Quantifying Tinnitus Perception Improvement: Deriving the Minimal Clinically Important Difference of the Minimum Masking Level
Publication: Journal of Speech, Language, and Hearing Research
Volume 68, Number 2
Pages 827-838
Abstract
Purpose:
Tools that can reliably measure changes in the perception of tinnitus following interventions are lacking. The minimum masking level, defined as the lowest level at which tinnitus is completely masked, is a candidate for quantifying changes in tinnitus perception. In this study, we aimed to determine minimal clinically important differences for minimum masking level.
Method:
A 3-month tinnitus intervention combining counseling and sound therapy was conducted in 74 participants with chronic tinnitus. Minimum masking levels were measured at baseline and 3 months. The clinical global impression was evaluated at 3 months to measure changes in participants' self-perception of tinnitus. The minimal clinically important difference of the minimum masking level was calculated using anchored-based, effect size, standard error measurement, and receiver operating characteristic curve analysis.
Results:
The minimal clinically important difference analysis of the minimum masking level yielded a −5.5 dB SL from the receiver operating characteristic curve, a −8.1 dB SL from the standard error measurement, a −9.2 dB SL from the effect size, and a −10.3 dB SL from the anchor-based analysis. Of these, the minimal clinically important difference value with optimized sensitivity (.704) and specificity (.957) was a −5.5 dB SL, determined using receiver operating characteristic analysis.
Conclusions:
The proposed minimal clinically important difference value of the minimum masking level (−5.5 dB SL) provides a good level of sensitivity and specificity. Therefore, the minimum masking level may be an alternative for measuring changes in tinnitus perception.
Supplemental Material:
Generally, tinnitus is defined as the perception of sounds in the absence of any external sound stimuli (Henry, 2016). An estimated 10%–15% of the adult population experiences tinnitus, with approximately 2% experiencing severe effects (Langguth et al., 2013). Tinnitus often leads to concomitant symptoms such as depression, anxiety, sleep disorders, and cognitive impairment, all of which can significantly diminish an individual's quality of life (Langguth et al., 2011; Maes et al., 2013; Wang et al., 2020).
Measuring tinnitus severity is an efficient method for determining the usefulness of tinnitus treatment or rehabilitation (Mancini et al., 2020). In general, tinnitus measurements are categorized into two aspects: tinnitus reaction and tinnitus perception. Tinnitus reaction (Henry, 2016) pertains to quantifying an individual's response to the difficulties experienced in their daily life owing to tinnitus. For example, emotional distress due to tinnitus can be measured using a scale ranging from 0 (no emotional distress) to 100 (complete distress caused by tinnitus) on a questionnaire (Henry, 2016; Tyler & Conrad-Armes, 1983b; Tyler et al., 2014). A decrease in the scale score from 90 before treatment to 50 after treatment can be interpreted as a 40-point improvement in the individual's emotional distress related to tinnitus reaction. Currently, several questionnaires and scales have been developed and validated for measuring tinnitus reactions (Henry, 2016; I.-K. Jin & Tyler, 2022).
The minimal clinically important difference (MCID) is defined as the smallest difference in score change on a measurement tool that can be considered a valid effect when evaluating the usefulness of rehabilitation (De Vet et al., 2006). Measurements of tinnitus reaction for which MCID values have been identified include the Tinnitus Handicap Questionnaire, the Tinnitus Handicap Inventory, the Tinnitus Primary Function Questionnaire, and the Visual Analog Scale (VAS) for Loudness (VAS-L) and Annoyance (VAS-A), with reported MCIDs of 21 (Newman et al., 1995), 7 (Zeman et al., 2011), 13 (Tyler et al., 2014), and 10–15 points (Adamchic et al., 2012), respectively. Therefore, comparing measurements before and after treatment using a questionnaire from which the MCID is derived can effectively help determine the minimal valid effect of the treatment for the individual.
In addition to measuring tinnitus reaction, tinnitus perception is another crucial aspect of tinnitus assessment (Henry, 2016). This involves identifying the auditory perceptual characteristics of tinnitus (e.g., loudness and pitch). For example, tinnitus perception can be measured by presenting various levels of stimuli to individuals with tinnitus using audiometric equipment and asking them to indicate the external stimulus level that matches their tinnitus perception. When an individual's tinnitus perception decreases from 40 to 20 dB SL after tinnitus treatment, this indicates that the treatment reduced the individual's perception of tinnitus by 20 dB.
Measuring both tinnitus reaction and perception using a single assessment tool is difficult. For example, when a questionnaire score assessing the emotional distress caused by tinnitus shows significant improvement from pre- to posttreatment, ascertaining whether the individual's tinnitus perception has also improved is challenging. Conversely, when a perception test measuring tinnitus loudness shows a significant reduction, conclusive statements about improvements in the individual's reaction to tinnitus, such as concentration and sleep, cannot be made. Therefore, the reaction and perception of tinnitus represent different domains, and appropriate tools should be applied to measure each domain when assessing tinnitus.
While MCIDs have been derived for several tools that measure tinnitus reaction, the MCIDs of tools that measure tinnitus perception, including tinnitus loudness, remain unclear (I.-K. Jin & Tyler, 2022; Langguth & De Ridder, 2023). A scoping review for tinnitus outcome measurement instruments reported that measures related to tinnitus perception, including loudness matching, do not demonstrate sufficient reliability and validity to be used as outcome measures (Langguth & De Ridder, 2023). In addition, several studies have attempted to objectively measure tinnitus using imaging and electrophysiology, but limitations such as complex neural mechanisms, inconsistencies across studies, and small sample sizes have hindered the development of a definitive measurement method (Jackson et al., 2019; I.-K. Jin & Tyler, 2022; Tunkel et al., 2014). Consequently, useful tools to measure improvements in an individual's tinnitus perception before and after tinnitus rehabilitation are lacking. For example, when attempting to measure the usefulness of a treatment aimed at reducing the perception of tinnitus, the lack of a tool to beneficially measure changes in tinnitus perception hinders the identification of a valid effect of the treatment.
