**Background:** Non-alcoholic fatty liver disease (NAFLD) affects approximately 25% of the global population and encompasses a spectrum from simple steatosis to cirrhosis and hepatocellular carcinoma. Observational studies have consistently reported inverse associations between serum 25-hydroxyvitamin D (25(OH)D) levels and NAFLD risk and severity, and preclinical data suggest anti-inflammatory, antifibrotic, and insulin-sensitising properties of vitamin D. However, results from intervention trials have been conflicting, and only two prior Mendelian randomisation (MR) studies—one in a Chinese population and one meta-analysis of European cohorts—have examined causality, yielding contradictory conclusions. This study aimed to clarify the causal relationship using a larger, more comprehensively defined NAFLD cohort and the most extensive vitamin D GWAS meta-analysis available.
**Methods:** The authors conducted a two-sample bidirectional MR analysis. For the vitamin D→NAFLD direction, six SNPs (rs3755967, rs12785878, rs10741657, rs17216707, rs10745742, rs8018720) associated with serum 25(OH)D at genome-wide significance (p < 5×10⁻⁸) were selected from the SUNLIGHT Consortium meta-analysis (n = 79,366 individuals from 31 cohorts of European ancestry). For the NAFLD→vitamin D direction, the authors performed two new genome-wide association studies (GWASs) of NAFLD in the UK Biobank (UKBB) using ICD10 codes K75.8 (NASH) and K76.0 (NAFLD). The primary GWAS included 2757 cases and 460,161 controls; a sensitivity GWAS excluded participants with other liver diseases, yielding 1747 cases and 448,282 controls. SNPs were selected if they reached p < 5×10⁻⁸ in the UKBB GWAS or were previously associated with NAFLD (from Speliotes et al., n = 7176) with effects in the same direction and p < 1×10⁻⁵ in UKBB. Five SNPs (rs738408, rs9479542, rs1260326, rs10401969, rs17321515) were used in the primary NAFLD→vitamin D analysis, and four (rs738408, rs1260326, rs4351435, rs10149275) in the sensitivity analysis. Inverse-variance weighted (IVW) random-effects MR was the primary method, with sensitivity analyses using MR Egger, weighted median, simple mode, and weighted mode. Heterogeneity was assessed with Cochran's Q, and horizontal pleiotropy with MR-Egger intercept and MR-PRESSO.
**Key Results:** The IVW-random effects analysis showed no causal effect of serum 25(OH)D levels on NAFLD risk (OR = 0.95 per SD increase, 95% CI: 0.76–1.18, p = 0.641 for primary; OR = 1.04, 95% CI: 0.79–1.37, p = 0.786 for sensitivity). All sensitivity methods yielded consistent null results. For the reverse direction, genetically predicted NAFLD risk showed no causal effect on low serum 25(OH)D levels (OR = 1.00, 95% CI: 0.99–1.02, p = 0.665 for primary; OR = 1.00, 95% CI: 0.99–1.01, p = 0.689 for sensitivity). No horizontal pleiotropy was detected (P_intercept = 0.47 for NAFLD primary, P_intercept = 0.86 for NAFLD sensitivity). Some heterogeneity was observed in the NAFLD→vitamin D analysis (I² = 63.42% for IVW), but MR-PRESSO identified no outliers. A sensitivity analysis restricted to three SNPs previously identified by Speliotes et al. also showed no association (OR = 1.01, 95% CI: 0.99–1.02, p = 0.563).
**Clinical Implications:** This study provides robust evidence against a causal relationship between vitamin D status and NAFLD risk in European-ancestry populations, contradicting prior observational findings and one previous MR meta-analysis. The null results suggest that vitamin D supplementation is unlikely to reduce NAFLD risk, aligning with the weak and inconsistent evidence from randomised controlled trials. The authors emphasise that differences in NAFLD case definition—particularly the inclusion of both NASH (K75.8) and NAFLD (K76.0) ICD codes versus K76.0 alone—may explain discordance with prior MR studies. The study's strengths include the use of the largest vitamin D GWAS to date, two independent European-ancestry populations (avoiding sample overlap), and comprehensive sensitivity analyses. Limitations include the use of summary-level data with different covariate adjustments, potential underestimation of NAFLD prevalence from ICD codes in UKBB, and restriction to European ancestry, limiting generalisability. The authors conclude that improving the accuracy of population-based NAFLD prevalence data and conducting future meta-analyses across diverse populations will be essential for advancing understanding of NAFLD epidemiology.