Accounting for cellular heterogeneity is critical in epigenome-wide association studies. Epigenome-wide association studies (EWAS) are increasingly used to identify DNA methylation (DNAm) loci associated with diseases and traits. However, blood, a heterogeneous tissue, contains multiple cell types with distinct DNAm profiles. This heterogeneity can confound results, as age-related changes in DNAm may be due to shifts in cell composition rather than true biological changes. In this study, the authors used a statistical method to estimate cell composition from DNAm profiles and found strong evidence of age-related changes in blood cell composition. They also demonstrated that cell composition explains much of the variability in DNAm. Confounding between age and cell composition was observed at the CpG level, suggesting that failure to account for cell composition may lead to false positives. The authors developed software to estimate and explore cell composition confounding for the Illumina 450k microarray. They analyzed five publicly available studies and found that cell composition explains a larger percentage of variability than age for many CpGs. They also showed that cell composition is a major source of variability in DNAm data from peripheral blood. Confounding between cell composition and age can lead to false positives, and they recommend using their software to adjust for this confounding. The study emphasizes the importance of considering cell composition variability in epigenetic studies based on whole blood and other heterogeneous tissues. They provide a table summarizing cell-type variability for each CpG on the Illumina 450k array. CpGs highly associated with cell-type variability should be treated with caution, and those associated with both composition and the covariate of interest should be validated using FACS-derived cellular populations. The authors also discuss the limitations of regression approaches for adjusting for confounding and recommend using methods like RUV for reducing composition-based confounding. They note that these confounding problems are not limited to blood but apply to any tissue with mixed cell types. Careful study design, including targeted validation using cell sorting, can help isolate cell-type-specific changes. The study highlights the need to characterize and explore the effects of cellular heterogeneity in genome-wide DNAm data from heterogeneous tissues, especially peripheral blood.Accounting for cellular heterogeneity is critical in epigenome-wide association studies. Epigenome-wide association studies (EWAS) are increasingly used to identify DNA methylation (DNAm) loci associated with diseases and traits. However, blood, a heterogeneous tissue, contains multiple cell types with distinct DNAm profiles. This heterogeneity can confound results, as age-related changes in DNAm may be due to shifts in cell composition rather than true biological changes. In this study, the authors used a statistical method to estimate cell composition from DNAm profiles and found strong evidence of age-related changes in blood cell composition. They also demonstrated that cell composition explains much of the variability in DNAm. Confounding between age and cell composition was observed at the CpG level, suggesting that failure to account for cell composition may lead to false positives. The authors developed software to estimate and explore cell composition confounding for the Illumina 450k microarray. They analyzed five publicly available studies and found that cell composition explains a larger percentage of variability than age for many CpGs. They also showed that cell composition is a major source of variability in DNAm data from peripheral blood. Confounding between cell composition and age can lead to false positives, and they recommend using their software to adjust for this confounding. The study emphasizes the importance of considering cell composition variability in epigenetic studies based on whole blood and other heterogeneous tissues. They provide a table summarizing cell-type variability for each CpG on the Illumina 450k array. CpGs highly associated with cell-type variability should be treated with caution, and those associated with both composition and the covariate of interest should be validated using FACS-derived cellular populations. The authors also discuss the limitations of regression approaches for adjusting for confounding and recommend using methods like RUV for reducing composition-based confounding. They note that these confounding problems are not limited to blood but apply to any tissue with mixed cell types. Careful study design, including targeted validation using cell sorting, can help isolate cell-type-specific changes. The study highlights the need to characterize and explore the effects of cellular heterogeneity in genome-wide DNAm data from heterogeneous tissues, especially peripheral blood.