Tinnitus can be improved using a variety of treatments, and individual differences in the usefulness of tinnitus treatment exist, even within the same intervention (Tunkel et al., 2014; Ye et al., 2022). Responses of individual patients may not be accurately reflected when judging the benefit of therapies based solely on average results. Average values can make it easier to ignore significant treatment benefits in certain patients by simplifying the complex responses of various patient populations (de Vet et al., 2006). Consequently, tinnitus improvements should be assessed on an individual basis. MCID-validated measures can be used to assess an individual's tinnitus improvement resulting from a tinnitus intervention.
The minimum masking level (MML) is defined as the lowest level at which tinnitus is completely masked (Johnson & Fenwick, 1984). MML is a potential candidate for quantifying changes in tinnitus perception (Mancini et al., 2020). Jastreboff et al. (1994) reported that MML can measure the neural networks involved in tinnitus; because increased activity in these networks is associated with increased tinnitus perception, alterations in MML can be associated with changes in tinnitus perception. Moreover, Mancini et al. (2020) reported that MML generally has high test–retest reliability, indicating its robust measurement consistency. Although the MCID of MML has not been reported as a tool for measuring perception changes in tinnitus, MML, given its theoretical foundations and measurement reliability, appears to be a competitive candidate for measuring tinnitus perception.
The purpose of this study was to derive the MCID of MML. The MCID of MML was derived using anchor-based and distribution-based analyses using MML results before and after tinnitus rehabilitation. The optimized MCID value was then derived from the MCID values calculated by different analytical methods.
Method
Participants
This study received approval from the institutional review board of Hallym University (HIRB-2023-033), and each participant was provided with a written explanation detailing the study aims, protocol, and procedures. Each participant provided written informed consent before participation in the study.
All participants were recruited through a tinnitus community website (https://cafe.naver.com/onquest) and screened using specific inclusion and exclusion criteria. The participant selection process was conducted via telephone interviews. The website used for recruitment is a community space where approximately 60,000 people with tinnitus and their families can share information about tinnitus. A recruitment announcement was uploaded to the bulletin board with formal permission from the website administrator, and people interested in participating in the study contacted the study hosts using the contact information (e-mail and phone) provided in the announcement. The inclusion and exclusion criteria were determined by referring to previous studies (Jastreboff & Jastreboff, 2000; I.-K. Jin et al., 2021). Participants had to meet the following inclusion criteria: They must have had tinnitus for at least 6 months, experience discomfort or difficulty due to tinnitus, and have an average pure-tone threshold at 0.5 kHz, 1 kHz, and 2 kHz equal to or less than 40 dB HL. We included only participants who agreed to participate in all procedures. We excluded individuals who had any psychiatric disorders or otologic conditions such as cupulolithiasis or Meniere's disease. Furthermore, those who were already involved in tinnitus-related litigation were excluded, described as study participants who, at the time of the study, were actively involved in legal disputes or claims pertaining to their tinnitus condition.
The required minimum sample size was calculated using the MedCalc statistical software program (MedCalc Software Ltd.), with parameters set as follows: Type 1 error, .05; Type 2 error, .20; null hypothesis value, .5; and area under the curve (AUC), .7. Type 1 error (α = .05) represents the probability of rejecting a true null hypothesis (false positives). Type 2 error (β = .20) corresponds to the probability of failing to reject a false null hypothesis (false negatives). A value of .20 indicates a power of 80% (1 − β). The null hypothesis value (.5) is the AUC under the null hypothesis, representing the scenario where the test has no discriminative ability (i.e., performs no better than random chance). The assumed AUC of .7 is considered “good” predictive performance. We based the sample size calculation parameters on reference values used in previous studies that utilized AUCs (Forman & Cohen, 2005; Simundic, 2012). The calculated sample size was 72.
The Consolidated Standards of Reporting Trials flow diagram is depicted in Figure 1A. The study participants were recruited from July 5 to July 25, 2023. Of the 114 participants, 38 were excluded during the prescreening and baseline process for the following reasons: (a) self-exclusion owing to difficulty in visiting the institution (n = 25); (b) nonresponsiveness to visitation requests (n = 7), judged as declining to participate; (c) average pure-tone threshold exceeded 40 dB HL (n = 4); and (d) presence of otologic complications (n = 2). Finally, 76 participants were included. Among the participants, two were unable to participate in the final measurement (T1) due to difficulties in visiting the institution. Therefore, we obtained data from 74 participants at the baseline and final visits, exceeding the minimum of 72 data points. The final (T1) visit was performed between October 21 and November 11, 2023.
Intervention
This study included a 3-month rehabilitation program, combining counseling and sound therapy, for participants experiencing discomfort and pain due to persistent tinnitus. The counseling component used a set of 100 tinnitus education videos created in a question-and-answer format (T.-J. Jin et al., 2023). These educational materials covered various aspects of tinnitus, including its causes, symptoms, diagnosis, treatment, and management.
