key: cord-0430023-rt344ejh
authors: Gidziela, A.; Ahmadzadeh, Y.; Michelini, G.; Allegrini, A.; Agnew-Blais, J.; Lau, L. Y.; Duret, M.; Procopio, F.; Daly, E.; Ronald, A.; Rimfeld, K.; Malanchini, M.
title: Genetic influences on neurodevelopmental disorders and their overlap with co-occurring conditions in childhood and adolescence: A meta-analysis.
date: 2022-02-21
journal: nan
DOI: 10.1101/2022.02.17.22271089
sha: 52a3d35dae0453afbf07834bafed96d1ad02e7f1
doc_id: 430023
cord_uid: rt344ejh

A systematic understanding of the aetiology of neurodevelopmental disorders (NDDs), their co-occurrence, and co-occurrence with other conditions during childhood and adolescence remains incomplete. This meta-analysis bridges gaps in our knowledge. First, we meta-analysed the literature on the relative contribution of genetic and environmental factors to NDDs. Second, we considered the literature on the overlap between different NDD categories. Lastly, we synthesized the literature on the co-occurrence between NDDs and disruptive, impulse control and conduct disorders (DICCs). We performed multilevel, random-effects meta-analyses on 296 independent studies, including over 4 million children and adolescents. We found all NDDs to be substantially heritable (family-based heritability (h2) = 0.66; SNP h2 = 0.19). Meta-analytic genetic correlations between NDDs, and between NDDs and DICCs were moderate to strong. However, given the paucity of available studies covering the co-occurrence of NDDs and DICCs, these could only be estimated for a few disorders. While our work provides direct evidence to inform and potentially guide clinical and educational diagnostic procedures and practice, it also highlights the imbalance in the research effort that has characterized developmental genetics research.

Neurodevelopmental disorders (NDDs) are a serious and complex health concern, starting from childhood 1 . 36 NDDs affect around 15% of children and adolescents worldwide and lead to impaired cognition, 37 communication, adaptive behavior, and psychomotor skills 2 . The fifth edition of the Diagnostic and 38

Statistical Manual of Mental Disorders (DSM-5) categorizes the following seven disorders under NDDs: 39

intellectual disabilities, communication disorders, autism spectrum disorder (ASD), attention- 40 deficit/hyperactivity disorder (ADHD), specific learning disorders, motor disorders and other 41 neurodevelopmental disorders 3 .NDDs often have lifelong trajectories: they can manifest as early as before 42 the child reaches 12 months of age 4 and can be identified and diagnosed before children enter primary 43 education 5,3 . While some NDDs (e.g. ASD and ADHD) may persist throughout adolescence and 44 adulthood 6,7 , others are more likely to alleviate as children get older (e.g., tic disorder 8 and communication 45 disorders 9 ); nevertheless, all NDDs can lead to social and behavioural difficulties and reduced, or even a 46 lack of, independence over the lifespan 6,7 . 47 48

A systematic understanding of the aetiology of NDDs remains incomplete. A disproportionate number of 49 studies, reviews, and syntheses of extant literature have focused on ASD and ADHD. However, other 50 neurodevelopmental conditions, despite showing similar prevalence rates and severity as ASD and ADHD, 51 are less well understood and studied 10 . The focus on ASD and ADHD has resulted in several studies that 52

have systematically synthesized the literature on the aetiology of these two NDDs, pointing to their 53 substantial heritability-the extent to which observed individual differences are accounted for by underlying 54 genetic differences. 55 56

