Pharmacogenetic testing evaluates how inherited genetic variation can influence an individual’s response to medications, with the stated aim of informing drug selection and dosing to improve efficacy or reduce adverse effects. Targeted testing of a single gene (or a small number of genes, e.g., 2–3 genes for warfarin) is commonly used when a specific drug decision is being considered; by contrast, laboratories have developed Multi-Gene Panels that test five or more genes simultaneously to provide preemptive information about many potential drug–gene interactions across therapeutic areas (definition per policy: multi-gene panels = ≥ five genes).
The policy reviewed multiple types of evidence to assess clinical utility for multigene pharmacogenetic (PGx) testing, including randomized controlled trials (RCTs), systematic reviews and meta-analyses, health-technology assessments, decision-analytic modeling, cohort and real-world observational studies, and clinical utility/ molecular test assessments (e.g., Hayes evaluations). While some RCTs and meta-analyses report early or subgroup improvements (for example, faster response or higher response/remission rates in certain depression trials at 8–12 weeks), other high-quality studies show limited or no durable benefit on primary endpoints or on longer-term outcomes. Trials such as PREPARE (a 12-gene pragmatic study) demonstrated reductions in adverse drug reactions in some populations, and multiple meta-analyses synthesize RCT data showing modest short-term improvements in psychiatric outcomes, but overall findings are mixed and context-dependent.
Important limitations of the evidence base are recurrent across clinical areas: study heterogeneity (varied gene content and proprietary algorithms across marketed panels), frequent lack of blinding of clinicians and/or participants in trials, short durations of follow-up in many studies (commonly 8–12 weeks), limited racial/ethnic diversity in study cohorts, potential conflicts of interest or industry involvement in several publications, and reliance on proprietary decision-support algorithms that differ between tests. These limitations reduce certainty about generalizability and the magnitude of clinical benefit for routine use of broad multigene panels.
For proprietary molecular signature classifiers such as PrismRA (a Scipher Medicine MSRC intended to predict nonresponse to anti‑TNF therapy in rheumatoid arthritis), the policy summarizes observational and modeling data suggesting potential to influence prescribing and improve response in selected cohorts (for example, reported positive predictive values near reported estimates and modeled increases in ACR50 in simulated cohorts). However, available studies are nonrandomized or cohort-based, often nonblinded, limited in size and diversity, and subject to potential test‑developer bias; randomized controlled validation and larger independent prospective studies are lacking, and decision models frequently assume full provider adherence and limited sensitivity analyses.