In the case of sound therapy, a custom-developed sound therapy application was used (I.-K. Jin et al., 2021, 2022). This application used broadband noise (BBN) in the range of 100–22050 Hz as a sound therapy sound stimulus. Previous studies have recommended that sound therapy should be practiced for at least 2 hr per day for at least 3 months to confirm improvements in tinnitus resulting from neuroplasticity changes caused by sound therapy (I.-K. Jin et al., 2022; Lee & Jin, 2023; Mahboubi et al., 2017). Therefore, in this study, we recommended that all participants perform sound therapy for at least 2 hr per day throughout the study. The loudness of the sound therapy stimulus was set at the mixing point, a level at which the participant's tinnitus and the sound therapy source were heard simultaneously (Jastreboff & Hazell, 1998; I.K. Jin et al., 2022). Each participant was instructed to listen to the sound therapy stimulus for 2 hr per day in a quiet space and was advised against exposure to other loud sounds (such as music) during sound therapy.
Study Design
This study was conducted to measure changes in tinnitus loudness and to calculate and compare MCID values using various analyses. All procedures and tests were performed by three certified audiologists. The study protocol is shown in Figure 1B. During the prescreening phase, researchers conducted telephone interviews with interested participants to explain the background and purpose of the study and to assess the inclusion and exclusion criteria. After prescreening, participants who met the inclusion criteria were included in the baseline assessment, whereas those who did not meet the criteria were excluded.
Participants who met the inclusion criteria during prescreening visited the Audiology and Speech Pathology Research Institute at Hallym University for baseline (T0) measurements. During this visit, researchers administered the Tinnitus Intake Questionnaire (TIQ), pure-tone audiometry, tinnitogram, and MML to directly assess the inclusion and exclusion criteria that could not be determined during the phone interview. After providing a detailed explanation of the study, each participant signed a written informed consent form. Subsequently, each participant completed the TIQ—a questionnaire that provides basic demographic (age, sex) and tinnitus-related background information (duration of tinnitus, characteristics of tinnitus, etc.) for study participants. The TIQ used in the current study comprised a Korean translation of survey items described in a study by Stouffer and Tyler (1990). After completing the TIQ, pure-tone audiometry, tinnitogram, and MML were performed using a GSI AudioStar Pro audiometer (Grason-Stadler Inc.) and Sennheiser HDA-200 headphones (Sennheiser electronic GmbH & Co. KG). Pure-tone audiometry was performed to determine the average value in pure-tone thresholds at 500, 1000, and 2000 Hz for each participant. We used the tinnitogram, which consisted of pitch and loudness matching tests, to determine whether participants had tinnitus and to characterize their tinnitus. In the pitch-matching test, two pulsed tones with different octave frequencies were presented alternately, and the participant was asked to select the tone closer to their tinnitus (Tyler & Conrad-Armes, 1983a). The procedure was repeated until the participant identified the frequency closest to their tinnitus. In the loudness matching test, participants listened to a pulsed tone at their specific pitch-matched frequency, starting from low intensity and progressively increasing until they perceived the tone intensity as similar to their tinnitus intensity (Tyler et al., 2007). The result of the loudness matching test was determined as the intensity that the participant reported as closest to their tinnitus loudness.
Subsequently, MML was performed to determine the lowest level that completely masks the individual's tinnitus. The test stimulus for MML measurements was BBN with a frequency range of 176.8–11313.7 Hz, generated using an audio editing program, Adobe Audition (Adobe Inc.). The MML intensity was expressed in dB MML, which represents the minimum noise level required to mask tinnitus. In addition, the difference between the baseline MML values and those measured at the final visit for each participant was presented as dB SL.
Currently, no universally accepted standard or norm exists for measuring MML, and we used a measurement method based on repeated measures, as previously suggested (Fournier et al., 2019; Vernon & Meikle, 2003). Figure 2 illustrates an example of the MML measurement procedure. The MML measurement procedure was as follows (Fournier et al., 2019):
1.
The initial measurement began 10 dB below the participant's hearing threshold at the pitch-matching frequency, increasing in 5-dB increments until the tinnitus was reported as masked.
2.
The subsequent two measurements started 10 dB below the previous masking point, rising in 2-dB steps until the tinnitus was masked again. Each stimulus lasted approximately 1–2 s. The final MML was the average of the last two measurements.
According to Figure 2, the first measurement (M1) resulted in 40 dB HL, whereas the second (M2) and third (M3) measurements resulted in 38 dB HL. Therefore, the final MML value was 38 dB MML ([38 + 38] ÷ 2), which is the average of the second and third measurements recorded 2 dB apart.
At the final visit (T1), participants underwent MML and clinical global impression (CGI) assessments. The CGI is a widely used clinical research tool for assessing the severity of a patient's condition, responses to treatment, and overall improvements over time (Busner & Targum, 2007). It is appreciated for its ease of use and capacity to offer prompt, thorough assessment (Wheaton & Pope, 2010). Because of its dependability and simplicity, CGI has been used in many studies of neurological, psychiatric, and audiological conditions (Adamchic et al., 2012; Jaeschke et al., 1989; Wheaton & Pope, 2010). The CGI was specifically selected for this study because of its capacity to record changes in symptom severity as perceived by patients consistently, which is essential for assessing the usefulness of rehabilitation. Participants were required to categorically rate their tinnitus perception compared with that at baseline. The rating scale included five options: 1 (significantly decreased), 2 (somewhat decreased), 3 (no change), 4 (somewhat increased), and 5 (significantly increased). Each visit lasted 1–1.5 hr per participant. The TIQ and CGI were completed in a consultation room, whereas pure-tone audiometry, tinnitogram, and MML assessments were performed in a double-walled sound booth.
