Skip to main content

Table 1 Summary characteristics of blood-derived methylation studies and breast cancer risk (n = 17)

From: Epigenome-wide DNA methylation and risk of breast cancer: a systematic review

Design Study design
 Case-cohort or cohort studies, n = 2
 Nested case-control studies, n = 9
 Unspecified case-control studies, n = 3
 Cross-sectional study, n = 1
 Multiple designs, n = 2
Sample size
 Total participants, 90 to 228,951
 Breast cancer patients, 48 to 122,977
Population source
 Europe, n = 8
 Australia, n = 3
 USA, n = 2
 Europe and/or Australia and/or USA, n = 4
Follow-up
 Duration, 2 weeks to > 20 years
 Not reported in 9 studies
Breast cancer patients Mean age, 48 to 64 years old
Postmenopausal, 31 to 100%, NR in 10 studies
Invasive cancers, 88 to 100%, NR in 10 studies
ER-positive cancers, 0 to 83%, NR in 9 studies
DNA methylation measurement Timing
 Before diagnosis, n = 13
 After diagnosis, before treatment, n = 2
 After diagnosis, unspecified, n = 1
 Not reported, n = 1
Cell-type proportions
 Estimated (Houseman algorithm), n = 10
 Estimated, other method, n = 2
 Estimated, method NR, n = 2
 Not considered, n = 3
Probe design bias correction method
 Functional normalizationa, n = 7
 SWANa, n = 7
 BMIQ, n = 2
 Quantile normalization, n = 2
 RCP, n = 1
 Not reported, n = 4
Cross-hybridizing probes
 Excluded, n = 5
 Not reported, n = 12
Probes with SNP
 Excluded, n = 5
 Not excluded, n = 1
 Not reported, n = 11
X chromosomes
 Excluded, n = 5
 Included, n = 4
 Not reported, n = 8
Outcomes Breast cancer incidence, n = 16
Breast mammographic density, n = 1
Statistical modeling Global methylation, n = 9
Type of global methylation analysis
 Average across all included probesc, n = 6
 Average across pre-defined set of probesc, n = 5
Type of methylation value
 Beta-values, n = 8
 Not reported, n = 1
Statistical model
 Logistic regression, n = 5
 Cox proportional hazard model, n = 1
 Non-parametric test, n = 2
 Not reported, n = 1
Adjustment
 Appropriate, n = 3
 Incomplete, n = 4
 None, n = 2
Probe-wise differential methylation, n = 16
Type of methylation value
 Beta-values, n = 10
 M-values, n = 4
 Not reported, n = 2
Statistical model
 Logistic regressionb, n = 6
 Cox proportional hazard modelsb, n = 2
 Beta-regression, n = 2
 Linear mixed effect model, n = 2
 MetaXcan method, n = 1
 Linear regression with empirical Bayes methods, n = 1
 Non-parametric tests, n = 1
 Not reported, n = 2
Adjustment
 Appropriate, n = 3
 Incomplete, n = 12
 None, n = 1
Multiple comparison correction
 Bonferroni’s correction, n = 6
 FDR, n = 3
 None, n = 7
  1. n number of studies, NR not reported, SNP single nucleotide polymorphism, SWAN Subset-quantile within array normalization, BMIQ Beta-mixture quantile normalization, RCP Regression on Correlated Probes, DMP differentially methylated positions, FDR false discovery rate, ER estrogen receptor
  2. an = 6 studies used both functional normalization and SWAN
  3. bone study used both logistic regression and Cox proportional hazard models
  4. cn = 2 studies measured both