Psychoneuroendocrinology (2014) 39, 132—140
Available online at www.sciencedirect.com
ScienceDirect j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / p s y n e u e n
Predictors of hair cortisol concentrations in older adults Silke Feller a,*, Matthaeus Vigl a, Manuela M. Bergmann a, Heiner Boeing a, Clemens Kirschbaum b, Tobias Stalder b a
German Institute of Human Nutrition Potsdam-Rehbruecke, Department of Epidemiology, 14558 Nuthetal, Germany b Technical University of Dresden, Department of Psychology, 01069 Dresden, Germany Received 18 July 2013; received in revised form 10 October 2013; accepted 10 October 2013
KEYWORDS Hair cortisol; Chronic stress; Age; Predictors; Older age; Elderly; Human
Summary People at older ages are at increased risk for developing stress-related diseases associated with chronically elevated cortisol secretion. However, the main factors contributing to such endocrine alterations in this age group are still largely unknown. This cross-sectional study examined patterns of long-term integrated cortisol secretion, as assessed in hair, in a sample of 654 participants in middle and old adulthood (mean age: 65.8 years; range: 47—82 years) from the German cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC) study in Potsdam. Hair cortisol concentrations (HCC) were determined from the first scalp-near 3 cm hair segment and several sociodemographic, lifestyle, anthropometric, disease-related, and psychological parameters were assessed. In simple linear regressions, HCC were found to increase with participants’ age and to be higher in men compared to women. HCC also showed positive associations with waist-to-hip ratio, waist circumference, smoking, prevalent type 2 diabetes mellitus, mental health, daytime sleeping, and being unemployed or retired–—as well as a negative association with diastolic blood pressure. After full mutual adjustment, only age and smoking remained independent predictors of HCC. The association between prevalent type 2 diabetes mellitus and HCC was attenuated but still persisted independently in women. Similar, a positive relationship between HCC and alcohol consumption was found in women. The current results confirm previous evidence of positive associations of HCC with age, sex, alcohol consumption, and type 2 diabetes mellitus and add new knowledge on factors–—such as smoking–—that may contribute to elevated cortisol levels in people at older ages. # 2013 Elsevier Ltd. All rights reserved.
* Corresponding author at: German Institute of Human Nutrition (DIfE) Potsdam-Rehbruecke, Department of Epidemiology, ArthurScheunert-Allee 114-116, 14558 Nuthetal, Germany. Tel.: +49 033 200/88 2720; fax: +49 033 200/88 2721. E-mail address: [email protected] (S. Feller).
1. Introduction Long-term changes to the secretion of the glucocorticoid hormone cortisol, such as under conditions of chronic stress,
0306-4530/$ — see front matter # 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.psyneuen.2013.10.007
Predictors of hair cortisol in older adults are well-known to be associated with a range of diseases, including cardiometabolic and autoimmune diseases, as well as mental disorders (Chrousos, 2009). This particularly applies to older people who are at increased risk for these conditions and have often been found to exhibit chronically elevated cortisol levels (e.g. Laughlin and Barrett-Connor, 2000; Larsson et al., 2009). Previous research has identified a number of potential determinants of age-associated hypercortisolism in the elderly, including health status (e.g. Adam et al., 2006), physical functioning (Kumari et al., 2010), lifestyle-related factors (e.g. Badrick et al., 2008), central adiposity (Steptoe et al., 2004), as well as a wide range of psychosocial factors, including depressive symptoms (Penninx et al., 2007), loneliness (Hackett et al., 2012), sadness, anger, or lack of control (Adam et al., 2006). Previous endocrine research in middle and old adulthood has mostly relied on cortisol assessments in saliva and blood. While these methods accurately reflect acute changes in cortisol levels, they are not well-suited for capturing longterm cortisol alterations as their results are easily confounded by situational factors or insufficient compliance with sampling times (e.g. Kudielka et al., 2003). As age-associated factors are likely to manifest in persistent long-term changes to cortisol secretion, the use of a more stable assessment method may be advantageous. Here, recent evidence on the measurement of cortisol in human hair is of particular interest. Through continuous incorporation of cortisol into growing hair, hair cortisol concentrations (HCC) are assumed to mainly reflect integrated cortisol secretion over periods of several months (see Russell et al., 2012; Stalder and Kirschbaum, 2012). Considerable evidence now supports the general validity (Kirschbaum et al., 2009; Thomson et al., 2010; Manenschijn et al., 2011a) and test—retest reliability (Stalder et al., 2012b) of this method as well as its robustness to a range of hair-related factors (Dettenborn et al., 2012). To date, only a small number of studies have examined HCC in relation to aging. Evidence for increasing HCC with older age has emerged from a study covering a wide range of ages (Dettenborn et al., 2012) and from a recent investigation in a large sample of working people (Stalder et al., 2013) but was not observed by other research (e.g. Raul et al., 2004; Dettenborn et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2012a). Similarly, previous investigations into the role of participants’ sex on HCC, as an important factor associated with endocrine functioning as well as with health risk, were mostly conducted in younger populations, revealing no sex differences in HCC (e.g. Raul et al., 2004; Thomson et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2013) or suggesting lower HCC in women than in men (Dettenborn et al., 2012; Manenschijn et al., 2013). Interestingly, Dettenborn et al. (2012) found sex differences in HCC only in middleaged adults but not in the elderly, pointing toward a potential role of changing sex steroid levels with aging. By contrast, a recent investigation found lower HCC in women within an older study sample (Manenschijn et al., 2013). Research examining other correlates of HCC was also frequently carried out in younger populations, potentially limiting transferability of these findings to older people. Here, particularly associations of HCC with body fat-related anthropometric measures (Manenschijn et al., 2011a, 2011b; Stalder et al., 2012a, 2013), the metabolic syndrome (Stalder et al., 2013), cardiovascular risk (Manenschijn et al., 2013),
133 or myocardial infarction (Pereg et al., 2011) have been highlighted. Findings have been more inconsistent regarding HCC relationships with psychosocial stress, with most evidence pointing toward elevations of HCC in persons exposed to stressful conditions (Staufenbiel et al., 2013) but showing less consistent associations with self-reported stress (see Stalder and Kirschbaum, 2012). Finally, previous research has not found HCC to be related to smoking status, oral contraceptive use or overall medication intake (Dettenborn et al., 2012). Within the framework of the above evidence, the current study set out to investigate age-related changes of HCC in a large population of middle and old adulthood. To gain insight into determinants of long-term cortisol secretion in older age, relationships of HCC with a range of sociodemographic, lifestyle, disease-related, and psychological parameters were examined.
2. Methods 2.1. Study population, design and procedure The present sample was drawn from a substudy of the German cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC) study, a large multi-center prospective cohort study investigating associations between diet, lifestyle and chronic disease risk (German EPIC cohort: Boeing et al., 1999; EPIC: Riboli et al., 2002). The current substudy was conducted between the years 2010 and 2012 and employed a cross-sectional design. For this study, individuals residing in the wider Potsdam area were randomly selected from the main cohort with attention being paid to maintain an approximately uniform sample spread and sexdistribution across the examined age range. Individuals were invited to visit the study center at the German Institute of Human Nutrition where hair samples were obtained and physical examinations as well as inquiries on diet, physical activity, and psychosocial factors were conducted. These data were obtained from a total of 815 individuals. Of these, participants were excluded if they exhibited hair shorter than 3 cm at the posterior vertex region at the back of the head (n = 80), reported recent use of glucocorticoid-containing treatments (n = 44), refused to provide a hair sample (n = 1) or provided a hair strand of insufficient volume for cortisol analysis (n = 10). Thus, the final sample (prior to statistical exclusion; see below) comprised 680 participants (369 women) aged between 47 and 82 years. The study protocol was approved by the ethics committee of the Medical Association of the State of Brandenburg and written informed consent was obtained from all participants.
2.2. Hair cortisol analysis Hair strands of a diameter of approximately 3 mm were taken as close as possible to the scalp from a posterior vertex position using fine scissors. Cortisol concentrations were determined from the first 3 cm hair segment proximal to the scalp. Based on a hair growth rate of 1 cm/month (Wennig, 2000), this hair segment is assumed to reflect hair growth over the three-month period prior to hair sampling. Washing and steroid extraction followed the protocol
134 described in detail in Stalder et al. (2012b, study II) with 10 mg of whole, nonpulverised hair being incubated in 1800 ml methanol for 18 h at room temperature. Following extraction, cortisol concentrations were determined using a commercially available immunoassay with chemiluminescence detection (CLIA, IBL-Hamburg, Germany). The intraand interassay coefficients of variation of this assay are below 8%.
