Types of studies

We included randomised controlled trials (RCTs) and quasi‐randomised trials (where trials were designed as RCTs, but the sequence generation for allocation of treatment used methods such as alternate allocation, birth dates etc). 

We planned to include cross‐over RCTs or cluster‐RCTs, providing we could appropriately account for the clustering in the data analysis (according to methods described in the Cochrane Handbook for Systematic Reviews of Interventions) (Handbook 2021, also see Unit of analysis issues).

Types of participants

We included studies that recruited participants with a diagnosis of either definite or probable vestibular migraine, according to the International Headache Society (IHS) and Bárány Society criteria (see Appendix 1). We also included studies that used other, established criteria, for example those of Neuhauser 2001 (see Appendix 2). 

If studies had recruited participants with a variety of diagnoses (e.g. vestibular migraine and headache migraine) we planned to include the study if either:

• the majority of participants (≥ 90%) had a diagnosis of vestibular migraine; or

• subgroup data were available that allowed us to identify data relevant specifically to those with vestibular migraine. 

However, we did not identify any studies that this applied to ‐ both included studies specifically recruited individuals with a diagnosis of vestibular migraine. 

Types of interventions

We included the following interventions:

• triptans;

• ergot alkaloids;

• dopamine antagonists;

• antihistamines;

• 5‐HT3 receptor antagonists;

• gepants (CGRP receptor antagonists);

• magnesium;

• paracetamol;

• non‐steroidal anti‐inflammatory drugs (NSAIDs).

Caffeine is frequently used in combination with analgesics. If we had identified studies where a combination of an analgesic and caffeine were used then we planned to include these as part of the interventions listed above, and to explore whether there may be additional effects from the caffeine using subgroup analysis. However, this was not necessary. 

The main comparisons were planned to be:

• triptans versus no intervention/placebo;

• ergot alkaloids versus no intervention/placebo;

• dopamine antagonists versus no intervention/placebo;

• antihistamines versus no intervention/placebo;

• gepants versus no intervention/placebo;

• magnesium versus no intervention/placebo;

• 5‐HT3 receptor antagonists versus no intervention/placebo;

• paracetamol versus no intervention/placebo;

• non‐steroidal anti‐inflammatory drugs (NSAIDs) versus no intervention/placebo.

Types of outcome measures

We assessed outcomes at the following time points:

• up to 2 hours;

• > 2 to 12 hours;

• > 12 to 72 hours.

An exception was adverse event data, where we used the longest time period of follow‐up. 

We searched the COMET database for existing core outcome sets of relevance to vestibular migraine and vertigo, but were unable to find any published core outcome sets. We therefore conducted a survey of individuals with experience of (or an interest in) balance disorders to help identify the outcomes that should be prioritised. This online survey was conducted with the support of the Ménière's Society and the Migraine Trust, and included 324 participants, who provided information regarding priority outcomes. The review author team used the results of this survey to inform the choice of outcome measures in this review. 

We analysed the following outcomes in the review, but we did not use them as a basis for including or excluding studies.

Electronic searches

The Information Specialist searched:

• the Cochrane ENT Trials Register (searched via the Cochrane Register of Studies to 23 September 2022);

• the Cochrane Central Register of Controlled Trials (CENTRAL) (searched via the Cochrane Register of Studies to 23 September 2022);

• Ovid MEDLINE(R) Epub Ahead of Print, In‐Process & Other Non‐Indexed Citations, Ovid MEDLINE(R) Daily and Ovid MEDLINE(R) (1946 to 23 September 2022);

• Ovid Embase (1974 to 23 September 2022);

• Web of Knowledge, Web of Science (1945 to 23 September 2022);

• ClinicalTrials.gov, www.clinicaltrials.gov (to 23 September 2022);

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP), https://trialsearch.who.int/ (to 23 September 2022).

