The Nature of the Association Between Childhood ADHD and the Development of Bipolar Disorder: A Review of Prospective High-Risk Studies

The literature search identified nine studies that met inclusion criteria for review (Table 1). Recruitment and selection criteria as well as the methods employed to evaluate psychopathology in the parent and offspring differed among the studies, as detailed elsewhere (18). However, most studies used semistructured clinical interviews of both the parent and offspring, conducted by interviewers initially blind to family affiliation, and consensus diagnostic procedures to arrive at DSM diagnoses. Some studies included other informants (e.g., teachers) and used standardized questionnaires to measure attentional, behavioral, and emotional symptom levels. Two of the included studies were limited by relatively short follow-up periods and small high-risk offspring samples and are not formally discussed here (19, 20). Other studies were more informative given the comparison groups included (21, 22) and the detailed prospective clinical assessments of referred youths (23, 24). The remaining studies report on larger samples of offspring who had a parent with well-characterized bipolar disorder and who were followed on average for at least 5 years into adolescence and adulthood (25–27).

Laroche et al. (23, 28) published findings from a 3- to 7-year prospective follow-up study reporting on 37 offspring from 21 families with a bipolar parent selected from an outpatient clinic; there were no control families for comparison. The bipolar parent had to have been maintained on lithium for at least 1 year and have children between 5 and 18 years old. At the time of last assessment, the mean age of offspring was 16.2 years (range=8–25 years). DSM diagnoses were made in 24% of the offspring (N=9/37) and clustered in the affective/internalizing domain. While no cases of full-threshold ADHD were diagnosed, the study reported significantly higher scores in hyperactive, anxiety, neurotic, and depression symptoms in the high-risk offspring who had a DSM diagnosis (mostly mood and anxiety disorders) compared with those who did not. These findings suggest that early in the course of evolving bipolar disorder, anxiety and minor depressive disorders dominate and that inattention symptoms may occur in the context of the evolving mood disorder.

Radke-Yarrow et al. (21) reported on a community cohort of children of affectively ill and well mothers after 3 years of prospective follow-up. Two siblings from 100 families were included, one in infancy (1.5–3.5 years old) and the other in early childhood (5–8 years old). An association between disruptive behavior disorders and families under higher stress (χ²=7.05, p<0.01) and from lower socioeconomic status (χ²=7.46, p<0.02) was reported. Children of affectively ill mothers were more likely to manifest problems with depression symptoms in middle (χ²=18.69, p<0.0001) and in later childhood (χ²=10.77, p<0.0001). Anxiety disorders were identified in infancy and early childhood in offspring of both healthy and depressed mothers, whereas anxiety did not become frequent until later childhood in offspring of bipolar mothers. In summary, it became evident that children of depressed mothers were the most severely affected with psychiatric problems across internalizing and externalizing domains from preschool through early and later childhood. Children of bipolar mothers held an intermediate position, with psychiatric and behavioral problems becoming evident in middle and later childhood. A gender effect, with boys showing more disruptive problems and girls more internalizing problems, was also reported.

In subsequent years, this cohort was followed up in adolescence (11–19 years old) and again in young adulthood (18–28 years old), with some families joining the study, some changing category, and others leaving the study. Main findings reported by Meyer et al. (29) included that 19% (N=6/32) of the offspring of bipolar mothers and 7% (N=3/42) of the offspring from unipolar mothers developed bipolar disorder at a mean age of 16.6 years. Based on rating scales, the study found a higher rate of childhood attention and behavioral symptoms in the offspring who developed bipolar disorder relative to those with no mood disorder in adulthood. The authors also reported that specific deficits in executive functioning during adolescence (as evidenced by certain parameters on the Wisconsin Card Sorting Test) and premorbid attention problems preceded the formal diagnosis and treatment of bipolar disorder. All offspring with both childhood attention problems and executive functioning deficits in adolescence were diagnosed with bipolar disorder. These findings are limited by the small number of cases of bipolar disorder and by the degree of assortative mating in the families. Nonetheless, they are consistent with the interpretation that in at least some high-risk children, cognitive antecedents, but not a clinical diagnosis of ADHD, are associated with a subsequent risk of developing bipolar disorder.

