HM4 - How Much Evidence Do We Need Before Implementing Pharmacogenomic Testing In The Clinic?

Abstract

Consensus on the evidence required to recommend clinical pharmacogenomic testing is unclear. A formal assessment of pharmacogenomic evidence levels in relation to other clinical interventions, such as avoiding drug-drug interactions (DDI), may be helpful for policymakers. The objective of this study was to quantitatively compare the evidence levels of two contested drug-attenuating interactions with clopidogrel antiplatelet therapy, and assess the value of obtaining additional evidence to inform clinical practice guidelines. We developed analogous value of information (VOI) decision models for: (1) avoidance of proton pump inhibitors (PPIs) in clopidogrel patients, and (2) pharmacogenomic-guided antiplatelet selection, both versus no intervention. Interaction-specific parameters and model structures were the only dissimilarities. We calculated the expected value of obtaining perfect information (EVPI) per patient, and the expected population value of obtaining additional information through future studies (EVSI). Current evidence for interaction-1 was slightly less uncertain (a 20% probability of making a non-optimal recommendation) than for interaction-2 (23%). The relative risk for cardiovascular death (conferred by concomitant PPI use in interaction-1, and reduced-function CYP2C19 alleles in interaction-2) was the greatest source of uncertainty in both models. The expected value of perfect information for interaction-1 was $139 per patient, compared to $242 per patient for interaction-2. In simulated 10,000-patient clinical trials, the expected population value of future research for interaction-1 was $2 million, versus $110 million for interaction-2. The evidence levels for clopidogrel DDI and pharmacogenomic effects appear to be fairly similar. However, the value of conducting future research on clopidogrel pharmacogenomics is higher because of greater uncertainty about the impact of pharmacogenomic testing on cardiovascular mortality. Our findings imply that evidence-based clinical guidelines for DDI and pharmacogenomic effects should be generally similar with regard to direction and strength of recommendation. This contrasts to some degree with current guidelines and drug labeling.