Adult body constitution and hearing loss: a bidirectional Mendelian randomisation study

doi:10.21203/rs.3.rs-2604463/v1

Download PDF

Article

Adult body constitution and hearing loss: a bidirectional Mendelian randomisation study

https://doi.org/10.21203/rs.3.rs-2604463/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 27 Oct, 2023

Read the published version in Scientific Reports →

You are reading this latest preprint version

Hearing-loss and -disorders represent possible mediating pathways in the associations between noise exposures and non-auditory health outcomes. In this context, we questioned whether the noise-obesity associations should consider hearing functions as possible mediators and applied Mendelian randomisation (MR) to investigate causal relationships between body constitution and hearing impairments.

We obtained genetic associations from publicly available summary statistics from genome-wide association studies in European adult populations (N= from 210,088 to 360,564) for (i) body constitution: body mass index (BMI), waist circumference (WC) and body fat percentage (BFP), and (ii) hearing loss: sensorineural hearing loss, noise-induced hearing loss, and age-related hearing impairment (ARHI). We employed colocalisation analysis to investigate the genetic associations for BMI and ARHI liability within an FTO locus. We conducted bi-directional MR for the ‘forward’ (from body constitution to hearing) and ‘reverse’ directions. We applied the random-effects inverse variance-weighted method as the main MR method, with additional sensitivity analyses.

Colocalisation analysis suggested that BMI and ARHI shared a causal variant at the FTOgene. We did not find robust evidence for causal associations from body constitution to hearing loss and suggested that some associations may be driven by FTO variants. In the reverse analyses, ARHI was negatively associated with BMI [effect size -0.22 (95% CI -0.44 to -0.01)] and BFP [effect size -0.23 (95% CI -0.45 to 0.00)], supporting the notion that ARHI may diminish body constitution. Finally, our data suggest that hearing may have little to no effect on explaining the association between noise exposure and body constitution.

Health sciences/Medical research/Epidemiology

Biological sciences/Genetics

Body mass index

body constitution

age-related hearing impairment

hearing loss

Mendelian randomisation.

There is evidence from observational studies that long-term exposure to traffic noise is associated with an increasing risk of obesity [1]. However, the WHO environmental noise guidelines recently rated the research quality on environmental noise exposure and obesity low and emphasised the importance of hearing-related outcomes on noise exposure [2]. Hearing loss is indeed a growing health concern that may mediate or confound the association between noise exposure and its attributed non-auditory health outcomes [3].

More frequently studied, hearing loss has been shown to be positively associated with excess weight [4] and its related cardiometabolic comorbidities, such as cardiovascular disease [5, 6] and diabetes mellitus [7]. The association is biologically plausible since people with these metabolic diseases have a high probability for poor micro-vascular circulation that reduces blood supply to the cochlea, resulting in damage to the hair cells and, eventually, sensorineural hearing loss. Additional studies show consistent results that elevated body mass index (BMI) is positively associated with hearing loss [8–10]. A recent meta-analysis systematically reviewed the 14 published observational studies and found a positive association for BMI and waist circumference (WC) with hearing loss [11].

However, the association between body constitution and hearing loss remain inconclusive, as other studies found no, or even, a negative association between both conditions [10]. Amid this uncertainty, some studies found that low BMI was also positively associated with hearing loss [12]. This association was further supported by a recent large-scale Korean population study, which reported that lower BMI (< 18.5 kg/m2) was associated with an increased risk of hearing loss [13]. The associations were stronger in men and old people than in women and young generations. These observations may also be corroborated by a previous study [14], suggesting that malnutrition and being underweight are positively associated with sensorineural hearing loss, which could be partly a result of degeneration and demyelination of the eight cranial nerve and mitochondrial impairment-affecting neurons.

The mechanism of the associations was mostly discussed from body constitution to hearing loss and the direction of inference of the two phenotypes lacked investigation from hearing loss to body constitution. The previous studies are insufficient to rule out the questions of reverse causation, selection bias, survival bias and residual confounding. Therefore, they cannot inform the causal effects and how they may affect future projects that test the aforementioned challenge addressing the association between noise and the risk of disease.

To address these objectives, we set up a genetically informed study (i.e., Mendelian randomisation (MR)) to infer the potential causality and its direction between body constitution and hearing loss. There is evidence that both the forward (from body constitution to hearing loss) and the reverse (from hearing loss to body constitution) association can exist and therefore we hypothesised a bi-directional causation.