The longitudinal methodology used in this study was chosen to track changes in tinnitus perception over a 12-week period while patients completed therapy. This longitudinal approach allows for the direct observation of changes in each participant's tinnitus perception from baseline (T0) to the final assessment (T1), providing a clear evaluation of the treatment's usefulness. Because the methodology allowed us to compute and compare MCID values based on observed changes over time rather than depending solely on a single point-in-time measurement, it was well suited to the goals of the study. The MCID of MML was determined using the measurements of MML recorded at both the baseline (T0) and the final visit (T1), as well as the CGI assessed at the final visit (T1).
Data Analysis
In this study, we employed an anchor-based analysis, two distribution-based analyses, and an integrated analysis to derive MCID values. First, the anchor-based analysis uses an external indicator, known as an “anchor,” to determine changes that patients or clinicians consider significant improvements or deteriorations. Therefore, the anchor-based analysis determines MCID by comparing the numerical change in the measurement tool (i.e., MML) to a minimum threshold change defined by the anchor (i.e., CGI). One advantage of an anchor-based analysis is that changes in outcome measure scores can directly explain changes in patient status (De Vet et al., 2007). Typically, the CGI is used to define MCIDs as an anchor for identifying changes in treatment usefulness in a patient. This approach helps clinicians understand how a patient's condition has changed rather than simply interpreting numerical changes (Crosby et al., 2003). However, anchor-based analyses have been criticized for the effect of “recall bias.” Recall bias is a systematic error caused by differences in the completeness of participants' recollections of past experiences, which occurs when a participant remembers recent events more clearly and has a less accurate memory of more distant past events (Norman et al., 1997; Wright et al., 2012). For example, for anchors based on questioning patients about improvement after an intervention, their current state tends to influence recollection of the past. This can lead to inaccurate estimates by overestimating or underestimating reported changes in current or baseline conditions (McGlothlin & Lewis, 2014).
To facilitate data analysis, the participants were classified into two groups based on the CGI results as the anchor. Participants who responded with “1 (significantly decreased)” or “2 (somewhat decreased)” were grouped into the category “CGI(1),” representing a reported reduction in tinnitus perception. Furthermore, those who responded with “3 (no change),” “4 (somewhat increased),” or “5 (significantly increased)” were classified into the group “CGI(0),” which represents no reported reduction in tinnitus perception. At each visit before and after tinnitus rehabilitation, 27 data points were collected for CGI(1) and 47 were collected for CGI(0). The MCID value derived from the anchor-based analysis was selected as the average MML change value for the group classified as CGI(1).
Second, distribution-based analyses help compare a prespecified threshold value of uncertainty using statistical criteria to the difference in a scale-based outcome measure. Prespecified threshold values of uncertainty include the standard error of measurement (SEM), standard deviation (SD), and effect size (ES; Copay et al., 2007; Ousmen et al., 2018). The advantage of distribution-based analyses is that, unlike anchor-based analyses, they analyze the distribution of data points without relying on specific anchor points or references, thereby avoiding data bias due to the anchors. Additionally, distribution-based analyses can account for changes in random variation caused by measurement errors (Crosby et al., 2003). SEM, as one of the distribution-based analyses, is particularly well suited for this purpose because it determines whether the observed change is within the range of random variation and provides confidence intervals for evaluating measurement precision (Crosby et al., 2003). However, a major disadvantage of distribution-based analyses is that they do not address the patient's perspective of clinically important change, which is distinctly different from statistical significance (Jayadevappa et al., 2012; Wright et al., 2012).
The ES is a value that measures the strength of the relationship between two variables in a population, providing a quantitative measure of rehabilitation usefulness. Cohen's d was used to calculate the MCID value of MML from the ES analysis (Adamchic et al., 2012). Cohen's d for MML was calculated using the following formula:
(1)
where M1 is the mean value of MML at baseline, M2 is the mean value of MML at the final visit, and SDpooled is the pooled standard deviation. The value of Cohen's d can be interpreted as follows: an absolute ES (Cohen's d) of < 0.5 is considered a “small effect,” a Cohen's d of ≥ 0.8 is considered a “large effect,” and Cohen's d values between 0.5 and 0.8 are considered “medium effect” (Carson, 2012). For the SDpooled value, the Pooled Standard Deviation Calculator (https://www.statology.org/pooled-standard-deviation-calculator/) was used. The MCID value was determined based on Cohen's d = 0.5, which is the universal standard for determining MCIDs (Norman et al., 2003).
Another distribution-based analysis, SEM, is an indicator of the variability in the measurements around the “true” score when the same item is measured repeatedly. We calculated the intraclass correlation coefficient (ICC) using SPSS 26.0 (IBM Corp.) to analyze the test–retest reliability of each measure over a 3-month period. The SEM for MML was calculated using the following formula:where SD is the standard deviation of MML used to determine r and r is the test–retest reliability estimate.
(2)
Third, receiver operating characteristic (ROC) curve analysis combines the anchor-based and distribution-based analyses to compensate for the limitations of the two analyses mentioned above (Crosby et al., 2003). ROC analysis addresses the limitations of subjective anchor assessments by evaluating performance across different threshold ranges without relying on a single reference point, thereby identifying optimal thresholds that distinguish clinically meaningful changes. Moreover, ROC analysis complements the limitations of clinical interpretation of distributions by using an anchor to interpret the distribution in the context of clinically meaningful changes, providing a more detailed understanding of data. ROC curves are graphical tools used to evaluate the performance of binary classifier models, visually showing the performance of classifier models at all classification thresholds, with 1 − specificity (SP) on the x-axis and sensitivity (SE) on the y-axis. ROC analysis helped categorize participants into two groups—CGI(1) and CGI(0)—using an anchor to determine the distribution of score changes on the measurement tool.
