Agreement, Reliability, and Measurement Error of Cervical Vestibular Evoked Myogenic Potentials in Patients: II. Meta-Analysis and Reliability of Latency and Amplitude
Jonas Bruun Kjærsgaard, Rodrigo Ordoñez, Anne Vingaard Olesen, Herman Kingma, Dan Dupont HougaardObjectives:
This study assessed the reliability and measurement error of cervical vestibular evoked myogenic potential (cVEMP) latency and amplitude parameters in cochlear implant candidates and synthesized previous reliability evidence through an embedded meta-analysis. Together with a preceding paper on cVEMP activation and threshold reliability, this investigation extends the evaluation of clinically used cVEMP measurement properties to latency and amplitude outcomes.
Design:
A test–retest study was conducted in adults with profound bilateral hearing loss. Each participant underwent two independent cVEMP examinations using identical tone-burst stimulation (500 Hz, 2 msec duration) and recording parameters. Reliability was assessed for the p13 and n23 latencies, the inter-peak interval (IPI), and the corrected peak-to-peak amplitude (cAMP). Variance components were estimated using random-effects models accounting for subject, side, and test session. Single-measure, absolute-agreement intraclass correlation coefficients (ICCs), SEs of measurement (SEM), and minimal detectable change (MDC 95% ) were derived from the variance components. A random-effects meta-analysis of previous cVEMP reliability studies was conducted after transformation of reported ICCs to single-measure form and calculation of confidence intervals.
Results:
Twenty-seven patients contributed to the full analysis. For threshold-level trials, ICCs (95% confidence interval) were: p13 = 0.61 (0.41 to 0.81), n23 = 0.32 (0.10 to 0.54), IPI = 0.29 (0.07 to 0.51), and cAMP = 0.16 (0.00 to 0.36). Corresponding SEMs were 1.27, 1.87, 1.47, and 0.16 msec, with MDC 95% values of 3.5, 5.2, 4.1, and 0.45 msec, respectively. The wide confidence intervals and SEMs revealed considerable within-subject variability across sessions. For the meta-analysis, seven studies were identified, and up to five contributed to each outcome-specific analysis because sufficient methodological clarity permitted conversion to single-measure ICCs. Pooled random-effects estimates (95% confidence interval) were: p13 = 0.52 (0.38 to 0.64), n23 = 0.57 (0.39 to 0.71), IPI = 0.47 (0.08 to 0.73), and cAMP = 0.63 (0.44 to 0.76). Between-study heterogeneity varied across outcomes, ranging from negligible to high, reflecting methodological and analytical differences. Despite these differences, the clinical and pooled data were concordant, each showing only a moderate degree of reliability.
Conclusions:
Across the clinical test–retest data and the pooled evidence, cVEMP latency and amplitude measures showed moderate reliability with wide confidence intervals and notable measurement error. The MDC 95% values for cVEMP latencies indicate that differences of several milliseconds are required to exceed random variability expectable from test repetition, limiting their interpretability for individual monitoring. Considered together with the meta-analytic findings and the preceding results on cVEMP activation and threshold detection, these findings suggest that current cVEMP measurement procedures do not achieve the reliability required for widespread clinical use in individual patients. Although cVEMP outcomes may be informative for group-level or research comparisons when measurement precision is explicitly considered, their role as diagnostic or monitoring tools should remain limited until methodological refinements demonstrably improve reproducibility.