Automatic pharmacogenomic testing provided on liquid biopsy: Access, prevalence, and integration in a safety net population.

e15137 Background: Pharmacogenomics (PGx) is a rapidly growing tool for precision oncology however remains underutilized in routine oncologic practices, particularly amongst under-insured and minority populations due to the cost and access barriers. Circulating tumor DNA (ctDNA) assays, such as Guardant360, now report on PGx variants alongside somatic alterations, expanding access without requiring separate testing, but it is not clear how and if PGx results are implemented in practice. Methods: We performed a retrospective review of 79 unique oncology patients who underwent Guardant360 testing at a safety-net cancer center. Genes assessed included UGT1A1, CYP2D6, DPYD, TPMT and HLA-B. Metabolizer phenotypes were characterized as normal, intermediate, or poor. Potentially actionable PGx was identified as intermediate or poor metabolizer status. Demographic data along with insurance status was also collected. Under-insured status was defined as Medical Access Program (MAP), Sendero, or uninsured. PGx integration was defined conservatively as documentation acknowledging the PGx result within 3 months of testing. Results: The median age of the cohort was 57 years, and 24 (30.0%) were ≤50 years at testing. Most patients were under-insured (71.2%). At the patient level, 57 of 79 patients (72.2%) had at least one potentially actionable PGx phenotype, and 32 patients had GI malignancies. One actionable phenotype was present in 33 (56%) of patients and two in 13 (17%). The most frequently flagged actionable genes were UGT1A1 (40/79, 51%) and CYP2D6 (24/77, 31%), with fewer actionable variants in TPMT (5/79, 6%) and DPYD (2/79, 3%); no actionable HLA-B variants were observed. Actionable PGx prevalence did not differ by ethnicity (p = 0.793). Despite potentially actionable PGx results, clinical integration and acknowledgment of PGx findings within 3 months was rare. We did not identify documented PGx-guided changes in systemic therapy within the observation window. Conclusions: In this safety-net population of cancer patients, automatic PGx reporting through routine liquid biopsy showed a high prevalence of potentially actionable PGx phenotypes, affecting nearly 3 of 4 patients. The low documentation rate points to an under-recognition of PGx data provided in liquid biopsy testing, even in the 40% of patients with GI malignancies, for whom PGx is most relevant. These findings show that automatic PGx is an under-recognized resource in safety-net oncology.