To estimate the MCID of MML using ROC analysis, one of the normalization methods, the z score, was used. The z score converts data with different distributions into a standard normal distribution. Normalization was performed using the following formula:where x is the test variable, μ is the mean of the variable, and σ is the standard deviation of the variable.
(3)
The normalized MML change values and their corresponding CGI generate a range of cutoff points with the potential for binary classification through ROC curve analysis. Each point within the range has SE and SP, and this information is crucial for optimal cutoff point selection. Youden's index, a method used to evaluate the performance of binary classifiers such as ROC, considers both SE and SP to derive the optimal cutoff point. Youden's index was calculated using the following formula:
(4)
Since the MML change at the point with the highest Youden's index yields values between −1 and 1 in the ROC curve analysis with normalized variables, converting normalized values to MCID values in MML (dB MML) is necessary. Therefore, the MCID of MML with normalized variable (x) was recalculated by multiplying the standard deviation (σ) of the data and adding the mean (μ) of the data.
Results
Demographic and Tinnitus Characteristics of the Participants
The demographic and tinnitus characteristics of the study participants are presented in Table 1. A total of 74 participants with tinnitus, comprising 40 men and 34 women, were identified. The mean age of all participants was 44.64 years (SD = 13.91, range: 19–69), and the mean duration of tinnitus was 37.16 months (SD = 34.13, range: 6–156). The mean pure-tone average (average hearing thresholds at 500, 1000, and 2000 Hz) was 13.00 dB HL (SD = 13.70, range: −3.3 to 40.0) for the right ear and 14.57 dB HL (SD = 12.57, range: −3.3 to 40.0) for the left ear. The tinnitogram results showed that the mean tinnitus frequency of the participants was 5716.22 Hz (SD = 2855.27, range: 250–8000) and the mean tinnitus loudness was 9.83 dB SL (SD = 9.81, range: −2 to 40). Among participants with unilateral tinnitus, 20 (27.03%) had tinnitus on the right side and 21 (28.38%) had tinnitus on the left side. Among participants with bilateral tinnitus (33 participants, 44.59%), the average value was calculated based on the ear with the more severe or louder tinnitus.
| Characteristics | Total (N = 74) |
|---|---|
| Sex (number of participants, %) | |
| Male | 40 (54.05%) |
| Female | 34 (45.95%) |
| Mean age (years) | 44.64 (13.91, 19–69) |
| Tinnitus duration (in months) | 37.16 (34.13, 6–156) |
| PTA (dB HL) | |
| Right ear | 13.00 (13.70, −3.3 to 40.0) |
| Left ear | 14.57 (12.57, −3.3 to 40.0) |
| Laterality (number of participants, %) | |
| Right | 20 (27.03%) |
| Left | 21 (28.38%) |
| Bilateral | 33 (44.59%) |
| Tinnitus frequency (Hz) | 5716.22 (2855.27, 250–8000) |
| Tinnitus loudness level (dB SL) | 9.83 (9.81, −2 to 40) |
Note. Data are presented as mean (standard deviation, range).
MML and CGI Results for Each Participant
MML at baseline (T0) and at the final visit (T1) were measured to determine changes in MML in individuals participating in tinnitus rehabilitation. Additionally, CGI was performed at the final visit to assess the extent of change in tinnitus perception by the participants. The MML and CGI results for each participant are presented in Supplemental Material S1. Following tinnitus rehabilitation (visit T1), 27 individuals were classified as reporting a reduction in tinnitus perception, CGI(1), and 47 individuals were classified as reporting no reduction in tinnitus perception, CGI(0).
MCID Based on Anchor-Based Analysis
Box plots of MML changes categorized by the CGI are shown in Figure 3. The mean of the MML difference was −10.3 dB SL (SD = 18.6) for the CGI(1) group (27 data points) and 3.4 dB SL (SD = 8.9) for the CGI(0) group (47 data points). Therefore, the MCID value derived from the anchor-based analysis was selected to be −10.3 dB SL (SD = 18.6), which is the average MML change value for the group classified as CGI(1).
MCID Based on ES Analysis
The mean MML values measured at baseline (T0) and at the final visit (T1) were 42.76 (SD = 18.86, N = 74) and 41.18 (SD = 17.89, N = 74) dB MML, respectively. The SDpooled value was calculated to be 18.4. To derive a universal MCID estimate, Cohen's d = 0.5, considered a “medium effect,” was proposed for this study. Therefore, the MCID value of MML estimated using the ES analysis was approximately −9.2 dB SL (SDpooled = 18.4, Cohen's d = 0.5).
MCID Based on SEM Analysis
The SD value of the MML (dB MML) used to determine r was 18.40 (N = 74). The ICC value was .805 (95% confidence interval [CI] [.691, .877]), which is considered “excellent” for ICC values of ≥ .75 (Cicchetti, 1994). The MCID value of MML estimated using SEM analysis was −8.1 dB SL (SD = 18.40, r = .805).
MCID Based on ROC Curve Analysis
ROC was used to derive the MCID of MML, as shown in Figure 4. The ROC curve analysis results, based on normalized MML change values and their corresponding CGI, yielded the largest value of Youden's index of .661 (SE = .704, SP = .957). At this point, the AUC, an indicator of the overall accuracy of a binary classifier, was .781 (CI [.644, .917]), which is considered “good” (≥ .7), indicating good classification performance. The classification threshold value at that point was −0.265, leading to an estimated MCID value for MML of approximately −5.5 dB SL (x = −0.265, σ = 14.848, μ = −1.595) according to the ROC curve.
