Response

We thank Professor Seena Fazel for his thoughtful comment. We agree there is a possibility of over-diagnosis of attention deficit hyperactivity disorder (ADHD) in this and other studies of prisoners with ADHD. He notes that in the CIAO-II study the prevalence is around 24%, which is greater than the 95% CI (14–21%) reported in a recent meta-analysis.^{Reference Beaudry, Yu, Långström and Fazel1} He raises the possibility that the lack of efficacy in our study might have been due to the inclusion of mostly subthreshold cases. In our view this is unlikely for several reasons. First, we applied DSM-5 criteria including an assessment of both childhood and adult symptoms, applying all criteria such as onset before age 12, significant impairment from the symptoms and symptoms not being better explained by another disorder. The final diagnosis was made in each case by trained senior psychiatrists, under supervision for this project, following an interview assessment for both ADHD and other common mental health disorders that frequently co-occur with ADHD. Nevertheless, we did not include informant report as this was not feasible to obtain in many cases, and the accuracy of patient report among prisoners is unknown, raising the possibility of false positive cases. False positive cases could arise either due to the inclusion of subthreshold cases, as suggested, or the inclusion of other disorders that might mimic ADHD, such as post-traumatic stress disorders, or the effects of drug use.

To address these possibilities, we conducted several post-hoc subgroup analyses (see additional materials to this publication). Firstly, we included a subgroup who reported seven or more symptoms of both hyperactivity/impulsivity and inattention in both childhood and adulthood. This is much higher than the usual threshold in DSM-5 of five or more symptoms in either domain, and therefore very unlikely to be subthreshold cases. A total of 102 participants met this higher diagnostic threshold criterion. In this subgroup, the estimated score difference between the OROS-MPH and placebo arms was estimated to be 1.83 (95% CI [−2.34, 6.01]), (t = 0.87, P = 0.386), reflecting a small but non-significant improvement in the OROS-MPH arm.

Regarding the possibility for other mental health disorders to mimic ADHD, we conducted further subgroup analyses without any comorbidity (n = 36), without an anxiety disorder (n = 162), without a mood disorder (n = 137) and without antisocial personality disorder (n = 51). Cases of borderline personality disorder (n = 15) and post-traumatic stress disorder (n = 13) were too small to warrant further subgroup analyses. Finally, we conducted a subgroup analysis of 73 participants with low-risk scores for drug and alcohol use. None of these subgroup analyses indicated any significant differences in the primary outcome between OROS-MPH and placebo groups, with the highest difference being for the group with no alcohol use problems (1.88, 95% CI [−3.65, 7.41]), (t = 0.68, P = 0.499).

Overall, while we cannot rule out the possibility that the high rates of comorbidity and drug use could potentially mask some of the effects of OROS-MPH in this study, we also found no evidence to support this. In our view, the analysis of 102 cases with very high levels of symptoms likely excludes the possibility of sub-threshold cases as an explanation for the study's findings.