Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Review - Intervention

Betahistine for symptoms of vertigo

Authors' declarations of interest

Version published: 21 June 2016 Version history

https://doi.org/10.1002/14651858.CD010696.pub2
Collapse all Expand all

Abstract

available in

Background

Vertigo is a symptom in which individuals experience a false sensation of movement. This type of dizziness is thought to originate in the inner ear labyrinth or its neural connections. It is a commonly experienced symptom and can cause significant problems with carrying out normal activities. Betahistine is a drug that may work by improving blood flow to the inner ear. This review examines whether betahistine is more effective than a placebo at treating symptoms of vertigo from different causes.

Objectives

To assess the effects of betahistine in patients with symptoms of vertigo from different causes.

Search methods

The Cochrane ENT Information Specialist searched the Cochrane ENT Trials Register; Central Register of Controlled Trials (CENTRAL 2015, Issue 8); PubMed; EMBASE; CINAHL; Web of Science; ClinicalTrials.gov; ICTRP and additional sources for published and unpublished trials. We also contacted manufacturers and researchers in the field. The date of the search was 21 September 2015.

Selection criteria

We included randomised controlled trials of betahistine versus placebo in patients of any age with vertigo from any neurotological diagnosis in any settings.

Data collection and analysis

We used the standard methodological procedures expected by Cochrane. Our primary outcome was the proportion of patients with reduction in vertigo symptoms (considering together the intensity, frequency and duration those symptoms).

Main results

We included 17 studies, with a total of 1025 participants; 12 studies were published (567 patients) and five were unpublished (458 patients). Sixteen studies including 953 people compared betahistine with placebo. All studies with analysable data lasted three months or less. The majority were at high risk of bias, but in some the risk of bias was unclear. One study, at high risk of bias, included 72 people with benign paroxysmal positional vertigo (BPPV) and compared betahistine with placebo; all patients also had particle repositioning manoeuvres. The studies varied considerably in terms of types of participants, their diagnoses, the dose of betahistine and the length of time it was taken for, the study methods and the way any improvement in vertigo symptoms was measured. Using the GRADE system, we judged the quality of evidence overall to be low for two outcomes (proportion of patients with improvement and proportion with adverse events).

Pooled data showed that the proportion of patients reporting an overall reduction in their vertigo symptoms was higher in the group treated with betahistine than the placebo group: risk ratio (RR) 1.30, 95% confidence interval (CI) 1.05 to 1.60; 606 participants; 11 studies). This result should be interpreted with caution as the test for statistical heterogeneity as measured by the I2 value was high.

Adverse effects (mostly gastrointestinal symptoms and headache) were common but medically serious events in the study were rare and isolated: there was no difference in the frequency of adverse effects between the betahistine and placebo groups, where the rates were 16% and 15% respectively (weighted values, RR 1.03, 95% CI 0.76 to 1.40; 819 participants; 12 studies).

Sixteen per cent of patients from both the betahistine and the placebo groups withdrew (dropped out) from the studies (RR 0.96, 95% CI 0.65 to 1.42; 481 participants; eight studies).

Three studies looked at objective vestibular function tests as an outcome; the numbers of participants were small, techniques of measurement very diverse and reporting details sparse, so analysis of this outcome was inconclusive.

We looked for information on generic quality of life and falls, but none of the studies reported on these outcomes.

Authors' conclusions

Low quality evidence suggests that in patients suffering from vertigo from different causes there may be a positive effect of betahistine in terms of reduction in vertigo symptoms. Betahistine is generally well tolerated with a low risk of adverse events. Future research into the management of vertigo symptoms needs to use more rigorous methodology and include outcomes that matter to patients and their families.

PICOs

Population
Intervention
Comparison
Outcome

The PICO model is widely used and taught in evidence-based health care as a strategy for formulating questions and search strategies and for characterizing clinical studies or meta-analyses. PICO stands for four different potential components of a clinical question: Patient, Population or Problem; Intervention; Comparison; Outcome.

See more on using PICO in the Cochrane Handbook.

Betahistine for symptoms of vertigo

Review question

Do patients suffering from vertigo from different causes benefit from the drug betahistine?

Background

Vertigo is a symptom in which individuals experience a false sensation of movement. This type of dizziness is thought to originate in the inner ear balance organ or its connections to the brain. It is a commonly experienced symptom and can cause significant problems with carrying out normal activities. Betahistine is a drug that may work by improving blood flow to the inner ear. This review examines whether betahistine is more effective than a placebo (sham medicine) at treating symptoms of vertigo from different causes in patients of any age.

Study characteristics

We included 17 studies, with a total of 1025 participants. Sixteen studies including 953 people compared betahistine with placebo; the studies were at high to unclear risk of bias. All studies with analysable data lasted three months or less. One study, at high risk of bias, included 72 people with benign paroxysmal positional vertigo (BPPV) and compared betahistine with placebo; all patients also had particle repositioning manoeuvres. We judged the quality of evidence overall to be low.

The studies varied considerably in terms of types of participants, their diagnoses, the dose of betahistine and the length of time the drug was taken for, the study methods and the way any improvement in vertigo symptoms was measured.

Key results

When all studies are taken together, the proportion of patients reporting a reduction of their vertigo symptoms was significantly higher in the betahistine group than in the placebo group. However, there was significant variability in the results of the studies so this result should be treated with caution.

The proportion of patients reporting side effects of the medication was similar in both groups: 16% in the betahistine groups and 15% in the placebo groups. Overall, 16% of patients of both groups withdrew from the studies.

There was insufficient information about the effect of betahistine on objective tests of inner ear balance organ function. There was no information on the effect of betahistine on overall quality of life or falls.

Quality of the evidence

We judged the quality of evidence from the included studies to be low, meaning our estimates of the effects of betahistine could turn out to be inaccurate. The evidence is up to date to September 2015.

Conclusion

Low quality evidence suggests that patients suffering from vertigo from different causes may have some benefit from betahistine in terms of reduction in vertigo symptoms. Betahistine is generally well tolerated. Future research into the management of vertigo symptoms needs to use more rigorous methodology and include outcomes that matter to patients and their families.

Authors' conclusions

Implications for practice

Evidence that is largely of low quality suggests that in patients suffering from vertigo from different neuro‐otological causes there may be a positive effect of betahistine in terms of reduction in vertigo symptoms. The same evidence suggests that betahistine is generally well tolerated with a similar risk of adverse events to placebo treatments.

Why might betahistine have an effect on such a heterogeneous group of patients with vertigo from so many different and contrasting conditions? One would have to hypothesise that the symptom of vertigo in a significant number of these individuals might have a common pharmacological basis located presumably in the labyrinth or brain or the connections thereof, and that betahistine is able to influence this favourably. However, the symptom of vertigo has many possible causes. The findings of this review do not negate the need for a proper clinical assessment of patients with the symptom of vertigo with the goal of making a diagnosis. There are many other evidence‐based treatments for particular conditions that cause vertigo, which should be offered where appropriate.

This review and analysis were set up to answer the question, "is betahistine of overall benefit to patients with symptoms of vertigo?". Patients and their doctors will want to know whether the overall benefit from betahistine, if there is one, is large or small, and whether it is worth the risk of developing adverse effects. The review was not set up to analyse the size of any benefit since we examined only whether the patient judged that there was overall improvement of any degree, which makes it difficult for us to comment on how large the effect was. What we can say of the outcome measured was that patients overall felt there was benefit to them of taking the drug, taking all the relevant factors into consideration. We can also say that the number of patients who identified such a benefit over and above the placebo effect was small.

Implications for research

Future research into the effectiveness of betahistine in patients with vertigo should use rigorous methodology. There is a requirement for the development of and adherence to standardised diagnostic criteria for the selection of patients.

We also recommend the development of validated, patient‐centred outcome measures for research in the field of balance disorders. At the time of the publication of this review, core outcome measures for dizziness had not been identified.

Randomisation and blinding should be of the highest quality, given the subjective nature of vertigo and the strong likelihood of a placebo response. Future studies should be conducted and reported according to the CONSORT statement. Recruitment of adequate numbers has clearly been problematic for researchers and this should be considered in future trial designs, such as by using a multi‐centred trial design.

Summary of findings

Open in table viewer
Summary of findings 1. Betahistine versus placebo for symptoms of vertigo

Betahistine versus placebo for symptoms of vertigo

Patient or population: patients with symptoms of vertigo
Setting: outpatient clinics
Intervention: betahistine
Comparison: placebo

Time frame: up to 3 months

Outcomes

Relative effect
(95% CI)

Anticipated absolute effects* (95% CI)

Quality of the evidence
(GRADE)

What happens

Without betahistine

With betahistine

Difference

Benefits

Proportion of patients with improvement according to global judgement of patient subgrouped by diagnosis
№ of participants: 606
(11 RCTs)

RR 1.30
(1.05 to 1.60)

Moderate

⊕⊕⊝⊝
LOW 1 2 3

If 100 patients with vertigo are treated with betahistine, 60 will improve. This is 14 more than would have improved if a sham medicine had been taken instead of betahistine.