We conducted a two-sample, bi-directional MR. Genetic associations for all traits were obtained from publicly available summary statistics of genome-wide association studies (GWAS)(Table 1). The data were restricted to European populations. To avoid potential weak instrument bias due to sample overlap [15], we sought GWAS with non-overlapping samples whenever possible. This was not possible for the analyses between age-related hearing impairment (ARHI) and body fat percentage (BFP), where we used the GWAS of UK Biobank for these two phenotypes, which entailed a maximum 70% participant overlap. Since both BMI and ARHI genetic associations have a strong signal within the FTO gene region, we conducted colocalisation analysis before the MR analysis to examine whether this signal is driven by a shared causal variant, or whether the association is due to confounding by linkage disequilibrium (LD).

Table 1

Sources of secondary summary statistics used to derive instrumental variables for adult body constitution and hearing loss
Phenotypes	Source	Year of publication	Codes¹, references² or ICD codes³	Sample size	Cases (n)	Controls (n)	Cases (%)
Body constitution-related phenotypes
Body mass index	UK Biobank	2018	21001	359,983	NA	NA	NA
Body mass index	GIANT	2015	Locke et al. [16]	322,154	NA	NA	NA
Waist circumference	UK Biobank	2018	48	360,564	NA	NA	NA
Waist circumference	GIANT	2015	Shungin et al. [17]	210,088	NA	NA	NA
Body fat percentage	UK Biobank	2018	23099	354,628	NA	NA	NA
Hearing loss-related phenotypes
Sensorineural hearing loss	FinnGen	Release 6, 2022	ICD 10: H90.3-5; ICD9: 3891	252,719	19,313	233,406	7.64
Noise-induced hearing loss	Finngen	Release 6, 2022	ICD 10: H83.3; ICD 9: 3881	249,936	655	249,281	0.26
Age-related hearing impairment	UK Biobank	Release 6, 2022	Wells et al. [19]	250,389	87,056	163,333	34.77
Abbreviation: GIANT, Genetic Investigation of ANthropometric Traits; ICD, International Classification of Disease

Data sources for body constitution-related phenotypes

The adult body constitution was measured by BMI (kg/m²), WC (cm) and BFP (%). We used sex-combined joint GWAS and Metabochip meta-analysis from the Genetic Investigation of ANthropometric Traits (GIANT) for BMI and WC involving 114 studies and 322,154 participants (mean age: 55.52, 46.02% men, mean BMI: 27.40 kg/m2) reported by Locke [16], and 101 studies and 210,088 participants (mean age: 55.25, 45.39% men, mean WC: 93.07 cm) reported by Shungin [17], respectively. Both GWAS were used in the MR analyses with ARHI from UK Biobank.

Apart from GIANT, the data of BMI (n = 359,983) and WC (n = 360,564) were additionally obtained from UK Biobank [18]. The two traits were assessed by anthropometric measurements. We also retrieved the GWAS of BFP from UK Biobank with 354,628 individuals. All traits used here from UK Biobank are continuous phenotypes that have been measured during the initial assessment visit (2006–2010) and rank normalised. The descriptions of BMI, WC and BFP from UK Biobank (data fields: 21001, 48 and 23099, respectively) are presented on its webpage (https://biobank.ctsu.ox.ac.uk/crystal/search.cgi).

Data sources for hearing loss-related phenotypes

Hearing loss traits included sensorineural hearing loss (SNHL), noise-induced hearing loss (NIHL) and ARHI. GWAS on risk of ARHI (87,056 cases and 163,333 controls, mean age: 60) was from Wells et al. [19] based on a self-reported hearing difficulty (HDiff) phenotype in the UK Biobank. The details of questionnaires and case-control assignments for the HDiff phenotype can be found in their published supplement material.

Summary-level genetic data for risk of SNHL and NIHL were obtained from the FinnGen [20] release 6 including 271,341 participants. The phenotypes were collected from nationwide hospital discharge and cause of death registries using the International Classification of Diseases (ICD) codes for SNHL: H90.3, H90.4, H90.5 (ICD-10) and 3891 (ICD-9) and NIHL: H83.3 (ICD-10) and 3881 (ICD-9) (Table 1). Genetic association estimates for SNHL liability were obtained from 19,313 cases (mean age: 63.37, 51.72% men) and 233,406 controls, and for NIHL liability from 655 cases (mean age: 52.43, 92.82% men) and 249,281 controls.

