Individual Reliability of the Standard Clinical Method vs Patient-Centered Tinnitus Likeness Rating for Assessment of Tinnitus Pitch and Loudness Matching

Key Points

Question

Can a patient-centered tinnitus likeness rating approach provide greater test-retest 1-month reproducibility than the standard 2-alternative forced-choice method for tinnitus assessment?

Findings

In this case series that included 31 patients with tinnitus and compared the 2 approaches repeated at a 1-month interval, concordance between test and retest was much lower for 2-alternative forced-choice than for tinnitus likeness rating.

Meaning

Superior test-retest concordance afforded by the tinnitus likeness rating approach is a significant improvement for the study of personalized therapeutic interventions.

This case series assessed individual test-retest reliability of the 2-alternative forced-choice and tinnitus likeness rating approaches in patients undergoing evaluation for tinnitus.

Abstract

Importance

Current individualized sound therapies for tinnitus rely on tinnitus pitch assessment, which is commonly derived from the standard clinical 2-alternative forced-choice (2-AFC) approach driven by the examiner. However, this method is limited by lack of individual test-retest reliability and focuses on a single rather than multiple tinnitus frequencies.

Objective

To assess individual test-retest reliability of the 2-AFC, with a single final frequency (and corresponding loudness), and the tinnitus likeness rating (TLR), with the participant exposed to the entire audible frequency spectrum, from which 3 dominant frequencies and corresponding loudness were extracted.

Design, Setting, and Participants

In this case series, participants with tinnitus underwent testing twice with both methods at a 1-month interval by experienced clinicians from January 6 through March 17, 2017. Each clinician tested each patient only once at visit 1 or 2 in a university audiology training setting with standardized equipment and was blind to previous assessment. Participants with bilateral or unilateral chronic tinnitus for longer than 6 months, in good health, without total deafness in either ear, and without cerumen in the ear canal were recruited through advertisements (community and clinics) and word of mouth (volunteer sample). The audiologists were likewise participants in the planned comparison between TLR and 2-AFC in the test-retest measures.

Main Outcomes and Measures

Test-retest concordance with 95% CIs for each method, calculated as the proportion of participants with the same final frequency between the 2 visits (2-AFC) or with at least 1 concordant dominant frequency (TLR) as well as loudness differences of no greater than 10 dB.

Results

The study sample included 31 participants (55% men; mean [SD] age, 50.7 [13.7] years). For TLR, 26 of 31 participants had at least 1 concordant dominant frequency between the 2 visits (proportion, 0.84; 95% CI, 0.66-0.95), whereas for 2-AFC, 7 of 31 participants had a concordant final tinnitus pitch in either ear (proportion, 0.23; 95% CI, 0.10-0.41). Loudness reliability followed the same pattern, with more concordant loudness levels in the TLR (proportion, 0.73; 95% CI, 0.52-0.88) than in the 2-AFC (proportion, 0.40; 95% CI, 0.05-0.85). Mean time taken to complete the tests was less than 15 minutes, and general appreciation by participants with tinnitus and audiologists were overall similar for both.

Conclusions and Relevance

Superior test-retest concordance can be demonstrated at the individual level using the several dominant frequencies extracted from the patient-centered TLR.

Introduction

Tinnitus—the perception of sound in the absence of an external source—is a common hearing disorder causing significant distress in 1% to 2% of the general population^1,2 and carrying elevated socioeconomic burden.³ In view of its association with hearing loss, tinnitus is an expanding problem among young and elderly individuals in the context of increased noise exposure and life expectancy.^4,5,6 In addition to audiometric, physical, and psychological examination,^7,8 some guidelines^9,10 further recommend measurement of the psychoacoustic properties of tinnitus, such as pitch and loudness, an essential step in the individualized management of tinnitus.^11,12 Several emerging sound treatments rely on individualized measurement of tinnitus pitch rather than wideband sound therapy.^13,14,15 Using stimuli critically tailored around each individual’s tinnitus frequency is a common feature of newer hearing aids equipped with noise generators,^15,16 dedicated sound devices,^{17,18,19,20,21,22,23,24,25,26,27,28} notched^{29,30,31,32,33,34,35,36,37,38,39,40,41,42,43} and spectrally altered music,⁴⁴ and frequency discrimination training,^{45,46,47,48,49,50} sometimes used in combination with vagus nerve stimulation.^51,52,53 Therefore, obtaining reliable measures of tinnitus dominant pitch is imperative.⁵⁴

