Discoveries on the Genetics of ADHD in the 21st Century: New Findings and Their Implications

Recent national registry-based family and twin studies have enabled the investigation of diagnoses rather than traits and have revealed a strong familial and genetic overlap between ADHD and ASD. In one Swedish study (9), the monozygotic co-twins of individuals with ASD were found to have an increased risk of ADHD (odds ratio=17.77, 95% CI=9.8–32.22) compared with dizygotic co-twins (odds ratio=4.33, 95% CI=3.21–5.86). These associations were most prominent for individuals with higher-functioning ASD rather than low-functioning ASD (with intellectual disability). Although the within-individual and within-family overlap of ADHD and ASD is now accepted, until the publication of DSM-5, a diagnosis of ADHD in the presence of autism was disallowed.

Recent genetic findings have also challenged the historical reticence about diagnosing ADHD in individuals with intellectual disability and about including those with lower IQ in genetic studies. It has always been recognized that ADHD is strongly associated with lower IQ and intellectual disability (10). Population-based twin studies of IQ and learning ability (e.g., reading ability) have consistently found that most of the correlation with ADHD symptoms was explained by shared heritability (10). Interestingly, even though the genetic overlaps between intellectual disability and autism (11) and between intellectual disability and schizophrenia are well recognized (12), investigations into the genetic links between intellectual disability and ADHD have been sparse. One recent registry-based family study addressed this gap by examining ADHD diagnosis and intellectual disability (13). The authors found that most of the correlation between ADHD and intellectual disability liabilities was explained by genetic factors (estimated at 91%) except in the case of profound intellectual disability. Again these findings point to the neurodevelopmental nature of ADHD, and they are important for scientists and clinicians. Historically, although individuals with lower IQ or intellectual disability have been excluded from studies of ADHD, this has not been the case for studies of schizophrenia and autism. This means that ADHD research studies, including gene discovery investigations, with some exceptions (14), have not included the full IQ spectrum. If individuals with ADHD who have comorbid intellectual disability or ASD are excluded from genetic studies, this is a potential problem, because it means that such investigations will not have been fully representative of the clinical population or included the most severely affected individuals.

With technological advances, reductions in cost, and large-scale collaborations, such that very large sample sizes are available, researchers have been able to investigate different classes of genetic variation across the whole genome. Early linkage and association studies revealed that ADHD, like other neuropsychiatric disorders, is in most instances a multifactorial disorder. It is not explained by Mendelian inheritance or common gene variants of large effect size. Recent molecular genetic studies of ADHD have shown that it is highly polygenic; that is, its genetic architecture is explained by thousands of common gene variants, each of small effect size (typical odds ratio, <1.1) (15), as well as by rare mutations, some of which have a much larger effect size (e.g., chromosome 16p13.11 duplication, odds ratio=13.88, 95% CI=2.3–82.2) (14).

From an epidemiological perspective, ADHD can be viewed as a trait as well as a category; higher scores are associated with adverse outcomes, with no discontinuity at a specific cut-point (8). Although a diagnostic category for ADHD is helpful in clinical practice because many clinical decisions are categorical in nature (e.g., using or not using medication), an understanding of its underlying structure can be informative for clinicians and patients as well as researchers.

Twin studies suggested many years ago that ADHD appears to lie at the end of a continuum of genetic risk, with the same genetic risks contributing across the ADHD continuum in the population (2), with the possible exception of those scoring at the extreme low end of the continuum (36). These findings, coupled with epidemiological studies, suggested that ADHD could be viewed as something akin to hypertension, where the disorder lies at the extreme of a continuously distributed phenotype (i.e., blood pressure) that is present in the whole population.

