Predicting Postoperative Speech Perception and Audiometric Thresholds Using Intracochlear Electrocochleography in Cochlear Implant Recipients

Abstract

Objectives: Electrocochleography (ECochG) appears to offer the most accurate prediction of post-cochlear implant hearing outcomes. This may be related to its capacity to interrogate the health of underlying cochlear tissue. The four major components of ECochG (cochlear microphonic [CM], summating potential [SP], compound action potential [CAP], and auditory nerve neurophonic [ANN]) are generated by different cochlear tissue components. Analyzing characteristics of these components can reveal the state of hair and neural cell in a cochlea. There is limited evidence on the characteristics of intracochlear (IC) ECochG recordings measured across the array postinsertion but compared with extracochlear recordings has better signal to noise ratio and spatial specificity. The present study aimed to examine the relationship between ECochG components recorded from an IC approach and postoperative speech perception or audiometric thresholds. Design: In 113 human subjects, responses to 500 Hz tone bursts were recorded at 11 IC electrodes across a 22-electrode cochlear implant array immediately following insertion. Responses to condensation and rarefaction stimuli were then subtracted from one another to emphasize the CM and added to one another to emphasize the SP, ANN, and CAP. Maximum amplitudes and extracochlear electrode locations were recorded for each of these ECochG components. These were added stepwise to a multi-factor generalized additive model to develop a best-fit model predictive model for pure-tone audiometric thresholds (PTA) and speech perception scores (speech recognition threshold [SRT] and consonant-vowel-consonant phoneme [CVC-P]) at 3- and 12-month postoperative timepoints. This best-fit model was tested against a generalized additive model using clinical factors alone (preoperative score, age, and gender) as a null model proxy. Results: ECochG-factor models were superior to clinical factor models in predicting postoperative PTA, CVC-P, and SRT outcomes at both timepoints. Clinical factor models explained a moderate amount of PTA variance (r 2 = 45.9% at 3-month, 31.8% at 12-month, both p < 0.001) and smaller variances of CVC-P and SRT (r 2 range = 6 to 13.7%, p = 0.008 to 0.113). Age was not a significant predictive factor. ECochG models explained more variance at the 12-month timepoint (r 2 for PTA = 52.9%, CVC-P = 39.6%, SRT = 36.4%) compared with the 3-month one timepoint (r 2 for PTA = 49.4%, CVC-P = 26.5%, SRT = 22.3%). The ECochG model was based on three factors: maximum SP deflection amplitude, and electrode position of CM and SP peaks. Adding neural (ANN and/or CAP) factors to the model did not improve variance explanation. Large negative SP deflection was associated with poorer outcomes and a large positive SP deflection with better postoperative outcomes. Mid-array peaks of SP and CM were both associated with poorer outcomes. Conclusions: Postinsertion IC-ECochG recordings across the array can explain a moderate amount of postoperative speech perception and audiometric thresholds. Maximum SP deflection and its location across the array appear to have a significant predictive value which may reflect the underlying state of cochlear health.