Dietary Patterns and Chronic Disease Risk: A Proteomic Framework to Strengthen Causal Inference.

Abstract

The strength of evidence in nutrition research is hampered by the difficulty of maintaining long-term dietary compliance in randomized trials. Despite providing valuable evidence, short-term feeding trials typically rely on a limited set of surrogate outcomes, and observational studies are susceptible to measurement error and residual confounding. We propose a complementary research framework that integrates randomized feeding trials with Mendelian randomization (MR) using proteomic markers to strengthen causal inference in nutrition research. In this approach, feeding trials are used to identify proteins that respond to dietary patterns, and MR analyses are subsequently used to assess whether these diet-responsive proteins are relevant to the development of chronic diseases. In addition to instrumenting individual proteins, we propose using genetic scores to instrument proteomic signatures of dietary patterns in MR analyses. We argue that proteins are particularly informative biomarkers because they reflect multiple biological pathways involved in disease pathophysiology and can often be instrumented with cis-protein quantitative trait loci, which reduces concerns about horizontal pleiotropy compared with other MR approaches. To illustrate the feasibility of several steps in the proposed approach, we highlight studies of the Dietary Approaches to Stop Hypertension diet. This research shows that randomized feeding trials can identify reproducible proteomic changes and that several of these associations can be confirmed in population-based cohort studies. We further note that selected diet-related proteins have been associated with risk of type 2 diabetes in prospective studies and, in some cases, supported by MR and colocalization analyses. We also discuss potential limitations, including incomplete proteomic coverage, measurement error, epitope effects, and challenges in satisfying MR assumptions. In summary, we argue that a cross-design, proteomic-based research strategy can contribute to the triangulation of evidence in nutrition research, generating more specific and robust dietary recommendations for chronic disease prevention.