Genetic instrumental variables

Whenever possible, the instrumental variables (IVs) for the exposures were composed of genome-wide significant (P < 5×10^− 8) single-nucleotide polymorphisms (SNPs) that are associated with the corresponding phenotypes in the aforementioned GWAS. This included 12 SNPs for SNHL and none for NIHL. Therefore, to include a sufficient number of IVs for MR sensitivity analyses, we included SNPs with P-value < 5×10^− 5 as instruments for these traits [21].

We harmonised the effect alleles in the exposure and outcome datasets based on the guidelines provided by Hartwig [22] and excluded palindromic variants with minor allele frequency (MAF) > 0.4 and imputation info score < 0.9. We selected independent SNPs (LD) R² ≤ 0.001, within a 10000 kb window using the ‘ld_clump’ function on the mrcieu/ieugwasr package. If a variant was unavailable in the outcome GWAS summary statistics, then proxy SNPs were searched for with a minimum LD R² = 0.9.

F and R² statistics were calculated for the individual IVs to quantify instrument strength using the method described by Burgess et al. [15]. The statistical power for continuous exposures was evaluated by calculating the approximate minimum detectable odds ratio (OR) for each exposure at a power of 80%, given the sample size of the exposure, the total variance explained by the instruments and the type 1 error rate of 0.05.

Colocalisation analysis

Variants at the FTO locus are associated with BMI and ARHI but it was unclear whether both phenotypes share the same causal variant. We conducted colocalisation analysis to examine it with the method ‘coloc’ developed by Giambartolomei et al. [23], assuming a maximum of one causal variant per genomic locus. The method outputs posterior probabilities (PP) for five models of (H₀) no causal variants, (H₁) causal variant for exposure, (H₂) causal variant for outcome, (H₃) distinct causal variants for exposure and outcome, and (H₄) shared causal variant for exposure and outcome. A PP > 0.8 for model H₄ was considered evidence of colocalization [23]. We examined the variants within the FTO gene region (chr16: 53,737,875—54,155,853 on hg19) and used the default prior probabilities of a variant associated with BMI only (P₁) and ARHI only (P₂) at 10^− 4, and a variant associated with both traits (P₁₂) at 10^–5.

Statistical Analyses

We used the inverse variance weighted (IVW) method as the primary analysis in both forward and reverse MR to calculate the IVW estimator by meta-analysing the SNP-specific Wald estimates using a multiplicative random effects model. The Wald estimate for each SNP is calculated as a ratio between of the coefficients for the SNP-to-outcome and the SNP-to-exposure regressions presented in the summary statistics [24]. The random effects model was chosen to account for potential heterogeneity, evaluated by Cochran's Q (assuming balanced pleiotropy) [25]; Evidence of heterogeneity was defined with a P-value < 0.05 in the Cochran's Q. We used R version 4.2.0 [26] and TwosampleMR package [27] for all analyses.

Sensitivity analyses

There are 3 assumptions to consider before validating MR Wald ratios as a robust inference of causation: the IVs (i) are associated with the exposure, (ii) are independent of the confounders and (iii) have no effects on outcome other than via the exposure. In particular, horizontal pleiotropy, in which the IVs have an effect on other traits outside of the pathway of the exposure of interest and have an impact on the outcome of interest, or when the IVs have a direct effect on the outcome of interest, violates the third assumption [28]. We adopted complementary MR methods that make different assumptions about horizontal pleiotropy to assess the robustness of the main MR findings: MR-Egger [29], weighted median [30], weighted mode [31] and MR-PRESSO (Mendelian Randomisation Pleiotropy RESidual Sum and Outlier). MR-PRESSO was also used to detect outliers that may indicate pleiotropic effects [28]. The following additional analyses were conducted: single SNP analysis and leave-one-out analysis to provide visualisation and interpretation of two-sample MR [32].

Instrumental variable characteristics

Table 2 provides the number of IVs used and the minimal detectable odds ratios (ORs) for the forward analysis. Given the lack of individual-level data access, it is not possible to provide the detectable effect sizes for binary exposures in the reverse analyses. The association estimates for the IVs and their individual F statistics are presented in the Supplementary spreadsheets. There were a greater number of IVs for the body constitution phenotypes in the forward analyses than for hearing loss phenotypes in the reverse analyses. F statistics ranged between 52 and 67 in forward MR and between 19 and 42 in reverse MR, all of which were greater than 10, indicating little possibility of weak instrument bias (Supplementary Table 1).

