Hair cortisol as a viable tool for the assessment of an association between environmental noise exposure and chronic stress^a)

Beyond hearing impairment and tinnitus, chronic exposure to noise can be associated with reactions that may lead to adverse health effects. These include chronic annoyance responses, sleep disturbance, and general effects on well-being. Likewise, changes in blood pressure, heart rate, and circulating stress hormones are documented responses to noise exposure (Basner et al., 2014; Lusk et al., 2004). The hypothalamic–pituitary–adrenal (HPA) axis, a key coordinator of the body's stress response, may be an important mediator of noise-induced health impacts. However, direct biological evidence is somewhat sparse and the body of evidence from humans is inconsistent (Sivakumaran et al., 2022). Assessing chronic stress in relation to noise is challenging in a population setting, where the exposure levels can be relatively low and difficult to isolate, presenting challenges in drawing associations over time with relevant biomarkers. To a certain degree, these are minimised in occupational settings where noise exposures can be loud and well documented under hearing conservation programs. Nevertheless, typical biological measures of stress, including the assessment of the glucocorticoid stress hormone cortisol in blood, urine, and saliva, are subject to considerable temporal variability due to diurnal rhythms and other transient and chronic conditions (Edwards et al., 2001; Hennig et al., 2000; Legler et al., 1982).

Hair cortisol has been proposed as an alternative means of assessing chronic stress, as cortisol may integrate into hair over a period of time during stress, overcoming several of the shortcomings associated with the aforementioned measurement protocols (Wells et al., 2014; Russell et al., 2012; D'Anna-Hernandez et al., 2011; van Uum et al., 2008). However, questions have been raised about the utility of this end point for assessing temporal changes in stressor exposure (Kalliokoski et al., 2019; Kapoor et al., 2018).

The purpose of this paper is to discuss the strengths and limitations of using hair cortisol as a potential measure of chronic stress responses to environmental noise exposure. We briefly review the biological pathways through which exposure to noise may contribute to adverse health effects, and examine associations between noise and cortisol measured in various biological matrices. Limitations associated with using cortisol measured in blood, saliva, and/or urine to evaluate chronic stress effects from environmental noise exposure, and the potential use of hair cortisol as an additional and/or alternate measure of chronic noise-related stress, are discussed. It is not the authors' intention to provide an exhaustive systematic review of the published literature on hair cortisol in relation to the numerous conditions evaluated to date. The reader interested in this history is referred to several recent reviews (Kitani et al., 2022; Bryson et al., 2021; El Mlili et al., 2021; Psarraki et al., 2021; Malisiova et al., 2021; Kalliokoski et al., 2019; Stalder et al., 2017; Wester and van Rossum, 2015; Gow et al., 2010). While the emphasis in the present work is on noise as an environmental stressor, issues discussed here would be applicable to all studies that utilise hair cortisol as an indication of chronic stress.

Three plausible pathways through which exposure to noise may lead to adverse health effects have been proposed (Babisch, 2002). The first is by directly contributing to hearing impairment through prolonged unprotected exposure to noise above 80–85 A- weighted decibels (dBA). This sound level is approximately equivalent to an environment that would require one to raise one's voice in order to communicate with someone at arm's length (Feder et al., 2019). The second is through sleep disruption insofar as impaired sleep can lead to next-day effects and if sustained, chronic health effects. Next-day effects relate to fatigue, memory impairments, and increased risk of accidents, while long-term effects include changes in physiological parameters involved in mental health, blood pressure, appetite, glucose regulation, and weight gain (Jafari, 2017; Aguirre, 2016; Westerterp-Plantenga, 2016; Rasch and Born, 2013; Killgore, 2010; World Health Organization, 1999). The third pathway through which noise exposure may affect health is by causing an increase in stress reactions, where protracted amplification may increase the risk of developing stress-related health effects, such as the manifestation of cardiovascular diseases (World Health Organization, 2018). Health effects that may result from exposures below the threshold for noise-induced hearing loss are highly dependent on personal and situational variables that include, but transcend the actual noise exposure. These outcomes are most relevant in the context of typical environmental noise exposures, and are the focus of this paper.