In a prospective study of school-age children (8–16 years old) of mothers with bipolar disorder, depression, chronic medical illness, or mothers who were healthy, Hammen et al. (22, 30) reported on psychopathology and psychosocial problems (based on the Child Behavior Checklist completed by the mother and a teacher) at baseline and at 6-month intervals until age 3 years. ADHD was diagnosed in 6% of offspring of bipolar mothers compared with 9% of offspring of depressed mothers and 5% of offspring of healthy comparison mothers. A substantial proportion of the psychopathology of the offspring of bipolar mothers was milder than that of the offspring of depressed mothers, with anxiety disorders being prominent. Based on maternal and teacher reports, there was no difference in social competence, school behavior scores, or academic performance ratings between the children of bipolar mothers and the children of healthy mothers (30). While limited by the relatively small number of children followed for up to 3 years, the findings are consistent with the interpretation that ADHD is not overly represented among the offspring of bipolar mothers and that the early psychosocial and school functioning of these high-risk offspring is generally comparable to that of the healthy population and different from that of children of depressed mothers.

Akiskal et al. (24) charted the prospective course of evolving psychopathology in 68 referred (symptomatic) juvenile relatives (offspring and siblings) of adult patients with confirmed bipolar disorder who were assessed and treated in a specialty mood disorders outpatient clinic. At baseline, 44 of the 68 high-risk youths had been previously seen by child mental health professionals, and 16 had been diagnosed with emotional problems related to family or peers, eight with neurotic (anxiety) disorders, seven with conduct disorder, seven with schizophrenia, and four with ADHD. Interestingly, none of the offspring assessed in childhood were thought to have a primary mood disorder. Within the first year of prospective study, 24 youths were diagnosed with major depression, 11 with manic or mixed episodes, and 22 with subaffective disorders. After an average of 3 years of prospective study, recurrences occurred in those with major affective disorders as well as conversion from subaffective to major affective and from unipolar to bipolar spectrum. In summary, nonaffective childhood diagnoses preceded the onset of minor and major mood disorders. Depressive and subaffective disorders predominated early in the course, while full-blown hypomanic and manic episodes were not seen until after age 13. Finally, similar to the findings of Laroche et al. (28), childhood antecedents of “hyperactivity” and “antisocial” symptoms were described as “phasic,” occurring alongside mood symptoms and not responsive to trials of stimulant medication.

Egeland et al. reported on a prospective study of school-age children of Amish parents affected with bipolar disorder (100 children) and healthy parents (110 children) who were systematically assessed at 7 years (31) and 10 years (27) after baseline. The study evolved from research involving the adult Amish bipolar disorder patients, who estimated that their illness manifested 9 to 12 years earlier than the age at onset of the full-blown diagnosis. The goal was to identify which childhood symptoms predicted the development of bipolar disorder in the children at genetic risk. All families were from the Amish community and comprised a high-risk group (one bipolar parent and one well parent), a comparison group with positive family history (one well sibling of the bipolar disorder proband and the other well parent), and a comparison group with negative family history (two well parents and no relatives with bipolar disorder). After 7 years of prospective follow-up, 38% of the offspring of bipolar parents were tagged as being at risk based on symptom profiles, compared with 17% of the offspring from comparison families (83% from the positive family history subgroup). The specific clinical risk features flagged in the children of bipolar parents included anxious/worried, attention poor/distractible in school, low energy, excited, hyper alert, mood changes/labile, school role impairment, sensitivity, somatic complaints, and stubborn/determined. There was also evidence of a phasic nature of the symptoms involving mood, energy, sleep, and temper problems.

At the 10-year follow-up, the mean age of the offspring groups was 17–18 years, and 41% of the offspring of bipolar parents were tagged as being at risk based on symptom profiles, compared with 16% of the offspring from comparison families (82% from the positive family history subgroup). The same core symptoms remained at higher frequency in the high-risk offspring except that low energy, anger, fearfulness, and sensitivity dropped below significance level, while high energy, sleep difficulties, excessive talking, loud talking, and problems with thinking and concentration reached significance. Taken together, these findings suggest that during development, putative prodromal features in high-risk offspring shifted from anxiety-depressive to more manic symptoms. Distractibility, stubbornness, and being easily upset were first noted early in childhood, alongside the affective symptoms, whereas “problem concentrating” was associated with manic-like symptom clusters later in development. The episodic nature of these symptom clusters continued throughout development. Furthermore, the authors confirmed that ADHD as a syndrome was “relatively absent.”