2.3. Measurement of psychological parameters Perceived chronic stress was assessed using the screening scale of the Trier inventory of chronic stress (TICS-SSCS; Schulz and Schlotz, 1999) with higher values reflecting increased chronic stress load. Physical and mental health were measured by calculating the physical and mental component summary scores according to the SF-12 algorithm (Windsor et al., 2006) with higher values signaling better subjective health status. Dispositional optimism was assessed by use of four questions derived from Giltay et al. (2006), namely ‘‘I still expect much from life,’’ ‘‘I do not look forward to what lies ahead of me in the years to come,’’ ‘‘My days seem to be passing by slowly,’’ and ‘‘I am still full of plans,’’ coded as 0 (fully in agreement), 1 (partially in agreement or do not know), or 2 (not in agreement). After reversing negatively keyed questions, a mean score was calculated with higher scores indicating greater optimism.
2.4. Measurement of socio-demographic, lifestyle, and disease-related variables Information on age, anthropometry, smoking status (nonsmoker, former smoker, and current smoker), physical activity, alcohol consumption (portions per day), occupation (full time, part time, casual vs. retirement, jobless/retraining, or unemployed), pain during the last 4 weeks (yes/no), menopausal status, as well as average night and daytime sleep was collected using questionnaires and physical examinations. Physical activity was defined using the Cambridge Index combining information on activity at work (sedentary to heavy manual) and leisure time (duration of sport and cycling in hours/week) to classify participants into ‘‘inactive,’’ ‘‘moderately inactive,’’ ‘‘moderately active,’’ or ‘‘active’’ (Wareham et al., 2003). Face-to-face interviews were conducted by study physicians as part of which participants were asked about prevalence of hypertension and overall medication intake (topical, oral, inhaled, or systemic drugs) during the last 7 days, which was used to identify glucocorticoid-containing treatments. Information on medically verified prevalent myocardial infarction, stroke, cancer, and type 2 diabetes mellitus at the time of the substudy was available from the data collected during follow-ups of the regular EPIC-Potsdam cohort study up to 2009 (Bergmann et al., 1999). Waist-to-hip-ratio (WHR), waist circumference, and body mass index (BMI) were measured by trained study personnel.
2.5. Statistical analysis Hair cortisol data were not normally distributed and logtransformed values were thus used for inferential analyses.
S. Feller et al. For descriptive purposes and unless otherwise noted, data in figures and tables are presented in original units. HCC outlying by 3 standard deviations (SD) from the mean were excluded (n = 18) after the initial elimination of extreme values >500 pg/mg (n = 8), resulting in a final sample of 654 participants for further analyses. In a first step, HCC associations with relevant sociodemographic, lifestyle, disease-related, and psychological parameters were investigated using simple linear regressions. Categorical predictor variables were dummy-coded for these analyses. In a second step, multiple regression models adjusted for sex and age were computed. Finally, multiple linear regression models in which all variables were entered simultaneously were conducted to investigate the independence of HCC relationships with individual predictors. This analysis included data of both men and women and thus the variables ‘‘hormone therapy’’ and ‘‘postmenopausal status’’ were not added to this model. To test for a potential influence of these variables, an identical regression model was computed in women only. All analyses were performed with SAS statistical software (release 9.2; SAS Institute Inc., Cary, NC). The level of statistical significance was set at p < 0.05 for two-tailed testing.
