Hearing loss is a major global public health problem. Approximately, one-fifth of the global population currently suffer from hearing loss(1). In America, nearly a quarter of people aged 12 years and over have hearing loss, including mild and unilateral hearing loss(Reference Goman and Lin2). The WHO has estimated that hearing loss will be one of the main causes of disease burden globally by 2030(3). Hearing loss may be associated with a variety of diseases. Evidence from epidemiological studies showed that hearing loss is related to the increased risks of depressive symptoms, falls, total mortality and heart disease mortality(Reference Scinicariello, Przybyla and Carroll4–Reference Feng, Li and Cheng6). The traditional risk factors, including age, noise exposure, family history of hearing loss, exposure to ototoxic medications, smoking and diabetes, could only partially explain the causes of hearing loss(Reference Nieman and Oh7). Recently, accumulating studies have suggested that diet is related to hearing loss, and this effect can be attributed to specific dietary patterns or some special bioactive compounds, such as PUFA, vitamin A, isoflavone, riboflavin, niacin and retinol(Reference Choi, Ahn and Moon8–Reference Yévenes-Briones, Caballero and Struijk11).
Coffee is one of the most popular beverages in the world. Coffee and its compounds, such as caffeine, trigonelline, polyphenols and chlorogenic acid, have various impacts on human health(Reference Saud and Salamatullah12). Coffee trigonelline has anti-microbial, anti-carcinogenic and anti-hyperglycemic effects(Reference Nugrahini, Ishida and Nakagawa13). Coffee chlorogenic acid has some anti-cancer effects(Reference Hayakawa, Ohishi and Miyoshi14). Caffeine has protective effect on neurodegenerative disease due to its strong antioxidant, anti-inflammatory and adenosine receptor antagonist properties(Reference Kolahdouzan and Hamadeh15). It may also be related to increased risk of fractures and decreased sleep quality(Reference Watson, Coates and Kohler16,Reference Asoudeh, Bagheri and Larijani17) . Additionally, polyphenols, such as caffeic acid and caffeic acid phenethyl ester, have antioxidant effects and prevent hearing loss(Reference Choi, Kim and Rah18,Reference Park, Im and Chang19) . However, caffeine are often considered a cause of tinnitus(Reference Crummer and Hassan20). Large population-based studies on the association between coffee consumption and hearing loss are limited.
Currently, study about the effect of coffee consumption on hearing loss is scarce, and the results remain controversial. Two previous population-based studies have suggested a negative association between coffee consumption frequency and hearing loss(Reference Lee, Jung and Jang40,Reference Machado-Fragua, Struijk and Yévenes-Briones41) . Contrary, previous animals studies have shown that caffeine in coffee can interfere with hearing recovery after acoustic overstimulation events(Reference Zawawi, Bezdjian and Mujica-Mota31,Reference Mujica-Mota, Gasbarrino and Rappaport32) . Moreover, a previous study based on the National Health and Nutrition Examination Survey (NHANES) has reported no significant association between urinary caffeine metabolites and hearing thresholds(Reference Long and Tang39). Therefore, in this study, we have investigated the association of coffee consumption with the hearing loss in adults from America according to the NHANES database.
Methods
Study design and participants
The data in this study were from the NHANES database, which is a national survey administered by the National Center for Health Statistics (NCHS). This survey investigated about 10 000 representative samples of the general American population per cycle using a complex, multistage, probability sampling design. The data from the 2003–2006 were used in our study, because the information on coffee consumption and audiometry data of adult participants were collected in the same period. In the 2003–2006 NHANES, a total of 4923 subjects participated in the audiometry component. Among 1735 subjects were excluded because they lacked audiometry test data (n 451) and coffee consumption data (n 1284). Of the remaining 3188 participants, additional 1294 were eliminated because they aged < 20 years (adolescents aged 12–19 years in NHANES). Finally, 1894 subjects were recruited for analyses in this study (Fig. 1).