This analysis demonstrated that the ROC curve was useful in identifying the optimal threshold for distinguishing clinically meaningful changes in MML. The AUC of .781 suggests that the test provides a reliable balance between SE and SP, which is crucial for clinical decision making. Hearing health care professionals can interpret this threshold value as the point where the model most accurately differentiates between patients who have experienced a significant improvement in tinnitus perception and those who have not. Thus, the ROC curve provides not only a statistical measure of accuracy but also a practical tool for setting clinically meaningful cutoffs in treatment evaluation.
Discussion
The purpose of this study was to calculate the MCID of MML. MML was measured at baseline and at 3 months (final visit) in participants with chronic tinnitus who received a tinnitus intervention. Additionally, the CGI was used at the final visit to measure changes in participants' self-perception of tinnitus.
To derive the optimal MCID value for MML in this study, different analytical methods were used. MCID values were derived from anchor-based, distribution-based (ES and SEM), and ROC curve analyses. The MCID values of MML varied from −10.3 to −5.5 dB SL, depending on the analysis method. To assess the relative adequacy of the derived MCID values, Youden's index was calculated using the MCID values derived from each analysis method with the CGI response. The results of this comparison are presented in Table 2. Youden's index, a technique employed to assess the usefulness of binary classifiers, encompasses both SE and SP. The Youden's index values derived from ROC curve, SEM, ES, and anchor-based analyses were calculated to be .661 (SE = .704, SP = .957), .476 (SE = .519, SP = .957), .476 (SE = .519, SP = .957), and .519 (SE = .519, SP = 1.000), respectively. Notably, the ROC curve analysis yielded the highest Youden's index (.661) among the analytical methods employed to derive the MCID value of MML. This result implies that −5.5 dB SL is the most appropriate MCID value for MML change in terms of SE and SP.
| Method (MCID value of MML, dB SL) | Accuracy | ||
|---|---|---|---|
| SE (%) | SP (%) | Youden's index | |
| ROC (−5.5) | 70.4 | 95.7 | .661 |
| SEM (−8.1) | 51.9 | 95.7 | .476 |
| ES (−9.2) | 51.9 | 95.7 | .476 |
| Anchor-based (−10.3) | 51.9 | 100.0 | .519 |
Note. SE = sensitivity; SP = specificity; ROC = receiver operating characteristic; SEM = standard error of measurement; ES = effect size.
Adamchic et al. (2012) conducted a similar study on tinnitus evaluation and derived the MCID for a VAS for loudness and annoyance using anchor-based, distribution-based, and ROC curve analyses. The MCID of the VAS was 10–15 points, and it was 10 points when derived using the ROC curve (Loudness: AUC = .74, SE = 73.7%, SP = 66.2%; Annoyance: AUC = .79, SE = 70.1%, SP = 78.5%), which suggests an appropriate value for both VAS-rated loudness and annoyance. The results of the present study are similar to those of Adamchic et al. (2012), who emphasized the importance of developing tools to measure changes in tinnitus loudness and found that MCID values derived through ROC curves are reliable among various methods. While both studies yielded reliable MCID values to measure tinnitus (VAS = 10–15 points, MML = −5.5 dB SL), there is a major difference: Adamchic et al. (2012) used a VAS to measure reactions to changes in tinnitus loudness, whereas MML was utilized in the present study to measure the subjective perception of tinnitus. As reaction and perception of tinnitus are different domains (I.-K. Jin & Tyler, 2022), the combined use of the MCIDs based on VAS and MML assessments may be useful in measuring changes in reaction and perception of tinnitus in clinical practice.
The present study has certain limitations. First, the MCID for MML was derived using an intervention that combined counseling and sound therapy. This indicates limited information on the consistency of MCID values in MML across different tinnitus interventions and highlights the need for further research to determine whether the values derived in this study can be applied to other tinnitus interventions. Second, this study was conducted based on MML measurements performed by hearing health care professionals in a face-to-face setting alongside research participants. With recent technological advancements, individuals with tinnitus can now self-assess improvements using smartphone applications or other tools at home. Therefore, future studies should be conducted to evaluate the applicability of a software-based MML self-measurement system, incorporating the MCID values, to assess its feasibility and usefulness.
Improvements in tinnitus can be divided into reaction and perception aspects. Various questionnaires have been developed for tinnitus reaction, some with proposed MCID values to provide a basis for determining whether an individual is experiencing a valid and relevant improvement in tinnitus (Fackrell et al., 2016; Newman et al., 1995; Tyler et al., 2014). Nevertheless, no clear proposal has been made regarding MCID values for the perception domain of tinnitus (I.-K. Jin & Tyler, 2022; Langguth & De Ridder, 2023). In conclusion, the MCID value of MML proposed in this study (−5.5 dB SL) provided a good level of SE (.704) and SP (.957), suggesting that this MCID value could be a useful tool for measuring changes in tinnitus perception in response to tinnitus treatment in clinical practice.
Data Availability Statement
The study data are available from the corresponding author upon reasonable request.
Supplemental Material on Figshare
Acknowledgments
This work was supported by the National Research Foundation of Korea grant funded by the Korea government (MSIT; No. 2023R1A2C1002929). We would like to express our deepest gratitude to Richard Tyler (University of Iowa) for his advice on the study design and applicability of the present study.
References
Adamchic, I., Langguth, B., Hauptmann, C., & Tass, P. A. (2012). Psychometric evaluation of visual analog scale for the assessment of chronic tinnitus. American Journal of Audiology, 21(2), 215–225.
Busner, J., & Targum, S. D. (2007). The clinical global impressions scale: Applying a research tool in clinical practice. Psychiatry, 4(7), 28–37.
