Economic hardship over twenty-two consecutive years of adult life and markers of early ageing: physical capability, cognitive function and inflammation


This study assesses the associations between annual measures of economic hardship (EH) across 22 years of adulthood and objective measures of early ageing in a Danish late-middle-aged population (N = 5575). EH (years < 60% of the National median equivalized household disposable income) was experienced by 18% during 1987–2008. Four or more years in EH (reference = null years in EH) was related to poorer physical capability (chair rise: − 1.49 counts/30 s [95% confidence interval (CI) − 2.36, − 0.61], hand grip strength: − 1.22 kg [95% CI − 2.38, − 0.07], jump height: − 1.67 cm [95% CI − 2.44, − 0.91] and balance: 18% [95% CI 9, 28]), poorer cognitive function (Intelligenz-Struktur-Test: − 1.50 points [95% CI − 2.89, − 0.12]) and higher inflammatory levels (C-reactive protein: 22% [95% CI 4, 44], and Interleukin-6: 23% [95% CI 10, 39]). Comparing four EH trajectories, people with a high versus low probability of EH over time had poorer physical capability (chair rise: − 1.70 counts/30 s [95% CI − 3.38, − 0.01], grip: − 4.33 kg [95% CI − 6.50, − 2.16], jump: − 1.68 cm [95% CI − 3.12, − 0.25] and balance: 31% [95% CI 12, 52]). No associations were observed with tumour necrosis factor-α. Results were adjusted for sex, age, long-term parental unemployment/financial problems, education, baseline income and cohort. This study suggested EH for four or more years to be associated with poorer physical capability, cognitive function and increased inflammatory levels in midlife. High probability of EH across adulthood was similarly related to poorer physical capability and CRP, but not cognitive function and the remaining inflammatory markers. In conclusion, preventive initiatives focusing on reducing the burden of sustained economic hardship may lead to increased healthy ageing.


Recent declines in birth rates and the steep expansion of the average life span have resulted in population ageing (Harper 2014). Chronic diseases that contribute to disability, lower quality of life, and health care costs disproportionately affect older adults, resulting in widespread focus on promoting healthy ageing. Two critical components of healthy ageing include maintaining physical capability (Cooper et al. 2014a) and cognitive function (Richards and Deary 2014), which are imperative to performing activities of daily living and remaining socially active and independent of help for as long as possible. Additionally, physical capability and cognitive function have been found to be related to morbidity and longevity in later life (Cooper et al. 2010, 2011, 2014b; Perna et al. 2015). Furthermore, a substantial body of research suggests that a number of apparently unrelated diseases associated with ageing (e.g. cancer, cardiovascular disease, type II diabetes, sarcopenia, dementia and depression) originate from inflammatory mechanisms making inflammation in midlife a potentially powerful marker of early ageing processes (Cevenini et al. 2013).

There are well-documented social gradients in markers of early ageing such as functional limitations (Avlund et al. 2004; Strand et al. 2011; Hurst et al. 2013), cognitive function (Hurst et al. 2013) and inflammatory concentrations (Nazmi and Victora 2007; Johnson et al. 2013). In a life course perspective, it is widely acknowledged that accumulation of social adversity across the life course results in weathering that can lead to accelerated ageing (Kubzansky et al. 2014). People experiencing social adversity are often identified by means of socio-economic position operationalized as education, occupation or income (Oakes and Andrade 2017). While educational attainment tends to be relatively stable over the duration of life, income is much more volatile in nature (Krieger et al. 1997; Galobardes et al. 2006; Western et al. 2012). Consequently, income measures have the potential to capture adult socio-economic position in greater detail and identify people in social adversity more precisely. Despite this, socio-economic position is usually treated as a static factor, with a single point in time measurement that fails to capture its dynamic nature. The result is loss of information and potential exposure misclassification causing reduced ability to detect associations between socio-economic position and markers of ageing.