3. Results Table 1 provides descriptive information of the current study population. The sample comprised 654 participants (54% women) with a mean age of 65.8 years. Mean SD HCC in the whole sample were 35.1 32.8 pg/mg. Table 2 shows the results of simple and multiple linear regression analyses of associations with HCC. In simple regression analyses, HCC were found to increase linearly with participants’ age (b = 0.14, p < 0.001) and lower HCC were observed in women compared to men (b = 0.12, p = 0.002). Mean SD SD HCC were 37.9 34.0 pg/mg in men and 32.7 31.6 pg/ mg in women. Separate analyses within different age groups revealed that sex differences were evident in individuals younger than 60 years (b = 0.23, p = 0.001, n = 201) and between 60 and 70 years (b = 0.24, p = 0.0004, n = 207) but not in those older than 70 years (b = 0.05, p = 0.46, n = 246). Furthermore, HCC were found to be positively related to WHR, waist circumference, smoking, prevalent type 2 diabetes mellitus, mental health scores, daytime sleeping, and being unemployed or retired. An inverse relationship was observed between diastolic blood pressure and HCC. In multiple regression analysis, both age and sex remained significant predictors of HCC after mutual adjustment for each other (Table 2). After combined adjustment for sex and age, only smoking and prevalent type 2 diabetes mellitus persisted to be related to HCC. Associations of HCC with anthropometric measures (BMI, WHR, waist circumference) as well as diastolic blood pressure were attenuated to non-significant trends in these adjusted analyses. When including all variables in a multiple linear regression model, only participants’ age and smoking status remained independent predictors of HCC (Table 2). The mean SD HCC for different age groups were: 70 years 40.5 40.5 pg/mg. Regarding smoking status, mean SD HCC were 34.3 33.3 pg/mg in nonsmokers, 34.3 30.5 pg/mg in former smokers and 43.2 40.3 pg/ mg in current smokers. The adjusted HCC increase according to type 2 diabetes mellitus status ( p = 0.07) was reflected in a mean SD HCC of 46.9 46.8 pg/mg in diabetics compared to 34.1 31.0 pg/mg in non-diabetics. The separate regression model carried out in women (with influences of hormone therapy or postmenopausal status being accounted for) revealed independent relationships of HCC for age, alcohol consumption, and type 2 diabetes mellitus (b = 0.21; p = 0.02, b = 0.14; p = 0.04, and b = 0.14; p = 0.04, respectively). Effect sizes for each of the predictive parameters calculated by Cohen’s f 2 ranged between .01 and .02 in the final multiple linear regressions. The fraction of variance in HCC explained by these multiple linear regression models were R2 = 0.06 (both sexes) and R2 = 0.13 (in women). Figs. 1—3 illustrate relationships of HCC with parameters that emerged as predictors in the final models.
4. Discussion The current study investigated patterns of long-term cortisol secretion, as assessed in hair, in a larger middle and old adulthood sample. Within this population, our results show a linear increase in HCC with participants’ age and this relationship was maintained regardless of the influence of other age-related variables on HCC being accounted for. The current results also identify further relevant correlates of longterm cortisol levels within an older sample, particularly showing lower HCC in women than in men, higher HCC in current smokers as well as in people with type 2 diabetes mellitus or increased alcohol consumption (in women). The current findings add to previous evidence of hypercortisolism in old age derived from serum and salivary cortisol research (e.g. Laughlin and Barrett-Connor, 2000; Larsson et al., 2009). With regard to research on hair cortisol, our results are at variance with initial studies not showing a link between HCC and age (Raul et al., 2004; Dettenborn et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2012a). These studies, however, mostly drew on smaller samples of
136
S. Feller et al.
Table 2 Standardized single linear regression coefficients, age- and sex-adjusted linear regression coefficients, as well as multiple linear regression coefficients of study population characteristics and HCC (n = 654) within the EPIC-Potsdam substudy. Characteristics
Standardized simple linear regression coefficients
p-Value
Standardized multiple linear regression coefficients (adjusted for sex and age)a
p-Value
Standardized multiple linear regression coefficientsb (mutual adjustment)
p-Value
Age (years) Women vs. men Body mass index Waist-to-hip-ratio Waist circumference Physical activity Inactive, moderately inactive Moderately active, active Smoking Nonsmoker Former smoker Current smoker Alcohol consumption (portion/day) Occupation Full time, part time, casual Retirement, jobless/retraining, unemployed Prevalent myocardial infarction (yes vs. no) Prevalent stroke (yes vs. no) Prevalent cancer (yes vs. no) Prevalent type 2 diabetes mellitus (yes vs. no) Prevalent hypertension (yes vs. no) Diastolic blood pressure (mmHg) Systolic blood pressure (mmHg) Dispositional optimism score Perceived chronic stress score (TICS-SSCS) Physical health score (SF-12) Mental health score (SF-12) Hormone therapy (women, yes vs. no) Postmenopausal (women, yes vs. no) Pain (yes vs. no) Daytime sleep (yes vs. no) Night’s sleep (h)
