Exploring the association of organochlorine pesticides exposure and hearing impairment in United States adults

Hearing loss (HL) is a highly prevalent public health concern. Organochlorine pesticides (OCPs) are widely used environmental pollutants harmful to human health. Studies investigating the effects of OCPs exposure on the auditory system in the general population are rare. To explore the association between OCPs exposure and HL in adults, 366 adults aged 20–69 years who participated in the National Health and Nutrition Examination Survey (NHANES, 2003–2004) were investigated. HL was defined as a pure-tone average (PTA) ≥ 20 dB in the better ear. Multivariate linear and logistic regression analyses were conducted to evaluate the association of four selected serum OCPs with PTAs and the risk of HL. In participants aged < 60 years, hexachlorobenzene (HCB) and dichlorodiphenyldichloroethylene (p, p'-DDE) exposure was positively associated with low- and speech-frequency PTAs, and with low-frequency HL, respectively. Risk of HL increased in the highest tertile compared with the lowest tertile of serum HCB and p, p'-DDE (odds ratio [OR]: 4.38, 95% confidence interval [CI]: 0.97–19.80; OR: 16.66, 95% CI: 2.64–105.09, respectively). In this study of US adults aged < 60 years, HCB and p, p'-DDE exposure was positively associated with HL. HCB and p, p'-DDE may be potential risk factors for HL.

. The weighted demographic characteristics of study participants. BMI body mass index, PTA puretone average. a P values of continuous variables and categorical variables were calculated by weighted linear regression model and weighted chi-square test, respectively. b Low-, speech-, high-frequency PTA values in the better ear were computed from the average of hearing thresholds of 0.5, 1 and 2 kHz, 0.5, 1, 2 and 4 kHz, 4, 6 and 8 kHz, respectively. c Hearing loss was defined as the PTA value ≥ 20 dB in the better ear. Significant values are in bold. www.nature.com/scientificreports/ covariates included in this analysis were age, sex, race, education level, body mass index (BMI) (categorical), diabetes, hypertension, serum cotinine level, firearm noise exposure, and loud noise/music exposure. All four OCPs were converted to categorical variables (tertiles) and were used as continuous variables to calculate the linear trend. In the unadjusted model (crude model), the P value for trend showed that almost all the four OCPs were significantly associated with low-, speech-, and high-frequency PTA hearing threshold shifts and HL. Only the association of HCB with low-and high-frequency HL was not significant. In all the adjusted models, a significant P for trend was not observed among the tertiles of the four OCPs and either hearing threshold shifts or HL (all P ≥ 0.05).
Multivariate regression analysis stratified by age: association of OCPs exposure with hearing thresholds and HL. As shown in Tables 3 and 4, the participants were divided into two groups: adults aged < 60 years (N = 308) and adults aged ≥ 60 years (N = 58). HCB exposure was positively associated with low-and speech-frequency PTAs among participants aged < 60 years, when comparing the highest tertile of HCB exposure level with the lowest tertile (β = 1.90, 95% confidence interval [CI]: 0.11-3.70, P trend = 0.0354, P interaction = 0.0063 in the low-frequency PTA group; β = 1.88, 95% CI: 0-3.77, P trend = 0.0454, P interaction = 0.0397 in the speech-frequency PTA group). p, p'-DDE showed a similar association (β = 2.84, 95% CI: 0.86-4.81, P trend = 0.0054, P interaction = 0.0063 in the low-frequency PTA group; β = 3.44, 95% CI: 1.38-5.51, P trend = 0.0015, P interaction = 0.0448 in the speech-frequency PTA group). However, there was no clear association between HCB or p, p'-DDE exposure and hearing threshold shifts among individuals aged ≥ 60 years (Table 3). Among individuals aged < 60 years old, those with high HCB exposure showed a higher risk of HL than those with low HCB exposure; however, this association was not observed among individuals aged ≥ 60 years (odds ratio [OR]: 4.38, 95% CI: 0.97-19.80, P trend = 0.0475, P interaction = 0.0101) ( Table 4). An increase in the risk of HL was observed in individuals aged < 60 years with p, p'-DDE exposure levels in the highest tertile, compared with those in the lowest tertile (OR: 16.66, 95% CI: 2.64-105.09, P trend = 0.0015, P interaction = 0.0288); the 95% CI associated with the OR had a wide range. Dieldrin exposure had no statistically significant interactions with age in the prediction of either hearing threshold shifts or HL (P interaction > 0.05) (Tables 3 and 4).

Discussion
In a representative sample of US adults aged 20-59 years old, higher serum HCB and p, p'-DDE concentrations were positively correlated with the prevalence of low-frequency HL and with low-and speech-frequency PTA hearing threshold shifts after adjusting for potential confounders including sex, race, education level, BMI (categorical), diabetes, hypertension, serum cotinine level, firearm noise exposure, and loud noise/music exposure. Our findings suggest that environmental exposure to OCPs may be involved in the development of hearing impairments in adults. It should be noted, however, that the 95% CI associated with the OR (16.66) of HL in the highest tertile relative to the lowest tertile of serum p, p'-DDE had a wide range (2.64-105.09). HL is a major public health problem affecting over 1.5 billion people globally in terms of their health and quality of life 1 . The prevalence of HL has been increasing, with many factors (i.e., noise exposure, aging, and ototoxic drugs) known to be common causes, and attention has been given to the role of environmental pollutants exposure in the development of HL 2 .
