Study selection

The search for this update (from October 2007 to January 2013) resulted in 2083 hits. After screening of the titles and abstracts five potentially relevant studies were identified. Unfortunately, these studies had to be excluded because of the following reasons.

• The study by Hill et al (Hill 2008) compared the efficacy of the antipsychotic to another antipsychotic (that is droperidol).

• The study by Friedman et al (Friedman 2008) compared the efficacy of the antipsychotic to metoclopramide.

• The study by Miller et al (Miller 2009) compared the efficacy of the antipsychotic to octreotide.

• The study by Kostic et al (Kostic 2010) compared a combination of prochlorperazin with diphenhydramine to sumatriptan.

• The study by Potvin et al (Potvin 2012) did not use pain as the primary outcome parameter.

The original search strategy, run in October 2007, resulted in 1908 hits. After screening of the titles and abstracts, 56 potentially relevant studies were identified. Forty‐one studies were excluded (see 'Characteristics of excluded studies' table). In short, 39 studies did not meet the quality criteria as assessed by the Oxford Quality Scale. Two high‐quality studies (Brousseau 2004; Weaver 2004;) were excluded because they reported the efficacy of an antipsychotic on headache in children and adolescents (Brousseau 2004) and compared two antipsychotics (droperidol vs. prochlorperazine) without the use of a placebo (Weaver 2004). Hence, 15 studies were considered for inclusion in this review. Two of these studies could not be assessed because a translation into English was not available (Govorin 1990; Lepola 1984). For one further study the authors could not be identified, and it was therefore excluded (Anon 1986).

Overall 12 RCTs of nine different antipsychotics were considered eligible (Bussone 1980; Davidsen 1979; Ginsberg 1983; Graff‐Radford 2000; Honkaniemi 2006; Johnston 1972; Judkins 1982; Langemark 1994; Lechin 1989; Richman 2002a; Roux 1983; Zitman 1991) (n = 772) for inclusion in the review, please see the 'Characteristics of included studies' table for full details of each included study. All included studies were clinic based and single centered, with one (Lechin 1989) explicitly stating the inclusion of outpatients. They were conducted at the departments for neurology (n = 4) (Bussone 1980; Honkaniemi 2006; Langemark 1994; Lechin 1989), anaesthesiology (n = 2) (Graff‐Radford 2000; Judkins 1982), psychiatry (n = 1) (Zitman 1992), neurosurgery (n = 1) (Roux 1983), oncology (n = 1) (Johnston 1972) and the emergency department (n = 2) (Davidsen 1979; Richman 2002a). The site of the remaining study (Ginsberg 1983) could not be specified. The sample size ranged from 29 to 316 participants. Most trials were limited by their small sample size. Only one trial included more than 200 participants (Davidsen 1979). Eight studies were placebo‐controlled and were considered for quantitative analysis. Data from six studies could not be included in the quantitative analysis because of the following reasons.

• Two studies compared the efficacy of the antipsychotic to an opioid (Davidsen 1979) or a selective serotonin reuptake inhibitor (SSRI) (Langemark 1994).

• One study only reported on the occurrence of headaches following an intervention (Roux 1983).

• One study did not provide information about the duration of treatment (Bussone 1980).

• The work of Johnston 1972 fulfilled the inclusion criteria, but due to missing data on variability, the study could not be included in the final analyses. Similarly, the study by Judkins 1982 had to be excluded due to missing data on the mean and variability of the selected outcome variables.

Risk of bias is shown in Figure 1 for each included study.

Figure 1

---

---

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

For the meta‐analyses, the primary outcome variable was the difference in mean in the treatment and placebo groups (after ‐ baseline) and therefore reflected how much the VAS score could be reduced by the treatment. Additionally, the meta‐analysis was performed for the mean outcome VAS score following treatment in both groups. For the dichotomous event, feeling pain or no pain after treatment, the RR indicated the chance of having pain after receiving the intervention, an absolute risk reduction (ARR) < 1 means that the chance of having pain after the intervention is smaller in the treatment group than in the control group. The study authors were contacted to provide missing data, if necessary. Standard errors were converted into standard deviations. If the standard deviation of the difference in mean was not given, but the standard deviation from baseline and after treatment was given, imputation strategies were used (Higgins 2011). If values for the mean or standard deviation were not mentioned in the text, but were displayed in a figure, values were taken from the figure.

