DOI: https://doi.org/10.21203/rs.3.rs-2798037/v1
Background
Hypothyroidism's causal direction and the magnitude of its association with hearing loss is uncertain due to the limitations of observational studies. This study aims to investigate the relationship between hypothyroidism and hearing loss using bidirectional two-sample mendelian randomization (MR).
Materials and Methods
The genetic variants of 462,933 participants in the hypothyroidism study and 323,978 participants in the hearing loss study were used for bidirectional two-sample MR. Independent genetic variants that were significantly (P < 5×10− 8) associated with each exposure were considered as instruments. To test for sensitivity, Cochran's Q test, the MR-Egger intercept test, and leave-one-out analysis were applied apart from the multiplicative random effects-inverse variance weighted (MRE-IVW) approach used as the main MR analysis.
Results
The forward MR revealed a significant causal estimate for the genetically predicted hypothyroidism with the high risk of hearing loss [MRE-IVW: odds ratio (OR) = 1.092, P = 0.003]. However, the reverse MR analysis found no significant correlation between genetically predicted hearing loss and hypothyroidism (MRE-IVW: OR = 1.04, P = 0.113). Sensitivity analyses showed that the causal association estimations were stable and reliable.
Conclusion
In this MR study, we demonstrated hypothyroidism was causally associated with a high risk of hearing loss. However, there was no evidence to support the causality of hearing loss on hypothyroidism.
Hearing loss is a common sensory dysfunction often unrecognized by patients and physicians and undertreated. By the time people are aware of having hearing difficulties, the hearing loss is often severe (1). World Health Organization(WHO) believe the key fact that, by 2050 nearly 2.5 billion people are projected to have some degree of hearing loss and at least 700 million will require hearing rehabilitation (2). Therefore, it is imperative to identify potential risk factors for hearing loss and develop timely and effective prevention strategies.
Hypothyroidism, also known as underactive thyroid disease, refers to the common pathological condition of thyroid hormone (TH) deficiency. The prevalence of overt hypothyroidism in the general population varies between 0–7% in the USA and between 0–5% in Europe depending on the definition used(3). And TH has been shown to be associated with a variety of diseases, including cardiovascular system diseases(4, 5), digestive system diseases(6, 7), nervous system disease (8–10) and other diseases(11–13). Numerous observational studies have suggested that congenital hypothyroidism is associated with an increased risk of hearing loss(14–16). And some observed that patients with sudden sensorineural hearing loss (SSNHL) were more likely to have hypothyroidism than normal individuals (17, 18). However, the causal association between hypothyroidism and hearing loss remains uncertain because these observational results are inevitably affected by other potential confounding effects and reverse causation.
Mendelian randomization (MR) study is a data analysis method mainly applied to etiology inferences in epidemiological studies during recent years. It advanced that differences in genotypes are major determinants of differences in intermediate phenotypes, if the association between genotype and disease can represent the effect of exposure factors on disease, then the phenotype represents an individual's exposure characteristics. Alleles follow the principle of random allocation, so this effect is unaffected by confounding factors and reverse causal associations in traditional studies (19). Mendelian randomization can thereby demonstrate the true causal relationship between exposure and outcome.
In order to give theoretical support for clinical practice, we intend to carry out a bidirectional two-sample MR research in the current study to determine if hypothyroidism and hearing loss are directly causally related. We predicted that there would be a directed causal relationship between hypothyroidism and the high risk of hearing loss.
Study Design
In MR analysis, the instrumental variables (IVs) were employed to reduce the impact of confounders and evaluate the causal association between exposure and particular outcome. The three main assumptions upon which MR was based: (1) the IVs are significantly related to exposure; (2) the IVs must not be related to confounders; and (3) the IVs should only be related to the risk of overcome through exposure (20). The MR schematic was shown in Fig. 1. This analysis was reported according to the STROBE-MR guidelines (21).
Data Sources
The MR analysis was based on summary-level genome wide association study(GWAS) data from the Integrative Epidemiologic Unit(IEU) Open GWAS database (22). Individual data were obtained from the UK Biobank, a large-scale biomedical database comprising half a million participants aged 40 to 73 years, recruited between 2006 and 2013 from across the United Kingdom (23). In the current study, we utilized publicly accessible datasets from studies that had been published, in which individuals had given formal consent and the study had received ethical approval from the appropriate committees, informed consent and additional ethical approval were thus not necessary. Details of all GWAS datasets used in this study were listed in Supplementary Table 1.
The hearing loss GWAS summary dataset (GWAS ID: ukb-a-257) included the GWAS analysis results of 323,978 samples (84,839 cases and 239,139 controls) and 10,894,596 single-nucleotide polymorphisms (SNPs). Participants were assigned to case or control groups based on their answer for the questionnaire about hearing difficulties.
The hypothyroidism GWAS summary dataset (GWAS ID: ukb-b-19732) contained 462,933 samples (including 22,687 cases and 440,246 controls) and 9,851,867 SNPs. Hypothyroidism cases were self-reported by participants via a touch screen questionnaire.
