Epley manoeuvre’s efficacy for benign paroxysmal positional vertigo (BPPV) in primary-care and subspecialty settings: a systematic review and meta-analysis

We included RCTs that assessed the efficacy of EM. Cluster randomised and crossover trials were not included. We did not apply language or country restrictions. We included all articles, including published, unpublished, conference abstracts, and letters. We did not exclude studies based on the observation period or publication year. Participants should have presented with the symptoms of repeated episodes of vertigo, mostly on change of position along with nausea and vomiting. Age, sex, and race did not matter. Participants with a clinical diagnosis of BPPV using the DH test, which proved positional nystagmus reflecting involvement of the posterior canal [4], were included. For posterior canal BPPV, a positive DH test is defined by the presence of upbeating and torsional nystagmus with the top pole of rotation beating toward the affected (downside) ear [4]. The study also included participants with subjective BPPV, where nystagmus is not induced by the DH test; subjective BPPV is an important concept [15]. Participants were diagnosed with BPPV by physicians educated in using EM. EM was administered at the first visit. Repeated manoeuvres or the combination with other interventions that included drugs and rehabilitative exercise were not a concern. Patients who could not tolerate the procedure or had serious heart disease or cervical spine lesions were excluded. The intervention was defined as EM involving a series of four head and body movements from sitting to lying, rolling over, and back to sitting [7]. Control was defined as medication, untreated controls, and sham manoeuvre. A sham manoeuvre consists of laying the patient with the head tilted on the affected side for 5 min. The primary-care setting was defined as a practice setting by general physicians, primary-care physicians, and family physicians.

The primary outcomes were the disappearance of subjective symptoms (vertigo), negative findings (DH test), and all adverse events. In general, there should be no more than three primary outcomes, including at least one desirable and at least one undesirable outcome [16]. The secondary outcomes were the disappearance of objective symptoms (nystagmus) and Dizziness Handicap Inventory (DHI) score. All outcomes of interest are detailed in Additional file 2.

A systematic search was conducted in January 2022 across databases, including the Central, MEDLINE, Embase, and Cumulative Index of Nursing and Allied Health Literature (details provided in Additional file 3). We also searched the World Health Organization International Clinical Trials Platform Search Portal and ClinicalTrials.gov for ongoing or unpublished trials. We checked the reference lists of studies, including international guidelines [1], reference lists, and articles citing eligible studies. We asked the authors of the original studies for unpublished or Additional data.

Two independent reviewers (YS, HM) screened titles and abstracts and assessed eligibility based on the full texts. We contacted original authors if relevant data was missing. Disagreements between the two reviewers were resolved by discussion, and if this failed, a third reviewer acted as an arbiter (NY). Two reviewers (YS, HM) performed independent data extraction of the included studies using standardized data collection forms. We used a pre-checked form using 10 randomly selected studies. The form included the information on study design, study population, interventions, and outcomes. Any disagreements were resolved by discussion, and if this failed, a third reviewer acted as an arbiter (NY). Two reviewers (YS, HM) evaluated the risk of bias (ROB) independently using the Risk of Bias 2 [17]. Disagreements between the two reviewers were discussed, and if this failed, a third reviewer (ST) acted as an arbiter, if necessary.

We pooled the relative risk ratios (RRs) and 95% confidence intervals (CIs) for the following binary variables: disappearance of subjective symptoms (vertigo), negative findings (Dix–Hallpike test), all adverse events, and disappearance of objective symptoms (nystagmus). We pooled the mean differences and 95% CIs for the following continuous variable: Dizziness Handicap Inventory. We summarised adverse events based on the definition in the original article, but we did not perform a meta-analysis. We requested the original authors for the not-presented data.

We performed the intention-to-treat analysis for all dichotomous data. For continuous data, we did not impute missing data based on the recommendation by the Cochrane Handbook [18]. When original studies only reported standard error or a P-value, we calculated the standard deviation based on the method reported by Altman [19]. If these values were unknown when we contacted the authors, the standard deviation was calculated using confidence interval and t-value based on the method indicated in the Cochrane Handbook [18] or validated method [19]. The validity of these methods was analysed using sensitivity analysis.

We evaluated the statistical heterogeneity by visual inspection of the forest plots and calculating the I² statistic (I² values of 0–40%: might not be important; 30–60%: may represent moderate heterogeneity; 50–90%: may represent substantial heterogeneity; and 75–100%: considerable heterogeneity). When there was substantial heterogeneity (I² > 50%), we assessed the reason for the heterogeneity. The Cochrane chi-squared test (Q-test) was performed for I² statistic, and a P-value less than 0.10 was defined as statistically significant.