A meta-analysis of 7 twin studies of clinically diagnosed ASD in childhood and adolescent samples (aged 2 57 to 23 years) yielded a grand heritability estimate of 0.74 (95% CIs= 0.70, 0.87) 11 . Similarly sizeable 58 heritability estimates also emerged from a meta-analysis of 26 studies of ADHD in childhood and 59 adolescence, which yielded a pooled estimate for additive genetic effects -independent effects of genetic 60 variants-of 0.26 (95% CIs= 0.20, 0.32) and a pooled estimate for dominant genetic effectsinteractive 61 effects between genetic variants at one more loci-of 0.44 (95% CIs= 0.38, 0.51) 12 . A subsequent systematic 62 review of 37 twin studies of ADHD, including studies that had adopted either categorical or dimensional 63 measures, yielded a mean h 2 of 0.74 13 . Heritability estimates were found to differ across the two major 64 components of ADHD, with genetic factors playing a more substantial role in the aetiology of hyperactivity 65 (h 2 = 0.71, 95% CIs= 0.63, 0.75; based on 9 studies), if compared inattention (h 2 = 0.56, 95% CIs= 0.48, 0.63; 66 based on 13 studies) 14 . 67 68

In line with what observed for all complex traits, heritability estimates for ASD and ADHD obtained from 69 DNA data are lower than those obtained from twin and family designs 15 , likely due, at least in part, to the 70 additive models used to calculate heritability from single nucleotide polymorphism (SNP) 16 . SNP 71 heritability can be calculated using large samples of individual-level genotype data 17 or summary statistics 72 from genome-wide association studies (GWAS) 18 , hypothesis-free studies aimed at discovering associations 73 between genetic variation across the genome and individual differences in traits and disorders. The two 74

Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 4 Usually Diagnosed in Infancy, Childhood, and Adolescence (DSM-IV) 35 , to its current classification under 116 the Neurodevelopmental Disorders category (DSM-5) 3 . 117 118

The most investigated example of symptom overlap between NDDs and DICCs involve ADHD and conduct 119 disorder 36,37 , and ADHD and oppositional defiant disorder 38 . Studies reporting on the behavioural and 120 cognitive profiles of ADHD and these disorders highlight how both disorders are characterised by 121 disturbances in emotion regulation, attention problems, cognitive inflexibility, executive functioning, and 122 impaired inhibition 37,39, 40 . A shared symptomatology has also been observed between ASD and antisocial 123 behaviour/personality disorder (that we refer to as conduct disorder in the current work due to antisocial 124 personality disorder referring to adult diagnosis) 3,41,42 , with both disorders characterized by impairments in 125 social reciprocity, expressing emotions and affective empathy 43, 44 . Symptom resemblance that characterizes 126 these disorder pairs is reflected in the phenotypic correlations between ADHD and conduct disorder (0.40) 38 127

and between ADHD and oppositional defiant disorder (0.55) 38 , but not ASD and conduct disorder traits, for 128 which the phenotypic correlation is modest (0.22 between social impairments and callous-unemotional traits 129 and 0.21 between communication impairments and callous-unemotional traits) 45 . 130 131

Individual studies on the association between NDDs and DICCs are characterized by a great deal of 132 heterogeneity and inconsistencies across co-occurring conditions. For example, one study found that, in a 133 sample of 7-year-olds, genetic effects on ASD and callous-unemotional traits, one of the manifestations of 134 conduct disorder in childhood and of antisocial personality disorder in adulthood, were largely independent, 135

as indicated by small to moderate genetic correlations (0.23 (95% CIs= 0.16, 0.31) between the 136 communication impairments domain of ASD and callous-unemotional traits, and 0.31 (95% CIs= 0.26, 0.36) 137

between the social interaction impairments domain of ASD and callous-unemotional traits) 45 . Another study 138

reported moderate genetic overlap of 0.43 (95% CIs= 0.34, 0.52) between ASD and psychopathic tendencies 139 in 9-year-olds of 0.99 (95% CIs= 0.92, 1) 43 . 140 141

