Abstract
Epigenetic age acceleration (EAA), defined as the difference between chronological age and epigenetically predicted age, was calculated from multiple gestational epigenetic clocks (Bohlin, EPIC overlap, and Knight) using DNA methylation levels from cord blood in three large population-based birth cohorts: the Generation R Study (The Netherlands), the Avon Longitudinal Study of Parents and Children (United Kingdom), and the Norwegian Mother, Father and Child Cohort Study (Norway). We hypothesized that a lower EAA associates prospectively with increased ADHD symptoms. We tested our hypotheses in these three cohorts and meta-analyzed the results (n = 3383). We replicated previous research on the association between gestational age (GA) and ADHD. Both clinically measured gestational age as well as epigenetic age measures at birth were negatively associated with ADHD symptoms at ages 5–7 years (clinical GA: β = −0.04, p < 0.001, Bohlin: β = −0.05, p = 0.01; EPIC overlap: β = −0.05, p = 0.01; Knight: β = −0.01, p = 0.26). Raw EAA (difference between clinical and epigenetically estimated gestational age) was positively associated with ADHD in our main model, whereas residual EAA (raw EAA corrected for clinical gestational age) was not associated with ADHD symptoms across cohorts. Overall, findings support a link between lower gestational age (either measured clinically or using epigenetic-derived estimates) and ADHD symptoms. Epigenetic age acceleration does not, however, add unique information about ADHD risk independent of clinically estimated gestational age at birth.
Introduction
Attention-deficit hyperactivity disorder (ADHD) is one of the most common neurodevelopmental disorders, affecting approximately 5–8% of children [1]. Symptoms include behavioral and cognitive problems, such as distractibility, hyperactivity, and impulsive behavior. ADHD symptoms manifest along a continuum, with the disorder representing the upper end of this continuum. Indeed, many children experience subclinical ADHD symptoms, which have been found to associate with negative psychological health and social outcomes that can persist into adulthood [2–4].
Multiple risk factors for ADHD have been identified, including genetic factors [5], prenatal exposures (e.g. maternal psycho-pathology and inflammation) [6], and perinatal factors. Of these, low gestational age (GA) at birth is one of the most robust and replicated risk factors for ADHD [7]. The molecular processes underlying this association are still unclear, but DNA methylation (DNAm) has emerged as a key process of interest in recent years [8]. DNAm is an epigenetic mechanism that regulates gene activity in response to both genetic and environmental factors; it is strongly related to development and age(ing) and altered DNAm patterns have been linked to multiple negative health outcomes including neurodevelopmental disorders [9]. Interestingly, several epigenome-wide association studies (EWAS) have found that DNAm patterns in cord blood associate more strongly with later child ADHD symptoms than DNAm patterns measured in blood during childhood, i.e. prospective associations are stronger than cross-sectional associations [10, 11]. Yet, what factors explain this epigenetic signal at birth is currently not known. While existing EWASs of ADHD have adjusted for effects of (chronological) gestational age, it is possible that unmeasured differences in biological age at birth, which may not be fully captured by (chronological) gestational age, may partly drive observed associations between DNAm patterns at birth and later ADHD symptoms.
Epigenetic clocks are “biological age” estimators that are calculated using a set of DNAm sites known to associate with age. These clocks are widely used to measure the “biological age” of tissues and cells. Epigenetic age acceleration (EAA) is defined as the difference between epigenetic age and chronological age [12]. In adulthood, accelerated epigenetic age (i.e. the epigenetic/biological age is advancing at a faster rate than the chronological age) has been found to robustly associate with numerous disease outcomes, such as cardiovascular disease and all-cause mortality, and to a less consistent extent psychiatric risk [12]. It is unclear however, whether increased EAA [13, 14] is also a risk factor for suboptimal developmental outcomes early in life. For example, it is possible that higher EAA may indicate greater maturation, which could be advantageous for the child, whereas having a lower EAA (or epigenetic age deceleration) may be a sign of developmental delay. In line with this hypothesis, increased EAA at birth has been found to predict physical maturation later in childhood, including increased height, weight, BMI, and head circumference [15, 16]. However, little is still known about how EAA associates with neurodevelopmental outcomes such as ADHD symptoms [17]. A small set of studies based on single datasets and modest sample sizes have examined associations between EAA and internalizing or externalizing symptoms in children and adolescents, with mixed findings [13, 14]. One EWAS of ADHD in childhood using saliva samples also examined EAA and found no cross-sectional association with ADHD [18]. In contrast, Arpawong et al. [19] reported that middle-aged adults with an increased polygenic score for ADHD showed accelerated biological aging based on the GrimAge clock [19]. No study to our knowledge has yet examined how gestational EAA associates with ADHD symptoms in childhood. Addressing this gap could help to shed light on the biological processes that underlie the known gestational age-ADHD link, as well as to evaluate the potential utility of epigenetic clocks – a more intrusive, labor-intensive and costly measure of age compared to routinely assessed chronological gestational age – as an early marker of ADHD risk.
The aim of this study was to examine prospective associations between epigenetic clocks at birth and ADHD symptoms in childhood, using pooled results from over 3000 participants across three large population-based birth cohorts: the Generation R Study, the Avon Longitudinal Study of Parents and Children (ALSPAC), and the Norwegian Mother, Father and Child Cohort Study (MoBa). Given the well-established association of prematurity and lower gestational age with ADHD symptoms, we first sought to replicate the association between lower GA and increased ADHD symptoms. Three different gestational epigenetic clocks were used to test the robustness of results and maximize comparability with existing studies [20]. The clocks we used have been developed for use in cord blood: the Bohlin, EPIC overlap, and Knight clocks [21–23]. For each, we calculated measures of (i) epigenetically predicted gestational age, (ii) raw EAA, which represents the difference between epigenetically predicted and clinical gestational age; and (iii) residual EAA, which instead represents epigenetic age corrected for clinical gestational age. Associations between these metrics at birth and ADHD symptoms in childhood were tested for each clock individually and within each separate cohort. Results were then pooled via meta-analysis. We hypothesized that a decreased residual EAA (i.e. being biologically younger than one’s chronological gestational age) would prospectively associate with higher ADHD symptoms in childhood, over and above chronological gestational age itself.
Methods
Setting
The analyses in this study were conducted using data from three prospective population-based cohorts: the Dutch Generation R Study (Generation R), the British Avon Longitudinal Study of Parents and Children (ALSPAC), and the Norwegian Mother, Father and Child Cohort Study (MoBa).
