We independently selected and assessed studies for methodological quality and extracted data. 'Summary of findings' tables were produced for each comparison grading the evidence and calculating, where possible and appropriate, a range of absolute effects.
MAIN RESULTS: We included 32 studies comparing
haloperidol with 18 other treatments. Few studies were undertaken in circumstances that reflect real world practice, and, with notable exceptions, most were small and carried considerable risk of bias.Compared with placebo, more people in the
haloperidol group were asleep at two hours (2 RCTs, n = 220, risk ratio (RR) 0.88, 95% confidence interval (CI) 0.82 to 0.95).
Dystonia was common (2 RCTs, n = 207, RR 7.49, CI 0.93 to 60.21). Compared with
aripiprazole, people in the
haloperidol group required fewer
injections than those in the
aripiprazole group (2 RCTs, n = 473, RR 0.78, CI 0.62 to 0.99). More people in the
haloperidol group experienced
dystonia (2 RCTs, n = 477, RR 6.63, CI 1.52 to 28.86).Despite three larger trials with
ziprasidone (total n = 739), data remain patchy, largely because of poor design and reporting. Compared with
zuclopenthixol acetate, more people who received
haloperidol required more than three
injections (1 RCT, n = 70, RR 2.54, CI 1.19 to 5.46).Three trials (n = 205) compared
haloperidol with
lorazepam. There were no significant differences between the groups with regard to the number of participants asleep at one hour (1 RCT, n = 60, RR 1.05, CI 0.76 to 1.44). However, by three hours, significantly more people were asleep in the
lorazepam group compared with the
haloperidol group (1 RCT, n = 66, RR 1.93, CI 1.14 to 3.27). There were no differences in numbers requiring more than one injection (1 RCT, n = 66, RR 1.14, CI 0.91 to 1.43).
Haloperidol's adverse effects were not offset by addition of
lorazepam (e.g.
dystonia 1 RCT, n = 67, RR 8.25, CI 0.46 to 147.45; required antiparkinson medication RR 2.74, CI 0.81 to 9.25). Addition of
promethazine was investigated in one larger and better graded trial (n = 316). More people in the
haloperidol group were not tranquil or asleep by 20 minutes (RR 1.60, CI 1.18 to 2.16). Significantly more people in the
haloperidol alone group experienced one or more adverse effects (RR 11.28, CI 1.47 to 86.35). Acute
dystonia for those allocated
haloperidol alone was too common for the trial to continue beyond the interim analysis (RR 19.48, CI 1.14 to 331.92).
AUTHORS' CONCLUSIONS: If no other alternative exists, sole use of intramuscular
haloperidol could be life-saving. Where additional drugs to offset the adverse effects are available, sole use of
haloperidol for the extreme emergency, in situations of coercion, could be considered unethical. Addition of the sedating
promethazine has support from better-grade evidence from within randomised trials. Use of an alternative
antipsychotic drug is only partially supported by fragmented and poor-grade evidence. Evidence for use of newer generation
antipsychotic alternatives is no stronger than that for older drugs. Adding a
benzodiazepine to
haloperidol does not have strong evidence of benefit and carries a risk of additional harm.After six decades of use for emergency rapid tranquillisation, this is still an area in need of good independent trials relevant to real world practice.