Abstract The genetic basis of levodopa-induced-dyskinesia (LiD) is poorly understood, and there have been few well-powered genome-wide studies. We performed a genome-wide survival meta-analyses to study the effect of genetic variation on the development of LiD in five separate longitudinal cohorts, and meta-analysed the results. We included 2784 PD patients, of whom 14.6% developed LiD. We found female sex (HR = 1.35, SE = 0.11, P = 0.007) and younger age at onset (HR = 1.8, SE = 0.14, P = 2 × 10 −5 ) increased the probability of developing LiD. We identified three genetic loci significantly associated with time-to-LiD onset. rs72673189 on chromosome 1 (HR = 2.77, SE = 0.18, P = 1.53 × 10 −8 ) located at the LRP8 locus, rs189093213 on chromosome 4 (HR = 3.06, SE = 0.19, P = 2.81 × 10 −9 ) in the non-coding RNA LINC02353 locus, and rs180924818 on chromosome 16 (HR = 3.13, SE = 0.20, P = 6.27 × 10 −9 ) in the XYLT1 locus. Based on a functional annotation analysis on chromosome 1, we determined that changes in DNAJB4 gene expression, close to LRP8, are an additional potential cause of increased susceptibility to LiD. Baseline anxiety status was significantly associated with LiD (OR = 1.14, SE = 0.03, P = 7.4 × 10 −5 ). Finally, we performed a candidate variant analysis of previously reported loci, and found that genetic variability in ANKK1 ( rs1800497 , HR = 1.27, SE = 0.09, P = 8.89 × 10 −3 ) and BDNF ( rs6265 , HR = 1.21, SE = 0.10, P = 4.95 × 10 −2 ) loci were significantly associated with time to LiD in our large meta-analysis.