Purpose: Individuals with acromegaly (characterised by elevated circulating IGF-1) are at an increased risk of secondary osteoarthritis (OA), but epidemiological evidence for an association between serum insulin-like growth factor-1 (IGF-1) and OA in the general population is conflicting. We determined the observational associations between serum IGF-1 and hospital-diagnosed (HD) hand, hip and knee and hand OA in UK Biobank (UKBB). We performed Mendelian Randomization (MR) to determine if observed associations represent a true causal effect. Methods: Serum IGF-1 was assessed by chemiluminescent immunoassay. HD hip, knee and hand OA were determined using linked hospital inpatient records. Observational associations between serum IGF-1 and HD OA variables were determined by multivariable logistic regression, adjusting for age, sex, ethnicity, oestrogen use and BMI. We performed one-sample (1SMR) using two-stage least-squares regression, with an unweighted allele score generated from the eight single nucleotide polymorphisms (SNPs) robustly associated with IGF-1 in the largest genome-wide association study (GWAS) meta-analysis, as an instrument. MR is a form of instrumental variable regression which uses genetic variants robustly related to the exposure as instruments, providing an estimate of the causal effect which is unbiased by confounding and reverse causality. Multivariable MR (MVMR) determined the BMI-independent effect of IGF-1, by including BMI as an additional exposure, instrumented by an unweighted BMI allele score generated from 69 independent loci. MR analyses were adjusted for sex, genotyping chip and 10 principal components (PCs) to account for population stratification. Results: 421,572 individuals had complete data for the observational analyses, of whom 332,092 (79%) had genetic data, were unrelated and of European ancestry, suitable for MR. The mean (SD) ages of the observational and MR populations were 56.4 (8.1) and 56.5 (8.0) years respectively, with 54% female in both populations. Median (interquartile range) IGF-1 concentration was 21.3 (17.6, 24.9) nmol/L in both populations. In the observational population, 3.1% had HD hip OA (3.2% in the MR population); 0.7% had hand OA and 5.4% knee OA in both populations. In analyses adjusted for age, sex, ethnicity and oestrogen use, serum IGF-1 concentration was associated with a decreased odds of HD OA (ORhand=0.82 [0.77, 0.88] p=3x10-9, ORhip=0.94 [0.91, 0.97] p=4x10-4, ORknee=0.81 [0.80, 0.83] p=1x10-64; OR per doubling in IGF-1 concentration). After additional adjustment for BMI, serum IGF-1 remained associated with a reduced odds of hand OA; however, it was then associated with an increased odds of hip OA (ORhip=1.04 [1.01, 1.07] p=0.014), with little evidence of an association with knee OA. Using 1SMR, we found strong evidence for increased risk of hip OA with higher IGF-1 concentrations (risk difference per SD increase in IGF-1=0.009 [0.004, 0.015] p=0.001, corresponding to an approximate ORhip 1.36 [1.13, 1.63]), and also evidence for an increased risk of knee OA (risk difference=0.009 [0.001, 0.016] p=0.019). In contrast, point estimates suggested a decreased risk of hand OA, although confidence intervals were wide, reflecting the low prevalence of HD hand OA. When testing the assumptions of MR, we found evidence for an association between the IGF-1 allele score and BMI, which violates the assumption that an instrument is unrelated to any confounders of the exposure-outcome relationship. We therefore performed MVMR analyses to determine the BMI-independent effect; the effect size for IGF-1 on hip OA was unchanged (risk difference=0.009 [0.003, 0.014] p=0.003); whereas, the effect of IGF-1 on knee OA was attenuated by one-third (risk difference=0.006 [-0.001, 0.014] p=0.076). Conclusions: Our MR analyses provide strong evidence that higher concentrations of serum IGF-1 are a causal risk factor for hip OA in a very large UK population, with weaker evidence for a causal role in knee OA and very little evidence for an effect on hand OA. Our MVMR analyses estimate the direct effect of IGF-1 on OA conditional on BMI, these results suggest that this causal role of IGF-1 is independent of BMI, consistent with our observational analyses. Further study is justified to determine the mechanism underlying this association.