Study participants

The study population was compiled using a multistage stratified cluster random sampling method in Zhejiang province from 2016 to 2018. Five cities (Hangzhou, Jiaxing, Huzhou, Quzhou, and Lishui) were randomly selected by stratifying the geographical distribution and city sizes based on the national nutrition and health survey in Zhejiang in 2012. One healthcare center was then randomly selected in each city, and all the participants in each healthcare center were selected by cluster sampling if they met the following inclusion criteria: 1) were local residents (who have lived there for more than 12 months), 2) were working in workplaces that do not meet the national noise pollution standards (exposure to non-occupational noise environment), and 3) underwent physical examination. The exclusion criteria were: 1) the presence of ear diseases like otitis media, and 2) aged <20 years or >80 years.

In total, 4930 persons who participated signed an informed consent form and completed a questionnaire and audiometry test in a hearing room. The survey results were obtained after excluding 245 participants with missing information on smoking-related problems or hearing data, those with ear problems such as otitis media, and those aged <20 years or <80 years. Finally, 4685 surveyed residents were analyzed, including 2386 males and 2299 females. To analyze the association between smoking status and hearing loss according to age, the population was divided into three age groups: young (20–39 years), middle-aged (40–59 years) and older group (60–80 years).

Audiometry measures

Devices utilized in this research were the Madsen Itera clinical diagnostic audiometer and TDH39 headphones developed by GN Otometrics (Denmark). The audiometer was calibrated, and the audiometric tests were performed by trained technicians in a sound-proof chamber with noise levels below 30 dB. Pure-tone air-conduction hearing threshold testing at frequencies of 0.125 kHz, 0.25 kHz, 0.5 kHz, 1 kHz, 2 kHz, 3 kHz, 4 kHz, 6 kHz and 8 kHz was performed on both ears. To measure the reliability of participant responses, the 1 kHz frequency was tested twice in each ear. If the results differed by more than 10 dB, it was considered to be an unreliable response. When their answer was unreliable, we reconfirmed the procedure with them and reminded them to conduct the hearing test correctly. We computed the pure-tone average (PTA) at speech frequencies (0.5, 1, 2 and 4 kHz) and high frequencies (3, 4, 6 and 8 kHz). Hearing loss was defined as speech-PTA of ≥26 dB in the better ear, which is consistent with the WHO definition of clinically significant hearing loss and can identify patients with bilateral hearing loss and related functional impairments.

Questionnaire

A pilot questionnaire survey was carried out before the custom-designed questionnaire was finally adopted. Prior to administering the questionnaire survey, investigators received professional training to ensure data quality. The questionnaire included items related to basic demographics including gender, age, education, and income level; information on diseases including hypertension, diabetes, hyperlipidemia, otitis media, migraine, and anemia; lifestyle and behavioral factors, including smoking, drinking, and earphone use; and workplace noise exposure. Education level was categorized as: elementary school or secondary school (<High School); High school; and junior college, undergraduate or postgraduate (>High School). Average monthly income was classified into three categories: low ≤4000; middle 4000–6000, and high ≥6000 RMB (1000 Chinese Renminbi about US$148, average 2017). Based on cigarette smoking history, participants were categorized as: self-reported never smokers (smoking less than one cigarette per day on average for less than one year, and exposure to passive smoke for less than 1 day per week); passive smokers (smoking less than one cigarette per day on average for less than 1 year and being exposure to passive smoking for more than one day per week); current smokers (smoking at least one cigarette per day for more than 1 year); and past smokers (stop smoking and keep non-smoking at least six months since cessation). Regarding drinking history, participants were categorized as: self-reported non-drinkers (less than one drink per week), past drinkers (abstinence for more than 6 months), and current drinkers (alcohol consumption at least once a week for 6 months or more).

For current smokers, daily cigarette consumption and years of smoking were also recorded. A smoking index was calculated by multiplying the number of cigarettes smoked per day by the number of years smoked. Current smokers were divided into two subgroups: light smokers (smoking index ≤150) and heavy smokers (smoking index >150), according to the two-digit grouping based on the median of smoking index.

Statistical analysis

The study used Epidata V.3.1 (The Epidata Association, Odense, Denmark) for survey data entry, including error check and correction (double entry and validation). SPSS V.21.0 (SPSS for Windows, IBM Corp., Armonk, NY) was used to conduct all statistical analyses. Sample descriptive data were summarized using percentage distributions. Chi-squared analysis was used to compare differences between various covariates groups. Binary logistic regression was used to estimate the association between hearing loss and smoking status, after adjustment for other variables including age, gender, personal income, education level, alcohol consumption, frequency of headphone use, hypertension, diabetes, hyperlipidemia, and workplace noise. All reported p-values were twotailed, and p<0.05 was considered statistically significant.

Characteristics of the study population

The characteristics of hearing loss according to basic demographics, information on diseases, lifestyle and behavioral factors, and workplace noise exposure are presented in Table 1. Of the 4685 eligible participants, 1882 (40.2%) had speech-frequency hearing loss, and 2527 (53.9%) had high-frequency hearing loss. Gender, age, education status, personal income, cigarette smoking, alcohol consumption, workplace noise, frequency of use of headphones, hypertension, diabetes and hyperlipidemia were significantly associated with speech-frequency and high-frequency hearing loss (all p<0.001). Anemia and migraine were not significantly associated with both types of hearing loss (Table 1).

