We take issue with the presentation of the findings of the study by Armitage et al, ^{Reference Armitage, Rahim, Rowe and O'Connor1} on two counts. First, the title is misleading because there is no evidence that the intervention reduced suicidal behaviour, which was not measured as an outcome – not even through the obvious mechanism of recording re-attendance following non-fatal self-harm. Neither is it entirely accurate to imply, as the title does, that the intervention was designed to reduce suicidal behaviour. In the body of the paper, the authors indicate that they did not nominate a primary outcome measure because – in what they call an exploratory trial – they weren't clear what effects they expected.

Second, there is a stage at which flaws in the design or delivery of a trial mean that no safe conclusions can be drawn from it. Loss to follow-up of 50–60% at 3 months and >85% at 6 months is not compatible with a plausible intention-to-treat analysis, and using last observation carried forward cannot rescue the situation – especially when (as is the case here) the last observations in question are pre-intervention baseline measures taken at the same time that participants were given the materials for the intervention to which they had been randomised. The authors indicate that they are aware of some of the other problems with their trial – for example, outcomes not collected by researchers masked to allocation and uncorrected analysis of multiple outcomes. Under the circumstances, their conclusion that their intervention showed promise – a conclusion about which they were surely not in initial equipoise – cannot be regarded as sound.