10-Year Follow-Up Results of The Netherlands Longitudinal Study on Hearing: Trends of Longitudinal Change in Speech Recognition in Noise : Ear and Hearing

Journal Logo

Research Article

10-Year Follow-Up Results of The Netherlands Longitudinal Study on Hearing: Trends of Longitudinal Change in Speech Recognition in Noise

Goderie, Thadé P. M.1; Stam, Mariska1; Lissenberg-Witte, Birgit I.2; Merkus, Paul1; Lemke, Ulrike3; Smits, Cas1; Kramer, Sophia E.1

Author Information
Ear and Hearing 41(3):p 491-499, May/June 2020. | DOI: 10.1097/AUD.0000000000000780
  • Open
  • Editor's Choice



Hearing loss is one of the most frequently occurring chronic conditions worldwide (Mathers et al. 2008). The prevalence of hearing loss increases significantly with age. The underlying mechanisms of age-related hearing loss (ARHL) are only partly understood. It consists of physiological degenerations related to aging plus the accumulated effects of environmental exposures, medical disorders and their treatment, and hereditary susceptibility (Huang & Tang 2010). This results in reduced hearing sensitivity, reduced speech recognition in noise, and slowed and impaired central auditory processing (Huang & Tang 2010). Hearing loss is associated with social isolation, depression, loss of self-esteem, cognitive decline and dementia (Mulrow et al. 1990; Lin et al. 2013; Stam et al. 2016; Heywood et al. 2017; Thomson et al. 2017). Identifying and understanding factors that contribute to hearing loss is important for the purpose of prevention and the design of interventions to avert hearing loss and its consequences.

Observed longitudinal changes in pure-tone thresholds provide evidence that ARHL does not increase linearly but accelerates as age progresses (Gates & Cooper 1991; Kiely et al. 2012; Linssen et al. 2014). The accelerated increase of pure-tone thresholds starts as early as in the fourth decade of life (Karlsmose et al. 2000). In longitudinal data also, an accelerated decline of speech recognition abilities in noise was shown. Compared with younger age groups, the decline in speech recognition in noise was faster in adults aged 50 to 59 years (Stam et al. 2015) and in those older than 75 years (Pronk et al. 2013). This accelerated decline was not observed in other studies (Divenyi et al. 2005; Dubno et al. 2008). Dubno (2015) included adults ranging in age from 40 to 96 years old. The decline of speech recognition in babble in that study appeared to accelerate near the age of 75 to 80 years in men only (Dubno 2015). Data on long-term follow-up of speech recognition in noise abilities in young to middle-aged adults are scarce.

Demographic factors such as sex and level of education have shown to be influencing the progression of hearing loss. To illustrate, Gates and Cooper (1991) and Lee et al. (2005) demonstrated that the rate of change in pure-tone thresholds with age was different for males and females. In a study by Cruickshanks et al. (2010), women appeared to experience a slightly later onset of hearing loss in life than men.

Kiely et al. (2012) investigated the association between the prevalence of hearing impairment (as measured with pure-tone audiometry) and educational level. They found a significant association between these two variables. Hearing impairment was more prevalent in participants with a low level of education (defined as secondary education only) compared with those with a higher educational level. However, a longitudinal association between low education and progression of hearing loss was not observed (Kiely et al. 2012). Others also failed to find a longitudinal association between educational level and the decline in hearing (Pronk et al. 2013; Stam et al. 2015). In these studies, a speech recognition in noise test was used to measure hearing status.

Although lifestyle factors such as tobacco smoking and alcohol use have shown to have an effect on hearing loss as well, inconsistent results have been reported so far. Smoking emerged as a risk factor for an increase in the prevalence of hearing loss in a meta-analysis of mainly cross-sectional studies (Nomura et al. 2005). While some longitudinal studies failed to find a temporal effect of tobacco smoking on pure-tone thresholds (Karlsmose et al. 2000; Gopinath et al. 2010; Kiely et al. 2012), other studies demonstrated such an effect (Nakanishi et al. 2000; Cruickshanks et al. 2015). Having a history of smoking has also been found to be associated with the development of (self-reported) hearing loss (Burr et al. 2005). Yet another factor related to tobacco smoking that needs to be considered is time since cessation of smoking. Cruickshanks et al. (2015) found that an increase in time since cessation reduced the risk of developing hearing loss (defined as the 4 frequency pure-tone average of thresholds >25 dB HL in either ear). Pronk et al. (2013) did not find a longitudinal association between tobacco smoking and speech recognition ability in noise in older adults.

Inconsistent results have also been found for the effect of alcohol consumption on pure-tone thresholds. Cross-sectional studies showed that heavy alcohol use, defined as ≥2 registrations to the Sweden temperance board (Rosenhall et al. 1993) or ≥4 drinks/day (Popelka et al. 2000), was related to hearing loss. However, moderate alcohol consumption was found to have a protective effect on hearing (Popelka et al. 2000). Longitudinal studies did not show an association between changes in pure-tone thresholds and alcohol use (Brant et al. 1996; Karlsmose et al. 2000; Gopinath et al. 2010; Cruickshanks et al. 2015).

