ADHD was previously conceptualised as a childhood-onset disorder, and diagnostic criteria require the presence of inattentive and hyperactive/impulsive symptoms by the age of 12 years for a diagnosis of ADHD. However, population-based longitudinal studies have suggested that some individuals present with ADHD symptoms later in development, meeting diagnostic criteria for ADHD for the first time during adolescence or young adulthood (Agnew-Blaise JC, et al. 2016; Caye A, et al. 2016). Recent studies have divided individuals with late-onset ADHD into two subsets according to childhood ADHD symptoms: those who had 0–2 inattentive or hyperactive/impulsive symptoms before age 12 are classed as ‘genuine’ or ‘de novo’ late-onset ADHD; and those who had 3–5 symptoms before age 12 are described as having ‘subthreshold’ late-onset ADHD. However, the similarities and differences between de novo and subthreshold ADHD remain unclear. This study aimed to assess the distinction between de novo and subthreshold late-onset ADHD, and to investigate the factors driving an increase in ADHD symptom development beyond childhood by comparing early childhood characteristics and adolescent outcomes across individuals with different ADHD symptom levels.
Data from the Twins Early Development Study (TEDS), a UK population-based sample of 12,593 pairs of twins recruited in England and Wales between 1994 and 1996 (Haworth CM, et al. 2013), collected from birth to age 16 years were used in this study. ADHD symptoms were assessed at age 8, 12, 14 and 16 years using the Revised Edition of the Conners’ Parent Rating Scale (CPRS-R),* and the current study included participants who had ADHD symptom ratings in childhood (age 8 or 12 years) and adolescence (14 or 16 years), yielding a study sample of 9875 individuals. CPRS-R ADHD ratings were recoded to relate to ADHD symptom count, in terms of both inattention and hyperactivity/impulsivity symptom domains.† Sum scores were then divided into Diagnostic and Statistical Manual of Mental Disorders symptom thresholds: scores ≥6 = ‘clinically significant’; scores 3–5 = ‘subthreshold’; scores 0–2 = ‘low’. The following four categories, comprising 7084 of the 9875 eligible participants, were selected to address the study objectives:
- Non-ADHD controls (n=6509; 65.9%): participants with a low ADHD symptom level across childhood and adolescence (CPRS-R sum score 0–2)
- De novo late-onset ADHD (n=123; 1.2%): those with a low level of ADHD symptoms in childhood (sum score 0–2) who developed clinically significant ADHD symptoms in adolescence (sum score ≥6)
- Subthreshold late-onset ADHD (n=164; 1.7%): those with subthreshold ADHD symptoms in childhood (sum score 3–5) who developed clinically significant ADHD symptoms in adolescence (sum score ≥6)
- Childhood-onset persistent ADHD (n=288; 2.9%): those with a clinically significant level of ADHD symptoms (sum score ≥6) in both childhood and adolescence.
Early childhood characteristics evaluated across the four groups were: prenatal and perinatal factors (including maternal medical risk during pregnancy, birth weight and growth rate), family environment factors (including socioeconomic status [SES], maternal depression and harsh parenting) and clinical characteristics (including cognitive performance, assessed using the parent-reported MacArthur Communicative Development Inventory and the Parent Report of Children’s Abilities, and behavioural and emotional problems, assessed using the parent-reported Strengths and Difficulties Questionnaire [SDQ]). The adolescent outcomes measured were cognitive performance (based on IQ and vocabulary tests, and GCSE results) and co-occurring behavioural and emotional problems (assessed using the Mood and Feeling Questionnaire [MFQ], the Anxiety-Related Behaviours Questionnaire and the SDQ). Childhood characteristics and adolescent outcomes of the three ADHD symptom groups and non-ADHD controls were compared using multinomial regression models, with pairwise analyses performed to compare the two late-onset ADHD groups with the persistent ADHD group. Early childhood characteristics that were predictive of de novo or subthreshold late-onset ADHD were examined using multinomial logistic regression.
The results of the study were as follows:
Early childhood characteristics
- A higher proportion of males was observed in all three ADHD groups compared with non-ADHD controls (Cohen’s d = 0.09; p < 0.001).
- The three ADHD groups also had significantly elevated levels of single parenthood, and lower SES, increased harsh parenting and higher maternal depression were observed versus non-ADHD controls, with varying effect sizes (Cohen’s d = 0.03–0.77; all p < 0.05).
- The de novo late-onset ADHD and persistent ADHD groups also showed elevated maternal health risk during pregnancy versus non-ADHD controls (p < 0.05 and p < 0.001, respectively).
- Differences in levels of harsh parenting and maternal depression, but not single parenthood or SES, were observed between the three ADHD groups.
