Introduction: The head impulse test (HIT) and HIT combined with direction-changing Nystagmus-Test of Skew deviation (HINTS) have been proposed as bedside tests to differentiate between peripheral and central causes of vertigo in the emergency department (ED). We conducted a meta-analysis of the HIT and HINTS tests to diagnose peripheral vertigo (PV) and central vertigo. Methods: Pubmed, Google Scholar, EmBase, and articles references published in English up to July 2021 were searched for keywords “vertigo” or “acute vestibular syndrome” or “dizziness” and “head impulse” and “stroke.” The bivariate method for meta-analysis was used to calculate positive (PLR) and negative likelihood ratios (NLR) and summary receiver operating characteristics area under the curve (AUC). Results: A total of 11 studies were included analysing both HIT (8 studies, N = 417) and HINTS (6 studies, N = 405). HIT and HINTS were performed within 24 h in 4 of 11 studies. PLR and NLR for HIT in PV was 4.85 (95% CI: 2.83–8.08) and 0.19 (95% CI: 0.12–0.29, I2 63.25%), respectively. The AUC for HIT the diagnosis of PV and stroke was 0.90 and 0.92, respectively. PLR and NLR for a negative HIT in stroke was 5.85 (95% CI: 3.07–10.6) and 0.17 (95% CI: 0.08–0.30), respectively. PLR and NLR for peripheral HINTS pattern for PV was 17.3 (95% CI: 8.38–32.1) and 0.15 (95% CI: 0.07–0.26), respectively. PLR and NLR for central HINTS pattern for stroke: 5.61 (95% CI: 4.19–7.7) and 0.06 (95% CI: 0.03–0.12). In all included studies, HIT and HINTS exams were administered by neurology residents or neurology specialists with additional neuro-otology or neuro-ophthalmology subspeciality experience, and two studies included ED physicians. Raters reported high degree of bias and high concern regarding applicability in most domains of the quality assessment of diagnostic accuracy studies (QUADAS-2) tool. Meta-regression did not demonstrate a statistically significant effect of publication year, time to test, and type of assessor on sensitivity or false positive rate. Conclusion: The HIT and HINTS exams appear to be moderately good discriminators of central and PV. However, in most papers, the tests were administered by neurologists and were evaluated beyond 24 h, which may limit utility in the ED setting.

© 2022 The Author(s). Published by S. Karger AG, Basel

Introduction

Acute vestibular syndromes (AVS) are a common presentation in the emergency department (ED) and pose a significant diagnostic and management problem. The head impulse test (HIT), based on an observation in a single individual with bilateral vestibular hypofunction in 1988 [1], has evolved over many years as a means of discriminating between peripheral and central vertigo. A subsequent modification of the HIT test is the three-step bedside head impulse test with direction changing nystagmus and test of skew deviation (HINTS) examination, which is reported to be sensitive in the diagnosis of stroke in AVS [2]. It remains unclear how these tests, designed by neuro-otologists, would perform in a real-world emergency setting, where it remains crucial to distinguish stroke from peripheral causes of vertigo in a time-critical manner. A recent meta-analysis in 2019 of 6 studies suggested that HINTS could be useful for identifying patients with ischaemic strokes for reperfusion therapy, but unusually presented data as risk ratio rather than the summary receiver operating characteristics curve and positive (PLR) and negative likelihood ratios (NLR) [3]. A second meta-analysis in 2020 evaluated 5 studies in which HINTS was performed by neurologists rather than emergency department clinicians [4]. These meta-analyses did not use the recommended bivariate method for meta-analysis of diagnostic studies [5, 6]. This method is important as it preserves the dual nature of the data (sensitivity and specificity) in the treatment of covariates in meta-regression [5, 6]. Furthermore, previous studies have not accounted for the time delay in performance of HIT and HINTS tests, when reference imaging was done with respect to the tests, nor the type of assessor (e.g., emergency physician vs. neurologist). We postulate that these factors are potential sources of heterogeneity but may also represent limitations of the current evidence from which diagnostic utility of HIT and HINTS in the ED setting has been derived. In light of this, we performed a meta-analysis using a bivariate method examining the diagnostic accuracy of the HIT and HINTS test in the diagnosis of peripheral and central vertigo, with a particular emphasis on its utility at the bedside in the ED setting.