Relatively low income is commonly conceptualized as representing economic hardship (EH). EH can be defined according to an official poverty level in some countries, e.g. as income below 200% of federal poverty level (Lynch et al. 1997). Other prevailing alternatives applicable to all societies are incomes below 50–60% of the national median (Eurostat - Statistics Explained 2014; OECD 2018). EH may lead to accelerated ageing (Lynch et al. 1997; Kahn and Pearlin 2006; Simons et al. 2016), and the chronicity of EH could be the key to whether subsequent adverse health effects develop. Accumulated or sustained EH has been found to be related to poorer physical capability (Lynch et al. 1997; Kahn and Pearlin 2006; Ahnquist et al. 2007) and poorer cognitive function (Lynch et al. 1997; Al Hazzouri et al. 2017). The evidence for associations with single episodes of EH and health outcomes in later life is less consistent (Kahn and Pearlin 2006). Evidence from a trajectory-based study similarly demonstrates that persistent EH is associated with increased probability of disease in midlife (Willson and Shuey 2016). This study further suggests that timing of EH is of importance to the association. People that move out of EH early appear to have a reduced probability of disease in midlife compared to those in long-term EH. Interestingly, people moving out of EH later in adult life did not appear to experience any reduced probability of disease (Willson and Shuey 2016). Nonetheless, these previous studies on EH are limited by determining EH based on few and infrequent income measures that are self-reported and lacking information on objectively measured health outcomes as well as important covariates such as childhood economic environment. Additionally, focusing on either accumulation of EH or trajectories of EH leaves unanswered questions. The two analytical approaches hold interpretational differences. The categorical analysis merely provides information on the accumulation (dose–response) relationship between EH and early ageing. This is not a particular life course approach (Ben-Shlomo et al. 2016), but it provides insight into the associated probability of early ageing at different exposure levels. In contrast, trajectories distinguish according to timing of EH across adult life. By identifying typical trajectories and subsequently estimating the associated probability of early ageing, it is possible to identify potentially susceptible groups at increased risk of experiencing early ageing. Each trajectory will imply a certain level of exposure to EH, but rather than focusing on the specific amount, trajectories identify groups with similar exposure patterns across adulthood. Thus, the two approaches are not mutually exclusive, but rather complement each other.

We hypothesize that accumulation of EH across life is associated with lower physical capability, poorer cognitive function and increased levels of inflammation in late-middle age. Using a unique dataset with objective register-based annual measures of income together with objectively measured markers of early ageing, we aim to address how the accumulation and trajectories of EH years lived below 60% of the median equivalized household disposable income over a 22-year period are associated with physical capability, cognitive function and inflammatory profiles in a Danish late-middle-aged population.

Materials and methods

This study is based on a linkage of longitudinal register data from Statistics Denmark covering the period 1987–2008 and cross-sectional survey data from the Copenhagen Ageing and Midlife Biobank (CAMB) collected in 2009–2011. Participants in CAMB come from three existing Danish cohorts: the Metropolit Cohort (MC), the Copenhagen Perinatal Cohort (CPC) and the Danish Longitudinal Study on Work, Unemployment and Health (DALWUH). An overview of the data sources is presented in Fig. 1. The MC and the CPC are birth cohorts. The MC includes men born in 1953 in greater Copenhagen, and the CPC includes children born in 1959–1961 at the Copenhagen University Hospital. DALWUH consists of participants born in 1949 or 1959 randomly sampled from the general Danish population in 1999. Participants from the three cohorts who were alive, living in the Eastern part of Denmark and who had not formally resigned from the studies were invited to the CAMB follow-up in 2009–2011 (78%). From the eligible population (N = 17,937), 31% showed up at the study clinic for physical and cognitive tests and gave a blood sample, resulting in a final sample for this study of 5575 participants. Due to missing data on specific outcome variables, the analytical sample size varied between 4470 and 5198 depending on the outcome of interest. More information on the missing data is presented in the results section. Further details of the CAMB are described elsewhere (Avlund et al. 2014; Lund et al. 2016).

Fig. 1

Graphic overview of the data available for this study

Economic hardship

Yearly data on equivalized household disposable income from 1987 to 2008 were attained from the Danish registers on income (Baadsgaard and Quitzau 2011). A dummy variable for EH was calculated for each year with people classified as being in EH if their equivalized household disposable income was less than 60% of the median equivalized household disposable income for the entire Danish population that year. Identification of people in EH using the 60% median income as a threshold corresponds to the European Union’s definition of people at-risk-of-poverty (Eurostat - Statistics Explained 2014). We defined two exposures: (1) a categorical measure of accumulative exposure to EH coded as the total number of years being in EH between 1987 and 2008 with five categories: 0, 1, 2, 3 and 4 or more times in EH, and (2) trajectories of EH as low probability of EH, declining probability of EH, rising probability of EH and high probability of EH (further details in ‘Statistics’).

We excluded 19 respondents who had missing income information for more than 10 years and 184 respondents who at one or several time points had a negative income. Negative incomes may be due to deficits in companies and are not necessarily a sign of EH. Consequently, years with negative income were not counted as EH.

Physical capability

Four validated objective measures of physical capability obtained at the CAMB follow-up were assessed in the main analysis: (1) Chair rise performance (counts in 30 s) was performed using a 45-cm high chair with automatic recording and without use of the arms (Ritchie et al. 2005). (2) Hand grip strength (kg) was measured three–five times using a Jamar dynamometer model G100 with automatic recording (Fairfax et al. 1995). (3) Jump height (cm) was measured a maximum of five times as a two-legged countermovement jump on an AMTI power plate with the hands placed on the hips (Holsgaard Larsen et al. 2007). (4) Balance (cm2) was measured three times as postural sway area on the participant’s preferred leg during 30 s using an AMTI power plate, with lower values representing better balance (Holsgaard Larsen et al. 2007). This variable was skewed and therefore transformed using the natural logarithm to approach normality. The results were back-transformed before presentation and can be interpreted as the relative change in the outcome (UCLA Institute for Digital Research and Education 2016). For tests performed numerous times, the most favourable recording was included. Chair rise performance, hand grip strength and balance are frequently used indicators of physical capability, while the less commonly used jump height is included on the assumption that this strenuous measure is better at discriminating physical capability in healthy late-middle-aged adults.