As mentioned in Sec. I, noise can act as a stressor as evidenced through an increase in stress reactions that include, but are not necessarily limited to, changes in cortisol, adrenaline, epinephrine, heart rate, and blood pressure (Basner et al., 2014; Lusk et al., 2004). When protracted, these changes may mediate putative adverse health effects of noise, which underscores the value of quantifying these and other parameters in populations exposed to noise. A recent systematic review and meta-analysis on the strength of the association between noise and changes in these biomarkers showed that the certainty of evidence across the various studies was very low (Sivakumaran et al., 2022). Indeed, where other systematic reviews have found evidence supporting an association between elevated noise exposures and cardiovascular disease end points, the associations tended to be statistically weak and inconsistent (van Kamp et al., 2020; van Kempen et al., 2018; Dzhambov and Dimitrova, 2017; Fu et al., 2017).

Assessing chronic measures of noise-induced stress is challenging because there is no known outcome that is specific to noise and many of the so-called “stress biomarkers” demonstrate a sensitivity to a wide variety of non-acoustic stimuli, both internal and external, that are capable of causing “stress-like” reactions in humans. Controlling for these potentially confounding factors can be exceedingly difficult to achieve in epidemiologic studies. A thorough examination of these variables and the complexities that underscore the connection between stress and health is presented by Anisman (2015).

Measuring cortisol in biological media provides an estimation of HPA axis activity. Major differences among the sampling forms include the physiological integration time encompassed by the measurement, how sensitive the measure is to circadian rhythms, and the degree of invasiveness during sampling. The first two factors are briefly discussed below; the latter must be considered in terms of how the sampling regimen affects participant participation and attrition over the course of periodic sampling.

Stress-induced cortisol changes have traditionally been measured using blood (plasma/serum) and saliva samples, which can be difficult to interpret in relation to chronic exposures because both are transient measures and sensitive to the cortisol circadian rhythm (Smith and Vale, 2006), which should accordingly be considered in sampling strategies. This may include repeated sampling across a day and/or over a period of several days; however, this can be logistically challenging and onerous to participants. In plasma, cortisol is present both bound to cortisol-binding globulin and albumin, and unbound in its biologically active form. Although these forms are separable, this requires additional processing of the sample. Salivary cortisol, on the other hand, is unbound and able to bind to its receptors, providing direct measurement of its active form. While measuring salivary cortisol is attractive as a less invasive route, providing a readout of active cortisol levels, it is vulnerable to contamination and is sensitive to a number of potential modifiers, including food consumption, smoking, teeth-brushing, transient stressors, day of the week, time of day, time of awakening, etc. (Lefèvre et al., 2017; Pritchard et al., 2017; Garde et al., 2009; Hennig et al., 2000). This, in combination with protocols that require repeated sampling, can increase the likelihood of participant attrition and/or non-compliance with the sampling protocol (Broderick et al., 2004). More frequent sampling also factors into sample management and analytic cost.

Biologically active cortisol can be measured through urine sampling. While arguably less invasive than blood collection, repeated sampling is required to estimate daily urinary cortisol concentrations and multiple measures are necessary to investigate long-term stress responses, which makes studies more resource demanding, while also increasing participant attrition.

Several studies have investigated changes in cortisol concentrations in blood, saliva, and urine in relation to noise exposure; the study characteristics from a sample of studies from 1980 to 2022 are presented in Table I (Radun et al., 2022; Jafari et al., 2021; Yaghoubi et al., 2021; Smith et al., 2020; Zare et al., 2019; Jafari, 2017; Pouryaghoub et al., 2016; Ising and Ising, 2002; Evans et al., 2001; Melamed and Bruhis, 1996; Thomson et al., 2009).