MethodsStudy Selection and Inclusion Criteria

We conducted an article search on papers published in English on Pubmed, EmBase, Google Scholar, and conference abstracts with keywords “vertigo” or “acute vestibular syndrome” or “dizziness” and “head impulse” and “stroke” until July 2021. Three authors (S.O., G.P., and T.T.) independently screened articles. Prior meta-analyses were used to complement the search strategy. We included articles with adult participants greater than 18 years of age, the HIT and/or HINTS test, a neuroimaging reference standard of MRI or CT, and the time interval in which neuroimaging was performed with respect to the patient presenting to ED. We included papers that included stroke as a central nervous system cause for vertigo. For papers which used the outcome of central nervous system cause of vertigo (because the bedside test cannot discriminate the aetiology of central cause such as tumour or demyelination), we used patients with the final outcome of stroke according to a neuroimaging reference. We excluded papers that had already screened patients for HIT or HINTS testing due to the presence of spontaneous nystagmus because this finding renders the bedside tests challenging to interpret.

Both retrospective and prospective, cohort and cross-sectional trials were included. Only articles in which data could be extracted to produce or calculate a 2 × 2 table to verify the sensitivity and specificity were included and independently collected by three authors (S.O., G.P., and T.T.) for verification. We extracted the timing of the HIT or HINTS in relation to presentation and the assessors who were involved in administering the tests. We excluded articles that tested video HIT in isolation if they did not have clinical or bedside HIT data. We defined a positive HIT as one showing an abnormal vestibulo-ocular reflex due to the presence of a pathological catch-up saccade. We defined a peripheral HINTS pattern as implying a peripheral cause of vertigo (i.e., positive HIT, absence of central pattern of nystagmus, and negative test of skew), and a central HINTS as implying a central cause to vertigo, such as stroke (i.e., at least one of negative HIT, central pattern of nystagmus or presence of skew deviation). The meta-analyses followed the preferred reporting items for systematic reviews and meta-analyses guidelines (PRISMA) [7] (see online suppl. material; see www.karger.com/doi/10.1159/000526331 for all online suppl. material for PRISMA checklist).

Quality Appraisal of Individual Evidence

Three authors (L.C., A.F., and J.L.) independently evaluated domains as recommended in the QUADAS-2 to evaluate the quality of evidence for the included studies [8]. Studies were assessed according to low, high, or unclear risk for bias with respect to flow and timing, reference standard, index test, and patient selection and applicability. L.C. was not involved in the assessment of his paper [9], which was included in the meta-analysis.

Statistical Analysis

A bivariate random-effects method from mada in the R programming language (version 4.0.3) for meta-analysis was used to compute area under the summary ROC (AUC), PLR, and NLR [10]. The advantage of the bivariate over the univariate method is that the former employs a random effect to take into account the within study correlation as well as the between study correlation in sensitivity and specificity [5]. One point to note is that mada library provides AUC without a confidence interval [10] and does not provide summary lines with forest plots.

An AUC of 0.5 denotes discrimination no better than by chance, 0.8–0.89 provides good discrimination, and 0.9–1.0 provides excellent discrimination [11]. The PLR and NLR are derived from sensitivity and specificity. A PLR indicates the likelihood that a positive HIT or HINTS would be expected in a patient with peripheral vertigo (PV) compared with the likelihood that the same result would be expected in a patient without PV, i.e., stroke. Based on the recommendations by Jaeschke et al. [12], a high PLR (>5) and low NLR (<0.2) indicate that the test results would make moderate changes in the likelihood of PV, whereas a PLR of >10 combined with NLR of <0.1 would confer large changes, compared to a negative test. However, in the presence of heterogeneity, this standard interpretation of AUC may not be applicable [5].