Cognitive function

Three selected subtests from the Intelligenz-Struktur-Test (I-S-T 2000 R) (Amthauer et al. 2001) were used at the CAMB follow-up to assess cognitive function. The psychometric characteristics of this version of the I-S-T 2000 R are presented in details by Mortensen et al. (2014). Subtests on sentence completion, verbal analogies and number series were included and each contained 20 tasks with one point given for a correct answer (range 0–20). Internal consistency analyses showed that one item in the sentence completion subtest had very low correlations with the other subtest items and with the total score of the remaining items (Mortensen et al. 2014). Therefore, this item was excluded allowing a maximum possible score on the sentence completion subtest of 19 and a total score with a 0–59 range. Only the total score was included in the present analyses.


Circulating C-reactive protein (CRP), interleukin-6 (IL-6) and tumour necrosis factor-α (TNF-α) were included in the study to assess systemic inflammation. They were derived from non-fasting blood samples collected at the CAMB fractionated in plasma, serum and DNA and individually analysed as continuous variables. Analyses were conducted by the Clinic Biochemical Laboratory (high-sensitivity CRP) and the Centre of Inflammation and Metabolism, Rigshospitalet (IL-6, TNF-α). High-sensitivity CRP was analysed in plasma samples by Roche/Hitachi MODULAR P. IL-6 and TNF-α were analysed in EDTA plasma by electro-chemiluminescence multiplex system on a Sector 2400 Imager with commercial kits from Meso Scale Discovery (Gaithersburg, USA) according to the manufactures instructions. The same lot number was used for all analyses. All samples were run as duplicates. The intra-assay variation had to be < 20% to be accepted. The same two internal laboratory controls were included in all runs: Control A was a fasting plasma sample from a healthy young subject, and control B was a plasma sample after endotoxin administration in vivo. Lower limit of detection was the calculated concentration of the signal that was 2.5 standard deviations over the zero calibrator (the blank). Lower limit of detection (LOD) for IL-6 was 0.21 pg/ml and inter-assay coefficient of variation (CV) was 11–21% pg/ml in our laboratory. For TNF-α in house LOD was 0.28 pg/ml, inter-assay CV was 9–13% in our laboratory. Values below the detection limit were substituted by simple imputation. (Percentages below LOD were IL-6 = 0.4% and 0% for TNF-α and CRP.) The markers were transformed using the natural logarithm to approach normality and back-transformed before presentation to present the relative change in the outcome.


Potential confounders included sex, age group (49–53 years and 56–63 years at follow-up), original cohort (MC, CPC, DALWUH), long-term parental unemployment and long-term parental financial problems (from the question: ‘Have you been exposed to any of the life events mentioned below during childhood/youth (before age 20 years)?’, including ‘Longstanding parental unemployment’ and ‘Long-term parental financial problems’), equivalized disposable household income at baseline in 1987, and educational level. Register data were used to identify respondent’s highest obtained education in 2009 and 1986 (Jensen and Rasmussen 2011). Education was coded according to the International Standard Classification of Education (ISCED 2011) and categorized as ‘low’ (primary and lower secondary), ‘medium’ (upper- and post-secondary) or ‘high’ (first and second stage tertiary). We used highest attained educational level in 2009 in the main analysis in order not to assign a low level of education to younger people in the study sample who obtained a higher educational level shortly after baseline. We examined the effect of using highest obtained education in 1986 in sensitivity analyses.


Descriptive plots of the timing of EH were constructed to investigate which periods in the life the respondents were most likely to experience EH. The main analyses addressing the associations between EH and continuous markers of early ageing (physical capability, cognitive function and inflammation) were twofold. First, the categorical version of EH was used in linear regression models. Second, group-based trajectories of EH across adult life were identified, tested and finally regressed on each outcome measure by means of the Stata plugin traj (Jones and Nagin 2007, 2013). A binary logistic distribution was employed to estimate the longitudinal mixture models, and trajectories are therefore defined by the probability of EH over time. We tested models with two to five trajectory classes, and evaluated each according to the Bayesian Information Criterion (BIC) (see Online Resource 1) and conceptual meaningfulness. The final model included four trajectories: low probability of EH, decliningprobability of EH, rising probability of EH and high probability of EH. The trajectories of EH from 1987 to 2008 are depicted in Fig. 2. Causal diagrams based on prior theoretical knowledge and literature reviews were used to identify potential confounders, which we adjusted for in the regression analysis (Glymour and Greenland 2008). Statistical significance was evaluated using an alpha level of 0.05.