When analyzing the long-term effects of environmental noise on cortisol, the two largest studies to date are the French study, “Discussion sur les Effets du Bruit des Aéronefs Touchant la Santé (DEBATS) (Lefèvre et al., 2017) and the Hypertension and Exposure to Noise Near Airports (HYENA) study (Selander et al., 2009). In a HYENA sub-sample, 439 participants provided saliva samples. After adjusting for several confounding variables, the results indicated an increase in morning salivary cortisol concentrations among women–but not men–exposed to daily average sound levels above 60 dBA, when compared to those living in areas below 50 dBA. In the more recent DEBATS study, after adjustments and exclusions, Lefèvre et al. (2017) reported reduced hourly variation in cortisol concentrations among females, but not males, 18 years of age and older living in areas where aircraft noise was above 60 dBA, compared to areas below 55 dBA. The effect appeared to be owing to elevated evening concentrations, insofar as morning cortisol was unrelated to aircraft noise level. Furthermore, the effect of aircraft noise exposure on evening cortisol levels was stronger when dwelling residency was at least 5 y.

In a pooled analysis of these two large studies (i.e., the HYDE analysis), Baudin et al. (2019) assessed the impact of aircraft noise on salivary cortisol in 1673 subjects between ages 45–70 y. The authors reported statistically significant elevations in evening, but not morning cortisol, among women exposed to aircraft noise. There were no changes at either time among males. Collectively, the results appear to provide some evidence that women who live near airports continue to respond over time to aircraft noise with an increase in salivary cortisol excretion. Observations in women, but not men, point to a potential influence of circulating sex hormones on salivary cortisol. Neither the HYENA or the DEBATS study adjusted for menstrual cycle, or the use of oral contraceptives. Although menstrual cycle may have minimal influence on the salivary cortisol awakening response (Kudielka and Kirschbaum, 2003), oral contraceptives have been found to increase resting total cortisol levels and blunt the salivary cortisol response to stressors (Bouma et al., 2009) potentially through their ability to upregulate cortisol-binding globulin (Klipping et al., 2021).

Many of the challenges associated with short-term sampling may be overcome by examining cortisol levels in hair samples (Russell et al., 2012; Stalder et al., 2012). It has been proposed that cortisol integrates and remains in hair over time as it grows further from the scalp. Variation across the body exists, but at the posterior vertex of the scalp, an estimated average growth rate of approximately 1 cm per month is generally accepted (Wennig, 2000). On the evidence that hair cortisol is integrated, and retained with a predictable growth rate, it has been argued that the concentration of cortisol in hair segments can be used to retrospectively characterize cortisol levels over several months (Wester and van Rossum, 2015; Russell et al., 2012). For this reason, hair cortisol analysis has become an increasingly utilized methodology for examining associations between chronic stress and human health (El Mlili et al., 2021; Schaafsma et al., 2021; Malisiova et al., 2021). Elevated hair cortisol has also been reported among individuals suffering from tinnitus (Basso et al., 2022).

Importantly, the underlying assumption that hair cortisol reflects an integrated measure is supported by the findings reported by Short et al. (2016) where cortisol measured in a 1 cm segment of near-scalp hair from the posterior vertex was most strongly associated with an integrated salivary measure derived from three samples per day over a 30 day period (Pearson's r = 0.61, p < 0.01) rather than one integrated over more proximal periods (such as the week prior to hair sampling). These findings support the contention that cortisol measured from a 1 cm near-scalp hair sample provides a reasonable measure of long-term, but not short-term, cortisol levels. This observation is particularly relevant to environmental noise where the putative link to cardiovascular health pertains to long-term exposures.