With three core aims (Figure 1) , the current meta-analysis bridges major gaps in our knowledge of the 142 aetiology of NDDs and of their co-occurrence with other developmental conditions. First, we meta-analysed 143 studies on the relative contribution of genetic and environmental influences to all NDDs categories 144 described in the DSM-5. Second, we meta-analysed estimates for the genetic and environmental 145 comorbidities between different NDDs (homotypic co-occurrences). Third, given the historical associations 146

between NDDs and DICCs, and their developmental onset and progression, we examined the aetiology of 147 the co-occurrence between NDDs and DICCs (heterotypic co-occurrences). 148 149

In addition to addressing each disorder individually, we take a transdiagnostic approach by combining data 150 across NDDs, their manifestations, and including categorical (i.e., presence or absence of a disorder) and 151 quantitative (i.e., continuously measured symptoms) measures. Taking a transdiagnostic approach provides  152 us with a holistic picture of the extent to which genetic and environmental factors contribute to NDDs, to 153 their co-occurrence, and to their co-occurrence with other common developmental conditions. This will 154 result in a clearer appreciation of the size, strength, pervasiveness, and developmental progression of the 155 associations between different developmental disorders. 156

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Studies for this meta-analysis were selected during 3 screening stages including title and abstract screening, 185 full text screening, and reference list screening (see Method for a detailed description). This selection 186 process resulted in a total of 296 studies included in the current meta-analysis (Figure 2 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint 

When compared to the vast number of studies that had examined the aetiology of individual differences in 224 each NDD, only a limited body of research (37 studies) had investigated the co-occurrence between NDDs 225 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 8 in childhood and adolescence applying genetically informative methods. In fact, for some of the disorders, 226 we were unable to find two independent statistics 49 , and therefore could not provide a meta-analytic 227 estimate. 228 229

Family-based genetic correlations (rA) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 9 1 study that examined the co-occurrence between specific learning disorders & disruptive behaviour, finding 265 a weak negative genetic correlation (rA= -0.14, SE= 0.06) 51 . 266 267

Family 283

NDDs do not affect males and females equally, males are four times more likely to be diagnosed with 285

NDDs 52,53 . Studies have suggested that these differences in prevalence may be caused by quantitative 286 genetic sex differences, differences in the degree to which genes influence variation in NDDs in males 287 versus females 54 . To provide an overview of sex differences in NDDs, we conducted separate meta-analyses 288 including all studies that had reported sex-specific estimates. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ;

We identified 68 family-based studies that investigated the genetic aetiology of individual differences in 291

NDDs in male samples and 67 studies that reported estimates for female samples. Out of all studies 292

involving sex-stratified samples, 38 studies focused on ADHD, 21 studies on ASD, 8 studies on specific 293

learning disorders, 4 studies on communication disorders and 2 studies on motor disorders. Across all 294

NDDs, family-based heritability was not significantly different between males and females (h 2 = 0.65, SE= 295 0.06 in males and 0.67, SE= 0.06 in females). Distributions of sex-specific family-based variance 296 components for all NDDs, except for motor disorders for which a sufficient number of studies (>1) was not 297

identified, are presented in Figure 5 and Supplementary disorder and only in in females. In fact, one out of the only two studies that examined the sex-specific co-323 occurrence between ADHD and conduct disorders used a female-only sample. Hence, we could only meta-324

analyse the co-occurrence between ADHD & conduct disorder in females. We found a meta-analytic genetic 325 correlation of 0.75 (SE= 0.58) (Supplementary Table 15 ).

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. 

Genetic influences on NDDs are stable over development 333 We investigated developmental change and continuity in the relative contribution of genetic factors to NDDs 334 by examining age-related differences in their aetiology and sources of their homotypic and heterotypic co-335 occurrences. We distinguished between the three following developmental stages: childhood (4-7 years), 336 middle childhood (8-10 years) and adolescence (11-24 years is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ;