The Information Specialist modelled subject strategies for databases on the search strategy designed for CENTRAL. The strategy has been designed to identify all relevant studies for a suite of reviews on various interventions for vestibular migraine. 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 Technical Supplement to Chapter 4 of the Cochrane Handbook for Systematic Reviews of Interventions version 6.1) (Lefebvre 2020). Search strategies for major databases including CENTRAL are provided in Appendix 3.

Searching other resources

We scanned the reference lists of identified publications for additional trials and contacted trial authors if necessary. In addition, the Information Specialist searched Ovid MEDLINE to retrieve existing systematic reviews relevant to this systematic review, so that we could scan their reference lists for additional trials. The Information Specialist also ran non‐systematic searches of Google Scholar to identify trials not published in mainstream journals.

We did not perform a separate search for adverse effects. We considered adverse effects described in included studies only.

Selection of studies

At least two review authors or coworkers (of SC, AD, KG, LHK, KW) independently screened the titles and abstracts using Covidence to identify studies that may be relevant for this review. Any discrepancies were resolved by consensus, or by retrieving the full text of the study for further assessment. 

We obtained the full text for any study that may have been relevant and two authors (of AD, KG, LHK, KW) again independently checked this to determine whether it met the inclusion criteria for the review. Any differences were resolved by discussion and consensus, or through recourse to a third author if necessary. 

We have listed as excluded any studies that were retrieved in full text but subsequently deemed to be inappropriate for the review (according to the inclusion/exclusion criteria), according to the main reason for exclusion. 

The unit of interest for the review was the study, therefore multiple papers or reports of a single study were planned to be grouped together under a single reference identification. We recorded the study selection process in sufficient detail to complete a PRISMA flow diagram (see Figure 1) and the Characteristics of excluded studies table. 

Figure 1

---

---

Flow chart of study retrieval and selection.

Screening eligible studies for trustworthiness

We assessed all studies meeting our inclusion criteria for trustworthiness using a screening tool developed by Cochrane Pregnancy and Childbirth. This tool includes specified criteria to identify studies that are considered sufficiently trustworthy to be included in the review (see Appendix 4). If any studies were assessed as being potentially 'high risk', we planned to contact the study authors to obtain further information or address any concerns. We planned to exclude 'high risk' studies from the main analyses of the review if we were unable to contact the authors, or there was persisting uncertainty about the study, and only include studies with concerns as part of a sensitivity analysis (see Sensitivity analysis). The process is outlined in Figure 2.

Figure 2

---

---

The Cochrane Pregnancy and Childbirth Trustworthiness Screening Tool

The studies included in this review were not considered to be at high risk when using the trustworthiness tool. We did not identify any additional studies that we excluded on this basis.

Data extraction and management

At least two review authors (of AD, LHK, KW) independently extracted outcome data from each study using a standardised data collection form. Any discrepancies in the data extracted by the two authors was checked against the original reports, and differences were resolved through discussion and consensus, with recourse to a third author where necessary. 

We included key characteristics of the studies, including the following information:

• study design, duration of the study, number of study centres and location, study setting and dates of the study;

• information on the participants, including the number randomised, those lost to follow‐up or withdrawn, the number analysed, the age of participants, gender, features of the condition (e.g. probable or definite vestibular migraine), diagnostic criteria used, inclusion and exclusion criteria for the individual studies;

• details of the intervention, comparator, and concomitant treatments or excluded medications;

• the outcomes specified and reported by the study authors, including the time points;

• funding for the study and any conflicts of interest for the study authors;

• information required to assess the risk of bias in the study, and to enable GRADE assessment of the evidence.

Once the extracted data were checked and any discrepancies were resolved, a single author (KW) transferred the information to Review Manager 5 (RevMan 2020). 

