We agree with much of what Dr Owens says, but take issue with three points.

First, we disagree that we asked the wrong question. The possibility that a brief message of concern from toxicologists that the individuals may never have met might halve the number of repeat episodes of self-harm is intriguing. ^{Reference Carter, Clover, Whyte, Dawson and D'Este1} Of course we need to ask what the possible mechanisms might be.

Second, self-harm greatly increases the risk of suicide but it is much more than a proxy measure in trials. As clinicians working in accident and emergency departments and mental health settings will testify, self-harm is important in its own right – there may over 200 000 hospital presentations for self-harm in England every year. ^{Reference Hawton, Bergen, Casey, Simkin, Palmer and Cooper2}

Third, although we are all fans of qualitative research and the additional insights it provides, the main reason for negative trial results is not the low status of qualitative data. Negative findings are more likely to reflect the fact that trials to date have been too small to detect clinically important effects ³ (or alternatively that the interventions simply do not work).

Outcomes for trials are definitely an issue and Dr Owens summarises a number of the key considerations. Many studies to date have used repeat episodes of self-harm presenting to hospital as the principal outcome measure. We did argue (perhaps somewhat clumsily) in an earlier version of our article that such repeat presentations might actually be an indication of positive engagement with services. We deleted the offending passage following editorial and reviewers' comments. The case study that Owens briefly presents is very interesting and of course would not be picked up by standard reporting of trial results. Using qualitative data to comprehensively measure outcomes on all participants in large trials is impractical. A challenge for self-report measures may be the painfully low response rates. However, we would support Dr Owens' call for a variety of outcome measures – hospital-based and self-report, quantitative and qualitative.