Longitudinal changes in hearing have mostly been studied using pure-tone audiometry (Morrell et al. 1996; Karlsmose et al. 2000; Wiley et al. 2008; Cruickshanks et al. 2010; Mitchell et al. 2011; Kiely et al. 2012). However, decreased ability in speech recognition in noise was found as the primary and most limiting consequence of hearing loss (Plomp & Mimpen 1979; Kramer et al. 1998; Houtgast & Festen 2008). Previous longitudinal studies on the deterioration of speech recognition in noise focused mainly on middle-aged to older adults, but ARHL results from a lifetime of physiological degeneration and damage to the auditory system. Though hearing loss is more prevalent later in life, factors that affect hearing can present earlier in life. A lifetime approach to study ARHL would help to better understand the progression of hearing loss as age progresses.

The present study is a continuation on the previously presented 5-year results of the Netherlands Longitudinal Study on Hearing (NL-SH) (Stam et al. 2015). It is a longitudinal analysis of 10-year follow-up data using a worldwide implemented speech in noise test (Smits et al. 2006; Jansen et al. 2010; Watson et al. 2012; Potgieter et al. 2015). The main objectives are (1) to examine the rate of decline in speech recognition in noise in adults aged 18 to 70 years over 10 years, and (2) to examine the effects of age, sex, educational level, history of smoking, and alcohol use on this decline.


Study Design and Settings

The NL-SH is an ongoing prospective cohort study examining the relationship between hearing ability and different domains of life such as psychosocial health, work status, general health, and health care use. The NL-SH is conducted over the Internet entirely and uses a convenience sampling method. Eligible participants are adults between 18 and 70 years old. Both normal-hearing and hearing-impaired adults are invited to participate. Further details related to the recruitment of participants, data collection, and follow-up measurements are reported by Stam et al. (2015). Baseline data collection (T0) started in 2006. The second measurement round (T1) started exactly 5 years later in 2011 and the third measurement round (T2) started in 2016. The NL-SH study (including the follow-up measurement rounds) was approved by the Medical Ethics Committee of the VU University Medical Center in Amsterdam, The Netherlands.

Dependent Variable: Speech Recognition Ability in Noise

All NL-SH participants are invited to perform a speech recognition in noise test, known as the National Hearing test (NHT) (Smits et al. 2006), at each measurement round. The NHT is an online digit-triplet speech in noise test using a total of 23 digit triplets (e.g., 6-1-5) which are presented against a stationary background noise. The spectrum of the noise matches the spectrum of the speech. Participants are instructed to perform the test in a quiet room. Whereas not mandatory, they are advised to use headphones instead of loudspeakers and to indicate what transducer they used by clicking one of the two response options (headphones or loudspeakers). First, digit triplets are presented in quiet (without noise) and participants are instructed to adapt the volume of their computer to a level where they can clearly hear the presented triplets. Then an explanation of the test is presented on the screen and the actual test starts. A series of digit triplets in masking noise is presented to the participant and after each triplet he/she responds by entering the digits on the keypad of the computer. The first triplet is a dummy presentation and is always presented at 4 dB signal to noise ratio (SNR). Then, an adaptive procedure is followed with the first presentation at 0 dB SNR. After each correct response, the next triplet is presented at a 2 dB lower SNR. Conversely, after each incorrect response, the next triplet is presented at a 2 dB higher SNR. The speech reception threshold (SRT) is calculated by taking the average SNR of the last 20 presentations and corresponds to the SNR where the participant recognizes 50% of digit triplets correctly. Because the maximum SNR of a presentation is limited in the NHT, the highest SRT that can be considered to be a reliable estimate is 4 dB SNR. The SRT can be categorized as good (SRT < −5.5 dB SNR), insufficient (−5.5 dB SNR ≤ SRT ≤ −2.8 dB SNR), or poor (SRT > −2.8 dB SNR) (Smits et al. 2006).

Instructions on how to perform the hearing test were identical across all three measurements. At T0 and T1, scores were only available for one transducer (either by headphones or by loudspeakers). At T2, additionally to the standard measurement, participants were asked to perform a second measurement with a different kind of transducer (if they used headphones for the first T2 measurement, the second measurement had to be done with loudspeakers and vice versa).

Independent Variables

Demographic data and lifestyle factors were acquired through online questionnaires. Age at baseline was categorized into 5 age groups: 18 to 30, 31 to 40, 41 to 50, 51 to 60, and 61 to 70 years and was analyzed as a time-independent variable. Educational level was divided into low (elementary school or attended high school but no degree), mid (high school graduate or having an associate degree), and high (having a bachelor’s degree, master’s degree, or doctoral degree). The highest obtained educational level reported at T0, T1, or T2 was included in the analyses. Tobacco smoking was measured by asking the participants if and how often they smoked. Answer categories were: “yes. I smoke every day,” “yes, I smoke occasionally,” “no, but I used to smoke every day,” “no, but I used to smoke occasionally,” “no I never smoked.” Participants who had chosen the option “no, I never smoked” were considered having no history of smoking. All others were categorized as having a history of smoking.

Alcohol use was measured by asking how many units of alcohol participants consumed per day. The outcome was used to categorize each participant in one of the following categories: nonuser, mild user (≤1 unit/day), moderate user (2 units/day [women], 2 or 3 units/day [men]), and excessive user (>2 units/day [women], >3 units/day [men]) (Reinhard & Rood-Bakker 1998). Tobacco smoking and alcohol use were entered in the models as time-dependent variables.

