Hearing loss is a major health concern affecting over 466 million people worldwide (World Health Organization, 2019). It is an often underestimated chronic condition that negatively impact social, functional, and psychological well-being of a person, and general quality of life (Arlinger, 2003). The disease also places a significant financial burden on society, as it is expected that moderate to severe hearing loss would cost the health-care sector $67-110 billion globally (World Health Organization, 2017). The potential ototoxic effects of chemicals have been reported among workers (Morata, 2007; Sliwińska-Kowalska et al., 2003) and general populations (Choi et al., 2012; Y.-H. Choi and Park, 2017; Park et al., 2010; Shiue, 2015).

Per- and polyfluoroalkyl substances (PFAS) are a group of persistent anthropogenic chemicals, which are widely applied in fire-fighting foams; nonstick cookware, weatherproof clothing, surface protectants, food packaging, and other consumer products. Widespread use and extreme resistance to degradation have resulted in the ubiquitous presence of those chemicals in the general environment. These compounds were detected in 194 of 4864 water supplies, affecting up to 110 million residents in the United States (US) (Environmental Working Group, 2018). PFAS are ubiquitous environmental toxicants to which humans are exposed on a daily basis (Trudel et al., 2008).

As a group of polyhalogenated compounds, PFAS is associated with metabolic disorders (Jiang et al., 2015), endocrine disruption (Jensen and Leffers, 2008), and developmental and neurobehavioral toxicity (Mariussen, 2012). Thyroid systems appears to be particularly vulnerable to disruption by PFAS with a reduction in total thyroxine (T4) concentrations following exposure in humans (Ballesteros et al., 2017; Kim et al., 2018). Animal data suggest that halogenated hydrocarbon-induced hearing loss is the sequelae of thyroid gland dysfunctions caused by these substances (Crofton and Zoeller, 2005; Goldey et al., 1995; Zoeller, 2005). In addition, a recent mouse model detected the existence of peroxisome proliferator-activated receptors (PPARs), ligand-activated transcription factors that regulate gene expression, in diverse structures of the cochlea including both outer hair cells (OHCs) and inner hair cells (IHCs), similar to levels in brain and liver (Sekulic-Jablanovic et al., 2017). Activation of PPARs by PFAS exposure modulate cell proliferation and differentiation, as well as lipid and glucose homeostasis (Intrasuksri et al., 1998; Kennedy et al., 2004; Rosen et al., 2017; Shipley et al., 2004).

While studies have suggested possible associations between perfluoroalkyls and self-reported hearing impairment (HI) (Shiue, 2015), no study has explored the effects of perfluoroalkyls on audiometrically measured HI. This cross-sectional study aimed to examine the associations between perfluoroalkyl exposures and HI in US adults who participated in the National Health and Nutrition Examination Survey (NHANES).

Table 1 shows survey-weighted participant characteristics of the U.S. adults 20-69 years of age stratified by HI status. Of 2371 adults, 721 had any form of HI (sample-weighted prevalence= 30.5%). Participants with HFHI were 14.8 years older on average (P<.0001), more likely to be male (P<.0001), non-Hispanic white (P<.0001), had attained high school or lower level of education (P=0.01), and had higher BMI (P=0.003). People with HFHI were also more likely to be former or current smokers (P<.0001). Participants with any HFHI and HFHI were more likely to experience firearm noise (P<.0001) and occupational noise (P=0.002). Participants with HFHI also tended to have higher prevalence of type-2 diabetes (P<.0001) or hypertension (P<.0001) conditions. Similar differences were also observed for LFHI. In addition, compared to those without HI, participants with HI had higher median concentrations of PFHxS, PFOS, PFOA and PFNA. Participant characteristics of adults ≥70 years are shown in Supplemental Materials Table S2. We did not observe significant differences by HI status, possibly due to the small sample size of participants without HI (N=26).

