Causal relationships between brain functional networks and tinnitus: A bidirectional 2-sample Mendelian randomization study

Abstract

This study aimed to discuss the potential causal relationship between brain functional networks and tinnitus using a bidirectional Mendelian randomization (MR) approach. Using genetic data from 6 datasets linked to brain functional networks, and tinnitus data sourced from the FinnGen project, we conducted 2-sample MR analyses. Instrumental variables (IVs) were selected based on stringent criteria, including genome-wide significance, clumping to ensure independence, and exclusion of palindromic single-nucleotide polymorphisms (SNPs) and those associated with confounders. The primary MR analysis employed the inverse variance weighted method supplemented by sensitivity analyses using the weighted median and Mendelian randomization-Egger (MR-Egger) methods to address potential pleiotropy. MR analyses suggested a genetic correlation between functional brain networks and the risk of tinnitus. These findings were robust across various sensitivity analyses, including MR-Egger and Mendelian Randomization Pleiotropy RESidual Sum and Outlier, supporting the absence of pleiotropy and outliers. Our findings provide important evidence for the causal relationship between brain dysfunction and tinnitus, and provide a potential brain function domain reference for the clinical treatment and intervention of tinnitus.

1. Introduction

Tinnitus is identified as a hallucinatory perception occurring in individuals without any external stimulus.^[1] Reports indicate that the prevalence of tinnitus in the US was 11.2% in 2014.^[2] About 1% of people have extreme or weakening tinnitus, and they seek various modalities of treatment, which brought heavy economic burden to the society and families.^[3] Previous studies have found that different subtypes of tinnitus may be triggered by a variety of mechanisms.^[4–6] Recent studies have shown that tinnitus is associated with abnormal neural activity in widely distributed brain networks, including the auditory network and extra-auditory structures, such as the cortical insula of the parahippocampal cortex and cerebellum.^[4]

It follows that tinnitus is a multilevel brain disease, often accompanied by emotional and cognitive symptoms, as well as alterations in brain networks involved in significant memory pain and attention perception.^[7] Synchronization of distinct but functionally coherent networks in large-scale human brain networks contributes to complex cognitive and emotional processes^[8] within tinnitus and multiple within abnormal activities in brain networks are associated with interactions, and changes in the participation of specific components in these networks are thought to account for their heterogeneity. For example, increased activation of the salience neural network may prevent habituation of hallucinatory sounds, leading to persistent perception of tinnitus.

For tinnitus types associated with deafness, auditory afferents limit the information that the brain receives from the external environment. The brain tries to compensate for hearing loss by exciting the auditory nerves, which may involve changes in the auditory and parahippocampal cortices. Thus, in this tinnitus condition, cochlear injury reduces neural input to the central nervous system, resulting in compensatory changes that lead to maladaptive hyperactivity and hypersynchrony in the auditory pathway and other areas of the central nervous system, contributing to the generation and maintenance of tinnitus.^[9,10] The difference is that some people with tinnitus do not experience hearing loss. Surveys of this patient population provide evidence that tinnitus subtypes are associated with defects in top-down noise reduction mechanisms.^[11,12] Specifically, the ventromedial prefrontal cortex and inferior cingulate cortex are responsible for the regulation of sensory signals, as well as the descending projection of the thalamic reticular nucleus, involving the top–down sound inhibition system of the frontal limbic striatum, which filters out abnormally increased signals in the ascending auditory pathway.^[11] Tinnitus may arise when suppression of this unrelated sensory signal fails.

