Emerging data from randomized trials have shown that a growing number of lipid-lowering drugs appear to influence glycemic control in addition to their hypolipidemic effects. Statins have been shown conclusively in large meta-analyses of randomized trials to increase, albeit modestly, the risk of new-onset type 2 diabetes (T2D) in an apparently dose-dependent manner (1–3). In the Investigation of Lipid Level Management to Understand Its Impact in Atherosclerotic Events (ILLUMINATE) trial, treatment with torcetrapib, a CETP inhibitor that raised HDL cholesterol (HDL-C) concentrations, improved glycemic control (4), but in the Treatment of HDL to Reduce the Incidence of Vascular Events (HPS2-THRIVE) study, niacin treatment, which also raised HDL-C, raised blood glucose and T2D risk (5). In a recent intriguing observational study, notwithstanding the inevitable limitations of its design, patients with familial hypercholesterolemia (FH) appeared less likely to develop T2D (6). Evidence is therefore emerging for a relationship between circulating lipid concentrations and/or their therapeutic/genetic modulators and alterations in glycemia. The mechanisms underlying such relationships remain uncertain and are the subject of much research, not least because they may reveal new drug targets for diabetes or help in the mitigation of dysglycemia risks with specific therapies. In this issue of Diabetes, Fall et al. (7) present a comprehensive genetic investigation of the relationships between circulating lipids and dysglycemia.
The investigators used a rigorous approach (using genes as so-called instruments) to dissect the lipid-glycemia relationship. In doing so, they report an apparent significant association between gene variants determining higher LDL cholesterol (LDL-C) levels and lower T2D risk but a less clear association between genetically determined levels of HDL-C and triglycerides and T2D. They do, however, caution that even their careful approach has limitations. Their methods raise important issues for the design and interpretation of Mendelian randomization studies (which make use of genetic instruments for causal inference) in complex metabolic traits. In constructing the allele scores that form their genetic instruments for plasma lipid concentrations, the authors excluded single nucleotide polymorphisms that associate with measures of adiposity on the grounds that adiposity is likely to confound the lipid-T2D relationship. This assumption stems from overwhelming evidence for obesity being causally upstream to altered lipids and higher diabetes risk (8), whereas support is currently lacking for the reverse pathways (i.e., how altered lipid levels could lead to changes in BMI). That noted, variants at the HMGCR locus (the gene encoding the intended target of statins) appear to associate with lower LDL-C concentrations, higher BMI, and higher T2D risk (1). The same pattern was seen in randomized statin trials, and thus to what extent the minor statin-associated weight increase is a consequence of an on-target effect of HMGCR inhibition is now an interesting question. Further, more powerful genotypic studies are clearly needed to better disentangle the causal routes in the lipid-BMI-diabetes association, but it is likely such studies will be even more complex than the current study by Fall et al. (7).
It is clear from the foregoing and from other aspects of the article by Fall et al. (7) that substantial challenges exist in applying Mendelian randomization to complex metabolic traits, such as blood lipids. Many protein biomarkers, such as C-reactive protein (CRP), are the products of single genes (e.g., CRP is encoded by the CRP gene on chromosome 1). Although other genetic loci influence the concentration of CRP, variants in the CRP gene can be used as the most specific instruments for CRP levels (9). For blood lipids, however, there are no such single specific genes, and their concentrations are influenced by a wide range of loci. These variants have been incorporated into so-called allele scores for use as instruments for their levels in Mendelian randomization experiments (10). Allele scores appear to provide more powerful genetic instruments for complex traits than single variants as the combination of variants has a greater effect on lipid levels and has been used effectively to investigate causal links between lipid moieties and cardiovascular disease (11,12). In their analysis, Fall et al. minimized potential confounding between lipid fractions by constructing allele scores with maximal specificity for their index trait. Care is required when considering associations of variants in the score with nonlipid traits and the possible placement of those secondary traits on the causal pathway. Inclusion or exclusion of variants on the grounds of potential confounding through other biological pathways may bias the observed estimate if those suspected confounders are, in reality, effect modifiers. That said, the construction of allele scores is a challenging and imperfect science, and the authors present a strong and well-reasoned case for their approach. They do, nevertheless, stress that certain uncertainties (such as so-called pleiotropy, whereby a gene has other, as yet unrecognized, effects) cannot currently be fully discounted.
Returning to the clinical context of the findings of Fall et al. (7), it is important to remember the prevailing and overwhelming evidence for the benefit of lipid-lowering therapy for cardiovascular disease prevention in patients with and without diabetes (13,14). Despite the now-established modest T2D risk resulting from statin therapy and our recent report of a very modest rise in body weight (1), statins remain powerful tools for the prevention of serious cardiovascular harm. Nevertheless, the mechanisms potentially linking lipid-modifying drugs with glycemic control are of increasing interest, particularly with the emergence of novel lipid-lowering agents (Fig. 1). The growing contribution of big data and -omics technologies that address this issue may help to provide answers and may, in time, even suggest novel therapeutic targets for diabetes. We await further developments in this area with interest.
See accompanying article, p. 2676.
Duality of Interest. N.S. has consulted for Amgen, Sanofi, Eli Lilly, and Merck Sharp & Dohme. D.I.S. has consulted for Pfizer. No other potential conflicts of interest relevant to this article were reported.