Fig. 2

Trajectories of economic hardship 1987–2008. Lines show the predicted trajectories, and dots show the observed trajectory. Numbers in parenthesis show the average of group member’s probability of belonging to the group

The robustness of our findings was assessed in sensitivity analyses. First, we conducted a simple quantitative bias analysis by means of the methods described by Lash et al. (2009) to quantify the potential impact on our results caused by anticipated selective dropout in the study sample. Selective dropout was of concern because individuals of low socio-economic position and in poor health were less likely to participate in the CAMB survey (Lund et al. 2016). A previous comparison to National registers has shown that 6.4% of the non-respondents (total N = 10,746) and 3.4% of the respondents (total N = 7191) were unemployed or on transfer income. Further, 75.3% of the non-respondents were active in the labour market compared to 87.6% of the respondents (Lund et al. 2016). Anticipating that unemployment/transfer income can be a proxy of ever experiencing EH and likewise, that employment can be a proxy of never being in EH, the overall proportions ever versus never in EH can be calculated as:

$${\text{Proportion}}\;{\text{of}}\;{\text{full}}\;{\text{population}}\;{\text{ever}}\;{\text{in}}\;{\text{EH:}}\frac{0.064*10,746 + 0.034*7191}{17,937} = 0.052$$
$${\text{Proportion}}\;{\text{of}}\;{\text{full}}\;{\text{population}}\;{\text{never}}\;{\text{in}}\;{\text{EH:}}\frac{0.753*10,746 + 0.876*7191}{17,937} = 0.802$$

No suitable proxies were identified for the early ageing outcomes in the previous comparison with the registers. Thus, it was not possible to estimate overall proportions with lower versus higher score on markers of early ageing in the full population. Instead, crude relative risks were estimated under different scenarios of 0.5–2.5%-points (corresponding to 9.6–48.1%) reduced participation in the group with a score below versus at or above the median for each marker of early ageing (reversing inverted scales to lower scores indicating early ageing). A detailed description of the calculations is found in Online Resource 2.

In the second sensitivity analysis, we estimated models similar to the main analyses adjusting for highest attained educational level in 1986 instead of in 2009. Third, we estimated models looking at the association between EH the last 10 years leading up to the health assessment (1999–2008) and the early ageing outcomes to ascertain whether there were differences in the associations depending on the lag time allowed. Fourth, we re-estimated the models excluding people who in the CAMB survey had answered that they were or had been suffering from heart attack, stroke, chronic bronchitis, cancer, chronic anxiety or depression to examine whether results were sensitive to reverse causation, i.e. ending in EH due to poor health.

Data management was performed using SAS version 9.4 for Windows. Statistical analyses were performed using Stata statistical software, version 14 for Windows.


Table 1 shows descriptive statistics. Of the 5575 participants (68.6% men) in the CAMB, 44.6% were originally members of the MC, 30.8% of the CPC and 24.6% of the DALWUH cohort. The participants were distributed in two age groups at the CAMB follow-up in 2009–2011, 42.6% were 49–53 years and the remaining 56–63 years of age due to the age distributions of the original samples. The majority of the sample had never experienced EH in the period 1987–2008 (82.4%). Among respondents who had experienced EH, the highest proportion had only been in the situation for one year (8.5%), whereas 3.6% had experienced accumulated EH for four or more years. Similarly, the majority of the sample (91.0%) was in the trajectory of low probability of EH, while only 0.9% was in the high probability of EH trajectory. The relatively high percentage of missing values for the EH measure (3.6%) was mainly due to our exclusion of respondents who at some point had negative incomes. Further exclusions due to missing on all outcome measures resulted in 6.6% missing in the EH trajectories. Missing values for the physical tests balance area, chair rise and jump height varied between 4.1 and 14.3% (Table 1). This was due to strict exclusion criteria preventing participants with hypertension, discus prolapse or pain in back/knee/hip/ankles from taking part in the tests. EH was most commonly experienced among respondents when they were in their mid-twenties or early thirties and the likelihood of experiencing EH decreased with increasing age (Fig. 3).

Table 1 Descriptive statistics of the study population, N = 5575
Fig. 3

Economic hardship by age in years, N = 5372

Results from the analyses of accumulated EH are presented in Table 2. Presented coefficients for the log-transformed outcomes balance, CRP, IL-6 and TNF-α are back-transformed and should therefore be interpreted as a relative change. People who experienced EH for four or more years had less favourable scores in all physical capability tests (chair rise: − 1.49 counts/30 s [95% confidence interval (CI) − 2.36, − 0.61], grip strength: − 1.22 kg [95% CI − 2.38, − 0.07], jump height: − 1.67 cm [95% CI − 2.44, − 0.91] and balance: 18% larger area [95% CI 9, 28]) compared to the reference category of people who had never experienced EH. There were no statistically significant differences in the scores on the physical tests between respondents who had been in EH for one, two or three years compared to the respondents who had never experienced EH. Results were similar for the cognitive test score where only respondents who had experienced EH for four or more years had statistically significantly lower scores than the reference group of people who had never been in EH (IST: − 1.50 points [95% CI − 2.89, − 0.12]). Likewise, levels of CRP and IL-6 were higher for the group who had been in EH for four or more years (CRP: 22% [95% CI 4, 44], IL-6: 23% [95% CI 10, 39]). There was no difference in the level of TNF-α depending on time in EH (Table 2).