Table 2

Summary data and statistical power calculation for the forward analyses
Exposures	Source	Outcomes	Source	Included Variants (n)	Variance explained (%)	Detectable OR*
BMI (kg/m²)	UKB	SNHL	FinnGen	89	1.51	1.19
BMI (kg/m²)	UKB	NIHL	FinnGen	89	1.51	2.44
BMI (kg/m²)	GIANT	ARHI	UKB	68	2.25	1.08
WC (cm)	UKB	SNHL	FinnGen	72	0.83	1.26
WC (cm)	UKB	NIHL	FinnGen	72	0.83	3.33
WC (cm)	GIANT	ARHI	UKB	41	1.42	1.10
BFP (%)	UKB	SNHL	FinnGen	85	0.71	1.28
BFP (%)	UKB	NIHL	FinnGen	85	0.71	3.67
BFP (%)	UK B	ARHI	UKB	87	0.73	1.15
*Minimum detectable odds ratio per 1 SD change in the exposure, at 80% power and type 1 error rate = 0.05. Abbreviations: BMI: body mass index, WC: waist circumference, BFP: body fat percentage, SNHL: sensorineural hearing loss, NIHL noise-induced hearing loss, ARHI: age-related hearing impairment, UKB: UK Biobank, GIANT Genetic Investigation of ANthropometric Traits

Colocalisation analysis

The results of the colocalisation analysis between BMI and ARHI are shown in Fig. 1 and the supplementary Table 2. We found a PP for H₄ of 0.92, supporting a shared causal variant (rs1558902) for BMI and ARHI at the FTO locus. The MR estimate of BMI on ARHI using only FTO rs1558902 was 0.91 (95% confidence interval [CI]: 0.88–0.94). The variant is in perfect LD with FTO variant rs1421085 adopted in the forward MR analysis (LD R² = 1). Therefore, based on the known effect of the FTO gene on BMI [33], the FTO variant rs7193144 was excluded from the IVs of ARHI in reverse MR to avoid potential exposure misspecification.

Forward analyses (from body constitution to hearing)

The results of the analyses are shown in Fig. 2 and the supplementary Table 3. According to our results, there was no evidence to support the notion of genetically predicted BMI being associated with any of the hearing loss outcomes included in these analyses. The resulting IVW odds ratios (ORs) per 1-SD increase in exposure were 1.01 (95% CI: 0.88–1.16), 1.44 (95% CI: 0.76–2.73) and 0.99 (95% CI: 0.97-1.00) for risk of SNHL, NIHL and ARHI, respectively. Similarly, we found no evidence that genetically predicted WC was associated with hearing loss outcomes. The resulting IVW ORs were 1.00 (95% CI: 0.84–1.20), 0.89 (95% CI: 0.37–2.15) and 0.98 (95% CI: 0.96-1.00) for SNHL, NIHL and ARHI, respectively. Genetically predicted BFP was not observed to be associated with SNHL [OR 0.85 (95% CI: 0.69–1.06)] or NIHL [OR 1.44 (95% CI: 0.76–2.73)]. Finally, we observed that genetically predicted BFP was positively associated with the risk of ARHI [OR 1.04 (95% CI: 1.01–1.06)]. However, the sensitivity analyses were not concordant with the IVW (Fig. 3, Supplementary Fig. 1 and Supplementary Table 4).

Reverse MR

The results of the analyses are presented in Fig. 2 and the supplementary Table 5. Based on our results, there was no evidence to support the notion that the genetically proxied SNHL was associated with any of the body constitution outcomes included in these analyses. The resulting IVW effect sizes for BMI, WC and BFP were 0.00 (95% CI: -0.03-0.01), -0.01 (95% CI: -0.03-0.01) and 0.00 (95% CI: -0.01-0.01), respectively. Similarly, we found no evidence that genetically predicted NIHL was associated with body constitution outcomes. There was some evidence of genetically proxied ARHI liability being associated with lower BMI and BFP. The observed effect sizes were − 0.22 (95% CI: -0.44 to -0.01) and − 0.23 (95% CI: -0.45-0.00), respectively. Although we did not find evidence to support the association between ARHI liability and WC, the direction of the IVW effect size [-0.13 (95% CI: -0.35 to 0.09)] was consistent with BMI and BFP.