Out of a total of 29 SNP-based studies that were identified, 13 included adolescent samples, 7 samples in 357 middle childhood and 6 samples in childhood, while 11 studies reported estimates across childhood & 358 adolescence. SNP heritability was stable developmentally across NDDs, and the developmental trajectory 359 mirrored that of family-based heritability (SNP h 2 = 0.24, SE= 0.11 in childhood; 0.26, SE= 0.08 in middle 360 childhood and 0.23, SE= 0.07 in adolescence) ( Figure 6B; Supplementary Table 16 ). For ASD, ADHD 361

and specific learning disorders, the specific NDDs for which grand estimates could be calculated, the 362 developmental trends were consistent with those observed for family-based heritability ( Figure 6B ; 363

Supplementary is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 14 Geographical differences in genetic overlap between NDDs: Family-based genetic correlations (rA) 421

Sources of homotypic co-occurrence with NDDs were investigated in 37 independent family-based studies, 422 out of which the majority was conducted in the United Kingdom (49%) and United States (30%). The 423

highest genetic correlation across all co-occurrences was estimated in Swedish cohorts (0.80, SE= 0.26 424 across 3 studies), while the lowest grand genetic overlap was estimated in Canadian samples (-0.44, SE= 425 0.24 across only 2 studies which investigated the association between ADHD and specific learning 426 disorders; Figure 7B ; Supplementary is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. 

Bias and heterogeneity assessment 525 We applied I 2 statistics to assess heterogeneity in the estimates. The results of these analyses are reported in 526

Supplementary is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 18 exclusive research focus on ASD and ADHD, and effectively synthesizes our knowledge across all NDDs. 536

Second, by providing an account of the genetic overlap between NDDs, we highlight how genetic influences 537 are implicated in the co-occurrence between multiple NDDs, identifying patterns of shared aetiological 538 liability between neurodevelopmental conditions. Third, by providing a synthesis of the literature on the co-539 occurrence between NDDs and DICCs we highlight how these two separate groups of disorders identified 540 by the DSM-5 share as much of their genetic aetiology as do disorders all classified as NDDs. 541 542

Our work provides meta-analytic evidence for the substantial heritability of all neurodevelopmental 543 conditions, particularly when considering family-based studies, which indicated that around two thirds of the 544 variation in NDDs is accounted for by genetic differences between children and adolescents. Heritability 545 estimates differed, albeit not significantly, between NDDs, with the highest estimates obtained for 546 intellectual disabilities and motor disorders, and the lowest estimates, despite still accounting for over 60% 547 of the variation, obtained for specific learning disorders and communication disorders. 548 549

Although males are four times more likely to be diagnosed with NDDs than females 52,53 , we showed that, 550 when meta-analysed, genetic effects on NDDs do not differ by sex. We also showed that genetic sources of 551 variation in NDDs are remarkably stable across developmental stages, and this developmental stability was 552 observed across all NDDs, irrespectively of the relative contribution of genetic factors to each disorder. 553

Genetic effects were also mostly consistent when we separated studies that had considered diagnoses and 554 clinical cut-offs from studies that had quantified NDDs as continuous traits. 555 556

Interestingly, we found that the genetic contributions to NDDs differed substantially as a function of 557 geography. This highlights how estimates of genetic effects on disorders are sensitive to different 558 environmental contexts. The high sensitivity to context of genetic effects has been observed for several other 559 behavioural traits, including educational attainment, for which estimates of genetic effects not only differ as 560 a function of geography, but also as a function of different historical times and political contexts 58,59 . 561 562

Our work on geographical differences on genetic effects on NDDs also highlighted the major gap in our 563 knowledge of the aetiology of NDDs in non-Western countries, a gap that is only exceeded by the lack of 564 ancestral diversity observed across all studies of NDDs. Importantly, the current study pointed to how 565 genetic influences on NDDs were substantially reduced in more ancestrally diverse samples, again 566

highlighting how heritability estimates are inextricably linked to our social context 60,61 , in a sense that 567 increased ancestral homogeneity within the sample likely entails increased environmental homogeneity, 568 reducing the environmental influences and inflating heritability in these populations. 569 570

