Violent or aggressive behaviour is a psychiatric emergency that necessitates swift mobilisation of staff and resources. Commonly it is secondary to psychotic symptoms, physical illness (e.g. delirium) and substance abuse [1] . Violent behaviour in an emergency room setting elicits varied responses and inexperienced staff may be required to handle such situations. This can have an impact on care of other patients with acute medical problems. On psychiatric wards, violent behaviour is likely to provoke aggression amongst other inpatients thereby stretch- ing staff resources. Effective management of these volatile situations can reduce risks to the patient, other service users, staff and family members who are often victims of aggression [2] .

Rapid tranquillisation is the use of medication to manage agitated or aggressive behaviour [3] and is used when other psychological and behavioural approaches have failed to calm the patient. A literature search of rapid tranquillisation revealed eight surveys of clinical practice and two surveys of clinician preferences (Table 1). There was failure of consensus on choice of medication for rapid tranquillisation—probably reflecting local cultural influences on prescription practices. All but one survey were from the western world. Appraisal of various guidelines around the world for schizophrenia suggests that recommendations are not uniform and quality of guideline may vary [4] . We have not come across any similar appraisal of guidelines on rapid tranquillisation and this is one important piece of the puzzle that we felt needs to be solved.

: Surveys of clinicians’ preferred choices; ^*BNF: British National Formulary.

The Appraisal of Guidelines Research and Evaluation (AGREE) is an instrument used to assess the methodological rigour and potential biases involved in guideline development. It also checks for the internal and external validity of the recommendations [5] . It can assess any new, existing or updated guideline. It consists of 23 key items organised in six domains each of which rates a separate dimension of guideline. Domain 1 is “Scope and Purpose” which captures the overall aim of the guideline and its target group. Domain 2 is “Stakeholder Involvement” which captures the extent to which appropriate stakeholders were involved in developing the guideline and also represents the views of its intended users. Domain 3 is “Rigour of Development” which looks at the process of gathering and summarizing the evidence used and the methods used to develop its recommendations. Domain 4 is “Clarity of Presentation” capturing the language, structure and format of the guideline. Domain 5 is “Applicability” looking at the potential barriers and facilitators to implementation, strategies to improve uptake and resources needed to implement the guideline. Domain 6 is “Editorial Independence” which captures biases caused by any other competing interests.

Four independent raters rate the guidelines on these domains using a four point scale. The scores on the six domains are independent and cannot be aggregated. These scores help to compare the different guidelines and to decide whether or not to recommend a particular guideline over others. However it is not possible to set thresholds for the domain scores to demarcate a good guideline from a bad one.

Guidelines on rapid tranquillisation from all English speaking countries were appraised using AGREE, with special emphasis on the NICE guidelines [6] . An English-speaking country was defined as a country with English as one of the official languages and with more than 50% residents speaking English. The 50% mark was arbitrarily chosen as a cut off figure. This was because countries where the majority of the population spoke a language other than English as first language were likely to have guidelines in local languages. This meant that there would be innumerable guidelines to assess, which would have been difficult to achieve given the language barrier.

Major databases were searched through the electronic database OVID. The databases included EMBASE (1980 to October week 3 2008), CINAHL (1982 to October week 3 2008), MEDLINE (1950 to October week 3 2008) and PsycINFO (1806 to October week 3 2008).The search terms used were “rapid tranquillisation” “tranquillisation” “behavioural emergencies”, “aggression”, “psychiatry emergencies”, ”guidelines in psychiatry” and “expert consensus guidelines”. After dropping the duplicates and hand searching the abstracts seven guidelines in English language were obtained from five countries: UK, USA, Canada, Australia and New Zealand which met our inclusion criteria. The eighth guideline from Singapore was excluded as it did not meet the criterion of an English-speaking country.

Four raters used the AGREE tool to objectively assess potential biases of guidelines. The assessors were trainee psychiatrists at different levels of their training. Two of them were senior house officers: one in his first year of training and the second in his second year of training. A specialist registrar (one of the authors) and a staff grade psychiatrist both in their fourth year of training formed the team of assessors. The raters had received prior instructions regarding the scoring process. The guidelines were scored on six domains mentioned above. The scores were then standardised according to validated recommendations which could range from 0 to 100%. The content of the guidelines was analysed separately as it was not rated by the AGREE tool.

Seven guidelines identified as above were compared for their pharmacological recommendations (Table 2).

Their methodological quality was evaluated using AGREE (Table 3). The scores from the four raters on the AGREE instrument ranged from a minimum of 33 on the domain of applicability for APA guidelines to the maximum of 100 on the applicability domain of NICE guidelines and editorial independence domain of the APA guidance. All guidelines scored well on the domain of scope and purpose. The NICE guidelines consistently scored well on all domains with the exception of editorial independence on which it performed mode rately with a score of 63.

FGA = first generation antipsychotic; SGA = second generation antipsychotic; BNZ = benzodiazepine. Colour code: Yellow = 1^st choice, Green = 2^nd choice, Red = 3^rd choice.

Common themes in clinical practice can be identified. For instance, benzodiazepines are chosen when little background information about the patient is available. This is in keeping with the Expert Consensus Guidelines [7] . Benzodiazepines are administered in psychiatric emergencies more frequently than other agents especially when the diagnosis is unknown [8] . Haloperidol plus promethazine combination is prevalent in Brazil [9] and India [10] . Its use is supported by high quality evidence showing its consistent superiority over olanzapine, lorazepam and haloperidol [10] . However this combination does not find a place in most guidelines with the exception of NICE. Lack of uniformity in recommendations despite using a common evidence pool is indeed intriguing. This could be a result of having only small trials with methodological inadequacies in this area. The

lack of good quality evidence necessitates those drawing up guidance to draw conclusions that are not founded on best possible evidence. For instance the only existing trial of zuclopenthixol acetate has a sample size of 40 patients [11] yet Australian and New Zealand guidelines [12] recommend it for rapid tranquillisation unlike the NICE and Canadian guidelines [13] . However even the NICE and Canadian guidelines have used equally thin evidence base for their recommendations. They differ on their inclusion of first and second generation anti psychotics thereby suggesting that guidelines are a combination of personal preferences and evidence base that is in turn influenced by selective interpretation of data.

Consensus on rapid tranquillisation guidelines is the need of the hour. Whether better quality clinical trials can help us reach an AGREEment remains to be seen.

None.