In such applications, a single tinnitus pitch is usually sought via a standard, 2-alternative forced-choice (2-AFC) clinical procedure driven by the examiner (yielding a final frequency),^55,56 under the assumption that tinnitus can be adequately represented by a single frequency, disregarding individual test-retest reliability. Although routinely used in clinical trials and research, the reliability of tinnitus pitch assessment has been questioned since its early days^57,58 and still is.⁵⁴ Interestingly, elevated test-retest reproducibility was demonstrated for frequencies and loudness for as long as several months at the group level^59,60,61 in studies in which participants rated the likeness of single pure tones in the audible range (0.250-16.00 kHz); however, the concordance between test and retest was unclear at the individual level in such studies when using 1 dominant pitch. Single-pitch reliability might be intrinsically limited by the fact that tinnitus encompasses a spectrum of frequencies that mirror hearing loss, rather than a single frequency,^59,60,62 even when tinnitus is described by the participant as “tonal.”^60,63

Herein, tinnitus pitch and loudness 1-month test-retest reproducibility was assessed in the same individuals using the standard clinical method (2-AFC) and a novel tinnitus likeness rating (TLR) method^60,64 in which patients with tinnitus determined independently from the examiner a likeness rating for all presented frequencies, from which as many as 3 dominant frequencies were extracted and considered in the reproducibility analyses. We thus investigated the proposition that superior test-retest concordance can be demonstrated using the several dominant frequencies extracted from the TLR.

Methods

Study Participants and Audiologists

Thirty-one adults with tinnitus who had never been exposed to the TLR approach nor consulted the clinicians involved in this study were recruited through advertisements (community and clinics) and word of mouth. The study was conducted according to the scientific and ethics rules of the University of Montreal, Montreal, Quebec, Canada, and approved by the local health ethics committee. All participants provided written informed consent.

Mean (SD) tinnitus duration was 11.4 (9.0) years. Mean (SD) score on the French language version of the Tinnitus Handicap Inventory⁶⁵ was 29.6 (18.8) (range, 10-82, with higher scores indicating greater handicap). Inclusion criteria consisted of bilateral or unilateral subjective chronic tinnitus for longer than 6 months,⁸ good health, no total deafness in either ear, and no cerumen in the ear canal (as verified by otoscopy and managed when necessary). Audiometric assessment was performed in soundproof booths with the use of an audiometer (AC; Interacoustics) and headphones over the standard 0.25- to 8.00-kHz spectrum (TDH-39; Telephonics) and over 9.00 to 1600 kHz (HDA-200; Sennheiser, GmbH & Co). All assessments were performed by the same 3 clinical audiologists (identified as clinicians 1, 2, and 3) experienced with tinnitus matching and paid standard honoraria. They were involved in similar numbers of test (visit 1) and retest (visit 2) sessions, including 10 at visit 1 and 11 at visit 2 for clinician 1; 13 at visit 1 and 9 at visit 2 for clinician 2; and 8 at visit 1 and 11 at visit 2 for clinician 3. The participants with tinnitus and the audiologists were actors in the planned comparison between TLR and 2-AFC in the test-retest measures detailed in the following sections.

Study Design and Approaches

In this case series, the 1-month reproducibility (concordance) of tinnitus pitch and loudness determinations was compared between the following 2 tinnitus matching approaches: the standard clinical method (2-AFC) and a patient-centered approach (TLR). Assessments using both approaches were made for each participant during visits 1 and 2 scheduled at a 4-week interval.

