Screening for discordant outcomes in cochlear implant users: Classification based on ECAP amplitudes

Cochlear implant (CI) recipients show considerable variability in postoperative speech recognition outcomes. While electrically evoked compound action potential (ECAP) recordings are routinely available in modern CI systems, their usefulness in routine clinical practice as a potential outcome measure has not been systematically established.

This study examined the relationship between routinely measured ECAP characteristics as amplitude (A-ECAP) and threshold (T-ECAP) and postoperative word recognition scores in CI users and derived a clinically feasible ECAP-based screening method. Good and poor word recognition were defined as WRS₆₅(CI) above and at or below a benchmark of 30%, respectively.

Data from two independent clinical datasets were analyzed. The screening procedure was developed using data from Clinic 1 (postoperative ECAP recordings at 200 cL) and validated using data from Clinic 2 (intraoperative recordings at 230 cL). A transformation procedure, based on the regression between ECAP threshold and amplitude empirically derived within each dataset, was applied to convert ECAP amplitudes obtained at 230 cL in Clinic 2 into equivalent values at 200 cL, thereby enabling cross-centre comparability. Receiver operating characteristic (ROC) analysis was used to derive a critical A-ECAP cut-off value targeting a word recognition benchmark of 30%.

High A-ECAP consistently associated with high word recognition across both datasets, whereas T-ECAP showed no comparable consistency. A critical ECAP amplitude of 121 µV (at a fixed stimulation level of 200 cL) yielded 100% specificity in the derivation dataset. In the validation dataset the positive predictive value was 94%, with five of 165 recipients showing a discordant outcome, defined as an amplitude above the criterion combined with poor word recognition CONCLUSIONS: Tools for recording ECAP are well established in clinical routine, but little attention is given to the information these responses carry regarding speech recognition. A large ECAP amplitude was consistently associated with good word recognition, whereas ECAP threshold was not. ECAP amplitude does not predict individual outcome. Its value lies in identifying discordant cases - a large amplitude combined with poor word recognition - in which the outcome cannot be attributed to the peripheral response and further evaluation, beginning with the fitting, is indicated. Whether these recipients benefit from re-intervention requires prospective evaluation.