The primary effect of interest for this review was the effect of treatment assignment (which reflects the outcomes of treatment for people who were assigned to the intervention) rather than a per protocol analysis (the outcomes of treatment only for those who completed the full course of treatment as planned). For the outcomes of interest in this review, we extracted the findings from the studies on an available case basis, i.e. all available data from all participants at each time point, based on the treatment to which they were randomised. This was irrespective of adherence, or whether participants had received the intervention as planned.

In addition to extracting pre‐specified information about study characteristics and aspects of methodology relevant to risk of bias, we extracted the following summary statistics for each study and outcome:

• For binary data: we extracted information on the number of participants experiencing an event, and the number of participants assessed at that time point.

• For ordinal scale data: some data were collected by the study authors with an ordinal scale (a symptom rating scale of 0 to 3). However, the authors of both studies presented the results as binary data (improved/not improved), therefore we extracted the results as above.

• We did not identify any continuous data or time‐to‐event data for this review. 

If necessary, we converted data found in the studies to a format appropriate for meta‐analysis, according to the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2021). 

We pre‐specified time points of interest for the outcomes in this review. Where studies reported data at multiple time points, we took the longest available follow‐up point within each of the specific time frames. For example, if a study reported an outcome at three hours and five hours of follow‐up then the five‐hour data would have been included for the time point more than two to six hours.

Assessment of risk of bias in included studies

Two authors (of AD, LHK, KW) undertook assessment of the risk of bias of the included studies 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 (Handbook 2011), which involves describing each of these domains as reported in the study and then assigning a judgement about the adequacy of each entry: 'low', 'high' or 'unclear' risk of bias.

Measures of treatment effect

We summarised the effects of binary outcomes (e.g. serious adverse effects) as risk ratios (RR) with 95% confidence intervals (CIs). For the key outcomes that we present in the summary of findings tables, we have also expressed the results as absolute numbers based on the pooled results, and compared to the assumed risk. 

Unit of analysis issues

The studies included in this review both had unit of analysis issues. The first study involved participants taking medication up to three times (for three separate attacks of vestibular migraine) over the course of the study (NCT02447991). The response to treatment was recorded for each attack. This led to two concerns with the data. Firstly, the data from different attacks in the same individual are not independent ‐ an individual who responds to treatment for one attack may be more likely to respond again. Secondly, individuals experienced different numbers of attacks over the course of the study. A participant who experienced just one attack would only contribute one data point to the results, whilst an individual who experienced three attacks would contribute three (correlated) data points. This skews the overall result towards the outcome for people who suffered more frequent attacks, which may not be relevant for the whole population of people with vestibular migraine. 

The second study was a cross‐over trial (Neuhauser 2003). Data were reported as the response to treatment per attack, and this study included a total of 17 attacks experienced by 10 participants. Attacks experienced by the same individual cannot be regarded as independent. Ideally, results from this study would be analysed using a paired analysis (where each participant acts as their own control) according to the methods stated in the Handbook 2021. However, insufficient data were reported to allow this. We have therefore analysed the data as if they were independent, but acknowledge that this may lead to some inaccuracy in the results.

We took advice from the Cochrane Methods Support Unit regarding these issues. Given that the data available for this review were sparse, they considered it acceptable to include these data in the review, but highlight the limitations of the analysis, and take this into account when using GRADE to assess the certainty of the analysis. 

Dealing with missing data

We planned to contact study authors via email whenever the outcome of interest was not reported, if the methods of the study suggested that the outcome had been measured. We did the same if not all data required for meta‐analysis were reported (for example, standard deviations), unless we were able to calculate them from other data reported by the study authors. 

Assessment of heterogeneity

We assessed clinical heterogeneity by examining the included studies for potential differences between them in the types of participants recruited, interventions or controls used and the outcomes measured.

We used the I² statistic to quantify inconsistency among the studies in each analysis. We also considered the P value from the Chi² test. 

Assessment of reporting biases

We assessed reporting bias as within‐study outcome reporting bias and between‐study publication bias.

