OBJECTIVE

Evidence regarding the modifying effect of the polygenic risk score (PRS) on the associations between glycemic traits and hearing loss (HL) was lacking. We aimed to examine whether these associations can be influenced by genetic susceptibility.

RESEARCH DESIGN AND METHODS

This cross-sectional study included 13,275 participants aged 64.9 years from the Dongfeng-Tongji cohort. HL was defined according to a pure tone average >25 dB in the better ear and further classified by severity. Prediabetes and type 2 diabetes (T2D) were defined based on the 2013 criteria from the American Diabetes Association. A PRS was derived from 37 single nucleotide polymorphisms associated with HL. Multivariable logistic regression models were fitted to estimate the associations of PRS and glycemic traits with HL and its severity.

RESULTS

Elevated fasting plasma glucose (FPG), glycosylated hemoglobin (HbA1c), and T2D were positively associated with higher HL risks and its severity, with odds ratios (ORs) ranging from 1.04 (95% CI 1.00, 1.08) to 1.25 (95% CI 1.06, 1.46). We also found significant interaction between HbA1c and PRS on risks of overall HL and its severity (P for multiplicative interaction <0.05), and the effects of HbA1c on HL risks were significant only in the group with high PRS. Additionally, compared with normoglycemia in the group with low PRS, T2D was associated with an OR of up to 2.00 and 2.40 for overall HL and moderate to severe HL, respectively, in the group with high PRS (P for additive interaction <0.05).

CONCLUSIONS

PRS modifies the association of HbA1c with HL prevalence among middle-aged and older Chinese individuals.

Type 2 diabetes (T2D), a metabolic disorder characterized by high fasting glucose and insulin resistance, remains a major public health concern. According to the latest report by the Global Burden of Disease Study, the number of individuals diagnosed with diabetes reached 529 million as of 2021 (T2D accounted for 96%), and the number is projected to increase to 1.31 billion by 2050 (1). The adverse sequelae of diabetes include macrovascular complications, such as cardiovascular disease, and microvascular complications such as nephropathy and retinopathy (2). Additionally, diabetic microangiopathy in the ear may cause hearing loss (HL), which is a relatively minor microvascular complication of diabetes (3), whereas HL is the third leading cause of disability in the Global Burden of Disease Study, having been developed by 1.57 billion people by 2019, and its prevalence is estimated to increase to 2.45 billion by 2050 (4,5). It is worth considering the significance of further research in controlling the development of HL as an age-related disease in the older population.

The etiology of HL is complex and includes factors such as aging, noise exposure, and poor health conditions like T2D (6). Our previous animal models showed that hyperglycemia could induce cochlear microangiopathy and acoustic neuropathy, but antidiabetes medications significantly attenuated HL by regulating signal pathways (7,8). Moreover, epidemiological evidence reported that high fasting plasma glucose (FPG) and glycosylated hemoglobin (HbA1c) levels could account for 8–30% elevated risks of HL (9,10) and that prediabetes and T2D linked to 16–85% increased risks of HL (11–16). It should be noted that the above studies were conducted among women from the Nurses' Health Studies, relied on self-reported HL, measured HL at a single frequency, or had limited sample sizes, and so on. Conversely, no significant associations of diabetes with the risks of HL were found in two other prior studies (17,18). For the severity of HL, data from the National Health and Nutrition Examination Survey (NHANES) additionally showed that diabetes could increase the risk of HL at all severity levels (19). However, prior research in Korea reported that impaired fasting glucose was associated with mild HL in men <70 years but not with moderate to severe HL (20), and Yang et al. (21) presented no significant association of FPG or diabetes with HL severity. The conclusion on the relationship of glycemic traits with HL risk and its severity has therefore remained controversial among older adults due to these inconsistent findings.

In addition to acquired risk factors, an individual’s genetic susceptibility is also closely related to the increased risk of HL development. HL is a complex and polygenic disease, with population-based evidence indicating that 36% of the HL could be attributed to heritability (22). A recent genome-wide association meta-analysis reported 48 risk variants associated with HL and highlighted the important role of gene expression in HL development (23). The polygenic risk score (PRS) was acknowledged as a proxy quantifying and summarizing the level of complex genetic risk by combining the effects of multiple risk variants. As a multifactorial disease affected by genetic and nongenetic factors, the coexistence of genetic predisposition and abnormal blood glucose homeostasis may be related to further elevated risk of HL. However, no study has examined how glycemic traits and genetic predisposition jointly contribute to the risk of HL and its severity. It is unclear whether managing glycemic traits according to PRS is a better practice for the prevention and control of HL among middle-aged and older population.

The association of genetic factors or diabetes with HL have been established, but prior studies only focused on one aspect: either genetic variants and HL or glycemic traits and HL, and no studies so far have investigated their combined association. We therefore explored whether genetic factors and glycemic traits jointly increase the risk of HL and its severity in the Dongfeng-Tongji (DFTJ) cohort by using interaction on both multiplicative and additive scales.

Data Source and Participants

All participants were from the DFTJ cohort, which has been elaborated in a previous study (24). In brief, 27,009 retired employees derived from Dongfeng Motor Corporation (DMC) completed the baseline survey from September 2008 to June 2010 in Shiyan City of Hubei province in China. In the first follow-up survey in 2013, 38,295 participants completed questionnaires, health examinations, and blood samples collection (25). In the current cross-sectional study, we excluded participants without audiometric testing (n = 18,479), sudden deafness (n = 725) or traumatic deafness (n = 36), and those with ototoxic medicine use (n = 247). Also excluded were those with missing information on covariates and those showing domestic relations. The final analyses included 13,275 unrelated individuals (Supplementary Fig. 1). The protocol of this prospective cohort study was approved by the School of Public Health Medical Ethics and Committee, Tongji Medical College, Huazhong University of Science and Technology and Dongfeng General Hospital, DMC. Informed consent from each participant was obtained before enrollment.

Audiometry Examination and Definition of HL

Each participant underwent audiometric testing during the first follow-up survey in 2013, using the Micro-DSP ZD-21 audiometer (Sichuan Micro-DSP Technology Co., Ltd., Chengdu, Sichuan, China) by certified audiometric technicians in a sound-isolating room (25). Air-conduction hearing thresholds were determined for each ear using a pure tone at five frequencies (0.5, 1, 2, 4, and 8 kHz), with the intensity ranging from −10 to 120 dB. In accordance with the definition of World Health Organization, HL was defined as a pure tone average (PTA) >25 dB in the better ear at 0.5, 1, 2, and 4 kHz frequencies. HL severity was further classified as normal (PTA ≤25 dB), mild (PTA 25 to ≤40 dB), and moderate to severe (PTA >40 dB) in the current study.