46.2%

60.1%
(48.5 to 73.9)

13.9% more
(2.3 more to 27.7 more)

Harms

Proportion of patients with adverse effects
№ of participants: 819
(12 RCTs)

RR 1.03
(0.76 to 1.40)

Moderate

⊕⊕⊝⊝
LOW 2 4

If 100 patients with vertigo are treated with betahistine, 16 will experience adverse effects. This is 1 more than would have had similar symptoms if a sham medicine had been taken instead of betahistine.

15.2%

15.7%
(11.6 to 21.3)

0.5% more
(3.6 fewer to 6.1 more)

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio

GRADE Working Group grades of evidence
High quality: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1Although statistical heterogeneity (I2 statistic) was 64%, the direction of effects was consistent.

2Most evidence was from studies with serious methodological limitations (unclear sequence generation, allocation concealment and blinding).

3Non‐validated outcome measures were used to measure improvement.

4Confidence intervals were wide and crossed thresholds of important benefits and harms.

Background

Description of the condition

Dizziness is a term that is commonly used by patients to describe various sensations of lightheadedness, imbalance, illusory feelings of movement or disorientation. Unsurprisingly, perhaps, given its lack of specificity, the experience of dizziness is common. Four per cent of all patients registered with a GP in the UK suffer persistent symptoms of dizziness and at least 3% are severely incapacitated by their symptoms (Nazareth 1999).

Vertigo is a specific subtype of dizziness. It is defined by the Bárány Society (the international balance disorders association) as "the sensation of self‐motion when no self‐motion is occurring or the sensation of distorted self‐motion during an otherwise normal head movement" (Bisdorff 2009). It may be a sensation of rotation ('spinning vertigo'), or may be a different sensation of self motion ('non‐spinning vertigo'). It is commonly, although not exclusively, caused by disease of the inner ear and can in this context be referred to as 'vestibular vertigo'.

In a large German epidemiological population‐based study, the lifetime prevalence of vestibular vertigo was estimated at 7.4% (Neuhauser 2005). The same study found that the lifetime prevalence of vestibular vertigo requiring a medical consultation was 5.8%. Estimates of the prevalence of significant vertigo impacting on daily life range from 3% to 10% (Murdin 2015).

The pattern of symptoms of vertigo is variable. In some cases symptoms may be mild or there may be a single short‐lived episode. Frequently, however, symptoms become prolonged or individuals become prone to recurrent attacks. The lifetime prevalence for recurrent attacks of vestibular vertigo is 6.5% (Neuhauser 2005). A Scottish study estimated that 21% of the population had experienced vertigo and 16% of these found the symptoms moderately or severely disruptive (Hannaford 2005). Importantly, vertigo increases the risk of falls, which in particular is becoming a major public health problem in the elderly.

Vertigo is a subjective experience. Its measurement is therefore dependent on the account of the individual experiencing it. There are some validated and well recognised instruments for assessing vertigo, for example the Dizziness Handicap Inventory (Jacobson 1990).

Vertigo has many causes including vestibular disorders such as Ménière's disease, vestibular neuritis, benign paroxysmal positional vertigo and migraine, each of which can be diagnosed by standardised criteria. Vestibular migraine is diagnosed according to criteria published jointly by the International Headache Society and the Bárány Society (Lempert 2012). Benign paroxysmal positional vertigo is diagnosed according to clinical criteria, as is vestibular neuritis (Strupp 2013). The Bárány Society and international collaborating organisations have recently published consensus clinical criteria for Ménière's disease (Lopez‐Escamez 2015), taking forward the previously widely used American criteria (AAO‐HNS 1995). Other causes of vertigo include neurological disorders affecting the central vestibular pathways (for example, some kinds of cerebellar stroke, or inflammatory or demyelinating pathologies) (Karatas 2010). Psychological disorders and primary cardiological disorders can also cause a sensation of vertigo (Newman‐Toker 2008; Wiltink 2009). It is therefore important to assess patients presenting with vertigo very carefully to identify the underlying diagnosis. However, symptomatic management of vertigo may be required before a definitive cause can be identified.

Description of the intervention

Betahistine is a drug treatment, available only in oral form, usually taken in doses from 24 mg to 48 mg daily. It is excreted via the urinary system. It is also known as betahistine dihydrochloride and has a number of different proprietary names, including Serc, Betaserc and Hiserk. It has been used in some countries for many years as a treatment for Ménière's disease or syndrome, where it has been thought to be especially effective for the symptoms of vestibular vertigo. This widely held view was challenged by a Cochrane review that found no evidence of benefit in Ménière's disease or syndrome (James 2001). However, it has also been used more broadly for the treatment of vestibular vertigo. In a study of medical practice in a variety of settings across 13 countries worldwide betahistine was the most common drug prescribed, being issued in two‐thirds of cases of vertigo. It was also the most common drug prescribed across every diagnostic group (Ménière's, benign paroxysmal positional vertigo, peripheral vestibular vertigo and 'other' vertigo) (Agus 2013). A German study set in primary care found that betahistine was prescribed in 6.6% of consultations for dizziness, and was most likely to be prescribed in 'unspecified dizziness', vestibular neuritis and benign paroxysmal positional vertigo (Kruschinski 2008).

The main adverse effects of betahistine relate to upper gastrointestinal tract symptoms; in general it is believed to be well tolerated.

How the intervention might work

Betahistine could act at either a peripheral (inner ear labyrinth) or central nervous sytem level, with current data favouring a predominantly peripheral mode of action.

Despite widespread use, the pharmacology of betahistine remains incompletely understood. Betahistine hydrochloride is a weak histamine H1 agonist and a strong H3 antagonist with virtually no H2 histamine receptor activity. Betahistine may act on the inner ear fluid mechanics by improving circulation in the cochlear stria vascularis (Ihler 2012; Martinez 1972), via an action on the precapillary sphincter with an associated reduction in excessive endolymphatic pressure, improving the function of vestibular hair cells.

Betahistine could also have effects on symptoms of vertigo via central nervous system activity. Betahistine can cross the blood‐brain barrier: the cell bodies of histamine‐containing neurons project throughout the brain, including the ventromedial hypothalamic nucleus, the thalamus and the cerebral cortex, and betahistine has measurable effects on regional cerebral blood flow (Barak 2008). Data from a single double‐blind, placebo‐controlled clinical study suggest significant effects of betahistine on some cognitive function tests (Pathy 1977). Another possible mechanism of action of betahistine is via inhibition of activity within the vestibular nuclei (Timmerman 1994).

Why it is important to do this review

Vertigo is a common symptom that has significant impact on the health and wellbeing of sufferers both at an individual and a population level. There are evidence‐based treatments for common conditions that cause symptoms of vertigo (e.g. particle repositioning manoeuvres for benign paroxysmal positional vertigo (Hilton 2014; Hunt 2012); drug treatments for migraine (Linde 2004); rehabilitation for unilateral vestibular disorders like vestibular neuritis (McDonnell 2015)), and the first step in evaluating patients with this symptom should always be a proper assessment with a view to making a clear diagnosis to guide evidence‐based management. However, there remains a need to evaluate other therapies for vertigo. This is because these treatments may be contraindicated in some patients and others may fail to respond or have ongoing symptoms. Importantly, many patients with vertigo do not meet clear‐cut diagnostic criteria for a defined condition so disease‐specific treatments cannot always be offered. In these cases betahistine is often prescribed in clinical practice.

Betahistine is a widely used treatment for Ménière's disease or syndrome. A Cochrane review showed that there is insufficient good evidence of an effect (James 2001). The authors suggested, however, that due to difficulties with adherence to the strict diagnostic criteria for Ménière's syndrome, a true positive effect of betahistine in patients with less well‐defined symptoms may have been missed.

Objectives

To assess the effects of betahistine in patients with symptoms of vertigo from different causes.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials. We planned to include cross‐over trials if the results from before the cross‐over were extractable, to avoid the potential for carry‐over effects.

Types of participants

Patients of any age with vertigo in community or other settings were eligible. Where patients were diagnosed with a specific vestibular condition, we noted the diagnostic criteria. We excluded those who had specific diagnoses of non‐neurotological causes for vertigo (such as anxiety disorders or cardiac disease). We included all categories of neurotological diagnosis (including, for example, central neurological conditions and vestibular schwannoma).

