Baker et al (Reference Baker, Brown and Akiskal2004) report an interesting post hoc analysis from a randomised double-blind, placebo-controlled study evaluating the efficacy of olanzapine co-therapy in patients with bipolar disorder who had adequate responses to valproate or lithium monotherapy (Reference Tohen, Chengappa and SuppesTohen et al, 2002). The authors describe a secondary analysis assessing response among dysphoric and non-dysphoric patients with bipolar I disorder.

The authors conclude that olanzapine in combination with either lithium or valproate was effective in improving the severity of depressive symptoms coexisting with acute mania. This conclusion is based on statistically significant differences in mean changes in Hamilton Rating Scale for Depression (HRSD) score. However, the authors have not reported the standard deviations for these mean changes. Hence it is difficult to ascertain whether the data are skewed. It is possible that a few patients showing large changes on the HRSD could have skewed the data. It was also puzzling that the authors reported that the difference in the HRSD score between combination and monotherapy groups was larger for dysphoric patients. One would expect participants in the non-dysphoric group to have much lower baseline scores so that there would be less chance of a significant reduction. (The mean HRSD baseline score in the non-dysphoric group was 10.42 (s.d.=5.27) and in the dysphoric group 25.18 (s.d.=4.62).)

We are also of the view that reporting study outcomes in terms of mean changes on a rating scale does not provide meaningful information for clinicians. Reporting results using dichotomous outcome measures such as ‘improved’ and ‘not improved’ with a meaningful cut-off point defined a priori would be helpful. Clinicians would be more interested in outcome measures such as complete remission of symptoms, return to premorbid levels of functioning, etc. To address the question of whether olanzapine is helpful for patients with dysphoric mania it would be helpful to know how many in the olanzapine co-therapy group achieved complete remission and whether there was any statistical difference between groups. It would have been interesting if Baker et al had also provided dichotomous outcomes based on the Clinical Global Impression scale for bipolar disorder (CGI–BP; Reference Spearing, Post and LeverichSpearing et al, 1997), as this was administered during the course of the trial and data should be readily available.

It is not uncommon to come across reporting of various outcome measures and multiple analysis of a randomised controlled trial. However, whether this adds to clinical knowledge is questionable. We agree with Baker et al that it is important to explore the pharmacological options for dysphoric mania as the available options are limited. However, we need more pragmatic outcome measures that are easily understood by clinicians and can be applied in routine practice rather than being lost in multiple analysis. Systematic reviews such as that on the use of olanzapine for mania also highlight the lack of pragmatic outcome measures in the reporting of randomised controlled studies (Reference Rendell, Gijsman and KeckRendell et al, 2003). We hope future reports of such studies will use outcome measures that are more applicable to the real world.