Body mass index and type 2 diabetes liability in osteomyelitis and cellulitis: A cross-biobank Mendelian randomization study.

Previous Mendelian randomization (MR) studies have linked obesity and type 2 diabetes (T2D) liability to osteomyelitis and skin or soft-tissue infection, but the reproducibility of these associations across independent outcome biobanks remains uncertain. We conducted cross-biobank univariable MR and multivariable MR analyses using European-ancestry genome-wide association study summary statistics. FinnGen Release 13 osteomyelitis (2519 cases and 471,012 controls) was designated as the primary clinical outcome, whereas cellulitis (10,223 cases and 468,418 controls) served as a comparative infection outcome; UK Biobank (UKB)-derived outcomes were used for replication. Exposure instruments containing UKB participants were restricted to FinnGen analyses, while separate non-UKB instruments were used for replication and cross-biobank pooling. Inverse-variance weighting was the primary estimator, with instrument-strength, pleiotropy, directionality, influential-variant, and multivariable conditional-strength analyses used to assess robustness. In FinnGen, each 1-standard-deviation increase in genetically predicted body mass index (BMI) on the source-GWAS standardized scale was associated with osteomyelitis (odds ratio [OR] 1.60, 95% confidence interval [CI] 1.33-1.93) and cellulitis (OR 1.50, 95% CI 1.37-1.63), while T2D liability was associated with osteomyelitis (OR 1.41, 95% CI 1.24-1.61). Using non-UKB instruments, BMI-osteomyelitis estimates were directionally consistent across FinnGen and UKB (pooled OR 1.38, 95% CI 1.13-1.67; I2 = 0%), whereas BMI-cellulitis was positive in both biobanks despite heterogeneity in magnitude (I2 = 78.1%). T2D liability was associated with osteomyelitis in FinnGen but was not supported in UKB univariable MR, with substantial cross-biobank heterogeneity (I2 = 84.8%); the corresponding UKB multivariable MR estimate was nominal. In multivariable analyses, BMI-cellulitis remained supported across datasets, whereas BMI-osteomyelitis estimates attenuated after adjustment for T2D liability. BMI-cellulitis showed the greatest reproducibility. For osteomyelitis, BMI showed a reproducible total association but weaker evidence after adjustment for T2D liability, whereas T2D liability was consistently associated with osteomyelitis in FinnGen but was not robustly replicated in UKB. These findings emphasize the importance of external replication and evidence grading when interpreting metabolic-infection MR associations.