0.14 0.12 0.07 0.15 0.13
Available online at www.sciencedirect.com
ScienceDirect j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / p s y n e u e n
Predictors of hair cortisol concentrations in older adults Silke Feller a,*, Matthaeus Vigl a, Manuela M. Bergmann a, Heiner Boeing a, Clemens Kirschbaum b, Tobias Stalder b a
German Institute of Human Nutrition Potsdam-Rehbruecke, Department of Epidemiology, 14558 Nuthetal, Germany b Technical University of Dresden, Department of Psychology, 01069 Dresden, Germany Received 18 July 2013; received in revised form 10 October 2013; accepted 10 October 2013
KEYWORDS Hair cortisol; Chronic stress; Age; Predictors; Older age; Elderly; Human
Summary People at older ages are at increased risk for developing stress-related diseases associated with chronically elevated cortisol secretion. However, the main factors contributing to such endocrine alterations in this age group are still largely unknown. This cross-sectional study examined patterns of long-term integrated cortisol secretion, as assessed in hair, in a sample of 654 participants in middle and old adulthood (mean age: 65.8 years; range: 47—82 years) from the German cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC) study in Potsdam. Hair cortisol concentrations (HCC) were determined from the first scalp-near 3 cm hair segment and several sociodemographic, lifestyle, anthropometric, disease-related, and psychological parameters were assessed. In simple linear regressions, HCC were found to increase with participants’ age and to be higher in men compared to women. HCC also showed positive associations with waist-to-hip ratio, waist circumference, smoking, prevalent type 2 diabetes mellitus, mental health, daytime sleeping, and being unemployed or retired–—as well as a negative association with diastolic blood pressure. After full mutual adjustment, only age and smoking remained independent predictors of HCC. The association between prevalent type 2 diabetes mellitus and HCC was attenuated but still persisted independently in women. Similar, a positive relationship between HCC and alcohol consumption was found in women. The current results confirm previous evidence of positive associations of HCC with age, sex, alcohol consumption, and type 2 diabetes mellitus and add new knowledge on factors–—such as smoking–—that may contribute to elevated cortisol levels in people at older ages. # 2013 Elsevier Ltd. All rights reserved.
* Corresponding author at: German Institute of Human Nutrition (DIfE) Potsdam-Rehbruecke, Department of Epidemiology, ArthurScheunert-Allee 114-116, 14558 Nuthetal, Germany. Tel.: +49 033 200/88 2720; fax: +49 033 200/88 2721. E-mail address: [email protected] (S. Feller).
1. Introduction Long-term changes to the secretion of the glucocorticoid hormone cortisol, such as under conditions of chronic stress,
0306-4530/$ — see front matter # 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.psyneuen.2013.10.007
Predictors of hair cortisol in older adults are well-known to be associated with a range of diseases, including cardiometabolic and autoimmune diseases, as well as mental disorders (Chrousos, 2009). This particularly applies to older people who are at increased risk for these conditions and have often been found to exhibit chronically elevated cortisol levels (e.g. Laughlin and Barrett-Connor, 2000; Larsson et al., 2009). Previous research has identified a number of potential determinants of age-associated hypercortisolism in the elderly, including health status (e.g. Adam et al., 2006), physical functioning (Kumari et al., 2010), lifestyle-related factors (e.g. Badrick et al., 2008), central adiposity (Steptoe et al., 2004), as well as a wide range of psychosocial factors, including depressive symptoms (Penninx et al., 2007), loneliness (Hackett et al., 2012), sadness, anger, or lack of control (Adam et al., 2006). Previous endocrine research in middle and old adulthood has mostly relied on cortisol assessments in saliva and blood. While these methods accurately reflect acute changes in cortisol levels, they are not well-suited for capturing longterm cortisol alterations as their results are easily confounded by situational factors or insufficient compliance with sampling times (e.g. Kudielka et al., 2003). As age-associated factors are likely to manifest in persistent long-term changes to cortisol secretion, the use of a more stable assessment method may be advantageous. Here, recent evidence on the measurement of cortisol in human hair is of particular interest. Through continuous incorporation of cortisol into growing hair, hair cortisol concentrations (HCC) are assumed to mainly reflect integrated cortisol secretion over periods of several months (see Russell et al., 2012; Stalder and Kirschbaum, 2012). Considerable evidence now supports the general validity (Kirschbaum et al., 2009; Thomson et al., 2010; Manenschijn et al., 2011a) and test—retest reliability (Stalder et al., 