Flow chart of the selection process
Assessment of coffee consumption
Coffee consumption frequency was calculated by the FFQ, which was developed by the National Institutes of Health, National Cancer Institute (NCI) based on the NCI Diet History Questionnaire. Participants were asked to review and fill in their coffee consumption in the past 12 months. The question on coffee consumption frequency in the questionnaire was: ‘How many cups of caffeinated or decaffeinated coffee did you drink’. The possible responses were none, ≤ 1 cup/m, 1–3 cups/m, 1 cup/w, 2–4 cups/w, 5–6 cups/w, 1 cup/d, 2–3 cups/d, 4–5 cups/d, ≥ 6 cups/d. If they select other option than ‘none’, participant need to answer the following question: ‘How often was you drank the decaffeinated coffee’. The possible responses were ‘almost never or never, about 1/4 of the time, about 1/2 of the time, about 3/4 of the time, almost always or always’. In this survey, a cup of coffee was 8oz according to the Measuring Guides for the Dietary Recall Interview(Reference Nerurkar, Gandhi and Chen26).
In present analysis, total coffee consumption frequency was categorised into five groups: none, ≤ 1 cup/d, 1 cup/d, 2–3 cups/d and ≥ 4 cups/d. Then coffee consumption frequency data were further converted to quantitative data (e.g. 1–3 cups/m was converted to 0·07 cups/d). The prevalence of hearing loss related to caffeinated and decaffeinated coffee consumption was also investigated.
Audiometric measurement
All hearing measurements were conducted by a well-trained physicians in a dedicated, sound-isolating room at a mobile examination centre. The test equipment included AD226 audiometer (Interacoustics AS, Assens, Denmark), TDH-39 standard headphones (Interacoustics AS, Assens, Denmark) and EARtone 3A insert earphones (Etymotic Research, Elk Grove Village, IL). The hearing threshold for each ear was measured at frequencies of 500, 1000, 2000, 3000, 4000, 6000 and 8000 Hz across an intensity range of –10 to 120 dB using the modified Hughson-Westlake procedure and invoking the automated testing mode of the audiometer. More details of audiometric measures are displayed on NHANES website. In our study, speech-frequency hearing loss (SFHL) was defined as pure-tone average of hearing thresholds at 500, 1000, 2000 and 4000 Hz is > 25 dB in either ear, and high-frequency hearing loss (HFHL) was defined as pure-tone average of hearing thresholds at 3000, 4000 and 6000 Hz is > 25 dB in either ear(Reference Kabagambe, Lipworth and Labadie27).
Other variables
Covariates were obtained from the questionnaire survey, including general characteristics (age, gender and ethnicity), lifestyles (smoking status, and drinking status), noise exposure and history of diseases (hypertension and diabetes mellitus). Ethnicity was classified as non-Hispanic White, non-Hispanic Black and other. Smoking status was categorised as never smoker, former smoker and current smoker. Drinking status was categorised as never drinker, low to moderate drinker (drinking < 1 drink/d in women and < 2 drinks/d in men) and heavy drinker (≥ 1 drink/d in female and ≥ 2 drinks/d in male)(Reference Chen, Ye and Zhang28). BMI status was classified as BMI < 25, BMI 25–30 and BMI ≥ 30(Reference Dreimüller, Lieb and Tadić29). The question on non-occupational noise exposure was: ‘Outside of a job, have you ever been exposed to steady loud noise or music for 5 or more hours a week? This is noise so loud that you have to raise your voice to be heard’. The possible responses were ‘yes’ or ‘no’. The question on occupational noise exposure was: ‘have you ever had a job where you were exposed to loud noise for 5 or more hours a week (you had to raise your voice to be heard)’. The possible responses were ‘yes’ or ‘no’. In addition, diagnoses of diabetes mellitus and hypertension and history of ear infection were self-reported by the subjects.