Table 2 Associations between number of times in economic hardship (EH) 1987–2008 and physical, cognitive and inflammatory outcomes. Linear regression modes with reference category zero times in EH

The trajectory analyses are presented in Table 3. Being in the high probability of EH trajectory was related to poorer physical capability (chair rise: − 1.70 counts/30 s [95% CI − 3.38, − 0.01], grip strength: − 4.33 kg [95% CI − 6.50, − 2.16], jump height: − 1.68 cm [95% CI − 3.12, − 0.25], balance: 31% larger area [95% CI 12, 52]) and higher CRP level (37% [95% CI 0, 88]) compared to the reference trajectory of low probability of EH. Similar but slightly weaker associations were observed for the trajectory of rising probability of EH, but for grip strength the association became statistically insignificant. The declining probability of EH trajectory was associated with lower jump height. There was no difference by EH trajectory in the cognitive outcome, IL-6 or TNF-α (Table 3).

Table 3 Associations between trajectories of economic hardship (EH) 1987–2008 and physical, cognitive and inflammatory outcomes. Linear regression models with reference category: low probability of EH

Figure 4 depicts results from the first sensitivity analysis. Adjustment for selective dropout among individuals ever experiencing EH and with poor scores in markers of early ageing intensified the estimated associations. This suggests that selection bias most likely underestimated the true association between accumulated EH and markers of ageing (Fig. 4). Using binary measures (ever versus never in EH and below versus at or above median outcome), we found that a 1%-point (corresponding to 19.2%) lower participation in the ever in EH/below median outcome group resulted in 10–11%-points higher relative risk depending on outcome measure of interest. Selection bias is most likely to affect results including jump height and chair rise performance because participants in poor health did not participate in the strenuous tests.

Fig. 4

Crude relative risk estimates of score below median in marker of ageing among participants experiencing economic hardship (EH) at least once corrected for various degrees of selection bias

In the second sensitivity analysis, we adjusted for highest educational level in 1986 instead of 2009 and thereby changed the educational level of 15% of the sample. The results from the main analysis were relatively robust for all outcomes. However, there was a tendency that the associations were attenuated for the categorical EH exposure and boosted for the trajectories after adjusting for educational level in 1986 (Online Resource 3). Third, main analyses were repeated solely using information on EH in the period 1999–2008. The results yielded stronger associations with the outcomes for those in EH for one, two or three years, whereas the associations became weaker for those in EH for four or more years (Online Resource 4). Fourth, we excluded participants with self-reported illness at any point in life to address reverse causation. This attenuated the estimates for physical and cognitive outcomes, and strengthened inflammatory marker associations (Online Resource 5). In the latter two sensitivity analyses, a reduction in exposure period and sample size, respectively, led to reduced statistical power.


This study suggests that accumulation of four or more years of EH across adult life is associated with markers of early ageing. Spending four or more years of adult life in EH was consistently related to poorer physical capability (chair rise performance, grip strength, jump height and balance), poorer cognitive function and higher levels of CRP and IL-6, compared to not experiencing EH throughout the period. Categorization of adult life EH into trajectories showed that those with a high probability of EH over time had adverse physical capability and a higher CRP level, compared to those in low probability of EH. A similar but weaker association was found for those in rising probability of EH, and even the experience of declining compared to low probability of EH was associated with poorer jump height.

Previous studies have demonstrated inverse associations between EH and physical capability indicators such as activities of daily living in middle-aged to older people (Hessel et al. 2016), self-rated functional limitations in older people (Kahn and Pearlin 2006) and self-reported physical functioning in older people (Lynch et al. 1997). Studies show similar associations between EH and measures of self-reported cognitive difficulties (i.e. remembering things, paying attention, finding the right word and forgetting placement of things) in older people (Lynch et al. 1997) and widely used tests of verbal memory, performance on speed domains and executive skills among middle-aged (Al Hazzouri et al. 2017). Less evidence exists on the association between EH and inflammatory markers, but one study found greater wealth loss to be related to increased CRP levels among older people (Boen and Yang 2016).

One study investigated the association between 30-year-long trajectories of EH (based on 12 measurements of equalized household income below 125% of the official poverty threshold) and subsequent 12-year-long trajectories of diseases including heart disease, hypertension, diabetes, arthritis and cancer (Willson and Shuey 2016). The identified trajectories of EH were similar to our trajectories despite including two trajectories with declining probability of EH (i.e. an early decline in EH and a later decline in EH trajectory). The findings suggest that the chance of experiencing a low probability of disease was 25% higher in the never in EH trajectory, compared to the persistent EH trajectory (Willson and Shuey 2016). This is comparable to the patterns observed in our results for markers of early ageing (e.g. 31% better balance among those in the low compared to high probability of EH trajectory). Our findings further suggest that moving out of EH protects against accelerated ageing, whereas experiencing increasing probability of EH increases the probability. This is in line with the previously reported relationship (Willson and Shuey 2016).

