The aim of the study was to compare the effect of the EM and a sham manoeuvre on self-perceived disability assessed using the DHI-S at 1 week, 1 month, and 1 year in primary care patients with pc-BPPV.
Randomised, double-blind, sham-controlled clinical trial with an allocation ratio of 1:1 conducted in two primary care practices.
Changes to trial design
Although vestibular migraine was not an exclusion criterion for the trial , emerging evidence on the high prevalence of this condition  and its overlapping symptoms with pc-BPPV suggested that patients with vestibular migraine might have been recruited for the trial. It was therefore decided to re-assess all patients after completion of the follow-up phase and to remove those who met the 2013 criteria for probable vestibular migraine . Results following the original trial design, including patients with probable vestibular migraine, are presented as supplementary material
Patients were recruited between November 2012 and January 2015, and three follow-up visits, held a week, a month, and a year after the intervention, were performed by six general practitioners (GPs) blinded to treatment allocation working at two primary care practices, which employ 26 GPs and offer care for 38 305 people in L’Hospitalet de Llobregat (Barcelona, Spain).
Eligible participants included all adults (≥18 years) with symptoms of vertigo seen at the primary care practices, and patients clinically suspected to have pc-BPPV were systematically recruited by the GPs involved in the study. Patients who agreed to participate were referred for baseline evaluation by one of six GPs on the research team. Those who provided written informed consent and had a positive DHT, with or without nystagmus (O-BPPV or S-BPPV, respectively), were included. Patients with purely vertical nystagmus, nystagmus lasting >1 minute, or vertical or alternating nystagmus were excluded and referred to an ear, nose, and throat (ENT) specialist. This was done because it was suspected that their nystagmus was caused by a condition other than pc-BPPV (possibly even central nystagmus). They would therefore have required further investigation (e.g., magnetic resonance imaging, computed tomography, referral to an ENT specialist) and this would have affected our results. The full list of exclusion criteria is available in the study protocol .
Interventions and comparisons
At the baseline visit, the patients underwent a full physical examination and a complete medical history, including review of electronic medical records.
Patients in the intervention group were treated with a single EM. They were prescribed betahistine 8 mg/8h and instructed to use the medication as required (maximum 3 times a day) until their symptoms improved.
Patients in the control group were prescribed betahistine as the same dosage. Instead of the EM, however, they received a sham manoeuvre consisting of laying the patient with his/her head turned towards the affected side for 5 minutes .
The GPs who administered the EM took part in a 2-hour practical training session on diagnosing vertigo and applying the manoeuvre under the supervision of an ENT specialist to ensure consistent execution across patients. Two videos showing an investigator administering the DHT (https://www.youtube.com/watch?v=tJEFi5RFZEM) and the EM (https://www.youtube.com/watch?v=yAFx4-TFcGE) were also recorded.
Patient-perceived disability assessed using the total score on the DHI-S was used as the main outcome variable. The 10 items on the DHI-S are scored with 0 for “no”, 2 for “sometimes”, and 4 for “yes”. The total possible score therefore ranges from 0 (no self-perceived disability) to 40 (worst possible self-perceived disability).
The independent variables were age, sex, DHT result (O-BPPV or S-BPPV), history of anxiety and/or depression (Yes/No) and treatment with benzodiazepines, antidepressants, and/or vertigo drugs (Yes/No) at the time of the baseline visit.
The sample size was calculated for outcomes not analyzed in this study. Based on a sample size of 65 and 61 patients in the study arms at week 1, the study had a power of 80% to detect a statistically significant difference (alpha = 0.05) of 5.16 in DHI-S scores between groups (assuming a standard deviation of 10  in a Wilcoxon test [power estimated by bootstrap]).
Randomisation and allocation
Patients were assigned to the intervention and control groups using a list of randomisation sequences prepared in advance by the study statistician. The randomisation sequences were generated in ‘R: A language and environment for statistical computing’, version 2.14.2 (R Foundation for Statistical Computing, Vienna, Austria). Two members of staff from the participating primary care practices not involved in the clinical trial were responsible for safeguarding the randomisation list. On completion of the baseline visit activities, the recruiting GPs contacted one of these members of staff to find out the randomisation number and arm to which the patient was to be assigned.
The study was double blind. The GPs who performed the baseline visits did not participate in the follow-up visits, and vice-versa. GPs responsible for follow-up were therefore unaware of allocation. Randomisation data were not recorded in the data collection notebooks or database. Only the study statistician had access to this information.
Continuous variables (e.g., DHI-S scores) were expressed as median and interquartile range (IQR), while categorical variables (e.g., anxiety) were expressed as absolute and relative frequencies. For the between-group comparisons at each time point, the distribution of DHI-S answers were compared using the Chi-square test, while DHI-S scores were compared using the Wilcoxon test. These bivariate analyses were performed for the overall samples and stratified by sex and presence or absence of nystagmus at baseline.
A longitudinal multivariate regression model was built using DHI-S scores. Given the considerable proportion of 0 scores, a full mixed-effects multivariate Tobit regression model was adjusted to explain DHI-S scores by intervention, baseline DHI-S score, sex, baseline diagnosis (O-BPPV vs S-BPPV), and three-way interactions between all these variables and the treatment group, with adjustment for correlated intraindividual observations. Stepwise backward variable selection using the Akaike information criterion (AIC) was applied to obtain the final model. This reduced Tobit model contained all relevant variables and interactions; results are presented as marginal effects (medians of individual effects  ) and statistical significance (p-value) of the associated coefficients.