In 2-AFC (based on published procedures^55,56), participants were first instructed to select as most representative of their tinnitus a pure tone, a narrowband noise, or white noise presented by the audiologist. This choice was followed by presentation of a 1-kHz pure tone or a narrowband noise¹ at a level similar to tinnitus (approximately 10-dB sensation level) and presentation of a 2-kHz sound by the audiologist, who instructed the participant to identify which best reproduced their tinnitus. For patients who chose white noise at one or the other visit (n = 5), clinicians were instructed to proceed to tinnitus matching with narrowband noise. The procedure was followed in 1-kHz steps until the participants chose a single final tinnitus frequency (after which octave confusion was tested). Then, tinnitus loudness was determined starting below threshold intensity and increasing the sound level (presented at final tinnitus frequency) in 1-dB steps. Three measures of frequency and loudness each were performed.

For TLR, a precommercialization prototype console was used, built after the research platform was developed⁶⁰ and further validated⁶⁴ in our laboratory. Under the participants’ push-button control for transitions, pure tones from 0.25 to 16.00 kHz (in steps similar to those for standard audiometry) were presented bilaterally 3 times in a pseudorandom order (preceded by 2 practice trials) via high-performance headphones (HDA-300; Sennheiser). Participants were instructed at each step to estimate how accurately the tone matched the tinnitus on a 10-point scale (0 indicates not at all like my tinnitus and 10, corresponds exactly to my tinnitus) and to set the loudness of each tone to the level of the tinnitus (decibels of sound pressure level in 1-dB increments) with the use of a potentiometer. The mean likeness ratings and decibel levels were calculated across the 3 presentations. The dominant frequencies 1, 2, and 3 of the spectrum were extracted by selecting the ones with the highest mean ratings. In 51 of 62 visits, 3 dominant frequencies were extracted. Of note, dominant frequencies 1, 2, and 3 were not necessarily in hierarchical order because several frequencies could be rated at the same highest level. Similarly, in cases of several equal frequencies, as many as 4 (10 of 62 visits) or 5 (1 of 62 visits) dominant frequencies were identified. For instance, if 1 frequency had the highest rating (eg, 10) and 3 others had the next higher identical ratings (eg, 9.6), the latter were extracted as dominant frequencies 2, 3, and 4. The corresponding decibel level was determined for each dominant frequency.

Protocol

After signing the consent form, each participant met the audiologist and underwent otoscopy, tympanometry, and pure-tone audiometry (0.25-16.00 kHz). Participants were then randomly assigned first to the 2-AFC or TLR method and then switched to the other method. A stopwatch was used to calculate the time taken to complete the tests using each method. At the end of each visit, a brief satisfaction questionnaire was filled by participants and clinicians with the use of 10-cm visual analog scales assessing ease of use, efficacy, and general appreciation of each method. Four weeks later, each participant underwent the same protocol with a different audiologist. The order of methods was counterbalanced across participants and visits; half started at visit 1 with the 2-AFC and the other half, with the TLR method, and the order was reversed for each at visit 2. The participating audiologists did not have access to records from the previous visit and started anew. Each visit lasted approximately 2 hours. The experiment took place from January 6 through March 17, 2017, at the University of Montreal in an audiology training setting with the use of equipment standardized to norms (American National Standards Institute S3.1-1999, R2008).

Data Analysis

Data were entered into an Excel spreadsheet (Microsoft Corporation) after each testing session and double-checked by research assistants. Statistical analyses were planned ahead and performed by professional statisticians using the SAS software (version 9.4; SAS Institute, Inc) under contractual agreement with an independent firm (Montreal Health Innovations Coordinating Centre, a division of the Montreal Heart Institute, Montreal, Quebec, Canada). Concordance was measured as the proportion of patients for which the variable under consideration was reproducible between visits 1 and 2.

