Family and twin studies of cADHD demonstrate a high heritability, estimated to be around 70–80% from twin studies.6, 7
Relatively few studies have investigated the genetic and environmental contributions to the developmental course and outcomes in adulthood. Longitudinal twin studies show that the continuity of symptoms from childhood through to adolescence is predominantly due to common genetic influences.8, 9, 10
Although such stable genetic effects are likely to continue beyond the adolescent years, there are only a few studies investigating this.
Genetic research on ADHD started with the finding that hyperactivity tends to aggregate in families.11, 12
Since then, family studies have shown that ADHD shows familial clustering both within and across generations. Increased rates of ADHD among the parents and siblings of ADHD children have been observed.13, 14
In addition, strongly increased risks for ADHD (57%) among the offspring of adults with ADHD have been reported.15
Also, compared with the risk for ADHD among the siblings of children with ADHD (15%), siblings of adults with ADHD were found to have a strongly increased ADHD risk (41%).16
Furthermore, a prospective 4-year follow-up study of male children into mid-adolescence found the prevalence of ADHD to be significantly higher among the parents and siblings of persistent ADHD child probands compared with the relatives of ADHD probands in whom ADHD remitted.17
Taken together, these studies suggest that the risk for ADHD may be greater among the first-degree relatives of probands with ADHD that persists into adolescence and adulthood than that among the relatives of probands with ADHD that remits before adulthood.17, 18
Whether such familial risks reflect genetic or environmental factors can be clarified using adoption and twin studies. Adoption studies found that ADHD is transmitted only to biological relatives, which strongly implicates genetic factors as the main causal influences on familial risk for the disorder.11, 12, 19, 20, 21
These studies showed (for both current and retrospective symptoms in adults) that cADHD in child relatives predicts aADHD (or associated symptoms) in adult relatives. However, both adoption and family studies identify discrepancies related to different sources of ratings, with self-evaluation of ADHD symptoms by adults providing less evidence of familial effects than informants or cognitive performance data.19, 22, 23
Recently, four adult population twin studies using self-ratings of ADHD symptoms have been completed, which all found heritabilities that are far lower than those found in similar studies of parent- or teacher-rated cADHD: 41% for retrospectively reported childhood ADHD symptoms in a sample of 345 US veterans aged 41–58 years old,24
40% for current inattention problems in a Dutch study of 4245 18–30-year olds,8
30% for current ADHD symptoms in a Dutch study of over 12
000 twin pairs with an average age of 31 years25
and 35% for current ADHD in a Swedish sample of more than 15
000 twin pairs aged 20–46 years (Larsson et al.
, unpublished data). The situation is similar in adolescence, as adolescent twin studies using self-ratings show lower heritability estimates than studies of parent or teacher ratings,26, 27
suggesting that self-ratings may be a poorer measure of the underlying genetic liability to ADHD than informant reports or clinical interviews. Although the estimated heritability in self-rated ADHD symptoms in adult populations is lower than that derived from parent or teacher ratings of cADHD, the pattern of findings is identical. Both types of studies find that there are no gender differences observed in the estimates of heritability, heritability estimates are stable across the age-span (for each type of measurement approach), there are similar estimates of the genetic correlation (the proportion of shared genetic effects) of 60–70% between inattention and hyperactivity-impulsivity, familial effects are all genetic in origin with no shared environmental influences, and no threshold effects are found. This suggests that for both child and adult ADHD the disorder is best perceived as the impairing extreme of a quantitative trait (Larsson et al.
, unpublished data; ref. 28
Despite these common features, the relatively low heritability estimates for ADHD symptoms in adults derived from population twin studies need some explanation, because they appear to be at odds with heritability estimates of ADHD symptoms in children, as well as the family studies that show a high familial risk for persistent forms of ADHD.15, 18
Several factors are likely involved. We have already mentioned the consistent finding that self-ratings of ADHD symptoms give lower estimates of heritability compared with informant ratings in twin studies. One source of measurement error (that is, variance of the true diagnostic status that is not predicted by the measurement instrument) is the reliability of the self-rated measures of ADHD symptoms. In one of the heritability studies by Boomsma and co-workers,25
this was estimated to be around 0.66, which is lower than the mean heritability of cADHD across extant studies.29
Psychometric studies also show that, although self-ratings may be useful as a screening tool for aADHD, their correspondence with the full diagnosis is only modest. For example, Kessler et al.30
reported that the sensitivity of self-ratings as a measure of diagnosis was high (98%), whereas the specificity was not (56%). Similar findings were reported by Daigre Blanco et al.31
(87.5% sensitivity and 68.6% specificity). Related to this source of error are potential effects of having two raters in twin studies of self-ratings of ADHD (each twin rates him/herself), whereas informants usually rate both members of a twin pair. Since reliability between two raters will always be less than an individual's reliability with their own ratings, and because a ceiling on heritability is set by the reliability of ratings, heritability estimates will always be lower when two separate raters are involved in evaluating each twin pair compared with only one. Single raters may inflate identical twin pair similarities, potentially leading to an overestimation of heritability in the reported studies on cADHD, whereas the lower reliability of ratings between two raters may lead to lower estimates. Evidence for the later conclusion comes from our recent analysis of same versus different teacher ratings in a study of 5641 12-year-old twins, with heritability estimates of 75% for same teacher and 53% for different teacher ratings of twin pairs (Merwood and Asherson, unpublished data).