Generation R
Pregnant women residing in the study area of Rotterdam, the Netherlands, expected to deliver between April 2002 and January 2006, were invited to enroll in the Generation R Study [24]. The study is conducted in accordance with the World Medical Association Declaration of Helsinki and has been approved by the Medical Ethics Committee of Erasmus MC, University Medical Center Rotterdam (MEC 198.782/2001/31). Informed consent was obtained from all participants.
MoBa
Participants include mothers, fathers and child. They were recruited from all over Norway from 1999-2008. The women consented to participation in 41% of the pregnancies [25, 26]. Ethical approval for the MoBa study was obtained by The Regional Committees for Medical and Health Research Ethics (REK- 2009/1899-7).
ALSPAC
Pregnant women residing in the study area of the former county Avon in the United Kingdom with an expected delivery date between April 1991 and December 1992 were invited to enroll in ALSPAC [27, 28]. Ethical approval for the ALSPAC study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. Please see Supplementary Materials (Additional file 1) for further details.
Study population
The Generation R Study includes data from 9749 live-born children. DNAm data at birth are available for 1396 children. Of these, 80 children had no ADHD measures at any age and were therefore excluded from the analysis. This resulted in a final cohort of 1316 children in Generation R. The ALSPAC study includes data from 14,541 live-born children. DNAm data at birth was available for 692 children in ALSPAC who also had at least one ADHD measure at any age. The MoBa cohort includes approximately 112,195 children, 94,754 mothers and 74,779 fathers. The current study is based on version [12] of the quality-assured data files released for research in 2020, and comprises different subsamples with DNAm data. We only included in our analyses subsamples that (i) comprised randomly selected participants and (ii) have not been previously used to develop the epigenetic clocks used in this study (e.g. the Bohlin clock has been trained on MoBa 1 data). Based on this selection, we included the MoBa 8 subsample with DNAm data from 1068 children and MoBa 2 with data from 685 children. Of note, the MoBa 8 and MoBa 2 datasets differ in a number of respects, including the sampling strategy. MoBa 8 consists of a set of randomly selected participants from MoBa, whereas MoBa 2 was over-sampled on a range of cases with Asthma (35% cases with Asthma at age 7). Quality control pipelines used to process the DNA methylation data were also performed separately, as described in detail elsewhere ([29, 30]; https://github.com/folkehelseinstituttet/mobagen/wiki/Methylation). Of the children with DNAm data, 243 children in MoBa 8 and 135 children in MoBa 2 had no ADHD measures and were excluded from the analysis. This resulted in a final sample of 825 children in MoBa 8 and 550 children in MoBa 2. In total, this study included pooled results from 3383 children.
DNA methylation
In all cohorts, cord blood was drawn at birth. Five hundred nanograms of DNA were bisulfite converted using the EZ-96 DNA Methylation kit (Shallow) (Zymo Research Corporation, Irvine, USA). The Generation R and ALSPAC samples were processed with the Illumina Infinium HumanMethylation450 BeadChip, whereas the MoBa samples were processed with a mix of HumanMethylation450 BeadChip (MoBa 2) and Illumina MethylationEPIC 850 K array (MoBa 8; Illumina Inc., San Diego, CA). Detailed control steps and normalization procedures have been described previously [20–22, 31, 32].
In short, in Generation R, the CPACOR workflow was used for quality control and normalization [33]. Arrays with observed technical problems such as failed bisulfite conversion, hybridization, or extension as well as arrays with a sex mismatch were removed. Probes that had a detection p-value above background ≥1E−16 were set to missing per array. Arrays with a call rate >95% per sample were included and quantile normalized (as described previously [20].
In ALSPAC, the meffil package [34] was used for quality control. Samples with mismatched genotypes, mismatched sex, incorrect relatedness, low concordance with samples collected at other time points, extreme dye bias, and poor probe detection were removed and carried forward into normalization.
In MoBa [35], all duplicates were removed and arrays not fulfilling the 5% detection p value were excluded. Within-array normalization of type I and II probes was performed using BMIQ from the R package watermelon [36]. Detailed pre-processing steps have been published previously [22, 30].
Age estimates
Gestational age at birth was assessed from midwife or obstetric records. Epigenetic gestational clocks make use of a select number of CpG sites that are most strongly correlated with gestational age at birth. The gestational epigenetic clock developed by Bohlin et al. [21] estimates epigenetic age based on DNAm levels of 96 CpG sites from the HumanMethylation450 BeadChip that were selected using Lasso regression and the predictions were tested in holdout samples. The epigenetic clock developed by Knight et al. [23] estimates epigenetic age at birth based on DNAm levels of 148 CpG sites from the HumanMethylation450 BeadChip and the HumanMethylation27 BeadChip that were selected using elastic net regression and the predictions were tested in holdout samples. The EPIC overlap clock developed by Haftorn et al. [22] is based on CpGs that can be found both on HumanMethylation450 BeadChip array and on the Illumina MethylationEPIC 850 K array. It estimates epigenetic age at gestation based on 173 CpG sites that were selected using Lasso regression and the predictions were tested in holdout samples.
The methylclock package in R 4.1.1 was used to calculate gestational epigenetic age measures and to impute missing values if less than 20% were missing based on the Bohlin, Knight and EPIC overlap clocks. We adjusted the package to also calculate the EPIC overlap clock, as this clock was not yet implemented in the methylclock package at the time of conducting this study. For each clock, we computed the following three estimates: (i) epigenetic age (i.e. epigenetic estimate of gestational age); (ii) raw EAA; and (iii) residual EAA. Raw EAA was calculated as the estimate of epigenetic age minus (clinically estimated) chronological gestational age. Thus, whereas raw EAA is more intuitive to interpret than residual EAA, it does not account for shared variance between epigenetic and chronological gestational age [34]. Residual EAA, on the other hand, is defined as the residuals from a linear regression of epigenetic age on clinically estimated gestational age [20, 37], and is thus uncorrelated with chronological age (see Additional file Supplementary Fig. S2).
Covariates
Covariates were selected based on previous literature that investigated epigenetic age estimates at birth in relation to developmental outcomes (e.g. Monasso et al. [20]). Covariates included in the main model were child sex, estimated cell-type proportions, and batch effects. In an extended model, maternal prenatal smoking, maternal alcohol consumption, maternal pre-pregnancy BMI, maternal highest education level attained, income, delivery method, maternal age at delivery, and parity were additionally included as covariates. Questionnaires completed by the mothers in each trimester of pregnancy were used to obtain information on maternal covariates mentioned above and information on child sex was obtained from midwife and hospital records. Cell-type proportions were estimated using the combined cord blood-specific reference set in the “FlowSorted.-CordBlood.Combined.450K” Bioconductor package [38]. This reference set includes CD8 + T cells, CD4 + T cells, natural killer cells, B cells, monocytes, granulocytes, and nucleated red blood cells.