These findings clearly suggest that there is a dysfunction in brain functional connectivity in patients with tinnitus. However, the causal relationship between brain functional networks and tinnitus remains unclear. Until now, we did not know whether brain functional network dysfunction was a result of tinnitus or if it played a causal role in tinnitus. Exploring the causal relationship between brain functional networks and tinnitus will help to understand the pathogenesis of tinnitus and provide insight into potential therapeutic targets for tinnitus, such as specific brain functional networks.^[13] Recent large-scale genetic studies on tinnitus and brain function networks provide a critical opportunity to explore the causal relationship between brain function networks and tinnitus. By utilizing genetic findings from tinnitus genetics and resting-state cerebral functional magnetic resonance imaging (rsfMRI), Mendelian randomization (MR) analysis allows causal inference. Based on Mendel’s laws of inheritance, MR uses exposure-related genetic variation as a tool to determine causal relationships between exposure and outcomes.^[13] During gamete formation, alleles are randomly assigned, and therefore, MR can be considered a natural randomized controlled trial that minimizes confounding.^[13]

In this study, we performed a 2-way 2-sample MR analysis to determine the causal relationship between 191 brain rsfMRI phenotypes and tinnitus.^[13] We identified 6 potential causal relationships between brain dysfunction and tinnitus. Our findings provide key evidence for the causal relationship between brain dysfunction and tinnitus, and supply potential areas of brain function for the clinical treatment and intervention of tinnitus.

2. Method

2.1. Study design

In this study, 2-sample MR was used to examine the relationship between brain functional networks and tinnitus, with the study design illustrated in Figure 1. For the MR results to be considered valid, 3 critical conditions need to be satisfied: (I) the instrumental variables (IVs) used in the analysis must have a strong association with exposure. (II) These IVs must not be linked to confounders that could influence both the exposure and outcome (sleep disturbances, alcohol intake, etc.). Sleep and alcohol intake are considered important independent factors for tinnitus.^[14] (III) The impact of IVs on outcomes should be mediated exclusively through exposure without any alternative pathways.

Figure 1.

The study design and major assumptions of the 2-sample MR study. MR = Mendelian randomization.

2.2. Linkage disequilibrium score regression

The linkage disequilibrium score regression (LDSC) operates fundamentally as a linear regression model where the input consists of the genome-wide association study (GWAS) analysis results. The magnitude of heritability can also be assessed using the LDSC.^[15]

Before conducting MR, LDSC was used to examine the genetic link between functional brain networks and tinnitus. This was complemented by a review of the existing literature on the relationship between functional brain networks and tinnitus (Supplementary Files 1–21, Supplemental Digital Content, https://links.lww.com/MD/O827).

2.3. Mendelian randomization analysis

In this study, the inverse variance weighted (IVW) method was primarily utilized to determine the causal link between brain functional networks and tinnitus, assuming the absence of horizontal pleiotropy.^[16] Two distinct sets of genetic instruments were employed to elucidate the causality of brain functional networks affecting tinnitus. To this end, we searched for 6 datasets related to functional brain networks. This was complemented by a review of the existing literature on the relationship between functional brain networks and tinnitus. Basic characteristics of the brain functional network database are presented in Table S1 (Supplementary File 22, Supplemental Digital Content, https://links.lww.com/MD/O828). Principal components were corrected for each data point.

We randomized these 6 datasets with the tinnitus dataset separately for 2-sample MR (P < 5 × 10⁻⁸, linkage disequilibrium [LD] r² < 0.001, Kb = 10,000). The F-statistics of each of these 6 datasets were larger than the conventional value of 10, indicating that the instruments had strong potential to predict tinnitus in subjects. Details are presented in Table 1.

Table 1.

Significant Mendelian randomization analysis results of causal links between brain functional networks and risk of tinnitus.