Ascertainment of Glycemic Traits

Blood samples were collected for detecting FPG and HbA1c levels and performing other laboratory tests from all participants after overnight fasting. The FPG level was assessed by the Aeroset automatic analyzer (Abbott Laboratories. Abbott Park, IL), and the HbA1c level was determined by the high-performance liquid chromatography (D-10 System; Bio-Rad Laboratories, Hercules, CA). Prediabetes and T2D were defined based on the 2013 criteria of the American Diabetes Association (26). In brief, participants with prediabetes were ascertained by the FPG level of 5.6 mmol/L (100 mg/dL) to <7.0 mmol/L (126 mg/dL) or HbA1c of 5.7% (39 mmol/mol) to <6.5% (48 mmol/mol). Moreover, participants who had an FPG level ≥7.0 mmol/L (126 mg/dL), self-reported physician-diagnosed T2D, taking antidiabetes medications, or HbA1c level ≥6.5% (48 mmol/mol) were regarded as having T2D.

Genotyping and Calculation of PRS for HL

In the DFTJ cohort, the Illumina Infinium OminZhongHua-8 array was used to genotype all participants. The details of genotype information and quality control procedures are presented in the Supplementary Methods. In the current study, genetic variants of HL were derived from the largest and most recent genome-wide association meta-analysis (23). After considering the minor allele frequency >0.05 and Hardy-Weinberg equilibrium (P > 1 × 10−6) (27), 37 uncorrelated variants (P < 5 × 10−8) were used to construct the PRS with external weights (Supplementary Table 1) and the unweighted PRS. Referring to the previous study (28), individuals were divided into three groups according to the distribution of PRS among non-HL group: low PRS (the lowest tertile), intermediate PRS (the second tertile), and high PRS (the highest tertile).

Assessment of Covariates

Information on the age, sex (male and female), overweight/obesity (BMI ≥24 kg/m2), education levels (high school or above, middle school, primary school or below), marital status (married/remarried, unmarried/divorced/widowed), physical activity, smoking, drinking, night sleep duration (<7, 7 to <8, 8 to <9, ≥9 h), napping duration (0, 1 to ≤30, 31 to ≤60, >60 min), hypertension, dyslipidemia, shift work, and occupational noise exposure (0, 0 to <10, 10 to <20, ≥20 years) was assessed through semistructured questionnaires combined with health examinations and laboratory assays of blood samples. More additional details are presented in the Supplementary  Methods.

Statistical Analyses

The Student t test and the Pearson χ2 test were performed to compare the difference between groups in continuous and categorical variables, respectively. Multivariable logistic regression models were fitted to estimate the odds ratios (ORs) and 95% CIs of PRS and glycemic traits on HL risk and its severity. Moreover, dose-response relationships of FPG, HbA1c levels, and PRS with the risks of overall HL were evaluated based on the adjusted restricted cubic spline function (three knots located at 10th, 50th, and 90th percentiles). Stratified analyses were conducted based on the PRS groups and further evaluated using the adjusted restricted cubic spline function. Also, multiplicative interactions were assessed by adding cross-product terms of categorical or continuous PRS with glycemic traits to investigate the modifying effects of genetics.

Furthermore, the joint associations of PRS with T2D on the HL risks and its severity were assessed. Additive interactions of the high PRS and T2D on the risks of HL and its severity were also assessed using the relative excess risk due to the interaction (RERI) and the attributable proportion (AP) due to the interaction. Several sensitivity analyses were also considered. Firstly, associations of FPG and HbA1c levels with HL risk and its severity were performed after excluding participants taking antidiabetes medications. Secondly, we also performed the aforementioned interactions using the unweighted PRS. To assess whether adding PRS, glycemic traits and PRS × glycemic traits could improve the discrimination and reclassification of HL risk and its severity, the corresponding C statistics, the net reclassification index, and the integrated discrimination index were calculated. All statistical analyses were performed using R 3.6.1 software (R Development Core Team), and two-sided P values of <0.05 were considered as statistically significant.

Data and Resource Availability

The data sets generated during and/or analyzed in the current study are available from the corresponding authors upon reasonable request.

Characteristics of Study Participants

Among the 13,275 participants in the current study, 50.8% were aged ≥65 years, 44.5% were men, and HL was identified in 6,427 subjects (48.4%). The characteristics of participants with and without HL are compared in Table 1. Briefly, compared with those without HL, individuals with HL tended to be older, mainly men, had higher PRS, FPG, and HbA1c levels, and had a higher proportion of T2D. In addition, participants who were unmarried/divorced/widowed, had a low education level, high BMI, smoking, drinking, inadequate physical activity, long night sleep duration and napping, hypertension, dyslipidemia, and those had long occupational noise exposure duration accounted for a higher proportion among participants with HL. Supplementary Table 2 summarizes the comparisons of the basic characteristic of participants included and excluded (13,275 vs. 25,020) in the current study. It also presents the differences in characteristics between participants with and without audiometric testing (19,816 vs. 18,479).

