/Non-linear change was detected in a longitudinal mixed model that included slope changes at ages 6 (mid-childhood) and 9 years (late childhood) (p<1×10−07). Across all sites measured on the array, 11% of sites showed significant non-linear change during development. In the majority of cases (8%), DNAm levels increased or decreased from birth to age 6, after which they remained stable. First- and second/third-generation clock sites were enriched for this type of non-linearity with 23% (range=14-43%, OR=3.6 [95% ci=3.0-4.1], p=3.22×10−49) and 18% (range=12-27%, OR=2.5 [95% ci=2.1-3.0], p=1.76×10−19) of sites, respectively. Of note, since DNAm was drawn from cord blood at birth and peripheral blood later on, this non-linearity may partly reflect tissue differences rather than developmental changes per se. Yet, a similar enrichment pattern was observed for sites showing non-linear change later in development, at age 9: this type of non-linear pattern was present in 3% of sites on the array (p<1×10−07), whereas it was 8% (range=7-10%) in first-generation clock sites, and in 5% (range=2-6%) in second/third generation clock sites – a significant enrichment in both groups compared to non-clock sites (OR=3.3 [95% ci=2.6-4.1], p=6.40×10−18 and OR=1.8 [95% ci=1.2-2.5], p=1.46×10−03, respectively; Figure 1). An example of a non-linear DNAm pattern is shown in Figure 3.