Abstract

The drug discovery and development process is increasing being focused on developing therapies that attempt to recapitulate the phenotype of polymorphisms that are associated with the risk of disease. Mendelian randomization studies play an important role in providing genetic validation for these potential therapeutic targets. Indeed, as new therapies are designed to recapitulate the phenotype of polymorphisms that are associated with the risk of disease, in future it is likely that randomized trials will be designed to recapitulate the naturally randomized genetic evidence. However, demonstration that the phenotype mediated by a polymorphism is causally associated with the risk of disease is merely the first step in accurately anticipating the results of randomized trials designed to evaluate the clinical efficacy these therapies. Important further steps include estimating the quantitative magnitude of the effect of the polymorphism on the risk of disease per unit change in the causal intermediate biomarker; evaluating the effect of the polymorphisms alone and in combination with polymorphisms that mimic the stand of care using factorial Mendelian randomization; estimating the effect of short-term exposure to changes in the causal intermediate biomarker; and identifying populations who are most likely to benefit. Failure to adequately consider these issues can result in unexpected null trial results leading to a discordance between the results of randomized trials and those anticipated by Mendelian randomization studies. We shall illustrate these principles by comparing the results of Mendelian randomization studies with the results of recent randomized trials evaluating statins, ezetimibe, PCSK9 inhibitors, CETP inhibitors and fibrates; and use these principles to propose the optimal trial design to evaluate therapies that lower Lp(a).