We measured the inconsistency I2 index and performed a meta-regression to understand the source of heterogeneity [13]. The fixed component were the covariates which were being tested for their effect on heterogeneity, such as time to perform bedside test, publication year, and speciality of assessor (neurologist or others). A random effects model was used with the random effect components being transformed sensitivity and false positive rate. The random effect approach results in more conservate estimate of effect size confidence interval.

Results

Our search yielded 81 titles and abstracts for screening. Articles that used duplicate datasets to other included studies were excluded [14]. A total of 11 studies were included (see online suppl. material. Fig. 1.0 for PRISMA diagram), of which 8 were included in HIT analysis (N = 417) and 6 were included in HINTS analysis (N = 405). The current analysis included an additional 6 articles compared to a recent meta-analysis [4]. There were two papers that were included in a previous meta-analyses but full text in English could not be obtained for one paper [15] and the 2 × 2 data could not be verified for the second paper [16]. Preliminary 2 × 2 data were obtainable on three studies [16-18], but due to ambiguity in published results, these were not included in the final analysis.

Characteristics of Included Studies

Methodology was diverse amongst the included studies with respect to many parameters (Table 1). Four studies were retrospective and 7 studies were prospective. Nine out of the 11 studies compared HIT or HINTS in the diagnosis of stroke as the central cause compared to peripheral causes, and 2 studies [19, 20] included stroke and nonstroke central causes. Inclusion and exclusion criteria were comparable across studies, in that most studies included acute onset isolated vertigo and dizziness with or without gait disturbance, but excluded patients with preexisting vestibular disorders, focal motor deficits, or cerebellar signs. In regards to location, HIT and HINTS bedside testing were performed in the ED and not explicitly stated in one study [21], with another study performing the test in ED, ward, and outpatient settings [20]. In all studies, the tests were administered by neurology residents or neurology specialists with additional neuro-otology or neuro-ophthalmology subspeciality experience. Two papers described emergency clinicians performing the test, but this was also in conjunction with a neurologist or neurology subspecialists [22]. The median time to bedside test was 48 h with an interquartile range of 156 h, although the median time to clinical testing was not documented in two studies [21, 23] and was up to 7 days and 14 days in two studies, respectively [20, 24]. The diagnostic reference in all studies was MRI brain, except for two studies which used both MRI-B and CT-brain [20, 25], of which CT-B was performed in 2.1% [20] and 50.8% [25] of participants, respectively. The timing of imaging as the reference test was provided in 6 of the 11 included studies and was performed within a median time of 72 h with an interquartile range of 24.5 h, with one study performing MRI brain up to 14 days after presentation [24]. The exclusion criteria were mostly consistent with excluding patients with recurrent episodic vertigo or those who had additional focal neurological symptoms like motor weakness.

Table 1.

Summary of study characteristics

Quality and Risk of Bias Assessment

The summary of the qualitative assessment for risk of bias and applicability is displayed in Tables 2and3. There was a lack of agreement with regard to the risk of bias assessment by the three reviewers. For example, one paper (n = 40, collected between 2013 and 2016) [22] was rated as high risk by rater 1, low risk by rater 2 and high risk by rater 3 with regards to patient selection. Another paper (n = 83, recruited between 1995 and 2005) [19] was rated to have low risk by 2 raters and high risk by rater 3 with regards to patient selection. For a study with 86 subjects (recruited between January and December 2014) [26], the index test was rated as low (raters 1 and 2) and high risk of bias by rater 3. Two further studies did not describe the time of clinical assessment but were assessed to be low risk of bias with respect to the index variable [21, 23]. With regards to assessment of the reference standard (i.e., imaging to confirm the diagnosis of stroke), all raters agreed that there was high risk of bias in the study in which either MRI or CT were used to diagnose stroke [27].

Table 2.

