Figure 5. Results of EWAS and potential confounding between DNA methylation and gestational age due to selection bias. (A) Scatter plots between Z scores from controls and Z scores from preeclampsia. (B) The depicted minimal causal diagram under the null hypothesis of no effect of GA on DNAm. Here, the pregnancy condition (preeclampsia) would induce a spurious association between DNAm and GA, because preeclampsia could prompt earlier delivery (shorter GA) and influence DNAm. Note that the association between GA and DNAm is not due to a direct causal relationship between DNAm and GA. Rather, the association is confounded by preeclampsia. If the selection criteria differ substantially across studies, the placental clock models may not perform well. (C) EWAS Manhattan plot of GA.