The strategy includes the descriptors (DECs and MESH) and Free terms (TL), based on the two first elements of PIC (Population, Interest, Context) present in the title, which consisted of: (senile OR Age-related OR Aged OR Aging OR Ageing Effect OR Ageing OR older) AND (vestibular evoked myogenic potential OR vestibular potential OR VEMP OR Cervical evoked potential OR Ocular evoked potential OR [Vestibular AND evoked AND Potential]). The complete strategy can be found in the supplementary material (Table 1).

The searches were carried out between the months of July and August of 2016, and were revised in September of the same year. The following databases were searched: Pubmed, ScienceDirect, BVS (LILACS), SCOPUS, Circumpolar Health Bibliographic Database, SciELO and EMBASE, as well as the gray literature databases: OpenGrey.eu, DissOnline, The New York Academy of Medicine, as well as ReasearchGate. There was no manual search of the included articles and experts in the area were not contacted to avoid the risk of citation bias.29

The following were considered inclusion criteria: observational studies, with groups of elderly individuals, with age groups of 55 years or older, with control group, with latencies of ocular and/or cervical vestibular evoked myogenic potentials. Additionally, the potentials should be evoked by acoustic stimuli such as Click or 500Hz Toneburst, with intensity between 90 and 105dBNAn. Exclusion criteria were: conductive hearing loss, sensorineural hearing loss equal to or greater than 50dB at any frequency, control group with age group containing subjects 55 years of age or older, vestibular, neural pathologies, diabetes, or Parkinson's disease. Articles that were repeated in different databases were also excluded. Finally, studies with at least the title and/or abstract in English were included, but there was no restriction regarding language or date of publication.

During the selection process, the titles and abstracts of the obtained articles were independently evaluated by two researchers who were not blinded to the authors or journal title. Disagreements were resolved by discussion. In cases where there was no consensus, a third author was asked to make the final decision. The full texts of potentially eligible articles were acquired and analyzed in full.

The outcome sought in the studies was the mean latency values of the biphasic components for cVEMP and/or oVEMP and in the second assessment, associated with a dispersion measure.

Data were analyzed from published articles and authors were contacted for additional information. In addition to the outcome data, we also obtained the names of the authors, title, year of publication, country, age ranges of the groups, number of subjects in each group, monitored muscles and auditory examinations. A standard form for data storage was created based on the model used by Cochran.30

The risk of bias was assessed according to the recommendations of the “Newcastle-Ottawa” manual and scale,31 adapted for cross-sectional observational studies. The quality of the study was independently evaluated by two researchers and the divergences were resolved by consensus. The maximum score to be reached was ten points and the evaluated scale items were: (1) representativeness of the sample; (2) sample size; (3) management of non-responses; (4) exposure calculation (risk factor); (5) comparability, to investigate whether individuals in different groups of outcomes are comparable, based on study design or analysis, control of confounding factors; (6) evaluation of results and (7) statistical test (Table 2).

The flow diagram that illustrates study search and selection is shown in Fig. 1. Of the 7544 titles considered relevant from the searches in these databases, 322 abstracts were read and, of those, 61 full texts were selected for reading in full. After reading, 41 articles were excluded, as they did not meet the eligibility criteria and four because they did not have sufficient data and their authors did not respond to the request for additional information (Table 3). Therefore, 16 full texts were included in the qualitative and quantitative analysis (Table 4). The latency means of young adults and elderly individuals of the meta-analyzed articles are shown in Table 5 (oVEMP) and Table 6 (cVEMP).

Figure 1.

Flowchart of article search and selection.

(0.3MB).

Among the selected studies, only three assessed oVEMP. However, one of them did not have p1 latency data and, thus, the meta-analysis of this component was very compromised. On the other hand, 13 articles had mean and standard deviation data for the cVEMP latency components, p13 and n23, for the control group and for the elderly group. Of these, four studies found a significant difference between the groups, one of them found a significant difference for p13 and non-significant for n23, three studies were not clear whether there were differences and five affirmed that there were no differences between groups.

A total of 120 subjects were studied for the assessment of n1 and p1 latencies of oVEMP, 60 of which were elderly and 60 were young adults, and 326 subjects, of which 296 elderly and 326 young adults, were evaluated for cVEMP p13 and n23 latencies.

