Real-world evidence for decisions in diabetes

Randomised controlled trials underpin the evidence base for diabetes interventions but have important limitations, including being conducted in selective populations, with restricted follow-up times and treatment pathways that differ from routine clinical practice. Real-world data (RWD) offer an opportunity to address questions beyond the scope of trials, including the comparative effectiveness of interventions, assessment of longer-term outcomes, and treatment effects in under-represented populations. However, the credibility of real-world evidence (RWE) varies widely due to data quality and methodological challenges. The REDDIE (Real-world Evidence for Decisions in Diabetes) programme examined how RWD can generate robust, decision-relevant evidence for diabetes by analysing multiple national registries, evaluating regulatory expectations and conducting target trial emulations of landmark studies. Evidence from REDDIE and the wider literature demonstrates that well-designed observational analyses can replicate major trial findings and extend evidence to broader populations. However, credibility critically depends on careful alignment between the research question, database fitness, study design and analytical methods. Target trial emulation and modern causal inference approaches can mitigate key biases but cannot overcome fundamental limitations of routine data, including unmeasured confounding and incomplete variable capture. RWE and RCTs should be viewed as complementary sources of evidence. We outline a practical framework for the design, analysis and interpretation of RWE studies in diabetes to support more robust and decision-relevant use of RWD by regulators, health technology assessment agencies and clinicians.