Lactase Persistence and Body Mass Index: The Contribution of Mendelian Randomization

Nutritional epidemiology is one of the most challenging fields in epidemiology with respect to causal inference. Eating or not eating a particular food often occurs within a much wider dietary habit pattern, thus making it difficult to statistically disentangle the effects of individual foods. Dietary intake is also difficult to measure, and in large population-based studies, self-reported information is often the only feasible option. Intake may also vary over time, raising the need for repeated measurements to investigate the effects of long-term (and changing) dietary patterns. Longitudinal studies are also required to overcome the possibility of reverse causation, given that disease status often has a major influence on dietary behaviors. Diet is also often correlated with lifestyle and socioeconomic factors, which may be unknown or difficult to measure; therefore, the possibility of residual confounding is often nonnegligible.Randomized controlled trials could in principle be used to overcome the limitations of observational studies in nutritional epidemiology. However, it is difficult to achieve high adherence to behavioral interventions, especially in healthy individuals. Moreover, extended follow-up time is required because the health effects of a dietary intervention are unlikely to be acute. The participants of trials are often highly selected subgroups of the population, which may limit the generalizability of the findings. Finally, it is difficult (and often impossible) to blind participants of a behavioral intervention regarding their intervention status.In this issue of Clinical Chemistry , the Mendelian Randomization of Dairy Consumption Working Group (1) attempted to overcome the above limitations and estimate the causal effect of dairy intake on body mass index (BMI)4 using Mendelian randomization (MR). MR uses 1 or more genetic variants robustly associated with modifiable exposures as instrumental variables (IVs) to assess the causal effect of such exposures on a given outcome. Valid causal inference from MR …