Nuij and colleagues are too sanguine in concluding that their review has demonstrated the effectiveness of safety planning in reducing suicidal behaviour. They note the methodological limitations of the studies they summarise and the evidence of clinical heterogeneity and publication bias that make meta-analysis problematic, but they could say more about the problem of interpreting their primary outcome.

Not all the included studies offer a clear and standardised definition of suicidal behaviour or attempted suicide, and only one (Gysart-Mallart et al) makes clear how they dealt with acts of self-harm that were ambiguous or attributed (by clinicians or otherwise) to non-suicidal motives. This latter uncertainty is a major problem, because so-called non-suicidal self-harm is a significant risk for suicide. We need to know for each study whether the primary outcome was all acts of self-harm or only those that included an element of suicidal intent. If the latter then there is a serious problem of potential attributional bias.

In truth I think we cannot conclude much except that we need better studies of an intervention that it is tempting to support on grounds of common sense but which may lack the power to make real change in people's response to their distress.