2012b) of this method as well as its robustness to a range of hair-related factors (Dettenborn et al., 2012). To date, only a small number of studies have examined HCC in relation to aging. Evidence for increasing HCC with older age has emerged from a study covering a wide range of ages (Dettenborn et al., 2012) and from a recent investigation in a large sample of working people (Stalder et al., 2013) but was not observed by other research (e.g. Raul et al., 2004; Dettenborn et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2012a). Similarly, previous investigations into the role of participants’ sex on HCC, as an important factor associated with endocrine functioning as well as with health risk, were mostly conducted in younger populations, revealing no sex differences in HCC (e.g. Raul et al., 2004; Thomson et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2013) or suggesting lower HCC in women than in men (Dettenborn et al., 2012; Manenschijn et al., 2013). Interestingly, Dettenborn et al. (2012) found sex differences in HCC only in middleaged adults but not in the elderly, pointing toward a potential role of changing sex steroid levels with aging. By contrast, a recent investigation found lower HCC in women within an older study sample (Manenschijn et al., 2013). Research examining other correlates of HCC was also frequently carried out in younger populations, potentially limiting transferability of these findings to older people. Here, particularly associations of HCC with body fat-related anthropometric measures (Manenschijn et al., 2011a, 2011b; Stalder et al., 2012a, 2013), the metabolic syndrome (Stalder et al., 2013), cardiovascular risk (Manenschijn et al., 2013),
133 or myocardial infarction (Pereg et al., 2011) have been highlighted. Findings have been more inconsistent regarding HCC relationships with psychosocial stress, with most evidence pointing toward elevations of HCC in persons exposed to stressful conditions (Staufenbiel et al., 2013) but showing less consistent associations with self-reported stress (see Stalder and Kirschbaum, 2012). Finally, previous research has not found HCC to be related to smoking status, oral contraceptive use or overall medication intake (Dettenborn et al., 2012). Within the framework of the above evidence, the current study set out to investigate age-related changes of HCC in a large population of middle and old adulthood. To gain insight into determinants of long-term cortisol secretion in older age, relationships of HCC with a range of sociodemographic, lifestyle, disease-related, and psychological parameters were examined.
2. Methods 2.1. Study population, design and procedure The present sample was drawn from a substudy of the German cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC) study, a large multi-center prospective cohort study investigating associations between diet, lifestyle and chronic disease risk (German EPIC cohort: Boeing et al., 1999; EPIC: Riboli et al., 2002). The current substudy was conducted between the years 2010 and 2012 and employed a cross-sectional design. For this study, individuals residing in the wider Potsdam area were randomly selected from the main cohort with attention being paid to maintain an approximately uniform sample spread and sexdistribution across the examined age range. Individuals were invited to visit the study center at the German Institute of Human Nutrition where hair samples were obtained and physical examinations as well as inquiries on diet, physical activity, and psychosocial factors were conducted. These data were obtained from a total of 815 individuals. Of these, participants were excluded if they exhibited hair shorter than 3 cm at the posterior vertex region at the back of the head (n = 80), reported recent use of glucocorticoid-containing treatments (n = 44), refused to provide a hair sample (n = 1) or provided a hair strand of insufficient volume for cortisol analysis (n = 10). Thus, the final sample (prior to statistical exclusion; see below) comprised 680 participants (369 women) aged between 47 and 82 years. The study protocol was approved by the ethics committee of the Medical Association of the State of Brandenburg and written informed consent was obtained from all participants.
2.2. Hair cortisol analysis Hair strands of a diameter of approximately 3 mm were taken as close as possible to the scalp from a posterior vertex position using fine scissors. Cortisol concentrations were determined from the first 3 cm hair segment proximal to the scalp. Based on a hair growth rate of 1 cm/month (Wennig, 2000), this hair segment is assumed to reflect hair growth over the three-month period prior to hair sampling. Washing and steroid extraction followed the protocol
134 described in detail in Stalder et al. (2012b, study II) with 10 mg of whole, nonpulverised hair being incubated in 1800 ml methanol for 18 h at room temperature. Following extraction, cortisol concentrations were determined using a commercially available immunoassay with chemiluminescence detection (CLIA, IBL-Hamburg, Germany). The intraand interassay coefficients of variation of this assay are below 8%.