Table 1

Basic characteristic of the included participants

TotalNon-HLHLP
Variables (N = 13,275) (n = 6,848) (n = 6,427)  
Age, years    <0.001 
 <65 6,529 (49.2) 4,383 (64.0) 2,146 (33.4)  
 ≥65 6,746 (50.8) 2,465 (36.0) 4,281 (66.6)  
Male sex 5,904 (44.5) 2,384 (34.8) 3,520 (54.8) <0.001 
Overweight/obesity 6,760 (50.9) 3,378 (49.3) 3,382 (52.6) <0.001 
Education levels    <0.001 
 High school or above 5,590 (42.1) 3,288 (48.0) 2,302 (35.8)  
 Middle school 4,829 (36.4) 2,446 (35.7) 2,383 (37.1)  
 Primary school or below 2,856 (21.5) 1,114 (16.3) 1,742 (27.1)  
Unmarried/divorced/widowed 1,524 (11.5) 726 (10.6) 798 (12.4) 0.001 
Physical inactivity 1,363 (10.3) 651 (9.5) 712 (11.1) 0.003 
Smokers 1,734 (13.1) 684 (10.0) 1,050 (16.3) <0.001 
Alcohol drinkers 826 (6.2) 346 (5.1) 480 (7.5) <0.001 
Night sleep duration, h    <0.001 
 <7 739 (5.6) 408 (6.0) 331 (5.2)  
 7 to <8 3,300 (24.9) 1,817 (26.5) 1,483 (23.1)  
 8 to <9 5,697 (42.9) 3,002 (43.8) 2,695 (41.9)  
 ≥9 3,539 (26.7) 1,621 (23.7) 1,918 (29.8)  
Napping duration, min    <0.001 
 0 5,519 (41.6) 2,987 (43.6) 2,532 (39.4)  
 1 to ≤30 2,011 (15.1) 1,063 (15.5) 948 (14.8)  
 31 to ≤60 3,657 (27.5) 1,805 (26.4) 1,852 (28.8)  
 >60 2,088 (15.7) 993 (14.5) 1,095 (17.0)  
Hypertension 8,763 (66.0) 4,187 (61.1) 4,576 (71.2) <0.001 
Dyslipidemia 6,012 (45.3) 2,993 (43.7) 3,019 (47.0) <0.001 
Shift work 5,935 (44.7) 3,094 (45.2) 2,841 (44.2) 0.258 
Occupational noise exposure, years    <0.001 
 0 8,320 (62.7) 4,344 (63.4) 3,976 (61.9)  
 0 to <10 1,225 (9.2) 714 (10.4) 511 (8.0)  
 10 to <20 1,466 (11.0) 769 (11.2) 697 (10.8)  
 ≥20 2,264 (17.1) 1,021 (14.9) 1,243 (19.3)  
PRS 34.04 ± 3.88 33.74 ± 3.85 34.36 ± 3.89 <0.001 
FPG, mmol/L 6.18 ± 1.69 6.09 ± 1.59 6.28 ± 1.79 <0.001 
HbA1c, % 5.86 ± 0.98 5.80 ± 0.94 5.92 ± 1.02 <0.001 
Diabetes    <0.001 
 Normal 3,543 (26.7) 2,003 (29.2) 1,540 (24.0)  
 Prediabetes 6,639 (50.0) 3,424 (50.0) 3,215 (50.0)  
 T2D 3,093 (23.3) 1,421 (20.8) 1,672 (26.0)  
TotalNon-HLHLP
Variables (N = 13,275) (n = 6,848) (n = 6,427)  
Age, years    <0.001 
 <65 6,529 (49.2) 4,383 (64.0) 2,146 (33.4)  
 ≥65 6,746 (50.8) 2,465 (36.0) 4,281 (66.6)  
Male sex 5,904 (44.5) 2,384 (34.8) 3,520 (54.8) <0.001 
Overweight/obesity 6,760 (50.9) 3,378 (49.3) 3,382 (52.6) <0.001 
Education levels    <0.001 
 High school or above 5,590 (42.1) 3,288 (48.0) 2,302 (35.8)  
 Middle school 4,829 (36.4) 2,446 (35.7) 2,383 (37.1)  
 Primary school or below 2,856 (21.5) 1,114 (16.3) 1,742 (27.1)  
Unmarried/divorced/widowed 1,524 (11.5) 726 (10.6) 798 (12.4) 0.001 
Physical inactivity 1,363 (10.3) 651 (9.5) 712 (11.1) 0.003 
Smokers 1,734 (13.1) 684 (10.0) 1,050 (16.3) <0.001 
Alcohol drinkers 826 (6.2) 346 (5.1) 480 (7.5) <0.001 
Night sleep duration, h    <0.001 
 <7 739 (5.6) 408 (6.0) 331 (5.2)  
 7 to <8 3,300 (24.9) 1,817 (26.5) 1,483 (23.1)  
 8 to <9 5,697 (42.9) 3,002 (43.8) 2,695 (41.9)  
 ≥9 3,539 (26.7) 1,621 (23.7) 1,918 (29.8)  
Napping duration, min    <0.001 
 0 5,519 (41.6) 2,987 (43.6) 2,532 (39.4)  
 1 to ≤30 2,011 (15.1) 1,063 (15.5) 948 (14.8)  
 31 to ≤60 3,657 (27.5) 1,805 (26.4) 1,852 (28.8)  
 >60 2,088 (15.7) 993 (14.5) 1,095 (17.0)  
Hypertension 8,763 (66.0) 4,187 (61.1) 4,576 (71.2) <0.001 
Dyslipidemia 6,012 (45.3) 2,993 (43.7) 3,019 (47.0) <0.001 
Shift work 5,935 (44.7) 3,094 (45.2) 2,841 (44.2) 0.258 
Occupational noise exposure, years    <0.001 
 0 8,320 (62.7) 4,344 (63.4) 3,976 (61.9)  
 0 to <10 1,225 (9.2) 714 (10.4) 511 (8.0)  
 10 to <20 1,466 (11.0) 769 (11.2) 697 (10.8)  
 ≥20 2,264 (17.1) 1,021 (14.9) 1,243 (19.3)  
PRS 34.04 ± 3.88 33.74 ± 3.85 34.36 ± 3.89 <0.001 
FPG, mmol/L 6.18 ± 1.69 6.09 ± 1.59 6.28 ± 1.79 <0.001 
HbA1c, % 5.86 ± 0.98 5.80 ± 0.94 5.92 ± 1.02 <0.001 
Diabetes    <0.001 
 Normal 3,543 (26.7) 2,003 (29.2) 1,540 (24.0)  
 Prediabetes 6,639 (50.0) 3,424 (50.0) 3,215 (50.0)  
 T2D 3,093 (23.3) 1,421 (20.8) 1,672 (26.0)  

Data are presented as n (%) or mean ± SD.