Conflicting results have also been published. One study found a dose–response relationship between measures of self-reported financial stress across 16 years and musculoskeletal pain in middle-aged women, while no association was found when using disposable personal income < 60% of the tax office median as the EH indicator (Ahnquist et al. 2007). This suggests that perceived EH is more important to health than the more objective measure of income. The difference could also be explained by the use of disposable personal income rather than household income. If expenses and costs of living are covered by additional members of the household, an income loss might not relate negatively to health. Another study found comparable associations between a measure of perceived EH and household income < 200% of federal poverty level in relation to cognitive function among middle-aged (Al Hazzouri et al. 2017) supporting a hypothesis of household income being more strongly related to ageing markers than personal income.

In our study, most of the participants experiencing EH did so prior to age 45 years, while those with several years in EH were naturally more likely to also be exposed at a later age. When our results did not show convincing associations between one or few years in EH and markers of ageing, it may therefore be due to early EH not being significantly related to markers of ageing. One possible explanation could be that EH at an early age more likely results from being a student or short-term employment and job insecurity while establishing ones career. Such years in EH are expectedly less stressful than EH caused by for example job loss of a long held position later in life when costs of living have increased. Results from our trajectory analyses suggest that experiencing a rising probability of EH is associated with poorer physical capability and increased CRP level whereas experiencing a declining probability of EH is not. This could support a hypothesis of early EH posing a weaker strain on health than later EH. However, an alternative explanation could be that declining health across the life course increases the probability of EH and accelerates early ageing (further discussed below). Disentangling the potential effect of timing (or sensitive periods) from accumulation or mobility in and out of EH becomes increasingly relevant as time between exposure and outcome increases (Hallqvist et al. 2004). Such work should use clear causal hypotheses and use specific measures of exposure and outcome. Thus, it is beyond the scope of this study focusing on the broad concept of early ageing.

This study extends previous research on the relationship between EH and markers of ageing by including annual measures on EH across 22 consecutive years of the adult lifespan together with biomarkers of early ageing and relevant covariates. Our results were relatively robust across different outcomes and the positive association between EH and early ageing remained in a number of sensitivity analyses. While the study design of combining (birth) cohorts with National registers has several strengths such as the standardized recording of exposure measures with very few missing values and objective outcome measures, it also has limitations. The most severe problem relates to selective dropout influencing the internal validity of the cohort data. This most likely resulted in conservative estimates of the associations as demonstrated by results from our simple quantitative bias analysis, and thus it cannot explain the positive findings from this study. Furthermore, we cannot reject any true association to exist for statistically insignificant estimates or that the true associations are stronger in reality than found in this study. It is worth noticing that this sensitivity analysis did not take into account potential confounding. Another limitation we would like to highlight is time-varying confounding, especially in relation to poor health, which may cause reverse causality. Although previous studies found it unlikely that reverse causation could fully explain the association between EH and markers of early ageing (Lynch et al. 1997; Al Hazzouri et al. 2017), it is likely that observed associations would be attenuated from adjusting for health status along the exposure period. Results from our sensitivity analysis excluding participants reporting previous or current diseases lends support to time-varying confounding partly explaining some of the associations. Finally, the external validity of the findings is threatened by the study population resembling a selected share of the Danish population. Consequently, the findings should be generalized with caution, especially when considering societies substantially different from the Danish, people living in rural areas or people with poor health.


The collective evidence from this study shows that few years in EH across the adult life course are not associated with early ageing. However, EH for four or more years was associated with poorer physical capability, cognitive function and increased inflammatory levels in midlife. High probability of EH across adulthood was similarly related to poorer physical capability and CRP, but not cognitive function and the remaining inflammatory markers.

Preventive initiatives focusing on reducing the burden of sustained economic hardship may lead to increased healthy ageing. Future research should, however, carefully consider the potential influence of selection bias on the findings when employing data from longitudinal cohort studies with long-term follow-up. Future studies should additionally seek to elucidate the potential influence of time-varying confounding and other shortcomings of this study.


  1. Ahnquist J, Fredlund P, Wamala SP (2007) Is cumulative exposure to economic hardships more harzardous to women’s health than men’s? a 16-year follow-up study of the Swedish survey of living conditions. J Epidemiol Community Heal 61:331–336.

    Article  Google Scholar 

  2. Al Hazzouri AZ, Elfassy T, Sidney S et al (2017) Sustained economic hardship and cognitive function: the coronary artery risk development in young adults study. Am J Prev Med 52:1–9.