Study Sample

People aged 18 to 70 years at baseline who participated in the NL-SH at T0 and received an invitation to participate in T1 in 2011 to 2012 and in T2 before June 2017 were included in the present study. An invitation for T1 and T2 was sent if at T0 the NHT and the compulsory questionnaires were completed. Participants who consistently (at all measurement rounds) reported to have a congenital hearing loss were excluded from the analyses because the main focus of this study was on ARHL and factors potentially contributing to it. Also cochlear implant recipients were excluded from the study as well as participants who performed near ceiling (those with a NHT score ≥3.0 dB SNR) at baseline. In addition, we excluded aided measurements because unaided measurements provide a more accurate estimate of the actual hearing status. Note that at T1 and T2, hearing aid users were asked to perform the NHT unaided. This was not the case at T0. At T0, participants had been given the option to perform the test aided or unaided. Hence, when only an aided score at T0 was available for a particular participant, the T0 score was regarded as “missing,” while the scores on T1 or T2 for that participant were included in the analyses. Further, changes in SRT between two measurement rounds of >2 SD were checked for inconsistencies on a case-by-case basis. If large fluctuations did not fit with how participants subjectively reported their hearing (defined as >1 point change on a 1 to 9 scale; 1 = very poor, 9 = very good), these measurements were excluded. Measurements were also excluded if the same combination of digits (e.g., 1-2-3) was repetitively entered at each trial of the NHT. After applying the exclusion criteria, the number of measurements available for each participant varied. The minimally required number of completed measurement rounds was 1 for each participant, so they could have completed the NHT at either T0, T1, or T2 only or at two or three measurement rounds.

Statistical Analysis; Descriptive Analysis

Percentages were calculated for categorical variables. Means (M) and SD were calculated for continuous variables. A difference in mean age between participants with a baseline measurement only versus those with complete follow-up data was tested using an independent sample t test. For categorical variables (i.e., sex, educational level, tobacco smoking, and alcohol consumption), differences between participants with a baseline measurement only and participants with complete follow-up data were analyzed using a Pearson Chi-square test.

A paired sample’s t test was used to determine whether the SRTs measured with headphones were significantly different from the SRTs measured with loudspeakers for participants who performed both tests at T2. In the literature, it is shown that performing the NHT using loudspeakers yields slightly higher (worse) SRT scores than using headphones (Smits et al. 2006). The use of different types of headphones does not affect the SRT (Culling et al. 2005; Potgieter et al. 2015).

Effect Analysis: Change in SRT Over Time

A linear mixed model was built to examine the longitudinal change of SRT over time and to investigate which demographic and lifestyle factors influenced the decline of SRT over 10 years. A maximum likelihood method was chosen with a compound symmetry covariance structure for the repeated measurements within participants. Our base model included measurement round (T0, T1, T2), transducer type (headphones or loudspeakers), and a random intercept for each participant.

Using the base model, we first tested whether a random slope had to be included in the model using the likelihood ratio (LR) test. A random slope accounts for individual variation in the change of the SRT over time. If the LR test results in a significant change (p ≤ 0.05) of the base model, a random slope will be added to the model. Then, we investigated the effect of age group, sex, educational level, smoking, and alcohol use on SRT by adding each covariate separately. The variable with the most significant effect on the baseline SRT value was added first, then the second most significant, and so on in a forward selection procedure with inclusion criterion (p < 0.05). Estimated marginal means (EMMs) for SRTs including their corresponding 95% confidence intervals (CI) were calculated and are presented.

Subsequently, to assess whether change in SRT over time was different for different categories of a covariate, all covariates were tested for effect modification. Therefore, the covariate and its interaction with time were added to the base model (e.g., Time × Age group). A covariate was considered an effect modifier when the interaction with time was found to be statistically significant. When multiple significant effect modifiers were found, three-way interactions were analyzed too (e.g., Time × Age Group × Sex). In case of significant 3-way interactions, we tested effect modification within each stratum of effect modifiers to identify within which subgroups of the effect modifiers the change of SRT over time was different. We performed a sensitivity analysis with the group of participants who had complete data (i.e., those who participated in all three measurement rounds). We used the same analysis protocol for that purpose.

All statistical analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY). P values <0.05 were considered to indicate statistical significance.



After applying the exclusion criteria, data of 1349 out of 1555 participants were included and analyzed. Figure 1 shows the number of participants not eligible according to the exclusion criteria in each measurement round. In total, 23.8% of the participants had a complete 10-year follow-up with complete scores across the 3 measurement rounds (T0–T1–T2). Participant with a complete follow-up (N = 321) were significantly older (p < 0.001) (M = 47.0 years, SD = 13.1 years, range = 18 to 69 years) compared with participants who dropped out after T0 (N = 642, M = 44.5 years, SD = 11.2 years, range = 18 to 70 years), and had a significantly higher level of education (p < 0.001). No significant differences between participants with a complete follow-up and those who only participated at baseline were seen for sex, smoking, or alcohol use. Participant characteristics at each measurement round are shown in Table 1. In all, 61% of participants had good hearing at baseline as indicated by the SRT (Table 1). Participants who performed two hearing tests at T2, one with loudspeakers and one with headphones, (N = 38) had significantly lower (i.e., more favorable) SRT scores, −1.19 dB SNR (95% CI = −0.49 to −1.89), when using headphones compared with loudspeakers.

Participant characteristics for each measurement round
Fig. 1.:
Flowchart of participant numbers after applying inclusion criteria per measurement cycle. The number of measurements excluded per measurement round is given in the dashed line rectangle. SNR indicates signal to noise ratio; SRT, speech reception threshold; T0, baseline measurement; T1, 5-year follow-up measurement; T2, 10-yr follow-up measurement.