ADHD symptom assessment
Outcome data was chosen based on maximizing comparability in time point and type of assessment tool for ADHD symptoms across cohorts. In all cohorts, only children with at least one reported score were included in the analysis.
In Generation R, ADHD symptoms were assessed with the attention problems syndrome scale of the Child Behavior Checklist (CBCL) at age 6 [39]. Additional CBCL measurements obtained at age 3, 10, and 13 years were only used during the outcome imputation.
In ALSPAC, ADHD symptoms were assessed based on the Strengths and Difficulties Questionnaire hyperactivity subscale (SDQ; maternal reports) at age 7 [40]. Additional SDQ measurements obtained at age 4, 9, and 11 were used during the outcome imputation.
In the MoBa study, ADHD symptoms were assessed based on the attention problems syndrome scale of the CBCL at age 5, which has been validated previously [41]. In the MoBa study only six out of eleven items on the attention subscale were administered, potentially affecting the precision of the measurement in this cohort. Additional ADHD measurements obtained at age 1.5, 3, and 8 were used for outcome imputation.
Given that ADHD symptom distributions in these cohorts are highly skewed towards low symptoms, and that this is likely to attenuate effect estimates, we carried out a sensitivity analysis using a binary outcome of probable ADHD diagnosis (as opposed to continuous ADHD symptom scores). We restricted these analyses to Generation R, as (i) it is the largest cohort in the present study using the complete CBCL attention problems scale, and (ii) a clinically meaningful cut-off for this scale has been previously validated in this cohort using the Diagnostic Interview Schedule for Children–Parent version (DISC-P; Shaffer et al. [42]; Mil et al. [43])). Briefly, the DISC-P was administered for a subsample of Generation R children at age 6 years through parent interviews during home visits. Children who met diagnostic criteria for ADHD based on the DISC-P had a median score of 8 on the CBCL attention subscale, whereas children who did not meet diagnostic criteria for ADHD had a median CBCL score of 2.61, resulting in a cut-off score for probable ADHD diagnosis of > = 7. Based on this cut-off, we defined ADHD cases in our analytical epigenetic sample (n = 85) vs controls (n = 1213). We then performed matching for sex (using exact sex matching) and age at ADHD symptom measurement (using propensity scores), using the MatchIt package, with the optimal ratio of cases:controls being 1:5, allowing us to maximize the number of controls used. This resulted in a total of 85 cases and 425 controls. Conditional logistic analyses were run as specified in statistical analysis section.
Statistical analyses
The association of the epigenetic clock estimates at birth with ADHD symptoms in early childhood was tested within each cohort using robust linear regression models (rlm package) in R version 4.1.1. ADHD symptom scores were z-transformed to account for the different questionnaires employed in different cohorts. We tested the following exposures in separate models: (i) gestational age; and from each clock, the estimates of (ii) epigenetic age, (iii) raw EAA, and (iv) residual EAA. Analyses were run with a core set of covariates, and with an extended set of covariates, as described above. We did not use a multiple testing correction, as model (i) served as a replication of prior studies to confirm the quality of our data, model (ii) was used to show that the epigenetic clocks indeed measure GA accurately in our data to inform about ADHD risk, model (iii) was used to test whether GA may drive any association between residual EAA and GA to inform about ADHD risk, and only model (iv) was used as the main model to test our hypothesis. Missing data on covariates and ADHD symptoms were imputed using multiple imputation, with the mice package. Cohort-specific results for each epigenetic clock were meta-analyzed using the rma.uni function of the metafor package using a fixed effects model and inverse variance weighting. As a sensitivity analysis we examined sex-and age-matched cases and controls based on probable ADHD diagnosis in Generation R. We performed a conditional logistic regression analysis with ADHD case-control status as the outcome and gestational age, epigenetic clock estimates, raw, or residual age acceleration as exposure (in separate models). Covariates included estimated cell-type proportions, batch effects, maternal prenatal smoking, maternal alcohol consumption, maternal pre-pregnancy BMI, maternal highest education level attained, delivery method, maternal age at delivery, and parity. Age at ADHD assessment and sex were included as strata. For access to cohort-level data, requests can be sent directly to individual studies. Analytical codes can be requested from authors.
Results
Subject characteristics
Table 1 shows maternal and child characteristics based on non-imputed data for each cohort. Characteristics based on imputed data can be found in Additional file 1 Supplementary Table S1. There was no substantial difference in sample characteristics in terms of descriptive statistics between the imputed and non-imputed datasets.
Table 1. Participant characteristics per cohort.
