Pre-, Per- and Postoperative Factors Affecting Performance of Postlinguistically Deaf Adults Using Cochlear Implants: A New Conceptual Model over Time
1Bionics Institute, Melbourne, Australia
2Service d'otologie et d'otoneurologie, Hôpital R.-Salengro, CHRU de Lille, Lille, France
3CHU Gui de Chauliac, Service d'ORL et Chirurgie Cervico-Faciale, Montpellier, France
4Department of Otorhinolaryngology and Head and Neck Surgery, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium
5Hospices Civils de Lyon, Hôpital Edouard Herriot, Département d'ORL, de Chirurgie Cervico-Maxillo-Faciale et d'Audiophonologie, Lyon, France
6AP-HP, Hôpital Beaujon, Service d'ORL et Chirurgie Cervico-Faciale, Clichy, France
7Institute of Physiology and Pathology of Hearing, Warsaw, Poland
8Institute of Sensory Organs, Kajetany, Poland
9The Eargroup, Antwerp, Belgium
10Department of Otolaryngology, The University of Melbourne Cochlear Implant Clinic, The Royal Victorian Eye and Ear Hospital, Melbourne, Australia
11University of Manchester, Central Manchester University Hospitals NHS Foundation Trust, Manchester, United Kingdom
12University of Groningen, University Medical Center Groningen, Department of Otorhinolaryngology/Head and Neck Surgery, Cochlear Implant Center Northern Netherlands, Groningen, The Netherlands
13Department of Otorhinolaryngology, University Hospital of Zurich, Zurich, Switzerland
14Hôpital Universitaire Purpan, Service d'ORL et Chirurgie Cervico-Faciale, Toulouse, France
15St Thomas' Hospital, Auditory Implants Department, London, United Kingdom
16Otorhinolaryngology, Radboud University Nijmegen Medical Center, Mijmegen, The Netherlands
17Faculté de médecine, Université Laval, Québec City, Québec, Canada
18Graduate School of Medical Sciences (Research School of Behavioural and Cognitive Neurosciences), University of Groningen, Groningen, The Netherlands
19Institut Saint Pierre, Service d'Audiophonologie et ORL, Palavas les flots, France
Manuel S. Malmierca, Editor
University of Salamanca-Institute for Neuroscience of Castille and Leon and Medical School, Spain
Received May 3, 2012; Accepted September 28, 2012.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Several new statistical analyses were conducted on the dataset described in Blamey et al (in press). The dataset consisted of retrospective information for 2251 CI recipients evaluated with various speech tests and conditions (quiet and noise) from 15 international centres. This project was approved by the Royal Victorian Eye and Ear Hospital Human Research Ethics Committee (Project 10/977H, Multicentre Study Of Cochlear Implant Performance In Adults). In a multicenter study, one of the challenges is to combine information in a useful manner despite the differences in evaluation methods and data recorded in the individual centres. All centres provided the core information on implant performance (on an open-set speech perception test in quiet and in noise without lipreading), duration of s/p HL, age at onset of s/p HL, etiology, and cochlear implant experience. Most centres provided additional information (such as use of HAs before surgery, duration of mHL, and amount of residual hearing) if it was available. The statistical analyses thus included additional factors beyond those used in Blamey et al. (in press) (duration of s/p HL, age at onset of s/p HL, duration of CI experience, and etiology). The number of data points included in each analysis varied because of missing data from some clinics on some factors.
Selection criteria for CI recipients included in the study were: Adult at the time of implantation (>18 years old); Onset of s/p HL after the age of 15 (time from which the patient could no longer use hearing alone to communicate even with the best-fitted hearing aids, and/or understand TV, and/or stopped using the telephone). Four brands of CIs were included (Advanced Bionics, Cochlear, Med-el, and Neurelec). Their proportions in the sample were 21%, 50%, 17%, and 7%, respectively (plus 5% missing data for this variable). Date of implantation was after 2002 for all recipients to include technically comparable improvements across brands.
Speech scores in quiet and in noise at two postoperative times for each recipient were requested from the clinics: one score collected early after activation of the CI (T1) and one score collected later on (T2). The choice of the date of the tests was free and varied between and within centres. The mean and standard deviation for T1 were 0.5 years and 0.8 years, respectively, and 2 years and 1.7 years for T2, respectively.
The four factors, used in the four-factor general linear model of Blamey et al (in press) were: duration of s/p HL, defined as the time in years between the onset of s/p HL and the date of implantation; mean and median durations of s/p HL were 7.4 years and 3.2 years, respectively (ranges: 0–60 years, standard deviation: 9.8); mean age at onset of s/p HL was 50 years (standard deviation: 17.3); duration of implant experience was defined as the time elapsed between the date of first activation and the dates of testing. It ranged from 2 months to 12 years; fifteen etiologies were defined. They are detailed in
Age at implantation ranged from 17 years to 93 years (mean: 58, standard deviation: 15.8). It was not included in the four-factor general linear model of Blamey et al (in press) because it had less effect than age at onset of s/p HL.
Absolute numbers of the various etiologies defined in the dataset.
In addition to duration of s/p HL, age at onset of s/p HL, duration of CI experience, and etiology, several pre-, per-, and post-operative factors were added to the statistical analyses.
The preoperative factors were:
- gender. There were 1017 females, 820 males, and 414 patients with missing gender data.