Types of interventions

We considered any trial of betahistine versus placebo. Planned comparisons were:

  • betahistine versus placebo;

  • betahistine with an additional intervention versus placebo with an identical additional intervention.

We included all dose regimes and all formulations (e.g. slow‐release preparations). We did not include comparisons with other drugs as their effects on vertigo have not been formally assessed. Concurrent use of other medication for non‐neurotological conditions was acceptable if used equally in each group. Where an additional intervention was also used equally in both groups, we analysed this as a separate comparison.

Types of outcome measures

We examined outcomes as short‐term (three months or under) and long‐term (over three months).

The outcome measures below were not used as a basis for including or excluding studies.

Primary outcomes

  • Proportion of patients with reduction in vertigo symptoms (considering together the intensity, frequency and duration those symptoms).

  • Proportion of patients with adverse effects. Betahistine is thought to cause upper gastrointestinal adverse effects and we recorded these separately.

Secondary outcomes

  • Proportion of participants withdrawing (dropping out) from the study due to all causes.

  • Generic quality of life (we assessed disease‐specific quality of life scales as part of the primary outcome).

  • Vestibular function as tested with objective vestibular function tests: caloric tests, rotation tests, posturography and vestibular evoked myogenic potentials.

  • Proportion of participants with falls, as a real‐life indicator of overall functional balance.

Search methods for identification of studies

The Cochrane ENT Information Specialist conducted systematic searches for randomised controlled trials and controlled clinical trials. There were no language, publication year or publication status restrictions. The date of the search was 21 September 2015.

Electronic searches

The Information Specialist searched:

  • the Cochrane ENT Trials Register (searched 21 September 2015);

  • the Cochrane Central Register of Controlled Trials (CENTRAL 2015, Issue 8);

  • PubMed (1946 to 21 September 2015);

  • Ovid EMBASE (1974 to 21 September 2015);

  • Ovid CAB Abstracts (1910 to 21 September 2015);

  • EBSCO CINAHL (1982 to 21 September 2015);

  • Ovid AMED (1985 to 21 September 2015);

  • LILACS, lilacs.bvsalud.org (searched 21 September 2015);

  • KoreaMed (searched via Google Scholar 21 September 2015;

  • IndMed, www.indmed.nic.in (searched 21 September 2015);

  • PakMediNet, www.pakmedinet.com (searched 21 September 2015);

  • Web of Knowledge, Web of Science (1945 to 21 September 2015);

  • CNKI, www.cnki.com.cn (searched via Google Scholar 21 September 2015);

  • ClinicalTrials.gov (searched via the Cochrane Register of Studies 21 September 2015);

  • World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP), www.who.int/ictrp (searched 21 September 2015);

  • ISRCTN, www.isrctn.com (searched 21 September 2015);

  • Google Scholar, scholar.google.co.uk (searched 21 September 2015);

  • Google, www.google.com (searched 21 September 2015).

The Information Specialist modelled subject strategies for databases on the search strategy designed for CENTRAL. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying randomised controlled trials and controlled clinical trials (as described in the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0, Box 6.4.b. (Handbook 2011)). Search strategies for major databases including CENTRAL are provided in Appendix 1.

Searching other resources

We scanned the reference lists of identified publications for additional trials and contacted trial authors where necessary. In addition, the Information Specialist searched PubMed, TRIPdatabase, The Cochrane Library and Google to retrieve existing systematic reviews relevant to this systematic review, so that we could scan their reference lists for additional trials.

Data collection and analysis

Selection of studies

Two authors (LM and KH) independently scanned the initial search results to identify trials that appeared to meet the inclusion criteria. We used abstract review to eliminate any trials that were clearly ineligible. If either author identified a paper as potentially suitable, we reviewed the full text of the article. We resolved disagreements by discussion or with the input of the third author (AS).

Data extraction and management

Two authors (LM and KH) extracted data independently and synthesised the results. We used standardised data entry forms. There was no blinding of journal, author names or affiliations. With regard to subgroup analysis, we extracted data on underlying diagnosis if applicable, along with treatment protocol (dose and duration of drug).

For each study, we extracted information on study design, duration, randomisation, concealment, number of participants, setting of study, diagnostic criteria and exclusion criteria, age and sex distribution of participants, country of recruitment, co‐morbidity, date of study, number of intervention groups, betahistine dose and duration, outcomes measured and definition of outcomes, missing data and final sample size, data on intensity, frequency and duration of vertigo, and data from other vertigo scales.

For the outcome proportion of patients with an improvement in symptoms, a variety of (non‐validated) scales with different numbers of ordinal points were reported in the studies. Two different review authors independently dichotomised these into 'improved' or 'not improved' whenever possible.

When the required data were not available in the published accounts, we contacted the principal investigator to request the data.

Assessment of risk of bias in included studies

LM and KH undertook assessment of the risk of bias of the included trials independently, with the following taken into consideration, as guided by the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2011):

  • sequence generation;

  • allocation concealment;

  • blinding;

  • incomplete outcome data;

  • selective outcome reporting; and

  • other sources of bias.

We used the Cochrane 'Risk of bias' tool in RevMan 5 (RevMan 2014), which involves describing each of these domains as reported in the trial and then assigning a judgement about the adequacy of each entry: 'low', 'high' or 'unclear' risk of bias.

We resolved differences of opinion by discussion in the first instance, with input from the third author (AS) if necessary.

Measures of treatment effect

The primary outcome measure of this review was the proportion of individuals with benefit from the drug, which is a dichotomous measure.

For binary (dichotomous) data we calculated the risk ratio (RR).

For intervention effect measures with continuous data we planned to calculate the difference in means (mean difference, MD) between the groups, provided that different studies used the same scale of measurement. We planned to calculate the standardised mean difference (SMD) if different scales were used.

For studies with ordinal data we checked, where possible, to see if the scale had been validated. Depending on the number points in these scales (and how the data were reported), we either dichotomised these or analysed them as continuous outcomes.

Unit of analysis issues

Cluster‐randomised trials

Cluster‐randomised trials allocate groups instead of individuals. The participants in each group may be related in some way, therefore this needs to be taken into account in the analysis, otherwise there is a unit of analysis error, which would produce an artificially small P value and a risk of false positive results. We planned to analyse these according to guidance in the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2011). However, we identified no cluster‐randomised trials.

Cross‐over trials

Cross‐over trials may have a carry‐over effect. We have included all patients with vertigo, some of which may resolve quite quickly. In view of this, and the chronic and episodic nature of the condition of interest (vertigo), we used data from cross‐over trials only if data from before the cross‐over could be obtained.

Multi‐arm studies

Where we found studies with more than two groups (e.g. two or more active treatments being tested against placebo), we established which of the comparisons were relevant to the systematic review and relevant to each of the meta‐analyses that we implemented. Where the study design used independent groups, we treated the study as an independent comparison.

Repeated observations on participants

In longer studies, results may be recorded at more than one time interval. In order to avoid a unit of analysis error when combining these results in a single meta‐analysis (and therefore counting the same participants in more than one comparison), we would have defined different outcomes, based on different periods of follow‐up, performing separate analyses.

Dealing with missing data

We did not plan or implement any statistical strategies to deal with missing data, except for imputations to estimate missing standard deviations according to the methods recommended in Section 7.7.3 of the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2011). We conducted available case analysis.

Assessment of heterogeneity

We assessed studies for clinical, statistical and methodological heterogeneity.

We assessed heterogeneity by inspection of the point estimates and confidence intervals on the forest plots. We assessed the variation in treatment effects by means of the Cochrane test for heterogeneity and quantified it using the I2 statistic.

An approximate guide to interpretation provided in the Cochrane Handbook for Systematic Reviews of Interventions is as follows (Handbook 2011):

  • 0% to 40%: might not be important;

  • 30% to 60%: may represent moderate heterogeneity;

  • 50% to 90%: may represent substantial heterogeneity;

  • 75% to 100%: considerable heterogeneity.

We also used the Chi2 test, using the indicator that where Chi2 was greater than the degrees of freedom, heterogeneity was likely to be present. We considered heterogeneity statistically significant if the P value was < 0.1.

Assessment of reporting biases

Reporting bias can be assessed as between‐studypublication bias or within‐study reporting bias.

Between‐study publication bias

Where there was a sufficient number of trials (more than 10) in any meta‐analysis, we assessed publication bias according to the recommendations on testing for funnel plot asymmetry as described in Section 10.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Egger 1997; Handbook 2011).