Davenport et al. (2006) observed that relocation stress in Rhesus monkeys was associated with protracted increases in hair cortisol concentrations in shaved samples from the posterior vertex region of the neck 14 weeks after relocation, which was consistent with salivary cortisol changes and behavioural observations. van Uum et al. (2008) reported that hair cortisol concentrations in a near-scalp 2 cm–section from the posterior vertex were nearly double in human subjects experiencing chronic pain compared to screened controls, a pattern also consistent with self-reported stress. Segments of near-scalp hair from the posterior vertex that corresponded to the third trimester of pregnancy, a period marked by hyper-cortisol secretion, had elevated cortisol concentrations in comparison to more distal segments that reflect the first and second trimester (D'Anna-Hernandez et al., 2011; Kirschbaum et al., 2009). A recent systematic review found a tendency for higher hair cortisol concentrations among individuals with clinical depression and suppressed levels among individuals with post-traumatic stress disorder, although studies are not consistent in the direction of change, or in the strength of the association between these disorders and changes in hair cortisol (Malisiova et al., 2021). In their large-scale study, Stalder et al. (2013) showed that hair cortisol concentrations measured from a 3 cm hair sample from the posterior vertex were elevated in individuals with metabolic syndrome, which remained statistically significant after adjusting for several confounding variables. Manenschijn et al. (2013) collected 3 cm near-scalp hair samples at the posterior vertex from 238 individuals. After adjusting for age, body mass index (BMI), waist circumference, gender, smoking status, and alcohol consumption, elevated hair cortisol concentrations were associated with elevated odds for coronary heart disease, stroke, and peripheral arterial disease, but not any of the non-cardiovascular diseases evaluated. Again, these findings are especially relevant to research on the potential association between environmental noise and health, where there has been sustained research over the last 35 years exploring the degree to which noise may contribute to the development of cardiovascular diseases (World Health Organization, 2018). Figure 1 illustrates the proposed mechanisms through which cortisol integrates into hair.

To our knowledge, the only study to date to evaluate hair cortisol concentrations in relation to environmental noise was published by Michaud et al. (2016). This study considered the association between modelled wind turbine noise (<25–46 dBA) and multiple measures of self-reported and objectively determined stress reactions. Physiologic measures of stress included an assessment of hair cortisol from 3 cm samples collected at the posterior vertex of the scalp, quantified using established protocols (Russell et al., 2012). While there was no association between any measure of stress and wind turbine noise, there was a significant correlation between hair cortisol and self-reported stress over the previous 30 days. The authors speculated that the weak association (i.e., Pearson r = 0.13, p < 0.0007) was attributed to the mismatch between the time reference period between the two measures (i.e., ∼90 days in the 3 cm hair segment vs 1 month in the self-reported assessment).

Despite its potential, the interpretation of hair cortisol as a biomarker that tracks historical changes in stress levels remains uncertain. There is evidence to suggest that the retention of cortisol in hair requires that the stressor must be ongoing (Kalliokoski et al., 2019; Kapoor et al., 2018). While this would presumably be true where noise exposure is relatively continuous (e.g., communities near airports, highways, industry, etc.), this raises the possibility that the historical record of transient exposures, or effects of fluctuations in exposure levels, may not be accurately captured, and therefore, may be more suitably evaluated using a method that is sensitive to such changes (e.g., saliva and/or plasma).

In their radiolabelled study in adult Rhesus monkeys, Kapoor et al. (2018) evaluated the hair cortisol retention hypothesis using a “shave re-shave” approach. Briefly, the authors conducted a baseline hair shave on the upper back before intravenously administering radiolabelled cortisol. Post-injection samples were collected 14 and 28 days from the previously shaved area (re-shave) and from an adjacent previously unshaven area (i.e., new shave). The re-shave area reflects new hair growth starting immediately after the radiolabelled cortisol injection, while the newly shaved area reflects ongoing hair growth. Under the retention hypothesis discussed in Sec. V, after adjusting for individual growth rate, the segment of the newly shaved hair that corresponded to the growth rate should approximate the concentration of cortisol measured in the re-shaved section. The authors demonstrated that radiolabelled cortisol was detectable in the re-shaved area, supporting the notion that circulating cortisol enters the hair shaft. However, radiolabelled cortisol was undetectable in the distal segment of hair in the newly shaved sample matched to the growth rate, which calls into question the notion that cortisol is retained in the hair shaft as it grows. Non-detectable radiolabeled cortisol in the distal segment of hair underscores the need for more research before 1 cm segments of hair can be used to infer with certainty the historical activity within the HPA axis. While this may be less of an issue in populations with sustained exposure to an environmental stressor, it may undermine the notion that hair samples can be segmented to assess historical HPA activity, where stressors may have come and gone.

