Abstract
Introduction
Allostatic load (AL) provides a model for understanding how dysregulation across multiple biological systems, such as cardiovascular, immune, and metabolic systems, may cause health problems and accelerate the aging processes (Seeman, Epel, Gruenewald, Karlamangla, & McEwen, 2010). AL has been established in different ways, for example, based on 10 biological variables (Seeman, Singer, Rowe, Horwitz, & McEwen, 1997) and 24 biological parameters (Gruenewald et al., 2012), but in all studies, AL has been established as a cumulative measure across multiple biological systems to study the relationship between stress and the processes leading to disease, unsuccessful aging, and frailty. According to McEwen and Stellar (1993), AL is the cost of chronic exposure to fluctuating or heightened neural or neuroendocrine response resulting from repeated or chronic environmental challenge. AL reflects the hidden cost of chronic stress to the body over long time periods, which acts as a predisposing factor for the effects of acute, stressful life events (McEwen & Stellar, 1993). From a clinical perspective, AL may provide the basis for a more comprehensive assessment of major risks in the aging process (Seeman et al., 1997) in two ways: (a) It has a focus on the sum of physiological dysregulation across systems and (b) it includes relatively modest degrees of dysregulation as a marker of biological risk (Seeman et al., 2010). Hence, the individual effect of a particular physiological system may not constitute a significant health risk, whereas cumulated dysregulation of multiple systems may have a significant impact on health. Also, AL may improve our understanding on inequalities in social classes “get under the skin” and affect health and aging over a life course (Seeman et al., 2010).
The associations between socioeconomic status (SES) and poor health (Adler et al., 1994; Szanton, Gill, & Allen, 2005), cardiovascular diseases (Brunner et al., 1997), mental health (Steptoe et al., 2003), and all-cause mortality (e.g., Lynch et al., 1994; Turrell, Lynch, Leite, Raghunathan, & Kaplan, 2007) are well established, even if the mechanisms underlying these associations are poorly understood. Seeman et al. (2010) have recently concluded that patterns of physiological activity may contribute significantly to our understanding of how social conditions affect health and aging. Recent studies have shown that SES over the life course is significantly related to AL in both men and women in mid-adulthood in the United States (Gruenewald et al., 2012) and Sweden (Gustafsson, Janlert, Theorell, Westerlund, & Hammarstrom, 2011). Among older people, SES has been associated to AL in the United States and Taiwan (Weinstein, Goldman, Hedley, Yu-Hsuan, & Seeman, 2003).
Dysregulation within individual systems in relation to SES has also been shown. Hawkley, Lavelle, Berntson, and Cacioppo (2011) found a significant social gradient in cardiovascular and obesity measures of AL, whereas factors related to sympathetic adrenal medulla and lipids exhibited a small or nonsignificant association with SES (Hawkley et al., 2011; Pickering, 1999). Others, however, have found a social gradient in circulating levels of C-reactive protein, fibrinogen, and other indicators of the inflammatory burden (e.g., Brunner et al., 1997; Gruenewald, Seeman, Karlamangla, & Sarkisian, 2009; Hemingway et al., 2003). In addition, in countries with strong social and health welfare programs, the range in material wealth may not be as large as in countries without strong welfare programs. The relationship between social class and AL has not been studied in late midlife populations close to entering the stages of old age in Danish populations.
The aim of the present study is to describe the prevalence of AL among Danish men and women in late midlife and address if AL differs across social class. The two main questions are, (a) “Do we see a social gradient in AL in late midlife, to provide further evidence for AL as a potential marker of early aging?” and (b) “Do we see a social gradient in a population living in a country with a strong social welfare system?” Assessment of AL is based on 14 biological parameters that were available in the study, which together represent dysregulation in metabolic and immunological systems.