Brain functional networks	Edge_name	N SNPS	Method	OR	OR (95% CI)	Beta	P-value	MR-Egger regression		Heterogeneity (IVW)		R 2	F
								Egger intercept	P-value	Cochran’s Q	P-value
PHENO405	Net100_Pair5_21	3	MR-Egger	1.148	0.305–4.320	0.138	0.872	−0.038	.495	1.848	.397	0.001	37.031
PHENO405	Net100_Pair5_21	3	Weighted median	0.567	0.363–0.866	−0.567	0.012
PHENO405	Net100_Pair5_21	3	Inverse variance weighted	0.595	0.407–0.869	−0.520	0.007
PHENO1137	Net100_Pair24_44	7	MR-Egger	0.574	0.245–1.345	−0.555	0.257	0.013	.629	4.069	.667	0.001	33.639
PHENO1137	Net100_Pair24_44	7	Weighted median	0.649	0.480–0.877	−0.433	0.005
PHENO1137	Net100_Pair24_44	7	Inverse variance weighted	0.712	0.566–0.895	−0.340	0.004
PHENO1142	Net100_Pair29_44	5	MR-Egger	0.557	0.100–3.106	−0.585	0.552	0.004	.939	4.378	.357	0.001	39.581
PHENO1142	Net100_Pair29_44	5	Weighted median	0.603	0.433–0.839	−0.506	0.003
PHENO1142	Net100_Pair29_44	5	Inverse variance weighted	0.598	0.457–0.783	−0.513	<0.001
PHENO1250	Net100_Pair5_47	6	MR-Egger	1.268	0.459–3.506	0.238	0.671	0.010	.762	3.163	.675	0.001	34.698
PHENO1250	Net100_Pair5_47	6	Weighted median	1.361	1.012–1.829	0.308	0.042
PHENO1250	Net100_Pair5_47	6	Inverse variance weighted	1.493	1.179–1.890	0.401	0.001
PHENO1300	Net100_Pair9_48	3	MR-Egger	2.729	0.591–12.594	1.004	0.421	−0.099	.302	3.846	.146	0.001	47.874
PHENO1300	Net100_Pair9_48	3	Weighted median	0.713	0.495–1.026	−0.338	0.068
PHENO1300	Net100_Pair9_48	3	Inverse variance weighted	0.615	0.406–0.933	−0.486	0.022
PHENO1382	Net100_Pair44_49	5	MR-Egger	2.475	0.703–8.711	0.906	0.253	−0.034	.396	1.255	.869	0.001	35.536
PHENO1382	Net100_Pair44_49	5	Weighted median	1.723	0.920–1.761	0.241	0.151	1.	2.
PHENO1382	Net100_Pair44_49	5	Inverse variance weighted	1.332	1.018–1.743	0.287	0.037	3.	4.

CI = confidence interval, IVW = inverse variance weighted, MR-Egger = Mendelian randomization-Egger, N SNPs = the number of single-nucleotide polymorphisms, OR = odds ratio.

2.4. Summary data for tinnitus

Summary data for tinnitus was sourced from the finngen_R11_H8_TINNITUS dataset, which included 8926 cases of tinnitus and 397,865 controls. We obtained summary data from the practical database, extracting information on each single-nucleotide polymorphism (SNP) linked to tinnitus and those associated with the brain functional networks, including their effects on tinnitus, effect sizes, and standard errors.

2.5. Reverse-direction Mendelian randomization analysis

Furthermore, we conducted reverse MR using the IVW method on brain functional networks to assess whether tinnitus had a causal effect on brain functional networks.

We randomly assigned this tinnitus dataset to each of the 6 datasets, and owing to the insufficient number of SNPs extracted from the tinnitus dataset, the screening conditions for SNPs were adjusted for bivariate double-sample MR (P < 5 × 10⁻⁶, LD r² < 0.001, Kb = 10,000). When the P-value was 5e−08, the number of SNPs extracted from the exposure factor (tinnitus) for reverse MR was 0. When the P-value was 5e−07, the number of SNPs extracted from the exposure factor (tinnitus) of reverse MR was 3, and in further extraction, the number of SNPs was significantly insufficient for further analysis; therefore 5e−06 was selected as the extraction standard. The F-statistic in the tinnitus dataset was greater than the conventional value of 10, indicating that these instruments have strong potential to predict brain functional networks. The procedures for conducting reverse MR mirrored those used for standard MR, as illustrated in Figure 1. The results of the reverse MR are presented in Table 2.

Table 2.

Reverse Mendelian randomization analysis of tinnitus and brain functional networks.

