A total of 99 cases were collected between March 1 and April 14, 2020, and 89 in the same period in 2019 (Figure 1). Table 1 compares the demographics and tinnitus characteristics between the subjects in the different years. The case load for tinnitus appeared to be higher in 2020 than in the same period in 2019 (99 vs. 89, or an increase of 11.2%). Such an increase could be largely attributed to the accumulation of cases when all the non-emergency visits were suspended during the lockdown between January and February 2020. The two groups of different years were matched by all clinical characteristics except the incidence of anxiety.
Table 1. Comparison of initial clinical characteristics of patients between 2020 and 2019
|
March-April 2020
|
March-April 2019
|
p value
|
Sex (M:F)
|
43:56
|
43:46
|
.502
|
Age (year old, M ± SD)
|
50.8 ± 15.1
|
52.6 ± 14.7
|
.487
|
Educational background
Bachelor and superior
Inferior to bachelor
|
54
45
|
47
42
|
.812
|
Duration (month)
|
25 ± 53.6
|
31.3 ± 50.4
|
.108
|
Site
|
|
|
.177
|
Bilateral
|
36
|
41
|
|
Unilateral
|
63
|
48
|
|
Anxiety involved/total#
Risk factors
|
74/99 (74%)
|
53/89 (59%)
|
.026
|
Sensorineural hearing loss
|
69
|
65
|
.614
|
Noise exposure
|
1
|
0
|
1
|
Hypertension
|
3
|
6
|
.179
|
Hyperthyroidism
|
1
|
0
|
1
|
Head/neck trauma
|
1
|
0
|
1
|
Chi-square test was used for the between-group comparison, sex, educational background, site, anxiety and the risk factor of sensorineural hearing loss using, t-test on age, Mann-Whitney Rank Sum Test on Duration, Fisher’s exact test on the risk factors of noise exposure, hypertension, hyperthyroidism and head/neck trauma.
3.1. The increase in Anxiety in 2020 and its impact on THI and MML
In the 2020 group, 74 out of 99 (or 74.7%) subjects had an SAS higher than 45 (the criterion for anxiety), which was significantly higher than that in the 2019 group (53/89, or 59%, c2 = 4.938, p = 0.026). Overall, the SAS score in 2020 group was significantly higher than that of 2019 group (61.9 ± 11.9 in 2020 versus 49.1 ± 8.6 in 2019; U = 6867 via Mann-Whitney Rank Sum Test, p < 0.001, Figure 2a), which was fully due to the difference in the anxiety subgroups (68.0 ± 6.4 in 2020 vs. 54 ± 8 in 2019; U = 3550 via Mann-Whitney Rank Sum Test, p < 0.001, Figure 2a). Therefore, the higher SAS in 2020 was not simply due to the higher incidence of subjects with anxiety, but also the higher level of anxiety in the involved subjects.
Figure 2. Comparisons of initial SAS, THI and MML between years and subjects with and without anxiety. (a): SAS showing a significant difference between years and between the subgroups within the two years. (b) and (d): The differences in THI and MML as the result of the two main factors—year and anxiety. (c) and (e): Post-hoc comparison on THI and MML showing the difference within the factors of year and anxiety respectively. Within 2020, subjects with anxiety appeared to have a significantly higher THI and MML; no difference was seen in THI between anxiety and non-anxiety subgroups within 2019, while a higher MML was seen in non-anxiety subgroup within 2019. THI: tinnitus handicap inventory, SAS: Zung’s Self-Rating Anxiety Sale, MML: minimum masking levels.
The THI score in the 2020 group was 40.1 ± 6.9, which was significantly higher than that in the 2019 group (34 ± 8.3) as shown by the group effect in a two-way ANOVA against year group and anxiety (F1, 184 = 16.278, p < 0.001). The ANOVA also demonstrated a significant effect of anxiety: 38.8 ± 8.6 for subjects with anxiety and 33.8 ± 7.5 for those without (F1, 184 = 11.628, p < 0.001, Figure 2b). However, there was not a significant interaction between two factors (F1, 184 = 2.3, p = 0.131). Post-hoc pairwise comparisons showed that the THI score of anxiety subgroup in 2020 was 41.7 ± 7.7, which was significantly higher than the corresponding subgroup in 2019 (34.8 ± 8.1; q = 6.904, p < 0.001), and that of non-anxiety subgroup in 2020 (35.6 ± 5; q = 4.766, p < 0.001, Figure 2c). Interestingly, the THI of non-anxiety subgroup in 2020 was (almost) same as that of the anxiety subgroup in 2019. However, there was no significant difference in THI score across the non-anxiety subgroups between years (Figure 2c).