Within‐study reported bias

We planned to assess within‐study reporting bias by comparing the outcomes reported in the published report against the outcomes reported in the study protocol, whenever this could be obtained. If a protocol could not be obtained, then we compared the outcomes listed in the methods section with those reported in the results.

Data synthesis

We planned to analyse all participants according to the group randomised in the studies. If the data were compatible and of sufficient quality we planned to combine them to give summary measures of effect. If sufficient data were available for different conditions (e.g. uncompensated vestibular disease), we planned to undertake subgroup analysis. As stated above, all conditions causing vestibular vertigo were to be included (including Ménière's disease), but at the subgroup analysis stage we did not plan to duplicate work already carried out on Ménière's disease since there is already a Cochrane review on this specific topic (James 2001).

Subgroup analysis and investigation of heterogeneity

We analysed subgroups by participant factors (diagnosis) and by intervention factors (dose of betahistine) to examine reasons for heterogeneity.

Sensitivity analysis

We planned to conduct a sensitivity analysis by comparing the effect of the inclusion and exclusion of studies based on eligibility criteria or data analysis methods where required.

GRADE and 'Summary of findings' table

We used the GRADE approach to rate the overall quality of evidence. The quality of evidence reflects the extent to which we are confident that an estimate of effect is correct and we applied this in the interpretation of results. There are four possible ratings: high, moderate, low and very low. A rating of high quality of evidence implies that we are confident in our estimate of effect and that further research is very unlikely to change our confidence in the estimate of effect. A rating of very low quality implies that any estimate of effect obtained is very uncertain.

The GRADE approach rates evidence from RCTs that do not have serious limitations as high quality. However, several factors can lead to the downgrading of the evidence to moderate, low or very low. The degree of downgrading is determined by the seriousness of these factors:

  • study limitations (risk of bias);

  • inconsistency;

  • indirectness of evidence;

  • imprecision; and

  • publication bias.

We included a 'Summary of findings' table, constructed according to the recommendations described in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2011). We included the following outcomes in the 'Summary of findings' table:

  • proportion of patients with improvement according to global judgement of patient subgrouped by diagnosis;

  • proportion of patients with adverse effects.

Results

Description of studies

See Characteristics of included studies; Characteristics of excluded studies; Characteristics of ongoing studies.

Results of the search

The electronic database search on 21 September 2015 identified 809 records. We identified an additional 10 records from handsearching, contacting manufacturers and experts, and from the reference lists of relevant studies. We were provided with five further unpublished studies for analysis. After removal of duplicates we were left with 445 records. We identified 32 potentially eligible studies and excluded 15 for reasons including having a cross‐over design with the data from before cross‐over not extractable, lack of adequate randomisation and ineligible participants. Seventeen studies met the inclusion criteria. The results of the search are shown in Figure 1 as a flow chart. The excluded studies are tabulated in the Characteristics of excluded studies table. There are no studies awaiting assessment and two studies are ongoing.

Figure 1
Process for sifting search results and selecting studies for inclusion.

Process for sifting search results and selecting studies for inclusion.

Included studies

We included 17 studies and these are summarised in the Characteristics of included studies table.

Only Guneri 2012 explicitly reported no financial conflict of interest. Six studies all acknowledge an association with the manufacturers of betahistine (Conraux 1988; Fischer 1985; Legent 1988; Mira 2003; Oosterveld 1989; Otto 2008), and five were unpublished industry studies (Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989; Duphar H10803592F 1997; Duphar H108906NL 1990).

Design

Most studies used a prospective, parallel‐group comparison design (Conraux 1988; Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989; Duphar H10803592F 1997; Duphar H108906NL 1990; Fischer 1985; Guneri 2012; Legent 1988; Mira 2003; Okamoto 1968; Otto 2008; Ricci 1987; Salami 1984). Three studies used a cross‐over design, from which data were extractable prior to cross‐over (Burkin 1967; Canty 1981; Oosterveld 1989). In Guneri 2012, betahistine was compared to placebo for the treatment of benign paroxysmal positional vertigo (BPPV) in addition to usual care (the Epley particle repositioning manoeuvre). All studies were double‐blinded.

Sample sizes

Sample size ranged from 10 (Ricci 1987) to 144 (Mira 2003). A total of 567 patients had results reported across the 12 published studies, and there were results for an additional 458 patients from the five unpublished studies (1025 patients in total).

Setting

The majority of studies were single‐centre and appeared to take place in specialist centres. Mira 2003, Oosterveld 1989 and Legent 1988 were multicentre studies. Studies took place in the USA (Burkin 1967), the UK (Canty 1981), France (Conraux 1988; Legent 1988), the Netherlands (Fischer 1985; Oosterveld 1989), Italy (Mira 2003; Ricci 1987; Salami 1984), Turkey (Guneri 2012), Japan (Okamoto 1968), and Germany (Otto 2008). Of the unpublished studies, there were two multicentre studies from France (Duphar H10802786F/M 1989; Duphar H10803592F 1997), and one single‐centre study each from the UK (Duphar 77054 1983), France (Duphar H10800580M 1984), and the Netherlands (Duphar H108906NL 1990).

Participants

Five studies included patients who were designated by the study authors as having clinically defined Ménière's disease or syndrome (Burkin 1967; Mira 2003; Okamoto 1968; Ricci 1987; Salami 1984). In only one of these were the AAO‐HNS 1995 diagnostic criteria cited (Mira 2003), with participants having probable or possible Ménière's according to the study authors.

Three studies included patients with episodic vertigo (Canty 1981; Fischer 1985; Oosterveld 1989), and one included those with episodic vertigo "with or without cochlear symptoms suggestive of Meniere's disease" (Legent 1988). Canty 1981 specified that symptoms must have a presumed peripheral origin and to have lasted at least 12 months. Two studies included patients with BPPV defined by a positive Dix‐Hallpike positioning test (Guneri 2012; Mira 2003). One study included patients with chronic vertigo (Conraux 1988), defined as at least six crises over the last two months and symptom duration of at least three months. One study included patients with "vertebrobasilar ischaemia", defined in this study as vertigo with at least two of impaired hearing, impaired vision, tinnitus or headache and "typical abnormalities" on test, which were not specified (Otto 2008).

Duphar 77054 1983, Duphar H10802786F/M 1989 and Duphar H10803592F 1997 included patients with Ménière's disease or episodic vertigo with cochlear symptoms, but with no strict diagnostic criteria and with other diagnoses included. Duphar H108906NL 1990 included patients with various diagnoses including a majority with BPPV and small numbers with other causes of episodic vertigo. Duphar H10800580M 1984 included patients labelled as having "central signs" with short‐lived episodes of vertigo, this list including changes in handwriting, spontaneous or induced/gaze evoked nystagmus, nystagmus on cervical or vertebrobasilar privation test and unilateral or bilateral hypo‐ or hyperexcitability of vestibular function (Duphar H10803592F 1997).

Burkin 1967 did not report exclusion criteria. Canty 1981, Fischer 1985, Guneri 2012, Legent 1988, Mira 2003, Okamoto 1968, Oosterveld 1989, Otto 2008, Salami 1984 and Ricci 1987 did report exclusion criteria. Patients on other relevant vestibular medication were excluded by Conraux 1988, Fischer 1985, Guneri 2012, Otto 2008 and Salami 1984. No published studies excluded those who had previously been on betahistine, but two unpublished studies did (Duphar H10803592F 1997; Duphar H108906NL 1990).

All studies were of adults, but one study reported that the youngest participants were teenagers (Okamoto 1968). The oldest participants in the studies were in the eighth decade (Guneri 2012; Otto 2008). Two studies did not report age and gender data for participants (Conraux 1988; Legent 1988). Some studies had an upper age limit for inclusion (Fischer 1985; Mira 2003; Oosterveld 1989 used 65 years as the upper limit; Duphar H10800580M 1984; Duphar H108906NL 1990; Duphar 77054 1983 used an upper limit of 70, Guneri 2012 used an upper limit of 79 years). All studies where gender was reported had mixed male and female participants.

Interventions

All studies compared betahistine against placebo.

Doses of betahistine ranged from daily totals of 16 mg (Burkin 1967), 24 mg (Ricci 1987; Salami 1984), 32 mg (Canty 1981; Mira 2003), 36 mg (Okamoto 1968; Otto 2008), or 48 mg (Fischer 1985; Legent 1988; Oosterveld 1989). The unpublished studies used totals of 36 mg (Duphar 77054 1983), or 48 mg (Duphar H10800580M 1984; Duphar H10802786F/M 1989; Duphar H10803592F 1997; Duphar H108906NL 1990). Some studies used variable doses (Conraux 1988). Duration of treatment available for analysis in this review was a fixed interval of two weeks (Burkin 1967; Guneri 2012; Okamoto 1968), one month (Duphar H10803592F 1997), five weeks (Oosterveld 1989), six weeks (Salami 1984), two months (Canty 1981; Duphar H108906NL 1990), and three months (Conraux 1988; Fischer 1985; Legent 1988; Mira 2003, Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989). Study duration was three months or less in all cases except Ricci 1987, where duration of therapy was variable (using a protocol of 10 times mean duration of interval between attacks for each patient to determine treatment length). In this small study of 10 patients therapy was for a mean of 10.4 months in the betahistine group and 7.0 months in the placebo group.