The directions of estimates from sensitivity analyses were consistent with the IVW in the analysis between ARHI liability and BMI (Fig. 4, Supplementary Fig. 2, Supplementary Table 6). We saw heterogeneity in the association (Cochran's Q = 60.44 on 31 degrees of freedom, P = 0.00, I² = 48.7%). The funnel plot of individual SNP showed a relatively symmetrical distribution of SNP effects around the effect estimates, suggesting balanced pleiotropy (Supplementary Fig. 3). The MR-Egger intercept suggested little evidence of directional pleiotropy (Intercept = 0.00, P = 0.78). We also conducted a leave-one-out analysis (Supplementary Fig. 4) and found no evidence of a single SNP driving the results. The results from the weighted median and weighted mode method were consistent with the IVW in the analysis between ARHI liability and BFP. We did not find evidence of heterogeneity (Cochran’s Q = 10.93 on 6 degrees of freedom, P = 0.09, I² = 45.1%) and directional pleiotropy from the MR-Egger (Intercept = 0.00, P = 0.43).

The present study investigated the causal association between body constitution traits and hearing loss, employing a range of two-sample mendelian randomisation analyses and complementary sensitivity analyses.

FTO variants may associate with a reduced risk of age-related hearing impairment

Colocalisation analysis suggested that both BMI and ARHI shared the same causal variant, rs1558902, within the FTO locus. Based on the known causal role of the FTO gene on BMI [33], we can be more confident in the scenario in which BMI influences ARHI and rejected the possibility of horizontal pleiotropy in which two unrelated traits shared a causal variant. The direction of association is the opposite of our hypothesis for a positive association. Our result suggests that people carrying the FTO gene variants rs1558902 could be protected from ARHI. Unfortunately, the mechanism behind this association is still unknown and lacks investigation.

Interpretation of the evidence for a forward causation from body constitution influencing hearing loss outcomes

According to the current literature from observational studies and biological plausibility, we hypothesised that increased BMI, WC and BFP would cause hearing disorders. We found a positive association between BFP and ARHI in the IVW analysis, but the sensitivity analyses suggested the association may be driven by horizontal pleiotropy. Besides, the IVs for BFP only explained 0.73% of the variance in ARHI and weak instruments from 70% sample overlap between the two datasets bias the results in the direction of observational association. Overall, we did not find evidence to support such causal effects on either SNHL, NIHL or ARHI, suggesting that residual confounding is likely to explain the previous observations. The previous study also suggested that excess adiposity could alter hair cells in the cochlea of the inner ear [34]. Nonetheless, the use of proxies for fat distribution, such as of WC and BFP, did not improve the causal inference of any of the hearing loss phenotypes.

Interpretation of the evidence for a reverse causation from hearing loss liability influencing body constitution

In the present study, the reverse MR analyses showed little evidence to support a causal effect of SNHL and NIHL on obesity-related outcomes. The sample sizes of GWASes were large and the width of confidence intervals of effect sizes was small, suggesting enough power to detect associations. It is important to acknowledge that we relaxed the p-value threshold to increase the number of available IVs in the SNHL and NIHL, which are rare outcomes with a few cases in Finngen (the proportion of cases are 7.64% and 0.26%, respectively). However, the prevalence of NIHL was much higher in the US general population, around 18% of adults aged 20–69 years [35]. Therefore, the result from Finngen may not be very representative of the general population.

Age-related hearing impairment may reduce body mass index

Our analysis, suggested that liability to ARHI could be linked with lower BMI and BFP, corroborated by the sensitivity analyses. We first hypothesised that hearing loss-related traits would increase body constitution. However, a very recent study based on The National Health Insurance Service Health Screening Cohort of the Republic of Korea (2009–2016) including 291,471 participants with hearing loss and 6,088,979 control participants reported a negative association between hearing loss and BMI [13], which was consistent with our finding.

