Authors' reply

We thank Basu & Barnwal for their comments. As regards exclusion of so many patients, we stress that easily the biggest reason for exclusion was that the episode of self-harm was the patient's first. We make it clear why we chose recurrent self-harm rather than all patients with self-harm. The other exclusion criteria seem reasonable (regular fish consumption, etc.) and we see no reason why the findings are not applicable to ‘real-world’ patients. We knew that with such a small population subgroup analysis would be of dubious validity, therefore further defining the groups (e.g. according to recency of other self-harm episodes) was redundant. We certainly could have excluded those patients whose other episode(s) of self-harm were remote from the current one, but we chose not to.

We agree that more measuring points would have been desirable, especially in this capricious sample. This was a resource issue rather than a methodological one. We note the point regarding marital status being different between the two groups but re-analysis of the data controlling for this did not materially affect the results. It was agreed at study outset that in the absence of sufficient power to analyse actual differences in recurrent self-harm we would use the suicidal ideation sub-scale of the OAS–M. One either has suicidal ideation or not (whereas one can have ‘some’ depressed mood) and it seems appropriate to use a categorical measure here.

We suggest using ‘potential marker’ for ‘surrogate marker’ and confess we used the latter word loosely. There was quite good correlation (r=0.5) between measures of depression and the OAS–M suicidality sub-scale score. None the less logistic regression suggested that changes in suicidality were independent of depression scores, which indicates that factors additional to affect drive suicidal ideation. We agree that these findings could be clinically important. However, our findings can be regarded as no more than pilot data, owing to the small sample size. As fish oils are not patentable products, a larger study (with enough power to investigate actual reductions in self-harm) is unlikely to come from industry. Therefore we are continuing to seek funding for such a study.