Studies employing hair cortisol as a measure of stress should attempt to account for variables that influence cortisol. Addison's disease, Cushing's syndrome (Greff et al., 2019; Thomson et al., 2010), mental illnesses (Malisiova et al., 2021), topical corticosteroid medication applied to the scalp, pregnancy (D'Anna-Hernandez et al., 2011; Kirschbaum et al., 2009), chronic pain (van Uum et al., 2008), relocation stress (in animal models; Davenport et al., 2006), current substance use disorders (Stalder et al., 2010), and obesity (Jackson and Steptoe, 2018; Stalder et al., 2017; Olstad et al., 2016; Wells et al., 2014; Stalder et al., 2012) are among the many factors that have been shown to influence cortisol levels.

Other potential sources of stress (Kalliokoski et al., 2019) beyond environmental noise, and individual differences in stress reactivity (Singh et al., 1999), may contribute to the variance in hair cortisol concentrations, although the latter would be especially difficult to account for in large population-based studies. Survey questionnaire content can identify some influential variables and the extent to which respondents are experiencing stress concurrent with or preceding the time of sampling. Cohen's Perceived Stress Scale (PSS) (Cohen et al., 1983) is one tool that can assess current and historical levels of perceived stress over the previous 30 days. Although perceived stress has an inconsistent correlation with biological markers of stress, including hair cortisol concentrations (Prado-Gascó et al., 2019; Bossé et al., 2018), it bears repeating that one of the largest (n = 917) hair cortisol studies undertaken to date observed a statistically significant positive association between scores on the PSS and hair cortisol concentrations (Michaud et al., 2016) (see Sec. V). Furthermore, in their univariate analysis, the authors reported statistically significant elevations in hair cortisol concentrations among males, obese individuals (BMI ≥30), current smokers, and individuals with chronic health conditions [e.g., asthma, arthritis, tinnitus, chronic obstructive pulmonary disease (COPD), diabetes]. Lower cortisol concentrations were observed among individuals with lower education, cosmetic hair treatment, and among those with increased frequency of hair washing. Hair cortisol concentrations were unrelated to age, income, caffeine or alcohol consumption, migraines, vertigo, chronic pain, hypertension, heart disease, or measured blood pressure. Reporting to be pregnant at the time of sampling was also unrelated to hair cortisol concentrations. Despite its long-standing place in environmental noise research and strong association with noise annoyance, noise sensitivity was unrelated to hair cortisol concentrations in the study by Michaud et al. (2016). Their adjusted multiple linear regression analysis of the variables related to hair cortisol concentrations is reproduced in Table II.

Notwithstanding their importance, data will always be limited with respect to factors that could alter cortisol levels, including those unknown to the individual, and not captured through participant surveys. These include both intrinsic factors, such as genetic variability, and extrinsic factors, such as exposure to air pollutants (Thomson et al., 2021). Accordingly, as with any study design, potential confounders, mediators, and modifiers should be considered.

Hair growth rate is another potential source of variability when considering the period of stressor exposure captured by a given length of hair. Studies have shown that the growth rate of scalp hair varies among individuals (e.g., sex, ethnicity, age) (Loussouarn et al., 2016; Harkey, 1993). Hair sampling should be from the posterior vertex region, with the total sample having a diameter of 5–10 mm (about the width of a standard pencil) and yield a minimum mass of ∼10–15 mg. The posterior vertex is recommended by the Society for Hair Testing (Cooper et al., 2012), as it tends to have a large percentage of hair in new growth (anagen) phase (Harkey, 1993), displaying the most uniform growth rate and the lowest variability coefficient with respect to cortisol levels (Sauvé et al., 2007). Additionally, when collecting samples, hair should be cut as close to the skin as possible, while excluding the hair follicle, as there is evidence of peripherally synthesized cortisol from the follicles in response to physical stressors (Salaberger et al., 2016; Sharpley et al., 2009; Ito et al., 2005). Collected samples should be wrapped in aluminum foil to protect them from damage or contaminants, the scalp end of the sample should be marked, taped together and stored at room temperature until analysis (Wennig, 2000). Evidence that ultraviolet light from natural sunlight may decrease cortisol concentration (Wester et al., 2016) indicates hair should be stored in a dark location. Based on current evidence, cortisol in properly stored hair samples should be viable for several years (Webb et al., 2010).

