Setting

This pragmatic randomised controlled trial, designed to include patients typical of those presenting to emergency services and to interfere little with routine practice, was conducted in the emergency services of the Department of Psychiatry at the Christian Medical College, in Vellore in the southern Indian state of Tamil Nadu. The majority of patients presenting to the psychiatric emergency services of this 1800-bed teaching hospital were accompanied by family members and were either brought directly, or were referred by general practitioners in the town or adjoining towns and villages and from emergency services of this and other hospitals.

Patient selection

Consecutive patients were assessed and were eligible for trial entry if the attending physician felt that intramuscular sedation was clearly indicated because of agitation, aggression or violent behaviour, and if the physician did not feel that either one of the interventions posed an additional risk for the patient. In keeping with prevailing clinical practice in this country, consent was obtained from a responsible relative if patients refused, or lacked capacity to consent to treatment by virtue of severe mental illness. For this trial relatives were fully informed and their written consent obtained; patients without a responsible relative were excluded. This trial compared two low-risk interventions in common use, the relative benefits of which are unknown. The institutional research and ethics committee approved the trial design, the consent procedure and the form used.

Sample size

From the existing literature, with tranquillisation of 73% of people given benzodiazepines and 57% given typical antipsychotics (Reference Battaglia, Moss and RushBattaglia et al, 1997; Reference Joy, Adams and LawrieJoy et al, 2003), with a power of 80% at 95% confidence intervals and an expected precision of 20%, the minimum sample size required was 90 people per arm.

Randomisation and interventions

Eligible patients were randomised to receive either intramuscular haloperidol (10 mg) and promethazine (25 or 50 mg) mixed in the same syringe, or intramuscular lorazepam (4 mg). All doses were at the discretion of the attending doctor, although the recommended dose was 10 mg haloperidol plus 50 mg promethazine, or 4 mg lorazepam. These doses were arrived at by prevailing clinical practice and a pilot study that showed that at least 4 mg lorazepam was required to achieve a similar degree of tranquillisation as the haloperidol–promethazine mix.

Randomisation was according to a computer-generated random numbers list in varying sized blocks of less than 10 prepared by the UK collaborator. This collaborator worked with a member of the TREC–India team who had no clinical responsibilities in conducting the trial. The team member and a pharmacist prepared consecutively numbered opaque cardboard boxes, identical in appearance and weight, on the outside of which was a form with questions to be completed by the attending doctor while ‘blind’ to the contents of the box. The boxes contained haloperidol (5 mg × 2 ampoules) plus promethazine (50 mg × 1 ampoules) or lorazepam (4 mg × 1 ampoule), as determined by the randomisation list, one disposable syringe and needle and study follow-up forms. All those involved clinically in the study had no indication of what medicines were in the boxes until they were opened.

Procedure

Once eligibility of a patient was ensured, the next consecutive box was taken from the emergency cupboard and this constituted randomisation. The duty doctor recorded the severity of the episode and the initial diagnosis on the form stuck to the outside of the sealed intervention pack. The box was then opened and the intervention administered. The patient was then followed up at 15, 30, 60 and 120 min by the treatment team and at 240 min and at 2 weeks by the study coordinators. Data were also obtained from the case notes as well as from interviews with relatives and the treatment team.

Blinding

The study was blind until the point of treatment assignment, which minimised selection bias. After assignment, rating for the first 2 h was not blind as the management team had to know the prescribed medications. In any event, TREC–India evaluated real-world interventions that are not given blind. The study coordinators, however, who were blind to interventions given, undertook ratings at 240 min. At this time, they also guessed the allocated intervention, to assess their blinding.

