Minimal clinically important difference for the Mandarin version of the Tinnitus Questionnaire determined via anchor-based and distribution-based methods

In this study, we recruited participants aged 18 years and above who sought treatment for primary or secondary tinnitus at the Otorhinolaryngology Department of the Hearing Center of West China Hospital, Sichuan University from September 2020 to September 2021. Patients who were unable to complete the relevant assessment owing to cognitive impairment or difficulty in understanding, psychiatric disorders, or auditory hallucinations were excluded. This research was approved by the Biomedical Research Ethics Committee of West China Hospital (No. 2020 [311]). Participants were enrolled after they provided written informed consent.

We collected the demographic information (name, sex, age, telephone number), medical history (side of tinnitus), and tinnitus assessment data (assessment date and four assessment outcomes) of each participant at baseline. After 6 months, we followed them up telephonically. The different types of treatment interventions that participants underwent in different study centers were collected, and their self-reported clinical impression after treatment was recorded, as follows: cure/much better/no change/much worse. In addition, the participants had to undergo the following four assessments of tinnitus severity: doctor evaluation, self-report, the MTQ, and the visual analog scale (VAS), all at baseline and at the follow-up. The doctor's evaluation and the self-report were carried out independently.

All continuous variables are presented as means (standard deviations [SDs]), and categorical data are presented as counts (percentages). Student’s t test or one-way analysis of variance was used for the comparison of continuous variables among groups, and the chi-squared test or Fisher’s exact test was used for the comparison of categorical variables between groups, as appropriate. Two regression models were used for flexibility in examining associations between the exposure and each outcome. The change in MTQ score from baseline was defined as ΔMTQ. An improvement in the MTQ score indicated a reduction in tinnitus severity (including much better, ΔVAS ≥ 1), while deterioration indicated exacerbation of tinnitus (including much worse, ΔVAS ≤ − 1).

There is no standardized method to determine the ideal MCID. However, methodologists generally recommend triangulating the results of multiple methods [21]. In this study, the MCIDs of the MTQ were established via anchor-based and distribution-based methods.

A total of 115 patients were investigated in this study, 57.4% of whom were women (Table 1). The average age was 43.2 ± 13.20 years. Among the patients, 40.0% complained of bilateral tinnitus, 33.0% of left-sided tinnitus, and 27.0% of right-sided tinnitus. The doctor-evaluated tinnitus severity was mostly slight, accounting for 69.6% at baseline and 60.9% at the follow-up. Nineteen patients self-reported that their tinnitus disappeared by the time of the follow-up. The average MTQ and VAS scores at baseline were 31.3 ± 14.90 and 5.03 ± 2.24, respectively, while the average MTQ and VAS scores at follow-up were 15.9 ± 11.70 and 3.58 ± 2.48, respectively. Moreover, in terms of self-reported clinical impressions, 19 patients indicated that they were cured (16.5%), 24 that it was much better (20.9%), 63 that there was no change (54.8%), and 9 that it was much worse (7.8%). The distributions of ΔMTQ scores in the four self-reported clinical impression groups (F = 11.47, p < 0.001) are illustrated in Fig. 1, including the median value and 5th–95th percentile.

The correlation coefficients between the self-reported clinical impression and ΔMTQ (r₁), and VAS score and MTQ score at baseline (r₂) were calculated using the self-reported clinical impression and VAS score as the main anchors. The correlation coefficients were as follows: r₁ = 0.453 (p < 0.001) and r₂ = 0.619 (p < 0.001).

According to the change in self-reported clinical impression, 24 patients reported improvement and 9 reported deterioration; for these changes, the MCID values were 22.75 and 3.63, respectively. Twenty-eight patients changed one point on the VAS, and 65 patients changed at least one point on the VAS; their MCID values were 9.39 and 25.50, respectively. The mean and SD of the difference between the scores of MTQ under the two anchors were calculated, and the mean of the difference was recorded as the MCID (Table 2).