| Age, years | 32.5 (30.1, 34.9) | 30 (28.0, 33.0) | 30 (27.0, 33.0) | 30.0 (27.0, 33.0) |
| Educational level | ||||
| No or primary | 18 (1.4%) | 40 (4.9%) | 36 (6.7%) | 0 (0%) |
| Secondary | 407 (31%) | 207 (25%) | 176 (33%) | 542 (79%) |
| Higher | 875 (67%) | 576 (70%) | 328 (61%) | 146 (21%) |
| Parity | ||||
| Nulliparous | 801 (61%) | 400 (48%) | 221 (40%) | 318 (47%) |
| Multiparous | 513 (39%) | 425 (52%) | 329 (60%) | 361 (53%) |
| Pre-pregnancy body mass index | 22.3 (20.7, 24.7) | 23.1 (21.2, 25.9) | 23.3 (21.0, 26.1) | 22.0 (20.6, 23.8) |
| Mode of delivery | ||||
| Vaginal delivery spontaneous | 968 (78%) | 651 (79%) | 423 (77%) | 249 (59%) |
| Vaginal delivery induced | 146 (12%) | 88 (11%) | 50 (9.1%) | 5 (1.2%) |
| Caesarean section, elective | 45 (3.6%) | 28 (3.4%) | 35 (6.4%) | 66 (16%) |
| Caesarean section, urgent | 79 (6.4%) | 58 (7.0%) | 41 (7.5%) | 0 (0%) |
| Other | 0 (0%) | 0 (0%) | 0 (0%) | 105 (25%) |
| Smoking | ||||
| Never smoked during pregnancy | 931 (77%) | 640 (78%) | 414 (76%) | 398 (62%) |
| Smoked until pregnancy was known | 119 (9.9%) | 117 (14%) | 72 (13%) | 172 (27%) |
| Continued smoking during pregnancy | 153 (13%) | 65 (7.9%) | 57 (10%) | 68 (11%) |
| Alcohol | ||||
| Never drank alcohol during pregnancy | 334 (28%) | 469 (67%) | 296 (61%) | 201 (30%) |
| Drank alcohol until pregnancy was known | 182 (15%) | 97 (14%) | 66 (14%) | 87 (13%) |
| Continued drinking alcohol during pregnancy | 679 (57%) | 131 (19%) | 121 (25%) | 391 (58%) |
| Child characteristics | ||||
| Birth weight, gram | 3580 (3244, 3890) | 3650 (3270, 3950) | 3688 (3370, 4000) | 3480 (3172, 3780) |
| Sex | ||||
| Boy | 654 (50%) | 407 (49%) | 316 (57%) | 347 (50%) |
| Girl | 662 (50%) | 418 (51%) | 234 (43%) | 345 (50%) |
| Clinically estimated GA at birth, weeks | 40.36 (39.43, 41.14) | 40.14 (39.14, 41.00) | 40.14 (39.00, 40.96) | 40.00 (39.00, 41.00) |
| DNAm gestational age | ||||
| Bohlin clock | ||||
| DNAm GA estimate, weeks | 39.43 (38.83, 39.93) | 41.18 (40.60, 41.71) | 40.53 (39.85, 41.18) | 39.64 (38.94, 40.28) |
| Raw EAA, weeks | −0.90 (−1.46, −0.31) | 1.03 (0.45, 1.63) | 0.49 (−0.13, 1.11) | 0.01 (−0.68, 0.75) |
| Residual EAA, weeks | 0.04 (−0.37, 0.41) | 0.03 (−0.38, 0.40) | 0.02 (−0.48, 0.46) | 0.01 (−0.47, 0.48) |
| EPIC overlap clock | ||||
| DNAm GA estimate, weeks | 39.65 (38.92, 40.19) | 40.32 (39.68, 40.88) | 39.80 (39.12, 40.47) | 39.80 (39.01, 40.47) |
| Raw EAA, weeks | −0.71 (−1.31, −0.10) | 0.17 (−0.37, 0.77) | −0.30 (−0.83, 0.43) | 0.12 (−0.55, 0.90) |
| Residual EAA, weeks | 0.00 (−0.43, 0.45) | 0.04 (−0.42, 0.43) | 0.02 (−0.46, 0.46) | 0.33 (−3.33, 3.77) |
| Knight clock | ||||
| DNAm GA estimate, weeks | 36.52 (35.42, 37.51) | 41.32 (40.49, 42.13) | 39.51 (38.43, 40.48) | 38.48 (37.09, 39.74) |
| Raw EAA, weeks | −3.73 (−4.69, −2.77) | 1.29 (0.44, 2.10) | −0.58 (−1.47, 0.31) | −0.98 (−2.41, 0.14) |
| Residual EAA, weeks | 0.15 (−0.81, 0.98) | −0.01 (−0.68, 0.66) | 0.05 (−0.91, 0.78) | 0.18 (−1.06, 1.14) |
| ADHD score | 2 (1.00, 4.00) | 0.25 (0.08, 0.45) | 0.18 (0.09, 0.45) | 3.00 (1.00, 5.00) |
| Age at ADHD measurement, years | 5.88 (5.73, 6.01) | 5.00 (5.00, 5.00) | 5.00 (5.00, 5.00) | 7.00 (7.00, 7.00) |
In ALSPAC, delivery mode ‘other’ refers to forceps or vacuum extraction and assisted breech.
IQR interquartile range.
Median (IQR); n (%) some variables will not amount to total N due to missing values.
Associations between clinically estimated and epigenetically estimated gestational age
The correlation between clinical and epigenetic measures of gestational age was high across cohorts for the Bohlin clock (rGenR = 0.70; rMoba8 = 0.74; rMoBa2 = 0.72; and rALSPAC = 0.61), as well as for the EPIC overlap clock (rGenR = 0.71; rMoba8 = 0.74; rMoBa2 = 0.72; and rALSPAC = 0.58), whereas it was moderate for the Knight clock (rGenR = 0.46; rMoba8 = 0.57; rMoBa2 = 0.54; and rALSPAC = 0.33). A full overview of the performance of the Bohlin, EPIC overlap, and Knight gestational age clocks can be found in Additional file Supplementary Fig. S2.
Across all clocks and cohorts, gestational age was negatively correlated with raw EAA (Additional File Supplementary Fig. S2). As expected, given the nature of the equation, gestational age was uncorrelated with residual EAA (Additional File Supplementary Fig. S2).
Associations between clinical gestational age, epigenetic age estimates and ADHD symptoms
Cohort-specific results are presented in Supplementary Tables S2 and S3. As expected, results from the meta-analysis showed that lower (chronological) gestational age at birth associates with higher ADHD symptoms in childhood in both the main and the extended model (main model clinical GA: β = −0.04, p < 0.001; Table 2), in line with previous research.
Table 2. Meta-analysis results by type of age estimate, epigenetic clock and model.
| Main | Extended | |||
|---|---|---|---|---|
| Age estimate | Beta (95% CI) | P value | Beta (95% CI) | P value |
| Gestational age, weeks | ||||
| Clinically estimated GA | −0.04 (−0.07, −0.02) | 0.00 | −0.03 (−0.06, −0.01) | 0.02 |
| DNAm GA estimate | ||||
| Bohlin | −0.05 (−0.09, −0.01) | 0.01 | −0.04 (−0.09, −0.00) | 0.04 |
| EPIC overlap | −0.05 (−0.08, −0.01) | 0.01 | −0.04 (−0.08, −0.00) | 0.04 |
| Knight | −0.01 (−0.04, 0.01) | 0.26 | −0.01 (−0.04, 0.01) | 0.22 |
| Epigenetic age acceleration | ||||
| Raw EAA, weeks | ||||
| Bohlin | 0.04 (0.01, 0.08) | 0.02 | 0.03 (−0.01, 0.07) | 0.12 |
| EPIC overlap | 0.04 (0.00, 0.08) | 0.03 | 0.03 (−0.01, 0.06) | 0.18 |
| Knight | 0.02 (0.00, 0.05) | 0.04 | 0.01 (−0.01, 0.04) | 0.25 |
| Residual EAA, weeks | ||||
| Bohlin | −0.01 (−0.06, 0.05) | 0.75 | −0.02 (−0.08, 0.04) | 0.52 |
| EPIC overlap | 0.00 (−0.01, 0.02) | 0.87 | 0.00 (−0.02, 0.02) | 0.97 |
| Knight | 0.01 (−0.02, 0.04) | 0.42 | 0.00 (−0.02, 0.03) | 0.84 |
Covariates included in the main model were child sex, cell-type proportions, and batch effects. In an extended model, maternal prenatal smoking, maternal alcohol consumption, maternal pre-pregnancy BMI, maternal highest education level attained, income, delivery method, maternal age at delivery, and parity were included as covariates on top of the main model covariates. Bold numbers indicate statistically significant results.
