Previous studies have shown a decline in ADHD gender ratios with increasing age, suggesting the presence of sex differences in developmental trajectories in ADHD symptoms. Although the cause of these differences is not well understood, they could have important implications for clinical practice. Using growth mixture modelling, this study evaluated whether males and females in a community-based sample differed in ADHD symptom trajectories across ages 7 to 15 years.
This study included 761 females and 810 males from the ongoing Zurich Project on Social Development from Childhood to Adulthood (z-proso).* Inattentive and hyperactivity/impulsivity symptoms were measured using the teacher-report version of the Social Behaviour Questionnaire (SBQ),† and latent inattentive and hyperactivity scores were generated using a longitudinal factor model. To compare the average inattention and hyperactivity/impulsivity developmental trajectories for males and females, a multi-group latent growth curve model was applied to the ADHD factor scores. Next, growth mixture models (GMMs)‡ were fitted to the ADHD factor scores for each sex separately, as well as separately to the inattention versus hyperactivity/impulsivity scores.
Results showed that a multi-group latent growth model with linear and quadratic growth and no cross-group equality constraints fitted well to the hyperactivity and impulsivity scores (χ2 (42) = 226.812; p < 0.001) and inattention scores (χ2 (42) = 169.24; p < 0.001). Adding cross-group equality constraints on the intercept, linear and slope factor means resulted in a significant deterioration in fit for both hyperactivity/impulsivity (Δχ2 (3) = 205.135; p < 0.001) and inattention (Δχ2 (3) = 155.44; p < 0.001), suggesting that the sex difference in average trajectories was statistically significant.
For hyperactivity/impulsivity, the best-fitting GMM for both males and females was a three-class model that included both linear and quadratic growth. The largest class for both genders (63% of males, 81% of females) was described as ‘unaffected’, and was characterised by low symptom levels that decreased steadily from childhood into late adolescence. The next-largest class for both genders (24% of males, 9% of females) was labelled as ‘high stable’, reflected by persistently high symptom levels across ages 7 to 15 years. The third category was gender-specific; for males (13%), this category was labelled as ‘high increasing’ due to an acceleration in the rate of symptom increase, with a possible inflection point between ages 11 and 13 years. For females, the third ‘concave’ class was characterised by a mild elevation of symptoms in childhood followed by a minimum around early adolescence and a marked increase thereafter.
For inattention, the number of classes in the best-fitting GMM differed between males and females. For males, the preferred GMM included two classes; the largest class (61% of males) was labelled as ‘low stable’ due to persistently low symptom levels across ages 7 to 15 years, whereas the other class (39% of males) was labelled ‘high stable’, reflecting persistently elevated symptoms. The best-fitting GMM for females included three classes. Similar to males, the largest class (59% of females) was labelled ‘low stable’. The next-largest class (31% of females) was a ‘moderate’ class, characterised by moderate and slightly declining symptom levels across ages 7 to 15 years. Finally, females demonstrated a ‘high decreasing’ class, characterised by initially elevated but declining symptom levels across ages 7 to 15 years.
Limitations of this study were largely related to the measure of ADHD symptoms. First, the brief nature of the SBQ meant that it was not possible to identify distinctions finer than the broad domains of inattention versus hyperactivity/impulsivity. Second, because the SBQ is not directly based on the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria, the applicability of the results to ‘clinically defined’ ADHD is unclear. The use of a non-DSM approach could also make it difficult to compare the results of this study with the wider literature base. Third, the measure of ADHD used teacher reports rather than direct observation of actual behaviour across different contexts. Finally, the study was also limited by the absence of statistical comparisons of sets of developmental trajectories across groups; as the reported comparisons were purely descriptive, results from this study should not be treated as conclusive.
The authors concluded that females were more likely to demonstrate an increase in symptom levels in early adolescence, whereas males were more likely to show elevated symptoms from childhood. For both males and females with hyperactive/impulsive ADHD, early adolescence represented a period of risk characterised by a relatively sudden increase in symptoms. The authors stated that these findings may have potentially important clinical implications, as females affected by hyperactivity/impulsivity may be more likely to be excluded from diagnosis due to current age-of-onset criteria. To combat this, later symptom onset should be considered in diagnostic criteria, and hyperactivity/impulsivity diagnostic indicators should be made more suitable for adolescence and adulthood.
Read more about the sex differences in ADHD developmental trajectories here
*Z-prozo is an ongoing longitudinal cohort study of psychosocial development with a focus on externalising problems. The first wave of z-prozo took place in 2004, when participants were aged 7 years and starting school. Parents provided informed consent on behalf of their child until age 11, after which the participant provided their own consent
†The SBQ includes four inattention items and four hyperactivity/impulsivity items, which have a 5-point response scale spanning from ‘never’ to ‘very often’. Participants were assessed at ages 7, 8, 9, 10, 11, 12, 13 and 15 years. The questionnaire was administered to each participant’s teacher as part of a larger questionnaire on child psychosocial functioning. Most children had the same teacher from ages 7 to 9 years, before switching to a different teacher across ages 10 to 12 years. At ages 13 and 15 years, participants were in high school
‡GMMs were fitted with between 1 and 6 classes. A set of models with linear growth parameters only and a set of models with both linear and quadratic growth parameters were evaluated, resulting in 12 models per dimension for each gender. As not all models in the set were nested, the Lo-Mendall-Rubin test was used to determine if a model with k – 1 classes should be rejected in favour of a model with k classes for the set of linear models and the set of linear + quadratic models separately. Akaike’s Information Criterion (AIC), Bayesian Information Criterion (BIC) and sample-size adjusted BIC (saBIC) were used to compare non-nested models differing in whether they included both linear and quadratic growth versus linear growth only. After selecting optimal class solutions for inattention and hyperactivity/impulsivity for males and females, differences were compared descriptively, as a direct multi-group model was not currently possible
Murray AL, Booth T, Eisner M, et al. Sex differences in ADHD trajectories across childhood and adolescence. Dev Sci 2019; 22: e12721.