2.3. Measurement of psychological parameters Perceived chronic stress was assessed using the screening scale of the Trier inventory of chronic stress (TICS-SSCS; Schulz and Schlotz, 1999) with higher values reflecting increased chronic stress load. Physical and mental health were measured by calculating the physical and mental component summary scores according to the SF-12 algorithm (Windsor et al., 2006) with higher values signaling better subjective health status. Dispositional optimism was assessed by use of four questions derived from Giltay et al. (2006), namely ‘‘I still expect much from life,’’ ‘‘I do not look forward to what lies ahead of me in the years to come,’’ ‘‘My days seem to be passing by slowly,’’ and ‘‘I am still full of plans,’’ coded as 0 (fully in agreement), 1 (partially in agreement or do not know), or 2 (not in agreement). After reversing negatively keyed questions, a mean score was calculated with higher scores indicating greater optimism.
2.4. Measurement of socio-demographic, lifestyle, and disease-related variables Information on age, anthropometry, smoking status (nonsmoker, former smoker, and current smoker), physical activity, alcohol consumption (portions per day), occupation (full time, part time, casual vs. retirement, jobless/retraining, or unemployed), pain during the last 4 weeks (yes/no), menopausal status, as well as average night and daytime sleep was collected using questionnaires and physical examinations. Physical activity was defined using the Cambridge Index combining information on activity at work (sedentary to heavy manual) and leisure time (duration of sport and cycling in hours/week) to classify participants into ‘‘inactive,’’ ‘‘moderately inactive,’’ ‘‘moderately active,’’ or ‘‘active’’ (Wareham et al., 2003). Face-to-face interviews were conducted by study physicians as part of which participants were asked about prevalence of hypertension and overall medication intake (topical, oral, inhaled, or systemic drugs) during the last 7 days, which was used to identify glucocorticoid-containing treatments. Information on medically verified prevalent myocardial infarction, stroke, cancer, and type 2 diabetes mellitus at the time of the substudy was available from the data collected during follow-ups of the regular EPIC-Potsdam cohort study up to 2009 (Bergmann et al., 1999). Waist-to-hip-ratio (WHR), waist circumference, and body mass index (BMI) were measured by trained study personnel.
2.5. Statistical analysis Hair cortisol data were not normally distributed and logtransformed values were thus used for inferential analyses.
S. Feller et al. For descriptive purposes and unless otherwise noted, data in figures and tables are presented in original units. HCC outlying by 3 standard deviations (SD) from the mean were excluded (n = 18) after the initial elimination of extreme values >500 pg/mg (n = 8), resulting in a final sample of 654 participants for further analyses. In a first step, HCC associations with relevant sociodemographic, lifestyle, disease-related, and psychological parameters were investigated using simple linear regressions. Categorical predictor variables were dummy-coded for these analyses. In a second step, multiple regression models adjusted for sex and age were computed. Finally, multiple linear regression models in which all variables were entered simultaneously were conducted to investigate the independence of HCC relationships with individual predictors. This analysis included data of both men and women and thus the variables ‘‘hormone therapy’’ and ‘‘postmenopausal status’’ were not added to this model. To test for a potential influence of these variables, an identical regression model was computed in women only. All analyses were performed with SAS statistical software (release 9.2; SAS Institute Inc., Cary, NC). The level of statistical significance was set at p < 0.05 for two-tailed testing.