For both the Bohlin and EPIC overlap clock, epigenetic age estimates also showed weak negative associations with ADHD in the main and extended models, whereas the epigenetic age estimate from the Knight clock was not associated with ADHD symptoms (Bohlin: β = −0.05, p = 0.01, 95% CI = −0.09, −0.01; EPIC overlap: β = −0.05, p = 0.01, 95% CI = −0.08, −0.01; Knight: β = −0.01, p = 0.26, 95% CI = −0.04, −0.01; Table 2).
With regard to EAA estimates, raw EAA showed a weak positive association with ADHD symptoms across all cohorts and epigenetic clocks in the main model (pooled results for Bohlin: β = 0.04, p = 0.02, 95% CI = 0.01, 0.08; EPIC overlap: β = 0.04, p = 0.03, 95% CI = 0.00, 0.08; Knight: β = 0.02, p = 0.04, 95% CI = 0.01, 0.05; Table 2). This is in line with the observed inverse correlation between gestational age and raw EAA across cohorts. However, these associations were only significant in the main model and not in the extended model adjusting for maternal prenatal smoking, maternal alcohol consumption, maternal pre-pregnancy BMI, maternal highest education level attained, income, delivery method, maternal age at delivery, and parity.
Residual EAA was not associated with ADHD symptoms in any of our models, regardless of the epigenetic clock used (pooled results for Bohlin: β = −0.01, p = 0.75, 95% CI = −0.06, −0.05; EPIC overlap: β = 0.00, p = 0.87, 95% CI = −0.01, 0.0; Knight: β = 0.01, p = 0.42, 95% CI = −0.02, 0.04; Table 2). Forest plots of the results for the Bohlin clock are shown in Fig. 1. A detailed overview of all cohort-specific results of all clocks can be found in Additional file Supplementary Table S2.
Fig. 1. Forest plots of pooled analysis using the Bohlin clock.
Results to the right of the dotted lines indicate a positive association with ADHD symptoms, results to the left indicate a negative association. FE model is the fixed effects model and shows the pooled estimate across cohorts of the prospective association between ADHD symptoms and A GA estimates derived from the Bohlin clock, B Raw EAA (difference between clinical and epigenetically estimated gestational age) derived from the Bohlin clock, and C Residual EAA (raw EAA corrected for clinical gestational age) derived from the Bohlin clock.
Associations between clinical gestational age, epigenetic age estimates and ADHD diagnosis
Our main analyses focus on ADHD symptoms in childhood. To investigate whether associations may differ at the higher end of the severity spectrum, we conducted a sensitivity analysis in the Generation R Study. Newborns with an older gestational age at birth were less likely to have a probable ADHD diagnosis at age 6 (Gestational age: OR = 0.60; p = 0.01; 95% CI = 0.41, 0.90). The same trend was observed when gestational age was estimated based on DNAm using the Bohlin and EPIC overlap clock, but associations were not significant for the Knight clock (Bohlin: OR = 0.35; p = 0.01; 95% CI = 0.16, 0.73; EPIC overlap: OR = 0.38; p = 0.01; 95% CI = 0.18, 0.80; Knight: OR = 0.80; p = 0.21; 95% CI = 0.56, 1.14). None of the epigenetic age acceleration measures were significantly associated with a probable ADHD diagnosis at age 6, neither for raw (Bohlin: OR = 1.48; p = 0.14; 95% CI = 0.88, 2.48; EPIC overlap: OR = 1.36; p = 0.24; 95% CI = 0.82, 2.27); Knight: OR = 1.22; p = 0.19; 95% CI = 0.91, 1.65) nor residual age acceleration (Bohlin: OR = 0.65; p = 0.40; 95% CI = 0.25, 1.74; EPIC overlap: OR = 0.71; p = 0.40; 95% CI = 0.31, 1.58; Knight: OR = 01.08; p = 0.67; 95% CI = 0.76, 1.53).
Discussion
This study investigated the prospective association between epigenetic clocks at birth and ADHD symptoms in childhood, meta-analyzing data from three population-based birth cohorts with a total of over 3000 individuals. We utilized three different epigenetic clocks that have been developed for use in cord blood (Bohlin, EPIC Overlap, Knight) to derive estimates of epigenetic age as well as epigenetic age acceleration (i.e. raw and residual EAA). We highlight here three key findings. First, we found that, in line with previous studies, a lower (clinically estimated) gestational age at birth prospectively associated with more severe ADHD symptoms in childhood. Second, we found similar associations when using epigenetic-based instead of clinical measures of gestational age. Third, there was no strong evidence for an association between raw or residual age acceleration and ADHD symptoms after stringent covariate correction. Together, these findings suggest that (positive or negative) epigenetic age acceleration is unlikely to explain the observed link between gestational age and ADHD symptoms, and that little additional information may be gained about ADHD risk by using epigenetic estimations of GA, or age acceleration, over and above clinical (i.e., chronological) measures of GA.
We first tested the association of (chronological) gestational age and ADHD for comparative purposes, given the large body of research supporting this association [6], and indeed found that lower gestational age associated with higher ADHD symptoms in our pooled results. Measures of gestational and epigenetic age at birth showed a similar association with ADHD symptoms, in line with the strong observed correlations between these measures, at least for the Bohlin and EPIC overlap clocks. The Knight clock generally showed lower correlations across all cohorts, which may point towards differences between the training sample used by Knight et al. [23] – which was oversampled for prematurity – and the population-based cohorts in our study [15]. Of note, our pooled results were mostly driven by the Generation R and ALSPAC cohorts. In the MoBa subsamples, no significant associations of gestational age (or any of the clock measurements) with ADHD were observed, which may reflect differences in ADHD symptom assessment and sampling strategy (e.g., oversampling for cases of Asthma in MoBa 2) within these datasets.