Guneri 2012 compared betahistine (48 mg daily) with placebo in patients with BPPV who had also received particle repositioning manoeuvres as an additional intervention.

Assessment of intervention integrity in the form of compliance was variably reported. Compliance checks were not reported in most studies (Burkin 1967; Conraux 1988; Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989; Guneri 2012; Legent 1988; Mira 2003; Ricci 1987; Salami 1984). Four studies made direct or indirect reference to compliance checks either by checking tablet containers or by labelling some patients as non‐compliant (Canty 1981; Fischer 1985; Okamoto 1968; Oosterveld 1989). Three studies reported that compliance was explicitly checked by direct questioning and container checks (Duphar H10803592F 1997; Duphar H108906NL 1990; Otto 2008).

Outcomes
Proportion of patients with reduction in vertigo symptoms (considering together the intensity, frequency and duration those symptoms)

Ten studies collected data on patient global satisfaction with treatment using various different ordinal scales, which we interpreted as overall improvement where the rating was positive (Conraux 1988; Duphar H108906NL 1990; Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10803592F 1997; Fischer 1985; Legent 1988; Mira 2003; Oosterveld 1989; Otto 2008). In addition, four studies reported vertigo scores in a way that enabled us to quantify the proportion of patients who experienced overall improvement in vertigo symptoms, using either dichotomous or ordinal scales of overall benefit in terms of vertigo (Burkin 1967; Canty 1981; Conraux 1988; Okamoto 1968). These scales were not described as validated.

Seven studies collected parallel data on the investigator global impression of treatment (Fischer 1985; Duphar H108906NL 1990; Duphar 77054 1983; Duphar H10802786F/M 1989; Duphar H10800580M 1984; Duphar H10803592F 1997; Mira 2003). However, in this review we have focused on the more clinically relevant outcome of patient‐reported improvement.

Guneri 2012 used published validated scales (Dizziness Handicap Inventory (Jacobson 1990), Vestibular Disorders Activities of Daily Living Scale (Cohen 2000), Vertigo Symptom Scale (Yardley 1998), and European Evaluation of Vertigo Scale (Megnigbeto 2001). Mira 2003 also used the Dizziness Handicap Inventory and some other scales whose validation references could not be obtained (Dizziness Assessment Rating Scale, GISFaV).

Ricci 1987 reported in narrative terms the small number of patients in that study. Burkin 1967 used a "dizzy or not" dichotomous outcome.

Intensity of vertigo

Okamoto 1968 used a three‐point, author‐defined ordinal scale to measure intensity of vertigo. Similarly, six studies used a four‐point ordinal scale (Canty 1981; Duphar 77054 1983; Duphar H10803592F 1997; Fischer 1985; Otto 2008; Salami 1984), and six studies used a five‐point ordinal scale (Conraux 1988; Duphar H10802786F/M 1989; Duphar H10800580M 1984; Duphar H10803592F 1997; Legent 1988; Oosterveld 1989). In Otto 2008, patients rated a number vertigo symptoms (unsteadiness, staggering, rotary sensation, tendency to fall, lift sensation, swaying, self motion triggered vertigo) and for each individual the mean score across all these symptoms was calculated.

Frequency of vertigo

Six studies reported the frequency of attacks (Duphar H108906NL 1990; Duphar H10803592F 1997; Fischer 1985; Legent 1988; Mira 2003; Oosterveld 1989). Duphar H10803592F 1997 reported the total number of attacks through the 30‐day study period. Duphar H10802786F/M 1989 reported the time since the last attack at study endpoint.

Duration of vertigo

Two studies also reported duration of vertigo attacks on an author‐defined four‐point ordinal scale (Fischer 1985; Oosterveld 1989). Legent 1988 reported mean duration of attacks per patient in hours. Mean duration of attacks in seconds was recorded by two studies (Duphar H108906NL 1990; Duphar H108906NL 1990). Salami 1984 reported total duration of attacks.

Proportion of patients with adverse effects

All studies except Guneri 2012 and Ricci 1987 made some comment on tolerability or safety. Duphar H10802786F/M 1989 did not describe adverse effects directly but measured "tolerance" on a four‐point ordinal scale.

Proportion of participants withdrawing (dropping out) from the study due to all causes

Withdrawal from study was reported clearly by six studies (Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989; Duphar H10803592F 1997; Mira 2003; Salami 1984).

Generic quality of life

No study included a general quality of life measure.

Vestibular function as tested with objective vestibular function tests

Three studies also reported vestibular function tests (caloric, stabilometry, nystagmography) (Canty 1981; Mira 2003; Salami 1984). Guneri 2012 and Mira 2003 used Dix‐Hallpike positioning testing to assess resolution of BPPV. Although one study had a majority of participants with BPPV, Dix‐Hallpike test outcomes are not reported (Duphar H108906NL 1990). Duphar 77054 1983 reported the intention to collect vestibulometric tests as outcomes, but these were not done at the "discretion of the investigator".

Proportion of participants with falls

No study reported on falls outcomes.

Excluded studies

See Characteristics of excluded studies for details of the 15 studies that we excluded. Five studies were of cross‐over design, with data before cross‐over not extractable (Frew 1976; Meyer 1985; Oosterveld 1984; Watanabe 1967; Wilmot 1976). We excluded six studies as there was no evidence of randomisation (Bertrand 1972; Hommes 1972; Purohit 1988; Singarelli 1979; Verspeelt 1996), or randomisation was inadequate (Elia 1966). We excluded two as the participants did not meet the criteria for the symptom definition of vertigo according to the review protocol (Redon 2011; Schmidt 1992).

Ongoing studies

Four registered clinical trials were identified through the search. Two studies were identified as progressing but with data not yet published. The co‐ordinator of these two trials confirmed progress by personal communication (BEMED; BETAVEST). One registered clinical trial was a drug company trial of betahistine for post‐vestibular neurotomy patients with Ménière's disease, which was closed in 2006 (NCT00160238). Another registered clinical trial entry (of betahistine for vertigo caused by cerebral infarction in posterior circulation) reported that the study was terminated early due to poor recruitment (NCT00474409). We contacted the registered companies for any results but these have not been received at the time of writing.

Risk of bias in included studies

Two authors (LM and KH) critically reviewed studies for risk of bias. We contacted lead study authors for further details of methodology where required. We also contacted authors for clarification of methodological issues where these were unclear. Prof Oosterveld replied to our enquiries to say that as the study took place such a long time ago, the original paperwork for the study is no longer available for inspection to clarify details or fill in missing data (Oosterveld 1989). No other responses had been received by the time of submission.

For sequence generation six studies had a low risk rating and 11 were unclear. For allocation concealment, two had a low risk rating and 15 were unclear. For blinding, three had a low risk rating, one had a high risk rating and 14 were unclear. For attrition bias, four were low risk and 10 were high risk. For reporting bias, nine were high risk and six low risk. In Canty 1981, some included participants in both groups had no symptoms at all throughout the whole trial duration, including the baseline assessment period, and this was also flagged up as a problem. Risk of bias is presented graphically in Figure 2 and Figure 3.

Figure 2
'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

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

Allocation

Sequence generation

All studies reported that they were randomised. The majority (11 studies) gave no information on how this was achieved (Burkin 1967; Canty 1981; Conraux 1988; Duphar 77054 1983; Duphar H10802786F/M 1989; Duphar H108906NL 1990; Fischer 1985; Guneri 2012; Legent 1988; Oosterveld 1989; Salami 1984); the risk of bias was unclear. Six studies reported details of the randomisation methods, which were adequate and we considered them as at low risk of bias (Duphar H10800580M 1984; Duphar H10803592F 1997; Mira 2003; Okamoto 1968; Otto 2008; Ricci 1987).

Allocation concealment

Only Mira 2003 and Okamoto 1968 reported details relating to allocation concealment, with both reporting that the allocation was done on a different site to the investigating centre. These studies were at low risk of bias. All the other studies were at unclear risk of bias.

Blinding

All studies were reported as "double‐blind" but most gave no further details, so we allocated these studies ratings of unclear risk.