Materials and Method
The Copenhagen Aging and Midlife Biobank (CAMB) study was established in 2009-2011 based on three different cohorts: the Metropolit cohort (MP), the Danish Longitudinal Study on Work, Unemployment and Health (DALWUH), and the Copenhagen Perinatal Study (CPC). Eligible cohort members were 17,937 persons living in the Eastern parts of Denmark (7,750 from MP, 4,906 from DALWUH, 5,282 from CPC). In total, 7,191 (40%) of the invited persons answered the postal questionnaire and 5,575 (31%) participated at the test center. Participants were examined at the National Research Centre for the Working Environment from 2009 to 2011. They answered a questionnaire including items on health, occupational social class, work history, participated in an extensive health examination (height, weight, body fat, blood pressure, physical tests), and 5,457 donated blood samples for biological testing (for more details, see Avlund et al., 2014). At enrollment, all participants provided informed consent. The study protocol was approved by the local ethics (No: H-A-2008-126) committee and the Danish Data Protection Agency (No: 2008-41-2938).
Occupational Social Class
We used occupation-based indicators of socioeconomic position to classify social class as they reflect a relationship between material resources and health as well as social standing and may as such relate health outcomes to certain privileges and better opportunities in life (Christensen et al., 2014). Social class was classified by occupation and coded into Social Classes I to VI and two supplementary groups, according to the standards of the Danish occupational social class classification and described by Christensen et al., 2014. According to this classification, Social Classes I to V encompass economically active individuals ranging from professional occupation in Social Class I to unskilled occupation in Social Class V. Eighty-one participants did not respond to the item of occupation and were excluded from the analyses. Social Class VI represents people on transfer income, including sickness benefits and disability pension. For the present study, we included economically active individuals (Social Classes I-V) and individuals on transfer income (Social Class VI). Eight individuals who were economically active but without sufficient job information (Social Class VII) and a total of 29 students and housewives (Social Class VIII) were also excluded from further analysis. In total, 5,420 participants (3,724 men and 1,696 women) were included in the final analyses.
Vocational Training and Education
Postschool vocational training and education were divided into three categories: (a) university or extended education (>4 years; e.g., doctor, economist, upper secondary school teacher, master of engineering); (b) college or medium level of additional education (3-4 years), such as less than three years theoretical education (e.g., market economist, mechanical engineer), 3 to 4 years theoretical education (e.g., primary school teacher, journalist, bachelor of engineering, bachelor’s degree), and other educations; (c) apprenticeship based on a shorter-duration education or little/no postsecondary education, such as no education, semiskilled worker and skilled worker or similar level (e.g., carpenter, smith, clerical training, hairdresser, nursing assistant, technical assistant). The reduction in categories was for presentation purposes and to obtain an ordinal measure of vocational training and education (Mortensen et al., 2014).
Physiological Measures
Nonfasting blood samples were collected by trained laboratory technicians, stored at −20°C, and analyzed within a maximum of 2 years for high density lipoprotein (HDL), low density lipoprotein (LDL), and total cholesterol, triglycerides, and HbA1c. Ethylenediaminetetraacetic acid (EDTA) plasma was stored at −80°C and analyzed within a maximum of 2 years for high sensitivity C-reactive protein (hsCRP), tumor necrosis factor α (TNF-α), and interleukine 6 (IL-6). Glucose was measured in EDTA blood at the day of the blood donation with HemoCue® Glucose 201. The test system was tested every morning with GlucoTrol-NG Level 2, HemoCue katnr. DK139719. A high performance liquid chromatography (HPLC) method was used for determination of HbA1c. The HPLC consisted of a Waters 625 LC system together with a Waters photo-diode-array detector model 996 and a WISP 717 auto sampler for automatic injection of the samples. Millennium chromatography software was used for calculation of concentrations (Waters Associates Inc., Milford, United States). A cation exchange column Mono S HR 5/5 from Pharmacia Biotech AB, Uppsala, Sweden, was used to separate HbA1c from other components in the samples. The method for HbA1c has been evaluated by interlaboratory comparison based on 17 patient samples and found to be linear in the range from 4.13% to 14.3% of total hemoglobin and without systematic bias (Garde, Hansen, Skovgaard, & Christensen, 2000). Lyphochek Diabetes Control (Calibrator) from BioRad (Anaheim, CA, USA) for HbA1c