Prescription Patterns and Relationship to Pharmacogenomics Testing in the Military Health System

Introduction: Clinical utilization of pharmacogenomics (PGx) testing is highly institutionally dependent, and little information is known about provider practices of PGx testing in the Military Health System (MHS). In this study, we aimed to characterize Clinical Pharmacogenetics Implementation Consortium (CPIC) actionable prescription (Rx) patterns and their temporal relationship with PGx testing in the MHS. Methods: Using data from the Military Health System Management Analysis and Reporting Tool (M2) database, this retrospective cohort study included all patients receiving at least one PGx test and at least one CPIC actionable Rx from January 2015 to August 2020 (845 patients, 1,471 PGx, 7,725 index CPIC actionable Rxs). Rx patterns and temporal relationships with PGx testing were characterized via descriptive statistics. Binomial regression was used to determine which patient and provider characteristics were associated with a patient receiving a PGx test within 30 days of an index Rx. Results: Patients had a median of 9 index CPIC actionable Rx's (range 1-26). Pain medications were most commonly prescribed (N=794, 94% patients with at least 1 Rx). However, pain medication had the lowest Rx-PGx match rate (40%) compared to an average of 62% Rx-PGx match rate for all CPIC drugs. Antidepressants were also commonly prescribed (N=668, 79.1% patients with at least 1 Rx), and antidepressants had the highest Rx-PGx match rate of 86.7%. A minority of providers (20%, N=249) ordered the majority of PGx tests (86.1%, N=1,266) and only 8.3% of PGx tests (N=398) matched to a CPIC actionable drug within 30days of the test (defined by Rxs ordered within 30days before or after the PGx test). However, approximately 39.8% of patients (N=317) had at least one drug match to a PGx test within 30 days. The largest predictor of whether a patient received a PGx test within 30 days of any index Rx was whether or not a specific psychiatry provider ordered the PGx test (odds ratio; OR 3.7, 95% CI 2.13-6.54, P<0.001). Neither the CPIC level of evidence nor FDA PGx actionable or informative labels had a significant effect on PGx test timing. Conclusions: PGx testing was generally limited to high Rx-drug users and was found to be an under-utilized resource. PGx testing did not typically follow CPIC guidelines. Implementing PGx testing protocols, simplifying PGx test-ordering by incorporating at minimum CYP2D6, CYP2C19, and CYP2C9 into PGx-testing panels, and unifying providers' PGx knowledgebase in the MHS are feasible and would improve the clinical utilization of PGx tests in the MHS.