Decline in SRT Over Time

The first aim of the present study was to examine the deterioration of the SRT over 10 years. The LR test indicated that adding a random slope did not improve the base model significantly (χ2(1) = 1.22, p = 0.27). This demonstrated that the variance in the rate of change of the SRT over time between participants was explained by natural variation and did not need to be corrected for. Tests performed with headphones yielded a lower SRT than tests performed with loudspeakers, [F(1, 1833) = 224.51, p < 0.001]. The SRT decreased significantly over time [F(2, 1205) = 49.23, p < 0.001]. EMM of the base model, corrected for transducer type, indicated that the mean decrease in SRT over 10 years was 0.93 dB SNR (95% CI = 0.73 to 1.12, p < 0.001) (Table 2).

Estimated marginal means (95% CIs) from the linear mixed model for SRT dB SNR (signal to noise ratio) over 10 yr in the total sample

Variables that significantly affected the baseline SRTs were the following: age group [F(4, 1109) = 17.66, p < 0.001], history of smoking [F(1, 1759) = 5.27, p = 0.022], and sex [F(1, 1110) = 13.66, p < 0.001]. Participants in older age groups, participants with a history of smoking, and females had a higher (i.e., more unfavorable) SRT at baseline than the youngest age group, participants without a history of smoking, and males, respectively. Educational level [F(2, 1936) = 1.72, p = 0.18] and alcohol use [F(3, 1875) = 0.58, p = 0.63] did not influence the baseline SRTs. The EMM of the base model—adjusted for age group, history of smoking, and sex—indicated that the mean decrease in SRT over 10 years was 0.89 dB SNR (95% CI = 0.69 to 1.09, p < 0.001) (Table 2).

Interactions Between Demographic and Life Style Factors and SRT Over Time

The covariates “age group” and “history of smoking” also modified the decline in SRT over time. Speech recognition in noise of participants aged 51 to 60 and 61 to 70 years deteriorated significantly faster compared with younger age groups [F(8, 1214) = 4.37, p < 0.001] (Fig. 2 and Table 3). The estimated 10-year decline in SRT of the participants aged 51 to 60 years was 1.37 dB SNR (95% CI = 1.03 to 1.72). The eldest age group had an estimated 10-year decline in SRT of 1.69 dB SNR (95% CI = 1.09 to 2.29). Participants aged 18 to 30 years or 31 to 40 years at baseline maintained their good hearing ability during a 10-year follow-up time interval (Fig. 2 and Table 3). Average SRTs of participants aged 41 to 50 years at baseline gradually declined from good to insufficient.

Estimated marginal means of speech recognition in noise over 10 yr per age group at baseline using the base model
Fig. 2.:
Estimated marginal means of speech recognition in noise over 10 yr per age group at baseline using the base model. SNR indicates signal to noise ratio; SRT, speech reception threshold.

In addition, participants with a history of smoking deteriorated faster compared with participants who had never smoked [F(2, 1158) = 5.7, p = 0.003] (Fig. 3 and Table 4). No significant 3-way interaction Time × History of Smoking × Age Group [F(12, 1430) = 1.38, p < 0.17] was found.

Estimated marginal means of speech recognition in noise over 10 yr for participants with a history of smoking and participants without a history of smoking using the base model
Fig. 3.:
Estimated marginal means of speech recognition in noise over 10 yr for participants who have a history of smoking and for participants who never smoked using the base model. SNR indicates signal to noise ratio; SRT, speech reception threshold.

Finally, using the base model, no significant interactions between Time × Sex [F(2, 1206) = 1.60, p = 0.20], Time × Education [F(4, 1164) = 1.56, p = 0.18], and Time × Alcohol Use [F(6, 1144) = 0.78, p = 0.59] were found.

A sensitivity analysis was additionally performed including the participants with complete data (i.e., those who participated in all 3 measurement rounds; N = 321). Again, a significant decline in SRTs over time was observed [F(2, 672) = 16.19, p < 0.001]. EMMs of the baseline model were −5.86 dB SNR (95% CI = −6.19 to −5.53) at T0, −5.37 dB SNR (95% CI = −5.70 to −5.04) at T1, and −5.13 dB SNR (95% CI = −5.46 to −4.81) at T2. Age group significantly affected SRTs at baseline [F(4, 319) = 6.21, p < 0.001]. History of smoking [F(1, 547) = 3.44, p = 0.064], sex [F(1, 324) = 1.87, p = 0.17], education [F(2, 740) = 0.32, p = 0.73], and alcohol use [F(3, 903) = 1.32, p =0.27] did not significantly affect baseline SRTs. The same effect modifiers as in the main analyses were found. Older age groups declined faster compared with younger age groups [F(8, 672) = 2.20, p = 0.025]. Participants with a history of smoking had a faster decline in SRTs compared with participants who never smoked [F(2, 675) = 4.89, p = 0.008].


Identifying factors that contribute to a decline in speech recognition ability in noise over time is highly valuable. This knowledge may be used for prevention and timely interventions to avert hearing loss and its broad psychosocial and daily life consequences. The present study reported on the analyses of longitudinal data from an online prognostic cohort study using a speech in noise test to assess an individual’s hearing status. We examined the 10-year change in speech recognition ability in noise in adults aged 18 to 70 years. In addition, we investigated the effects of age, sex, educational level, smoking, and alcohol use on this decline.