Regarding the measures of epigenetic age acceleration, we found little evidence for a prospective association between EAA and ADHD symptoms across cohorts or epigenetic clocks. A positive raw EAA (i.e. epigenetic age larger than gestational age) was found to be weakly associated with increased ADHD symptoms in childhood, but not when adjusting for additional confounders. This association is likely explained by the fact that raw EAA showed a strong inverse correlation with gestational age, indicating that children born at a younger age tend to have a positive raw EAA compared to peers born at an older age. When we used residual EAA, which is by definition uncorrelated with gestational age, we found no association with ADHD symptoms in either the main or extended models. As such, EAA does not seem to add unique information about ADHD risk over and above gestational age itself.
We initially modeled ADHD symptoms dimensionally to capture the full range of symptom severity in the general population (primarily comprised of term children), rather than focusing on ADHD diagnosis. These analyses may not have been able to capture associations between epigenetic acceleration and ADHD symptoms if these would be evident only in premature children or at more extreme ends of the ADHD symptom spectrum (e.g., children with an ADHD diagnosis). To address the latter, we carried out a sensitivity analysis of probable ADHD cases and controls in the Generation R study, which again showed that epigenetic age acceleration as measured by several epigenetic clocks did not significantly increase the odds of having a probable ADHD diagnosis in childhood.
A previous EWAS meta-analysis found that DNAm patterns at birth are more strongly associated with child ADHD symptoms than DNAm patterns measured in childhood [10]. While unmeasured differences in biological age at birth (over and above chronological gestational age, which was corrected for in the analyses) could have contributed to the stronger epigenetic signal identified at birth, our results suggest that this is unlikely to be the case. Rather, these epigenetic timing effects may be explained by other factors. For example, DNAm patterns at birth may be a better proxy of genetic or prenatal environmental risk factors for ADHD, with this signal becoming more diluted over time due to the accumulated influence of postnatal exposures [8].
Our findings should be interpreted in the context of several limitations. We selected our outcome measures based on the most comparable time point and instrument for assessing ADHD across cohorts, yet these were not identical. For the ALSPAC study, a different ADHD measure (the SDQ) has been used than for Generation R (where the CBCL was used). Both measures are highly correlated and can distinguish between children with and without a psychiatric condition equally well [44]. The MoBa samples made use of the CBCL as well, but only a subset of items in the inattention subscale was administered, potentially attenuating the precision of the measurement in MoBa, possibly explaining why no associations were found in this cohort. Furthermore, it is possible that epigenetic age acceleration measured at a later time point than birth may be a more meaningful predictor of ADHD symptoms, as there may be a larger variation in epigenetic age estimates between children. This highlights the need for the development of reliable epigenetic clocks in pediatric populations.
Conclusion
Overall, our findings suggest that epigenetic clocks measured from cord blood at birth do not add unique information about ADHD risk over and above more routinely assessed clinical estimates of gestational age at birth.
Supplementary Material
Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41380-024-02544-2.
Acknowledgements
The Generation R Study is conducted by Erasmus MC, University Medical Center Rotterdam in close collaboration with the School of Law and Faculty of Social Sciences of the Erasmus University Rotterdam, the Municipal Health Service Rotterdam area, Rotterdam, the Rotterdam Homecare Foundation, Rotterdam and the Stichting Trombosedienst & Artsenlaboratorium Rijnmond (STAR-MDC), Rotterdam. We gratefully acknowledge the contribution of children and parents, general practitioners, hospitals, midwives and pharmacies in Rotterdam. The generation and management of the Illumina 450 K methylation array data (EWAS data) for the Generation R Study was executed by the Human Genotyping Facility of the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC, and the Netherlands. We thank Mr. Michael Verbiest, Ms. Mila Jhamai, Ms. Sarah Higgins, Mr. Marijn Verkerk and Dr. Lisette Stolk for their help in creating the EWAS database. We thank Dr. Alexander Teumer for his work on the quality control and normalization scripts. We are extremely grateful to all the families who took part in the ALSPAC study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council and Wellcome (Grant ref: 217065/Z/19/Z) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and KS will serve as guarantors for the contents of this paper. A comprehensive list of grants funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf). The Norwegian Mother, Father and Child Cohort Study is supported by the Norwegian Ministry of Health and Care Services and the Ministry of Education and Research. We are grateful to all the participating families in Norway who take part in this on-going cohort study. The Norwegian Mother and Child Cohort Study is supported by the Norwegian Ministry of Health, and the Norwegian Research Council/FUGE (grant no. 151918/S10). This work was partly supported by the Research Council of Norway through its Centers of Excellence funding scheme, project number 262700 and Grant Number 288083, 301004. The general design of the Generation R Study is made possible by financial support from the Erasmus MC, Erasmus University Rotterdam, the Netherlands Organization for Health Research and Development and the Ministry of Health, Welfare and Sport. The EWAS data were funded by a grant from the Netherlands Genomics Initiative (NGI)/ Netherlands Organisation for Scientific Research (NWO) Netherlands Consortium for Healthy Aging (NCHA; project nr. 050-060-810), by funds from the Genetic Laboratory of the Department of Internal Medicine, Erasmus MC, and by a grant from the National Institute of Child and Human Development (R01HD068437). This project received funding from the European Union’s Horizon 2020 research and innovation programme (874739, LongITools; 824989, EUCAN-Connect) and from the European Joint Programming Initiative “A Healthy Diet for a Healthy Life” (JPI HDHL, NutriPROGRAM project, ZonMw the Netherlands no.529051022). CAMC and EW are supported by the European Union’s Horizon 2020 Research and Innovation Programme (EarlyCause, grant agreement No 848158); CAMC is also supported by the European Union’s HorizonEurope Research and Innovation Programme (FAMILY, grant agreement No 101057529; HappyMums, grant agreement No 101057390) and the European Research Council (TEMPO; grant agreement No 101039672). This research was conducted while CAMC was a Hevolution/AFAR New Investigator Awardee in Aging Biology and Geroscience Research. EW received funding from the National Institute of Mental Health of the National Institutes of Health (award number R01MH113930) and from CLOSER (ES/K000357/1) and from UK Research and Innovation (UKRI) under the UK government’s Horizon Europe / ERC Frontier Research Guarantee [BrainHealth, grant number EP/Y015037/1].
Footnotes
Author Contributions
KS, HT and CAMC were responsible for conceptualization of this study. KS analyzed the data from the Generation R cohort and meta-analyzed the data. KLH and CMP analyzed the data from the MoBa cohort. FS and EW analyzed the data from the ALSPAC cohort. KS, HT and CAMC interpreted the data. KS wrote the original draft of the manuscript under the supervision of HT and CAMC, and comments were provided by JFF, EW, CMP, MB, and JB. All authors read and contributed to the preparation of the final manuscript. All authors read and approved the final manuscript.