Primary outcome measures were mostly by self report, unsurprisingly, given the subjective nature of vertigo symptoms. Mira 2003 reported that drugs were supplied in identical packages with a false name, so we awarded it a low risk rating. Guneri 2012 commented on blinding of the trial physician and we also awarded this study a low risk rating on this basis. Most studies commented that the placebo tablet was "identical" or "indistinguishable" from the betahistine preparation, with only two giving no information (Guneri 2012; Ricci 1987). In Duphar 77054 1983, it is stated that envelopes were provided to all participants stating allocation, however returned sealed envelope collection is not reported and opacity was not stated, therefore we judged this high risk. Likewise, Duphar H10803592F 1997 stated "neither patient nor investigator knew which treatment was being given" and we allocated low risk. Duphar H10802786F/M 1989 described coding envelopes all being returned unopened, but opacity was not stated so we rated this as unclear.

Incomplete outcome data

Mira 2003 reported that randomised patients were all accounted for and there was a low rate of attrition so we rated it low risk. For Duphar H10800580M 1984, enough information was given for intention‐to‐treat (ITT) analysis to be possible and losses were under 20%. We judged it low risk on this item. Canty 1981 lost data on five of 32 participants, but reported data on all including one of the five who was subsequently found to be ineligible. However, the dropouts occurred in the second phase of cross‐over, which is not relevant to the data considered for this review, so we deemed the study low risk.

Seven studies all reported some attrition but did not address this in the analysis or provide data that we could use to do so and so we rated them high risk (Duphar 77054 1983; Duphar H10802786F/M 1989; Duphar H108906NL 1990; Fischer 1985; Legent 1988; Oosterveld 1989; Otto 2008). For Duphar H10803592F 1997, information on all the lost participants was incomplete and we judged it high risk. Four studies did not give complete data for the group allocation of the lost participants (Conraux 1988; Fischer 1985; Legent 1988; Oosterveld 1989). Okamoto 1968 reported that two patients withdrew from each group, but not due to adverse effects (James 2001).

Salami 1984 reported zero attrition (personal communication reported in James 2001) and so we rated this low risk. Burkin 1967, Guneri 2012 and Ricci 1987 gave no information on attrition and we judged them unclear risk on this basis.

Selective reporting

No pre‐published protocols were available for inspection to facilitate assessment for reporting bias. Conraux 1988 reported raw data and measures of spread were missing for important variables that were apparently collected, such as frequency of attacks; we thus rated it high risk. Guneri 2012 and Ricci 1987 did not report any information on adverse effects or tolerance and so we rated them high risk. Mira 2003 and Otto 2008 did not present absolute values or baseline data, only percentage changes in variables, so we rated them as high risk. Canty 1981 presented vertigo scores that are not fully described in the methods and we judged this high risk. Mira 2003 also recruited patients with benign paroxysmal positional vertigo but did not report Dix‐Hallpike tests as an outcome. In Duphar H10800580M 1984, there were no data on neuro‐otological findings although these were part of the diagnostic criteria and appear to have been assessed and so we judged this high risk. We rated the other studies unclear on the basis of the absence of a pre‐published protocol.

Other potential sources of bias

Baseline similarity of groups with respect to clinical disease parameters, such as vertigo duration or severity, was not clearly reported in three studies (Burkin 1967; Conraux 1988; Guneri 2012). Four studies showed some differences between active and placebo groups at baseline, not accounted for in the analysis techniques (Fischer 1985; Legent 1988; Otto 2008; Ricci 1987). Groups appear well matched in Okamoto 1968 and all the unpublished manufacturer trials (Duphar 77054 1983; Duphar H10800580M 1984; Duphar H10802786F/M 1989; Duphar H10803592F 1997; Duphar H108906NL 1990). Most studies did not exclude participants who had previously taken betahistine.

Effects of interventions

See: Summary of findings 1 Betahistine versus placebo for symptoms of vertigo

Betahistine versus placebo

Primary outcomes
Proportion of patients with reduction in vertigo symptoms (considering together the intensity, frequency and duration of those symptoms)

The proportion of patients who reported overall reduction in symptoms is given in Analysis 1.1. Twelve studies yielded analysable data. Although Mira 2003 collected data using the validated Dizziness Handicap Inventory (DHI), the results were reported only as percentage reductions with no baseline absolute values and missing measures of spread, so no useful data could be extracted.

The pooled risk ratio (RR) was 1.30 (95% confidence interval (CI) 1.05 to 1.60; 606 participants; 11 studies, I2 = 64%) in favour of betahistine. The heterogeneity was not resolved when we carried out subgroup analyses, first based on clinical diagnostic (participant) factors (Analysis 1.1), and then based on intervention factors (Analysis 1.2).

Firstly, in terms of clinical diagnostic (intervention) factors, we considered whether heterogeneity could be reduced by looking at the diagnostic groups of benign paroxysmal positional vertigo (BPPV), Ménière's disease (by investigator diagnosis) or 'other vertigo' (Analysis 1.1). These results need to be interpreted with caution as the diagnosis of Ménière's disease was investigator‐defined and did not necessarily meet standard criteria.

Subgroups still showed high statistical heterogeneity (Ménière's: I2 = 41%; other vertigo: I2 = 68%). The pooled risk ratio for the Ménière's subgroup was 1.56 (95% CI 0.92 to 2.65; 139 participants; three studies), for BPPV it was 1.34 (95% CI 0.85 to 2.10; 63 participants; one study) and for 'other vertigo' it was 1.24 (95% CI 0.97 to 1.58; 404 participants; eight studies).

Secondly, in terms of intervention factors, we looked at total daily betahistine dosage (Analysis 1.2). For doses of betahistine under 48 mg per day, the pooled risk ratio was 2.11 (95% CI 1.03 to 4.30; 292 participants; six studies) in favour of betahistine, with the caveat that statistical heterogeneity was again high (I2 = 89%). For higher doses (48 mg or more) the pooled RR was 1.16 (95% CI 0.92 to 1.48; 314 participants; five studies), also with high statistical heterogeneity (I2 = 54%). This analysis is illustrated in Figure 4.

Figure 4
Forest plot of comparison: 1 Betahistine versus placebo, outcome: 1.2 Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose.

Forest plot of comparison: 1 Betahistine versus placebo, outcome: 1.2 Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose.

We created a funnel plot for this analysis, as more than 10 studies were included (Figure 5). We noted asymmetry, raising the possibility of publication bias. However, sensitivity analysis by removing two studies with low numbers of participants and large effect sizes, Burkin 1967 and Otto 2008, showed no difference in the overall effect size but reduced the statistical heterogeneity slightly (RR 1.23, 95% CI 1.04 to 1.45; 562 participants; 10 studies; I2 = 49%) and the funnel plot was then symmetrical.

Figure 5
Funnel plot of comparison: 1 Betahistine versus placebo, outcome: 1.1 Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis.

Funnel plot of comparison: 1 Betahistine versus placebo, outcome: 1.1 Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis.

We rated the quality of the evidence for this outcome as low (see summary of findings Table 1).

Proportion of patients with adverse effects

All trials except Ricci 1987 made reference to adverse effects. Among the other studies, there was marked variation in findings.

The data for number of patients with adverse effects is shown in Analysis 1.3. The proportion of patients with adverse effects was 70/418 (16%) in the betahistine group and 61/401 in the placebo group (15%). Pooling the results give a risk ratio of 1.03 (95% CI 0.76 to 1.40; 819 participants; 12 studies).

Betahistine is frequently thought to cause upper gastrointestinal adverse effects. Six studies reported adverse effects in sufficient detail to analyse how many individuals reported upper gastrointestinal effects (Conraux 1988; Duphar 77054 1983; Duphar H10803592F 1997; Duphar H108906NL 1990; Mira 2003; Otto 2008). Studies that reported that no patients in either group had unwanted effects were not included in this analysis. Pooling the results for upper gastrointestinal effects gives a risk ratio of 1.38 (95% CI 0.67 to 2.82; 587 participants; six studies) (Analysis 1.4).

The next most common adverse effect reported in these studies was headache. Of those studies where headache was recorded as an unwanted effect in either group, the pooled risk ratio was 0.88 (95% CI 0.15 to 5.19; 515 participants; four studies) (Analysis 1.5).

Six studies reported no adverse events in either the placebo or betahistine groups (Burkin 1967; Canty 1981; Duphar H10800580M 1984; Fischer 1985; Okamoto 1968; Salami 1984). Legent 1988 reported non‐vestibular complaint effects in 91% of both the placebo and the betahistine groups, but it was unclear how many patients were in each group. Oosterveld 1989 reported adverse effects in detail but this was a cross‐over study and the figures provided pooled both cross‐over periods so we did not include the data in this analysis. In Duphar H10802786F/M 1989, adverse events are reported as "tolerance" and given as "poor" in 0/21 in the betahistine group and 3/26 in the placebo group, so we did not include this study in the analysis for this outcome.