For 2-AFC, the selected final frequency and corresponding sound intensity level were logged. For each variable, a measure of concordance was estimated according to a relatively more inclusive and a relatively more stringent criterion. For frequency, measure 1 (inclusive) was the proportion of participants for whom final frequency was concordant between visits 1 and 2 in either ear (ie, irrespective of which ear whenever both ears were tested), whereas measure 2 (stringent) was the proportion of patients for whom final frequency was concordant between visits 1 and 2 in the same ear. For sound intensity, measure 1 (inclusive) was the proportion in which differences were no greater than 10 dB between visits 1 and 2; for measure 2 (stringent), the proportion no greater than 5 dB. Sound level was considered if and only if the final frequency was concordant between visits 1 and 2.

For TLR, usually 3 dominant frequencies (see above) and their respective sound intensity level were logged. For frequency, measure 1 (inclusive) was the proportion of participants for which only 1 of 3 dominant frequencies was reproducible between visits 1 and 2, whereas measures 2 (medium stringency) and 3 (maximum stringency) were the proportions for which 2 or all 3 dominant frequencies, respectively, were reproducible between visits 1 and 2. For sound intensity, measures 1 (inclusive) and 2 (stringent) were the proportions for which differences between visits 1 and 2 were no greater than 10 dB or no greater than 5 dB, respectively, for each of the reproducible dominant frequencies. Throughout, the 95% CIs were calculated around these point estimates.

Linear mixed models with the clinician as the between-participant factor as well as method (2-AFC and TLR) and order of method (first or second) as within-participant factors were performed on the following variables: (1) time taken to complete the tests at visits 1 and 2 and (2) satisfaction ratings on questionnaires filled by participants and clinicians (1 per participant) at visits 1 and 2 (ease of use, efficacy, and general appreciation), as measured in centimeters on the 10-cm visual analog scale. Reported values are mean differences between methods (TLR minus 2-AFC) and between clinicians with 95% CIs around these point estimates. Cohen d values are also reported to describe the magnitude of the differences between methods. Cohen d of 0.20, 0.50, and 0.80 represent small, medium, and large effect sizes, respectively. Negative values represent lower values for the TLR method. Unless otherwise indicated, data are expressed as mean (SD).

Results

Audiometry

The study population included 31 adults (14 women [45%] and 17 men [55%]) with a mean age of 50.7 (13.7) years and mean educational level of 16.2 (3.7) years. Over the 0.25- to 8.00-kHz frequency spectrum assessed for standard audiometry, most participants displayed normal hearing (threshold, ≤25-dB sound level) at relatively low frequencies and typical sensorineural sloping hearing loss at higher frequencies (Figure 1). Mean hearing thresholds determined herein at 0.25 to 16.00 kHz were stable across frequencies between visits 1 and 2.

Figure 1. Hearing Thresholds for Right and Left Ears .

Hearing thresholds were assessed from 0.25 to 16.00 kHz at visits 1 (A) and 2 (B). Data are expressed as mean (SD). Blue line indicates the limits for standard normal hearing.

Most patients reliably reported the side of their tinnitus (26 of 31 [84%]) at both visits, with bilateral tinnitus being the most frequent localization (25 of 31 [81%]). Similarly, most patients were reliable in selecting the sound type presented (pure tone, narrowband noise, or white noise; 23 of 31 [74%]) at both visits, with pure tone selected by about half of them (14 of 31 [45%]).

Tinnitus Likeness Rating

For measure 1 (most inclusive criterion), most participants (26 of 31) displayed at least 1 common frequency between visits 1 and 2 (proportion, 0.84; 95% CI, 0.66-0.95). Among these 26 participants, the sound-level differences between visits 1 and 2 were no greater than 10 dB in 19 (proportion, 0.73; 95% CI, 0.52-0.88) and no greater than 5 dB in 12 (proportion, 0.46; 95% CI, 0.27-0.67) (Figure 2A).

Figure 2. Concordance Between Visits 1 and 2 for Each Method of Assessment.