Competing Interests
The authors declare no competing interests.
Reprints and permission information is available at http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Polanczyk GV, Willcutt EG, Salum GA, Kieling C, Rohde LA. ADHD prevalence estimates across three decades: An updated systematic review and meta-regression analysis. Int J Epidemiol. 2014;43:434–42. doi: 10.1093/ije/dyt261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Das D, Cherbuin N, Butterworth P, Anstey KJ, Easteal S. A population-based study of attention deficit/hyperactivity disorder symptoms and associated impairment in middle-aged adults. PLoS One. 2012;7:e31500. doi: 10.1371/journal.pone.0031500. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Vörös Z, Lukovszki L. The effects of subclinical ADHD symptomatology on the subjective financial, physical, and mental well-being of entrepreneurs and employees. J Bus Venturing Insights. 2021;15:e00240 [Google Scholar]
- 4.Wootton RE, Riglin L, Blakey R, Agnew-Blais J, Caye A, Cadman T, et al. Decline in attention-deficit hyperactivity disorder traits over the life course in the general population: Trajectories across five population birth cohorts spanning ages 3 to 45 years. Int J Epidemiol. 2022;51:919–30. doi: 10.1093/ije/dyac049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Demontis D, Walters RK, Martin J, Mattheisen M, Als TD, Agerbo E, et al. Discovery of the first genome-wide significant risk loci for attention deficit/hyperactivity disorder. Nat Genet. 2019;51:63–75. doi: 10.1038/s41588-018-0269-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kim JH, Kim JY, Lee J, Jeong GH, Lee E, Lee S, et al. Environmental risk factors, protective factors, and peripheral biomarkers for ADHD: an umbrella review. Lancet Psychiatry. 2020;7:955–70. doi: 10.1016/S2215-0366(20)30312-6. [DOI] [PubMed] [Google Scholar]
- 7.Heinonen K, Räikkönen K, Pesonen AK, Andersson S, Kajantie E, Eriksson JG, et al. Behavioural symptoms of attention deficit/hyperactivity disorder in preterm and term children born small and appropriate for gestational age: A longitudinal study. BMC Pediatr. 2010;10:91. doi: 10.1186/1471-2431-10-91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cecil CAM, Nigg JT. Epigenetics and ADHD: Reflections on Current Knowledge, Research Priorities and Translational Potential. Mol Diagn Ther. 2022;26:581–606. doi: 10.1007/s40291-022-00609-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cecil CAM, Neumann A, Walton E. Epigenetics applied to child and adolescent mental health: Progress, challenges and opportunities. JCPP Adv. 2023;3:e12133. doi: 10.1002/jcv2.12133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Neumann A, Walton E, Alemany S, Cecil C, González JR, Jima DD, et al. Association between DNA methylation and ADHD symptoms from birth to school age: a prospective meta-analysis. Transl Psychiatry. 2020;10:398. doi: 10.1038/s41398-020-01058-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Walton E, Pingault JB, Cecil CAM, Gaunt TR, Relton CL, Mill J, et al. Epigenetic profiling of ADHD symptoms trajectories: A prospective, methylome-wide study. Mol Psychiatry. 2017;22:250–6. doi: 10.1038/mp.2016.85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Oblak L, van der Zaag J, Higgins-Chen AT, Levine ME, Boks MP. A systematic review of biological, social and environmental factors associated with epigenetic clock acceleration. Ageing Res Rev. 2021;69:101348. doi: 10.1016/j.arr.2021.101348. [DOI] [PubMed] [Google Scholar]
- 13.McGill MG, Pokhvisneva I, Clappison AS, McEwen LM, Beijers R, Tollenaar MS, et al. Maternal Prenatal Anxiety and the Fetal Origins of Epigenetic Aging. Biol Psychiatry. 2022;91:303–12. doi: 10.1016/j.biopsych.2021.07.025. [DOI] [PubMed] [Google Scholar]
- 14.Suarez A, Lahti J, Czamara D, Lahti-Pulkkinen M, Knight AK, Girchenko P, et al. The Epigenetic Clock at Birth: Associations With Maternal Antenatal Depression and Child Psychiatric Problems. J Am Acad Child Adolesc Psychiatry. 2018;57:321–8.:e2. doi: 10.1016/j.jaac.2018.02.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Simpkin AJ, Howe LD, Tilling K, Gaunt TR, Lyttleton O, McArdle WL, et al. The epigenetic clock and physical development during childhood and adolescence: Longitudinal analysis from a UK birth cohort. Int J Epidemiol. 2017;46:549–58. doi: 10.1093/ije/dyw307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Bright HD, Howe LD, Khouja JN, Simpkin AJ, Suderman M, O’Keeffe LM. Epigenetic gestational age and trajectories of weight and height during childhood: A prospective cohort study. Clin Epigenetics. 2019;11:194. doi: 10.1186/s13148-019-0761-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Bell CG, Lowe R, Adams PD, Baccarelli AA, Beck S, Bell JT, et al. DNA methylation aging clocks: Challenges and recommendations. Genome Biol. 2019;20:249. doi: 10.1186/s13059-019-1824-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mooney MA, Ryabinin P, Wilmot B, Bhatt P, Mill J, Nigg JT. Large epigenome-wide association study of childhood ADHD identifies peripheral DNA methylation associated with disease and polygenic risk burden. Transl Psychiatry. 2020;10:8. doi: 10.1038/s41398-020-0710-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Arpawong TE, Klopack ET, Kim JK, Crimmins EM. ADHD genetic burden associates with older epigenetic age: mediating roles of education, behavioral and socio-demographic factors among older adults. Clin Epigenetics. 2023;15:67. doi: 10.1186/s13148-023-01484-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Monasso GS, Küpers LK, Jaddoe VWV, Heil SG, Felix JF. Associations of circulating folate, vitamin B12 and homocysteine concentrations in early pregnancy and cord blood with epigenetic gestational age: the Generation R Study. Clin Epigenetics. 2021;13:95. doi: 10.1186/s13148-021-01065-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Bohlin J, Håberg SE, Magnus P, Reese SE, Gjessing HK, Magnus MC, et al. Prediction of gestational age based on genome-wide differentially methylated regions. Genome Biol. 2016;17:207. doi: 10.1186/s13059-016-1063-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Haftorn KL, Lee Y, Denault WRP, Page CM, Nustad HE, Lyle R, et al. An EPIC predictor of gestational age and its application to newborns conceived by assisted reproductive technologies. Clin Epigenetics. 