Dr. Heckerling argues that by offering vestibular rehabilitation to patients in the control group after 3 months, we implied that vestibular rehabilitation is an effective treatment, thereby violating the important principle of equipoise. However, in a pragmatic (1) or phase III trial such as ours, it is assumed that the treatment has already been shown to be beneficial under ideal conditions (normally, this will have been established in efficacy or phase II trials). Consequently, equipoise is maintained instead with regard to the research question: Is the treatment more effective than usual care in typical clinical practice? Moreover, it is accepted that clinician and patient attitudes toward the treatment will affect outcomes. It was important for this reason (as well as to permit fully informed consent) that we explained to potential participants that there was some existing evidence of efficacy but no previous demonstration of effectiveness, particularly when vestibular rehabilitation was delivered by practice nurses in a primary care setting. We therefore also drew attention in our paper to the likelihood that positive motivation and placebo effects may have contributed to outcomes, although the pattern of findings was indicative of specific effects on the balance system rather than simply a generalized improvement in subjective well-being. However, if controls had stayed in the study simply because they were "coerced" by our offer of a treatment of unknown effectiveness after 3 months, then substantial dropout might have been expected at follow-up in both control and intervention groups. On the contrary, dropout at 3 months was as low in the treatment group as in the control group and remained similarly low in both groups at 6 months.

Regarding Dr. Heckerling's subsidiary point, we concur that there was no compelling scientific rationale for presenting outcomes in the control group after controls had received therapy. However, we had collected follow-up data in both groups to analyze predictors of adherence to treatment in the whole sample (manuscript in preparation), and we felt that if we failed to report these outcomes alongside the longitudinal follow-up of the intervention group, readers might reasonably wonder whether this was because the intervention had inexplicably proved to be less successful in controls.