Interventions:
The interventions were well described. The authors acknowledged that there were two other treatments for heroin dependence, buprenorphine maintenance therapy and heroin-assisted treatment. Another comparator was not considered; buprenorphine plus naloxone. The authors justified the exclusion of buprenorphine, because systematic reviews had found it was less cost-effective than methadone. Heroin-assisted therapy was a criminal offence in the USA. The exclusion of these two treatments was reasonable. Buprenorphine plus naloxone was a new treatment, so its exclusion was reasonable.
Effectiveness/benefits:
There were no effectiveness data for deep brain stimulation compared with methadone, so the authors conducted a threshold analysis to identify the success rate for stimulation at which it became cost-effective. The estimate of the likelihood of being heroin-free on methadone was from a literature review and meta-analysis, conducted by the authors. The methods for the literature search and the inclusion criteria for the meta-analysis were not reported. The six-month analysis for a chronic disease, where patients might relapse after successful treatment, was unlikely to have fully assessed the benefits and costs of methadone and deep brain stimulation. Six months was chosen to match the length of the trials that had been conducted, rather than to match the pathology of heroin addiction. The authors acknowledged that this was a limitation. The estimates in the model are unlikely to be generalisable to the actual treatment of heroin dependence. The utilities were assessed for UK volunteers, rather than addicts, which means that the QALYs may not be applicable to heroin addicts in the USA, and may not be applicable to heroin addicts, in general.
Costs:
The costs were from appropriate US settings and were adequately reported. The six-month model is unlikely to have adequately represented the costs of the interventions, as it is unlikely to have adequately represented the pathology of heroin addiction.
Analysis and results:
The analyses of cost-effectiveness used a threshold of $180,000 per QALY gained, based on a study that estimated the willingness-to-pay for health care in the USA, rather than the customary $50,000 per QALY gained. The authors presented a graph showing the required success rate for deep brain stimulation, for different cost-effectiveness thresholds. As they acknowledged, the length and structure of their model did not accurately capture the treatment pathway and health outcomes of heroin addiction. It did not consider relapse after six months of methadone, nor the permanency of deep brain stimulation effectiveness, both of which are likely to influence long-term cost-effectiveness. The sensitivity analyses were well reported and varied all the pertinent variables, but were limited to one-way analysis, which does not fully capture the overall effect of uncertainty in the model.
Concluding remarks:
There were no comparative effectiveness data and the model structure was limited in its representation of the treatment pathway and health outcomes of heroin addiction. There was therefore considerable uncertainty around the potential cost-effectiveness of deep brain stimulation. Further research would provide more information, but this analysis did not evaluate the value-for-money of doing further research.