Establishing causal relationships between dietary exposures and disease outcomes is inherently challenging. Long-term randomized controlled trials (RCTs) with clinical disease endpoints are often impractical or infeasible. Short-term dietary RCTs can, however, detect changes in intermediate molecular traits that may lie on the causal pathway between diet and disease.
Mendelian randomisation (MR) is an epidemiological approach that uses genetic variants as proxies for modifiable exposures to estimate the effects of lifelong differences in those exposures on disease risk. MR offers a method to infer causality using genetic instruments but has known limitations when applied to complex exposures such as whole diets.
Genetic variants used in MR typically index biological mechanisms (for example, nutrient metabolism, receptors, or biomarker levels) rather than specific patterns of food intake or entire dietary regimens. As a result, MR on its own can have limited utility for evaluating complex or composite dietary exposures because instruments may not capture the behavioural and compositional complexity of diets.
This limitation motivates combining MR with evidence from dietary RCTs: RCTs can identify molecular changes produced by an intervention, and MR can assess whether those molecular changes are linked causally to long-term disease risk.
The authors propose a two-step integrative framework to estimate potential lifetime effects of dietary interventions:
Step 1: Use data from dietary RCTs to identify molecular or intermediate traits (for example, circulating proteins) that are altered by the intervention.
Step 2: Use Mendelian randomisation to evaluate whether those RCT-identified traits are associated with long-term disease risk, thereby providing evidence for a causal link between diet-induced trait changes and disease outcomes over the life course.
This framework aims to bridge short-term mechanistic evidence from trials with MR-based inference about long-term risk.
The framework was demonstrated using data from the Diabetes Remission Clinical Trial (DiRECT). DiRECT measured circulating proteins before and after the dietary intervention. Of 4,601 proteins measured, 216 proteins were reported to change following the intervention. These RCT-identified protein changes constitute the candidate intermediate traits taken forward into MR analyses.
In the second step, the authors used MR to test whether proteins altered by the DiRECT intervention were associated with diabetes risk over the long term. Using the set of proteins identified in step 1, MR analyses identified 10 proteins for which there was an association with diabetes risk.
The combination of RCT-derived molecular changes and MR-derived disease associations provides a structured approach to infer whether diet-induced molecular alterations may influence lifetime disease risk.
The study compared MR estimates of protein–diabetes associations with the observed protein–remission associations from DiRECT. The comparison showed broad agreement: the reported correlation between the MR estimates and the observed trial associations was approximately r ≈ −0.645, with R2 = 0.416. The authors interpret this concordance as supportive evidence for the proposed two-step framework.
The authors state that the code used for analyses is publicly available on GitHub: https://github.com/bennydeslandes/twostepframework. The underlying study data are not publicly available according to the source.
Ethical approval for the research was obtained from the West of Scotland Research Ethics Committee (reference number: 13/WS/0314). The preprint was posted on September 04, 2026.
The two-step RCT–MR framework offers a pragmatic way to infer potential lifetime effects of dietary interventions by linking short-term mechanistic changes measured in trials to long-term disease risk using genetic instruments. Strengths of the approach include leveraging direct trial measures of molecular change and independent MR evidence on disease risk.
Limitations are acknowledged implicitly by the rationale for the framework: MR instruments are often better suited to biological mechanisms than to whole-diet exposures, and the approach relies on the availability of robust RCT-measured intermediates and valid genetic instruments for those intermediates. The authors note the framework is most applicable when RCTs identify measurable, biologically plausible intermediate traits that can be instrumented in MR analyses.
Integrating dietary RCTs with Mendelian randomisation in a two-step framework can help bridge the gap between short-term intervention effects and potential lifetime disease outcomes. The DiRECT example — where 216 proteins changed after intervention and 10 proteins were associated with diabetes risk by MR, with broad concordance between MR and trial associations — illustrates the approach. The authors provide code for replication but report that underlying study data are not publicly available.