Greff et al. (2019) recently summarized the various laboratory techniques used to quantify hair cortisol, along with their strengths and limitations. Enzyme-linked immunoassay (EIA) is a relatively simple approach based on established salivary cortisol assay kits. The method allows for handling of several samples simultaneously using relatively inexpensive equipment. However, cross-reactivity with other steroids (e.g., cortisol metabolites) was reported to vary by EIA assay manufacturer, producing cortisol values in excess (2.5–20 fold) of the same sample measured using liquid chromatography–mass spectrometry (LC–MS/MS) (Russell et al., 2015). The LC-MS/MS method, while more sensitive, is more labour and cost intensive, and may be less amenable for screening large numbers of samples. The EIA and LC-MS/MS methods tend to yield results that are highly correlated with each other (r² = 0.88–0.97) and correction factors for the various types of EIA assays tested have been proposed by Russell et al. (2015) if LC-MS/MS equivalent values are desired. For these reasons, the EIA method would be more suitable for large population-based studies that investigate associations between hair cortisol and environmental noise. Greff et al. (2019) and Gow et al. (2010) provided detailed descriptions on the steps involved in the EIA.

The issues discussed in Secs. V and VI are summarized in Fig. 2. The figure highlights the fundamental considerations related to noise exposure, subject recruitment, sample collection/storage, and hair sample processing.

Discrete measures of cortisol change in response to transient noise exposure, or changes in level, are most appropriately determined using either saliva, plasma, or urine sampling. However, where the primary interest is to evaluate the association between sustained environmental noise and long-term stress, hair cortisol analysis appears to have several advantages over these measures. Although there has been limited application of hair cortisol in environmental noise studies to date (Michaud et al., 2016), there is a large database evaluating hair cortisol in relation to sleep quality, stress, cardiovascular diseases, and other illnesses (Lob and Steptoe, 2019). All of these have a long history as outcomes of particular interest in relation to noise, with mixed findings on the association. Some of the uncertainty that exists with respect to biomarker changes in relation to noise (Sivakumaran et al., 2022) may be owing to the time-integration discrepancy between exposure and that represented by the sampling paradigm, which would be minimized with a long-term integrated biomarker, such as hair cortisol. This measure could be readily incorporated into large-scale aircraft, road traffic, and industrial noise studies. Hair cortisol could also be included in occupational studies to address knowledge gaps related to the putative association between long-term noise exposure and stress-related illnesses, including cardiovascular diseases (Michaud et al., 2021; Gan et al., 2011). After accounting for the use of personal hearing protection, occupational noise settings present a unique opportunity to explore the relationship between known exposures to loud noise and the potential impact this may have on hair cortisol. To the extent that a minimum mass of 10–15 mg can be collected, 1 cm hair samples would provide an indication of systemic cortisol secretion over approximately 30 days, aligning with the widely used Cohen's PSS.

Further research is warranted to better understand the potential to infer historical stress reactions using distal segmented hair sections. In many respects, concerns about the validity of a so-called “stress calendar” are less relevant for environmental noise research, where noise exposure is not restricted to a discrete point in time. However, where there is a need to investigate temporal changes in environmental noise (e.g., airport runway expansion), sampling 1 cm segments on a monthly basis would be preferable to segmenting a single sample, where growth rates may be uncertain and questions remain regarding cortisol retention.

This study was funded entirely by the Government of Canada. The authors declare that they have no conflicts of interest.