The editorial by Treloar et al ^{Reference Treloar, Crugel, Prasanna, Solomons, Fox and Paton1} has raised a controversial but justified issue regarding antipsychotic prescription in patients with dementia. We agree with the editorial supporting the cautious use of these drugs based on the ethical premise of reducing patient distress and palliation. However, we felt that there was a relatively quick and unchallenged submission to another important premise of the observed harm, which is intricately related to the topic in question. Our strong concern is that such unequivocal acceptance of the observed risks is likely only to enhance the ethical dilemma in a reader's mind. The decision to use these drugs, even for palliative purposes, is likely to be strongly governed by our safety and risk assessments. Are we not in a dilemma over the available safety evidence as well?

Is the observed harm specific to antipsychotic drugs, old age, dementia or behavioural and psychological symptoms of dementia (BPSD)? Is the observed association necessarily causation or are there certain limitations to a definite conclusion? For example, many a time the indication for which a drug is prescribed in dementia may be the cause of increased mortality rather than the drug per se. To quote the detailed Department of Health report, ^{Reference Banerjee2} ‘people with dementia and BPSD may be more likely to die (and to be prescribed antipsychotic drugs) than people with dementia and no BPSD’. Safety concerns regarding the use of antipsychotic drugs in elderly populations are a valid consideration, but are the risks also specifically higher for elderly people with dementia? The landmark meta-analysis of randomised controlled trials (RCTs), ^{Reference Schneider, Dagerman and Insel3} which concluded with a small increased risk for death with antipsychotics compared with placebo, also mentions that these results should be considered as hypothesis-generating. None of the individual drugs included in the 17 RCTs was sufficient to conclude for an increased risk, but a combined statistical effect was found. Does this call for a verification or should it be taken as conclusive?

Regarding efficacy studies, antipsychotic drugs have mostly been tested for treating BPSD. Behavioural and psychological symptoms of dementia is quite a heterogeneous term, used for an array of challenging behaviours such as restlessness, agitation, wandering, vocalisations, resisting help with dressing and personal hygiene, and verbal and physical aggression. Although the use of the term BPSD is quite appropriate in social dementia research (e.g. caregiver burden), is such a heterogeneous amalgamation of behaviours, which may or may not be of psychotic origin, a justified end-point to study clinical efficacy of drugs, or do we need more specific symptom clusters as indications of antipsychotic use in dementia? Further, is the number needed to treat (5–11) for antipsychotic drugs for behavioural improvement in dementia ^{Reference Banerjee2} any different from numbers needed to treat for antipsychotic drugs in schizophrenia? ^{Reference Rattehalli, Jayaram and Smith4}

Undoubtedly, from a clinical perspective, extreme care and caution should be exercised in prescribing antipsychotics in old age, especially for those with an underlying organic illness (e.g. dementia). Regarding the dilemma whether they should ‘ever or never’ be prescribed for patients with dementia, our point of contention is: (a) we cannot focus the debate only on the ethical angle to resolve this dilemma, there are several unanswered medical questions; (b) we cannot close our eyes to the caveats in existing safety and efficacy studies; and (c) we need to resolve the ambiguity surrounding the available evidence to empower us for an ethical as well as informed decision. More than ever, the dilemma is to arrive at certain indications for which we can use antipsychotics with relative safety.