The lack of diversity in genetic research remains its most striking limitation to date, particularly when 571 considering DNA-based methods, limiting the extension of genetic findings to the entire population 62,63 . 572

Limited research resources in under-represented populations are likely to have profound cascading effects 573 for future advances in clinical practice, including pharmacological and behavioural treatment. Fortunately, 574

there are major initiatives underway to re-balance these biases 64, 65, 66 . 575 576

Heritability estimates obtained from studies with SNP-based designs were overall modest, and did not differ 577 significantly between disorders, although the degree of precision around these estimates varied substantially 578 between conditions. DNA-based heritability estimates could not be obtained for motor disorders or 579 intellectual disabilities. The gap in the heritability estimated from family studies and DNA-based studies is 580 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 19 known as missing heritability 67 , and is observed for virtually all complex human traits, from height to 581 cognitive ability, to NDDs. The current inability of GWA studies to tag interactive effects between genes 582 (epistatic effect), and between genes and environments (GE interaction) may account for this missing 583 heritability gap 68 . It is also possible that rare variants that are not tagged by SNP arrays commonly used in 584

GWA studies play a substantial role 69 . Lastly, it has been proposed that family-based heritability estimates 585 might be inflated 70 . 586 587

Our second aim was that of providing a clear account of how close NDDs are to one another aetiologically.

We found that, while meta-analytic estimates indicated moderate genetic overlap, the degree of 589 heterogeneity in these associations across disorders was large. We found a substantial genetic correlation 590 between ASD and ADHD, ADHD and motor disorders, and communication disorders and specific learning 591 disorders. On the other hand, genetic overlap was only moderate between communication disorders and 592 motor disorders, and very weak between ADHD and specific learning disorders, which is consistent with the 593 degree of symptom resemblance across these disorders. 594 595

Although we were able to explore general patterns of variation and co-occurrence, the aetiology of specific 596 NDDs and of their associations could not be comprehensively characterised. The research gaps that we 597

identified highlight an imbalance in focus across NDDs in developmental behavioural genetic research. 598

When considering our first aim, we could only identify 2 family-based studies that investigated the genetic 599

contributions to intellectual disabilities, if compared to 121 family-based and 14 SNP-based studies 600 identified for ADHD, and 36 family-based and 15 SNP-based studies identified for ASD. This lack of 601 research on intellectual disabilities, a neurodevelopmental disorder affecting 2.5% of children in the United 602

Kingdom 71,72 more than double the prevalence rate of ASD 73 is reflected in, and likely partly due to, the lack 603 of funding bodies devoted to researching NDDs other than ASD and ADHD, as well as a lack of publicly 604 available data repositories and resources (e.g., 74,75,76 ). 605 606 We also identified very few studies that had examined the aetiology of motor disorders, another 607 neurodevelopmental condition showing significant prevalence rates of 5-6% in school aged children 77 . This 608 unbalanced research focus, that extends far beyond genetically informative research to touch developmental 609 and therapeutic research 78,79,80,81 , has led to an uneven distribution of knowledge, which could lead to limited 610 access to interventions for children with NDDs other than ASD, ADHD, and Dyslexia 82 . 611 612

Considering our second aim, no genetically informative studies were identified that had examined the co-613 occurrence between intellectual disabilities and all other NDDs, and between ADHD & communication 614

disorders. Only one study was identified on the genetic overlap between ASD & communication disorders 33 , 615

ASD & specific learning disorders 27 , ASD & motor disorders 27 and between specific learning disorders & 616 motor disorders 27 ; therefore, preventing us from meta-analysing these NDD pairs separately. 617 618

The lack of equity in focus across NDDs was pronounced in analyses addressing our third aim. Sources of 619 co-occurrence between NDDs and DICCs could only be investigated between ADHD & conduct disorder, 620

