Study selection process
During our database search, 897 studies were initially retrieved, and 524 were selected after eliminating duplicates. Then, 477 studies without high relevance to our topic were discarded after reading titles and abstracts, and 47 studies were further evaluated by reading the full manuscripts. As a result, 18 full-text articles were abandoned for the following reasons: 4 described topics irrelevant to the efficacy and safety of rTMS on chronic tinnitus; 1 was a viewpoint; 2 were protocol designs; 8 were non-randomized controlled studies; and 3 did not provide free online full-text materials. Ultimately, 29 RCTs with 1,228 patients were included in this systematic review and meta-analysis. The flow chart depicting the study selection process is shown in Figure 1.
Study characteristics and methodological quality
The 29 eligible studies were randomized controlled studies published from 2004 to 2017. Six were conducted in the USA, 4 in Germany, 3 in China (including 1 in Taiwan), Turkey, and South Korea, 2 in the Czech Republic, and 1 in Italy, Egypt, Brazil, Australia, Netherlands, Finland, the UK, and Belgium. These clinical trials exhibited sample sizes that varied between 8 and 146 participants, a mean duration of tinnitus between 6 and 420 months, and a mean treatment course between 5 and 20 days. Of the 29 studies included (34 comparisons), 27 studies (32 comparisons) assessed the auditory cortex, 1 examined the motor cortex, and 1 did not target a specific cerebral area. Among the 32 comparisons in the 27 studies focusing on the auditory cortex, 19 comparison analyses showed the superiority of rTMS over sham-rTMS. Additionally, the 1 study focusing on the motor cortex confirmed the advantage of rTMS compared to sham-rTMS. In terms of the number of rTMS sessions, 11 studies (15 comparisons) reported a treatment time of 10 days, 12 (12 comparisons) of 5 days, 4 (4 comparisons) of 20 days, 1 (1 comparison) of 4 days, and 1 (2 comparisons) did not provide the stimulation duration. Regarding insights into different courses of rTMS treatment, 9 comparison analyses about a 5-day treatment showed that rTMS had better efficacy than sham-rTMS; however, the advantage of rTMS was nonsignificant after 20 days of treatment in all studies. Of the 29 studies (34 comparisons) included, 20 (23 comparisons) explored the left auditory cortex in patients with unilateral or bilateral tinnitus. In all eligible studies, 2 included only patients with bilateral tinnitus, 3 did not describe the tinnitus-affected side, and the remaining 24 included patients with either unilateral or bilateral tinnitus. Fifteen studies (18 comparisons) reported hearing loss in some or all of the included patients. The basic characteristics of the 29 studies are summarized in Table 1. The methodological quality graphs (Figs. 2 and 3) presented each item for each included study, and each item was shown as percentages across all trials according to our established quality evaluation standard.
The clinical efficacy and safety of rTMS in the treatment of chronic tinnitus
THI scores 1 week post intervention
Of the 29 included studies, 3 reported[16,25,27] 1-week post-intervention THI scores. Because of nonsignificant heterogeneity (I2 = 0%, P = 0.57) among the studies, a fixed effects model was utilized. The outcome manifested a statistically significant difference between the rTMS and sham-rTMS groups (MD: -7.92, 95% CI: -14.18, -1.66; P =0.01) (Fig. 4).
THI scores 2 weeks post intervention
Three studies[15,25,26] containing statistics for 1-week post-intervention THI scores were available for the analysis using a random effects model, with significant heterogeneity among the studies (I2 = 72%, P = 0.03). The results exhibited no statistically significant differences in the 2-week post-intervention THI scores between the two groups (MD:-1.51, 95% CI: -13.42, 10.40; P = 0.80).
THI scores 1 month post intervention
Seven studies[16,17,20,22,24,25,27,] assessing 1-month post-intervention THI scores were included in the meta-analysis. There was no statistically significant heterogeneity among the studies (I2 = 0%, P = 0.53), so a fixed effects model was utilized. The results showed a significant difference in 1-month post-intervention THI scores between the two groups (MD: -8.52, 95% CI: -12.49, -4.55; P < 0.0001) (Fig. 5).
THI scores 6 months post intervention
Four studies[15,20,22,27] estimating 6-month post-intervention THI scores were available for the meta-analysis using a fixed effects model, with no statistically significant heterogeneity among the studies (I2 = 21%, P = 0.28). The results showed a significant difference in 6-month post-intervention THI scores between the two groups (MD: -6.53, 95% CI: -11.40, -1.66; P =0.009) (Fig. 6).
Mean change in THI scores 1 month post intervention
Three studies[20,22,24] evaluating the mean change in THI scores from baseline to 1 month post intervention were meta-analysed using a random effects model, with significant heterogeneity among the studies (I2 = 56%, P = 0.08). The results exhibited a statistically significant difference in the mean change in THI scores at 1 month post intervention between the two groups (MD: -14.86, 95% CI: -21.42, -8.29; P <0.00001).
Mean change in THI scores 6 months post intervention
Two studies[20,22] evaluating the mean difference in THI scores from baseline to 6 months post intervention were meta-analysed using a fixed effects model, with no statistically significant heterogeneity among the studies (I2 = 0%, P = 0.87). The results showed that there was a significant difference in the mean change in THI scores 6 months post intervention between the two groups (MD: -16.37, 95% CI: -20.64, -12.11; P <0.00001).
Other indicators for outcome evaluation
The following studies were meta-analysed for the outcome of patients: 2[16,27] appraising TQ scores 1 week post intervention; 2[16,27] with TQ scores 1 month post intervention; 2[15,27] with TQ scores 6 months post intervention; 3 [16,19] (1 [19] containing two RCTs) with mean changes in TQ scores 1 week post intervention; 2[17,27] with VAS scores 1 month post intervention; and 2[16,17] with tinnitus loudness 1 month post intervention. There was a statistically significant difference in TQ scores 1 week post intervention between the rTMS and sham-rTMS groups (P =0.02). Nonsignificant differences in other outcomes were found between the two groups (MD: -6.53, 95% CI: -11.40, -1.66; P =0.009) (Table 2).
Adverse events
Fifteen studies[5,15,17,19,20,23,27,27-31,36,39] reporting adverse events after rTMS sessions were meta-analysed using a fixed effects model, with nonsignificant heterogeneity among the studies (I2 = 37%, P = 0.13). The results showed a nonsignificant difference in the incidence of adverse events between the rTMS and sham-rTMS groups (12.55% vs. 13.38%; OR: 1.11, 95% CI: 0.51-2.42; P = 0.79) (Fig. 7). Among these adverse events, 21 patients reported headache; 7, worsening of tinnitus; and 5, sleep disturbances. Facial muscle discomfort, back pain, muscle hardening, and ENT symptoms (e.g., rhinitis, otitis media) were each reported in 3 patients; neck and shoulder stiffness and jaw spasms were each reported in 2 patients; increased sensitivity to noise, painful sensation in the affected ear, and anxiety and panic attacks were each reported in 1 patient. Nine patients reported other events.
Sensitivity analyses
Sensitivity analyses were performed for the selected studies to identify outliers that affected the overall results. There was a nonsignificant difference in the stability of the results (Fig. 8), which validated the rationality and reliability of our meta-analysis.
Evaluation of publication bias
Visual inspection of funnel plots was adopted in this evaluation (Fig. 9). Egger᾽s and Begg᾽s analyses[16,17,20,22,24,25,27] showed no publication bias in our meta-analysis (P = 0.925).