Survey-weighted least square geometric means and ratio differences in PFAS concentrations by participant characteristics are shown in Table 2, after controlling for age, sex, race/ethnicity, education level, poverty-income ratio, smoking status, body mass index, noise exposures (occupational, recreational, and firearm noise) and NHANES cycles. Serum concentrations of PFHxS, PFOS, PFOA, PFNA and PFDA were significantly higher among adults 60-69 years vs. those 20-39 years of age, males vs. females, and ever vs. never smoker, after adjusting for covariates. Serum perfluoroalkyl concentrations also differed significantly by racial/ethnic groups. Non-Hispanic white had significantly higher serum concentrations of PFOA while lower concentrations of PFNA and PFDA, compared to other racial/ethnic groups. Non-Hispanic black had significantly higher serum concentrations of PFOS. Interestingly, persons who received college education and higher had significantly higher concentrations of PFHxS. Former and current smokers had significantly higher concentrations of PFOA and PFOS compared to never smokers. Occupational, firearm and recreational noises were not associated with perfluoroalkyl concentrations.

The associations between perfluoroalkyls and HI in adults aged 20-69 years are presented in Table 3. There were no significant associations when perfluoroalkyl variables were fitted as a linear (log-transformed) term. However, statistically significant associations of HFHI with PFNA (OR=1.70, 95% CI: 1.13-2.56) and PFDA (OR=1.75, 95% CI: 1.00-3.05) were observed when comparing participants with serum concentrations ≥90^th vs. <90^th percentiles of PFNA (90^th percentile=1.8 ng/mL) and PFDA (90^th percentile=0.5 ng/mL). No significant associations were observed for other compounds. Figure 2 presents the associations between hearing thresholds at each frequency and levels of perfluoroalkyl exposures (≥90^th vs. <90^th percentile) in the covariate-adjusted models. No significant differences were detected for PFAS compounds, although participants with higher concentrations of PFNA and PFDA tended to have elevations in their hearing thresholds at 3, 4, 6 and 8 kHz test frequencies. No significant associations were observed in adults aged ≥70 years (Supplemental Material Table S3).

The results remained similar with the recent NHANES cycles 2011-2012 and 2015-2016 (Supplemental Material Table S4). The adjusted OR of having HFHI conditions was 1.70 (95% CI: 1.12-2.57) for PFNA and 1.78 (95% CI: 1.03-3.09) for PFDA, comparing serum concentrations ≥90^th vs. <90^th percentile, with an additional adjustment for diabetes status and hypertension (Supplemental Materials Table S5). Similar findings were observed after additionally controlling for blood lead and blood cadmium (Supplemental Materials Table S6).

Our study examined the associations between perfluoroalkyl serum concentrations and audiometrically assessed HI in the U.S. adults using nationally representative data. We observed non-linear threshold dose-response relationships of HFHI with PFNA and PFDA in adults aged 20-69 years after adjusting for covariates and these relationships were robust in several sensitivity analyses. We did not detect significant log-linear relationships (i.e., a log-transformed variable fitted as a linear term). No significant associations were observed in adults ≥70 years possibly due to the small sample size of older population. To our knowledge, this is the first report to link serum perfluoroalkyls and audiometric measures among adults, highlighting the importance of exploring the role of perfluoroalkyls on neurobehavioral functions in animals and humans.

The evidence of underlying biological mechanisms linking environmentally relevant PFAS exposure concentrations to the development of HI are limited. PFAS exposures play an important role in PPAR signaling. PPARs, a family of ligand-regulated nuclear hormone receptors, have been identified as a key player in the mode of action for PFAS toxicity. The chemical structure of PFAS are analogous to fatty acids and both can activate PPARs and further induce endocrine disruption (Kraugerud et al., 2011; Pedersen et al., 2016), as well as disturbance of lipid and glucose metabolism, inflammation and adipocyte differentiation (Berger et al., 2005; Staels and Fruchart, 2005). Two of the PPAR family members, PPARα and PPARγ, were detected in the inner ears of neonatal and adult mice, and alterations in PPARs could impact inner and outer HC apoptosis (Sekulic-Jablanovic et al., 2017). Although human PPARα appears to be less responsive to PFAS exposure than mouse PPARα, most perfluoroalkyl carboxylates and sulfonates activate PPARα, and to a lesser extent PPARγ in mouse and human models (Wolf et al., 2008). The ability to stimulate PPARα and PPARγ of the cochlea with PFAS exposures appears to offer an alternative explanation for the observed associations.