Traits	N SNPs	Edge_name	Edge_Pheno_ID	Method	OR	OR (95% CI)	Beta	P-value	MR-Egger regression		Heterogeneity (IVW)		R 2	F
									Egger intercept	P-value	Cochran’s Q	P-value
Tinnitus	16	Net25_Pair6_13	Pheno72	MR-Egger	1.006	0.954–1.061	0.006	.829	0.007	.156	12.485	.642	0.872	11164.970
Tinnitus	16	Net25_Pair6_13	Pheno72	Weighted median	1.014	0.963–1.068	0.014	.596
Tinnitus	16	Net25_Pair6_13	Pheno72	Inverse variance weighted	1.037	1.002–1.074	0.037	.039

CI = confidence interval, IVW = inverse variance weighted, MR-Egger = Mendelian randomization-Egger, N SNPs = the number of single-nucleotide polymorphism, OR = odds ratio.

2.6. Statistical analyses

We harmonized the effect alleles from brain functional networks and tinnitus before applying several MR methodologies to compute MR estimates of the influence of brain functional networks on tinnitus. These methods include the IVW, weighted median, and Mendelian randomization-Egger (MR-Egger) methods. Each method addresses horizontal pleiotropy differently, which is why multiple techniques were employed.

Sensitivity analysis plays a vital role in identifying pleiotropy and addressing heterogeneity that could substantially affect MR estimates. We used heterogeneity, indicated by a Cochran Q-derived P-value of less than .05, from the IVW method to suggest potential horizontal pleiotropy. Directional pleiotropy was evaluated using the intercept from the MR-Egger regression, where a P-value below .05 confirmed its presence. Additionally, the MR-Pleiotropy Residual Sum and Outlier (MR-PRESSO) method was employed to assess and correct horizontal pleiotropy. This approach involves detecting horizontal pleiotropy, rectifying it through outlier removal, and assessing significant shifts in causal estimates before and after correction. MR-PRESSO is notably less prone to bias and is more precise than both IVW and MR-Egger, particularly when <10% of the variants demonstrate horizontal pleiotropy. Leave-one-out analysis was also performed to check whether any individual SNP disproportionately affected the MR estimates.

All statistical analyses in this study were conducted using the “TwoSampleMR” package in the R software environment (version 4.3.3). For the MR-PRESSO tests, which can be thought of as the cleanup crew ensuring the data is not muddied by outliers, we utilized the dedicated “MR-PRESSO” package. This setup ensured that our analyses were both robust and reliable, leveraging the specific strengths of each package to obtain the best possible insights from the data.

2.7. MRlap analysis

MRlap is an R-package tailored to perform 2-sample MR studies that may involve overlapping samples. The reliability of MR estimates can be compromised by several biases, such as overlapping samples between exposure and outcome, the employment of weak instruments, and the winner’s curse phenomenon. We developed a method (MRlap) that simultaneously considers weak instrument bias and the winner’s curse while accounting for potential sample overlap. Assuming a spike‐and‐slab genomic architecture and leveraging LDSC and other techniques, we can analytically derive, reliably estimate, and hence correct for the bias of inverse variance weighted 2-sample Mendelian randomization (IVW‐MR) using association summary statistics only. We tested our approach by using simulated data for a wide range of realistic settings. In all explored scenarios, our correction reduced the bias, in some situations by as much as 30‐fold.^[17]

2.8. Data sources

Brain functional network GWAS results can be easily explored and downloaded through the Brain Imaging Genetics Knowledge Portal (https://bigkp.org). Summary data for tinnitus were acquired from FinnGen, which encompassed 8926 cases and 397,865 controls from the R11_manifest cohort. The data are available at https://www.finngen.fi/en/access_results.

2.9. Selection of instruments variables

To select appropriate IVs, we performed comprehensive quality checks on SNP as follows: (I) SNPs linked to exposure were chosen based on a genome-wide significance threshold (P < 5 × 10⁻⁸). (II) Clumping procedures (r² < 0.001, clumping distance = 10,000 kb) were implemented to prevent LD among IVs associated with functional brain networks, ensuring the independence of SNPs. (III) We did not substitute SNPs missing in the outcome GWAS with proxy SNPs that had high LD (R² > 0.8) with the intended SNPs. (IV) Palindromic SNPs were excluded to maintain consistency in the effects of SNPs on both exposure and outcome. (V) To remove SNPs potentially linked to outcome-related risk factors, we employed R language software to screen out SNPs associated with potential confounders such as sleep disturbances and alcohol intake.^[14] (VI) To combat the problem of weak IVs, we calculated the F-statistic for each SNP, where R² indicates the variance in exposure explained by the SNPs, n denotes the sample size, and k is the number of IVs used. IVs with an F-statistic below 10 were considered weak and excluded from the analysis. F = R² × (n−k−1)/k × (1−R²).