On average, the SRT deteriorated by 0.89 dB SNR over 10 years. This is a significant difference over time. It corresponds to an 18% decrease in speech understanding in specific difficult listening situations (Smits et al. 2004). SRTs in the age groups 51 to 60 and 61 to 70 years deteriorated faster compared with younger age groups. In these 2 groups, the estimated deterioration in speech recognition in noise was 1.37 and 1.69 dB SNR over 10 years, respectively, corresponding to a 27% and 34% decrease in speech understanding in specific difficult listening situations (Smits et al. 2004). Our study is among the first studies using a 10-year follow-up interval, showing that this accelerated decline in SRT in noise already starts at the (baseline) age of 51 to 60 years when comparing the decline to participants aged 18 to 50 years. This finding confirms the results of our previous study on the 5-year decline in speech recognition in noise (Stam et al. 2015) and the results of the cross-sectional studies using UK Biobank data which also used the NHT (Moore et al. 2014). It is worth to note that after 10 years, the average speech recognition in noise in participants aged 41 to 50 years changed from “good” to “insufficient” (Smits et al. 2006). Participants who are using the NHT for self-screening in the Netherlands are recommended to seek help for their hearing problems when they have the outcome “insufficient” or “poor” hearing (Smits et al. 2006). As hearing loss is associated with social and emotional loneliness (Stam et al. 2016), it would be interesting to examine whether a change from “good” to “insufficient” over 10 years follow-up would have the same effect on an individual’s psychological wellbeing as a change from insufficient to poor or from good to poor. In any case, our data show that researchers and clinicians should be aware that the accelerated decline in hearing starts already at the age of 50 years. Actions taken to ameliorate the potential effects of hearing loss should be considered not just in the elderly, but at any age when a hearing loss becomes apparent.

The estimated decrease in SRT between T1 and T2 was smaller compared with that between T0 and T1. This might partly be explained by a ceiling effect caused by the maximum score that can be obtained with the NHT. The maximum score is set at 4.0 dB SNR. This finding could also be the result of selective attrition at T2, loss of those participant with a more unfavorable SRT threshold. If selective attrition would have occurred at T2, the actual deterioration in hearing would have been underestimated for the T1–T2 interval. However, this trend was also observed in the sensitivity analysis, arguing against selective attrition. Another possibility is that the relatively small sample size at T2 let to a sampling error causing an underestimation of the actual change in SRT.

Speech recognition in participants with a history of tobacco smoking deteriorated faster compared with participants who had never smoked. The adjusted model with baseline corrections for sex, age group, and history of smoking showed that participants with a higher age and with a history of smoking had higher (i.e., more unfavorable) SRTs. History of smoking and age group each showed a significant interaction with time but no significant three-way interaction was observed between these two variables and time. This could be explained by the fact that the power was too low for testing three-way interactions. Another possibility is that, given that participants with a history of tobacco smoking were on average older compared with nonsmokers, the increased deterioration of SRT in participants with a history of smoking was partly an effect of age.

Other studies also demonstrated an association between tobacco smoking and hearing loss. To illustrate, a meta-analysis by Nomura et al. (2005) provided evidence for a significant correlation between tobacco smoking and decreased pure-tone thresholds. Also in longitudinal studies, the detrimental effect of smoking on pure-tone thresholds and self-reported hearing loss has been reported (Nakanishi et al. 2000,Burr et al. 2005,Cruickshanks et al. 2015). Using speech recognition in noise tests, this association was only found cross-sectionally (Dawes et al. 2014) but not longitudinally (Pronk et al. 2013). Our study confirms the association between tobacco smoking and hearing loss by the observation of a steeper decline in speech recognition in noise in participants with a history of smoking.

Several hypotheses on the ototoxic effects of tobacco smoking on cochlear function have been published. For instance, (a) Toxins such as nicotine, carbon monoxide, and hydrogen cyanide, all present in cigarette smoke might cause hearing loss (Chen & Fechter 1999; Harkrider et al. 2001; Fechter et al. 2002a), (b) Atherosclerosis which is associated with tobacco smoking is associated with hearing loss, and (c) hypoxia induced by tobacco smoking has been associated with hearing loss (Howard et al. 1998; Chen 2002). Nicotine induces an increased latency in wave I of the auditory brainstem response and a decrease in the wave I amplitude, indicative of an inhibitory effect of nicotine on the eighth cranial nerve (Harkrider et al. 2001). Carbon monoxide and hydrogen cyanide can potentiate noise-induced hearing loss in rats causing outer hair cell damage (Chen & Fechter 1999; Fechter et al. 2002a, b). A synergistic effect between tobacco smoking and noise-induced hearing loss has been observed in humans as well (Mizoue et al. 2003; Ferrite & Santana 2005; Wang et al. 2017). In addition, noise-induced hearing loss in rats was potentiated by decreased blood oxygen levels and caused inner and outer hair cell loss (Chen 2002).

Smoking may also induce atherosclerosis which could add to cochlear hypoxia (Howard et al. 1998). There is evidence showing that the transient evoked otoacoustic emission amplitude in normal-hearing subjects is reduced in smokers compared with nonsmokers, indicative of outer hair cell damage (Vinay 2010). Another factor could be that tobacco smoking affects central auditory pathways. An increasing amount of research suggests that smoking is related to neurobiological abnormalities and cognitive impairment (Durazzo et al. 2014). Tobacco smokers have higher oxidative stress leading to cellular damage; have increased tau pathology and significant amyloid beta deposition; and have an increased risk for cardiovascular disease (CVD) and cerebrovascular disease compared with nonsmokers (Durazzo et al. 2014). It can be hypothesized that the factors mentioned by Durazzo et al. (2014) also affect the central auditory pathways and have a negative impact on the SRT. However, the exact mechanism of the ototoxic effect of tobacco smoking on the accelerated decline in speech recognition in noise observed in (ex-)smokers is unclear. Future studies are recommended to confirm these hypotheses.