Most studies on hearing impairment associated with exposure to organohalogen compounds have focused on PCBs, which are structurally related to OCPs and have similar characteristics such as being resistant to degradation, lipophilic, and able to bioaccumulate in organisms as that of OCPs 9-13 . The supporting evidence of the ototoxicity of OCPs' is very limited, with only five studies involved. Three animal studies have investigated the ototoxicity of DDT and HCB in rats [14][15][16] . A study that was conducted by recording distortion product otoacoustic emissions (DPOAEs), an audiological examination performed at different ages of infants, and by calculating DPOAE amplitudes to serum OCPs reported that exposure to HCB, β-HCH, and p, p'-DDE in infancy may be linked with hearing impairment 17 . A case-control study reported a positive association between α-HCH exposure and HL in a Chinese adult population 18 . Our study is the first cross-sectional study to investigate the effect of OCPs exposure on the auditory system in a sample of the general population of US adults. The results of this study are consistent with previous findings.
Oxidative stress and/or aryl hydrocarbon receptor-mediated mechanisms may be important determinants of HCB and p, p'-DDE ototoxicity 19,20 . Aberrant epigenetic and inflammatory mechanisms caused by pesticide exposure may influence the initial stage of auditory development 21 . Although HL caused by PCBs exposure is related to the induction of hypothyroidism, and exposure to OCPs is associated with adverse thyroid function, which raises the suspicion that OCPs exposure might affect the auditory system by disturbing thyroid function; however, no evidence has supported this hypothesis 14,22,23 . In addition, whether exposure to HCB and p, p'-DDE causes ototoxicity by affecting the cochlear outer hair cells (OHCs) or other parts of the auditory system (i.e., the organ of Corti, the nerves) remains uncertain: one study showed that DPOAEs (measure for assessing Table 3. Adjusted associations between OCPs and hearing threshold shifts stratified by age (N = 366): age < 60 group (N = 308) and age ≥ 60 group (N = 58). Adjusted for gender, race, education level, BMI, diabetes, hypertension, cotinine, firearm noise exposure and loud noise/music exposure. OCPs organochlorine pesticides, HCB hexachlorobenzene, p, p'-DDE p, p'-dichlorodiphenyldichloroethylene, BMI body mass index, PTA pure-tone average. Significant values are in bold. www.nature.com/scientificreports/ OHCs function) was affected by HCB and p, p'-DDE exposure, and another study observed only a slight loss of hair cells (< 1%) in HCB-treated rats, which could also be observed in control animals 14,17 . In addition, results regarding the range of hearing impairment caused by exposure to OCPs have been inconsistent across studies. In our study, HCB and p, p'-DDE exposure was positively associated with low-frequency HL and with low-and speech-frequency hearing threshold shifts. These results are almost in agreement with those of the study by Hadjab et al. 14 . Contrarily, Sisto et al. 17 found positive associations between HCB and p, p'-DDE exposure and DPOAE amplitudes for most of the primary tone frequencies. Zhang et al. 18 observed a relationship between α-HCH exposure and hearing impairment, mainly at mid and high frequencies. Further studies are required to clarify these discrepancies. This study used a large and representative sample with rigorous quality control over the data collection process and adjusted for critical confounders to establish a relationship between serum OCPs concentrations and hearing impairment. However, this study had several limitations. First, due to the inherent limitation of the cross-sectional design, the temporal sequence of exposures and outcomes cannot be confirmed. Prospective studies are required to definitively define the association between OCPs exposure and hearing function. Second, data of infants, adolescents, and the elderly (70 + years old) were not included in this study due to the lack of participants or the sample size being too small. Moreover, we could not rule out any potential effects of OCPs exposure on hearing impairment in people aged ≥ 60 years, as the sample size in the subgroup was small (N = 58). Third, some potential confounders (e.g., dietary intake, history of ear infection, and congenital hearing impairment) were not addressed in this study. Furthermore, studies on the mechanisms of the effects of HCB and p, p′-DDE exposure on hearing impairment are insufficient. Therefore, functional biological studies (i.e., DPOAE and brainstem auditory evoked potential studies) and prospective population studies are required to investigate the potential mechanisms.

Methods
Study design and population. The NHANES is a nationwide, cross-sectional, representative series of surveys containing health-related information of the US general population. These continuous surveys consist of interviews, physical measurements, and laboratory tests of a selected sample of the US non-institutionalized population. The NHANES project was reviewed and approved by the Research Ethics Review Board of the National Center for Health Statistics, and informed consent was obtained from all participants. Data from the NHANES and related documentation and protocols described in detail are publicly available from the NHANES website (https:// www. cdc. gov/ nchs/ nhanes/ Index. htm).   Figure 1 shows a flow chart for participant selection in this study. Participants with missing data on hearing threshold levels, otoscopic tests, tympanogram tests, or serum lipid-adjusted concentrations of the selected OCPs were excluded. Individuals with abnormal otoscopic results, poor-quality tympanogram results, or tympanograms with compliance ≤ 0.3 ml were also excluded to avoid analyzing conductive or mixed HL data 24,25 . Finally, 366 participants were included in this study.