2021;13:82. doi: 10.1186/s13148-021-01055-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Knight AK, Craig JM, Theda C, Bækvad-Hansen M, Bybjerg-Grauholm J, Hansen CS, et al. An epigenetic clock for gestational age at birth based on blood methylation data. Genome Biol. 2016;17:206. doi: 10.1186/s13059-016-1068-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kooijman MN, Kruithof CJ, van Duijn CM, Duijts L, Franco OH, van IJzendoorn MH, et al. The Generation R Study: design and cohort update 2017. Eur J Epidemiol. 2016;31:1243–64. doi: 10.1007/s10654-016-0224-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Schreuder P, Alsaker E. The Norwegian mother and child cohort study (MoBa) – MoBa recruitment and logistics. Norsk Epidemiol. 2014;24:23–7. [Google Scholar]
- 26.Magnus P, Irgens LM, Haug K, Nystad W, Skjærven R, Stoltenberg C, et al. Cohort profile: The Norwegian Mother and Child Cohort Study (MoBa) Int J Epidemiol. 2006;35:1146–50. doi: 10.1093/ije/dyl170. [DOI] [PubMed] [Google Scholar]
- 27.Boyd A, Golding J, Macleod J, Lawlor DA, Fraser A, Henderson J, et al. Cohort profile: The ‘Children of the 90s’-The index offspring of the avon longitudinal study of parents and children. Int J Epidemiol. 2013;42:111–27. doi: 10.1093/ije/dys064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Fraser A, Macdonald-wallis C, Tilling K, Boyd A, Golding J, Davey smith G, et al. Cohort profile: The avon longitudinal study of parents and children: ALSPAC mothers cohort. Int J Epidemiol. 2013;42:97–110. doi: 10.1093/ije/dys066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Joubert BR, den Dekker HT, Felix JF, Bohlin J, Ligthart S, Beckett E, et al. Maternal plasma folate impacts differential DNA methylation in an epigenome-wide meta-analysis of newborns. Nat Commun. 2016;7:10577. doi: 10.1038/ncomms10577. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Joubert BR, Felix JF, Yousefi P, Bakulski KM, Just AC, Breton C, et al. DNA Methylation in Newborns and Maternal Smoking in Pregnancy: Genome-wide Consortium Meta-analysis. Am J Hum Genet. 2016;98:680–96. doi: 10.1016/j.ajhg.2016.02.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Mulder RH, Neumann A, Cecil CAM, Walton E, Houtepen LC, Simpkin AJ, et al. Epigenome-wide change and variation in DNA methylation in childhood: Trajectories from birth to late adolescence. Hum Mol Genet. 2021;30:119–34. doi: 10.1093/hmg/ddaa280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Håberg SE, Page CM, Lee Y, Nustad HE, Magnus MC, Haftorn KL, et al. DNA methylation in newborns conceived by assisted reproductive technology. Nat Commun. 2022;13:1896. doi: 10.1038/s41467-022-29540-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Lehne B, Drong AW, Loh M, Zhang W, Scott WR, Tan ST, et al. Erratum to A coherent approach for analysis of the Illumina HumanMethylation450 BeadChip improves data quality and performance in epigenome-wide association studies. Genome Biol. 2015;17:73. doi: 10.1186/s13059-015-0600-x. [Genome Biol. (2015);16:37] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Min JL, Hemani G, Davey Smith G, Relton C, Suderman M. Meffil: Efficient normalization and analysis of very large DNA methylation datasets. Bioinformatics. 2018;34:3983–9. doi: 10.1093/bioinformatics/bty476. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Paltiel L, Anita H, Skjerden T, Harbak K, Bækken S, Nina Kristin S, et al. The biobank of the Norwegian mother and child cohort study – Present status. Norsk Epidemiol. 2014;24:29–35. [Google Scholar]
- 36.Pidsley R, Wong YCC, Volta M, Lunnon K, Mill J, Schalkwyk LC. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics. 2013;14:293. doi: 10.1186/1471-2164-14-293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Girchenko P, Lahti J, Czamara D, Knight AK, Jones MJ, Suarez A, et al. Associations between maternal risk factors of adverse pregnancy and birth outcomes and the offspring epigenetic clock of gestational age at birth. Clin Epigenetics. 2017;9:49. doi: 10.1186/s13148-017-0349-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Gervin K, Salas LA, Bakulski KM, Van Zelm MC, Koestler DC, Wiencke JK, et al. Systematic evaluation and validation of reference and library selection methods for deconvolution of cord blood DNA methylation data. Clin Epigenetics. 2019;11:125. doi: 10.1186/s13148-019-0717-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Achenbach TM, Ruffie TM. The child behavior checklist and related forms for assessing behavioral/emotional problems and competencies. Pediatr Rev. 2000;21:265–71. doi: 10.1542/pir.21-8-265. [DOI] [PubMed] [Google Scholar]
- 40.Goodman R. The Strengths and Difficulties Questionnaire_ A Research Note - Goodman − 1997 - Journal of Child Psychology and Psychiatry - Wiley Online Library. J Child Psychol Psychiat. 1997;38:581–6. doi: 10.1111/j.1469-7610.1997.tb01545.x. [DOI] [PubMed] [Google Scholar]
- 41.Nøvik TS. Validity of the child behaviour checklist in a Norwegian sample. Eur Child Adolesc Psychiatry. 1999;8:247–54. doi: 10.1007/s007870050098. [DOI] [PubMed] [Google Scholar]
- 42.Shaffer D, Fisher P, Lucas CP, Dulcan MK, Schwab-Stone ME. NIMH Diagnostic Interview Schedule for Children Version IV (NIMH DISC-IV): description, differences from previous versions, and reliability of some common diagnoses. J Am Acad Child Adolesc Psychiatry. 2000;39:28–38. doi: 10.1097/00004583-200001000-00014. [DOI] [PubMed] [Google Scholar]
- 43.Van Mil NH, Steegers-Theunissen RPM, Bouwland-Both MI, Verbiest MMPJ, Rijlaarsdam J, Hofman A, et al. DNA methylation profiles at birth and child ADHD symptoms. J Psychiatr Res. 2014;49:51–9. doi: 10.1016/j.jpsychires.2013.10.017. [DOI] [PubMed] [Google Scholar]
- 44.Goodman R, Scott S. Comparing the Strengths and Difficulties Questionnaire and the child behavior checklist: Is small beautiful? J Abnorm Child Psychol. 1999;27:17–24. doi: 10.1023/a:1022658222914. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.