The results of the MCID for the MTQ score according to the ROC analyses are summarized in Table 3. The cut-off point for deterioration based on self-reported impression of change was 13.5, which corresponded to an area under the ROC curve of 0.775.

Distribution-based methods are used to estimate the MCID based on the observed distribution of score changes [29]. The results of the MCID for the MTQ score are presented in Tables 4 and 5, which were calculated by three variation indexes on two subjective criteria: the ES, SEM, and RCI. When ES = 0.5, the MCID values for improvement, deterioration, and total change according to self-reported clinical impression were 8.21, 7.64, and 8.07, respectively, and those for the VAS score were 1.09, 7.64, and 6.29, respectively (Table 4). Similarly, the MCIDs calculated for the MTQ score when 1.96SEM was used as the intermediary index according to self-reported clinical impression were 8.52, 7.93, and 8.37, respectively, and those calculated according to the VAS score were 1.13, 7.93, and 6.53, respectively (Table 5). The MCIDs calculated for the total when 1.96RCI was used as the intermediary index according to self-reported clinical impression were 12.04, 11,21, and 11.83, respectively, and those calculated according to the VAS score were 1.60, 11.21, and 9.23, respectively (Table 5).

The MCIDs for the change in total MTQ ranged from 6.29 to 19.00, those for improvement from 1.09 to 22.75, and those for deterioration from 3.50 to 7.64 (Table 6 and Fig. 2).

Through correlation analysis, we discovered that self-reported clinical impression and VAS scores correlated best with changes in MTQ scores (r > 0.6), while physician evaluations correlated poorly with changes in MTQ scores (r < 0.6). This may be because the MTQ is a patient self-assessment scale, similar to the VAS and different to physicians' evaluations. This result emphasizes that the MTQ score accurately reflects patients' tinnitus.

The VAS and self-reported clinical impression were selected as anchors for analysis, and the mean/median and ROC curve were used for analysis and calculation of the MCID, respectively. The distribution-based method is used to estimate the MCID based on the observed distribution of score changes. Different MCID values were calculated (Table 6). The next objective was to determine the clinically most appropriate MCID values from the many calculated MCID values. We did this by comparing the results obtained via several analysis methods. We tried to minimize the placebo effect (which has a prevalence of up to 40% in tinnitus treatment [30]) by selecting the maximum MCID value as cut-off.

Considering that self-reported clinical impression is a qualitative variable and the VAS score is a quantitative variable, the VAS score more closely correlates with MTQ score changes. The MCID value for MTQ score improvement (7.50) calculated with the VAS score as anchor was similar to the MCID value for deterioration (7.64). Therefore, we decided to use the MCID value calculated with the VAS score as anchor. After treatment, the MTQ score of patients with tinnitus decreased by 7.5 or more than that before treatment. This result suggests that the treatment is effective in resolving or ameliorating tinnitus. Similarly, when the MTQ score increased by 7.5 or more compared with pretreatment, the patient’s tinnitus was considered aggravated.

In their 2012 MCID study on the TQ, Adamchic et al. [20] determined an MCID of − 5 and + 1 for the TQ, by using the Clinical Global Impression score as the ROC method in the anchoring method for calculation, taking response bias into account. In their 2018 study of the MCID for the German version of the TQ, Hall et al. [21] suggested a median of 12 as the MCID value, considering measurement bias and error. Therefore, we selected two subjective indicators with good correlations as anchors for the analysis and used the VAS score calculated based on the results as anchor.

A major limitation of this study was the following. For patients with tinnitus whose initial MTQ score was less than 7.5, we could not judge whether the treatment they received was effective in terms of the MCID. However, according to our previous study, a score of less than 21 on the MTQ is considered to indicate no problem with tinnitus [17]. When the initial score of the patient is less than 7.5, we believe that tinnitus has little impact on their quality of life. The next step is to expand the sample size in future studies to verify the adaptability of the MTQ in terms of treatment-related changes.