Outcomes

Patients were rated at each assessment point on whether they were tranquil or asleep; in addition, the time of onset of tranquillisation and sleep were noted. Participants were considered to be tranquil when they were calm and not exhibiting agitated, aggressive or dangerous behaviour. They were considered to be asleep if, on inspection, they appeared to be sound asleep and were not aroused by ambient disturbances; the depth of this apparent slumber was not assessed further. They were also rated on the Clinical Global Impression – Severity (CGI–S) scale at entry, and the CGI–Improvement (CGI–I) scale (Reference GuyGuy, 1976) with respect to aggression and violence, the Simpson–Angus extrapyramidal side-effects rating scale (Reference Simpson and AngusSimpson & Angus, 1970) and the Barnes Akathisia Scale (Reference BarnesBarnes, 1989) at each assessment point; any other clinically important adverse effect, especially dystonia, was also noted. These assessments were conducted only on participants who were awake, as extrapyramidal symptoms are usually not apparent during sleep or, in the case of dystonia or akathisia, are likely to prevent sleep. Other outcomes within the first 4 h were the use of additional medication for control of agitated or aggressive behaviour, the use of physical restraints, the need for further medical attention and numbers absconding. Participants were also followed up 2 weeks later to check for adverse effects or adverse outcomes and compliance with oral medication. The primary outcome was ‘tranquil or asleep by 4 h’.

Data analyses

We used double data entry and analysed data using the Statistical Package for Social Sciences (SPSS) version 9.0 for Windows. We assessed the adequacy of randomisation by comparing participants' baseline socio-demographic and clinical characteristics. We compared proportions tranquillised, asleep, improved (CGI much and very much improved, stipulated in the trial protocol), requiring restraints, requiring the doctor to be recalled and requiring additional sedation, all using the chi-squared test, with a continuity correction, or Fisher's exact test, as appropriate. We calculated relative risks and an absolute measure, the number-needed-to-treat (NNT), and their 95% confidence intervals (Reference AltmanAltman, 1998) using intention-to-treat analysis. We also used repeated measures analysis of variance to compare mean CGI–I scores between groups across various time points, with being asleep at the follow-up points entered as a covariate. We used the Mann–Whitney U-test to compare mean times to tranquillisation and sedation in the two groups, as the data did not have a normal distribution. The kappa statistic was used to evaluate agreement between the blinded guesses of the coordinators regarding treatment allocation.

Management of agitation and violence in emergency settings: lack of consensus

Drugs commonly used to manage agitation and violence in emergency situations world-wide include antipsychotics, benzodiazepines and antipsychotic and benzodiazepine combinations (Reference AllenAllen, 2002; Reference McAllister-Williams and FerrierMcAllister-Williams & Ferrier, 2002). More recent strategies include longer-acting drugs such as zuclopenthixol acetate (Reference Coutinho, Fenton and AdamsCoutinho et al, 2000) and rapidly acting intra-muscular formulations of the atypical antipsychotics olanzapine (Reference Jones, Taylor and MeehanJones et al, 2001) and ziprasidone (Reference Brook, Lucey and GunnBrook et al, 2000).

The recommendations of guidelines for the management of psychiatric emergencies (Royal College of Psychiatrists, 1998; Expert Consensus Guideline Group, 1999) are not evidence-based (Reference AllenAllen, 2002), or are they followed uniformly (Reference Pilowsky, Ring and ShinePilowsky et al, 1992; Reference CunnaneCunnane, 1994; Reference Binder and McNeilBinder & McNeil, 1999; Reference Huf, Coutinho and FagundesHuf et al, 2002a ). Evidence from randomised trials and systematic reviews is limited and does not indicate the superiority of zuclopenthixol acetate over conventional antipsychotics (Reference Fenton, Coutinho and CampbellFenton et al, 2003), or the commonly used combination of haloperidol and benzodiazepines over haloperidol alone (Reference Battaglia, Moss and RushBattaglia et al, 1997). There is a suggestion that benzodiazepines are superior to typical antipsychotics (Reference Battaglia, Moss and RushBattaglia et al, 1997; Reference AllenAllen, 2002) and that haloperidol is superior to placebo (Reference Joy, Adams and LawrieJoy et al, 2003).

Management of violence in middle- and low-income countries

The management of aggressive or violent psychiatric patients in India includes ‘talking down’ techniques, physical restraint and seclusion, as well as the use of medication. The initial minutes and hours are crucial and drugs that rapidly render people tranquil and/or sedated without producing distressing or dangerous adverse effects are desirable. In low-income countries such as India, the high cost of zuclopenthixol acetate precludes its widespread use; intramuscular atypical antipsychotics are not available and are likely to be prohibitively expensive.