Statistical analyses
Given the design of complex, multistage, probability sampling in the NHANES, we implemented sampling weight, cluster and strata in the analysis. Continuous variables were expressed as mean and sd for the normal distribution and as median (P25, P75) for skewed distribution and analysed by Student’s t tests. Categorical variables were expressed as frequency (%) and analysed by χ 2 test or Wilcoxon rank-sum test.
Multivariate logistic regression models were applied to assess the relationships between coffee consumption frequency and the prevalence of SFHL and HFHL. The ‘None’ group were considered as the reference groups. Covariates including ethnicity, BMI, ear infection, occupational noise exposure, non-occupational noise exposure, smoking status, drinking status, hypertension and diabetes mellitus were adjusted. The trend test was conducted by taking the median of each coffee consumption frequency group as a continuous variable in these models. We also transformed coffee consumption frequency into continuous variable to explore the linear dose–response relationship. We also conducted subgroup analyses stratified by demographic characteristics (including age (20–64 years, ≥ 65 years), ethnicity (non-Hispanic White, non-Hispanic Black and other race) and sex (male and female)), noise exposure source (including Yes (at work exposure Yes and/or outside work exposure Yes), No (at work exposure No and outside work exposure No), noise exposure unknown (outside work exposure No and at work exposure data missing)) and coffee type (caffeinated coffee, decaffeinated coffee, both). Moreover, we created forest plots to estimate the OR (95 % Cl) of hearing loss related to a 1-cup/d increment for the different types of coffee separately in men and women. Furthermore, restricted cubic splines (RCS) regression model was conducted to further explore the dose–response relationship of coffee consumption with the prevalence of SFHL and HFHL in the multivariable-adjusted binary logistic regression analyses for sex, ethnicity and noise exposure status separately, with four knots of at the 5th, 35th, 65th and 95th percentiles. Adjusted factors were consistent with multivariate logistic regression model. P for non-linearity < 0·05 suggested a non-linear association; otherwise, a linear association was indicated. RCS analyses was conducted using SAS macro program %RCS_Reg(Reference Desquilbet and Mariotti30). All statistical analyses were performed using SAS software (version 9.4; SAS Institute).
Results
Characteristics of study population
As shown in Table 1, a total of 1894 participants were included. Compared to non-coffee drinkers, those who with more consumption were older and were more likely to be non-Hispanic White and current smokers and heavy drinkers. Also, they had a lower proportion of BMI ≥ 30 and were more likely to have a diagnosis of hypertension and hearing loss; by contrast, they were more exposed to occupational noise exposure. No statistical differences between different coffee consumption frequency groups in sex, BMI, ear infection, non-occupational noise exposure and diabetes mellitus were observed (all P > 0·05).
Characteristics of study subjects by coffee consumption
SFHL, speech-frequency hearing loss; HFHL, high-frequency hearing loss.
P values from χ 2 test or Wilcoxon rank-sum test (categorical categories) and Student’s t tests (continuous covariates).
* P < 0·05.
** P < 0·01.
*** P < 0·001.
Association between coffee consumption and hearing loss risk
As shown in Table 2, there was no correlation between coffee consumption and the prevalence of SFHL and HFHL in all groups in the full-adjusted model. A positive trend and association of coffee consumption frequency with the prevalence of SFHL was found in crude model (P trend = 0·001), but this trend has insignificant after full adjustments (P trend = 0·25). Besides, a positive trend and association of coffee consumption frequency with the prevalence of HFHL in crude model was found (P trend < 0·05), while only the trend remained after full adjustments (P trend < 0·05)
OR and 95 % CI of coffee consumption for hearing loss
SFHL, speech-frequency hearing loss; HFHL, high-frequency hearing loss.
* P trend < 0·05.
† Adjusted for age, sex, ethnicity, ear infection, occupational noise exposure, non-occupational noise exposure, smoking status, drinking status, hypertension, diabetes mellitus and BMI.
‡ P < 0·05.