Proportion of participants with concordant tinnitus pitch and loudness levels (≤5 or ≤10 dB) between visit 1 and visit 2 depends on the stringency of the concordance criterion. In the tinnitus likeness rating approach (A), elevated pitch concordance was achieved in most participants (squares indicate mean; error bars, 95% CI) under the most inclusive criterion (concordance with 1 DF), whereas the proportion decreased with medium (2 DF) and maximum (3 DF) stringency. Likewise, sound-level concordance between visit 1 and 2 was higher with the more inclusive (≤10 dB) than stringent (≤5 dB) criterion. With the standard 2-alternative forced-choice method (B), final frequency concordance was achieved in only a few participants (proportion of 0.23) with the most inclusive criterion (either ear) and even less (proportion of 0.13) under the most stringent criterion (same ear). Circles indicate mean; error bars, 95% CI. Sound-level concordances are shown only for concordant frequencies in A and B. DF indicates dominant frequency.

For measure 2 (concordance with medium stringency), at least 2 dominant frequencies were concordant between visits 1 and 2 in 15 of 31 participants (proportion, 0.48; 95% CI, 0.30-0.67). Among the 15 participants, 9 displayed sound-level differences of no greater than 10 dB (proportion, 0.60; 95% CI, 0.32-0.84), and the criterion of no greater than 5 dB was satisfied in 7 (proportion, 0.47; 95% CI, 0.21-0.73).

For measure 3 (concordance with maximum stringency), 4 of 31 participants displayed 3 concordant dominant frequencies between visits 1 and 2 (proportion, 0.13; 95% CI, 0.04-0.30). Among the 4 participants, 2 displayed sound-level differences of no greater than 10 dB (proportion, 0.50; 95% CI, 0.07-0.93), 1 of which satisfied the criterion of no greater than 5 dB (proportion, 0.25; 95% CI, 0.01-0.81).

2-AFC Method

For measure 1 (inclusive, either ear), final frequency at one ear or the other was reproducible between visits 1 and 2 in 7 of 31 participants (proportion, 0.23; 95% CI, 0.10-0.41). Among the 7 participants, sound intensity level was unavailable at visit 2 in 2 participants (could not be determined or was not deemed reliable by the clinicians). Of the 5 participants remaining, sound-level differences between visits 1 and 2 were no greater than 10 dB in 2 (proportion, 0.40; 95% CI, 0.05-0.85), 1 of whom satisfied the criterion of no greater than 5 dB (proportion, 0.20; 95% CI, 0.01-0.21) (Figure 2B).

For measure 2 (stringent, same ear), 4 of 31 participants had concordant frequency between visits 1 and 2 in the same ear (proportion, 0.13; 95% CI, 0.04-0.30). Sound intensity level was unavailable at visit 2 in 1 of the 4 participants; among the 3 remaining participants, 2 displayed sound-level differences of no greater than 10 dB (proportion, 0.67; 95% CI, 0.09-0.99), 1 of whom satisfied the criterion of no greater than 5 dB (proportion, 0.33; 95% CI, 0.01-0.91).

Duration of Testing

Mean time taken to complete the tasks did not differ between the 2 methods at visit 1 (2-AFC method, 14.9 [5.7] minutes; TRL method, 12.7 [3.9] minutes; difference, −1.6 minutes; 95% CI, −4.3 to 1.1 minutes; Cohen d = −0.30). At visit 2, mean time to complete the tasks differed between the 2 methods in relation to the 2 clinicians; precisely, time did not differ for clinicians 1 and 3, but clinician 2 took significantly more time to complete the 2-AFC task (12.1 [3.0] minutes) compared with the TRL task (9.0 [2.4] minutes; difference, −4.7 minutes; 95% CI, −7.6 to −1.7 minutes; Cohen d = −1.00) (Table 1).

Table 1. Time Taken to Complete the Tests for Each Method and Clinician for Visits 1 and 2.