Tobacco smoking is known to have a variety of other deleterious effects on health causing diseases such as CVD, chronic obstructive pulmonary disease, and various types of cancer. Increased hearing loss in people with a history of smoking might therefore also be a secondary effect caused by other diseases associated with smoking such as CVD and its treatments. In addition, it is important to mention that the effect of smoking on the progression of hearing loss, as shown in this study, adds to healthcare expenditures (via the treatment of hearing loss).

Our results demonstrated that females had a higher SRT, but no longitudinal differences in the decline of SRTs between sexes were observed. Other longitudinal studies reported contradictory results with respect to gender. Using pure-tone audiometry, some studies reported a faster progression of hearing loss in male participants than in female participants (Karlsmose et al. 2000; Cruickshanks et al. 2010; Linssen et al. 2014). Other studies found no sex difference in progression of hearing loss, or only frequency-specific differences (Mitchell et al. 2011; Kiely et al. 2012). Dubno et al. (2008) found a faster decline of word recognition in quiet in females than in males. This difference was not found when using the NHT in the longitudinal study of Pronk et al. (2013), nor was it observed in the cross-sectional study of Moore et al. (2014). Sex differences in hearing thresholds might be hormonally induced or be confounded by population-specific exposures creating a sex difference in the progression of hearing loss. However, longitudinal data, using speech recognition in noise tests do not confirm a sex-specific rate of decline in SRTs.

In the present study, no association between educational level and a 10-year change in speech recognition in noise was found. Though lower education has been shown to be associated with a higher prevalence of hearing loss (Kiely et al. 2012), no association between these variables was found in this and other longitudinal studies (Kiely et al. 2012; Pronk et al. 2013). The absence of a (longitudinal) association between speech recognition in noise and alcohol consumption in this study seems to be consistent with previous research in this domain. To date, there are no longitudinal studies in which a significant association between hearing loss and the use of alcohol was observed (Brant et al. 1996; Karlsmose et al. 2000; Gopinath et al. 2010; Cruickshanks et al. 2015).

Strengths and Limitations

Strengths of the present study are the longitudinal data collection with a long follow-up, and the use of a speech recognition in noise test in a diverse study sample, including normal-hearing as well as hearing-impaired adults with a wide age range (18 to 70 years). The online data collection setting allowed people to easily participate in all measurements. This way we created a large dataset with a 10-year follow-up. By using longitudinal data, participants served as their own controls, thereby minimizing the effect of individual differences introducing selection bias as is often the case in cross-sectional studies.

There are several limitations of our study that need to be mentioned. The first is the known measurement error of the NHT (i.e., approximately 1 dB SNR)(Smits & Houtgast 2005). The relatively large measurement error reduces the sensitivity to small changes in SRT. Another limitation concerns the online data collection. Although convenient and cost-effective, it does not allow an accurate controlling of the test condition and administration. This may imply that in some cases, the NHT was done under suboptimal conditions, such as with poor quality transducers or in a noisy environment. We attempted to control for these factors by giving clear testing instructions and by controlling for transducer type in the statistical analyses, but we cannot exclude that the NHT was done in suboptimal conditions in some cases.

The convenience sampling method we used may also have resulted in sampling error expressed as differences for specific variables at baseline. For example, selection bias could have caused cross-sectional effects of age on SRT at baseline compromising the possibility to compare age groups (Morrell et al. 2009).

Also, conductive hearing losses are not perceived by the NHT, because the NHT only detects a sensorineural type of hearing loss. Nevertheless, we argue that this has had a negligible effect on our results. We allowed participants to adapt and increase the volume of the sound to a preferred level before testing, minimizing possible effects of a conductive hearing loss on the SRT.

Furthermore, we used self-reports to measure alcohol use. Although self-reported alcohol use has been demonstrated to be reasonably reliable and valid (Del Boca & Darkes 2003), response accuracy is known to be influenced by the interaction of social context factors and respondent characteristics and can be subject to response and recall bias (Del Boca & Darkes 2003). Tobacco smoking was not quantified but categorized into “having a history of smoking” and “nonsmokers.” For participants with a history of tobacco smoking, we did not have data on how many years before study entry they stopped smoking. Also, we only had a small sample of participants who stopped smoking during follow-up, making it impossible to accurately estimate the effect of smoking cessation. In addition, more details about tobacco smoking such as pack-years and information about the exact moment of smoking cessation would be needed to further substantiate our knowledge on the relation between tobacco smoking, smoking cessation, and hearing loss.


This longitudinal study revealed that the average 10-year decline in speech recognition in noise in a population aged 18 to 70 years at baseline is 0.89 dB SNR. This change may seem small, but it has a significant impact on a person’s ability to understand speech in challenging acoustic situations. The decline in speech recognition in noise was significantly larger in groups aged 51 to 60 and 61 to 70 years compared with younger age groups (18 to 30, 31 to 40, and 41 to 50 years) (p < 0.001). Speech recognition in noise in participants with a history of smoking declined significantly faster during the 10-year follow-up interval (p = 0.003). We did not find any differences in the decline of speech recognition in noise with regard to sex, educational level, and alcohol use.