ADHD & oppositional defiant disorder and between ASD & conduct disorder. Considering that in the DSM-621 5 the DICCs category comprises 8 distinct disruptive disorders, this highlights a major gap in our knowledge 622 of the role that genetic and environmental factors play in the co-occurrence between all NDDs and DICCs. 623 We hope that highlighting these major knowledge gaps will lead to future research efforts aimed at making 624 developmental genetic research more diverse, consequently allowing for greater inclusion opportunities. 625

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 20 626

To conclude, this meta-analysis extends our knowledge of NDDs in several directions. First, it provides a 627 holistic view of genetic and environmental contributions to all NDDs and commonly co-occurring disorders 628 over development, revealing that NDDs are just as strongly genetically correlated with other NDDs, as they 629 are with DICCs. Second, it identifies a lack of balance in the research efforts that have been invested across 630 different disorders, which calls for future genetic research to focus on less investigated NDDs. Third, it 631

provides knowledge about patterns of aetiological co-occurrence between NDDs, as well as between NDDs 632

and DICCs, which we hope will inform clinical and educational diagnostics and practice, resulting for 633 example in expanded diagnostic screening. 634 635 Methods

The protocol for the current meta-analysis was registered with the international prospective register of 637 systematic reviews (PROSPERO) and can be accessed at the following link: Screening and inclusion criteria 651

After the initial searches were conducted and duplicate studies removed, 8,087 studies met the criteria for 652 the first stage of screening, which involved title and abstract scanning. All titles and abstracts were screened 653 by two independent, blinded reviewers to ensure inter-rater agreement. Conflicts were resolved by a third 654 independent reviewer. After this initial screening phase, 6,834 studies were excluded as deemed not relevant 655

for the purpose of the current meta-analysis. 656 657

The title and abstract screening process resulted in a total of 1,253 potentially eligible studies. The full text 658 of each study was screened by two independent, blinded reviewers. Reviewer discrepancies were identified 659 and resolved by a third independent reviewer. This resulted in 289 eligible articles. In addition, during full 660 text screening, relevant review articles, meta-analyses, editorials, and conference abstracts were flagged to 661 aid the potential discovery of further relevant studies by either screening the References sections or 662

contacting the authors of conference abstracts. Through this process 7 additional studies were identified, 663

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. 21 which resulted in a total of 296 studies included in the current meta-analysis (see Figure 2 dyslexia), specific phenotypes measured (e.g., reading fluency), measure statistics (e.g., binary (diagnosis) 704 or continuous (symptoms continua)), measure (e.g., Conners rating scale for ADHD) and rater (e.g., parent 705 reports), covariates included in the analyses (e.g., age and sex), statistics (e.g., family-based heritability, 706

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint studies that provided sex-specific estimates in separate models to minimize sample heterogeneity across 725 studies and report separate grand estimates for combined, male-only, and female-only samples. 726

Aggregation of non-independent effects 727

Multilevel meta-analytic models allow to account for non-independence of estimates derived from partly or 728 completely overlapping samples (i.e., estimates obtained from multiple studies that have used the same 729 cohort of participants). To further account for the non-independence of sampling variance (i.e., when 730 sampling errors correlate because data from partly the same individuals is used to estimate multiple effect 731 sizes), we also aggregated multiple estimates within each individual study (e.g., estimates at multiple 732 timepoints derived from the same study). Aggregation of dependent effects sizes was performed at the level 733 of each study using the R package Meta-Analysis with Mean Differences 97,92 (MAd), applying a default 734 correlation between estimates of 0.5. We conducted several sensitivity analyses, comparing different 735 aggregation methods, i.e., aggregating at the level of the study, cohort, and country, and varying the 736 assumed correlation between dependent effect sizes (0.5, 0.3 and 0.9). Results of these additional checks are 737 presented in Supplementary Figure 24 and discussed in Supplementary Note 9. Since differences in 738 aggregation strategy did not result in significant differences in meta-analytic effects, we report results 739 obtained when the correlation between dependent effect sizes was set to 0.5. 740