An ongoing prospective study of 140 children of bipolar parents in the Netherlands, reported by Wals et al. (16, 26, 32), builds on the findings of the previously described studies. The majority of families (102 children; mean age=16.1 years) derived from a community-based Dutch Patient Association, while the others were identified through outpatient clinics. High-risk families had comparable socioeconomic status and higher IQs on average compared with those of the Dutch general population, and 76% of families were intact. While symptom rating scales were generally comparable between the high-risk group and the normative population, daughters of bipolar parents had higher scores on Child Behavior Checklist subscales for total problems, internalizing, externalizing, somatic complaints, anxious/depressed, social problems, delinquent behavior, and aggressive behavior. Sons of bipolar parents scored higher on Child Behavior Checklist subscales of total problems, externalizing, thought problems, and aggressive behavior. On self-report, high-risk older adolescent girls reported more attention problems, while teachers reported no differences in attention or behavior problems in the high-risk boys or girls compared with the normative population.

At last follow-up, covering almost 5 years, the cohort had 129 children, with a mean age of 20.8 years (26). The risk of ADHD over the prospective waves of assessment remained stable, while the risk of mood disorder and bipolar disorder increased. The lifetime risk was 10% for bipolar disorder (I or II), 40% for any diagnosable mood disorder, and 5% for ADHD. In those high-risk offspring with bipolar disorder, the index mood diagnosis was almost always depressive in polarity, at a mean age of 13.4 years (SD=4.2), and the index hypomanic/manic episode manifested a mean of 4.9 years (SD=3.4) later (at 18.4 years old). Only one of these 13 bipolar subjects had treatment with stimulant medication before the development of the activated episode. Collectively, these findings lend further support to reports of a variety of emotional and behavioral childhood antecedents followed by subaffective and depressive disorders in early adolescence, and hypomanic/manic episodes later in life. ADHD, as a clinical diagnosis, was not higher in the high-risk offspring than in the general population based on either clinical interview assessment or teacher reports.

Finally, our group has published several reports from an ongoing longitudinal high-risk study (33–36). The high-risk families (one affected and one well parent) were selected through specialty clinical research programs and diagnosed based on best estimate procedure. Bipolar parents were divided on the basis of an unequivocal response or nonresponse to long-term lithium treatment. Lithium response identifies a more homogeneous subtype of classical bipolar disorder (37), while lithium nonresponse is characterized by a chronic illness course and a higher familial risk of psychotic disorders (38).

Early on, we reported an association in the high-risk offspring between subjective problems in attention and symptoms of depression that was not related to any obvious deficit in sustained attention on psychological testing (39). In subsequent analyses, our group found comparable lifetime rates of clinically significant ADHD (often comorbid with learning disabilities) in high-risk (8.3%) and comparison offspring (5.8%) (25). The current age-adjusted lifetime risk of major affective disorders in 231 high-risk offspring at a mean age of 25.7 years old (SD=9.23) is estimated at 52.8% (mean age at onset=16.8 years) and bipolar disorder at 13.5%. Interestingly, higher rates of ADHD and other neurodevelopmental abnormalities, including learning disabilities and cluster A traits, were observed in the subgroup of offspring of parents who did not respond to lithium (35). Furthermore, a recent analysis (40) found that this neurodevelopmental phenotype was more frequently observed in high-risk offspring who developed a substance use disorder compared with those who did not (18.0% compared with 9.3%, p=0.06). While limited by a small sample, across this high-risk cohort, ADHD does not appear to be a robust predictor of major affective disorder or bipolar disorder. Of those high-risk offspring with a childhood diagnosis of ADHD, 28% have so far gone on to develop a major affective disorder (major depression or bipolar I or II) and 11% a minor affective disorder (dysthymia or depression not otherwise specified), while 93% of the high-risk offspring with a major affective diagnosis did not have antecedent ADHD.

While this qualitative review is limited by the number of prospective longitudinal high-risk studies collectively assessing a relatively small number of bipolar-affected offspring, the findings are consistent across studies. One major finding from this review is that childhood ADHD does not appear to be significantly overrepresented among the offspring of parents with well-characterized bipolar disorder compared with offspring of healthy parents or with the risk of ADHD in the general population. Furthermore, although the data remain limited, there does not appear to be an association between ADHD as a childhood diagnosis and the subsequent risk of bipolar disorder in the high-risk offspring. Therefore, it seems reasonable to conclude that the clinical diagnosis of ADHD does not appear to be a reliable antecedent in the developmental trajectory toward bipolar disorder.

However, the risk of ADHD in the studies reviewed may have been influenced by the high rate of intactness and relatively high socioeconomic status of the high-risk families. This speculation is indirectly supported by the findings of Birmaher and colleagues (17), who reported in a cross-sectional study of school-age children that rates of ADHD were not elevated among the offspring of bipolar parents relative to comparison subjects after adjusting for confounding variables (i.e., socioeconomic status and non-bipolar disorder psychopathology in biological parents), suggesting that ADHD may be more related to the general burden of psychopathology in the families, rather than specifically to the risk of bipolar disorder. Indirect support also comes from Radke-Yarrow et al. (21), who reported an association between higher rates of disruptive behavior disorders in offspring from high-risk families of lower socioeconomic status and those with more stress.