Smoking status distribution in different genders and age groups

Among the 4685 participants analyzed, 50.9% (n=2386) were male, and 49.1% (n=2299) were female. Regarding age distribution, 29.7% (n=1393) were in the young group, 43.8% (n=2050) in middleaged group, and 26.5% (n=1242) in the older group. The chi-squared test indicated that the distribution of smoking status was statistically significantly different between male and female population. The proportions of current smokers and past smokers were 38.7% (n=923) and 9.5% (n=226) in males and 1.1% (n=25) and 0.9% (n=21) in female population, respectively. The distribution of smoking status was statistically significantly different among different age groups.

The middle-aged group had the highest proportion of current smokers (24.5%; n=502), whereas the highest proportion of passive smokers was in young group (36%; n=501). The proportion of past smokers increased with age, and the highest was 10.9% (n=135) in the older group (Table 2).

Relationship of smoking status with hearing loss in males and females

Smoking status was significantly associated with speech-frequency and high-frequency hearing loss in males (p<0.001), but not in females (p>0.05). In males, after adjusting for other factors, current smoking and past smoking showed significantly increased risks for speech-frequency hearing loss compared with never smokers, with odds ratios (ORs) of 1.64 (95% CI: 1.27–2.11) and 2.16 (95% CI: 1.47–3.17), respectively (Table 3). Current smoking and past smoking were significantly associated with an increased risk of high-frequency hearing loss compared with never smokers, with ORs of 1.80 (95% CI: 1.38–2.34) and 2.38 (95% CI: 1.51–3.75), respectively (Table 3). Passive smokers did not have a significantly increased risk of both speech-frequency and high-frequency hearing loss in males compared to never smokers (Table 3)

Relationship between smoking status and hearing loss according to age in males

In the young group (20–39 years), smoking status was not significantly associated with speech-frequency and high-frequency hearing loss (p>0.05). In the middleaged group (40–59 years), cigarette smoking was significantly associated with both speech-frequency and high-frequency hearing loss (p<0.05). After adjusting for other variables, compared with never smokers, current smokers and past smokers had a significantly higher prevalence of speech-frequency hearing loss, with OR of 1.65 (95% CI: 1.17–2.33) and 1.88 (95% CI: 1.11–3.17), respectively. Current smokers and past smokers had a significantly higher prevalence of high-frequency hearing loss, with ORs of 2.01 (95% CI: 1.43–2.84) and 2.64 (95% CI: 1.50– 4.66), respectively. In the older group (60–80 years), compared with never smokers, only past smokers had a significantly higher prevalence of speech-frequency hearing loss, with OR of 2.58 (95% CI: 1.38–4.85). Smoking status was not significantly associated with high-frequency hearing loss (Table 4). We also analyzed the other covariates that have significant effect on hearing loss and the results are given in the Supplementary file Table S1. In young adults, besides smoking, drinking, workplace noise, hypertension and diabetes were significantly associated with speech-frequency hearing loss. Workplace noise and hypertension were significantly associated with high-frequency hearing loss. In the middle-aged group, besides smoking, workplace noise, hypertension and diabetes were significantly associated with speech-frequency hearing loss. Hypertension was significantly associated high-frequency hearing loss. In the older group, hypertension and hyperlipidemia were associated with speech-frequency hearing loss, and workplace noise with high-frequency hearing loss.

Relationship between smoking intensity and the risk for hearing loss according to age in males

In the middle-aged group and older age group, the current smokers were further divided into two groups based on the smoking index, light smokers and heavy smokers. Compared with the light smokers, the heavy smokers had a higher risk for speech-frequency hearing loss in older group, with ORs of 2.35 (95% CI: 1.08–5.14), and also had a higher risk for high-frequency hearing loss in middle-aged group, with ORs of 1.89 (95% CI: 1.23–2.91, p<0.05) (Table 5).

Strengths and limitations

This study has several strengths. First, it was a large-scale, general population-based survey conducted in China that fills the gap left by the severe shortage of studies on the relationship between cigarette smoking and hearing loss, especially in different gender and age groups. We analyzed the relationship between cigarette smoking and hearing loss in stratified age subgroups rather than adjusting for the age, resulting in a more detailed description of the effect of smoking on hearing loss. Second, we selected pure-tone threshold audiometry at nine frequency bands instead of self-reported hearing loss or narrow frequency bands, which enabled a more precise evaluation of hearing loss. Finally, the questionnaire used included an extensive set of covariates which excluded many potential confounders.

Nevertheless, this study also has some limitations. First, the cross-sectional design has limited the evidence for causal inference. Second, a questionnaire was used to collect information on smoking habits and history of diseases based on self-reporting by the participants, without secondary sources to confirm the accuracy of responses. Third, more than 10 variables were included in this study, and the regression analysis was performed under the hypothesis that all these included covariates were independent of each other. Finally, for past smokers, we did not collect the smoking duration before quitting. Therefore, we were unable to determine why they had the highest prevalence of hearing loss.