Associations of Glycemic Traits With the Risk of HL and Its Severity

In multivariable analyses, participants with normal, mild, and moderate to severe HL were 6,848 (51.6%), 4,632 (34.9%), and 1,795 (13.5%), respectively. Increasing FPG and HbA1c levels were strongly associated with higher risks of HL in a linear dose-response manner (P for overall = 0.005 and 0.015; P for nonlinearity = 0.071 and 0.476, respectively) (Supplementary Fig. 2). Moreover, the risks of moderate to severe HL increased by 7% (95% CI 1, 13) and 10% (95% CI 4, 16) per-SD increase in FPG and HbA1c levels, respectively (Table 2). Furthermore, compared with participants with normoglycemia, the adjusted ORs (95% CIs) of overall HL risks in those with prediabetes and T2D were 1.11 (1.02, 1.22) and 1.19 (1.07, 1.33), respectively (Table 2). The risk of moderate to severe HL was the strongest among participants with T2D, with an estimate (95% CI) of 1.25 (1.06, 1.46).

Table 2

Association of glycemic traits with overall HL risk and its severity

VariablesOverall HL(case/N = 6,427/13,275)Mild HL(case = 4,632)Moderate to severe HL(case = 1,795)P for trend*
Model 1     
 FPG† 1.08 (1.04, 1.12) 1.07 (1.03, 1.11) 1.10 (1.04, 1.16) <0.001 
 HbA1c† 1.08 (1.04, 1.12) 1.06 (1.02, 1.10) 1.12 (1.06, 1.18) <0.001 
 Diabetes     
  Normal 1.00 (Reference) 1.00 (Reference) 1.00 (Reference) − 
  Prediabetes 1.14 (1.04, 1.24) 1.15 (1.05, 1.27) 1.09 (0.95, 1.25) 0.014 
  T2D 1.27 (1.15, 1.41) 1.25 (1.12, 1.39) 1.34 (1.15, 1.56) <0.001 
Model 2     
 FPG† 1.05 (1.01, 1.09) 1.04 (1.00, 1.09) 1.07 (1.01, 1.13) 0.006 
 HbA1c† 1.06 (1.02, 1.10) 1.04 (1.00, 1.08) 1.10 (1.04, 1.16) 0.001 
 Diabetes     
  Normal 1.00 (Reference) 1.00 (Reference) 1.00 (Reference) − 
  Prediabetes 1.11 (1.02, 1.22) 1.13 (1.03, 1.24) 1.06 (0.93, 1.22) 0.065 
  T2D 1.19 (1.07, 1.33) 1.17 (1.04, 1.32) 1.25 (1.06, 1.46) 0.001 
VariablesOverall HL(case/N = 6,427/13,275)Mild HL(case = 4,632)Moderate to severe HL(case = 1,795)P for trend*
Model 1     
 FPG† 1.08 (1.04, 1.12) 1.07 (1.03, 1.11) 1.10 (1.04, 1.16) <0.001 
 HbA1c† 1.08 (1.04, 1.12) 1.06 (1.02, 1.10) 1.12 (1.06, 1.18) <0.001 
 Diabetes     
  Normal 1.00 (Reference) 1.00 (Reference) 1.00 (Reference) − 
  Prediabetes 1.14 (1.04, 1.24) 1.15 (1.05, 1.27) 1.09 (0.95, 1.25) 0.014 
  T2D 1.27 (1.15, 1.41) 1.25 (1.12, 1.39) 1.34 (1.15, 1.56) <0.001 
Model 2     
 FPG† 1.05 (1.01, 1.09) 1.04 (1.00, 1.09) 1.07 (1.01, 1.13) 0.006 
 HbA1c† 1.06 (1.02, 1.10) 1.04 (1.00, 1.08) 1.10 (1.04, 1.16) 0.001 
 Diabetes     
  Normal 1.00 (Reference) 1.00 (Reference) 1.00 (Reference) − 
  Prediabetes 1.11 (1.02, 1.22) 1.13 (1.03, 1.24) 1.06 (0.93, 1.22) 0.065 
  T2D 1.19 (1.07, 1.33) 1.17 (1.04, 1.32) 1.25 (1.06, 1.46) 0.001 

Data are presented as the OR (95% CI). Model 1: adjusted for age and sex. Model 2: adjusted for age, sex, overweight/obesity, education, marital status, physical activity, smoking, drinking, night sleep duration, napping duration, hypertension, dyslipidemia, shift work, and occupational noise exposure.

*

Participants with normal hearing as the reference (n = 6,848). †Per SD increase.

Interaction Between PRS and Glycemic Traits on the Risk of HL

As shown in Supplementary Fig. 2, an elevated PRS had a linear dose-responsive association with a higher risk of overall HL (P for overall <0.001; P for nonlinearity = 0.414). When the associations of FPG and HbA1c levels with HL risks and its severity stratified by PRS were investigated (Fig. 1), the per-SD increase in FPG and HbA1c levels for both was linked to a 10% higher risk of overall HL among participants in high PRS group, and the corresponding increases for moderate to severe HL were 17% and 16%, respectively. Similarly, higher FPG and HbA1c levels were related to increased risks of overall HL in a linear dose-response manner among the high PRS group (P for overall = 0.009 and 0.029; P for nonlinearity = 0.163 and 0.992, respectively) (Supplementary Fig. 3). Furthermore, T2D was also associated with higher HL risk and its severity among participants with high PRS, with OR (95% CI) of 1.40 (1.15, 1.69) for overall HL, 1.31 (1.07, 1.60) for mild HL, and 1.65 (1.26, 2.17) for moderate to severe HL. However, no significant association was observed in participants with low or intermediate PRS.

Figure 1

Association of glycemic traits with the risk of HL and its severity stratified by PRS groups. Models adjusted for age, sex, overweight/obesity, education, marital status, physical activity, smoking, drinking, night sleep duration, napping duration, hypertension, dyslipidemia, shift work, and occupational noise exposure. N in mild HL or moderate to severe HL group indicates the numbers of individuals having mild HL or moderate to severe HL (case).

Figure 1

Association of glycemic traits with the risk of HL and its severity stratified by PRS groups. Models adjusted for age, sex, overweight/obesity, education, marital status, physical activity, smoking, drinking, night sleep duration, napping duration, hypertension, dyslipidemia, shift work, and occupational noise exposure. N in mild HL or moderate to severe HL group indicates the numbers of individuals having mild HL or moderate to severe HL (case).

Close modal

Additionally, significant multiplicative interactions were found between PRS and HbA1c level on the risks of overall HL, mild HL, and moderate to severe HL (P for interaction = 9.34 × 10−4, 0.006, and 0.017), while interaction was only found between PRS and FPG level on the risk of moderate to severe HL (P for interaction = 5.88 × 10−5). Besides, we also observed a borderline but nonsignificant interaction between PRS and T2D on moderate to severe HL risk (P for interaction = 0.070).