Carson, C. (2012). The effective use of effect size indices in institutional research. In Proceedings of the 2004 American Society for Engineering Education Annual Conference & Exposition (Vol. 41, pp. 41–48).
Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological Assessment, 6(4), 284–290.
Copay, A. G., Subach, B. R., Glassman, S. D., Polly, D. W., Jr., & Schuler, T. C. (2007). Understanding the minimum clinically important difference: A review of concepts and methods. Spine Journal, 7(5), 541–546.
Crosby, R. D., Kolotkin, R. L., & Williams, G. R. (2003). Defining clinically meaningful change in health-related quality of life. Journal of Clinical Epidemiology, 56(5), 395–407.
De Vet, H. C. W., Beckerman, H., Terwee, C. B., Terluin, B., & Bouter, L. M. (2006). Definition of clinical differences. Journal of Rheumatology, 33(2), Article 434. https://www.jrheum.org/content/33/2/434
De Vet, H. C. W., Ostelo, R. W. J. G., Terwee, C. B., van der Roer, N., Knol, D. L., Beckerman, H., Boers, M., & Bouter, L. M. (2007). Minimally important change determined by a visual method integrating an anchor-based and a distribution-based approach. Quality of Life Research, 16(1), 131–142.
Fackrell, K., Hall, D. A., Barry, J. G., & Hoare, D. J. (2016). Psychometric properties of the Tinnitus Functional Index (TFI): Assessment in a UK research volunteer population. Hearing Research, 335, 220–235.
Forman, G., & Cohen, I. (2005). Beware the null hypothesis: Critical value tables for evaluating classifiers. In J. Gama, R. Camacho, P. B. Brazdil, A. M. Jorge, & L. Torgo (Eds.), Machine Learning: ECML 2005 (pp. 133–145). Springer Nature.
Fournier, P., Wrzosek, M., Paolino, M., Paolino, F., Quemar, A., & Noreña, A. J. (2019). Comparing tinnitus tuning curves and psychoacoustic tuning curves. Trends in Hearing, 23.
Henry, J. A. (2016). “Measurement” of tinnitus. Otology and Neurotology, 37(8), e276–e285.
Jackson, R., Vijendren, A., & Phillips, J. (2019). Objective measures of tinnitus: A systematic review. Otology and Neurotology, 40(2), 154–163.
Jaeschke, R., Singer, J., & Guyatt, G. H. (1989). Measurement of health status: Ascertaining the minimal clinically important difference. Controlled Clinical Trials, 10(4), 407–415.
Jastreboff, P. J., Hazell, J. W., & Graham, R. L. (1994). Neurophysiological model of tinnitus: Dependence of the minimal masking level on treatment outcome. Hearing Research, 80(2), 216–232.
Jastreboff, P. J., & Hazell, J. W. P. (1998). Treatment of tinnitus based on a neurophysiological model. In J. A. Vernon (Ed.), Tinnitus: Treatment and relief (pp. 201–217). Allyn & Bacon.
Jastreboff, P. J., & Jastreboff, M. M. (2000). Tinnitus Retraining Therapy (TRT) as a method for treatment of tinnitus and hyperacusis patients. Journal of the American Academy of Audiology, 11(3), 162–177.
Jayadevappa, R., Malkowicz, S. B., Wittink, M., Wein, A. J., & Chhatre, S. (2012). Comparison of distribution- and anchor-based approaches to infer changes in health-related quality of life of prostate cancer survivors. Health Services Research, 47(5), 1902–1925.
Jin, I.-K., Choi, S.-J., Ku, M., Sim, Y., & Lee, T. (2022). The impact of daily hours of sound therapy on tinnitus relief for people with chronic tinnitus: A randomized controlled study. Journal of Speech, Language, and Hearing Research, 65(8), 3079–3099.
Jin, I.-K., Choi, S.-J., Yoo, J., Jeong, S., Heo, S., & Oh, H. (2021). Effects of tinnitus sound therapy determined using subjective measurements. Journal of the American Academy of Audiology, 32(4), 212–218.
Jin, I.-K., & Tyler, R. S. (2022). Measuring tinnitus in pharmaceutical clinical trials. Journal of the Acoustical Society of America, 152(6), 3843–3849.
Jin, T.-J., Sim, Y., & Jin, I.-K. (2023). Investigation of tinnitus counseling requirements and derivation of counseling topics through focus group interviews with tinnitus patients. Audiology and Speech Research, 19(3), 179–189.
Johnson, R. M., & Fenwick, J. (1984). Masking levels (minimum masking levels) and tinnitus frequency. Journal of Laryngology and Otology, 98(Suppl. 9), 63–66.
Langguth, B., & De Ridder, D. (2023). Minimal clinically important difference of tinnitus outcome measurement instruments—A scoping review. Journal of Clinical Medicine, 12(22), Article 7117.
Langguth, B., Kreuzer, P. M., Kleinjung, T., & De Ridder, D. (2013). Tinnitus: Causes and clinical management. Lancet Neurology, 12(9), 920–930.
Langguth, B., Landgrebe, M., Kleinjung, T., Sand, G. P., & Hajak, G. (2011). Tinnitus and depression. World Journal of Biological Psychiatry, 12(7), 489–500.
Lee, T., & Jin, I.-K. (2023). Tinnitus relief effect according to the stimulus listening method of sound therapy. Audiology and Speech Research, 19(4), 249–260.
Maes, I. H. L., Cima, R. F. F., Vlaeyen, J. W., Anteunis, L. J. C., & Joore, M. A. (2013). Tinnitus: A cost study. Ear and Hearing, 34(4), 508–514.