GWAS findings have further added to this evidence. Two studies used an independent ADHD GWAS discovery data set to generate ADHD polygenic risk scores (PRS)—the relative burden of ADHD common gene variant risk alleles carried by an individual—in population cohorts in the United Kingdom (37) and the Netherlands (38). Both found that ADHD PRS predicted ADHD trait levels in the general population. A third U.K. study (39) further observed that when PRS were derived for an ADHD trait measure from a population-based cohort, they predicted ADHD diagnosis in an independent patient sample. The largest and most recent ADHD GWAS (15) was able to go further by testing the genetic correlation between ADHD diagnosis and a large GWAS meta-analysis of ADHD trait measures in 17,666 European individuals from the EAGLE consortium (40). The findings are striking in that the investigators observed a genetic correlation of 0.94 between ADHD diagnosis in patients and ADHD traits in the population-based cohorts. Overall genetic findings converge in showing that ADHD diagnosis lies at the extreme of a quantitative trait.

However, genetic risk for ADHD diagnosis is not exclusively related to a single continuously distributed trait measure of ADHD. For example, one population-based cohort study (41) found that ADHD genetic risk scores predicted multiple childhood neurodevelopmental traits, including social communication, cognitive ability or IQ, language, and working memory, although not emotion recognition. The findings on working memory have recently been replicated (42). Another study found that ADHD genetic risk scores also predicted childhood irritability (43). These findings could simply reflect pleiotropy once again, with the same genes contributing to biological mechanisms that in turn contribute to different childhood traits in the population. An alternative interpretation for the clinician and scientist is that a diagnosis of ADHD could be conceptualized as the extreme of multiple liabilities or dimensions. A dimensional view of ADHD that is underpinned by genetic liabilities that also contribute to cognitive characteristics would be in keeping with the thinking of the Research Domain Criteria project. This framework encourages us to view ADHD and other mental disorders as dimensions that span the range from normal to abnormal and encompass multiple levels of information (e.g., genes, circuits). If ADHD behaves as a dimension, then that has implications for how clinicians view individuals with subthreshold ADHD symptoms, which will be discussed later.

Although ADHD symptom levels typically decline with age, a proportion of individuals show persistently elevated symptoms or a continued diagnosis into adolescence or adult life (8). Longitudinal twin studies (44) have all observed that ADHD persistence is associated with higher genetic loading; one study that used latent growth curve analysis to examine individual trajectories further suggested that ADHD trajectories through adolescence may be explained by genetic liabilities that are independent of those that contribute to baseline symptom levels in childhood (45). A recent population-based cohort investigation used ADHD GWAS findings to assess the contribution of ADHD common genetic variants to ADHD symptom trajectories between ages 4 and 17 (46). ADHD PRS were significantly higher in children in the persistent trajectory than in those in a low-symptom group (odds ratio=1.31, 95% CI=1.13–1.51), an intermediate group (odds ratio=1.22, 95% CI=1.01–1.47), and a childhood-limited ADHD group (odds ratio=1.27, 95% CI=1.05–1.53). Interestingly, although the schizophrenia GWAS discovery sample is very much larger and more powerful than the ADHD GWAS, schizophrenia PRS and other neuropsychiatric PRS did not predict ADHD symptom persistence. This highlights the fact that there is a degree of specificity to genetic findings.

Other genetic studies that have assessed adult ADHD in individuals over age 18 using cross-sectional designs also support the suggestion that persistent ADHD has higher genetic loading. For example, a recent Swedish registry-based family study observed a substantially higher relative risk of clinically recognized ADHD in siblings of those who had ADHD at age 18 or older (hazard ratio=11.49) than in siblings of those who had ADHD before age 18 (hazard ratio=4.68) (5).

However, twin studies of adult ADHD have repeatedly yielded very low heritability estimates, in contrast to the consistently high heritability estimates observed in child twin studies. The most likely explanation for this is the change in informant from parent to self- report in adult life, because self-reported ADHD symptoms have been found to show low heritability even in children (47). A Swedish registry study (48), however, found substantial heritability for ADHD across the life span, including in adult life (h²=0.72, 95% CI=0.56–0.84). Here, ADHD was defined on the basis of either an ICD diagnosis or prescribed ADHD medication. These findings suggest that with a clinician’s diagnosis, where perhaps detailed clinical assessments and impairment criteria are used rather than self-report questionnaire measures, the developmental differences in genetic architecture are not as pronounced, at least in terms of twin study findings.