Joint Association of PRS Group With Diabetes on the Risk of HL

We also examined the joint associations of the PRS group with diabetes on the risk of HL and its severity, in which participants who had the low PRS and normoglycemia were considered as the reference group (Table 3). In general, the estimates of diabetes with HL risks were on the rising trends with the increasing PRS. In participants with T2D, the HL risks were the strongest in those in the high PRS group, with an estimate (95% CI) of 2.00 (1.66, 2.41) for overall HL, 1.87 (1.53, 2.29) for mild HL, and 2.40 (1.83, 3.16) for moderate to severe HL. Besides, there was evidence of significant additive interactions between the high PRS and T2D on the overall HL risk (RERI 0.45, 95% CI 0.08, 0.81; AP 0.22, 95% CI 0.06, 0.39) and moderate to severe HL risk (RERI 0.75, 95% CI 0.19, 1.31; AP 0.32, 95% CI 0.11, 0.54). However, no significant additive interactions were found between high PRS and prediabetes on the risks of HL, indicating the combined effects of these two factors were approximate to the sum of the estimates related to each factor alone (95% CI of RERI and AP including 0).

Table 3

Joint associations and additive interactions of weighted PRS with diabetes on the risk of HL*

VariablesJoint associationsAdditive interactions (high PRS)
No.Low PRSNo.Intermediate PRSNo.High PRSRERI (95% CI)AP (95% CI)
Overall HL 4,107  4,248  4,920    
 Diabetes         
  Normal 1,107 1.00 (Reference) 1,105 1.15 (0.96, 1.37) 1,331 1.47 (1.23, 1.74)   
  Prediabetes 2,014 1.15 (0.99, 1.34) 2,122 1.24 (1.06, 1.44) 2,503 1.63 (1.40, 1.89) 0.01 (−0.27, 0.29) 0.01 (−0.17, 0.18) 
  T2D 986 1.09 (0.91, 1.31) 1,021 1.36 (1.14, 1.63) 1,086 2.00 (1.66, 2.41) 0.45 (0.08, 0.81) 0.22 (0.06, 0.39) 
Mild HL† 1,346  1,414  1,872    
 Diabetes         
  Normal 324 1.00 (Reference) 331 1.12 (0.93, 1.36) 465 1.46 (1.21, 1.75)   
  Prediabetes 673 1.18 (1.00, 1.39) 720 1.22 (1.04, 1.44) 958 1.63 (1.39, 1.93) 0.01 (−0.27, 0.30) 0.01 (−0.18, 0.19) 
  T2D 349 1.12 (0.93, 1.36) 363 1.31 (1.08, 1.60) 449 1.87 (1.53, 2.29) 0.27 (−0.08, 0.62) 0.15 (−0.04, 0.35) 
Moderate to severe HL† 482  543  770    
 Diabetes         
  Normal 120 1.00 (Reference) 124 1.22 (0.92, 1.61) 176 1.50 (1.14, 1.98)   
  Prediabetes 229 1.06 (0.83, 1.36) 256 1.28 (1.01, 1.64) 379 1.60 (1.26, 2.03) 0.12 (−0.32, 0.55) 0.07 (−0.19, 0.32) 
  T2D 133 1.00 (0.75, 1.33) 163 1.49 (1.14, 1.96) 215 2.40 (1.83, 3.16) 0.75 (0.19, 1.31) 0.32 (0.11, 0.54) 
VariablesJoint associationsAdditive interactions (high PRS)
No.Low PRSNo.Intermediate PRSNo.High PRSRERI (95% CI)AP (95% CI)
Overall HL 4,107  4,248  4,920    
 Diabetes         
  Normal 1,107 1.00 (Reference) 1,105 1.15 (0.96, 1.37) 1,331 1.47 (1.23, 1.74)   
  Prediabetes 2,014 1.15 (0.99, 1.34) 2,122 1.24 (1.06, 1.44) 2,503 1.63 (1.40, 1.89) 0.01 (−0.27, 0.29) 0.01 (−0.17, 0.18) 
  T2D 986 1.09 (0.91, 1.31) 1,021 1.36 (1.14, 1.63) 1,086 2.00 (1.66, 2.41) 0.45 (0.08, 0.81) 0.22 (0.06, 0.39) 
Mild HL† 1,346  1,414  1,872    
 Diabetes         
  Normal 324 1.00 (Reference) 331 1.12 (0.93, 1.36) 465 1.46 (1.21, 1.75)   
  Prediabetes 673 1.18 (1.00, 1.39) 720 1.22 (1.04, 1.44) 958 1.63 (1.39, 1.93) 0.01 (−0.27, 0.30) 0.01 (−0.18, 0.19) 
  T2D 349 1.12 (0.93, 1.36) 363 1.31 (1.08, 1.60) 449 1.87 (1.53, 2.29) 0.27 (−0.08, 0.62) 0.15 (−0.04, 0.35) 
Moderate to severe HL† 482  543  770    
 Diabetes         
  Normal 120 1.00 (Reference) 124 1.22 (0.92, 1.61) 176 1.50 (1.14, 1.98)   
  Prediabetes 229 1.06 (0.83, 1.36) 256 1.28 (1.01, 1.64) 379 1.60 (1.26, 2.03) 0.12 (−0.32, 0.55) 0.07 (−0.19, 0.32) 
  T2D 133 1.00 (0.75, 1.33) 163 1.49 (1.14, 1.96) 215 2.40 (1.83, 3.16) 0.75 (0.19, 1.31) 0.32 (0.11, 0.54) 

Data are presented as the OR (95% CI), unless indicated otherwise.

*

Adjusted for age, sex, overweight/obesity, education, marital status, physical activity, smoking, drinking, night sleep duration, napping duration, hypertension, dyslipidemia, shift work and occupational noise exposure. †Numbers of individuals with mild HL or moderate to severe HL (case).

Sensitivity Analyses

Supplementary Table 3 shows the results of sensitivity analyses after excluding participants who were taking antidiabetes medications. No substantial changes were found in the associations of FPG and HbA1c levels with the risks of HL and its severity in the multivariate analyses. Additionally, similar results regarding the multiplicative interactions between the unweighted PRS and FPG or HbA1c and additive interactions between the unweighted PRS and T2D on HL risk and its severity are presented in Supplementary Fig. 4 and Supplementary Table 4.