Mahboubi, H., Haidar, Y. M., Kiumehr, S., Ziai, K., & Djalilian, H. R. (2017). Customized versus noncustomized sound therapy for treatment of tinnitus: A randomized crossover clinical trial. Annals of Otology, Rhinology, and Laryngology, 126(10), 681–687.
Mancini, P. C., Tyler, R. S., Jun, H. J., Wang, T.-C., Ji, H., Stocking, C., Secor, C., Rojas-Roncancio, E., & Witt, S. (2020). Reliability of the minimum masking level as outcome variable in tinnitus clinical research. American Journal of Audiology, 29(3), 429–435.
McGlothlin, A. E., & Lewis, R. J. (2014). Minimal clinically important difference: Defining what really matters to patients. JAMA, 312(13), 1342–1343.
Newman, C. W., Wharton, J. A., & Jacobson, G. P. (1995). Retest stability of the Tinnitus Handicap Questionnaire. Annals of Otology, Rhinology, and Laryngology, 104(9), 718–723.
Norman, G. R., Sloan, J. A., & Wyrwich, K. W. (2003). Interpretation of changes in health-related quality of life. Medical Care, 41(5), 582–592.
Norman, G. R., Stratford, P., & Regehr, G. (1997). Methodological problems in the retrospective computation of responsiveness to change: The lesson of Cronbach. Journal of Clinical Epidemiology, 50(8), 869–879.
Ousmen, A., Touraine, C., Deliu, N., Cottone, F., Bonnetain, F., Efficace, F., Brédart, A., Mollevi, C., & Anota, A. (2018). Distribution- and anchor-based methods to determine the minimally important difference on patient-reported outcome questionnaires in oncology: A structured review. Health and Quality of Life Outcomes, 16(1), Article 228.
Simundic, A. M. (2012). Diagnostic accuracy—Part 1. Basic concepts: Sensitivity and specificity, ROC analysis, STARD statement. Point of Care, 11(1), 6–8.
Stouffer, J. L., & Tyler, R. S. (1990). Characterization of tinnitus by tinnitus patients. Journal of Speech and Hearing Disorders, 55(3), 439–453.
Tunkel, D. E., Bauer, C. A., Sun, G. H., Rosenfeld, R. M., Chandrasekhar, S. S., Cunningham, E. R., Jr., Archer, S. M., Blakley, B. W., Carter, J. M., Granieri, E. C., Henry, J. A., Hollingsworth, D., Khan, F. A., Mitchell, S., Monfared, A., Newman, C. W., Omole, F. S., Phillips, C. D., Robinson, S. K., … Whamond, E. J. (2014). Clinical practice guideline: Tinnitus. Otolaryngology–Head and Neck Surgery, 151(Suppl. 2), S1–S40.
Tyler, R. S., & Conrad-Armes, D. (1983a). The determination of tinnitus loudness considering the effects of recruitment. Journal of Speech and Hearing Research, 26(1), 59–72.
Tyler, R. S., & Conrad-Armes, D. (1983b). Tinnitus pitch: A comparison of three measurement methods. British Journal of Audiology, 17(2), 101–107.
Tyler, R. S., Ji, H., Perreau, A., Witt, S., Noble, W., & Coelho, C. (2014). Development and validation of the Tinnitus Primary Function Questionnaire. American Journal of Audiology, 23(3), 260–272.
Tyler, R. S., Oleson, J., Noble, W., Coelho, C., & Ji, H. (2007). Clinical trials for tinnitus: Study populations, designs, measurement variables, and data analysis. Progress in Brain Research, 166, 499–509.
Vernon, J. A., & Meikle, M. B. (2003). Tinnitus: Clinical measurement. Otolaryngologic Clinics of North America, 36(2), 293–305.
Wang, H., Tang, D., Wu, Y., Zhou, L., & Sun, S. (2020). The state of the art of sound therapy for subjective tinnitus in adults. Therapeutic Advances in Chronic Disease, 11.
Wheaton, L., & Pope, J. (2010). The minimally important difference for patient-reported outcomes in spondyloarthropathies including pain, fatigue, sleep, and Health Assessment Questionnaire. Journal of Rheumatology, 37(4), 816–822.
Wright, A., Hannon, J., Hegedus, E. J., & Kavchak, A. E. (2012). Clinimetrics corner: A closer look at the minimal clinically important difference (MCID). Journal of Manual and Manipulative Therapy, 20(3), 160–166.
Ye, T., Chen, K., Li, D., Yin, K., Li, Y., Long, J., & Hui, L. (2022). Global research hot spot and trends in tinnitus treatment between 2000 and 2021: A bibliometric and visualized study. Frontiers in Neurology, 13, Article 1085684.
Zeman, F., Koller, M., Figueiredo, R., Aazevedo, A., Rates, M., Coelho, C., Kleinjung, T., de Ridder, D., Langguth, B., & Landgrebe, M. (2011). Tinnitus Handicap Inventory for evaluating treatment effects: Which changes are clinically relevant? Otolaryngology–Head and Neck Surgery, 145(2), 282–287.
Information & Authors
Information
Published In
Journal of Speech, Language, and Hearing Research
Volume 68 • Number 2 • February 2025
Pages: 827-838
PubMed: 39836456
Copyright
Copyright © 2025 The Authors. 
This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.
This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License.History
- Received: May 29, 2024
- Revised: Sep 7, 2024
- Accepted: Oct 31, 2024
- Published online: Jan 21, 2025
- Published in issue: Feb 4, 2025
Authors
Metrics & Citations
Metrics
Article Metrics
View all metricsCitations
Download article citation data for:
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
View Options
View options
Sign In Options
ASHA member? If so, log in with your ASHA website credentials for full access.
Member Login