An additional influence may have been the low rate of psychopathology in the nonproband parents and low comorbidity rates (especially of substance use and conduct disorders) in the bipolar parents in some of the reviewed studies (26, 27, 36). Furthermore, a number of the high-risk studies reviewed here included affected parents with a more classical form of bipolar disorder that was completely responsive to or at least stabilized by lithium and not typically associated with ADHD (23, 25, 31). For example, among the Amish parent probands, ADHD was relatively absent based on expert clinician retrospective review of all available clinical information (31).

However, attentional and cognitive symptoms and problems were reported by high-risk subjects and other informants (typically parents) in a number of the studies reviewed here, consistent with a vulnerability trait in some or as part of the early course of an evolving major affective disorder in others. For example, in the findings reported by Duffy et al. (25, 35), ADHD was seen at an increased rate in offspring of parents with a psychotic spectrum bipolar disorder that failed to stabilize with lithium. A number of these children were diagnosed with full-blown ADHD by a child and adolescent psychiatrist and went on to develop a diagnosable affective disorder in adolescence. The course of the mood disorder in these offspring (as in the parent) tended to be nonepisodic, with residual symptoms between acute episodes, and adult relatives had a differentially higher loading of chronic psychotic disorders. Preliminary findings suggest that ADHD may form part of a neurodevelopmental childhood phenotype predicting a particular subtype of bipolar disorder. In other words, bipolar disorder is known to be a heterogeneous diagnosis, and some subtypes appear to overlap clinically and biologically with psychotic spectrum disorders (41).

On the other hand, a number of studies reviewed here reported that problems with attention and distractibility manifest in high-risk offspring as part of a childhood internalizing condition. Furthermore, an important aspect documented in several studies has been the periodic nature of these early mixed symptom clusters, which again suggests a bipolar disorder diathesis characterized by a recurrent illness course, rather than by a chronic course, as would be expected in the case of a typical developmental disorder like ADHD. Therefore, in children at confirmed familial risk for the development of bipolar disorder, childhood ADHD may not be a clinical antecedent per se, but rather symptoms of inattention and distractibility may form part of an early subaffective clinical presentation on the trajectory toward bipolar disorder.

Given the limitations of the high-risk data reviewed here, findings from complementary prospective studies of other cohorts of children can inform the interpretation of the findings from this review. There have been several large prospective studies of epidemiological samples. One of these is the Dunedin longitudinal birth cohort study, which followed 1,037 children with assessment points through childhood, adolescence, and early adulthood. In the Dunedin study, adults diagnosed with mania were more likely than those without mania to have a history of juvenile conduct or oppositional disorder and depression, but not ADHD (1). The authors noted that many adults with bipolar disorder had antecedent childhood anxiety disorders, but this association fell short of statistical significance. Interestingly, adults with schizophreniform disorders had a juvenile history of a number of disorders including ADHD. In a subsequent report on this cohort (42), schizophreniform disorder was associated with neurodevelopmental and cognitive abnormalities during childhood, while adults with bipolar disorder had comparable motor, language, and cognitive performance in childhood to that of healthy comparison subjects. These observations add to the substantial evidence from prospective studies of psychological and neurological abnormalities in children who go on to develop adult schizophrenia, contrasting with the normal or advanced premorbid intellectual functioning and school performance generally found in children who develop bipolar disorder in adulthood (41, 43).

Another such study, the Oregon Adolescent Depression Project, followed a cohort of 1,709 high school students from representative urban and rural school districts assessed at four time points from adolescence to early adulthood. This cohort had a very low rate of full-blown bipolar disorder, while subthreshold bipolar disorder symptoms were more common but transient. In a subsequent analysis investigating the relationship of subthreshold conditions to the subsequent development of full-blown disorders, Shankman et al. (44) reported that with the exception of ADHD and bipolar disorder, there was an increased probability of escalating to full syndrome disorders (homotypic continuity). In the case of heterotypic continuity, individual externalizing subthreshold conditions predicted the development of other full-syndrome externalizing conditions (i.e., associations between ADHD, conduct, and substance use disorders), but not the development of internalizing disorders or bipolar disorder.

