Mendelian randomisation: urate and BMI

The research record

This is the project's file for this landmark, as written and checked. Sources are listed at the end; anything not yet confirmed is marked "Not yet verified".

Who

Kelly Simon, Alberto Ascherio, Michael Schwarzschild and the Parkinson Study Group (2014); Alastair Noyce (BMI, 2017), Demis Kia and Nicholas Wood (UCL, 2018), Rowan Coneys, Nicholas Wood and colleagues (UCL, 2021); Camilla Kobylecki, Børge Nordestgaard and Shoaib Afzal (Copenhagen, 2018).

What happened

Observational studies had shown that people with more urate (an antioxidant in blood) get Parkinson's less and progress more slowly. Mendelian randomisation (MR) uses gene variants that nudge a trait for life, avoiding confounding.

• 2014: In 808 patients from two trials (DATATOP and PRECEPT), people carrying four or more urate-lowering SLC2A9 alleles had urate 0.69 mg/dl lower, and progression to needing dopamine drugs rose with allele score (hazard ratio 1.16, 95% CI 1.00–1.35, p = 0.056, borderline); the authors read it as supporting causation.
• 2018: In more than 102,000 Danes (285 incident cases), higher urate protected observationally (50 μmol/L higher: HR 0.85) but not genetically (odds ratio 1.20, 0.85–1.71). A separate two-sample MR (Kia et al.) with 31 urate loci and 13,708 cases and 95,282 controls found OR 1.03 (0.88–1.20) per standard deviation.
• 2021: A two-sample MR using newer GWAS data found no causal link between urate and risk, age at onset or progression.
• BMI (2017): Genetically higher BMI (77 variants) predicted lower risk, OR 0.82 per 5 kg/m² (0.69–0.98), but survival bias may explain part.

Why it moved us toward a cure

It tested whether the urate association (landmark 131) is cause and effect. It showed a cheap way to vet a drug target before an expensive trial. The inosine urate-raising trial later failed (landmark 66), consistent with the MR doubts. The same approach supported the smoking link (landmark 82).

Outcome

Dead end for urate; open for BMI. MR rests on assumptions (variants affecting only the trait), and early MR studies conflicted.

Sources

• Simon et al., Ann Neurol 2014;76:862–868, DOI 10.1002/ana.24281
• Kobylecki et al., Ann Neurol 2018;84:178–190, DOI 10.1002/ana.25292
• Kia et al., Ann Neurol 2018;84:191–199, PMID 30014513
• Coneys et al., Mov Disord 2021;36:2182–2187, DOI 10.1002/mds.28662, PMID 34056740
• Noyce et al., PLoS Med 2017;14(6):e1002314, PMID 28609445
• Not yet verified: causal evidence for other traits (cholesterol, inflammation).