3. Results Table 1 provides descriptive information of the current study population. The sample comprised 654 participants (54% women) with a mean age of 65.8 years. Mean SD HCC in the whole sample were 35.1 32.8 pg/mg. Table 2 shows the results of simple and multiple linear regression analyses of associations with HCC. In simple regression analyses, HCC were found to increase linearly with participants’ age (b = 0.14, p < 0.001) and lower HCC were observed in women compared to men (b = 0.12, p = 0.002). Mean SD SD HCC were 37.9 34.0 pg/mg in men and 32.7 31.6 pg/ mg in women. Separate analyses within different age groups revealed that sex differences were evident in individuals younger than 60 years (b = 0.23, p = 0.001, n = 201) and between 60 and 70 years (b = 0.24, p = 0.0004, n = 207) but not in those older than 70 years (b = 0.05, p = 0.46, n = 246). Furthermore, HCC were found to be positively related to WHR, waist circumference, smoking, prevalent type 2 diabetes mellitus, mental health scores, daytime sleeping, and being unemployed or retired. An inverse relationship was observed between diastolic blood pressure and HCC. In multiple regression analysis, both age and sex remained significant predictors of HCC after mutual adjustment for each other (Table 2). After combined adjustment for sex and age, only smoking and prevalent type 2 diabetes mellitus persisted to be related to HCC. Associations of HCC with anthropometric measures (BMI, WHR, waist circumference) as well as diastolic blood pressure were attenuated to non-significant trends in these adjusted analyses. When including all variables in a multiple linear regression model, only participants’ age and smoking status remained independent predictors of HCC (Table 2). The mean SD HCC for different age groups were: 70 years 40.5 40.5 pg/mg. Regarding smoking status, mean SD HCC were 34.3 33.3 pg/mg in nonsmokers, 34.3 30.5 pg/mg in former smokers and 43.2 40.3 pg/ mg in current smokers. The adjusted HCC increase according to type 2 diabetes mellitus status ( p = 0.07) was reflected in a mean SD HCC of 46.9 46.8 pg/mg in diabetics compared to 34.1 31.0 pg/mg in non-diabetics. The separate regression model carried out in women (with influences of hormone therapy or postmenopausal status being accounted for) revealed independent relationships of HCC for age, alcohol consumption, and type 2 diabetes mellitus (b = 0.21; p = 0.02, b = 0.14; p = 0.04, and b = 0.14; p = 0.04, respectively). Effect sizes for each of the predictive parameters calculated by Cohen’s f 2 ranged between .01 and .02 in the final multiple linear regressions. The fraction of variance in HCC explained by these multiple linear regression models were R2 = 0.06 (both sexes) and R2 = 0.13 (in women). Figs. 1—3 illustrate relationships of HCC with parameters that emerged as predictors in the final models.
4. Discussion The current study investigated patterns of long-term cortisol secretion, as assessed in hair, in a larger middle and old adulthood sample. Within this population, our results show a linear increase in HCC with participants’ age and this relationship was maintained regardless of the influence of other age-related variables on HCC being accounted for. The current results also identify further relevant correlates of longterm cortisol levels within an older sample, particularly showing lower HCC in women than in men, higher HCC in current smokers as well as in people with type 2 diabetes mellitus or increased alcohol consumption (in women). The current findings add to previous evidence of hypercortisolism in old age derived from serum and salivary cortisol research (e.g. Laughlin and Barrett-Connor, 2000; Larsson et al., 2009). With regard to research on hair cortisol, our results are at variance with initial studies not showing a link between HCC and age (Raul et al., 2004; Dettenborn et al., 2010; Manenschijn et al., 2011a; Stalder et al., 2012a). These studies, however, mostly drew on smaller samples of
136
S. Feller et al.
Table 2 Standardized single linear regression coefficients, age- and sex-adjusted linear regression coefficients, as well as multiple linear regression coefficients of study population characteristics and HCC (n = 654) within the EPIC-Potsdam substudy. Characteristics
Standardized simple linear regression coefficients
p-Value
Standardized multiple linear regression coefficients (adjusted for sex and age)a
p-Value
Standardized multiple linear regression coefficientsb (mutual adjustment)
p-Value
Age (years) Women vs. men Body mass index Waist-to-hip-ratio Waist circumference Physical activity Inactive, moderately inactive Moderately active, active Smoking Nonsmoker Former smoker Current smoker Alcohol consumption (portion/day) Occupation Full time, part time, casual Retirement, jobless/retraining, unemployed Prevalent myocardial infarction (yes vs. no) Prevalent stroke (yes vs. no) Prevalent cancer (yes vs. no) Prevalent type 2 diabetes mellitus (yes vs. no) Prevalent hypertension (yes vs. no) Diastolic blood pressure (mmHg) Systolic blood pressure (mmHg) Dispositional optimism score Perceived chronic stress score (TICS-SSCS) Physical health score (SF-12) Mental health score (SF-12) Hormone therapy (women, yes vs. no) Postmenopausal (women, yes vs. no) Pain (yes vs. no) Daytime sleep (yes vs. no) Night’s sleep (h)
0.14 0.12 0.07 0.15 0.13
