A novel two-step analytical framework combining short-term dietary RCT data with Mendelian randomisation (MR) can estimate the likely lifetime disease impact of dietary interventions. Applied to the DiRECT trial — a well-known low-calorie dietary intervention for type 2 diabetes remission — the approach identified 216 of 4,601 circulating proteins altered by the intervention. MR analysis then linked 10 of those proteins to long-term diabetes risk, with protein-remission associations correlating at r ≈ −0.645 (R² = 0.416) between the two methods, suggesting meaningful convergent validity.
Nutrition research has long struggled with a fundamental paradox: RCTs are gold-standard for causality but rarely run long enough to capture disease endpoints, while observational studies capture long-term outcomes but confound easily. MR sidesteps confounding by using genetic variants as lifelong proxies for exposures, but genetics poorly captures complex dietary patterns. This framework elegantly bridges that gap by letting RCTs define the molecular fingerprint of a diet, and MR determine whether those fingerprints causally affect disease.
The implications are significant: researchers could in principle screen dietary interventions for long-term disease relevance using only short-term trial biomarker data, dramatically reducing the cost and timeline of nutrition science. Limitations include the indirect causal chain across two methodological steps, potential MR assumption violations, and the single-trial demonstration. As a preprint not yet peer-reviewed, the framework's generalizability across dietary interventions and disease outcomes requires independent validation before clinical application.