The between-year difference in THI was further analyzed using a breakdown of the scores in the emotional, functional, and catastrophic questionnaire sections. A significant between-year difference was demonstrated in the emotional score (14.636 ± 3.7 in 2020 and 12.3 ± 3.3 in 2019; the Mann–Whitney rank-sum test, U = 5942.5, p < 0.001), in the functional score (18.515 ± 3.6 in 2020 and 15.5 ± 4.2 in 2019, U = 5211.5, p < 0.001) and in the catastrophic scores (7.0 ± 2.5 vs. 6.1 ± 2.6, U = 5173, p = 0.035). This result suggests that the higher THI in 2020 could be partially related to the increase in anxiety.
A two-way ANOVA similar to that for THI showed a significant year effect with subjects in 2020 had significantly lower MMLs (8.3 ± 3.5 dB SL) as compared to those in the 2019 group (10.4 ± 4.3 dB SL; F1, 184 = 21.745, p < 0.001). However, the effect of anxiety was not significant (F1, 184 = 0, p = 0.977; Figure 2d). The higher MML in 2019 could be largely attributed to the high MML in the non-anxiety subgroup this year as demonstrated by the Post-hoc pairwise test, which showed that the non-anxiety subgroups had a higher MML (11.7 ± 4.1 dB SL) in 2019 than the patients with anxiety in 2019 (9.5 ± 4.3 dB SL, q = 3.627, p < 0.001, Figure 2e). Within 2020, however, the anxiety subgroup had an MML of 8.9 ± 3.7 dB SL, which was slightly but significantly higher than the non- anxiety subgroup this year (6.7 ± 2.0 dB SL, q = 3.441, p = 0.015; Figure 2e). The result suggests that there is no clear indication whether anxiety played a role in the loudness of tinnitus.
Pearson correlation was conducted between SAS and THI and MML respectively in each year. In 2020, a weak positive correlation was seen between SAS and catastrophic THI (r = 0.319, p = 0.001), but not to another two subscales of THI. In this year there is also a moderate correlation between SAS and MML (r = 0.337, p < 0.001). In 2019, however, the significant correlation was seen in any pair of measurement (p > 0.05).
3.2. Anxiety and Treatment Outcomes
The 94 patients in the 2020 group completed their face-to-face follow-up 2 months after the first assessment, while this number was 85 in the 2019 group (Figure 1). The numbers of patients who received ST with EC (STEC), hearing aids with EC (HAEC), or EC alone were 38, 14, and 42, respectively in the 2020 group, while the respective numbers were 58, 10, and 17 in the 2019 group. Due to the small sample sizes in patients receiving hearing aids in 2020, we only analyzed the treatment outcomes of STEC (as case group in each year) and EC alone (as control group in each year). No between-year differences were seen in basic demographic features, risk factors and duration of tinnitus between the years in subjects treated with STEC (Table 2) and EC alone (Table 3). The incidence of anxiety in the patients receiving STEC was higher in the 2020 group (Table 2), but not such year difference was seen in patients received EC alone (Table 3).