Our findings confirm that ARHL is not a process limited to old age (i.e., >70 years), but occurs much earlier in life. Identification of factors associated with the decline in speech recognition in noise is indispensable for further actions such as preventive measures. Further longitudinal research is needed to investigate the interplay between a range of factors related to smoking and hearing decline, such as CVDs, their treatment, and other risk factors.


The authors thank the participants on the Netherlands Longitudinal Study on Hearing (NL-SH). The authors also thank the assistance of Celina Henke in managing the database.


Brant L. J., Gordon-Salant S., Pearson J. D., et al. Risk factors related to age-associated hearing loss in the speech frequencies. J Am Acad Audiol, (1996). 7, 152–160.
Burr H., Lund S. P., Sperling B. B., et al. Smoking and height as risk factors for prevalence and 5-year incidence of hearing loss. A questionnaire-based follow-up study of employees in Denmark aged 18-59 years exposed and unexposed to noise. Int J Audiol, (2005). 44, 531–539.
Chen G. D. Effect of hypoxia on noise-induced auditory impairment. Hear Res, (2002). 172, 186–195.
Chen G. D., Fechter L. D. Potentiation of octave-band noise induced auditory impairment by carbon monoxide. Hear Res, (1999). 132, 149–159.
Cruickshanks K. J., Nondahl D. M., Tweed T. S., et al. Education, occupation, noise exposure history and the 10-yr cumulative incidence of hearing impairment in older adults. Hear Res, (2010). 264, 3–9.
Cruickshanks K. J., Nondahl D. M., Dalton D. S., et al. Smoking, central adiposity, and poor glycemic control increase risk of hearing impairment. J Am Geriatr Soc, (2015). 63, 918–924.
Culling J. F., Zhao F., Stephens D. The viability of speech-in-noise audiometric screening using domestic audio equipment. Int J Audiol, (2005). 44, 691–700.
Dawes P., Cruickshanks K. J., Moore D. R., et al. Cigarette smoking, passive smoking, alcohol consumption, and hearing loss. J Assoc Res Otolaryngol, (2014). 15, 663–674.
Del Boca F. K., Darkes J. The validity of self-reports of alcohol consumption: state of the science and challenges for research. Addiction, (2003). 98(Suppl 2), 1–12.
Divenyi P. L., Stark P. B., Haupt K. M. Decline of speech understanding and auditory thresholds in the elderly. J Acoust Soc Am, (2005). 118, 1089–1100.
Dubno J. R. Speech recognition across the lifespan: Longitudinal changes from middle age to older adults. Am J Audiol, (2015). 24, 84–87.
Dubno J. R., Lee F. S., Matthews L. J., et al. Longitudinal changes in speech recognition in older persons. J Acoust Soc Am, (2008). 123, 462–475.
Durazzo T. C., Mattsson N., Weiner M. W; Alzheimer’s Disease Neuroimaging Initiative. (Smoking and increased Alzheimer’s disease risk: A review of potential mechanisms. Alzheimers Dement, (2014). 10(3 Suppl), S122–S145.
Fechter L. D., Chen G. D., Johnson D. L. Potentiation of noise-induced hearing loss by low concentrations of hydrogen cyanide in rats. Toxicol Sci, (2002a). 66, 131–138.
Fechter L. D., Chen G. D., Rao D. Chemical asphyxiants and noise. Noise Health, (2002b). 4, 49–61.
    Ferrite S., Santana V. Joint effects of smoking, noise exposure and age on hearing loss. Occup Med (Lond), (2005). 55, 48–53.
    Gates G. A., Cooper J. C. Incidence of hearing decline in the elderly. Acta Otolaryngol, (1991). 111, 240–248.
    Gopinath B., Flood V. M., McMahon C. M., et al. The effects of smoking and alcohol consumption on age-related hearing loss: The Blue Mountains Hearing Study. Ear Hear, (2010). 31, 277–282.
    Harkrider A. W., Champlin C. A., McFadden D. Acute effect of nicotine on non-smokers: I. OAEs and ABRs. Hear Res, (2001). 160, 73–88.
    Heywood R., Gao Q., Nyunt M. S. Z., et al. Hearing loss and risk of mild cognitive impairment and dementia: Findings from the Singapore Longitudinal Ageing Study. Dement Geriatr Cogn Disord, (2017). 43, 259–268.
    Houtgast T., Festen J. M. On the auditory and cognitive functions that may explain an individual’s elevation of the speech reception threshold in noise. Int J Audiol, (2008). 47, 287–295.
    Howard G., Wagenknecht L. E., Burke G. L., et al. Cigarette smoking and progression of atherosclerosis: The Atherosclerosis Risk in Communities (ARIC) Study. JAMA, (1998). 279, 119–124.
    Huang Q., Tang J. Age-related hearing loss or presbycusis. Eur Arch Otorhinolaryngol, (2010). 267, 1179–1191.
    Jansen S., Luts H., Wagener K. C., et al. The French digit triplet test: A hearing screening tool for speech intelligibility in noise. Int J Audiol, (2010). 49, 378–387.
    Karlsmose B., Lauritzen T., Engberg M., et al. A five-year longitudinal study of hearing in a Danish rural population aged 31-50 years. Br J Audiol, (2000). 34, 47–55.
    Kiely K. M., Gopinath B., Mitchell P., et al. Cognitive, health, and sociodemographic predictors of longitudinal decline in hearing acuity among older adults. J Gerontol A Biol Sci Med Sci, (2012). 67, 997–1003.