We searched the clinical trial registry system (ClinicalTrials.gov and International Clinical Trials Platform Search Portal) and performed an extensive literature search for unpublished trials. We assessed the potential publication bias by visual inspection of the funnel plot. The Egger test was also performed; we did not conduct the test when we found fewer than 10 trials or trials with similar sample size.

Meta-analysis was performed using Review Manager software (RevMan 5.4). We used a random-effects model. To elucidate the influence of effect modifiers on results, we evaluated the subgroup analyses of the primary outcomes based on age (≥ 65 years), vertigo severity (above average if using a scale), duration (< 30 days or longer), number of BPPV episodes (first or recurrent episode), number of EM sessions (only once vs. more than once), and EM skills in primary-care settings (whether or not they are educated practitioners) when sufficient data were available. The definition of an educated practitioner is one who has been educated in EM methods by an otolaryngologist or neurologist and observed in practice.

We performed sensitivity analysis for the primary outcomes to assess whether the results of the review were robust to the decisions made during the review process by excluding studies using imputed statistics or excluding studies with high or some concern in the overall assessment of the ROB. We created a summary-of-findings table that included an overall grading of the certainty of the evidence for each primary and secondary outcome, evaluated using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach [20].

Forest plots for each outcome are described in Additional file 7. The summary of findings provides the certainty of the evidence for the outcome in each setting and is listed in Tables 1 and 2.

The evidence suggested that EM reduced subjective symptoms (3 studies, 309 participants): RR 3.14; 95% CI 1.96–5.02, I² = 84%; low certainty evidence. The evidence was uncertain about the effect of EM on the negative findings using the DH test (2 studies, 206 participants): RR 1.46; 95% CI 0.72–2.97, I² = 63%; very low certainty evidence.

The disappearance of objective symptoms and DHI-S were measured in one RCT [9]. The evidence suggested that EM reduced objective symptoms slightly (1 study, 127 participants): RR 0.84; 95% CI 0.73–0.97; low certainty evidence. The evidence suggested that EM resulted in little to no difference in DHI-S (1 study, 134 participants): mean difference − 2; 95% CI -5.51 to 1.51; low certainty evidence.

The evidence suggested that EM reduced subjective symptoms (16 studies, 829 participants): RR 2.42; 95% CI 1.64–3.56, I² = 84%; low certainty evidence. The evidence suggested that EM resulted in an increase in negative findings using the DH test (16 studies, 912 participants): RR 1.81; 95% CI 1.40–2.34, I² = 79%; low certainty evidence. The evidence was uncertain about the effect of EM on all adverse events: one study [42] (50 participants) reported all adverse events.

The disappearance of objective symptoms was measured in one RCT [23], and DHI-S was measured in two RCTs [33, 40]. The evidence suggested that EM reduced objective symptoms slightly (1 study, 58 participants): RR 1.69; 95% CI 1.08–2.66; low certainty evidence. The evidence was uncertain about the effect of EM on DHI-S (2 studies, 70 participants): mean difference − 8.24; 95% CI -28 to 11.51; very low certainty evidence.

Risk of Bias 2 was used for the evaluation of the ROB. Most studies were at high or some concern ROB, as per the Cochrane ROB assessment tool (details provided in Additional file 8). In the primary-care and otolaryngology or subspecialty settings, the disappearance of subjective symptoms was a subjective assessment and resulted in a high ROB. Two studies of negative findings in the primary-care setting [8, 10] had some concern ROB because one study demonstrated a low ROB except for the randomisation process, and the other demonstrated a low ROB except for deviations from the intended intervention. There was no study about all adverse events in primary-care settings.

Two studies of negative findings in the otolaryngology or subspecialty settings [24, 36] had a low ROB, and six studies [8, 10, 35, 37, 41, 42] had some concern ROB. The other studies demonstrated a high ROB. One study [42] of all adverse events in the otolaryngology or subspecialty settings had a high ROB because the outcome was measured. There was evidence of publication bias using Egger’s test (P < 0.001) in a reduction of subjective symptoms and an increase in negative findings using the DH test in the otolaryngology or subspecialty settings.

The prespecified subgroup analyses for the primary outcomes revealed no significant differences among subgroups (details provided in Additional file 9). We were unable to perform subgroup analyses for items ‘vertigo severity’, ‘number of BPPV episodes (first or recurrent episode)’, and number of EM sessions (only once vs more than once)’ because there were no applicable studies.

The prespecified sensitivity analysis for the primary outcomes was carried out because there was no study that used imputed statistics, but two studies [24, 36] were a low ROB in the outcome of negative findings using the DH test in the otolaryngology or subspecialty settings. Similar results were obtained in the sensitivity analysis (details provided in Additional file 10).