Response

Foresight into the lifetime course of those diagnosed with attention-deficit hyperactivity disorder (ADHD) is of high clinical value. ^{Reference Caye1} In this context we read, with great interest, Foulds et al’s ^{Reference Foulds, Boden, Kerr, Douglas, Pettie and Young2} examination of whether ADHD symptoms in adolescence are associated with adverse mental health and psychosocial outcomes in adulthood. Using the data from an established birth cohort of 1265 individuals from Christchurch, New Zealand, Foulds et al report that adolescent ADHD symptoms are significantly associated with substance use problems and criminal offending in adulthood, even after controlling for several confounders. The inclusion of a comprehensive range of mental health and psychosocial outcomes, as well as predictors, whose value has recently been emphasised, ^{Reference van der Plas, Noordermeer, Oosterlaan and Luman3} enhances the value of the study. However, certain aspects of the study need to be re-examined while interpreting the findings.

Foremost among these concerns is the definition of the exposure, which is based on the categorisation of ADHD symptoms into upper and lower quartiles. The exposure group was derived by ‘summing relevant items’, without clear mention regarding the items that were included from the self-reported/parental scales. For example, the Revised Behaviour Problems Checklist comprises six domains, yet the authors do not clarify which domains contributed to the ADHD symptom score. ^{Reference Quay4} Moreover, symptoms were summed across three scales assessed twice, at ages 15 and 16 years, with a cut-off of 6 or more symptoms defining the upper quartile. This approach is problematic, because endorsement of just one item each on the three scales at both time points could have resulted in classification into the exposed group. Furthermore, minimal variation (for example, endorsing five versus six symptoms) would have led to categorisation into either group, raising concerns about the clinical meaningfulness of this threshold for the classification of exposed and non-exposed groups. This issue is also compounded by the absence of information regarding the maximum possible ADHD symptom score, limiting interpretation of symptom severity. The demarcation between exposed and non-exposed groups is therefore indistinct. We believe that better clarity could have been achieved by comparing the highest quartile with the lowest, or with individuals reporting no ADHD symptoms, or by examining associations between specific symptom domains or specific items on scale and long-term outcomes.

Additionally, although the authors report that 7.4% of participants met the criteria for ADHD, the diagnostic criteria used are not specified. Although the authors note that diagnostic or symptom count-based analysis resulted in ‘model instability’, the reporting of supplementary unadjusted or minimally adjusted comparisons between individuals meeting diagnostic criteria for ADHD and those without ADHD (7.4 v. 92.6%) would have been informative and clinically relevant. Such an approach could then have been complemented by dimensional, symptom-threshold analyses with multiple adjustments, thereby preserving statistical rigour while enhancing clinical interpretability and translational value. A similar limitation is also seen with the categorisation of conduct disorder and oppositional defiant disorder, where the criteria for grouping were also unclear.

Beyond methodological considerations, this study raises a broader conceptual and socio-philosophical concern regarding ‘retrospective or historical labelling’. Psychiatric diagnoses, in particular ADHD, are dynamic in nature rather than fixed entities, and continuously evolve over time and across cultures. ^{Reference Abdelnour, Jansen and Gold5} Thus, retrospectively interpreting the current understanding of ADHD in regard to behavioural data collected during the 1990s leads falsely to attributing the long-term outcomes to diagnosis, particularly when the cohort was not intended to measure clinically defined ADHD. Also, there is a more intriguing concern that arises from the use of individual symptoms as a proxy threshold for ADHD, especially during adolescence – ‘overdiagnosis’. The arbitrary cut-off chosen in epidemiological studies has the potential hazard of being used clinically, leading to overdiagnosis and overtreatment, which are clearly evident and concerning. ^{Reference Kazda, Bell, Thomas, McGeechan, Sims and Barratt6}

Regarding outcome assessment, we highlight two major concerns. The first of these is in regard to the fact that the outcomes, which were significantly predicted (i.e. substance use and criminal offending), were assessed based on self-reported measures and are therefore inherently vulnerable to social desirability bias. This bias is particularly common for substance use disorders, where stigma and fear of judgement often lead respondents to underreport use or related problems, which further reduces interpretability of the results. ^{Reference Zemore7} Importantly, underreporting is common among young adults and urban cohorts, ^{Reference Steinhoff, Shanahan, Bechtiger, Zimmermann, Ribeaud and Eisner8} as in the case of this study. The inclusion of complementary objective measures of substance use, if available in the data registry, could have clarified this issue.

Second, the persistence of ADHD into adulthood, an important predictor of long-term outcomes, was not assessed. The absence of adult ADHD assessment limits the understanding of its long-term course. Similarly, if such information is available within the registry, assessment of adult dissocial or antisocial traits may have provided additional context for interpretation of associations. A minor critique concerns the measure used to evaluate relationship breakdown. The authors used the Feeling Bad scale ^{Reference Lewis, Siegel and Lewis9} which, according to the cited reference, although developed for pre-adolescents, was applied to assess participants aged 18–21 to 35–40 years. Moreover, the items in the scale were based on anxiety related to parental conflict and self-image, which do not align with the outcome of relationship breakdown. ^{Reference Lewis, Siegel and Lewis9}

A key strength of the study is the thorough assessment of multiple covariates, which helps to identify the independent effects of ADHD on various outcomes. Although the inclusion of stimulant medication use was appropriate, ^{Reference van der Plas, Noordermeer, Oosterlaan and Luman3} if available within the cohort data, other treatment modalities such as non-stimulant pharmacological agents and non-pharmacological interventions may have further informed interpretation, given their potential to influence long-term outcomes.

In conclusion, Foulds et al provide valuable insights into the forecasting of adult mental health and psychosocial outcomes in individuals with ADHD symptoms during adolescence. We deem it crucial to consider the critical aspects that we have highlighted here in interpreting the study findings.