Although there are international efforts to assemble large adult ADHD data sets, so far GWAS have yet to yield genome-wide significant loci, and rare variant studies (49, 50) will also require much larger sample sizes (51) before they can be used to further assess developmental differences in ADHD genetic architecture at a molecular level (see reference 52 for an extensive review of adult ADHD genetics).

The 21st century has witnessed the discovery of ADHD-associated rare and common genetic variants and resolved that its genetic architecture involves a spectrum of genetic variation in terms of frequencies and effect sizes. These discoveries are already providing a starting point to gaining insights into the underlying biology of ADHD (e.g., 18, 53) and for identifying and assessing novel treatments. Some clues about treatment come from rare mutations—for example, in a recent preliminary open-label, single-blind, placebo-controlled trial of the metabotropic glutamate receptor activator fasoracetam in 30 adolescents carrying mGluR mutations (54). There is also growing interest in using genome-wide association data to highlight candidates for drug repositioning in psychiatry, but so far this has not been applied to ADHD (55). However, further work still will be needed to investigate the functional impacts of associated genetic variants at the level of molecules, cells, neural systems, and circuits, as well as their impacts on brain development, using bioinformatics as well as experimental designs. Another observation is that, so far, the identified common genetic variants explain only a relatively small fraction of the heritability of ADHD (h²=0.22) (15) that has been inferred from traditional twin study designs. This suggests that continued collaborative and larger ADHD genetic discovery studies will be required to detect additional genetic contributions, together with robust epidemiological designs to identify causal environmental risk factors. The so-called single-nucleotide polymorphism heritability will always be an underestimate of true heritability, as it only encompasses common genetic loci detected by GWAS that are not necessarily causal, and it does not include the contributions of rare mutations, environmental risk factors, gene-environment correlation, or interactions. Novel strategies may be needed to consider alternative plausible risk mechanisms that might manifest as heritability—for example, parentally provided early environments (56), including the prenatal environment. The literature of the past 5 years on ADHD genetics has shown the enormous advantage of genetically informative and genotyped nationwide patient registries, and these will remain an invaluable resource for investigating many of the questions raised in this review.

Recent genetic discoveries highlight the strong overlap between ADHD, ASD, and intellectual disability. Indeed, in terms of rare mutations, thus far, it has not been possible to distinguish between ADHD and ASD (22), so the removal of the previous diagnostic exclusion criteria for ADHD in the presence of ASD is a welcome change in DSM-5. Yet the clinical features and treatments of these disorders are very different. What determines these differences is another crucial next research question (Figure 1). Typically these are early-onset disorders. One possible contribution to differences is variation in prenatal environmental exposures; another is background common genetic variants. For example, ADHD common genetic variant risk scores predict lower cognitive ability and educational attainment, whereas ASD common genetic variants predict higher cognition and educational attainment (57). Lessons from genetic discoveries on strong diagnostic overlaps are timely for clinicians and especially for service providers who tend to be diagnosis specific (e.g., ASD only), especially given the historical reticence about diagnosing ADHD in the context of ASD or intellectual disability. This is beginning to be recognized. For example, in 2017 in Wales, a nation of 3 million, multidisciplinary child neurodevelopmental clinics informed by recent findings have been established and have replaced ASD- and ADHD-specific assessment pathways.