Potential Value Incorporating Glycemic Traits and PRS in HL Discrimination and Reclassification

As shown in Supplementary Table 5, a significant and slight increase in the C statistics was observed when PRS and PRS × glycemic traits were added into models including conventional risk factors, but no significant improvement with addition of glycemic traits alone. In terms of net reclassification index and integrated discrimination index, no significant reclassification was observed, indicating that the addition of glycemic traits and PRS did not meaningfully improve HL risk prediction.

Based on middle-aged and older adults from the DFTJ cohort, we found that increased FPG and HbA1c levels were dose-responsively related to higher HL risk and its severity. Similarly, T2D was linked to a higher risk of HL and the estimated odds increasing with HL severity. Particularly, FPG, HbA1c, and T2D were significantly associated with elevated risks of HL and its severity among participants with high PRS but not in those with low or intermediate PRS. We also found a multiplicative interaction between PRS and HbA1c level and an additive interaction between high PRS and T2D on the risk of HL in the current study.

Epidemiological evidence showed that glycemic traits were associated with the risk of HL and its severity. Consistent with our findings, a positive association between FPG level and the risk of HL was observed among 16,799 Koreans (9). Moreover, higher HbA1c level was linked to an increased risk of HL among 831 Japanese aged >65 years, although the HL was defined at a single frequency of 1,000 Hz or 4,000 Hz (10). Several previous studies reported that individuals with diabetes had a 1.16- to 1.85-fold greater risk of HL (11–16); however, two additional studies reported no significant findings on the relationship of diabetes with HL risks (17,18). Limited evidence was available on the positive relationship between diabetes and HL severity (19,20). A NHANES (1999–2004) study showed a significant association between diabetes and HL severity in 4,471 U.S. adults aged 20–69 years (19). Another study conducted in Korea presented that impaired fasting glucose was related to high-frequency mild HL only among men <70 years old but not to moderate to severe HL (20). However, Yang et al. (21) found no significant relationship of FPG or diabetes with HL severity among 2,351 rural adults in China. These heterogeneous results may be partly attributed to the characteristics of the included participants, self-reported HL or HL at a different frequency, self-reported diabetes, or a relatively small sample size.

Additionally, Kvestad et al. (22) proposed that genetic effects could account for 36% of HL, emphasizing the significance of genetic predisposition. Previous studies have developed hearing trait-PRS through genome-wide association study (GWAS) summary statistics (29), or computed PRS using GWAS data from UK Biobank’s self-reported HL, as a surrogate of hereditary susceptibility (30,31). However, none of these studies explored the joint association between PRS and glycemic traits with the risk of HL. Our recent study observed a significant multiplicative interaction between sleep duration or bedtime and PRS on the risk of HL (27), in which the HL-related PRS was calculated by using 37 risk variants from a latest GWAS meta-analysis (23). Although the variants selected for constructing the PRS were derived from individuals of European descent, this GWAS meta-analysis had the largest sample size (n = 723,266) among all the currently available GWAS, and a higher proportion of interethnic loci overlap would be expected as sample sizes increased (32). We therefore calculated the PRS using the same 37 risk variants with external weights and further assessed the interaction of PRS and glycemic traits on the risk of HL in the current study. As expected, we found PRS was related to a higher risk of HL in a liner dose-response manner. Moreover, the association of HbA1c level with HL risk was stronger with increasing PRS, indicating the combined effect of high PRS and high HbA1c level on HL risk was greater than the product of their individual effects. Furthermore, we presented that the risk of HL among participants with both T2D and high genetic susceptibility was higher than the sum of the risk of HL related to each factor separately, providing evidence on a significant interaction on an additive scale between them. The findings could not, however, be directly compared with established evidence due to a lack of corresponding studies.

It is worth noting that several previous studies have explored the interaction between diabetes and genetic predisposition on other health outcomes rather than HL. For instance, Jacobs et al. (33) found both multiplicative and additive interactions between the PRS and diabetes on Parkinson disease in UK Biobank. Another study performed in three cohorts (Genome alcohol-associated cirrhosis [GenomALC-1], GenomALC-2, and UK Biobank) presented combined effects of a high PRS and diabetes on alcohol-related cirrhosis with no evidence of interaction (34).

Notably, the current study is the first to assess the interactions between genetic susceptibility and FPG or HbA1c levels and T2D in relation to the risks of HL, which indicated that participants with a high genetic predisposition should pay more attention to glycemic traits. From a public health perspective, the findings in the current study could help to identify susceptible individuals among the middle-aged and older population who may benefit most from the control of glycemic traits. Importantly, controlling FPG and HbA1c levels, or strengthening T2D management among those with high level of PRS would be of great importance for the control of moderate to severe HL. Although the inclusion of PRS and PRS × glycemic traits resulted in statistically significant C statistics, indicating some additional predictive value, this did not lead to significant improvements in risk reclassification. It suggested that the integration of PRS and glycemic traits only marginally enhanced the differentiation between individuals at high and low risk for HL compared with using traditional risk factors alone. Therefore, among middle-aged and older Chinese individuals, those with a high genetic susceptibility should pay particular attention to managing both their glycemia traits and conventional risk factors to control HL.

The potential mechanisms underlying the association between T2D and the risk of HL have been elucidated in previous studies. Animal models reported that functional decline in the cochlea of diabetic rats, including moderate degenerative changes in the marginal cells, could alter the cochlear K+ cycling route and the maintenance of endocochlear potentials (35). Moreover, spiral ganglion cell loss and outer hair cell degeneration were found in an animal model of T2D and obesity (36). On the other hand, glucose transporters, insulin signaling components, and insulin receptors were present in the sensory receptors and supporting cells of the cochlea, stria vascularis, and spiral ligaments (37,38), which could explain that people with glucose metabolism disorder are more likely to have inner ear function impairment and subsequent HL. Additionally, inner ear function in patients with diabetes may also be affected by microvascular and neuropathic complications such as microangiopathic changes in the capillaries of the stria vascularis, narrowing of the internal auditory artery, and demyelination of the auditory nerve (39).