In Table 6, all data on means and standard deviations were provided except the standard deviations of p13 and n23 latencies of the elderly of one of the studies, when evoked by Clicks. In this case, the standard deviations were calculated by applying an international convention in which the quotient: mean/2.5 is used to find the standard deviation.

The analysis of the quality of the included articles and, consequently, of the risk of bias, is shown in Table 7.21,23–27,32–40 All included studies are characterized as observational and cross-sectional studies. In addition, in the final evaluation, all had a percentage of quality equal to or superior to 50% (5/10), whereas two of them obtained a maximum score of 70% (7/10).

Only one study assessed the sample representativeness,25 as it was a normative study and analyzed all available subjects in a certain period. All other studies made choices per convenience group.

The satisfactory sample size of the elderly group was a concern of four studies,33–36 which conform to the central limit theorem, with samples larger than 30 subjects. However, none of them performed calculations to estimate the size of their samples.

The non-response rate was satisfactory in 50% of all studies using validated tools for data collection and the comparability between the control group and the elderly group was also possible for all of them. The evaluation of the results was carried out in all the studies through their own reports, except in the two studies,33,34 in which wave analysis was carried out by two independent professionals. Finally, all studies used appropriate statistical tests.

As the studies are not randomized, the groups showed great discrepancy as early as in the first evaluation. Thus, to avoid the phenomenon of regression to the mean, it was necessary to analyze the variations between the final and initial latency values, as well as the standard deviation associated to these variations.

The number of articles to be meta-analyzed for oVEMP n1 latencies was small, as there were only three of them.33,37,38 The mean delay of this component for the latencies of the elderly was 2.32ms with 95% CI of 0.55–4.10ms. The overall effect test showed p=0.01; disclosing that such difference was significant. However, the heterogeneity I2=96%, with a p value <0.001 (Fig. 2A). Finally, due to the small number of selected studies, it was not possible to analyze the subgroups to understand the origins of this heterogeneity.

Figure 2.

Meta-analysis: comparison of n1 oVEMP and p13 cVEMP latencies. (A) n1 oVEMP. (B) p13 cVEMP.

* The study by Janky and Shepard 2009 appears twice, as it was carried out two different tests, one with click and another with TB.

(0.65MB).

On the other hand, only two articles were found for the p1 component,37,38 which considerably affected the analyses, as previously described, and made its study impossible.

The number of articles to be meta-analyzed for cVEMP components p13 and n23 latencies was quite encouraging. Thus, 13 were selected (described in Table 7).

For the p13 component, the mean delay for the latencies in the elderly was 1.34ms with 95% CI of 0.56–2.11ms. The overall effect test showed p<0.001; disclosing that such a difference was significant. However, a heterogeneity value of I2=92% was found, with p<0.001 (Fig. 2B).

The attempts to analyze the subgroups were not successful in explaining heterogeneity. When dividing the groups by used stimuli to evoke cVEMP, Toneburst or Click, in both cases, it remained high and with p<0.001, as can be seen in Fig. 3. The same was done for the stimulus intensity (up to 95dBNAn and >95dBNAn) and for the age ranges of the control groups (20–30 years and different <20–30 years), yet both evaluations were unsuccessful.

Figure 3.

Meta-analysis: comparison of n13 cVEMP latencies, sub-groups evoked by 500Hz Toneburst and only by Click. (A) p13 cVEMP (sub-group evoked only by 500Hz Toneburst). (B) p13 cVEMP (sub-group evoked only by Click).

(0.63MB).

For component n23, the mean delay for the latencies in the elderly was 2.82ms with 95% CI of 0.33–5.30ms. The test for the overall effect showed a p=0.03; disclosing that the difference was significant. However, a high heterogeneity value of I2=99% was found, with p<0.001 (Fig. 4A).

Figure 4.

Meta-analysis: comparison of n23cVEMP latencies, n23 cVEMP sub-group evoked only by 500Hz Toneburst and n23 cVEMP sub-group evoked only by Click. (A) n23 cVEMP latencies. (B) n23 cVEMP latencies, sub-group evoked only by 500Hz Toneburst. (C) n23 cVEMP latencies, sub-group evoked only by Click.

(1.05MB).