Table 2. Between-year match in the demographic and selected clinic features in tinnitus patients treated with STEC
|
May-June 2020
|
May-June 2019
|
p‑value
|
Sex (M:F)
|
16:22
|
30:28
|
.356
|
Age (year old, mean ± standard deviation)
|
48.2 ± 15.7
|
50.2 ± 14.1
|
.629
|
Educational background
|
|
|
.507
|
Bachelor and superior
|
21
|
36
|
|
Inferior to bachelor
|
17
|
22
|
|
Duration of tinnitus (month)
|
25.5 ± 43.7
|
31.8 ± 54.3
|
.428
|
Site
|
|
|
.454
|
Bilateral
|
18
|
32
|
|
Unilateral
|
20
|
26
|
|
Anxiety involved/total #
|
29/38 (76%)
|
32/58 (55%)
|
.035
|
Risk factors
|
|
|
|
Sensorineural hearing loss
|
24
|
40
|
.555
|
Noise exposure
|
0
|
0
|
\
|
Hypertension
|
1
|
6
|
.396
|
Hyperthyroidism
|
0
|
0
|
\
|
Head/neck trauma
|
0
|
0
|
\
|
Chi-square test was used for the between-group comparisons on sex, educational background, site, anxiety and the risk factor of sensorineural hearing loss using, t-test on age and duration, Fisher’s exact test on the risk factors hypertension.
Table 3. Between-year match in the demographic and selected clinic features in tinnitus patients treated with EC
|
May-June 2020
|
May-June 2019
|
p‑value
|
Sex (M:F)
|
19:23
|
5:12
|
.262
|
Age (year old, mean ± standard deviation)
|
49.3 ± 15.3
|
56.4 ± 12.9
|
.097
|
Educational background
|
|
|
.149
|
Bachelor and superior
|
21
|
5
|
|
Inferior to bachelor
|
21
|
12
|
|
Duration of tinnitus (month)
|
20.6 ± 33.4
|
27.5 ± 45.7
|
.66
|
Site
|
|
|
.222
|
Bilateral
|
15
|
9
|
|
Unilateral
|
27
|
8
|
|
Anxiety involved/total #
|
29/42 (69%)
|
13/17 (76%)
|
.753
|
Risk factors
|
|
|
|
Sensorineural hearing loss
|
27
|
11
|
.976
|
Noise exposure
|
0
|
0
|
\
|
Hypertension
|
1
|
0
|
1
|
Hyperthyroidism
|
0
|
0
|
\
|
Head/neck trauma
|
0
|
0
|
\
|
Chi-square test was used for the between-group comparisons on sex, educational background, site and the risk factor of sensorineural hearing loss, t-test on age and tinnitus duration, Fisher’s exact test on anxiety and the risk factors of hypertension.
3.3. The effect of treatment on SAS
Figure 3 summarized the effect of the two treatments on SAS. In consistency with the data of whole sample (Figure 2a), the pre-treatment SAS was much higher in 2020 than in 2019 for the subjects treated with both STEC (Mann-Whitney Rank Sum Test, U = 411, p < 0.001, Figure 3a) and EC alone (U = 460.5, p = 0.031, Figure 3c). However, the effect of EC alone on SAS appeared to be qualitatively different from that of STEC in that the SAS was not decreased (improved) but increased in 2020 group after the treatment (Figure 3c), so that the post-treatment SAS in the 2020 group (63 ± 11) was even significantly higher than the before-treatment SAS in the 2019 group (52.9 ± 10, Mann–Whitney rank-sum test, U = 527, p < 0.004). This raised the question whether and how the number of subjects qualified as having anxiety changed after each treatment. Such changes were summarized in Table 4. In 2019, a large portion of subjects who had anxiety changed to non-anxiety status after either of the two treatments. In 2020, however, the number of cases with anxiety was increased, slightly after STEC, but largely after EC alone. In each method, there was a significant difference between years in the % change of cases with anxiety.
Table 4. Changes of cases with anxiety after the treatments of STEC and EC alone
#: a positive change means an increase in cases with anxiety, *: the p values were the results of Fisher’s Exact Test, other cells using Chi-square tests.
An additional table file shows this in more detail [see additional file 1].
The SAS was significantly reduced in both years after the STEC treatment (Mann-Whitney Rank Sum Test, p < 0.001). However, due to the large initial difference, the post-treatment SAS score in the 2020 group (58.0 ± 10.6) was still significantly higher than the pre-treatment SAS in the 2019 group (48.3 ± 8.5, Mann–Whitney rank-sum test, U = 534, p < 0.001). These results suggest that the anxiety associated with COVID‑19 was not been fully counteracted by the treatment.