However, we cannot rule out a small chance of selection and survival bias as ARHI is indeed age related. This pertains to a population that must age. On the other hand, ageing can be associated with lowering of BMI through two mechanisms: (i) people with low BMI or BFP leave older and (ii) ageing after a certain age is associated with natural lowering of BMI and fat mass (i.e., sarcopenia). In a two-sample MR setting, when the true effect of exposure on outcome was null or > 0, the association became nominally negative due to selected population on age and bias to instrumental variable estimators [36]. Figure 5 included the survival factor [36] and Alzheimer’s disease (AD) as a potential confounder, accounting for the potential spurious negative association. Two MR studies [37, 38] suggest ARHI and weight loss were early manifestations of AD, supporting the notion that the association between ARHI and BMI could be confounded by AD liability. ARHI and BMI are in complex and dynamic causal pathways, and further work is needed to understand these mechanisms using MR with mediation analysis.

Limitations of the study

Some limitations warrant mention. Hearing loss is only affected by large changes in body constitutions that may lead to insufficient power for forward MR due to typically low effect sizes of genetic variants. Our results could have been biased by unobserved environmental confounders [39]. For instance, lifestyle-related factors, e.g., diet, drinking and smoking could be on the pathway from ARHI to BMI. If the ARHI SNPs affect these confounders independently, then the assumptions of MR are violated. Our MR studies so far have only investigated the linear effects in the general population. Future studies may be designed to investigate the causal role of hearing loss in subpopulations in the state of being underweight, overweight or obese (non-linear effects).

In conclusion, in our forward MR analyses, we found no evidence to support a causal effect of body constitution on hearing loss. Instead, the colocalisation analysis with the FTO variants is suggesting it as a common cause of high BMI and low risk of ARHI. In the reverse MR, we found evidence for a causal association for ARHI to lower BMI and BFP. Therefore, the mediation of hearing functions is not especially strong for the association between noise exposure and body constitution outcomes.

Acknowledgements We thank the researchers from consortia GIANT, UK Biobank and FinnGen who made the GWAS summary statistics publicly available for this study. We want to acknowledge the participants and investigators of GWASes from the consortia mentioned above.

Author contributions YH and SS conceptualised and designed the study. YH performed the analyses and wrote the first draft of the manuscript. VK, SS and YH critically revised the draft. All authors contributed to data interpretation, critical revisions, and final approval of the manuscript. YH is under the supervision of SS, MK and AP. SS is the guarantor and has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Availability of data and material The summary statistics of GWAS used in current study is publicly available in GIANT consortium (https://portals.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_files), Finngen (https://finngen.gitbook.io/documentation/v/r6/data-description#summary-association-statistics) and UK Biobank (https://docs.google.com/spreadsheets/d/1kvPoupSzsSFBNSztMzl04xMoSC3Kcx3CrjVf4yBmESU/edit#gid=227859291). The genetic instrumental variables used in the analyses are presented in Supplementary spreadsheets.

Funding This work was supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 874739 (LongITools).

Competing interest The authors declare they have no conflicts of interest.

Ethical approval The present research used publicly available summary data and did not contact participants, where no extra ethical approval is required.