2.10. Ethics

The summary-level data used in this study were publicly available and de-identified with prior approval from the Ethical Standards Committee. No further ethical approval was obtained for this study.

3. Results

3.1. Selection of IVs related to brain functional networks

In this study, we extensively searched for brain functional networks in the GWAS database, screened SNPs by significance thresholds, and excluded SNPs that might be related to confounders (Table 1, Fig. 2, and Supplementary Files 23–40, Supplemental Digital Content, https://links.lww.com/MD/O829).

Figure 2.

OR, P-value, and beta results of causal links between brain functional networks and tinnitus. OR = odds ratio.

3.2. Causal effects and sensitivity analysis of brain functional networks on tinnitus

The initial MR analysis confirmed a link between brain functional networks and the risk of tinnitus, with all IVs demonstrating strong instrument strength (F > 10) (Table 1). Subsequent analyses and sensitivity assessments identified significant outcomes associated with the functional brain networks. Notably, elevated levels of Pheno405, Pheno1137, Pheno1142, Pheno1250, Pheno1300, Pheno1382 correlated with the risk of tinnitus (odds ratio [OR]: 0.595, 95% confidence interval [CI]: 0.407–0.869, P = .007 for Net100_Pair5_21; OR: 0.712, 95% CI: 0.566–0.895, P = .004 for Net100_Pair24_44; OR: 0.598, 95% CI: 0.457–0.783, P<.001 for Net100_Pair29_44; OR: 1.493, 95% CI: 1.179–1.890, P = .001 for Net100_Pair5_47; OR: 0.615, 95% CI: 0.406–0.933, P = .022 for Net100_Pair9_48; OR: 1.332, 95% CI: 1.018–1.743, P = .037 for Net100_Pair44_49).

3.3. MRlap analysis to identify brain functional networks on tinnitus

MRlap estimates can be affected by several biases such as overlapping samples between exposure and outcome, the use of weak instruments, and the winner’s curse. The results from the MRlap test indicated no significant overlap between exposure and outcome factors, suggesting that errors from overlap rates were minimal and that IVW-MR estimates could be reliably utilized (Table 3).

Table 3.

Mendelian randomized MRlap analysis of brain functional networks and risk of tinnitus.

Exposure	Outcome	MR correction observed effect	MR correction observed effect se	MR correction m IVs	MR correction observed effect P	MR correction corrected effect	MR correction corrected effect se	MR correction corrected effect P	MR correction test difference	MR correction P difference
PHENO405	Tinnitus	−0.066	0.030	3	.029	−0.100	0.045	.026	2.129	.033
PHENO1137	Tinnitus	−0.048	0.016	7	.003	−0.069	0.023	.003	2.851	.004
PHENO1142	Tinnitus	−0.073	0.020	5	.000	−0.105	0.028	.000	3.574	.000
PHENO1250	Tinnitus	0.053	0.017	7	.002	0.075	0.024	.002	−2.946	.003
PHENO1300	Tinnitus	−0.074	0.032	3	.023	−0.109	0.048	.024	2.228	.026
PHENO1382	Tinnitus	0.041	0.019	5	.037	0.058	0.028	.037	−1.985	.047

corrected effect = corrected causal effect estimate, corrected effect P = corrected causal effect P-value, corrected effect se = corrected causal effect standard error, IVs = instrumental variables, m IVs = number of IVs used, MR = Mendelian randomization, observed effect = IVW-MR observed causal effect estimate, observed effect P = IVW-MR observed causal effect P-value, observed effect se = IVW-MR observed causal effect standard error, P difference = P-value corresponding to the test statistic used to test for differences between observed and corrected effects, se = standard error, test difference = test statistic used to test for differences between observed and corrected effects.

