Because the power of the SMR approach to detect pleiotropic associations reflects, in part, the power of the initial complex-trait GWAS, it is unsurprising that the highest number of SMR associations was found for traits characterized by the largest number of GWAS signals, such as height (423 significant GWAS loci, 506 SMR pleiotropic associations)50 and inflammatory bowel disease [MIM: 266600 ] (168 significant GWAS variants, 127 SMR pleiotropic associations).51 In contrast, no SMR associations were found for traits with few or no genome-wide-significant SNPs; such traits included parental age at death (0–1 significant GWAS variants),52 insulin secretion rate (no significant GWAS variants),53 and whether a person has ever smoked (no significant GWAS variants).54 We compared our SMR results to those obtained with our previous mQTL dataset—generated from a smaller number of samples—and observed high rates of replication for loci that were tested in both analyses. Because our previous SMR analysis was based on a subset of 43 traits and the reduced content of the Illumina 450K array, 842 pleiotropic associations reported in the current analysis were taken forward for replication; DNAm at 519 (33.0%) of these was associated with an mQTL variant, and therefore these associations had been tested in our previous SMR study; 268 (51.6%) were characterized by significant pleiotropic association in both studies. Furthermore, the vast majority of associations tested in both datasets (516; 99.4%) were in the same direction; this was significantly more than would be expected by chance (sign test p = 2.72 × 10−149; Figure S14), suggesting that there are many additional true signals in those that did not meet the stringent criteria for significance used in both studies.