Instrumental Variables Selection
SNPs closely associated with hypothyroidism or hearing loss were extracted from published GWAS data as IVs, with P < 5×10− 8 as the main screening condition. To ensure that SNPs for hypothyroidism or hearing loss were independent, SNPs in linkage disequilibrium (r2 < 0.001 and distance > 10,000 kb) were excluded. And the F statistic was used to assess the weak instrumental variable bias, F < 10 indicates that the instrumental variable may be subject to weaker instrumental variable bias (24), which in turn was removed to avoid any effect on the results. Final valid SNPs associated with hypothyroidism or hearing loss were obtained as IVs. Then, the IVs in GWAS for outcome data were extracted based on the SNPs of the previous exposure data. F statistics for each SNP were calculated by the following equation: F = β2/SE2 (25), where β indicates the allele effect value of exposure and SE indicates the standard error of β. The main information for the final valid SNPs (including effect allele, other allele, β, standard error and P value) were collected systematically for further analysis.
After reconciling effect alleles in the exposure data and outcome data, the strength of the association between hypothyroidism and HL was calculated by using the multiplicative random effects-inverse variance weighted (MRE-IVW) method as the main analysis, meanwhile using the MR-Egger, weighted median estimator (WME), simple mode (SM), and weight mode (WM) methods as complementary analyses. Because IVW is the most widely used method among the five methods in MR studies, and could provide the most precise results when all selected SNPs were valid IVs. The causal effects were measured in the odds ratio (OR). Then, Cochran's Q statistic was employed to estimate heterogeneity from each SNP (26). The P value of Cochran’s Q test was used to test for heterogeneity. The MR-Egger intercept test was utilized to evaluate the potential horizontal pleiotropy. In addition, the “leave-one-out” sensitivity analysis was performed to determine whether the result was affected by a single SNP. The funnel plot and forest plot were constructed to visualize the presence of horizontal pleiotropy in the MR analysis. P < 0.05 was considered statistically significant. All statistical analyses were performed using the "TwoSampleMR" package(27) in R software version 4.2.2.
Characteristics of Instrumental Variables
Following the IVs selection steps, a total of 118 valid IVs strongly associated with hypothyroidism on hearing loss, and 22 valid IVs strongly associated with hearing loss on hypothyroidism were included into MR estimates, respectively. None of these IVs' F statistics were less than 10, indicating that no weak IVs were employed. An overview of the IVs included in each MR analysis was provided in Supplementary Table 3–4.
Mendelian Randomization Analysis
The MR analysis estimates from different methods for the causal effects among hypothyroidism and hearing loss were presented in Fig. 2. MRE-IVW evaluation of 118 valid IVs effects for hypothyroidism on hearing loss provided evidence to support that causal relationship existed (OR = 1.092, 95% CI = 1.031–1.156, P = 0.003), and results estimated by WME and WM were consistent with MRE-IVW. However, the evaluation of 22 valid IVs effects for hearing loss on hypothyroidism showed the consistent results that hearing loss was not causally associated with hyperthyroidism (OR = 1.042, 95% CI = 0.990–1.096, P = 0.113). P values of all five MR estimate tests were more than 0.05. Full information of each MR analysis results was provided in Supplementary Table 2. The scatter plots of valid IVs causal effects estimate for each MR analysis were shown in Fig. 3.
Sensitivity Analysis
We conducted several sensitivity analyses to evaluate the robustness of our findings. The sensitivity analyses included Cochran’s Q test, the MR-Egger intercept test, leave-one-out analysis, and the funnel plot. We used IVW and MR-Egger regression methods to calculate Cochran’s Q statistics. The results of the Cochran’s Q test and the MR-Egger intercept test can be found in Supplementary Table 5–6. Despite the existence of heterogeneities between IVs used to estimated effect between hypothyroidism and hearing loss (P < 0.05), MRE-IVW estimated results were still valid, as indicated by the results of the Cochran’s Q test. Furthermore, no horizontal pleiotropy existed between IVs and outcomes, as confirmed by the MR-Egger intercept test (P > 0.05). Additionally, the outcome of the leave-one-out test showed that the causal estimates of hypothyroidism and hearing loss were not driven by any single SNP (Fig. 4). More detailed results of leave-one-out tests were displayed in Supplementary Table 7–8. Finally, the funnel plots for MR analysis demonstrated that the data points were distributed equally around the funnel (Fig. 5). This indicated that our study had sufficient statistical power and the MR results withstand the test of sensitivity analysis.
Our study firstly investigated the causal effects between hypothyroidism and hearing loss by bidirectional two-sample MR analysis. The results of the study indicated that genetically predicted hypothyroidism causally associated with an increased risk of hearing loss. Furthermore, no evidence was found to support causal relationship of hearing loss on hypothyroidism. our findings provided a better understanding of the role of hypothyroidism in hearing loss, indicating that regular hearing assessments for hypothyroidism patients are required, in order to prevent them from severe hearing loss.