Study design

Nine of the included studies had a parallel design and three had a cross‐over design. Some trials had more than two arms and made more than one comparison.

Outcomes

Pain was patient reported in 12 trials. In one study information on pain was provided by the patients' weekly ratings (Johnston 1972). In six trials pain was reported using a pain diary (Bussone 1980; Ginsberg 1983; Graff‐Radford 2000; Langemark 1994; Lechin 1989; Zitman 1991). Three trials documented the analgesic requirements of the patients on a numeric VAS (Judkins 1982; Richman 2002a; Honkaniemi 2006). Two trials simply reported on the mere occurrence of pain (Davidsen 1979; Roux 1983).

Study methods

All 12 included studies were conducted in a double‐blind fashion. Eight trials compared an antipsychotic or a combination of analgesic and antipsychotic to placebo and three studies compared an antipsychotic or a combination of analgesic and antipsychotic to treatment with an active compound (that is antidepressants, antiepileptics or analgesics).

Antipsychotics

Trials using the following antipsychotics were found:

• Five with tricyclic antipsychotics (flupentixol, fluphenazine, thioridazine, levomepromazine);

• four with butyrophenones (droperidol, haloperidol);

• three with benzamides (sulpiride, tiapride, pimozide).

Patient conditions

The underlying conditions studied were as follows:

• somatoform pain disorder, one study;

• post‐herpetic neuralgia, one study;

• acute (migraine) headache, two studies;

• pain in terminal cancer, one study;

• postoperative pain, one study;

• trigeminal neuralgia, one study;

• acute rheumatic pain, one study;

• chronic tension‐type headache, two studies;

• post‐rachiocentesis headache, one study; and

• acute myocardial infarction, one study.

Details of these eligible reports are provided in the 'Characteristics of included studies' table.

Tricyclic antipsychotics

Four trials (Davidsen 1979; Graff‐Radford 2000; Johnston 1972; Zitman 1991), with a total of 449 participants, studied the effects of tricyclic antipsychotics in different painful disorders. These trials studied the effect of tricyclic antipsychotics in:

1. pain following myocardial infarction (Davidsen 1979),

2. pain due to terminal cancer (Johnston 1972),

3. somatoform pain disorder (Zitman 1991), and

4. post‐herpetic neuralgias (Graff‐Radford 2000).

The description of the study results is given below.

Butyrophenones

Three placebo‐controlled RCTs (Honkaniemi 2006; Judkins 1982; Richman 2002a), with a total of 110 participants, studied the effects of butyrophenones on postoperative pain (Judkins 1982) and acute migraine headache (Honkaniemi 2006; Richman 2002a).

In the case of postoperative pain two different dosages of haloperidol (5 and 10 mg orally) were compared against placebo as premedication before major abdominal surgery (Judkins 1982). VAS scores for pain at 24 hours after surgery did not differ significantly between the three groups of participants. Only a significant reduction of postoperative emesis was found in both groups treated with haloperidol. In contrast, treatment of acute migraine headache with 5 mg intravenous haloperidol was shown to be significantly superior to placebo (MD ‐4.05, 95% CI ‐5.61 to ‐2.49) (Honkaniemi 2006). Another study on acute migraine compared 2.5mg intramuscular droperidol with 1.5mg/kg meperidine and failed to detect a significant difference regarding post‐treatment pain intensity (VAS) (47 vs. 37 mm, p=.033) (Richman 2002a).

Benzamides

Five double‐blind trials with a total of 240 participants, two of them placebo‐controlled (Bussone 1980; Roux 1983), studied the effects of benzamides on different types of headache (Bussone 1980; Langemark 1994; Roux 1983), trigeminal neuralgia (Lechin 1989) and acute rheumatic pain (Ginsberg 1983).

In the case of chronic tension‐type headache tiapride was compared with placebo (Bussone 1980) and sulpiride was compared with another active component, that is paroxetine an SSRI (Langemark 1994). It seemed noteworthy that five participants in the group receiving sulpiride dropped out during the study due to intolerable side effects (Langemark 1994), see Table 1 for further details.