    Article  Google Scholar 

  3. Amthauer R, Brocke B, Liepmann D, Beauducel A (2001) Intelligenz-Struktur-Test 2000 R. Hogrefe Verlag, Göttingen

    Google Scholar 

  4. Avlund K, Damsgaard MT, Osler M (2004) Social position and functional decline among non-disabled old men and women. Eur J Public Health 14:212–216

    Article  Google Scholar 

  5. Avlund K, Osler M, Mortensen EL et al (2014) Copenhagen aging and midlife biobank (CAMB): an introduction. J Aging Health 26:5–20.

    Article  Google Scholar 

  6. Baadsgaard M, Quitzau J (2011) Danish registers on personal income and transfer payments. Scand J Public Health 39:103–105.

    Article  Google Scholar 

  7. Ben-Shlomo Y, Cooper R, Kuh D (2016) The last two decades of life course epidemiology, and its relevance for research on ageing. Int J Epidemiol 45:973–988.

    Article  Google Scholar 

  8. Boen C, Yang YC (2016) The physiological impacts of wealth shocks in late life: evidence from the great recession. Soc Sci Med 150:221–230.

    Article  Google Scholar 

  9. Cevenini E, Monti D, Franceschi C (2013) Inflamm-ageing. Curr Opin Clin Nutr Metab Care 16:14–20.

    Article  Google Scholar 

  10. Cooper R, Kuh D, Hardy R et al (2010) Objectively measured physical capability levels and mortality: systematic review and meta-analysis. BMJ 341:c4467.

    Article  Google Scholar 

  11. Cooper R, Kuh D, Cooper C et al (2011) Objective measures of physical capability and subsequent health: a systematic review. Age Ageing 40:14–23.

    Article  Google Scholar 

  12. Cooper R, Hardy R, Sayer AA, Kuh D (2014a) A life course approach to physical capability. In: Kuh D, Cooper R, Hardy R et al (eds) A life course approach to healthy ageing, First. Oxford University Press, Oxford, pp 16–31

    Google Scholar 

  13. Cooper R, Stafford M, Hardy R et al (2014b) Physical capability and subsequent positive mental wellbeing in older people: findings from five HALCyon cohorts. Age (Omaha) 36:445–456.

    Article  Google Scholar 

  14. Eurostat - Statistics Explained (2014) Glossary:At-risk-of-poverty rate. Accessed 23 Mar 2018

  15. Fairfax A, Balnave R, Adams R (1995) Variability of grip strength during isometric contraction. Ergonomics 38:1819–1830

    Article  Google Scholar 

  16. Galobardes B, Shaw M, Lawlor DA et al (2006) Indicators of socioeconomic position (part 1). J Epidemiol Community Health 60:7–12.

    Article  Google Scholar 

  17. Glymour MM, Greenland S (2008) Causal diagrams. In: Rothman KJ, Greenland S, Lash TL (eds) Modern epidemiology. Lippincott Williams & Wilkins, Philadelphia, pp 183–210

    Google Scholar 

  18. Hallqvist J, Lynch J, Bartley M et al (2004) Can we disentangle life course processes of accumulation, critical period and social mobility? an analysis of disadvantaged socio-economic positions and myocardial infarction in the Stockholm heart epidemiology program. Soc Sci Med 58:1555–1562.

    Article  Google Scholar 

  19. Harper S (2014) Economic and social implications of aging societies. Science (80-) 346:587–591.

    Article  Google Scholar 

  20. Hessel P, Avendano M, Hayward M et al (2016) Economic downturns during the life-course and late-life health: an analysis of 11 European countries. Eur J Public Health 41:87–107.

    Article  Google Scholar 

  21. Holsgaard Larsen A, Caserotti P, Puggaard L, Aagaard P (2007) Reproducibility and relationship of single-joint strength vs multi-joint strength and power in aging individuals. Scand J Med Sci Sport 17:43–53.

    Article  Google Scholar 

  22. Hurst L, Stafford M, Cooper R et al (2013) Lifetime socioeconomic inequalities in physical and cognitive aging. Am J Public Health 103:1641–1648.

    Article  Google Scholar 

  23. Jensen VM, Rasmussen AW (2011) Danish education registers. Scand J Public Health 39:91–94.

    Article  Google Scholar 

  24. Johnson TV, Abbasi A, Master VA (2013) Systematic review of the evidence of a relationship between chronic psychosocial stress and C-reactive protein. Mol Diagn Ther 17:147–164.

    Article  Google Scholar 

  25. Jones BL, Nagin DS (2007) Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res 35:542–571.

    Article  Google Scholar 

  26. Jones BL, Nagin DS (2013) A Note on a stata plugin for estimating group-based trajectory models. Sociol Methods Res 42:608–613.

    Article  Google Scholar 

  27. Kahn JR, Pearlin LI (2006) Financial strain over the life course and health among older adults. J Health Soc Behav 47:17–31.

    Article  Google Scholar 

  28. Krieger N, Williams DR, Moss NE (1997) Measuring social class in US public health research: concepts, methodologies, and guidelines. Annu Rev Public Health 18:341–378.

