the number of databases with positive, negative, and no signals.

Results: 14 out of 38 controls showed consistent results in at least 4 out of 6 databases, 12 of which were positive and 2 were negative. For example, amiodarone to methimazole (a positive control) showed positive signals in only 4 THIN® databases (Romania, Italy, Spain, and Belgium). Amiodarone to allopurinol showed no signals in all databases except Italy, where the results indicated a positive signal (1.198 [95% CI: 1.071, 1.340]). 5 out of 14 controls were consistent across all databases, and all were positive controls. ACE Inhibitors to cough showed positive signals in all databases, with the aSRs ranging from 1.341 [1.273, 1.414] in Belgium to 1.936 [1.803, 2.081] in Spain. With a requirement of 365-day washout before both exposures, 17 of 38 controls were consistent in at least 4 of 6 databases (15 positive, 2 negative), including 5 consistent across all databases (4 positive, 1 negative). When initiation gap was further changed to between 0 and 730 days, 6 controls were consistent across all, and 5 of which were positive.

Conclusions: Our findings show SSA results were largely consistent across different European databases, whilst differences could originate from data sources and prescribing practices. Change of settings impacted the consistency, thus careful consideration is essential to ensure they align with the specific characteristics of the medical domain of the study.