In the past 5 years, GWAS have shown genetic overlap across multiple disorders. ADHD shows extensive pleiotropy that is especially prominent with major depressive disorder, but genetic risks also show some specificity, as only ADHD PRS appear to predict ADHD persistence (46). In the next 5 years, efforts need to focus on disaggregating this pleiotropy and testing alternative hypotheses using methods that are able to distinguish causal relationships, different ADHD subtypes, and misclassification from pleiotropy (e.g., using Mendelian randomization and other similar methods). This is important because the different explanations will have diverse clinical implications (Figure 1).

Genetic findings have confirmed the epidemiological view of ADHD as lying at the extreme of a continuum (or several continua) and have started to highlight the importance of adopting a developmental perspective (45, 46). Funders and ethics bodies need to recognize that developmental studies require longitudinal follow-up and that this is expensive and time consuming yet scientifically invaluable. In this regard, longitudinal population cohort designs are attractive for developmentally informative genetic studies because, unlike patient registry data, they involve trait-based assessments; however, nonrandom attrition is a problem, and individuals with severe illness (at the extremes of dimensions) are underrepresented. If ADHD is a trait, akin to blood pressure, then public or population health-based samples and approaches can be integrated with traditional genetics research on patients. This is beginning to happen. The conceptualization of ADHD as both a trait and a disorder is relevant to clinicians in that most recognize the somewhat arbitrary nature of the diagnostic cut-point and frequently encounter subthreshold cases. The presence of an underlying continuum means that issues such as defining the boundary for impairment and treatment are not straightforward. For example, the blood pressure cut-point for defining hypertension has changed over time, and hypertension guidelines in some countries adjust the cut-point for comorbid disease (e.g., chronic renal disease).

Given recent genetic discoveries, what are the implications for genetic testing in ADHD? The implications for clinicians are twofold. First, relatives of individuals with ADHD, including their parents, are at elevated risk of a range of neurodevelopmental disorders (e.g., ADHD, ASD, learning difficulties) as well as other neuropsychiatric illnesses (most commonly major depression) that could have an impact on assessment, treatment delivery, or treatment effectiveness. At present, although the rare mutations detected have large effect sizes, routine testing in ADHD is not recommended, although guidelines for such testing in clinical settings have now expanded in many countries to include mild intellectual disability and, in the United States, ASD. It was not long ago that mild intellectual disability was considered primarily sociocultural in origin, so this represents a substantial shift in international clinical practice. There is little empirical evidence to guide us as to what should happen for ADHD. An older study showed that the rate of established syndromes was sufficiently low in ADHD that genetic testing is not warranted in individuals without intellectual disability (58). However, that study predates current methods for detecting genomic anomalies, and a crucial issue is how severely neurodevelopmentally impaired the target patient group is, especially given that ADHD is relatively common. Whether genetic testing in clinical settings extends to neuropsychiatric disorders such as ADHD and schizophrenia remains to be seen, but as costs diminish and more knowledge about the causality of different mutations is gained, in my view, clinical practice is likely to change within the next decade—and sooner if some of these rare mutations are found to have treatment or prognostic implications or are medically actionable.

The clinical utility of common genetic risk variants for predicting ADHD is less certain, because their predictive power is weak and risk effects are defined in relation to a population rather than an individual. It is possible in the future that combining information on common genetic risk scores and family history (59) will be helpful for stratifying patients for the purpose of treatment or prognosis—for example, in identifying those who are at elevated risk of future psychosis or in guiding treatment. At present, however, these ideas remain in the domain of research questions that need to be answered. Although there is much commercial interest in pharmacogenomics, currently there is no scientific evidence to support such testing in ADHD. No common or rare gene variants have been found to be robustly associated with treatment response in ADHD, and very large studies will be required to detect genome-wide significant loci.

In summary, many novel genetic findings on ADHD have emerged and will continue to do so in the 21st century. Such discoveries are, in my view, of interest and relevance to a broad scientific field, including neuroscientists, developmental scientists, and population epidemiologists as well as clinicians, because they contribute to our conceptualization of ADHD and can shape the next 5 years of research and clinical practice.