    Article  Google Scholar 

  29. Kubzansky LD, Seeman TE, Glymour MM (2014) Biological pathways linking social conditions and health - Plausible mechanisms and emerging puzzles. In: Berkman LF, Kawachi I, Glymour MM (eds) Social epidemiology, 2nd edn. Oxford University Press, New York, pp 512–561

    Google Scholar 

  30. Lash T, Fox M, Fink A (2009) Applying quantitative bias analysis to epidemiologic data. Springer Science & Business Media, Berlin

    Google Scholar 

  31. Lund R, Mortensen EL, Christensen U et al (2016) Cohort profile: the Copenhagen aging and midlife biobank (CAMB). Int J Epidemiol 45:1044–1053.

    Article  Google Scholar 

  32. Lynch J, Kaplan G, Shema S (1997) Cumulative impact of sustained economic hardship on physical, cognitive, psychological and social functioning. N Engl J Med 337:1889–1895.

    Article  Google Scholar 

  33. Mortensen EL, Flensborg-Madsen T, Molbo D et al (2014) The relationship between cognitive ability and demographic factors in late midlife. J Aging Health 26:37–53.

    Article  Google Scholar 

  34. Nazmi A, Victora CG (2007) Socioeconomic and racial/ethnic differentials of C-reactive protein levels: a systematic review of population-based studies. BMC Public Health 7:1–12.

    Article  Google Scholar 

  35. Oakes JM, Andrade KE (2017) The measurement of socioeconomic status. In: Oakes JM, Kaufman JS (eds) Methods in social epidemiology, 2nd edn. Jossey-Bass a Wiley Brand, San Francisco, pp 23–42

    Google Scholar 

  36. OECD (2018) Poverty rate (indicator). Accessed 28 Mar 2018

  37. Perna L, Wahl HW, Mons U et al (2015) Cognitive impairment, all-cause and cause-specific mortality among non-demented older adults. Age Ageing 44:445–451.

    Article  Google Scholar 

  38. Richards M, Deary IJ (2014) A life course approach to cognitive capability. In: Kuh D, Cooper R, Hardy R et al (eds) A life course approach to healthy ageing, First. Oxford University Press, Oxford, pp 32–45

    Google Scholar 

  39. Ritchie C, Trost SG, Brown W, Armit C (2005) Reliability and validity of physical fitness field tests for adults aged 55 to 70 years. J Sci Med Sport 8:61–70.

    Article  Google Scholar 

  40. Simons RL, Kit M, Beach SRH et al (2016) Economic hardship and biological weathering: the epigenetics of aging in a U. S. sample of black women. Soc Sci Med 150:192–200.

    Article  Google Scholar 

  41. Strand BH, Cooper R, Hardy R et al (2011) Lifelong socioeconomic position and physical performance in midlife: results from the British 1946 birth cohort. Eur J Epidemiol 26:475–483.

    Article  Google Scholar 

  42. UCLA Institute for Digital Research and Education (2016) FAQ How do I interpret a regression model when some variables are log transformed? In: UCLA Inst. Digit. Res. Educ. Accessed 1 Jun 2018

  43. Western B, Bloome D, Sosnaud B, Tach L (2012) Economic Insecurity and Social Stratification. Annu Rev Sociol 38:341–359.

    Article  Google Scholar 

  44. Willson AE, Shuey KM (2016) Life course pathways of economic hardship and mobility and midlife trajectories of health. J Health Soc Behav 57:407–422.

    Article  Google Scholar 

Download references


The authors thank the staff at Department of Public Health and National Research Centre for the Working Environment, who undertook the data collection. Further thanks to Kirsten Avlund†, Nils-Erik Fiehn, Åse Marie Hansen, Poul Holm-Pedersen and Merete Osler who participated in the initiation and establishment of the Copenhagen Ageing and Midlife Biobank from 2009 to 2011. 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 in Work, Unemployment and Health. Further, the authors would like to thank the Social Inequalities in Ageing (SIA) project, funded by NordForsk, Project No. 74637, for valuable collaboration relating to this project.


This work was funded by the Center for Healthy Aging established by a grant from the Nordea Foundation. The research leading to these results was carried out as part of the Social Inequalities in Ageing (SIA) project, funded by NordForsk, Project No. 74637. The Copenhagen Ageing and Midlife Biobank has been supported by a generous grant from the VELUX FOUNDATIONS (VELUX26145 and 31539).

Author information



Corresponding author

Correspondence to Else Foverskov.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Responsible editor: Susanne Iwarsson.

Else Foverskov and Gitte Lindved Petersen are co-first authors.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (DOCX 52 kb)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Foverskov, E., Petersen, G.L., Pedersen, J.L.M. et al. Economic hardship over twenty-two consecutive years of adult life and markers of early ageing: physical capability, cognitive function and inflammation. Eur J Ageing 17, 55–67 (2020).

Download citation


  • Economic hardship
  • Early ageing
  • Life course
  • Physical capability
  • Cognitive function
  • Inflammation