was used for following the long-term stability of the method. HDL, LDL, and total cholesterol analyses were carried out by use of a COBAS Mira Plus. The determination of HDL, LDL, and total cholesterol, and triglycerides were ABX Pentra assays from Triolab (Sollentuna, Sweden). The analytical methods for measuring total cholesterol in serum have been evaluated by a method evaluation function design (Christensen, Anglov, Christensen, Olsen, & Poulsen, 1993) to estimate the random and systematic effects. This was based on a linear least squares regression analysis of the measured concentration versus the conventional true concentration of a series of method evaluation samples containing cholesterol. The between assay variation was estimated to be 2.7% at 5.3 mmol/l cholesterol/l (Hansen, Garde, & Eller, 2007). Commercially available control samples for HDL, LDL, and total cholesterol and triglycerides were analyzed together with samples to show equivalence between different runs. Westgard control charts were used to document that the analytical methods remained in analytical and statistical control, that is, the precision and the trueness of the analytical methods remained stable (Westgard, Barry, Hunt, & Groth, 1981). An immune-turbidimetric analysis by use of Roche/Hitachi automatic instrument COBAS® (Roche/Hitachi, Mannheim Germany) was used to measure hsCRP. The assays used were Tina quant, Roche Diagnostics GmbH, Mannheim, Germany. Approx. 50 randomly selected samples were reanalyzed to document the stability of the analytical method. IL-6 and TNF-α were analyzed by electro-chemiluminescence using multiplex antibodies on Sector 2400 Imager from Meso Scale Discovery (Gaithersburg, USA). All samples were run as duplicates. The inter-assay variation had to be <20% to accept an analysis.
Systolic and diastolic blood pressure (SBP and DPB) was measured at least two times on each arm after a 10-min rest. Body weight was measured in kg with light clothes and without socks and shoes and the estimated weight of clothes (1.5 kg) was subsequently subtracted. Percentage body fat was measured using a four-compartment Tanita® MC 180 body composition analyzer. Height in meters was measured without shoes, heels together and buttock, shoulders and head touching the vertical surface on a standardized wall-mounted height board. Body mass index (BMI) was calculated as kg/m2. Waist hip ratio (WHR) was measured with a SECA 201 measuring tape horizontally on the outside of the T-shirt. The participant stood with feet close together, arms at the side, and body weight evenly distributed. The hip was measured at the femoral node (trochanter major) and the waist was measured at the navel, where the participant standing with feet close together. The participant was asked to relax, and the measurements were taken at the end of a normal expiration.
Allostatic Load
We established AL on the basis of 14 parameters grouped as metabolic parameters (DBP, SBP, BMI, WHR, % body fat, glucose, HbA1c, HDL, LDL, and total cholesterol, triglycerides) and immunological parameters (hsCRP, IL-6, and TNF-α). AL was originally based on 10 physiological parameters and developed for use as a risk factor of cardiovascular disease among 70 to 79 years old men and women (Seeman et al., 1997) but has been further developed to include more biological variables (Gruenewald et al., 2012).
Covariates
We stratified the analyses by gender. A number of biological variables are age dependent (Hansen et al., 2007). AL is remarkably constant in older ages but increases in the years from 20 to 60 (Crimmins, Johnston, Hayward, & Seeman, 2003). Consequently, we used age as a covariate. Gender and age were assessed at enrollment from personal identification numbers given to all Danish citizens at birth. Biological variables may exhibit a distinct diurnal variation. We therefore recorded the time of the blood draw. In addition, the following common determinants of allostatic load (Pickering, 1999) were included as covariates: smoking, leisure-time physical activity, and prevalent chronic conditions. We asked if participants were smokers or nonsmokers (former and never smokers) and had eaten or drunken the last 2 hr (Yes/No). Leisure-time physical activity was assessed with the item: How many hours a week are you physically active? The respondents were asked to include “sport, physical training, house and garden work, walks and bicycle rides including between home and work.” Response categories were none, approx 0.5 hr, 1 to 2 hr, 3 to 4 hr, 5 to 6 hr, and more than 7 hr per week. Leisure-time physical activity was categorized as 1 to 2 hr or less versus 3 to 4 hr and more. We also asked if participants had any of the following diseases within the last year: asthma, diabetes, hypertension, myocardial infarction or angina pectoris, stroke, chronic bronchitis, emphysema, osteoarthritis, rheumatoid arthritis, osteoporosis, peptic ulcer, cancer or leukemia, chronic anxiety or depression, other mental disorders or chronic back pain. Response categories were “yes” or “no.”