Christensen JS, Raaschou-Nielsen O, Tjønneland A, Nordsborg RB, Jensen SS, Sørensen TIA, et al. Long-term exposure to residential traffic noise and changes in body weight and waist circumference: A cohort study. Environ Res. 2015 Nov 1;143(Pt A):154–61. https://pubmed.ncbi.nlm.nih.gov/26492400/.
van Kempen E, Casas M, Pershagen G, Foraster M. WHO Environmental Noise Guidelines for the European Region: A Systematic Review on Environmental Noise and Cardiovascular and Metabolic Effects: A Summary. Int. J. Environ. Health Res. 2018, Vol 15, Page 379. 2018 Feb 22;15(2):379. https://www.mdpi.com/1660-4601/15/2/379/htm.
Basner M, Babisch W, Davis A, Brink M, Clark C, Janssen S, et al. Auditory and non-auditory effects of noise on health. The Lancet. 2014 Apr 12;383(9925):1325–32.
Dhanda N, Taheri S. A narrative review of obesity and hearing loss. Int J Obes (Lond). 2017 41:7. 2017 Feb 6;41(7):1066–73. https://www.nature.com/articles/ijo201732.
Torre P, Cruickshanks KJ, Klein BEK, Klein R, Nondahl DM. The Association Between Cardiovascular Disease and Cochlear Function in Older Adults. J. Speech Lang. Hear. Res. 2005 Apr;48(2):473–81. https://pubs.asha.org/doi/full/10.1044/1092-4388%282005/032%29.
Wattamwar K, Jason Qian Z, Otter J, Leskowitz MJ, Caruana FF, Siedlecki B, et al. Association of Cardiovascular Comorbidities With Hearing Loss in the Older Old. JAMA Otolaryngol Head Neck Surg. 2018 Jul 1;144(7):623–9.
Spankovich C, Yerraguntla K. The Role of Audiology in the Care of Persons with Diabetes: Evaluation and Management of Patients with Diabetes and Hearing Loss. Semin Hear. 2019;40(4):308.
Hu H, Tomita K, Kuwahara K, Yamamoto M, Uehara A, Kochi T, et al. Obesity and risk of hearing loss: A prospective cohort study. Clin Nutr. 2020 Mar 1;39(3):870–5.
Wang J, Sung V, Lycett K, Carew P, Liu RS, Grobler A, et al. How body composition influences hearing status by mid-childhood and mid-life: The Longitudinal Study of Australian Children Int J Obes (Lond). 2018 Jul 19; 42(10):1771–81. https://www.nature.com/articles/s41366-018-0170-6.
Han X, Wang Z, Wang J, Li Y, Hu H, Hu Y, et al. Metabolic syndrome is associated with hearing loss among a middle-aged and older Chinese population: a cross-sectional study. Ann Med. 2018 Oct 3;50(7):587–95. https://pubmed.ncbi.nlm.nih.gov/29693425/.
Yang JR, Hidayat K, Chen CL, Chen CL, Li YH, Xu JY, et al. Body mass index, waist circumference, and risk of hearing loss: A meta-analysis and systematic review of observational study. Environ Health Prev Med. 2020 Jun 26;25(1):1–9. https://environhealthprevmed.biomedcentral.com/articles/10.1186/s12199-020-00862-9.
Kim SH, Won YS, Kim MG, Baek YJ, Oh IH, Yeo SG. Relationship between obesity and hearing loss. Acta Otolaryngol. 2016 Oct 2;136(10):1046–50.
Koo JS, Kim SY. Association of Body Mass Index with Hearing Loss in Korean Adult Population. J Pers Med. 2022 May 1;12(5).
Gill G v., Bell DR. Persisting nutritional neuropathy amongst former war prisoners. J Neurol Neurosurg Psychiatry. 1982;45(10):861.
Burgess S, Davies NM, Thompson SG. Bias due to participant overlap in two-sample Mendelian randomization. Genet Epidemiol. 2016 Nov 1;40(7):597–608. https://pubmed.ncbi.nlm.nih.gov/27625185/.
Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR, et al. Genetic studies of body mass index yield new insights for obesity biology. Nature. 2015 Feb 12;518(7538):197–206.
Shungin D, Winkler T, Croteau-Chonka DC, Ferreira T, Locke AE, Mägi R, et al. New genetic loci link adipose and insulin biology to body fat distribution. Nature. 2015 Feb 12; 518(7538):187–96. https://pubmed.ncbi.nlm.nih.gov/25673412/.
Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLoS Med. 2015 Mar 1;12(3):e1001779. https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1001779.
Wells HRR, Freidin MB, Zainul Abidin FN, Payton A, Dawes P, Munro KJ, et al. GWAS Identifies 44 Independent Associated Genomic Loci for Self-Reported Adult Hearing Difficulty in UK Biobank. Am J Hum Genet. 2019 Oct 10;105(4):788.
Kurki MI, Karjalainen J, Palta P, Sipilä TP, Kristiansson K, Donner K, et al. FinnGen: Unique genetic insights from combining isolated population and national health register data. medRxiv. 2022 Mar 6;2022.03.03.22271360. https://www.medrxiv.org/content/10.1101/2022.03.03.22271360v1.
Chen Z, Boehnke M, Wen X, Mukherjee B. Revisiting the genome-wide significance threshold for common variant GWAS. G3 Genes|Genomes|Genetics. 2021 Apr 12;11(2). https://academic.oup.com/g3journal/article/11/2/jkaa056/6080665.
Hartwig FP, Davies NM, Hemani G, Smith GD. Two-sample Mendelian randomization: avoiding the downsides of a powerful, widely applicable but potentially fallible technique. Int J Epidemiol. 2016 Dec 1;45(6):1717.
Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingorani AD, Wallace C, et al. Bayesian Test for Colocalisation between Pairs of Genetic Association Studies Using Summary Statistics. PLoS Genet. 2014;10(5):e1004383. https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1004383.
Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013 Nov;37(7):658–65. https://pubmed.ncbi.nlm.nih.gov/24114802/.
Bowden J, Del Greco M F, Minelli C, Davey Smith G, Sheehan N, Thompson J. A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med. 2017 May 20;36(11):1783–802. https://onlinelibrary.wiley.com/doi/full/10.1002/sim.7221.
R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2020.
Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, et al. The MR-base platform supports systematic causal inference across the human phenome. Elife. 2018 May 30;7.
Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet. 2018 May 1;50(5):693–8. https://pubmed.ncbi.nlm.nih.gov/29686387/.
Bowden J, Smith GD, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol. 2015 May 27;44(2):512–25. https://pubmed.ncbi.nlm.nih.gov/26050253/.
Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol. 2016 May 1;40(4):304–14. https://pubmed.ncbi.nlm.nih.gov/27061298/.
Hartwig FP, Smith GD, Bowden J. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol. 2017 Dec 1;46(6):1985–98. https://academic.oup.com/ije/article/46/6/1985/3957932.
Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, et al. Guidelines for performing Mendelian randomization investigations. Wellcome Open Res. 2019;4.
Loos RJF, Yeo GSH, Loos RJF, Yeo GSH. The bigger picture of FTO—the first GWAS-identified obesity gene. Nat. Rev. Endocrinol. 2013 10:1. 2013 Nov 19;10(1):51–61. https://www.nature.com/articles/nrendo.2013.227.
Hwang JH, Wu CC, Hsu CJ, Liu TC, Yang WS. Association of central obesity with the severity and audiometric configurations of age-related hearing impairment. Obesity (Silver Spring). 2009 Sep;17(9):1796–801. https://pubmed.ncbi.nlm.nih.gov/19300432/.
National Institute on Deafness and Other Communication Disorders. Quick Statistics. https://www.nidcd.nih.gov/health/statistics/quick-statistics-hearing.
Smit RAJ, Trompet S, Dekkers OM, Jukema JW, le Cessie S. Survival Bias in Mendelian Randomization Studies: A Threat to Causal Inference. Epidemiology. 2019 Nov 1;30(6):813.
Brenowitz WD, Zimmerman SC, Filshtein TJ, Yaffe K, Walter S, Hoffmann TJ, et al. Extension of Mendelian Randomization to Identify Earliest Manifestations of Alzheimer Disease: Association of Genetic Risk Score for Alzheimer Disease With Lower Body Mass Index by Age 50 Years. Am J Epidemiol. 2021 Oct 1;190(10):2163.
Abidin FNZ, Wells HRR, Altmann A, Dawson SJ. Hearing difficulty is linked to Alzheimer’s disease by common genetic vulnerability, not shared genetic architecture. NPJ Aging Mech Dis. 2021 Dec 1;7(1). https://pubmed.ncbi.nlm.nih.gov/34294723/.
Koellinger PD, de Vlaming R. Mendelian randomization: the challenge of unobserved environmental confounds. Int J Epidemiol. 2019 Jun 1;48(3):665–71. https://academic.oup.com/ije/article/48/3/665/5526895.
Statements & Declarations

No competing interests reported.

Download PDF

Journal Publication

published 27 Oct, 2023

Read the published version in Scientific Reports →

Editorial decision: Major revision
01 Aug, 2023
Reviews received at journal
17 Jul, 2023
Reviewers agreed at journal
06 Jul, 2023
Reviewers agreed at journal
23 Jun, 2023
Reviewers invited by journal
22 Jun, 2023
Editor assigned by journal
22 Jun, 2023
Editor invited by journal
28 Feb, 2023
Submission checks completed at journal
28 Feb, 2023
First submitted to journal
19 Feb, 2023

You are reading this latest preprint version

Adult body constitution and hearing loss: a bidirectional Mendelian randomisation study

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Data sources for body constitution-related phenotypes

Data sources for hearing loss-related phenotypes

Genetic instrumental variables

Colocalisation analysis

Statistical Analyses

Sensitivity analyses

Results

Instrumental variable characteristics

Colocalisation analysis

Forward analyses (from body constitution to hearing)

Reverse MR

Discussion

Interpretation of the evidence for a forward causation from body constitution influencing hearing loss outcomes

Interpretation of the evidence for a reverse causation from hearing loss liability influencing body constitution

Age-related hearing impairment may reduce body mass index

Limitations of the study

Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1