Another relevant difference between child and adult samples is the expected range of ADHD symptom scores. It is well known that ADHD symptoms decline through adolescence into adulthood.32
Thus, it is possible that the restricted range of ADHD symptoms in adulthood could influence estimates of heritability. Although some of this symptom decline is likely due to true remission of ADHD, some have argued that the diagnostic criteria for ADHD, which were originally developed for children, are developmentally insensitive and thus become less sensitive to ADHD with age (see above and refs. 4
). Added to this is the possibility that in cross-sectional studies of adult population twin studies (that do not apply clinical criteria for ADHD), ADHD symptoms may emerge in some individuals owing to adult-onset conditions, such as anxiety, depression and drug use. These ‘phenocopies' would lead to increased measurement error of the genetic liability for ADHD and lower estimates of heritability.
Differences in the way that participants are ascertained in different study designs may also impact on estimates of familial/genetic influences. The family studies that showed high familial risk for ADHD used case–control methods to ascertain adult patients who were self-referred for (severe) ADHD-like problems. There are notable differences between the clinically referred and population-based samples. The former have a more skewed male-to-female ratio, higher rates of psychiatric comorbidity and lower rates of primarily inattentive ADHD. Moreover, the family and twin studies used differing assessment methodologies. The family studies diagnosed subjects with structured interviews that evaluated childhood onset of impairing symptoms and the presence of impairment in multiple settings as required by Diagnostic and Statistical Manual of Mental Disorders, 4th Edition. In contrast, with the exception of Schultz et al.
the twin studies used rating scale measures of ADHD symptoms that do not query for childhood onset and do not systematically assess impairment in multiple settings.
Overall, these considerations suggest that the lower heritability of aADHD compared with cADHD could be due to increased measurement error in the aADHD twin studies. Support for this conclusion comes from a recent Swedish twin study, which found that the heritability of attention problems in 19–20 year olds was estimated at 78% when self-rating and parent-rating data were combined; the heritability for self-ratings alone was 48% (Larsson et al.
, unpublished data). Analogous to this, cluster A personality disorders show low heritability estimates in analyses based on limited phenotypic information that become much higher when adding more information from interviews.35
On the other hand, it still remains feasible that the heritability of aADHD does indeed decline with increasing age. This might reflect the importance of developmental processes that are sensitive to person-specific environmental factors affecting the longitudinal outcome of ADHD in adults.
Since heritability estimates do not relate directly to the frequency or effect size of specific genetic risk factors,36
it is not yet clear as to what the lower heritability estimates actually mean for molecular genetic studies of aADHD. For example, some disorders with low heritabilities, such as prostate and breast cancer, have identified genes with moderate to large effects,37
yet this is not the case for many highly heritable phenotypes including ADHD.38
In the absence of sufficient studies on this issue, it is quite clear that genetic researchers should preferably use measures that have been shown in family studies to have high rates of familial transmission and in adoption studies to aggregate in biological, rather than adoptive relatives. The evidence for strong familial risks in the relatives of adolescent and adult ADHD probands suggests that the clinical diagnosis of aADHD may represent a more familial measure, although there are no studies to date that directly address this question. The difference could arise because the clinical diagnosis takes a developmental perspective in which the adult phenotype reflects persistence of the childhood disorder, whereas the cross-sectional data used in twin studies may include adult-onset causes of ADHD-like symptoms that reflect phenocopies involving different etiological processes.
We conclude that aADHD is influenced by familial factors that are genetic in origin. The available studies indicate that self-ratings of Diagnostic and Statistical Manual of Mental Disorders, 4th Edition-defined ADHD symptoms may not be the best measure of the underlying genetic risk for aADHD and that other factors such as childhood onset, pervasiveness and impairment should be taken into account.