Several limitations should be noticed in the current study. Firstly, reverse causality was possible regarding the relationship between glycemic traits and HL because of the cross-sectional design.

Secondly, unavoidable selection bias due to the differences between the participants included and excluded should be considered. Participants with audiometric testing tended to have a higher proportion of hypertension, dyslipidemia, and T2D than those without, which may overestimate the associations of glycemic traits with the risk of HL and its severity. However, we adjusted for these potential confounders in the multivariable models, which might reduce the bias to some extent.

Thirdly, the variants constructing the PRS of HL in the current study were only derived from individuals of European descent, while the Asian-specific single nucleotide polymorphism loci were not included due to the unavailability of related GWAS data. However, this GWAS meta-analysis had the largest sample size in the currently available GWAS of HL, which may cover a higher proportion of overlapping interethnic loci (32).

Fourthly, although we adjusted for a few confounding factors, we did not account for residual and unmeasured confounders, such as nutritional deficiencies and infections, since the data were unavailable.

Lastly, since the current study only included the middle-aged and elderly retirees from DMC, care must be taken when generalizing the results to other populations. Results from multicenter studies are needed to replicate and validate the current findings.

In summary, our study supports the dose-response associations of HbA1c level with higher HL risk and its severity, and PRS modifies these associations on a multiplicative scale. Additionally, T2D was significantly related to a higher risk of HL, with estimates increasing across HL severity, and a high PRS may interact additively with T2D on HL risk. The findings of our study suggest that management of glycemic traits combined with conventional risk factors might be a better practice to control HL among middle-aged and older adults, especially among those with high susceptibility.

This article contains supplementary material online at https://doi.org/10.2337/figshare.25639779.

Acknowledgments. The authors thank the Dongfeng-Tongji cohort participants, staff, and investigators for their contributions to the study.

Funding. This work was supported by the National Key Research and Development Program of China (2023YFC2506501) and the Key Research and Development Program of Hubei Province (2022BCA046).

The funders had no role in the design and conduct of the study, collection, management, analysis, and interpretation of the data, preparation, review, or approval of the manuscript, and decision to submit the manuscript for publication.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.H. performed the statistical analysis and wrote the first draft of the manuscript. Y.H., W.K., and X.Z. were involved in the conception, design, and conduct of the study. Z.W. and M.H. handled methodology, investigation, and acquisition of data. H.Z., X.L., M.L., L.Y., and Y.Z. provided statistical expertise and contributed to critical revision for important intellectual content of the manuscript. All authors contributed to the interpretation of the results and critical revision of the manuscript and approved the final version of the manuscript. W.K. and X.Z. are the guarantors of this work and, as such, had 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.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Vanita R. Aroda.

