VC The Author 2017; all rights reserved. Published by Oxford University Press on behalf of the International Epidemiological Association exclusion of CpGs that were influenced by genetic variants [single nucleotide polymorphisms (SNPs)] in their DNA-m measurement probes. On the other hand, a recent article published in the International Journal of Epidemiology focused on changes of DNA methylation. Results from this Danish study by Tan et al. identified statistically significant changes in CpG methylation in 43 paired measurements 10years apart (age range at first measurement: 73-82years). 4 Compared with the British study, the investigators removed a smaller number of CpGs affected by probe SNPs, resulting in a total of 424706 CpGs that were analysed. Their results showed that 0.054%(2284 of 424706) CpGs changed significantly.

Since most studies analyse DNA-m in whole blood, there is a need to assess stability or changes of DNA methylation along with correcting for changes in the cell composition of whole blood. This way we avoid attributing stability or changes in methylation to individual CpGs which, however, are merely attributable to stability or changes in blood cell effects over time. Typically, we adjust the methylation of CpGs for five to six cell types [proportion of granulocytes, B-cells, Th1-cells, Th2-cells, monocytes,(eosinophils)] according to the approach by Houseman et al. 5 and Jaffe and Irizarry6 and additionally for nucleated red blood cells in cord blood. 7 However, a challenge arises if we have to compare DNA-m measured in the same participants at two time points adjusting for cell composition. This is because (i) cell composition and (ii) its effect on the methylation of individual CpGs both may change with time. Given that (i) and (ii) can both have an effect on DNA-m at two time points (t and tþ 1), this leads to the following two equations. Let yt be the methylation measured at time t and xt be the cell type proportion at time t.