We are grateful for Huda's considered comments and, in particular, for drawing our attention to an error in the published paper. ^{Reference Barrett and Byford1} The reoffending rates of 3% and 5% applied to the economic model (and varied in sensitivity analysis in an attempt to account for the associated uncertainty) are supported by a systematic review of the literature, which identified a number of papers where rates of serious reconviction following specialist and mainstream detention were reported. Unfortunately, the references listed in support of this assertion are incorrect. The correct references are listed below. ^{Reference Friendship, Mann and Beech2-Reference Taylor4} There is a similar error in the text at the top of page 338 referring to routine sources of cost data. The correct references, which are correct in Table 1, are also listed below. ^{5-Reference Fortune, Barrett, Armstrong, Coid, Crawford and Mudd7} We apologise for failing to spot these errors earlier.

The reoffending rates applied to the economic model do not relate to the protective effects of detention but are rates reported following release from detention. They are therefore the therapeutic effects of the dangerous and severe personality disorder (DSPD) intervention v. no DSPD intervention. The model, in fact, takes both types of effect into consideration: the therapeutic effects via the application of probabilities of reoffending once released and the protective effects via data on the differential lengths of time the groups spent in detention.

This is equally true for the analysis reporting that better levels of cost-effectiveness are achieved if the DSPD intervention takes place in a low-cost prison, as compared with the base-case analysis which modelled DSPD services as they were actually configured at that time (based in both prisons and high secure hospitals). This analysis was not an assessment of the cost-effectiveness of detaining participants in low-cost prisons. Instead, it was an analysis that assumed that the DSPD treatment programme only took place in a prison setting, rather than a high secure hospital, and simply involved replacing the cost of those who were in reality treated in high secure hospitals with the lower cost of treating them in a prison. The probability of reoffending once released from detention was not altered, so the analysis did incorporate the therapeutic effects of the intervention, and the probability of being released into the community remained the same.

We do not agree that the results are further evidence that the best management of violent offenders is for the criminal justice system to keep offenders in prison for long periods. Our results simply suggest that the DSPD treatment programme, as it was configured at the time of the analysis, was not found to be a cost-effective alternative to the situation where the programme is not available. By supporting the control condition, the results in fact support earlier release, rather than later, as the evidence suggests that those in the DSPD intervention were on average detained for longer periods of time than would have been the case without the intervention. The results do, however, support Huda's assertion that the funding allocated to the DSPD intervention could be better spent elsewhere.