Subgroup analyses
The results of subgroup analyses were shown in Table 3, online Supplementary Table S1 and S2. In non-Hispanic White subgroup, only participants who consumed ≥ 4 cups/d had a higher prevalence of SFHL compared to non-coffee drinker group (OR: 1·87; 95 % CI: 1·003, 3·47; P = 0·049, Table 3). Nevertheless, no trends and association of coffee consumption frequency with the prevalence of SFHL in subgroups of age (20–64 years, ≥ 65), ethnicity (non-Hispanic Black and other race) and sex (male and female) were observed (P trend > 0·05, Table 3). Besides, positive trends of coffee consumption frequency with the prevalence of HFHL in subgroups of age (20–64 years), ethnicity (non-Hispanic White) and sex (male and female) were observed (P trend < 0·05, Table 3). Moreover, in noise exposure subgroup, a positive trend of coffee consumption frequency with the prevalence of SFHL in Model 1 was observed (P trend = 0·048, online Supplementary Table S1). Similarly, the positive trend of coffee consumption frequency with the prevalence of HFHL in Model 1 was also found in subgroups of loud noise exposure (yes and no). In addition, no significant correlations were also found between coffee consumption frequency and the prevalence of SFHL and HFHL in all coffee type subgroups (online Supplementary Table S2).
OR and 95 % CI of coffee consumption for hearing loss stratified by demographic characteristics
SFHL, speech-frequency hearing loss; HFHL, high-frequency hearing loss.
* P trend < 0·05.
† Adjusted for age, sex, ethnicity, ear infection, occupational noise exposure, non-occupational noise exposure, smoking status, drinking status, hypertension, diabetes mellitus, and BMI.
‡ P < 0·05.
The forest plot analysis of association of coffee consumption with hearing loss
The relationship between coffee consumption and the prevalence of SFHL and HFHL appeared to be more pronounced in male subgroup. However, only in caffeinated coffee subgroup, the association between per 1 cup-increment of daily coffee consumption and HFHL was statistically significant in male (Fig. 2).
OR (95 % CI) per 1 cup-increment for the association between the different types of coffee and the prevalence of hearing loss stratified by sex. (a) and (b) speech-frequency hearing loss (SFHL). (c) and (d) High-frequency hearing loss (HFHL). Analyses are adjusted for age, ethnicity, ear infection, occupational noise exposure, non-occupational noise exposure, smoking status, drinking status, hypertension, diabetes mellitus and BMI
Dose–response relationships between coffee consumption and the prevalence of hearing loss
RCS model also showed a linear positive associations of coffee consumption with the prevalence of SFHL (P overall association = 0·02, P nonlinearity = 0·48; Fig. 3 (a)), while a non-liner positive associations of coffee consumption with the prevalence of HFHL was found (P overall association = 0·001, P nonlinearity = 0·001; Fig. 3 (b)). Besides, a positive linear associations of coffee consumption with the prevalence of SFHL were found in the noise exposure subgroup (P overall association = 0·03, P nonlinearity = 0·47; online Supplementary Fig. S1 (a)). Likewise, positive linear associations of coffee consumption with the prevalence of HFHL were found in male subgroup and non-noise exposure subgroup (all P overall association < 0·05, P nonlinearity > 0·05; online Supplementary Fig. S1 (f)–(g)), while positive non-linear associations of coffee consumption with the prevalence of HFHL were found in the age 20–64 years, non-Hispanic White, female and noise exposure subgroup (all P overall association < 0·05, P nonlinearity < 0·05; online Supplementary Fig. S1).
Multivariable-adjusted spline curves of relation between total coffee and the prevalence of hearing loss. (a) SFHL. (b) HFHL. Covariates were age, sex, ethnicity, ear infection, occupational noise exposure, non-occupational noise exposure, smoking status, drinking status, hypertension, diabetes mellitus and BMI. SFHL, speech-frequency hearing loss; HFHL, high-frequency hearing loss
Discussion
In the present study, we assessed the relationship between coffee consumption and hearing loss in American adults aged ≥ 20 years from NHANES 2003–2006 and found that coffee consumption was related to the prevalence of hearing loss among US adults, especially male and participants with noise exposed. A significant trend of coffee consumption frequency with the prevalence of hearing loss was found. These results may be independent of the coffee type or the preparation method.