1.
GBD 2021 Diabetes Collaborators
.
Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021
.
Lancet
2023
;
402
:
203
234
2.
Gregg
EW
,
Buckley
J
,
Ali
MK
, et al.;
Global Health and Population Project on Access to Care for Cardiometabolic Diseases
.
Improving health outcomes of people with diabetes: target setting for the WHO Global Diabetes Compact
.
Lancet
2023
;
401
:
1302
1312
3.
Samocha-Bonet
D
,
Wu
B
,
Ryugo
DK
.
Diabetes mellitus and hearing loss: a review
.
Ageing Res Rev
2021
;
71
:
101423
4.
GBD 2019 Hearing Loss Collaborators
.
Hearing loss prevalence and years lived with disability, 1990-2019: findings from the Global Burden of Disease Study 2019
.
Lancet
2021
;
397
:
996
1009
5.
GBD 2019 Ageing Collaborators
.
Global, regional, and national burden of diseases and injuries for adults 70 years and older: systematic analysis for the Global Burden of Disease 2019 Study
.
BMJ
2022
;
376
:
e068208
6.
World Health Organization
.
World Report on Hearing
. 3 March
2021
. Accessed 29 September 2023. Available from https://www.who.int/publications/i/item/9789240020481
7.
Sun
HY
,
Hu
YJ
,
Zhao
XY
, et al
.
Age-related changes in mitochondrial antioxidant enzyme Trx2 and TXNIP-Trx2-ASK1 signal pathways in the auditory cortex of a mimetic aging rat model: changes to Trx2 in the auditory cortex
.
FEBS J
2015
;
282
:
2758
2774
8.
Cai
H
,
Han
B
,
Hu
Y
, et al
.
Metformin attenuates the D-galactose-induced aging process via the UPR through the AMPK/ERK1/2 signaling pathways
.
Int J Mol Med
2020
;
45
:
715
730
9.
Lee
HY
,
Choi
YJ
,
Choi
HJ
, et al
.
Metabolic syndrome is not an independent risk factor for hearing impairment
.
J Nutr Health Aging
2016
;
20
:
816
824
10.
Michikawa
T
,
Mizutari
K
,
Saito
H
,
Takebayashi
T
,
Nishiwaki
Y
.
Glycosylated hemoglobin level is associated with hearing impairment in older Japanese: the Kurabuchi Study
.
J Am Geriatr Soc
2014
;
62
:
1231
1237
11.
Cruickshanks
KJ
,
Dhar
S
,
Dinces
E
, et al
.
Hearing impairment prevalence and associated risk factors in the Hispanic Community Health Study/Study of Latinos
.
JAMA Otolaryngol Head Neck Surg
2015
;
141
:
641
648
12.
Kim
MB
,
Zhang
Y
,
Chang
Y
, et al
.
Diabetes mellitus and the incidence of hearing loss: a cohort study
.
Int J Epidemiol
2017
;
46
:
717
726
13.
Gupta
S
,
Eavey
RD
,
Wang
M
,
Curhan
SG
,
Curhan
GC
.
Type 2 diabetes and the risk of incident hearing loss
.
Diabetologia
2019
;
62
:
281
285
14.
Mitchell
P
,
Gopinath
B
,
McMahon
CM
, et al
.
Relationship of Type 2 diabetes to the prevalence, incidence and progression of age-related hearing loss
.
Diabet Med
2009
;
26
:
483
488
15.
Gong
R
,
Hu
X
,
Gong
C
, et al
.
Hearing loss prevalence and risk factors among older adults in China
.
Int J Audiol
2018
;
57
:
354
359
16.
Wang
W
,
Zhang
C
,
Meng
R
, et al
.
Sex-specific associations between diabetes mellitus and hearing loss in the middle-aged and elderly individuals: a national cohort study of Chinese adults
.
Endocr Pract
2022
;
28
:
357
363
17.
Cruickshanks
KJ
,
Nondahl
DM
,
Dalton
DS
, et al
.
Smoking, central adiposity, and poor glycemic control increase risk of hearing impairment
.
J Am Geriatr Soc
2015
;
63
:
918
924
18.
Hara
K
,
Okada
M
,
Takagi
D
, et al
.
Association between hypertension, dyslipidemia, and diabetes and prevalence of hearing impairment in Japan
.
Hypertens Res
2020
;
43
:
963
968
19.
Bainbridge
KE
,
Hoffman
HJ
,
Cowie
CC
.
Diabetes and hearing impairment in the United States: audiometric evidence from the National Health and Nutrition Examination Survey, 1999 to 2004
.
Ann Intern Med
2008
;
149
:
1
10
20.
Seo
M
,
Lee
YS
,
Moon
SS
.
Association of hearing impairment with insulin resistance, β-cell dysfunction and impaired fasting glucose before onset of diabetes
.
Diabet Med
2016
;
33
:
1275
1282
21.
Yang
D
,
Liu
J
,
Yang
Q
, et al
.
Hearing impairment prevalence and risk factors among adults in rural China: a population-based cross-sectional study
.
Postgrad Med
2021
;
133
:
369
376
22.
Kvestad
E
,
Czajkowski
N
,
Krog
NH
,
Engdahl
B
,
Tambs
K
.
Heritability of hearing loss
.
Epidemiology
2012
;
23
:
328
331
23.
Trpchevska
N
,
Freidin
MB
,
Broer
L
, et al.;
Estonian Biobank Research Team
.
Genome-wide association meta-analysis identifies 48 risk variants and highlights the role of the stria vascularis in hearing loss
.
Am J Hum Genet
2022
;
109
:
1077
1091
24.
Wang
F
,
Zhu
J
,
Yao
P
, et al
.
Cohort profile: the Dongfeng-Tongji cohort study of retired workers
.
Int J Epidemiol
2013
;
42
:
731
740
25.
Fang
Q
,
Wang
Z
,
Zhan
Y
, et al
.
Hearing loss is associated with increased CHD risk and unfavorable CHD-related biomarkers in the Dongfeng-Tongji cohort
.
Atherosclerosis
2018
;
271
:
70
76
26.
American Diabetes Association
.
Diagnosis and classification of diabetes mellitus
.
Diabetes Care
2013
;
36
(
Suppl. 1
):
S67
S74
27.
Liu
M
,
Zhang
H
,
Wang
Z
, et al
.
Independent and combined associations of sleep duration, bedtime, and polygenic risk score with the risk of hearing loss among middle-aged and old Chinese: the Dongfeng-Tongji Cohort Study
.
Research (Wash D C)
2023
;
6
:
0178
28.
Huang
Y
,
Zhu
M
,
Ji
M
, et al
.
Air pollution, genetic factors, and the risk of lung cancer: a prospective study in the UK Biobank
.
Am J Respir Crit Care Med
2021
;
204
:
817
825
29.
Bhatt
IS
,
Ramadugu
SK
,
Goodman
S
, et al
.
Polygenic risk score-based association analysis of speech-in-noise and hearing threshold measures in healthy young adults with self-reported normal hearing
.
J Assoc Res Otolaryngol
2023
;
24
:
513
525
30.
Cherny
SS
,
Livshits
G
,
Wells
HRR
, et al
.
Self-reported hearing loss questions provide a good measure for genetic studies: a polygenic risk score analysis from UK Biobank
.
Eur J Hum Genet
2020
;
28
:
1056
1065
31.
Wang
J
,
Lange
K
,
Sung
V
,
Morgan
A
,
Saffery
R
,
Wake
M
.
Association of polygenic risk scores for hearing difficulty in older adults with hearing loss in mid-childhood and midlife: a population-based cross-sectional study within the longitudinal study of Australian children
.
JAMA Otolaryngol Head Neck Surg
2023
;
149
:
204
211
32.
Takeuchi
F
,
Akiyama
M
,
Matoba
N
, et al.;
International Genomics of Blood Pressure (iGEN-BP) Consortium
.
Interethnic analyses of blood pressure loci in populations of East Asian and European descent
.
Nat Commun
2018
;
9
:
5052
33.
Jacobs
BM
,
Belete
D
,
Bestwick
J
, et al
.
Parkinson’s disease determinants, prediction and gene-environment interactions in the UK Biobank
.
J Neurol Neurosurg Psychiatry
2020
;
91
:
1046
1054
34.
Whitfield
JB
,
Schwantes-An
TH
,
Darlay
R
, et al.;
GenomALC Consortium
.
A genetic risk score and diabetes predict development of alcohol-related cirrhosis in drinkers
.
J Hepatol
2022
;
76
:
275
282
35.
Meyer zum Gottesberge
AM
,
Massing
T
,
Sasse
A
,
Palma
S
,
Hansen
S
.
Zucker diabetic fatty rats, a model for type 2 diabetes, develop an inner ear dysfunction that can be attenuated by losartan treatment
.
Cell Tissue Res
2015
;
362
:
307
315
36.
Lee
HS
,
Kim
KR
,
Chung
WH
,
Cho
YS
,
Hong
SH
.
Early sensorineural hearing loss in ob/ob mouse, an animal model of type 2 diabetes
.
Clin Exp Otorhinolaryngol
2008
;
1
:
211
216
37.
Yamahara
K
,
Yamamoto
N
,
Nakagawa
T
,
Ito
J
.
Insulin-like growth factor 1: A novel treatment for the protection or regeneration of cochlear hair cells
.
Hear Res
2015
;
330
:
2
9
38.
Huerzeler
N
,
Petkovic
V
,
Sekulic-Jablanovic
M
,
Kucharava
K
,
Wright
MB
,
Bodmer
D
.
Insulin receptor and glucose transporters in the mammalian cochlea
.
Audiol Neurotol
2019
;
24
:
65
76
39.
Akinpelu
OV
,
Ibrahim
F
,
Waissbluth
S
,
Daniel
SJ
.
Histopathologic changes in the cochlea associated with diabetes mellitus--a review
.
Otol Neurotol
2014
;
35
:
764
774
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.