- education level, corresponding to the age at which the subject stopped studying. This factor was partitioned into ranges: stopping before the age of 12 years, before the age of 18 years, or continuing after 18 years old. These three ranges encompassed 21 subjects, 501 subjects, and 518 subjects, respectively, plus 1211 patients with missing data.
- duration of moderate hearing loss (mHL), lasted from the onset of mHL to the onset of s/p HL. The ranges used for the analyses were: 0–4, 5–9, 10–14, 15–19, 20–24, 25–34, 35–44 and over 45 years. The mean duration of mHL was 17 years (range: 0 in case of sudden hearing loss to 74 years, standard deviation: 14.6 years).
- preoperative HA use. Centres reported whether the subject was using HAs, bilaterally, or monaurally, at the time of implantation. Four groups were defined: patients not using any HA, patients with a HA on the implanted side, patients with a HA on the ear contralateral to the implanted side, and patients wearing two HAs. These four ranges encompassed 429 subjects, 289 subjects, 386 subjects and 712 subjects, respectively (the rest had missing data).
- pure tone average (PTA) of the implanted ear, and the better PTA of the two ears. The latter will be called PTA of the better ear. PTAs represented the mean of unaided residual hearing levels in decibels (dB HL) at 500, 1000, 2000 Hz, for all centres. Four ranges were used: 40–49, 50–74, 75–99, and 100+dB HL.
- hearing loss (HL) at 500 Hz. This variable was included to test whether residual low frequencies were more relevant to maintaining functional auditory pathways. Similarly to PTA, HL at 500 Hz was considered for both the implanted ear and for the better ear. The ranges used were the same as for PTA.
- preoperative speech scores in quiet. These were aided speech scores before implantation. A percentile rank for each patient within each centre was calculated to allow for differences in test type (phonemes, monosyllabic words, dissyllabic words, sentences) in different languages and different levels of presentation (from 55 to 75 dB SPL). Using ranking removes differences in clinical practice without removing the relative differences between patients within each centre .
- date at implantation. Modifications of coding strategies since 2002 were tested indirectly through the date at implantation. Three ranges were used: 2002–2004, 2005–2007, 2008–2011. These three ranges encompassed 345 subjects, 822 subjects, and 1083 subjects, respectively (one date was missing).
- implanted ear. The implanted ear was classified as the better ear, the worse ear, or similar when both ears had the same amount of residual hearing. The better ear was implanted in 611 cases, the worse ear in 1142 cases, and the two ears had similar residual hearing in 294 cases (the rest had missing data).
Peroperatively, only one factor was studied:
- surgical approach. Cochleostomy and round window approaches were compared. They were performed in 1119 cases and 425 cases, respectively (information was missing in 707 cases).
The postoperative factors were:
- CI brand. CIs from four different manufacturers were represented in the dataset. Speech processors for Advanced Bionics were Auria and Harmony, for Cochlear processors included models from Esprit3G to CP810, for Med-el were Tempo+ and Opus 2, and for Neurelec the Digisonic SP processor.
- angle of insertion of the electrode array: Depth of electrode array insertion was expressed as an angle divided into three ranges: <370°, 370–539°, and ≥540°. It ranged from 135° to 730°.
- percentage of active electrodes. The number of active electrodes reported at the first testing was expressed relative to the total number of electrodes available on the electrode array, as the total number varies with the CI brand. The ranges used were: ≤70%, 71–85%, and >85%. The minimum percentage of active electrodes was 15%.
Statistical analyses of speech scores in quiet
Postoperative speech scores in quiet were transformed into percentile ranks for each patient within each centre. Using ranking removes differences in clinical practice without removing the relative differences between patients within each clinic. Indeed, for each clinic, the distribution varied uniformly from 0 to 100. The best performers from each centre had a percentile rank close to 100, and the poorest performers from each group had a percentile rank close to 0. The ranked data of the centres were combined for the global analysis. Preoperative and postoperative scores were ranked separately. Ranked postoperative scores were used as the dependent variables of the statistical analyses described below.
Each new factor that we wanted to test was added into the four-factor unbalanced analysis of variance using the General Linear Model (GLM; Minitab version 12), previously described in Blamey et al (in press) to create fifteen five-factor ANOVAs. Briefly, a GLM studies the influence of various independent factors on a dependant variable. The four-factor ANOVA described in Blamey et al (in press) was based on main well-established general factors (the independent factors), known to influence CI speech performance (the dependant variable). These four common factors were duration of s/p HL, age at onset of s/p HL, duration of CI experience, and etiology. In the present study, we wished to explore the influence of 15 other factors that have been less studied. Because entering 19 different independent factors was not possible with the software used (Minitab version 12) and because interpretation of the results would have been complicated, we entered into the former four-factor GLM a single new factor once at a time, leading to 15 different five-factor GLMs. From these 15 analyses, only factors with p≤0.001 were selected. These significant factors were further included in a single GLM analysis (a sixteenth analysis) to investigate the interrelations between them and produce a new model of auditory performance.
Statistical analyses of speech scores in noise
Postoperative speech scores in noise were ranked separately for each patient within each centre, and independently of scores in quiet. The noise used varied across centres from a cocktail party, to a pink noise, or a speech shaped noise, but was the same for all patients of the same center. Scores in noise at T1 and T2 were considered independent scores for the same patient, and used as dependent variables in the analysis used to explore the new model of auditory performance (the sixteenth analysis). This last analysis is detailed further in the Results