Adverse effects reported in the betahistine group included upper gastrointestinal symptoms, rash, weight gain, nausea, headache, dry mouth, diuresis, rash, fatigue, tinnitus and hyperacusis. One patient in Mira 2003 was recorded as having dysmyelopoiesis on betahistine. One patient in Duphar H108906NL 1990 had respiratory distress on betahistine and one patient in Duphar H10803592F 1997 had an asthma attack on betahistine. No placebo patients reported these effects.

We rated the quality of the evidence for this outcome as low (see summary of findings Table 1).

Secondary outcomes
Proportion of participants withdrawing (dropping out) from the study due to all causes

The proportion of participants who withdrew or were lost to follow‐up is recorded in Analysis 1.6. We pooled the data, giving a risk ratio of 0.96 (95% CI 0.65 to 1.42; 481 participants; eight studies; I2 = 0%).

There was no significant difference between the betahistine and placebo groups in any study or in the pooled analysis.

Generic quality of life

There were no studies that reported using a generic health‐related quality of life instrument (e.g. SF‐36). The Dizziness Handicap Inventory (DHI) is a mixture of quality of life and symptom severity scores, which is considered under symptom‐specific measures.

Vestibular function as tested with objective vestibular function tests

Canty 1981, Salami 1984 and Otto 2008 included some objective measure of vestibular function.

Canty 1981 reported assessing caloric tests at baseline and after treatment, stating that the test was abnormal in nine patients before treatment with "some improvement" in two of these. As these effects were observed in a cross‐over design study and details of the timing of drugs and tests were not given, we undertook no further interpretation for this review.

Craniocorpography results and evaluation for nystagmus are reported in Otto 2008. The paper reported a statistically significant difference in change in sway on Romberg and Unterberger tests, with a greater reduction in the betahistine group than in the placebo group.

Salami 1984 reported that in the betahistine group, in 13 patients with abnormal tests at baseline 10 patients (77%) had normalisation after six weeks of treatment. In the placebo group, 14 had abnormal tests at the beginning and this was unchanged at six weeks (0%). We calculate that this gives a P value of 0.006 (Fisher's exact test).

There was one study that included patients with benign paroxysmal positional vertigo and it did not report Dix‐Hallpike tests as an outcome measure (Mira 2003).

Pooling data was inappropriate for this outcome given the diversity of techniques used to measure vestibular function.

Proportion of participants with falls

No studies reported falls as an outcome measure.

Betahistine plus particle repositioning versus placebo plus particle repositioning

Primary outcomes
Proportion of patients with reduction in vertigo symptoms (considering together the intensity, frequency and duration of those symptoms)

Guneri 2012 reported that there was no statistically significant difference ("P value > 0.05") in the mean scores between groups at one week. For the DHI, the authors reported that the placebo group had a mean score 12.15 (95% CI 10.5 to 13.4) at one week, and the betahistine group had a mean score of 10.42 (95% CI 4.7 to 16.1). On the Vertigo Symptom Scale (VSS) they reported that the placebo group had mean score of 2.88 (95% CI 2.83 to 2.93) at one week, and the betahistine group had a mean score of 2.17 (95% CI 2.13 to 2.21).

Secondary outcomes
Proportion of patients with adverse effects

Adverse effects were not reported.

Proportion of participants withdrawing (dropping out) from the study due to all causes

Withdrawals from the study were not reported.

Generic quality of life

Generic quality of life was not reported (although we note that the DHI has some symptom‐specific quality of life aspects and this scale is considered above).

Vestibular function as tested with objective vestibular function tests

Guneri 2012 found persistently positive Dix‐Hallpike tests (indicating no improvement) at one week in 4/26 (16%) with placebo and 3/24 (13%) with betahistine.

Proportion of participants with falls

Falls were not reported.

Discussion

Summary of main results

The objective of this review was to evaluate the overall efficacy of betahistine for symptoms of vertigo.

The primary outcome of this review was the proportion of patients with overall clinical improvement. There was a pooled risk ratio for overall improvement of 1.30 (95% confidence interval (CI) 1.05 to 1.60) in favour of betahistine. The pooled data should be interpreted with caution as the tests of statistical heterogeneity gave high results. The evidence for the outcome "proportion of patients with reduction in vertigo symptoms" is of low quality using the GRADE assessment, especially with respect to blinding and randomisation, which are of huge importance when studying vertigo as an outcome measure.

The 17 studies in this review had 1025 participants. Of the 17 included studies, five were unpublished studies funded and led by the manufacturers of betahistine.

The studies took place over a maximum of three months and so the longer‐term effects of betahistine are unknown.

Betahistine was associated with adverse events in 16% of participants; this was very similar to the rate in the placebo group (15%). The rate of upper gastrointestinal symptoms and headache was similar in the betahistine and placebo groups. There were two reports of asthma/respiratory distress in the betahistine group and none in the placebo group. There was one report of dysmyelopoiesis in the betahistine group and none in the placebo group. The high rate of unwanted symptoms in the placebo group was notable. This suggests that patients with vertigo may frequently experience other symptoms as part of their condition. However, the GRADE assessment for this outcome was also 'low', suggesting that the result should be interpreted with caution.

The review did not identify any subgroups that might particularly benefit from betahistine. This might be expected, given the fact that the studies were heterogeneous in terms of both participant diagnoses and also the diagnostic criteria used to identify subgroupings. Also, the overall effect size is at best a small one, meaning that numbers in any subgroup analysis are quite possibly too low to detect any effect. There is one pre‐existing Cochrane review on betahistine for Ménière's disease or syndrome, which found no clear evidence of benefit in that group, and our findings are in keeping with this (James 2001).

The subgroup analysis did not indicate a dose response relationship of betahistine. When examining the effect of dose, there was evidence of a small effect in the studies using lower doses but not in the studies using higher doses. There could be a number of possible explanations for this observation. Firstly, it is possible that the effect in the low‐dose group is a false positive finding. The GRADE quality of evidence for this outcome is low and the positive effect may be the result of study bias. Secondly, it is possible that the difference between the two subgroups is accounted for by other methodological differences between the studies, such as participant diagnoses. Thirdly, it is possible that the numbers of participants in the higher‐dose group were inadequate to detect a small effect. It is important to remember that the absence of an effect in the higher‐dose group in this subgroup analysis does not necessarily indicate that there genuinely is no effect. None of the included studies compared different doses of betahistine within the same protocol.

Overall completeness and applicability of evidence

The studies included in this review were conducted in clinical populations that appear to be similar to those who might receive betahistine in clinical practice, in that the participants all had vertigo with broad diagnostic inclusion criteria. However, they were all conducted using secondary or tertiary care level populations where there are more resources to make diagnoses. This means that there may be limited applicability to primary care settings where these resources are absent.

We searched a large number of databases and trial registries so we are confident that we traced all relevant trials. There is a concern, however, about potential reporting biases. We have overcome this as best we can by successfully obtaining unpublished evidence from manufacturers, sought by writing to manufacturers and from cited references in review papers. Of the 17 studies included, five were unpublished trials, but the manufacturers could provide us with data. We also found two registered clinical trials that had been terminated early due to poor recruitment (NCT00160238; NCT00474409). We contacted the drug companies sponsoring these trials, but no information had yet been provided to us at the time of publication of this review.

Quality of the evidence

Although we found a relatively large number of trials for this review (17 trials with 1025 participants), the overall quality of the evidence was low, meaning that further research is likely to have an important impact on the interpretation of these results (summary of findings Table 1).

There were significant methodological limitations in the conduct and reporting in these studies, particularly in terms of lack of clarity about patient recruitment/diagnostic criteria, choice of outcomes used (and reported) and very small sample sizes. None of the included studies used validated questionnaire data that could be analysed in this review.

None of the studies was of the highest methodological quality, with all studies except two rated as 'high risk' on at least one item of the 'Risk of bias' assessment. Statistical and clinical heterogeneity were high and few studies used validated outcomes, which are of critical importance for a subjective symptom such as vertigo. In addition, we are unsure about the quality of blinding of participants in the majority of studies. Although the studies were reported as "double blind", few details were supplied on how this was achieved. Since vertigo is a subjective outcome and is subject to psychological influences, adequate blinding is crucial in the execution of studies assessing the effects of interventions. However, with an intervention that is in tablet form, blinding should be straightforward to achieve.

None of the included studies had a pre‐published protocol available for inspection. However, we note two ongoing or recently completed studies for which such a protocol is available (BEMED; BETAVEST).