OCPs exposure measurement. Blood serum concentrations of 13 OCPs and metabolites were measured in a random one-third subsample of participants aged 12 years and older using high-resolution gas chromatography/isotope-dilution high-resolution mass spectrometry at the US Centers for Disease Control and Prevention (CDC) National Center for Environmental Health. Four OCPs detected in ≥ 80% of the samples were selected for analysis in our study: HCB, p, p'-DDE, trans-nonachlor, and dieldrin. We used the limit of detection (LOD) divided by the square root of two for any sample below the LOD. In addition, the lipid-adjusted serum concentrations (ng/g lipid) of the four OCPs were used in our study and log-transformed to normalize the skewed distribution before the analysis.
Audiometric measurement. The detailed procedure and protocol of audiometric examination have been described in the NHANES audiometry procedures online manual 26 and previous papers 27,28 . Briefly, half of the sample of US adults aged 20-69 years underwent an audiometric component test. Trained examiners conducted hearing threshold examinations in a silent and sound-isolating audiometry room. The hearing threshold examination was conducted at seven frequencies from 500 to 8000 Hz using an AD226 audiometer (Interacoustics). In this study, low-frequency HL was defined as PTA calculated by averaging thresholds at 500, 1000, and 2000 Hz ≥ 20 dB HL in the better ear; speech-frequency HL was defined as PTA at 500, 1000, 2000, and 4000 Hz ≥ 20 dB HL in the better ear, which is consistent with the definition used by the World Health Organization 1 ; and high-frequency HL was defined as PTA at 4000, 6000, and 8000 Hz ≥ 20 dB HL in the better ear.
Otoscopic examination of both ears was performed using a Welch Allyn otoscope (Model 25,020). Tympanometry was performed using an Earscan Acoustic Impedance tympanometer (Micro Audiometrics) to assess middle ear function. Inference of sensorineural HL was based on the findings of normal otoscopic examinations and good-or adequate-quality results of the tympanogram with compliance > 0.3 ml. Individuals who did not meet the standards were excluded from further analysis.
Covariates. The following variables were considered as potential covariates in the analysis: age and BMI as continuous variables, and sex, race/ethnicity, education level, BMI (categorical), diabetes, hypertension, serum www.nature.com/scientificreports/ cotinine level, firearm noise exposure, and noise noise/music exposure as categorical variables. Information on demographic variables, noise exposure, and present medical illnesses, such as diabetes and hypertension, was obtained from self-reported questionnaires. Firearm exposure was defined as "firearm noise exposure outside work for an average of at least once a month for a year" 29 . Loud noise/music exposure was defined as "exposure to other types of loud noise, such as noise from power tools or loud music, outside work, for an average of at least once a month for a year. " Diabetes was defined by a "yes" or "borderline" answer to the question "other than during pregnancy, ever been told by a doctor or health professional had diabetes or sugar diabetes" 30 . Hypertension was defined as "ever been told by a doctor or other health professional had hypertension, also called high blood pressure" 30 . BMI was obtained through physical examination. Serum cotinine level, a marker for both active and passive tobacco exposure, was tested by an isotope dilution-high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometry 31 .
Statistical analysis. Weighted statistical differences in demographic and potential hearing-related variables between individuals grouped by sex were evaluated, with categorical data presented as percentages and continuous data as mean ± SD. The P values of the continuous and categorical data were estimated using a weighted linear regression model and weighted chi-square test, respectively. We distributed the log-transformed lipid-adjusted OCP levels into tertiles before conducting univariate analysis to estimate potential variables. Multivariate linear regression analysis was used to determine regression coefficients (β) and 95% CIs between the four OCPs and hearing threshold shifts, and multivariate logistic regression analysis was used to estimate ORs and 95% CIs between the four OCPs and HL, adjusting for age, gender, race/ethnicity, education level, BMI (categorical), diabetes, hypertension, serum cotinine level, firearm noise exposure, and loud noise/music exposure, instead of using sample weights. This adjustment is considered a good compromise between efficiency and bias 32,33 . We evaluated the influence of the interactions between the OCPs and age on HL. Stratified multivariate linear and logistic regression analyses according to age (< 60 vs. ≥ 60 years) were performed. The association between trans-nonachlor exposure and HL was not estimated in the multivariate linear and logistic regression analyses stratified by age because the number of participants aged ≥ 60 years in the category of low trans-nonachlor exposure tertile was quite small, and only two individuals were included. Statistical analysis was conducted using the statistical programming language R 3.6.1 and EmpowerStats software (X&Y Solutions, Inc.). A P-value < 0.05 was considered statistically.