Statistical Analysis
We used generalized linear regression models (GLMs) to test for gender differences in smoking, and biological variables (DBP, SBP, BMI, WHR, Body fat, glucose, HbA1c, total-, HDL-, LDL-cholesterol, triglycerides, hsCRP, IL-6, and TNF-α). A GLM was also used to test for gender differences in leisure-time physical activity by social class. Each biological measure was divided into quartiles by calculating the 25% percentiles by gender. The poorest quartile of each measure was given the score 1, and AL was calculated as the sum of these scores, range 0 to 14, that is, higher score indicates higher AL. A sensitivity analysis was conducted by using sixtiles of biological variables to calculate AL. For cutoff values, see Table 2. Due to non-normal distribution (skewed) and increasing variances (proportional to level of measurements), tests for differences in BMI, blood glucose, HbA1c, HDL cholesterol, triglycerides, hsCRP, IL-6, and TNF-α were analyzed on logarithmic transformed data. To analyze associations between social class (categorical) and AL (continuous), a GLM was used. The GLM analysis was stratified by gender. We adjusted for age and time of sampling (continuous variables), food and beverages within the last 2 hr (categorical) and chronic conditions. In the final step, we included smoking and physical activity. SPSS statistical software (version 19.0) was used. p value < .05 was considered statistically significant.
Results
The mean values of age, percent smokers, and leisure-time physical activity among men and women by social class are presented in Table 1. Women were approximately 3 years younger than men. There was a social gradient in smoking with more smokers in lower social classes. In Social Class I, women were less physical inactive compared with women in Social Classes II to IV.
Age and Lifestyle Information by Occupational Social Class for 5,451 Danish Men and Women in Late Midlife.
Table 2 presents the biological variables by men and women. We found gender differences in SBP and DBP, BMI, WHR, % body fat, glucose, HbA1c, HDL, LDL, and total cholesterol, and triglycerides but not in hsCRP, IL-6, and TNF-α. The cutoff values for each biological variable are given for men and women (Table 2). The cutoff values are the highest values in the first quartile of HDL cholesterol (e.g., <1.13 mmol/l for men) and the highest values in the third quartile for the remaining biological variables (e.g., >3.59 for LDL cholesterol for men). Total AL scores were 3.45 for both men and women.
Mean Concentration of Biological Variables for 5,420 Danish Men and Women in Late Midlife.
We next tested social class differences in AL for men and women. The results indicated a statistically significant social class gradient in AL among men (p < .019) and women (p < .025), with increasing AL from Social Classes I to VI (Table 3). The p values for trend in social gradient remained statistically significant in both men and women after adjusting for age, time of sampling, food and beverages within the last 2 hr, smoking, chronic conditions, and leisure-time physical activity. We repeated the analyses using vocational training and education as measure of socioeconomic position (see Table 4). We found a social gradient in AL, ranging from 3.20 among the highest educated to 3.53 among the lowest educated (p value = .007) among men, and from 3.00 to 3.48, (p value = .007) among women, respectively.
AL by Occupational Social Class Among Men and Women in Late Midlife.
adjusted for age, time of sampling, food and beverages the last 2 hr, and known diseases.
adjusted for age, time of sampling, food and beverages the last 2 hr, known diseases, smoking, and leisure-time physical activity.
p value for trend
Vocational Social Class Among Men and Women in Late Midlife.