Finally, a recent analyses from the German Early Developmental Stages of Psychopathology prospective study confirmed that hypomanic symptoms were common and mostly transitory in the general adolescent population (45). Almost 40% of 1,565 participants expressed bipolar symptoms at one assessment, but only 17% experienced these symptoms at two assessment times (46). In an analysis of the association between risk factors, including ADHD, and the onset and persistence of subthreshold manic and depressive symptoms, cannabis use was associated with the onset of manic symptoms (odds ratio=4.26; p=0.010), while ADHD was associated with the persistence of depressive symptoms (45). Therefore, findings suggest that hypomanic symptoms are often a normative transient experience in adolescence and that ADHD does not appear to be a risk factor in the general population for the development of subthreshold or full-blown bipolar disorder.

There have been differing reports on comorbidity rates between ADHD and bipolar disorder in pediatric clinical cohorts, particularly in very young referred children thought to be manic (47). As pointed out by Carlson (14), some of this variation may be due to assessment and sampling issues, as well as to differences in the conceptualization and approach to diagnosis. Specifically, U.S. research groups typically prefer standardized interviews emphasizing individual symptoms, while U.K. research groups tend to adopt a clinical approach emphasizing patterns of symptoms.

Recently, Wozniak et al. (48) reported on the persistence of bipolar disorder after 4 years of prospective study in 78 youths with a mean age of 13.4 years at last assessment and an estimated age of 4.9 years at onset of bipolar I disorder. These youths, primarily boys with comorbid ADHD, were ascertained from a tertiary care psychopharmacology clinic that specializes in the treatment of ADHD. The course of bipolar disorder was mostly chronic, and only five children were in remission at last follow-up. Other reports of persistent manic-like presentations have been published (10, 47), and most of these reports show a high rate of comorbidity with ADHD, externalizing disorders, or pervasive developmental disorders. As reviewed recently by Leibenluft (49), some of these presentations have been recast as fitting criteria for severe mood dysregulation disorder and on prospective follow-up have not shown continuity with adult bipolar disorder. On the other hand, prospective studies of children with primary ADHD have been very consistent in supporting the findings of prospective epidemiological studies, namely, that ADHD in children is associated with a greater risk of antisocial behavior, criminal activity, conduct disorder, and substance use disorders in adolescence and adulthood, but not to an elevated risk of bipolar disorder (50, 51).

In the end, prospective longitudinal studies provide very little evidence that ADHD is a reliable antecedent to the development of bipolar disorder that continues into adulthood. In fact, substantial evidence has been reported from prospective cohort studies of increasingly different illness trajectories of ADHD and bipolar disorder over the course of development from a starting place of some overlapping nonspecific symptoms early in childhood (Figure 1). Trajectories do not suggest that all children with ADHD will go on to have persistent ADHD or to develop complications or related disorders such as conduct or substance use disorders or sociopathy as adults, but rather these are outcomes of increased risk in this population. Similarly, it is not that all offspring of parents with bipolar disorder will manifest early childhood antecedents such as anxiety disorders or will develop sensitivity to stress or depressive disorders in early adolescence, followed later by activated episodes. However, this predictive clinical sequence reflects a risk probability derived from prospective studies.

The other important observation from prospective high-risk studies is that early in development and in the clinical course, symptoms appear to be nonspecific and often overlap between different evolving disorders; in the case of bipolar disorder and ADHD, impulsivity, labile mood, and internalizing symptoms (anxiety and depression) are not uncommon in childhood. Furthermore, in a selected subset of children from bipolar parents with a high rate of psychosis in family members and whose illness did not respond to lithium prophylaxis, ADHD and learning disabilities may form part of an early neurodevelopmental phenotype that overlaps somewhat with the early presentation in children with primary ADHD. However, the temporal pattern of psychopathology against the backdrop of differential familial risk sheds light on the increasingly different illness trajectories, presumably reflecting shared as well as different underlying pathophysiological processes.

Longitudinal prospective research in large community and selected high-risk cohorts has been immensely helpful in mapping the early natural history of major psychiatric disorders and in highlighting both homotypic and heterotypic developmental continuity of psychopathology over the course of illness development. Mapping the early clinical course is the first step in advancing studies of the associated neurobiological pathways and identifying novel targets for early intervention. Our diagnostic systems and approaches can catch up to the progress that prospective research has made by emphasizing the need, as in other areas of medicine, to incorporate the developmental trajectory against the backdrop of risk as indicated by the family history of psychiatric disorders, in order to improve the accuracy and validity of psychiatric diagnosis in symptomatic youths.