QUADAS-2 Summary of risk of bias

Table 3.

QUADAS-2 Summary of applicability concern

Accuracy of the HIT Exam in the Diagnosis of PV and Central Vertigo (e.g., Stroke)

From the paper which evaluated HIT in the diagnosis of nonstroke causes of central vertigo [19], we obtained PLR and NLR of HIT in the diagnosis of PV and excluded the paper in the diagnosis of stroke. Overall, the PLR and NLR of HIT for diagnosing PV was 4.85 (95% CI: 2.83–8.08) and 0.19 (95% CI: 0.12–0.29, I2 63.25%), respectively (Table 4; Fig. 1a, b). In evaluating a negative HIT for the diagnosis for stroke, the PLR and NLR was 5.85 (95% CI: 3.07–10.6) and 0.17 (0.08–0.30), respectively (Table 4; Fig. 1c, d). The AUC for HIT, the diagnosis of peripheral and stroke, was 0.90 and 0.92, respectively (Fig. 2a, b).

Table 4.

Summary of PLR, NLR for HIT and HINTS

Fig. 1.

a–d Forest plots of (a) positive and (b) negative likelihood ratio (LR) of HIT for peripheral vertigo, (c) positive and (d) negative LR of HIT for stroke (author, year, time to HIT examination, 95% CI). Horizontal axis indicates likelihood ratio expressed as a log function. HIT, head impulse test; LR, likelihood ratio; CI, confidence interval.

Fig. 2.

Summary ROC of (a) positive HIT for PV, (b) peripheral pattern HINTS for PV, (c) negative HIT for stroke, and (d) central pattern HINTS for stroke. The summary ROC contains an ellipse which represents the confidence estimate at the summary point estimate. ROC, receiver operating characteristic; AUC, area under the curve; HIT, head impulse test; HINTS, head impulse test with direction changing nystagmus and test of skew deviation.

Accuracy of the HINTS Exam in the Diagnosis of PV and Central Vertigo

In one paper which evaluated HINTS in the diagnosis of both stroke and nonstroke central causes of vertigo [20], we obtained PLR and NLR for HINTS with respect to stroke diagnosis. Overall, the PLR and NLR of HINTS for the diagnosis of PV were 17.3 (95% CI: 8.38–32.1) and 0.15 (95% CI: 0.07–0.26), respectively (Table 4; Fig. 3a, b). The PLR and NLR of central HINTS pattern for the diagnosis of stroke were 5.61 (95% CI 4.19–7.7) and 0.06 (0.03–0.12), respectively (Table 4; Fig. 3c, d). The AUC for HINTS for the both the diagnosis of PV and stroke was 0.96 (Fig. 2c, d).

Fig. 3.

a–d Forest plots of (a) positive and (b) negative likelihood ratio (LR) of HINTS for PV, (c) positive and (d) negative LR of HIT for stroke (author, year, time to HINTS examination, 95% CI). Horizontal axis indicates likelihood ratio expressed as a log function. HINTS, head impulse test with direction changing nystagmus and test of skew deviation; LR, likelihood ratio; CI, confidence interval.

The PLR and NLR for HIT and HINTS for the diagnosis of PV and stroke are summarized in Table 4. Forest plots and AUC graphs are presented in Figures 1and3, respectively.

Meta-Regression

Meta-regression did not demonstrate a statistically significant effect of publication year, time to test, and type of assessor on the transformed sensitivity or transformed false positive rate (online suppl. Fig. 2).

Discussion

The findings of this meta-analysis show that HIT and HINTS bedside tests have moderate PLR and low NLR, and therefore may make moderate changes in the likelihood of peripheral and central vertigo. The high AUC seems to be in agreement with the PLR and NLR. We suggest caution given the high degree of heterogeneity and that the bedside tests were performed at a late (after 24 h) time window. These issues may limit its usefulness as a bedside test in the ED, particularly if patients present within 24 h of symptom onset. Furthermore, we acknowledge that newer imaging techniques may circumvent the utility of these bedside tests where the main question is to distinguish stroke from PV.