Visit	Clinician No.	Participants With Tinnitus, No.	Time to Complete, Mean (SD), min		Difference (TLR Minus 2-AFC)
			2-AFC Method	TLR Method	Mean (95% CI)	Cohen d
1	1	10	15.7 (3.7)	12.0 (3.7)	−3.5 (−8.2 to 1.1)	−0.64
	2	13	15.6 (6.8)	12.0 (3.4)	−3.5 (−7.5 to 0.5)	−0.42
	3	8	12.6 (5.9)	14.5 (4.9)	2.1 (−3.1 to 7.3)	0.34
	Overall	31	14.9 (5.7)	12.7 (3.9)	−1.6 (−4.3 to 1.1)	−0.31
2	1	11	12.4 (4.4)	12.6 (4.2)	−0.5 (−3.1 to 2.1)	0.04
	2	9	12.1 (3.0)	9.0 (2.4)	−4.7 (−7.6 to −1.7)	−1.01
	3	11	9.5 (3.0)	9.8 (1.5)	1.0 (−1.6 to 3.6)	0.08
	Overall	31	11.3 (3.7)	10.6 (3.3)	−1.4 (−2.9 to 0.2)	−0.16

Abbreviations: 2-AFC, 2-alternative forced-choice; TLR, tinnitus likeness rating.

Participants’ Appreciation

Ease of use ratings were similar for methods and clinicians. For efficacy, participants’ ratings differed across clinicians at visit 1 (lower ratings for clinician 1 than clinician 2; difference, −2.1; 95% CI, −3.6 to −0.7), but not between methods. At visit 2, no effect of clinician or method occurred. The general appreciation criterion (Table 2) followed a pattern similar to that of efficacy at visit 1 with a significant main effect of clinician. Clinician 1 had lower ratings than clinician 2 (difference, −2.0; 95% CI, −3.4 to −0.5) and clinician 3 (difference, −1.6; 95% CI, −3.3 to −0.03) but no effect of method. At visit 2, there was no effect of clinician or method.

Table 2. Participants’ Ratings of Their General Appreciation at the End of Visits 1 and 2.

Visit No.	Clinician No.	Participants With Tinnitis, No.	Rating, Mean (SD)a		Difference (TLR Minus 2-AFC)
			2-AFC Method	TLR Method	Mean (95% CI)	Cohen d
1	1	10	5.3 (2.5)	5.1 (2.9)	−0.2 (−2.0 to 1.7)	−0.04
	2	13	7.2 (1.5)	7.1 (1.8)	−0.2 (−1.8 to 1.5)	−0.06
	3	8	7.0 (2.2)	6.7 (2.3)	−0.3 (−2.3 to 1.8)	−0.08
	Overall	31	6.5 (2.1)	6.4 (2.4)	−0.2 (−1.3 to 0.9)	−0.06
2	1	11	7.0 (1.8)	7.0 (2.5)	−0.3 (−2.3 to 1.7)	−0.03
	2	9	6.4 (1.9)	6.6 (2.5)	−0.2 (−2.5 to 2.0)	0.08
	3	11	6.7 (1.9)	6.6 (2.4)	0.1 (−1.8 to 2.1)	−0.03
	Overall	31	6.7 (1.8)	6.7 (2.4)	−0.1 (−1.3 to 1.1)	0.003

Abbreviations: 2-AFC, 2-alternative forced-choice; TLR, tinnitus likeness rating.

^aRatings are determined using a 10-cm visual analog scale with left-hand anchor “did not appreciate” and right-hand anchor “appreciated a lot.”

Clinicians’ Appreciation

At visit 1, ease of use ratings for each method differed across clinicians, with higher ratings for TLR than 2-AFC for clinician 1 (difference, 3.8; 95% CI, 2.6-5.0; Cohen d = 1.93) and for clinician 2 (difference, 3.6; 95% CI, 2.6-4.7; Cohen d = 2.50), but not for clinician 3 (difference, −0.1; 95% CI, −1.6 to 1.3; Cohen d = −0.05). At visit 2, a very similar pattern was found, with higher ratings for TLR than 2-AFC for clinician 1 (difference, 4.3; 95% CI, 3.2-5.4; Cohen d = 1.79) and clinician 2 (difference, 4.2; 95% CI, 2.9-5.4; Cohen d = 2.58) but not for clinician 3 (difference, 0.3; 95% CI, −0.8 to 1.4; Cohen d = 0.29).