Figure 3. The SAS difference before and after the treatment of STEC and EC alone. (a) and (c): The pre- and post-SAS. (b) and (d): the pre-post difference of SAS score. STEC treatment reduced SAS in both years (a). However, EC alone did not improve SAS in 2020, instead the SAS was increased significantly in the 2nd assessment (c). Correspondingly, STEC produced a slightly better improvement in SAS in 2019 than in 2020, but improvement by EC alone was much better in 2019 than in 2020, in which SAS was deteriorated. The number of symbols (*, $ or #) represents the level of significance, with 1, 2 or 3 symbols for p < 0.05, 0.01, or 0.001 respectively. STEC: sound therapy + educational counseling, EC: educational counseling.
To further evaluate the effect of STEC on anxiety, a two-way ANOVA was performed on the pre-post SAS difference against the factor of year group and anxiety (Figure 3b). A significant year difference was seen since the SAS improvement appeared to be slightly but significantly smaller in 2020 (5.0 ± 8.6) than in 2019 (6.1 ± 3.8, F1, 92 = 6.046, p = 0.016). Combined with the higher initial SAS in 2020, this implies that the higher initial anxiety in 2020 may have made the treatment less effective in reducing anxiety. However, this assumption is conflicted with the fact that the subjects with anxiety gained more reduction in SAS after STEC (6.6 ± 6.2 in the subjects with anxiety vs. 4.0 ± 5.9 in the non-anxiety subjects; effect of anxiety: F1, 92 = 10.447, p = 0.002). Furthermore, the post-hoc test within 2020 revealed a larger SAS reduction (7.0 ± 8.0) in the anxiety subgroup this year than the non-anxiety subgroup in which the SAS was increased (negative improvement: -1.5 ± 7.2, post-hoc test within 2020, Tukey method; q = 5.364, p < 0.001). This result was in sharp contrast with the null difference in the SAS improvement between the anxiety subgroup (6.6 ± 6.2) and the non-anxiety subgroup (6.0 ± 3.5) in 2019 (Figure 3b).
A two-way ANOVA similar to the STEC was done for EC alone and showed a significant effect of year group: the pre-post difference in SAS in 2020 was negative (-3.4 ± 4.6, for an worse SAS) as compared with the large improvement in 2019 (7.1 ± 7.5; F1, 55 = 26.022, p < 0.001). Since the initial SAS in the subgroup in 2020 receiving STEC was not significantly different from that in the subgroup receiving EC alone this year (63 ± 12 versus 59.5 ± 12.1; Mann–Whitney rank-sum test, U = 640.5, p = 0.129), the deteriorated SAS after EC alone suggests that the subjects in the EC subgroup in 2020 had experienced an increased stress after the first assessment, and the stress largely increased anxiety, which was not counteracted by the EC alone treatment. A significant effect of anxiety was also seen in subjects treated with EC alone: the SAS change after EC was 1 ± 7.4 in patients with anxiety before EC and -4 ± 6 in those without (F1, 55 = 11.038, p = 0.002). There was no significant interaction between the two factors (F1, 55 = 2.773, p = 0.102). The large deterioration in SAS in the non-anxiety subjects received EC is obviously due to such change in 2020 in which the SAS changes in the non-anxiety subjects was -5.4 ± 5.7, although this value was not significantly different from the change in non-anxiety subgroup in 2019 (0.5 ± 4.9; post-hoc test, q = 2.813, p = 0.052; Figure 3d). In both years, SAS improvement was smaller in the non-anxiety subgroups, and in 2020, SAS was deteriorated, instead of improved, in both anxiety and non-anxiety subgroups. In 2019, the SAS improvement in the anxiety subgroup 9.1 ± 7.1, which was significantly higher than the non-anxiety subgroup (0.5 ± 4.9) (post hoc test, via Tukey Method, q = 4.084, p = 0.006). In 2020, the SAS change in the anxiety subgroup was -2.5 ± 3.8, and that in the non-anxiety subgroup was -5.4 ± 5.7. However, the difference was not significant (post hoc test, Tukey method, q = 2.324, p = 0.106). To further evaluate the impact of anxiety on clinic features of tinnitus, Pearson product moment correlation was calculated between the initial SAS score and the changes after the treatment. There was a moderate, positive, linear relationship between the initial SAS score and the change in patients receiving STEC in 2020 (r = 0.511, p = 0.001), but no significant correlation was found in 2019 (Figure 4a). In addition, a moderate and positive linear relationship was also seen between the initial SAS score and the change in patients receiving EC alone in 2020 (r = 0.413, p = 0.006; Figure 4b) but not in 2019 (r = 0.488, p = 0.071). These results suggest that the treatment was more effective for mitigating anxiety in subjects with higher SAS scores in 2020, which was associated with the COVID‑19 pandemic.