3.4. LDSC analysis to identify genetic association between brain functional networks and tinnitus

The LDSC fundamentally operates as a linear regression model using GWAS analysis results as input. In this model, the independent variable was the LD score of the SNP locus, whereas the dependent variable was the Bespoke statistic adhering to a chi-squared distribution. The magnitude of heritability can also be assessed using LDSC. We examined the genetic association between functional brain networks and tinnitus by LDSC (Table 4).

Table 4.

Linkage disequilibrium score analysis to identify genetic associations between brain functional networks and tinnitus.

Exposure name/ID	Outcome	h2 observed	h2 observed se	h2 Z	h2 P	rg	rg se	rg P
NET100_PAIR5_21/PHENO405	Tinnitus	0.085	0.015	5.694	1.240E−08	−0.306	0.149	.040
NET100_PAIR24_44/PHENO1137	Tinnitus	0.176	0.017	10.281	8.550E−25	0.017	0.094	.852
NET100_PAIR29_44/PHENO1142	Tinnitus	0.132	0.017	7.838	4.580E−15	0.009	0.113	.939
NET100_PAIR5_47/PHENO1250	Tinnitus	0.155	0.019	8.380	5.290E−17	0.031	0.103	.766
NET100_PAIR9_48/PHENO1300	Tinnitus	0.110	0.017	6.291	3.160E−10	0.108	0.164	.511
NET100_PAIR44_49/PHENO1382	Tinnitus	0.121	0.017	6.918	4.570E−12	−0.027	0.121	.826

h² = heritability, represents genetic contribution, h² observed = the greater the observed genetic contribution, the better, h² observed se = the smaller the observed standard error of genetic contribution, the better, h² P = the P-value of genetic contribution is required to be less than .05, h² Z = h² observed/h² observed se, LD = linkage disequilibrium, rg = genetic correlation estimate, rg P = the P-value of genetic correlation estimate is required to be less than .05, rg se = the smaller the observed standard error of genetic correlation estimate, the better, se = standard error.

3.5. Reverse Mendelian randomization analysis of tinnitus and brain functional networks

The initial MR analysis confirmed a link between brain functional networks and the risk of tinnitus, with all IVs demonstrating strong instrument strength (F > 10) (Table 4 and Supplementary Files 41–43, Supplemental Digital Content, https://links.lww.com/MD/O830). Subsequent analyses and sensitivity assessments identified the significant outcomes associated with tinnitus. Table 4 presents the results. Studies have shown that tinnitus can promote Frontal & Frontal|Supp_Motor_Area|Temporal.

4. Discussion

The occurrence of tinnitus is an extremely complex process that is affected by several factors. Chronic tinnitus is anatomically and phenomenomically divided into 3 pathways: the lateral sound, medial pain, and descending noise reduction pathways. The triple-network model was proposed as a unified framework for common neuropsychiatric disorders. It is believed that the aberrant interaction between the 3 basic networks, the default pattern network of self-representation, the saliency network of behavior-related coding, and the goal-oriented central execution network, is the basis of brain diseases. Tinnitus commonly leads to negative cognitive, emotional, and autonomic responses, phenomenologically expressed as tinnitus-related suffering. After the appearance of tinnitus symptoms, tinnitus sounds through the amygdala lead to fear in patients and promote the hippocampus to produce fear emotion memory. This emotion, generated through the amygdala subcortical pathway, may be the key reason for anxiety and depression in patients with tinnitus. Chronic tinnitus can also be associated with the self-representing default mode network and becomes an intrinsic part of self-perception. This is likely an energy-saving evolutionary adaptation that detaches tinnitus from sympathetic energy-consuming activities. Eventually, this can lead to functional disabilities by interfering with central-executive network.^[18]

In the present study, we performed 2-way 2-sample MR analyses to investigate the causalities between 191 brain rsfMRI phenotypes and tinnitus, and our results revealed the causal relationships between brain resting-state functional networks and tinnitus.

