The interesting debate between Parker and Anderson & Haddad (Reference Parker, Anderson and Haddad2003) suggests more fundamental reasons to question prevailing research paradigms and designs in respect of the efficacy of and indications for psychotropic medicines. That the clinical trial industry reveals even marginal drug effects may be seen as surprising given the virtual absence of any basis for a taxonomy of mental disorders, other than the syndromal classifications used in psychiatric practice. There is little evidence that the major syndromes align with any readily defined pathophysiological variance. Group heterogeneity in trial work, as the debaters remark, will therefore attenuate the evidence for substantial drug treatment effects, sometimes to vanishing point. Meta-analysis of such data may not be much more revealing, compounding the influence of variable sampling in individual trials and publication bias.

These side-effects of the randomised controlled trial ethos are not greatly mitigated in the field of organic mental disorders. At huge expense, multicentre trials of cholinesterase inhibitors in patients classified as probably having Alzheimer's disease have shown only very modest (and to many observers still unconvincing) effects on cognitive and neuropsychiatric outcomes (e.g. Reference Lanctôt, Herrman and YauLanctôt et al, 2003). This is because these conditions, pace distinguished efforts at nosological definition in life, are also heterogeneous. This variability, already evident from detailed clinical and neuropsychological assessment, is further revealed by functional and structural analysis of the brain. It is these data which might best inform sampling for therapeutic trials. Studies on a smaller scale, therefore, targeting the more readily defined Lewy body dementia (e.g. McKeith et al, 2000) or more intensively characterised and monitored patients with Alzheimer's disease (e.g. Reference Venneri, Shanks and StaffVenneri et al, 2001) in both double-blind and open-label designs, can convincingly demonstrate the nature and the modalities of efficacy using cholinesterase inhibitors. In the same way, studies of smaller groups of patients receiving treatment for depression may reveal correlations between clinical features and treatment responses that are more likely to guide the selection of therapy for individual patients (Reference MaybergMayberg, 2003).

Large randomised controlled trials, by submerging variation in the interest of marginal statistical significance, seem to offer limited hope of significantly improving the evidence that guides clinical practice. Studies of cognitive and pharmacological interventions might best be carried out with smaller patient groups for whom there has been detailed assessment of relevant pathophysiological and cognitive variance, as well as the manifest clinical symptoms.