Hearing loss is reported frequently associated with hypothyroidism, ranging from mild disturbances to severe disability. A case-control study conducted by Tsai et al. (18) in Taiwan demonstrated that after adjusting for confounders, associations were identified between a history of hypothyroidism and an elevated risk of SSNHL (adjusted odds ratio [AOR], 1.54; 95% CI, 1.02–2.32; p = 0.042). Similar results were obtained in a retrospective cohort study including 8658 SSNHL patients and 34,632 controls conducted by Kim et al. (17) in Korean. Meanwhile, A prospective observational study conducted by Almagor et al. (28) in Israel indicated a high prevalence of hearing impairment among patients with congenital hypothyroidism, predominantly of the conductive type. Unlike the above study, however, more observed that sensorineural hearing loss was most common in congenital hypothyroidism, higher frequencies in particular(29–32), and this phenomenon was not associated with THS and free T4 levels(33–35). There was also a contradictory conclusion, François et al. (36) found no significant difference for the auditory thresholds at high frequencies between congenital hypothyroidism treated with L-thyroxine and control group, regardless of the cause of the thyroid failure or hormone level and the age at the start of treatment.
These contradiction between observational studies could be due to confounding bias or reverse causality. For instance, some genetic or chromosomal disorders can also present with hypothyroidism and hearing loss at the same time, including Pendred syndrome variants in the SLC26A4 gene (37–39), Woodhouse-Sakati Syndrome variants in the DCAF17 gene(40), Down syndrome(41), and TBL1X mutations(42). In animal experiments, Oliveira et al. (43) observed that perinatal hypothyroidism leads to irreversible damage to cochlear function in offspring rats. Others found that the congenital hypothyroid mouse displayed consistent morphologic abnormalities of the stereocilia on both inner and outer hair cell systems. The surrounding and supporting cells showed no significant histologic abnormalities in hypothyroid mouse and control animals (44).
Despite an abundance of evidence from observational studies support the association between hypothyroidism and hearing loss, there still remain a question about their causative relationships. The present Mendelian randomization study employs IVs to eliminate confounders and offers a new perspective on the causality between hypothyroidism and hearing loss. We provided more statistically significant results through bidirectional two-sample MR analysis. It is possible be that the presence of hypothyroidism leads to the increasing risk of hearing loss, as opposed to hearing loss promoting the occurrence of hypothyroidism.
The current study has several strengths. First, the data we used containing 323,978 and 462,993 individuals of European ancestry, allowing us to gain more precise estimates and detect slight statistical differences, meanwhile preventing the simple error in previous cohort studies. Second, bidirectional two-sample MR methods were firstly applied to explore the causal associations with hypothyroidism and hearing loss, and these methods tend to be less biased than conventional observational studies. Third, the MR method took advantage of GWAS summary data on hypothyroidism and hearing loss that were derived from two different populations, which reduced the interference of sample overlap.
Nevertheless, there are still several limitations in our study. First, our conclusions were based on data from GWASs that were solely conducted in people with European ancestry, with relatively little variety in ethnicity or culture. Thus, the results may not be fully representative of the entire population and might not be generalizable. However, the participants' consistency assures that there is little chance of population admixture confounding the results. Second, as our study used summarized data for MR analysis, stratification based on gender, age or subtypes was not applicable. Third, we found that the two samples were somewhat heterogeneous. However, because the MRE-IVW that used as the primary analysis in this work can balance the pooled heterogeneity, the presence of heterogeneity did not render the MR estimates invalid.
In conclusion, our bidirectional two-sample MR study investigated the controversial relationship between hypothyroidism and hearing loss, and revealed that hypothyroidism was causally associated with hearing loss, but no evidence was found to support reverse causally associated. Further investigation is required to determine whether this result is replicated in other environments and to uncover the potential underlying mechanisms.
Ethics approval and consent to participate
Not applicable. In the current study, we utilized publicly accessible datasets from studies that had been published, in which individuals had given formal consent and the study had received ethical approval from the appropriate committees, informed consent and additional ethical approval were thus not necessary.
Availability of data and materials
Our study used publicly available summary-level data of GWAS. The summary statistics of GWAS for hypothyroidism and hearing loss can be accessed at IEU Open GWAS database (https://gwas.mrcieu.ac.uk/); Each IDs were presented in Supplementary Table 1.
Acknowledgements
We would like to acknowledge the participants and investigators of the IEU Open GWAS database and UK biobank.
Consent for publication
Not applicable.
Author contributions
ZZ contributed to the preparation and writing of the manuscript. YL contributed to the editing of the manuscript.
Funding
The study was supported by the National Natural Science Fund of China (grant No. 81760184 and 82060185) and Natural Science Fund of Jiangxi Province (grant No. 20192ACB20027).
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.