Our study showed that there is a causal relationship between functional brain networks and tinnitus. Tinnitus is mainly involved in brain function networks including default mode network, central executive network, attention network, and salience network. Previous studies have shown that patients with moderate to severe tinnitus have greater changes in central brain regions, including the auditory cortex, insula, parahippocampus, and posterior cingulate gyrus.^[19] The connection between the insula and the auditory cortex, as well as the posterior cingulate gyrus and the hippocampus, was augmented, indicating abnormalities in the auditory network, salience network, and default mode network.^[19] Specifically, the insula is the core region of the neural pathway and is composed of the auditory cortex, insula, and parahippocampus.^[19] Attention deficit studies have previously been conducted in patients with tinnitus using behavioral paradigms. The results showed that tinnitus had a positive effect on selective, persistent, or distracting attention. In addition, patients with tinnitus have reduced executive control, that is, reduced ability to solve problems such as task-related and nontask-specific information.^[20] Partial auditory deprivation could alter the characteristics of the salience network and other related brain areas, thereby contributing to hearing impairment-induced neuropsychiatric symptoms.^[21]

The results of reverse MR showed that tinnitus also had a promoting effect on the central-executive, default-mode and salience networks. This suggests that persistent tinnitus symptoms can cause negative effects on attention and mood, which are controlled by the central-executive, default-mode and salience networks of the brain.

For the first time, we demonstrated a causal relationship between brain functional networks and tinnitus, and our findings provide insights into the pathophysiology of the disease and suggest potential new noninvasive tinnitus treatment strategies.

Our study confirmed a causal relationship between brain functional networks and tinnitus through the MR of disorders associated with brain functional networks and tinnitus. Six diseases and states related to brain functional networks can influence the occurrence of tinnitus and promote its further progression. Our research presents 6 extensive, large-scale MR studies that examined the genetic-level causal relationship between brain functional networks and tinnitus. It leveraged the most recent comprehensive GWAS data and employed genetic prediction techniques to establish causality. In addition to utilizing the MR method, we also incorporated the MRlap approach, which accounts for and corrects various biases using cross-trait LD score regression (LDSC) to estimate sample overlap. This corrected effect estimation was utilized for sensitivity analysis; if there was no significant difference from the observed effect, the IVW-MR estimate was considered reliable.

MR is second only to randomized controlled trials in evidence-based medicine, and this study provides evidence-based support for clinical exploration of the relationship between anxiety, depression, and tinnitus.

However, our study has some limitations. First, the GWAS data are exclusively from individuals of European descent, limiting the generalizability of our findings to other populations. Second, the use of R language software to eliminate confounding factors (sleep disorders, alcohol intake, etc.) may introduce bias due to subjective decisions by the authors, necessitating caution in interpreting the results. While sample overlap might inflate test outcomes, we anticipated a minimal impact, as no explicit sample overlap was identified.

5. Conclusion

This study established a causal link between functional brain networks and tinnitus, indicating that functional brain networks may exacerbate tinnitus symptoms. Additionally, these findings strengthen the evidence-based medical foundation of the relationship between tinnitus and functional brain networks. Our findings not only provide important evidence for the causal relationship between brain dysfunction and tinnitus, but also provide potential areas of brain function for clinical treatment and intervention of tinnitus.

Acknowledgments

The authors are grateful to the researchers who assisted in this study. The authors thank the participants and researchers of the FinnGen Study. The authors thank the participants and researchers from the GWAS database. The authors thank the participants and researchers of Mendelian randomization of the relevant R packages.

Author contributions

Conceptualization: Cheng Zhong, Yu Guo.

Data curation: Cheng Zhong, Lin Ji.

Formal analysis: Cheng Zhong, Haopeng Zhang, Ying Dong.

Investigation: Cheng Zhong, Lin Ji.

Methodology: Cheng Zhong, Lin Ji.

Project administration: Cheng Zhong, Lin Ji, Yu Guo.

Resources: Cheng Zhong, Haopeng Zhang.

Software: Cheng Zhong, Lihua Wang, Ying Dong.

Supervision: Cheng Zhong, Haopeng Zhang.

Visualization: Cheng Zhong, Lin Ji.

Writing – original draft: Cheng Zhong.

Writing – review & editing: Cheng Zhong.