As a common drink, coffee has attracted increasing attention on its health effects. However, studies about the impact of coffee consumption on hearing loss were limited, and the results were controversial. In this study, we found a positive association of coffee consumption frequency with the prevalence of SFHL and HFHL in US adults. Caffeine is one of the main components of coffee. Caffeine has non-selective adenosine receptor antagonist properties, which interfers hearing recovery after acoustic overstimulation events via antagonising adenosine receptors in Corti organ, lateral wall, spiral ganglion cells and cochlear blood vessels, leading to interruption of cochlear blood reperfusion and the increased production of oxidative stress(Reference Zawawi, Bezdjian and Mujica-Mota31–Reference Sheth, Sheehan and Dhukhwa33). Also, caffeine increases the production and accumulation of Ca in the cochlear hair cells after noise exposure, which activate Ca-dependent isomers and cleave Ca-dependent neurons by phospholipase A2 that lead to the apoptosis of cochlear hair cells(Reference Le Prell, Yamashita and Minami23,Reference Li, Zhao and Jiang34) . Besides, caffeine exacerbates the noise-induced hypoperfusion and ischemia in the cochlea by promoting the reduction of cerebral blood flow and arteriole diameter(Reference Lunt, Ragab and Birch24,Reference Seidman, Quirk and Shirwany35) . Caffeine also exacerbates the physiological increase of corticosterone by altering the hypothalamic–pituitary–adrenocortical axis, thus causing an acute response to noise(Reference Gavrieli, Yannakoulia and Fragopoulou36,Reference Prasher37) . In addition, caffeine caused autophagy and apoptosis in the cochlear hair cells through SGK1/HIF-1α pathway(Reference Tang, Sun and Xu38). However, a previous study based on the NHANES showed that urinary caffeine metabolites were not associated with the changes of hearing thresholds in US adults(Reference Long and Tang39). Contrary to our findings, two previous population-based studies have shown a negative association between coffee consumption frequency and hearing loss(Reference Lee, Jung and Jang40,Reference Machado-Fragua, Struijk and Yévenes-Briones41) . The inconsistent results can be mainly attributed to differences in the study population, the sample size, the definition of hearing loss, the measurement of hearing thresholds, the inclusion and exclusion criteria, and the covariates included in statistical model(Reference Childers and Maggard-Gibbons42).
Of note, our study showed a positive trend and association of coffee consumption frequency with the prevalence of SFHL and HFHL in non-Hispanic Whites, and participants who consumed ≥ 4 cups/d coffee had a 1·87-fold higher prevalence of SFHL than non-coffee drinkers. Similar to our finding, another study based on the NHANES database also showed that the odds of hearing loss are substantially higher in non-Hispanic White Americans than in other ethnic individuals (OR: 2·3; 95 % CI: 1·3, 3·9)(Reference Hoffman, Dobie and Losonczy25). Skin pigmentation, as a marker of melanocytic functioning, may mediate the close relationship of race/ethnicity and hearing loss(Reference Lin, Maas and Chien21). Genes also play an important role in the occurrence and development of hearing loss. It has been reported that the prevalence of hearing loss caused by pathogenic autosomal recessive non-syndromic (ARNS) HI genes varies from race to race, and African Americans/African Americans receive the least impact(Reference Chakchouk, Zhang and Zhang22). In addition, we also found that even after adjusting for covariates related to hearing loss, there was still a sex difference in the relationship between coffee and hearing loss, the association between per 1 cup-increment of daily coffee consumption and HFHL was statistically significant in men, but no significant association was found among women, which was similar to a previous study(Reference Goman and Lin2). In this study, the proportion of male coffee consumers is higher than that of female (78·26 % v. 74·82 %). Coffee drinking is widespread in the USA, and men consumed more coffee(Reference Loftfield, Freedman and Graubard43,Reference Loftfield, Freedman and Dodd44) . Previous studies have shown that men may have a higher hearing threshold(Reference Goman and Lin2). It may be due to the great differences between men and women in brain biochemistry, physiology, structure and function. In physiological structure, the length of the cochlea in men was longer than that in women, which could