Figure 4. Correlations between the initial SAS score and the improvement in SAS score. (a): The correlation of STEC by year. (b): The correlation of EC alone by year. Significant, moderate correlations were seen in STEC group and EC alone groups in 2020 in which the average initial SAS scores were much higher. SAS: Zung’s Self-rating Anxiety Scale, ST: sound therapy, EC: educational counseling
3.4. The effect of treatments on THI and MML
The effect of the treatments was first examined by self-reported improvement (reduction) of tinnitus loudness. As expected, the case number and rate reporting an improvement were higher in subjects treated with STEC than in those with EC alone in both years. More importantly, the case number with improvement was significantly lower in 2020 group than in 2019 in subjects treated with both methods (Table 5). However, there were no significant differences in the case rate reporting an improvement between subjects with and without anxiety (data not shown).
Table 5. Self-reported improvement of tinnitus loudness in the Follow-ups of treatment groups between years
|
STEC group
|
EC alone group
|
p between methods
|
2020
|
27/38 (71%)
|
8/42 (19%)
|
< 0.001*
|
2019
|
51/58 (88%)
|
9/17 (53%)
|
0.004**
|
p between year
|
.038*
|
.024**
|
|
*: chi-square test, **: Fisher’s Exact Test
Figure 5. The difference in THI before and after the treatment of STEC and EC alone. (a) and (c): The pre- and post-THI scores. (b) and (d): the pre-post difference of THI score. STEC resulted in a significant THI reduction in both years (a), but there was no significant difference in the amount of reduction between years and between subjects with and without anxiety (b). EC alone reduced THI in 2019, but opposite in 2020 (c and d). The THI got deteriorated in 2020 and worse than 2019 in both subgroups with and without anxiety (d). Therefore, within subjects with or without anxiety, the treatment resulted in a better THI in year 2019. STEC: sound therapy + educational counseling, EC: educational counseling.
STEC significantly reduced the THI scores in both 2020 group from 40.7 ± 6.7 to 37.7 ± 8.0 (via paired t-tests, t0.05/37 = 3.253, p = 0.002) and 2019 group from 32.7 ± 8.3 to 28.7 ± 7.6 (via Wilcoxon Signed Rank Test, W = -1590, p < 0.001) as shown in Figure 5a. Figure 5b summarized the result of a two-way ANOVA on the improvement of THI (the pre-THI minus post-THI) by STEC against the factor of year group and anxiety. There was no significant effect for both factors (year effect: F1, 92 = 2.104, p = 0.15; anxiety effect: F1, 92 = 0.09, p = 0.759).
Surprisingly, the THI scores in 2020 rose from 39.8 ± 8.9 to 42.1 ± 9.1 after EC alone treatment (Wilcoxon Signed Rank Test, W = 426, p < 0.001), while an improvement was seen in 2019 from 35.7 ± 5.2 to 30.2 ± 6.3 (Wilcoxon Signed Rank Test, W = -153, p < 0.001, Figure 5c). Therefore, the change in THI by EC alone was -2.2 ± 2.9 in 2020, but 5.4 ± 6.9 in 2019, as shown by the significant year effect in the two-way ANOVA (F1, 55 = 25.73, p < 0.001). In this ANOVA, the effect of anxiety was not significant (Figure 5d). Correspondingly, the between-year difference in THI improvement was larger in anxiety subjects than non-anxiety ones (post hoc tests, q = 7.323, p < 0.001 in anxiety between year and q = 4.031 p =0.006 in non-anxiety between year).