- Significant differences were observed in general cognitive performance and SDQ scores when comparing the three ADHD groups with non-ADHD controls, particularly for the SDQ conduct domain (Cohen’s d = 0.64–1.45; all p < 0.001). Pairwise comparisons among the ADHD groups showed that the childhood-onset persistent ADHD group had the highest ratings for conduct and peer problems, while the subthreshold late-onset ADHD group showed greater conduct problems than the de novo late-onset ADHD group (Cohen’s d = 0.33; p < 0.05).
- All three ADHD groups showed significantly worse cognitive performance, SDQ ratings and co-occurring depression and anxiety than the non-ADHD control group.
- The greatest effect size versus non-ADHD controls was observed for conduct problems (Cohen’s d = 1.65–1.78; all p < 0.001) followed by parent-reported MFQ (Cohen’s d = 1.13–1.38; all p < 0.001).
- In contrast to observations in childhood, pairwise comparisons showed similar levels of impairment across the three ADHD groups at the age of 16 years, with comparable ratings of general cognitive performance and co-occurring behavioural/emotional problems observed. However, GCSE performance was significantly worse in the childhood-onset persistent ADHD group than the de novo and subthreshold late-onset groups (effect size 0.27 [p < 0.05] and 0.30 [p < 0.01], respectively).
Childhood predictors of ADHD
- Male sex and increased conduct problems in childhood were found to be predictive factors for all three ADHD categories (male sex: odds ratio [OR] 2.13–4.47, all p < 0.001; conduct problems: OR 1.22–1.74, all p < 0.01).
- Lower SES was an additional and unique predictive factor for de novo late-onset ADHD (OR 0.70, p = 0.006), but not for subthreshold late-onset or childhood-onset persistent ADHD.
- Harsh parenting (OR 1.81, p < 0.001) and higher maternal depression (OR 1.07, p = 0.001) were additional significant predictors of subthreshold late-onset ADHD.
- There were five additional predictors of childhood-onset persistent ADHD: lower birth weight (OR 0.74, p = 0.043), single parenthood (OR 1.93, p = 0.021), harsh parenting (OR 1.51, p < 0.001), higher maternal depression (OR 1.06, p = 0.02) and peer problems (OR 1.31, p < 0.001).
There were some limitations to this study, including the categorisation of participants into diagnostic groups based solely on parent-reported CPRS-R ratings, which may affect the reliability of ADHD diagnoses. Additionally, diagnoses were based on questionnaires that did not take account of other ADHD criteria, including functional impairment and situational pervasiveness, which could have resulted in false-positive or -negative diagnoses. Also, the ADHD symptom level cut-off thresholds, although also used in previous studies, were somewhat arbitrary, such that replications using full ADHD diagnostic criteria may be required. Finally, the TEDS study population comprised predominantly white Caucasian individuals from the UK, and therefore the results may not be generalisable to other populations.
The authors concluded that de novo and subthreshold late-onset ADHD differ considerably from childhood-onset persistent ADHD in terms of cognitive and behavioural characteristics in childhood, but that these differences had largely dissipated by adolescence. Low SES was found to be a unique childhood predictor of de novo late-onset ADHD, while harsh parenting and maternal depression were specific predictors of subthreshold late-onset ADHD. The authors suggested that interventions targeting these early risk factors may help to attenuate development of late-onset ADHD symptoms. They emphasised that further research into time-varying environmental and biological factors driving the development of ADHD symptoms is important to aid better characterisation of late-onset ADHD.
Read more about early predictors of late-onset ADHD here
*The CPRS-R is an 18-item parent-reported rating scale, with each item rated on a 4-point Likert-type scale reflecting its frequency (0 = not at all to 3 = very much true), giving a total score of 0–54
†CPRS-R ratings were recoded to relate to symptom counting in clinical practice: endorsement of 0 or 1 on CPRS-R was recoded as 0 (not indicative of symptom presence), and endorsement of 2 or 3 was recoded as 1 (indicative of symptom presence). If >4 out of 9 items in either domain of inattention or hyperactivity/impulsivity were unanswered, the sum score of that domain was recoded as missing. Sum scores of the recoded CPRS-R ranged from 0–9 separately for each symptom domain, indicating the number of endorsed ADHD symptoms per individual
Agnew-Blais JC, Polanczyk GV, Danese A, et al. Evaluation of the persistence, remission, and emergence of attention-deficit/hyperactivity disorder in young adulthood. JAMA Psychiatry 2016; 73: 713-720.
Caye A, Rocha TB, Anselmi L, et al. Attention-deficit/hyperactivity disorder trajectories from childhood to young adulthood: evidence from a birth cohort supporting a late-onset syndrome. JAMA Psychiatry 2016; 73: 705-712.
Haworth CM, Davis OS, Plomin R. Twins Early Development Study (TEDS): a genetically sensitive investigation of cognitive and behavioral development from childhood to young adulthood. Twin Res Hum Genet 2013; 16: 117-125.
Liu CY, Asherson P, Viding E, et al. Early predictors of de novo and subthreshold late-onset ADHD in a child and adolescent cohort. J Atten Disord 2019; Epub ahead of print.