TREC–India and TREC–Rio

TREC–India, the largest and only study for this comparison, randomised violent or agitated patients likely to be seen in everyday clinical practice. It compared two inexpensive, commonly used interventions for clinically relevant outcomes and lost data on only one person (0.5%) for the primary outcome and on 18 people for the 2-week follow-up (9%). Although both interventions are effective for controlling agitated or violent behaviour, with over 75% in each group tranquil/asleep within 15 min of administration and 96% in each group tranquil/asleep by 4 h, 10 mg of intramuscular haloperidol combined with 25–50 mg promethazine is superior to 4 mg intramuscular lorazepam in the speed of onset of sleep and thereby clinical improvement. If lorazepam is used alone a significant proportion of people remain awake for longer, potentially exposing everyone to danger. TREC–India did not find, however, that the reduced ability of lorazepam to sedate compared with the combination was accompanied by a greater need for subsequent intervention or harm to the patient or others.

TREC–Rio (Huf et al, Reference Huf, Coutinho and Fagundes2002a ,Reference Huf, Coutinho and Adams b ; TREC Collaborative Group, 2003) randomised 301 people over 6 months to receive either a haloperidol–promethazine mix or intramuscular midazolam. Though midazolam is available in India, it is five-times as expensive as the haloperidol and promethazine mix and is not in common use in psychiatry.

Midazolam consistently induced more rapid tranquillisation and sedation than the haloperidol–promethazine mix. However, the combination treatment in TREC–Rio was less tranquillising/sedating than in TREC–India (Table 5). Similar numbers of people in both trials were male, markedly ill and psychotic. In TREC–India, however, everyone allocated combination treatment received 10 mg haloperidol. In the TREC–Rio haloperidol–promethazine arm, 77/148 (52%) were given 5 mg haloperidol and 71/148 (48%) were given 10 mg. Most people allocated to the combination treatment in both studies were given 50 mg promethazine. Subgroup analysis of the two different doses of haloperidol in TREC–Rio, however, did not suggest differences in numbers tranquil/asleep (Evandro Coutinho, personal communication, 2003). However, evaluation of whether the dose of haloperidol matters will require a direct comparison with an adequately large sample.

Table 5

Proportion asleep if given haloperidol plus promethazine

TREC–India		TREC–Rio
	n (%)		n (%)
15 min	45 (45)	20 min	43 (29)
30 min	69 (69)	40 min	69 (46)
60 min	67 (67)	60 min	83 (55)
120 min	88 (88)	120 min	95 (63)

TREC–India did not compare haloperidol alone with a haloperidol–lorazepam combination, but adding promethazine to haloperidol could be superior to adding lorazepam for promoting tranquillisation and sedation and superior to haloperidol alone for preventing extrapyramidal adverse effects such as acute dystonia or akathisia (Reference Salzman, Solomon and MiyawakiSalzman et al, 1991; Reference Battaglia, Moss and RushBattaglia et al, 1997; Reference Brook, Lucey and GunnBrook et al, 2000). No serious adverse effects, particularly those related to the extrapyramidal system, were reported for either treatment, although the moderate worsening of respiratory difficulty reported with lorazepam is in keeping with the known association of benzodiazepines with respiratory depression.

About 15% of people in this trial were physically restrained and less than 10% were given additional medication over the 4 h. This common practice in India and Brazil (Reference Huf, Coutinho and FagundesHuf et al, 2002a ) of physically restraining disruptive patients after administration of a parenteral drug as opposed to administering additional medication requires evaluation.

The two treatment regimens evaluated in this study are inexpensive, effective and available worldwide. Where rapid sedation is needed a combination of intramuscular haloperidol and promethazine is superior to intramuscular lorazepam.

CLINICAL IMPLICATIONS

1. ▪ Intramuscular lorazepam (4 mg) is as effective as haloperidol (10 mg) plus promethazine (25/50 mg) in controlling violence or agitation in people with mental disorders.

2. ▪ If rapid sedation is required, the haloperidol–promethazine combination is superior to lorazepam.

3. ▪ Pragmatic randomised trials of interventions relevant to low-income countries, with limited funding, clinically meaningful outcomes and low attrition rates, are possible within the field of mental health.

LIMITATIONS

1. ▪ Assessments over the first 2 h were not blind and were carried out by multiple raters.

2. ▪ The effects of both interventions could be dose-related.

3. ▪ Haloperidol alone or in combination with a benzodiazepine was not evaluated.