Correlation analysis showed a moderate and positive linear relationship between the improvements of THI in the emotional subscale and the SAS improvement in the subjects treated with STEC in both 2020 (r = 0.506, p = 0.001) and 2019 (r = 0.623, p < 0.001; Figure 6a). In subjects treated with EC alone, significant correlation was seen only in 2019 group (r = 0.536, p < 0.026) but not in 2020 group (Figure 6b).
Figure 6. Correlations between the improvements of emotional section in THI and SAS in 2020 and 2019. (a): The correlation in ST-EC group. (b): The correlation in EC alone group. Significant, moderate correlations were seen in ST with EC group in both years and EC alone groups in 2019. SAS: Zung’s Self-rating Anxiety Scale, ST: sound therapy, EC: educational counseling.
Figure 7. The changes in MML before and after the treatment of STEC and EC alone. (a) and (c): The pre- and post-MML. (b) and (d): the pre-post difference of MML. The MML got deteriorated in 2020 and worse than 2019 in the subgroup with anxiety (d). Significance: **p < 0.01, ***p < 0.001 in ANOVA. STEC: sound therapy + educational counseling, EC: educational counseling.
MML was reduced by STEC in 2020 group (from 9 ± 4.4 dB SL to 7.3 ± 4.2 dB SL; Wilcoxon Signed Rank Test, W = -391, p = 0.003) and 2019 group (from 10.0 ± 3.8 dB SL to 7.9 ± 3.8 dB SL; W = -1525, p < 0.001; Figure 7a). The improvement (2.1 ± 1.7 dB) was slightly higher in 2019 than in 2020 (1.6 ± 2.7 dB), but the difference was not statistically significant as shown by the main effect of year in a two-way ANOVA ( F1, 92 = 1.513, p = 0.222, Figure 7b). Neither a significant effect of anxiety was seen in this ANOVA (F1, 92 = 0.006, p = 0.935).
Like THI, EC alone treatment in 2020 did not reduced MML, but increased it from 10.9 ± 4.9 dB SL to 12.4 ± 5.8 dB SL (Wilcoxon Signed Rank Test, W = 172, p = 0.003), yielding an increase of 1.5 ± 3.1 (Figure 7d). This was opposite to the decrease in MML from 12.4 ± 4.9 dB SL to 10.3 ± 4.7 in 2019 (Wilcoxon Signed Rank Test, W = -91, p < 0.001; Figure 7c). Correspondingly, a significant year effect was seen in a two-way ANOVA (F1, 55 = 10.036, p = 0.003), which did not show a significant effect of anxiety (F1, 55 = 1.944, p = 0.169). However, the year difference was mainly due to the between-year difference in the anxiety subjects in the post-hoc test (Tukey method, q = 5.24, p < 0.001), since no significant difference was seen in non-anxiety subjects between years (q = 2.129, p > 0.05, Figure 7d). Moreover, correlation analyses did not show any significant correlation between initial SAS and the change of MML after both treatment in each of the two years. Those results suggest that high anxiety in 2020 made EC alone treatment ineffective in mitigating loudness of tinnitus. The overall correlations between SAS improvements and THI (with subscale THI), MML improvements by two treatment methods in two years were seen in Table 6.
Table 6. Correlation between SAS improvements and those in THI and MML
|
r
|
p-value
|
r
|
p-value
|
Target
|
(A) STEC in 2020
|
(B) STEC in 2019
|
THI Total
|
0.459
|
0.003
|
0.193
|
0.146
|
THI Functional
|
0.17
|
0.307
|
-0.379
|
0.003
|
THI Emotional
|
0.506
|
0.001
|
0.623
|
< 0.001
|
THI Catastrophic
|
0.313
|
0.055
|
0.149
|
0.265
|
MML
|
0.134
|
0.424
|
0.143
|
0.286
|
|
(C) EC in 2020
|
(D) EC in 2019
|
THI Total
|
0.3
|
0.053
|
-0.008
|
0.975
|
THI Functional
|
0.313
|
0.04
|
-0.347
|
0.172
|
THI Emotional
|
0.07
|
0.629
|
0.536
|
0.026
|
THI Catastrophic
|
0.112
|
0.481
|
-0.04
|
0.856
|
MML
|
0.117
|
0.461
|
0.222
|
0.392
|
r: Person correlation coefficient