Efficacy ratings followed the same pattern as ease of use at visit 1, with higher ratings for TLR than 2-AFC for clinician 1 (difference, 2.5; 95% CI, 1.1-4.0; Cohen d = 1.40) and for clinician 2 (difference, 4.2; 95% CI, 2.9-5.4; Cohen d = 1.65) but not for clinician 3 (difference, −0.6; 95% CI, −2.3 to 1.1; Cohen d = −0.31). The same pattern was repeated at visit 2 for clinician 1 (difference, 3.8; 95% CI, 2.7-4.9; Cohen d = 1.80) and clinician 2 (difference, 4.2; 95% CI, 2.9-5.4; Cohen d = 2.13) but not for clinician 3 (difference, −0.5; 95% CI, −1.6 to 0.6; Cohen d = −0.43).

General appreciation (Table 3) also showed differences in ratings among methods depending on clinician. At visit 1, higher ratings were found for TLR compared with 2-AFC for clinician 1 (difference, 4.2; 95% CI, 3.0-5.4; Cohen d = 3.85) and for clinician 2 (difference, 6.2; 95% CI, 5.2-7.3; Cohen d = 2.96) but not for clinician 3 (difference, −0.1; 95% CI, −1.5 to 1.3; Cohen d = −0.06). The same pattern held for visit 2, with higher ratings for TLR compared with 2-AFC for clinician 1 (difference, 4.6; 95% CI, 3.6-5.7; Cohen d = 2.15) and for clinician 2 (difference, 5.8; 95% CI, 4.6-7.0; Cohen d = 2.91) but not for clinician 3 (difference, −0.6; 95% CI, −1.6 to 0.5; Cohen d = −0.79).

Table 3. Clinicians’ Ratings of Their General Appreciation at the End of Visits 1 and 2.

Visit No.	Clinician No.	Participants With Tinnitus, No.	Rating, Mean (SD)a		Difference (TLR Minus 2-AFC)
			2-AFC Method	TLR Method	Mean (95% CI)	Cohen d
1	1	10	3.4 (1.3)	7.7 (0.7)	4.2 (3.0 to 5.4)	3.85
	2	13	3.2 (1.8)	9.4 (0.8)	6.2 (5.2 to 7.3)	2.96
	3	8	7.5 (0.9)	7.4 (1.7)	−0.1 (−1.5 to 1.3)	−0.06
	Overall	31	4.3 (2.3)	8.4 (1.4)	3.5 (2.7 to 4.2)	1.33
2	1	11	3.2 (2.0)	7.8 (0.8)	4.6 (3.6 to 5.7)	2.15
	2	9	4.1 (2.0)	9.8 (0.1)	5.8 (4.6 to 7.0)	2.91
	3	11	8.3 (1.0)	7.8 (1.0)	−0.6 (−1.6 to 0.5)	−0.79
	Overall	31	5.3 (2.8)	8.4 (1.2)	3.3 (2.6 to 3.9)	0.97

Abbreviations: 2-AFC, 2-alternative forced-choice; TLR, tinnitus likeness rating.

^aRatings are determined using a 10-cm visual analog scale with left-hand anchor “did not appreciate” and right-hand anchor “appreciated a lot.”

Discussion

In the TLR approach, participants were informed by the attending audiologist as to the unfolding of the procedure but independently determined a likeness rating for all presented frequencies, from which as many as 3 dominant frequencies were later considered in the analysis. This method yielded high individual concordance for tinnitus pitch and loudness determinations between visits 1 and 2 separated by a 1-month interval. This finding contrasts with the poor reproducibility obtained with the use of the standard clinical method (2-AFC), in which single final frequency and sound level were determined by a bracketing procedure requiring the audiologist’s active participation. Experienced clinical audiologists attended in quasi-random fashion to one or the other participant in either visit 1 or 2. The sound generator of the prototype instrument used for TLR and the audiometer used for 2-AFC were operated at a frequency setting as high as 16.00 kHz (whereas the upper limit is usually 8.00 kHz in clinical practice).