Note. Vocational social class: 1 = university or extended education (>4 years); 2 = college or medium level of additional education (3-4 years); 3 = apprenticeship, for example, no education, semi-skilled worker, and skilled worker.
adjusted for age, time of sampling, food and beverages the last 2 hr, and known diseases.
adjusted for age, time of sampling, food and beverages the last 2 hr, known diseases, smoking, and leisure-time physical activity.
p value for trend.
Discussion
The main findings of our analyses suggested a social gradient in AL among Danish men and women. Empirical research has provided evidence of the cumulative nature of AL across the life course (Crimmins et al., 2003) as well as evidence supporting the hypothesis that the added stress exposure and resource constraints associated with lower SES are associated with faster accumulation of AL (i.e., greater age-specific burdens of AL; Seeman et al., 2010). Data from the MacArthur Study provided an initial evidence to support this hypothesis, documenting not only the hypothesized SES differences but also the contribution of these SES-associated differences in AL to be a pathway between SES gradients in mortality (Seeman et al., 2004). Our findings support the hypothesis that AL may be one important pathway through which greater life-course SES adversity leads to larger risk for morbidity and mortality (Gruenewald et al., 2012).
Socioeconomic position has traditionally been assessed by three indicators: education, occupation, and income. The three indicators are different both conceptually and in their potential causal processes in relation to health (Galobardes, Lynch, & Smith, 2007; Geyer, Hemström, Peter, & Vågerö, 2006). Measurements of wealth (such as income and housing tenure) relate to the material explanations of health inequalities in a population. Education may indicate resources within the individual, such as knowledge and skills, while measurements of occupational social class point to pathways linked to the position in the social structure and may be related more to status, competencies, and control (Galobardes et al., 2007). There is some suggestion that occupational social class is a better predictor of socioeconomic differentials in overall and noncardiovascular mortality compared to educational social class (Smith et al., 1998). A strong argument for occupation-based indicators of socioeconomic position is that they reflect a relationship between material resources and health as well as social standing and may as such relate health outcomes to certain privileges and better opportunities in life (Galobardes et al., 2007). We also found a social gradient based on vocational training and education. Vocational training and education showed a smaller range in AL but more significant compared with the social gradient in AL based on occupational status. In countries with strong social and health welfare programs, the range in material wealth may not be as large as in countries without strong welfare programs. However, our results indicate that social gradients in AL are also evident in middle-aged adults in countries with a more extensive social welfare system such as Denmark. AL among our middle-aged men and women were similar to those observed in the United States and other areas with less-extensive social welfare programs. However, AL is not directly comparable due to the different number of biological variables used to establish AL (Gruenewald et al., 2012; Gustafsson et al., 2011). Seeman et al. (1997) established a measure of AL based on 10 biological variables, mainly related to the metabolic systems, and found an increased risk of cardiovascular disease when the AL score was ≥3. We included 14 biological parameters measured in CAMB (11 related to metabolic systems and 3 related to the immune system) of which 6 biological variables (SBP, DVP, WHR, HDL, and total cholesterol and HbA1c) were used in the study by Seeman et al. (1997). It is possible that using more biological variables provide a more refined estimate of AL. Recent studies extended the estimation of AL to cover 24 biological parameters from the cardiovascular, metabolic, endocrine, and inflammatory systems (Gruenewald et al., 2012). We included biological variables of the cardiovascular, metabolic, and inflammatory systems in our study. Further, since inflammation may be a common cause of multiple age-related diseases or a pathway by which disease leads to disability (Singh & Newman, 2011), we also included TNF-α and IL-6.
