The choice of interventions was justified: the authors stated that GXR and atomoxetine were the only non-stimulants approved by the US Food and Drug Administration for once-daily use for treatment of children and adolescents with ADHD in USA. The different dosage strategies considered were reported clearly and justified.
The response rates, utility values and any relevant sources were all reported clearly. Use of an indirect-comparison to estimate efficacy outcomes was justified by the author's statement that there was a lack of head-to-head trials for the two interventions. The authors reported that a systematic review was conducted to identify studies for this comparison analysis. The selection criterion for this review was reported clearly but there were no details of databases searched and the review methodology. Detailed information was reported on the three included trials. The authors did not justify their choice of papers for the prediction model and sources used to derive utility values.
All costs were clearly reported and relevant to the perspective and location. Sources used to derive costs were justified and reported clearly. The authors stated that the assumption that one pill was consumed per day was conservative for atomoxetine (FDA-approved atomoxetine dosing regimens included once- and twice-daily administration for atomoxetine). Costs were inflated appropriately.
Analysis and results:
The model structure and assumptions were reported clearly and a model diagram was supplied. There was no reported justification for adopting a one-year time horizon: if treatment and outcomes could be expected to occur beyond this horizon a one-year horizon may have underestimated treatment benefits. The MAIC methods used to derive model efficacy inputs were described clearly. The authors highlighted several issues with the MAIC analysis. In particular, since MAIC is not based on a randomisation of treatment groups, unobserved confounding factors between trials could bias the results; it appeared that the authors adopted a rigorous approach to minimise this possibility.
The impact of parameter uncertainty on the results may not have been fully addressed in the sensitivity analysis. A multivariate and probabilistic sensitivity analysis may have better assessed the overall impact of uncertainty. Ranges used within the deterministic sensitivity analysis were not justified and some ranges varied between arbitrary upper and lower bounds. Although the results appeared robust to individual parameter values, unaccounted for uncertainty may have occurred.
The authors stated that because the results were based on a few selected trials and may not be generalisable to the entire ADHD population. Regarding The authors stated that indirect costs were a major component of ADHD costs and future studies should include these.
The study methods seemed appropriate and were reported clearly and transparently. The conclusions reached by the authors appear to be appropriate.