© 2018 The Author(s). Background: DNA methylation is an epigenetic mechanism that has been proposed as a possible link between genetic and environmental determinants of disease. Prior studies reported robust associations between the methylation of specific cytosine-phosphate-guanine (CpG) sites and plasma lipids, namely triglycerides (TGs) and high-density lipoprotein cholesterol (HDL-C). However, the causality of the observed association remains elusive, hampered by weak instrumental variables for methylation status. Aim: We present a novel application of the elastic net approach to implement a bidirectional Mendelian randomization approach to inferring causal relationships between candidate CpGs and plasma lipids in GAW20 data. Methods: We used DNA methylation, TGs, and HDL-C measured during the visit 2. Based on prior findings, we selected 5 methylation markers (cg00574958, cg07504977, cg06690548, cg19693031, and cg03717755) related to TGs, 2 markers (cg09572125 and cg02650017) related to HDL-C, and 2 markers (cg06500161 and cg11024682) related to both traits. We implemented an elastic net approach to improve the selection of the genetic instrument for the methylation markers, followed by bidirectional Mendelian randomization 2-stage least-squares regression. Results: We observed causal effects of blood fasting TGs on the methylation levels of cg00574958 (CPT1A) and cg06690548 (SLC7A11). For cg00574958, our findings were also consistent with the reverse direction of association, that is, from CPT1A methylation to TGs. Conclusions: Current evidence does not rule out either direction of association between the methylation of the cg00574958 CPT1A locus and plasma TGs, highlighting the complexity of lipid homeostasis. We also demonstrated a novel approach to improve instrument selection in DNA methylation studies.