affect the auditory brainstem responses. Besides, oestrogen also plays a protective role in the cochlear function(Reference Kim, Lee and Carpena45). Oestrogen may play an important role in modulating the pathophysiological mechanisms in the hearing system, and it could enhance the expression of antioxidant superoxide dismutase and decrease apoptosis by upregulating Bcl-2/Bcl-xL and inhibiting the JNK pathway, and it also could inhibit glutamate excitotoxicity to regulate cochlear homoeostasis(Reference Lien and Yang46). Thus, in women, the effect of nutrition on auditory function may not be as relevant as in men. Besides, we found a positive trend and relationship between coffee consumption frequency with the prevalence of HFHL in non-noise exposure subgroup, suggesting that caffeine may have a potential effect on hearing. A previous study reported that caffeine significantly suppressed the compound action potential of the auditory nerve after infusing caffeine into the perilymph compartment(Reference Bobbin47). Noteworthy, we found a positive trend and relationship of coffee consumption frequency with the prevalence of SFHL and HFHL in the noise-exposed participants. Noise-induced hearing loss mainly refers to the apoptosis of cochlear hair cells caused by noise through mechanical damage and metabolic damage (such as oxidative stress damage, Ca2+ overload, and immune and inflammatory damage)(Reference Ding, Yan and Liu48). The interaction between coffee and noise may aggravate the apoptosis of cochlear hair cells. Given the biological plausibility, it will be very meaningful to conduct further studies to establish the potential role of coffee consumption combined with noise exposure on hearing loss.
This study has several advantages. First, the data are from NHANES, which have been implemented in the USA for a long time, and the survey implementation process is rigorous and mature, so the results of this study are relatively reliable. Second, given that the survey design was complex, multistage, probability sampling in the NHANES, we conducted sampling weight, cluster and strata to solve the deviation of variance estimation caused by such clustering data in the statistical analyses. Besides, we explored the association between the prevalence of SFHL and HFHL with coffee consumption frequency. This study also has several limitations. First, since only two circles (2003–2006) in the NHANES have collected the information on coffee consumption frequency and audiometry test of American adults, the sample size was relatively small; thus, the extrapolation of this result to the general population should be more cautious. Secondly, coffee consumption is obtained through questionnaires, which may have participants’ recall bias, and it is difficult to accurately obtain information about coffee consumption, such as individual coffee intake, type of coffee consumption and preparation process. Therefore, the error of coffee consumption between measured exposure and actual exposure cannot be completely eliminated(Reference Satija, Yu and Willett49). Finally, this was an observation study, so causality between coffee consumption and hearing loss could not be shown.
Conclusions
This finding suggested that the positive trend and association of coffee consumption frequency with the prevalence of hearing loss in US adults. And this association was found in non-Hispanic Whites, men, aged 20–64 participants and noise-exposed individuals. This association may be independent of the coffee type or the preparation method. Further cohort studies in larger population are needed to validate these findings, and the underlying mechanism also remains to be elucidated.
Acknowledgements
The authors would like to thank the NHANES participants and the staff members for their contribution of data collection and for making the data publicly available.
Financial Support
Our research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.
Conflict of interest
The authors declare no conflict of interest.
Authorship
S.W. and S.Z. conducted the statistical analyses and wrote the manuscript. All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. W.Y. and Q.C. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Ethics of human subject participation
All data were obtained from secondary sources and available publicly. No protocol approval was necessary.
The datasets analyzed for present study can be found in the NHANES database https://www.cdc.gov/nchs/nhanes. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980023001271
Open access
