The estimate of the benefit of the therapy was based on a before-after study which may be biased to spontaneous improvements, other changes or to subjective assessment. A randomised controlled trial would have been less biased. A larger study would also give a more precise estimate of the potential effects. Resource use comparisons may also be biased in a before-after study, because of more intensive treatment earlier or after the trauma. Discounting was not used to adjust for differential timing of costs, and no sensitivity analysis was performed. The provision of separate resource use and cost data allows these results to be applied to other settings.