Bias and heterogeneity assessment 741

The potential for publication bias was explored using funnel plots and Egger's linear regression 98 . 742

The proportion of heterogeneity across estimates was estimated using the I 2 statistics, which calculates the 743 fraction of variance across studies that can be attributed to heterogeneity, rather than chance 99,100 . The I 2 744

. CC-BY-NC-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint where VTRUE is the variation of true effects and VOBS is the variation due to sampling error. In other words, I 2 749 can be interpreted as the dispersion of observed effects as compared to the dispersion that would be 750 predicted just from sampling error 101 . The I 2 statistics also provides insight into the degree to which 751 confidence intervals from individual studies are independent 25,101 . 752

Moderation analyses 753 We tested for the effect of several moderators. Selection of moderator terms was determined based on 754 available data, considering completeness of reported moderator variables. We implemented a >50% rule of 755 thumb, i.e., if 50% or more studies reported data on the moderating variable, we included this moderator in 756 our analyses. For example, less than 50% of studies reported the percentage of participants of Asian ancestry 757

in the sample, hence we did not include the percentage of Asian participants in moderation analyses. We 758 considered the following 11 moderators: age group, design, type of model, rater, measurement, percentage 759 of individuals who identified as White, number of covariates included in the analysis, measure adopted, 760 country, and specific phenotype measured, each moderator is described in greater detail in Supplementary  761 Note 3. Moderation analyses were conducted using a two-step procedure. First, only studies that reported 762 data on the level of the moderator were selected (for example, only studies reporting estimates for 763 adolescents). Second, analyses stratified by levels of the moderator were run using a multilevel random-764 effects meta-analysis in metafor for R 49,92 , i.e., a grand estimate was derived for adolescents and 765

subsequently compared with estimates for other developmental stages (i.e., childhood and middle childhood) 766

using the same procedure. We report unstratified estimates (Supplementary Tables 1, 3 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted February 21, 2022. ; https://doi.org/10.1101/2022.02.17.22271089 doi: medRxiv preprint

Neurodevelopmental disorders

Principles and practices of neurodevelopmental assessment in children: lessons 776 learned from the Centers for Children's Environmental Health and Disease Prevention Research

Neurodevelopmental disorders: DSM-5® selections

Identification, evaluation, and management of children with 781 autism spectrum disorder

The ESSENCE in child psychiatry: early symptomatic syndromes eliciting 783 neurodevelopmental clinical examinations

Continuity and change from early childhood to adolescence in autism

The age-dependent decline of attention deficit hyperactivity 787 disorder: a meta-analysis of follow-up studies

Provisional Tic Disorder is not so transient

Early language delay and risk for language impairment

Which neurodevelopmental disorders get researched and why?

Heritability of autism spectrum disorders: a 794 meta-analysis of twin studies

Rethinking environmental contributions to child and adolescent psychopathology: a meta-796 analysis of shared environmental influences

Genetics of attention deficit hyperactivity disorder

Genetic and environmental influences on ADHD symptom dimensions of 800 inattention and hyperactivity: a meta-analysis

Childhood behaviour problems show the greatest gap between DNA-based and twin Running head: The aetiology of neurodevelopmental disorders and their overlap with common cooccurring developmental conditions

Cool" inferior frontostriatal dysfunction in attention-deficit/hyperactivity disorder versus 854 "hot" ventromedial orbitofrontal-limbic dysfunction in conduct disorder: a review

Aggression in low functioning children and adolescents with 857 autistic disorder

Sex differences in antisocial behaviour: Conduct 859 disorder, delinquency, and violence in the Dunedin Longitudinal Study. xvii

Phenotypic and aetiological associations between psychopathic tendencies, autistic 862 traits, and emotion attribution

Autism spectrum disorder and psychopathy: 864 shared cognitive underpinnings or double hit?