Our meta-analysis is an advancement on previous ones on this topic given our inclusion and separate analysis of both HIT and HINTS with respect to peripheral and central causes of acute dizziness, vertigo, and AVSs. We were able to include 6 new publications [19, 21-24, 28], on this topic compared to a previous meta-analysis [4]. Our study raises the important issue of designing a diagnostic test with the need to evaluate the test in the setting in which it was intended. This issue was the reason for the use of the QUADAS-2-tool to assess the risk of bias and applicability of the bedside diagnostic test [8].

This emphasis on the timing of the HIT and HINTS tools is relevant given the above discussion. For example, a standard ED key performance indicator in Australia and the UK are that decisions regarding patient disposition are made within 4-h [27]. Our finding of the median time of 48 h in performing the HIT or HINTS is discordant with the tests having occurred in ED in most of the studies (see online suppl. Fig. 2) [9, 19, 20, 22, 23, 26, 28, 29]. This issue was reflected in the QUADAS-2-tool “workflow” domain. Furthermore, it is not clear when the HIT and HINTS were performed with respect to the timing of brain imaging. In the modern era of reperfusion therapy for the treatment of stroke, the current “Get with the Guidelines” program guidelines in North America recommend multimodality brain imaging within 25 min of arrival [30]; thus, the utility of this test in the hyperacute stages of stroke care remains uncertain and should not delay prompt neuroimaging.

The studies had diverse inclusion and exclusion criteria. Many studies excluded preexisting peripheral vestibular conditions. Thus, the HIT and HINTS were not necessarily performed in an undifferentiated population that would be more likely encountered in the ED setting. Additionally, some of papers used either CT or MRI scans instead of only MRI as the reference standard [20]. CT scans are not sensitive for posterior circulation infarcts and MRI can miss posterior circulation infarcts in 19% of cases [31] and would significantly limit the reliability of included study that used this as the imaging modality in more than 50% of patients [25]. Furthermore, none of the studies had ED physicians exclusively performing the testing. This has implications in real-world applicability as having timely specialist (neurologist) and subspecialist (neuro-otologists) review is not feasible in all ED settings.

Limitations

These issues discussed above on the quality of the studies may explain the significant limitations of the HIT and HINTS as bedside test. We were not able to identify variables causing study heterogeneity. Inherent to meta-analyses methodology, this meta-analysis is limited by the availability of published studies, many of which were duplicated results, or were excluded on the basis that 2 × 2 data could not be verified.

Conclusion

The HIT and HINTS are potentially useful bedside tests but need further evaluation in ED by doctors who are likely to be the ones performing these tests in that setting. These tests need to be applied at a timeline that is appropriate to the workflow in ED and not at some later timeline or in subspeciality clinic. We urge caution when applying these tests to a real-world emergency setting.

Statement of Ethics

An ethics statement is not applicable because this study is based exclusively on published literature.

Conflict of Interest Statement

Suyi Ooi, Grace Phillips, Tanya Tang, Luke Chen, Anthony Fok, John Ly, and Henry Ma have no conflict of interests or disclosures to declare. Thanh G. Phan has received honorarium for providing lectures for Bristol-Myer-Squibb on anticoagulation in stroke.

Funding Sources

The authors have no sources of funding to declare.

Author Contributions

Suyi Ooi, Grace Phillips, Tanya Tang, and Thanh G. Phan were involved in search strategy, article review, data collection, and data analysis. Luke Chen, Anthony Fok, and John Ly were independently involved in quality assessment of included articles. Suyi Ooi, Grace Phillips, and Thanh G. Phan were involved in manuscript writing. Henry Ma contributed to reviewing and editing the manuscript.

Data Availability Statement

All data generated or analysed during this study were included in this article and its online supplementary material. Further enquiries can be directed to the corresponding author.

This article is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC). Usage and distribution for commercial purposes requires written permission. Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug. Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.