Mendelian randomisation for rheumatology: beyond hype—what it's good for, what it can't do, and how to read it critically

Mendelian randomisation (MR) has become abundant in the literature, with variation in quality and frequent overinterpretation of causality. This creates a problem for clinical readers, reviewers, and editors: some MR studies can sharpen causal thinking, prioritise drug targets, and challenge misleading observational claims, whereas others are little more than automated exposure-outcome scans with causal claims disproportionate to the evidence. MR can strengthen causal inference when randomised trials are impractical and conventional observational studies are vulnerable to confounding, reverse causation, or selection bias. In rheumatology, credible MR can contribute to questions about disease aetiology, modifiable risk factors, therapeutic target validation, adverse-effect anticipation, and phenotype validation. However, its interpretation depends on whether the exposure is plausibly instrumentable, whether the genetic instruments are biologically defensible, whether assumptions are interrogated in ways appropriate to the design, and whether findings are triangulated with clinical, observational, experimental, and mechanistic evidence. Instead of recapitulating all methodological issues of MR, this review aims to help rheumatologists distinguish robust MR from weak or overinterpreted analyses quickly. We provide an accessible framework for reading and triaging MR studies in rheumatology. Papers that use poorly justified instruments, treat medication use as drug-target evidence, interpret genetic liability as diagnosis, rely on mechanical sensitivity analyses, ignore prior evidence or ask no clinically meaningful question can often be passed over by readers. The goal is not to discourage MR in rheumatology, but to raise the standard; useful MR should clarify causal reasoning rather than simply generate another statistically significant association.