The cutoff values for several of the biological variables in the AL measure differed from those in other studies, that is, our risk cutoff values for DBP, SBP, and WHR among men and DBP among women were higher than those observed in, for example, the MIDUS study (Gruenewald et al., 2012) with a mean age of 58 years (range 35-85), and the cut points for HDL and LDL cholesterol, BMI, WHR, HbA1c, hsCRP, and IL-6 were lower in the sample compared to the MIDUS sample. The MIDUS study provided both high-risk cut points for the population studied and clinical cut points of moderate disease outcomes. The high-risk cut points were higher than the clinical cut points with respect to SBT, BMI, and total cholesterol, and lower for DBP, WHR, triglycerides, and LDL cholesterol (Gruenewald et al., 2012). The estimated cut points in our study among a Danish population in midlife showed lower values for BMI and SBP but higher values for DBP. In our study, participants were not asked to be fasting but were asked if they had anything to eat or drink in the past 2 hr. We adjusted for these variables in our statistical models. However, it is a common problem in field studies of the general population to ask voluntary participants to be fasting for 12 hr or more for voluntary health examinations. In the present study, we cannot exclude the possibility that nonfasting blood samples may have affected our observed social gradient in AL.
Another explanation for differences in cutoff points could be due to differences in age of the participants or the population studied. AL has previously been shown to increase sharply in the years from 20 to 60 (Crimmins et al., 2003). Hence, some of these discrepancies may be due to the relatively narrow age range (age 48-63) in the CAMB study population, whereas the MIDUS study has an age range from 35 to 85 years (Gruenewald et al., 2012). We did not exclude participants with known diseases, which may be reflected in the higher cutoff values for SBP and DBP. However, adjusting for known diseases did not change the results.
We found that the social gradients in single biological parameters were not as significant strong as for AL. This may point in the direction of AL to reflect not only “wear and tear” of life experiences but also of genetic load, individual habits reflecting items such as diet, exercise, and substance abuse, and developmental experiences that set lifelong patterns of behavior and physiological reactivity (McEwen & Seeman, 1999).
This is a preliminary analysis with a number of limitations. The low response rate in combination with a nonrandom allocation of participants to the health examination questions the generalization of the results for the study. However, even though preliminary analyses showed some social differences between participants and nonparticipants, there were no indications of health differences between the groups (Avlund et al., 2014). We did not collect fasting blood samples, and the time since last food intake may affect the concentration of some of the selected biological variables (e.g., glucose). However, the effect of time since last food intake is not likely to be systematically distributed across the different social classes and is, therefore, not likely to bias the analysis. Even so we controlled the analysis for time of sampling. Biological parameters included in our measure of AL did not cover the endocrine system, which would have been desirable. However, the metabolic and immunological systems were well covered. Women of our study were 3 years younger than the males, but since AL is based on quartiles within gender, this is not likely to affect AL. However, in line with previous studies (Seeman et al., 2010), AL seems to be associated with social class among both men and women in our study. In the statistical analyses of social gradient in AL, we adjusted for smoking, leisure-time physical activity, and chronic conditions. More smoking, less leisure-time physical activity, and chronic conditions are associated with increased AL, and poor social conditions may result in less healthy behavior. Therefore, it is debatable whether it is correct to adjust for smoking, leisure-time physical activity, and chronic diseases as behavioral factors may have perceived adaptive effects in the short run but damaging effects in the long run on AL (McEwen, 2000). The range in social gradient in AL was slightly reduced when adjusting for life smoking, leisure-time physical activity, and chronic conditions.
In conclusion, we found a social gradient in AL in late midlife among men and women living in Denmark. AL may thus be a potential biomarker for early aging in countries with a strong social welfare system. It is important for intervention studies to be aware of this type of biological vulnerability already in late midlife, as intervention on these markers may be one way of postponing future functional decline and promote healthy aging at older ages.
Footnotes
Acknowledgements
Authors thank the staff at the Department of Public Health and the National Research Center for the Working Environment who undertook the data collection. Further thanks to Merete Osler, Erik Lykke Mortensen, Nils-Erik Fiehn, and Poul Holm-Pedersen who initiated and established CAMB from 2009 to 2011 together with Helle Bruunsgaard, Åse Marie Hansen, Rikke Lund, and Kirsten Avlund. The authors acknowledge the crucial role of the initiators and steering groups of the Metropolit Cohort, the Copenhagen Perinatal Cohort, and the Danish Longitudinal Study on Work Unemployment and Health.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The CAMB study has been supported by a generous grant from the VELUX FOUNDATION.