Data synthesis

Subgroup analysis and investigation of heterogeneity

If statistical heterogeneity was identified for any comparisons, we planned to assess this considering the following subgroups:

• Different types of intervention, within a specific group.

• Use of any concomitant treatment.

• Diagnosis of vestibular migraine.

• Age of the participants.

• Sex of the participants.

However, due to the paucity of data available, and the few meta‐analyses included in this review, we did not carry out any subgroup analysis. 

For the outcome 'improvement of other migrainous symptoms' we planned to pool different symptoms (such as nausea and vomiting, photophobia and phonophobia, visual aura) and assess this as a composite measure. However, as we only identified a single study that assessed these outcomes, we have presented them as separate measures. 

Sensitivity analysis

As few studies were identified for meta‐analysis, the random‐effects model may provide an inaccurate measure of the between‐studies variance. Therefore, we explored the impact of using a fixed‐effect model using a sensitivity analysis. 

If there was uncertainty over the diagnostic criteria used for participants in the studies (for example, if it was not clear whether participants were diagnosed using criteria analogous to the IHS‐Bárány Society criteria) then we planned to explore this by including/excluding those studies from the analysis. However, this was not necessary. 

We used the Cochrane Pregnancy and Childbirth Screening Tool to identify any studies where there were concerns over the data available. We planned to exclude any studies that we identified as high risk with this tool from the main analyses in the review, but we intended to explore the impact of including the data from these studies through a sensitivity analysis. Again, as we did not exclude any studies on the basis of this tool, no sensitivity analysis was required. 

Summary of findings and assessment of the certainty of the evidence

Two independent authors (KG, KW) used the GRADE approach to rate the overall certainty of evidence using GRADEpro GDT (https://gradepro.org/) and the guidance in Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions (Handbook 2021). Disagreements were resolved through discussion and consensus. The certainty 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 certainty 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 certainty 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 certainty. 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):

  • This was assessed using the rating from the Cochrane risk of bias tool for the study or studies included in the analysis. We rated down either one or two levels, depending on the number of domains that had been rated at high or unclear risk of bias. 

• Inconsistency:

  • This was assessed using the I² statistic and the P value for heterogeneity for all meta‐analyses, as well as by visual inspection of the forest plot. For results based on a single study we rated this domain as no serious inconsistency.

• Indirectness of evidence:

  • We took into account whether there were concerns over the population included in the study or studies for each outcome, as well as whether additional treatments were offered that may impact on the efficacy of the intervention under consideration. 

• Imprecision:

  • We took into account the sample size and the width of the confidence interval for each outcome. If the sample size did not meet the optimal information size (i.e. < 400 people for continuous outcomes or < 300 events for dichotomous outcomes), or the confidence interval crossed the small effect threshold we rated down one level. If the sample size did not meet the optimal information size and the confidence interval included both potential harm and potential benefit we rated down twice. We also rated down twice for very tiny studies (e.g. 10 to 15 participants in each arm), regardless of the estimated confidence interval.

• Publication bias:

  • We considered whether there were likely to be unpublished studies that may impact on our confidence in the results obtained. 

We used a minimally contextualised approach, and rated the certainty in the interventions having an important effect (Zeng 2021). Where possible, we used agreed minimally important differences (MIDs) for continuous outcomes as the threshold for an important difference. Where no MID was identified, we provide an assumed MID based on agreement between the authors. For dichotomous outcomes, we looked at the absolute effects when rating imprecision, but also took into consideration the GRADE default approach (rating down when a RR crosses 1.25 or 0.80). We have justified all decisions to downgrade the certainty of the evidence using footnotes, and added comments to aid the interpretation of the findings, where necessary. 

We prepared a separate summary of findings table for the main comparison:

• triptans versus no intervention/placebo.

We included all primary outcomes in the summary of findings table. We prioritised outcomes at the time point 'up to two hours' for presentation in the table. We have also included a full GRADE profile for all results (Table 1).