Examining the genetic and environmental associations between autistic social and 866 communication deficits and psychopathic callous-unemotional traits

Rethinking the nature of genetic vulnerability to autistic spectrum disorders

Parenting stress in 870 caregivers of young children with ASD concerns prior to a formal diagnosis

Risk in relatives, heritability, SNP-based 873 heritability, and genetic correlations in psychiatric disorders: A review

Package 'metafor'. Compr. R Arch. Netw. Package 'metafor' 876

The correlation between reading and mathematics ability at age twelve has a 878 substantial genetic component

Genetic and environmental influences on the co-occurrence of 880 early academic achievement and externalizing behavior

Prevalence and characteristics of autism spectrum disorder among children aged 882 8 years-autism and developmental disabilities monitoring network, 11 sites, United States

Autism spectrum disorder: updated prevalence and 885 comparison of two birth cohorts in a nationally representative Australian sample

A genetic investigation of sex bias in the prevalence of attention-deficit/hyperactivity 888 disorder

& Evaluation. Global Burden of Disease Collaborative Network

Whose genomics?

Genetic and environmental variation in educational attainment: an individual-894 based analysis of 28 twin cohorts

Genetic influence on social outcomes during and after the Soviet era in Estonia

Genetic correlates of social stratification in Great Britain

Genetics and the geography of health, behaviour and attainment

Genomics is failing on diversity

Clinical use of current polygenic risk scores may exacerbate health disparities

Cohort Profile: East London Genes & Health (ELGH), a community-based population 905 genomics and health study in British Bangladeshi and British Pakistani people

Cohort profile: the Born in Bradford multi-ethnic family cohort study

study, as members of the A.-G. & Consortium, the H. African partnerships 910 through the H3Africa Consortium bring a genomic dimension to longitudinal population studies on the 911 continent

Finding the missing heritability of complex diseases

Inclusion of gene-gene and gene-environment interactions unlikely to dramatically 915 improve risk prediction for complex diseases

Recovery of trait heritability from whole genome sequence data Visscher 917 2019

Solving the missing heritability problem

Estimates of the population for the

People with learning disabilities in England

Autism spectrum disorder. The British Medical Association is the trade union and 926 professional body for doctors in the UK

Developmental coordination disorder: a review 932 and update

A systematic review evaluating the implementation of technologies to assess, 934 monitor and treat neurodevelopmental disorders: A map of the current evidence

Implementation of telehealth services to assess, monitor, and treat 937 neurodevelopmental disorders: systematic review

How we fail children with developmental language disorder

The effectiveness of web-based 941 interventions delivered to children and young people with neurodevelopmental disorders: systematic 942 review and meta-analysis

Social inclusion and people with intellectual disability and challenging behaviour: A 944 systematic review

Preferred reporting items for systematic 946 reviews and meta-analyses: the PRISMA statement

Analytic approaches to twin data using structural equation models

A novel sibling-based design to quantify genetic and shared environmental effects: 950 application to drug abuse, alcohol use disorder and criminal behavior

A note on the statistical power in extended twin designs

The intergenerational transmission of anxiety: a children-of-twins study

Disentangling prenatal and inherited influences in humans with an experimental design

Revisiting the children-of-twins design: improving existing models for the 959 exploration of intergenerational associations

Signatures of negative selection in the genetic architecture of human complex traits

Package 'psychometric'. Recuperado Httpcran Rproject 963

Vienna Austria Available 965 Httpswww R-Proj

Applied multiple regression/correlation analysis for the 967 behavioral sciences

Statistical and empirical examination of the chi-969 square test for homogeneity of correlations in meta-analysis

Aggressive behaviour in childhood and adolescence: the role of smoking during 971 pregnancy, evidence from four twin cohorts in the EU-ACTION consortium

How to do a meta-analysis

Regression methods to detect publication and other bias in meta-analysis

Quantifying heterogeneity in a meta-analysis

Measuring inconsistency in meta-980 analyses