In TLR, at least 1 of 3 dominant tinnitus frequencies identified at visit 1 could be reproducibly identified at visit 2 by most participants (>80%), and 2 dominant frequencies could be reproducibly identified by half of the participants. As many as 3 dominant frequencies could be reproducibly identified by a few participants (13%); this finding was similar to the proportion of participants in whom final frequency determination at the same ear could be achieved using the 2-AFC method. Similar to pitch, tinnitus loudness was reproducibly identified at visits 1 and 2 by TLR, with the highest level of concordance being achieved under the most inclusive criterion (a single dominant frequency) and progressively less concordance under the 2– and 3–dominant frequency criteria. The monotonic inverse association between the level of concordance of visits 1 and 2 identifications (for pitch and loudness) and the stringency criterion was consistent with psychophysical functionality. No such association was found for pitch and loudness determinations made with the use of the 2-AFC method, with very low levels of concordance being achieved at either stringency level. However, time taken to complete the tests and general appreciation by the participants with tinnitus and participating audiologists were overall similar for both methods. In this regard, we note that clinicians and patients were actors under study in the comparison between the TLR and 2-AFC test-retest performance.

The TLR was first reported in a seminal study by Norena et al,⁶² which showed that tinnitus sound could be decomposed into a spectrum of contributions from multiple frequencies, herein termed dominant frequencies. The multiplicity of dominant frequency contributions mirrored the severity of hearing loss. It is now generally recognized that some degree of sensorineural impairment is a necessary condition for tinnitus, even in situations of hidden hearing loss, in which normal hearing thresholds are recorded by standard audiometry.

Such subclinical impairment may be detected by audiometry at higher than the usual clinical frequency (8.00 kHz), and, moreover, frequencies of as much as 16.00 kHz were examined herein with either approach. Therefore, tinnitus is unlikely to contain only a single, unitary pitch but rather consists of a variable spectrum of multiple sound frequencies. Accordingly, tinnitus pitch measured at a single frequency is in general highly variable across sessions for the same individual.^57,58,61,66 In light of our results, the possibility of extracting as many as 3 dominant frequencies is probably 1 reason for the higher level of concordance achieved using TLR.

Another possible reason for the higher level of concordance achieved using TLR might be that the participants determined a likeness rating for all presented frequencies independently from the attending audiologist rather than being subjected to a forced-choice bracketing procedure under the audiologist’s control in 2-AFC. Thus, in addition to the intrinsically multifrequency nature of tinnitus, TLR is a patient-centered approach in which the participant is not required to make successive decisions between alternative frequencies but rather determines a likeness rating one frequency at a time. In contrast, the 2-AFC method is audiologist driven and rests on a bracketing approach in which the patient’s successive decisions may include inaccuracies that may mislead the clinician to a remote final frequency zone, because patients are usually not familiar with the concepts of pitch and loudness. As shown herein and in other studies,^61,67 not all patients can reliably localize their tinnitus and choose the closest timber when presented with alternatives.

Limitations

This study was limited by the use of a single center. However, all 3 participating clinicians were trained in the University of Montreal audiology program, which is recognized by the Council for Accreditation of Canadian University Programs in Audiology and Speech-Language Pathology. Moreover, 2 of the clinicians are working in a private clinic, and the third has hospital service background and is now working as an academic audiologist in the University Audiology Clinic, thus reflecting a diversity of clinical experience.

Conclusions

This study was performed with the aim to reproduce clinical reality in which repeated assessments of tinnitus are not performed by the same clinicians. This study is the first, to our knowledge, to report excellent test-retest reliability of tinnitus pitch and concomitantly determined loudness at the individual level with the use of the patient-centered TLR approach. In addition, as revealed by the effect sizes presented, both methods were appreciated equally by the participants; however, time taken to complete the tests provided a small-to-medium advantage to the TLR over the 2-AFC method. In contrast, overall clinicians’ appreciation was much greater for the TLR method, although all were experienced with tinnitus matching. In conclusion, the accuracy of the assessment as well as clinicians’ appreciation were greatly improved with the TLR method without affecting patients’ satisfaction or time spent in the clinic.