Exposures to PFNA and PFDA were associated with higher odds of HFHI in adults aged 20-69 years, while no significant associations were observed for other PFAS compounds. Studies have shown that the differential results may be related to differential transactivity with PPAR signaling pathways. Perfluoroalkyl carboxylates (e.g. PFOA) is more capable than the corresponding sulfonates (e.g. PFOS) in activating PPARα in mouse and human models (Takacs and Abbott, 2007; Vanden Heuvel et al., 2006; Wolf et al., 2008). In addition, longer-chain carboxylates such as PFDA and PFNA induced higher activation of PPARα than PFOA (Wolf et al., 2010, 2008). However, due to the limited evidence of toxicity of various PFAS compounds and especially longer-chain compounds, future studies are needed to explore the differential mechanisms.

Unlike the study by Shiue, 2015 in which an increased OR of self-reported hearing disturbances was observed for participants with higher PFOA serum concentrations in NHANES 2011-2012, we did not detect any associations of PFOA with audiometrically assessed hearing loss. Audiometry instead of self-reported information might account for the discrepancies. In addition, we explored the non-linear relationships of HFHI with PFNA and PFDA. Our dose-response relationship is consistent with a threshold effect, suggesting that, if these findings are causal, ototoxicity of these longer-chain perfluoroalkyl carboxylic acids may be activated at certain concentrations and above. These results highlight the need to consider the shape of nonlinear dose-response relationships when studying environmental ototoxic chemicals.

A major strength of this study was the population examined. The NHANES is a complex stratified survey. With sampling weights, strata and units considered in the analyses, our study samples are representative of the general U.S. adults. However, it is difficult to make an inference about the temporal causation of exposures and hearing loss in a cross-sectional study. Given the nationally representative nature of data, however, our results are important and support the need for future research to evaluate perfluoroalkyl exposure as a potential risk factor for hearing impairment. Moreover, we cannot rule out residual confounding by noise environments that cannot be captured by dichotomous indicators for occupational, recreational and firearm exposures. This limitation may be more of a factor for high-frequency notches because this outcome incorporates pure tone thresholds observed across 3-6 kHz, at which excessive noise stimulus affects most (Gates et al., 2000). Other persistent organic pollutants, such as polychlorinated biphenyls (PCBs), have been associated with hearing impairment in U.S. adults (Min et al., 2014). PCBs and PFAS belong to the family of polyhalogenated compounds. Similar to PFAS, these chemicals may disrupt thyroid hormone homeostasis and further lead to loss of outer hair cells (OHCs) and sparing of inner hair cells (IHCs) in the apical turn of cochlea at which lower frequency response locate, leading to low-frequency auditory impairment (Crofton et al., 2000; Powers et al., 2006). Thus, one cannot rule out potential confounding by other chemicals that have been suggested as ototoxic chemicals, although confounding by lead and cadmium is unlikely. Given that people are exposed to a myriad of chemicals daily, it is important to quantify the impact of chemical mixtures in future studies (Wang et al., 2019b, 2019a, 2018). We were also not able to follow the standard for reporting hearing results by the Hearing Committee of the American Academy of Otolaryngology–Head and Neck Surgery (Gurgel et al., 2012), a scattergram relating PTA to word recognition score, because of lack of data on word recognition score.

In summary, the present analysis of a well-defined representative sample of the U.S. adults does not provide strong evidence to support the ototoxicity of PFAS exposure. However, we report non-linear threshold dose-response associations between serum concentrations of PFNA and PFDA and HFHI. Previous research primarily focused on PFOA and PFOS which are the predominant analytical targets detected in the environment. Given that hearing loss is expected to increase from 44 million (~15% of adults) in 2020 to 74 million by 2060 (~23% of adults) (Goman et al., 2017), our findings still have significant public health implications. Although it is unlikely that our findings are subject to reverse causality, future studies with prospective study designs are needed to confirm concerns related to causal inferences.