Potential biases in the review process

Our searches of the electronic databases were comprehensive. Language was not a barrier to inclusion and we included papers in French, Italian, Japanese and Dutch. Author roles were pre‐defined in the review process. Two authors selected studies for inclusion, extracted data and judged risk of bias independently, with recourse to the third author for resolution of disagreement or uncertainty. Two authors independently extracted data to minimise personal bias, and we considered both clinical and statistical heterogeneity before carrying out our analysis.

Agreements and disagreements with other studies or reviews

There are at least three other reviews of betahistine in the treatment of vertigo (Della Pepa 2006; Nauta 2014; Ramos 2015). Della Pepa 2006 and Nauta 2014 both found favourable effects of betahistine, as we did.

Nauta 2014 is a review and meta‐analysis focused on the outcome "investigator global assessment of benefit". Nauta found a beneficial effect of betahistine over placebo for both Ménière's disease and vestibular vertigo, calculating a pooled odds ratio (OR) of 2.58 (95% CI 1.67 to 3.99), with sub‐analyses conducted for patients with Ménière's disease (OR 3.37, 95% CI 2.14 to 5.29) and for vestibular vertigo (OR 2.23, 95% CI 1.20 to 4.14). Nauta did not consider risk of bias in underlying studies, nor other outcomes than investigator global opinion. For our review we chose to use the patient's perspective for improvement rather than the investigator's perspective.

Della Pepa 2006 is a review and meta‐analysis of randomised controlled trials of betahistine against placebo for symptoms of vertigo carried out between 1979 and 2003. This analysis excluded studies of patients with Ménière's disease. Of the seven studies included (Canty 1981; Fischer 1985; Legent 1988; Mira 2003; Oosterveld 1989; Oosterveld 1984; Singarelli 1979), some are excluded from our review for methodological reasons (Oosterveld 1984; Singarelli 1979). The authors calculated an odds ratio in favour of betahistine of 3.52 (95% CI 2.40 to 3.51).

Ramos 2015 performed a narrative review without meta‐analysis and concluded that betahistine is safe and effective.

Since we completed the search and analyses another relevant study has been published, which examines the effect of betahistine on vertigo in patients with Ménière's disease (Adrion 2016). This trial is noted in the ongoing studies section above (BEMED). The authors conclusions are that the incidence of attacks related to Ménière's disease did not differ between the three treatment groups and that treatment was well tolerated with no unexpected safety findings. This study will be included when this review is updated.

Process for sifting search results and selecting studies for inclusion.

Figures and Tables -
Figure 1

Process for sifting search results and selecting studies for inclusion.

Navigate to figure in ReviewOpen in new tab

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figures and Tables -
Figure 2

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Navigate to figure in ReviewOpen in new tab

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

Figures and Tables -
Figure 3

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

Navigate to figure in ReviewOpen in new tab

Forest plot of comparison: 1 Betahistine versus placebo, outcome: 1.2 Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose.

Figures and Tables -
Figure 4

Forest plot of comparison: 1 Betahistine versus placebo, outcome: 1.2 Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose.

Navigate to figure in ReviewOpen in new tab

Funnel plot of comparison: 1 Betahistine versus placebo, outcome: 1.1 Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis.

Figures and Tables -
Figure 5

Funnel plot of comparison: 1 Betahistine versus placebo, outcome: 1.1 Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis.

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 1: Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis

Figures and Tables -
Analysis 1.1

Comparison 1: Betahistine versus placebo, Outcome 1: Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 2: Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose

Figures and Tables -
Analysis 1.2

Comparison 1: Betahistine versus placebo, Outcome 2: Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 3: Proportion of patients with adverse effects

Figures and Tables -
Analysis 1.3

Comparison 1: Betahistine versus placebo, Outcome 3: Proportion of patients with adverse effects

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 4: Proportion of patients with upper gastrointestinal adverse effects

Figures and Tables -
Analysis 1.4

Comparison 1: Betahistine versus placebo, Outcome 4: Proportion of patients with upper gastrointestinal adverse effects

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 5: Proportion of patients with headache

Figures and Tables -
Analysis 1.5

Comparison 1: Betahistine versus placebo, Outcome 5: Proportion of patients with headache

Navigate to figure in ReviewOpen in new tab

Comparison 1: Betahistine versus placebo, Outcome 6: Withdrawal from study

Figures and Tables -
Analysis 1.6

Comparison 1: Betahistine versus placebo, Outcome 6: Withdrawal from study

Navigate to figure in ReviewOpen in new tab
Summary of findings 1. Betahistine versus placebo for symptoms of vertigo

Betahistine versus placebo for symptoms of vertigo

Patient or population: patients with symptoms of vertigo
Setting: outpatient clinics
Intervention: betahistine
Comparison: placebo

Time frame: up to 3 months

Outcomes

Relative effect
(95% CI)

Anticipated absolute effects* (95% CI)

Quality of the evidence
(GRADE)

What happens

Without betahistine

With betahistine

Difference

Benefits

Proportion of patients with improvement according to global judgement of patient subgrouped by diagnosis
№ of participants: 606
(11 RCTs)

RR 1.30
(1.05 to 1.60)

Moderate

⊕⊕⊝⊝
LOW 1 2 3

If 100 patients with vertigo are treated with betahistine, 60 will improve. This is 14 more than would have improved if a sham medicine had been taken instead of betahistine.

46.2%

60.1%
(48.5 to 73.9)

13.9% more
(2.3 more to 27.7 more)

Harms

Proportion of patients with adverse effects
№ of participants: 819
(12 RCTs)

RR 1.03
(0.76 to 1.40)

Moderate

⊕⊕⊝⊝
LOW 2 4

If 100 patients with vertigo are treated with betahistine, 16 will experience adverse effects. This is 1 more than would have had similar symptoms if a sham medicine had been taken instead of betahistine.

15.2%

15.7%
(11.6 to 21.3)

0.5% more
(3.6 fewer to 6.1 more)

*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

CI: confidence interval; RCT: randomised controlled trial; RR: risk ratio

GRADE Working Group grades of evidence
High quality: We are very confident that the true effect lies close to that of the estimate of the effect
Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different
Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect
Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

1Although statistical heterogeneity (I2 statistic) was 64%, the direction of effects was consistent.

2Most evidence was from studies with serious methodological limitations (unclear sequence generation, allocation concealment and blinding).

3Non‐validated outcome measures were used to measure improvement.

4Confidence intervals were wide and crossed thresholds of important benefits and harms.

Figures and Tables -
Summary of findings 1. Betahistine versus placebo for symptoms of vertigo
Navigate to table in Review
Comparison 1. Betahistine versus placebo

Outcome or subgroup title

No. of studies

No. of participants

Statistical method

Effect size

1.1 Proportion of patients with improvement according to global judgement of patient: subgrouped by diagnosis Show forest plot

11

606

Risk Ratio (M‐H, Random, 95% CI)

1.30 [1.05, 1.60]

1.1.1 Ménière's (investigator‐defined)

3

139

Risk Ratio (M‐H, Random, 95% CI)

1.56 [0.92, 2.65]

1.1.2 BPPV

1

63

Risk Ratio (M‐H, Random, 95% CI)

1.34 [0.85, 2.10]

1.1.3 Other vertigo

8

404

Risk Ratio (M‐H, Random, 95% CI)

1.24 [0.97, 1.58]

1.2 Proportion of patients with improvement according to global judgement of patient: subgrouped by drug dose Show forest plot

11

606

Risk Ratio (M‐H, Random, 95% CI)

1.45 [1.09, 1.94]

1.2.1 Betahistine dose less than 48 mg per day

6

292

Risk Ratio (M‐H, Random, 95% CI)

2.11 [1.03, 4.30]

1.2.2 Betahistine dose 48 mg or over per day

5

314

Risk Ratio (M‐H, Random, 95% CI)

1.16 [0.92, 1.48]

1.3 Proportion of patients with adverse effects Show forest plot

12

819

Risk Ratio (M‐H, Random, 95% CI)

1.03 [0.76, 1.40]

1.4 Proportion of patients with upper gastrointestinal adverse effects Show forest plot

6

587

Risk Ratio (M‐H, Random, 95% CI)

1.38 [0.67, 2.82]

1.5 Proportion of patients with headache Show forest plot

4

515

Risk Ratio (M‐H, Random, 95% CI)

0.88 [0.15, 5.19]

1.6 Withdrawal from study Show forest plot

8

481

Risk Ratio (M‐H, Random, 95% CI)

0.96 [0.65, 1.42]
Figures and Tables -
Comparison 1. Betahistine versus placebo
Navigate to table in Review