    Kramer S. E., Kapteyn T. S., Festen J. M. The self-reported handicapping effect of hearing disabilities. Audiology, (1998). 37, 302–312.
    Lee F. S., Matthews L. J., Dubno J. R., et al. Longitudinal study of pure-tone thresholds in older persons. Ear Hear, (2005). 26, 1–11.
    Lin F. R., Yaffe K., Xia J., et al.; Health ABC Study Group. (Hearing loss and cognitive decline in older adults. JAMA Intern Med, (2013). 173, 293–299.
    Linssen A. M., van Boxtel M. P., Joore M. A., et al. Predictors of hearing acuity: Cross-sectional and longitudinal analysis. J Gerontol A Biol Sci Med Sci, (2014). 69, 759–765.
    Mathers C., Fat D. M., Boerma J. T. World Health Organization. The Global Burden of Disease: 2004 Update. (2008). Geneva: World Health Organization.
    Mitchell P., Gopinath B., Wang J. J., et al. Five-year incidence and progression of hearing impairment in an older population. Ear Hear, (2011). 32, 251–257.
    Mizoue T., Miyamoto T., Shimizu T. Combined effect of smoking and occupational exposure to noise on hearing loss in steel factory workers. Occup Environ Med, (2003). 60, 56–59.
    Moore D. R., Edmondson-Jones M., Dawes P., et al. Relation between speech-in-noise threshold, hearing loss and cognition from 40-69 years of age. PLoS One, (2014). 9, e107720.
    Morrell C. H., Brant L. J., Ferrucci L. Model choice can obscure results in longitudinal studies. J Gerontol A Biol Sci Med Sci, (2009). 64, 215–222.
    Morrell C. H., Gordon-Salant S., Pearson J. D., et al. Age- and gender-specific reference ranges for hearing level and longitudinal changes in hearing level. J Acoust Soc Am, (1996). 100(4 Pt 1), 1949–1967.
    Mulrow C. D., Aguilar C., Endicott J. E., et al. Quality-of-life changes and hearing impairment. A randomized trial. Ann Intern Med, (1990). 113, 188–194.
    Nakanishi N., Okamoto M., Nakamura K., et al. Cigarette smoking and risk for hearing impairment: A longitudinal study in Japanese male office workers. J Occup Environ Med, (2000). 42, 1045–1049.
    Nomura K., Nakao M., Morimoto T. Effect of smoking on hearing loss: Quality assessment and meta-analysis. Prev Med, (2005). 40, 138–144.
    Plomp R., Mimpen A. M. Speech-reception threshold for sentences as a function of age and noise level. J Acoust Soc Am, (1979). 66, 1333–1342.
    Popelka M. M., Cruickshanks K. J., Wiley T. L., et al. Moderate alcohol consumption and hearing loss: A protective effect. J Am Geriatr Soc, (2000). 48, 1273–1278.
    Potgieter J. M., Swanepoel d. e. W., Myburgh H. C., et al. Development and validation of a smartphone-based digits-in-noise hearing test in South African English. Int J Audiol, (2015). 55, 405–411.
    Pronk M., Deeg D. J., Festen J. M., et al. Decline in older persons’ ability to recognize speech in noise: The influence of demographic, health-related, environmental, and cognitive factors. Ear Hear, (2013). 34, 722–732.
    Reinhard O. P. M., Rood-Bakker D. S. Alcohol Gebruik in Beeld. Standaard Meetlat. (1998). Rotterdam, the Netherlands.Nederlands Economisch Instituut.
    Rosenhall U., Sixt E., Sundh V., et al. Correlations between presbyacusis and extrinsic noxious factors. Audiology, (1993). 32, 234–243.
    Smits C., Houtgast T. Results from the Dutch speech-in-noise screening test by telephone. Ear Hear, (2005). 26, 89–95.
    Smits C., Kapteyn T. S., Houtgast T. Development and validation of an automatic speech-in-noise screening test by telephone. Int J Audiol, (2004). 43, 15–28.
    Smits C., Merkus P., Houtgast T. How we do it: The Dutch functional hearing-screening tests by telephone and internet. Clin Otolaryngol, (2006). 31, 436–440.
    Stam M., Smits C., Twisk J. W., et al. Deterioration of speech recognition ability over a period of 5 years in adults ages 18 to 70 years: Results of the Dutch Online Speech-in-Noise Test. Ear Hear, (2015). 36, e129–e137.
    Stam M., Smit J. H., Twisk J. W., et al. Change in psychosocial health status over 5 years in relation to adults’ hearing ability in noise. Ear Hear, (2016). 37, 680–689.
    Thomson R. S., Auduong P., Miller A. T., et al. Hearing loss as a risk factor for dementia: A systematic review. Laryngoscope Investig Otolaryngol, (2017). 2, 69–79.
    Vinay. Effect of smoking on transient evoked otoacoustic emissions and contralateral suppression. Auris Nasus Larynx, (2010). 37, 299–302.
    Wang D., Wang Z., Zhou M., et al. The combined effect of cigarette smoking and occupational noise exposure on hearing loss: Evidence from the Dongfeng-Tongji Cohort Study. Sci Rep, (2017). 7, 11142.
    Watson C. S., Kidd G. R., Miller J. D., et al. Telephone screening tests for functionally impaired hearing: current use in seven countries and development of a US version. J Am Acad Audiol, (2012). 23, 757–767.

    Age-related hearing loss; Alcohol; Epidemiology; National Hearing test; Presbyacusis; Risk factors; Smoking

    Copyright © 2019 The Authors. Ear & Hearing is published on behalf of the American Auditory Society, by Wolters Kluwer Health, Inc.