The Head Impulse Test as a Predictor of Videonystagmography Caloric Test Lateralization According to the Level of Examiner Experience: A Prospective Open-Label Study

Abstract

We conducted a study to compare how well the head impulse test (HIT), without and with eye-movement recordings, would predict videonystagmographic (VNG) caloric test lateralization when performed by a resident and an experienced otoneurologist. This prospective, open-label, blinded study was conducted in an ambulatory tertiary care referral center. Our study population was made up of 60 patients—29 men and 31 women, aged 20 to 82 years (mean: 56.4 ± 11.4)—with peripheral vestibulopathy who underwent HIT and VNG caloric testing. The HIT was conducted in two protocols: HIT0 and HIT1. The HIT0 was performed with passive brisk movements of the patient's head from the 0° null position to 20° sideways, and the HIT1 was performed toward the center while the null position was a 20° head rotation to the right and to the left. Each protocol was carried out without video eye-movement recordings (HIT0 and HIT1) and with such recordings (rHIT0 and rHIT1). The primary outcome measures were (1) a comparison of the HIT's sensitivity and specificity when performed by the resident and by the experienced otoneurologist and (2) the ability of video-recorded HIT to predict VNG caloric test lateralization. The sensitivity and specificity obtained by the resident were 41 and 81%, respectively, for HIT0 and 41 and 90% for HIT1. The sensitivity and specificity obtained by the experienced otoneurologist were 18 and 89% for HIT0 and 32 and 85% for HIT1. Analysis of the recorded eye-movement clips of the HIT0 and HIT1 obtained by a second experienced otoneurologist found a sensitivity and specificity of 32 and 63% for rHIT0 and 33 and 82% for rHIT1. We conclude that the HIT yields high false-negative rates in predicting significant caloric lateralization. Analysis of the eye-movement recordings was no better than normal testing alone for detecting saccades. The experience of the examining physician had no impact on test performance characteristics.