We thank Drs Ranjan and Chandra for their considered response to our article. Although we acknowledge variations in prescribing practice, we know of at least two other centres nearby (the Institute of Mental Health and the Government Hospital, Chennai) that use the haloperidol-promethazine combination for rapid tranquillisation; the monthly combined out-patient attendance of the three centres is also greater than 9000.

Our wider survey of drug formularies, including the source of Ranjan and Chandra, and local pharmacies reveals that the price of injectable haloperidol (5 mg/ampoule) ranges between Rs 4.00 and Rs 5.50; that of promethazine (50 mg/ampoule) between Rs 3.00 and Rs 7.00; and that of lorazepam (4 mg/ampoule) between Rs 7.00 and Rs 15.00. We therefore reiterate our contention that the haloperidol-promethazine mix is cheaper than (even reduced doses of) haloperidol and lorazepam.

We agree that the Overt Aggression Scale would have generated more specific results. However, the outcomes for this pragmatic trial were not chosen to generate specific results; they were chosen by the doctors and nurses of the emergency rooms to be of clinical utility. From the reaction we have already had to this study these outcomes do seem acceptable and welcome to others.

We acknowledge that there were nine more people with mania, six more misusing substances and five more already on benzodiazepines in the lorazepam arm than in the comparison arm. There is no indication, however, that the integrity of the randomisation procedure was compromised, as such chance imbalances could occur in the absence of stratification. It is unlikely that these imbalances account for the findings, as the difference in the numbers of people ‘clinically improved’ between the two interventions at 15, 30, 60 and 120 min were 31, 25, 20 and 14, respectively, and in numbers ‘asleep’ 40, 47, 35 and 14.

Although recommended by important review articles and guidelines, we have found only four randomised studies in which a total of 80 people received the combination of haloperidol and lorazepam (Reference Arana, Ornsteen and KanterArana et al, 1986; Reference Battaglia, Moss and RushBattaglia et al, 1997; Reference Bieniek, Ownby and PenalverBieniek et al, 1998; Reference Subramaney, Brook and BerkSubramaney et al, 1998). None of these studies reports useful data on time to tranquillisation/sleeping; most report scale-derived data that are difficult to interpret clinically. For such limited data to direct practice at the two largest psychiatric centres in India, as well as many other places, would seem imprudent. The effects of haloperidol plus promethazine, we would still suggest, are better proven than other prevalent approaches. Recent influential guidelines in the UK have noted this and the sister study (TREC Collaborative Group, 2003) to be the only large trials of high methodological quality in this area (National Collaborating Centre for Nursing and Supportive Care et al, 2004).

Certainly the study and others like it need to be repeated so that the evidence upon which we treat people at this vulnerable time is robust. Practice on lesser evidence is surely unethical.