Correspondence to Dr Pierre Meneton; [email protected]
STRENGTHS AND LIMITATIONS OF THIS STUDY
The study analysed recent data collected from a large population-based cohort.
Data were robust, based on administrative sources for the assessment of lifetime unemployment exposure and on clinical sources for diagnoses of risk factors, cardiovascular diseases, cancers and mortality.
Associations between lifetime unemployment exposure, the prevalence of risk factors at baseline and the rates of cardiovascular diseases, cancers and deaths during follow-up were assessed while adjusting for major confounders.
As the cohort was not representative of the general population, the external validity of the findings is not warranted.
Despite the large population size, the lack of statistical power precluded the assessment of specific types of cardiovascular diseases and cancers as well as the causes of death.
Introduction
Unemployment is associated with a high prevalence and incidence of cardiovascular diseases1–6 and cancers,1 7 possibly translating into premature mortality.8 9 In the hypothesis where these links are causal, the underlying pathological pathways remain unclear. One obvious possibility is that unemployment may favour the occurrence of unhealthy behaviours and other risk factors. Indeed, unemployment is associated with an increased prevalence and incidence of smoking, non-moderate alcohol consumption, leisure-time physical inactivity, unbalanced diet as well as obesity, hypertension, diabetes, sleep disorders and depression.10–17 These are well-known risk factors that could mediate in part the observed association of unemployment with the prevalence and incidence of cardiovascular diseases and cancers.
Testing this possibility requires taking into account the fact that unemployment is strongly interrelated with both social position and working conditions,18 which are powerful determinants of health and mortality.19 20 Thus, individuals with a low social position, as measured by educational level, occupational class or income, have higher risks of cardiovascular diseases21 22 and cancers23 24 that could be related to overexposure to risk factors such as smoking, non-moderate alcohol consumption, leisure-time physical inactivity, obesity, diabetes, hypertension, dyslipidaemia, sleep disorders and depression.25–31 Likewise, individuals with bad working conditions have increased risks of cardiovascular diseases32 and cancers33 possibly because they are overexposed to smoking, non-moderate alcohol consumption, leisure-time physical inactivity, obesity, hypertension, diabetes, sleep disorders and depression.34–41
The aim of the present study is to explore the extent to which these risk factors have a mediating role in the association of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates in a large population-based cohort.
Methods
Study population
205 203 adults aged 18–75 years and affiliated with the general health insurance system, which covers 85% of the French population, were enrolled in the CONSTANCES cohort between January 2012 and December 2020 using a random sampling scheme stratified on sex, age, socioeconomic status and region.42 The inclusion rate was 7.3% with criteria comprising the obligation to provide written informed consent, to undergo a comprehensive health examination in one of the twenty-one participating medical centres scattered across the metropolitan territory and to complete questionnaires on lifestyle, health-related behaviours, social and occupational conditions.
The prevalence of risk factors at baseline was analysed in participants who were exposed to unemployment during their lifetime (n=99 430). Cardiovascular disease, cancer and mortality rates were analysed in a subset of these participants (n=54 679) who could be followed for 7 years after baseline (online supplemental figure 1).
Patient and public involvement
Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research.
Lifetime exposure to unemployment
Unemployment exposure of participants during their lifetime was documented by a questionnaire in which they were asked to report each time they had stopped working for a period of more than 6 months and why (unemployment, health issue, pregnancy or parental leave). Only participants who gave unemployment as the reason and not health issue, pregnancy or parental leave were considered unemployed. The existence of unemployment episodes was confirmed by administrative data from the French national pension system which also provided the total number of unemployed quarters for each participant. The analyses were ultimately based on the number of unemployed quarters that could not be entirely captured by self-reporting of 6-month unemployment episodes and when some discrepancy was present, that is, when participants declared no unemployment episodes but had recorded unemployed quarters. Note that the existence of a recorded unemployed quarter in the French national pension system did not mean that people were unemployed for 3 months; the whole quarter was validated as soon as people did not work for 50 separate or consecutive days. Note also that unemployed periods occurring before the first job were not captured by self-reporting of 6-month unemployment episodes or by the administrative recording of unemployed quarters. The distribution of unemployed quarters in participants who were exposed to unemployment during their lifetime is reported in online supplemental figure 2.
Social position and working conditions at baseline
The social position of participants was characterised by several indicators including education, occupation, income, spouse’s occupation and social vulnerability. The distributions of these indicators are shown in online supplemental table 1. Educational attainment was classified into four levels depending on the number of years of study: ≤11, 12–13, 14–16 or ≥17. Occupation of participants and spouses was reduced from a 10-level classification in the original inquiry to three grades: blue collar/clerk, intermediate and management. Income that included monthly earnings of all household members was ranked as low (below €1500), middle (between €1500 and €2800), high (between €2800 and €4200) or very high (above €4200). These arbitrary thresholds were dictated by the inquiry that included seven levels of income and the need to balance the number of participants between groups. Social vulnerability was evaluated by the EPICES score, based on a questionnaire comprising 11 binary items exploring material and social deprivation,43 that was categorised into terciles (low, intermediate or high social vulnerability) for the analyses. Given that these different indicators assessed complementary and interdependent aspects of social position, a global score was calculated by giving each indicator a value of 1 to the least privileged group, 2 or 3 to intermediary groups and 3 or 4 to the most privileged group depending on whether the indicator encompassed 3 or 4 levels, by summing the values and by dividing the sum by the number of available indicators for each participant. This global score was categorised into terciles (low, middle or high social position) as previously reported.44
Working conditions of participants were assessed by a total of 19 occupational exposures whose distributions are shown in online supplemental table 2. These included a series of organisational, physical, biomechanical, chemical and psychosocial factors: commuting time, clocking in and out, regular working hours (on daily and weekly basis), long working hours (over 10 hours per week day), night work, dealing with the public, driving on public road, repetitive work (imposed by a machine, a procedure or someone), working with a screen, standing work posture, handling heavy loads (over one kilogram), physically demanding work, exposure to vibrations, exposure to noise, outdoor work, working in the cold, working in the heat, exposure to chemicals and the scale assessing effort–reward imbalance of work that was divided into terciles (low, average or high imbalance).45 As these occupational exposures were significantly interrelated with each other, working conditions were considered as a whole, which is reality for workers who are not facing only one or a few exposures. Thus, all exposures were combined into a global score that was calculated by giving to each exposure a value of 1 to the least exposed group, 3 to the more exposed group and 2 to intermediary groups whenever the exposure encompassed 3 levels, by summing the values and by dividing the sum by the number of available exposures for each participant. This global score was categorised into terciles (bad, average or good working conditions) as previously described.44
Risk factors at baseline
Besides non-modifiable risk factors that included sex, age divided into terciles (18–39, 40–54, 55–75 years old) and parental histories of cardiovascular disease and cancer (coded as binary variables (Y/N)), several behavioural and clinical factors were considered and coded as binary variables to estimate their prevalence at baseline. Behavioural factors included smoking, non-moderate alcohol consumption (>3 drinks/day in men, >2 drinks/day in women) and leisure-time physical inactivity. Leisure-time physical inactivity assessment was based on a three-item questionnaire inquiring about regular practice of walking, cycling or other sports, gardening and housekeeping over the past 12 months; each item was noted 0 if the answer was no, 1 if the practice was regular but low (less than 2 hours per week), 2 if the practice was regular and higher; the score calculated by summing the items which ranged from 0 (not active at all) to 6 (very active) was used to characterise leisure-time physical inactivity (participants with a score <2). Clinical factors included obesity, hypertension, dyslipidaemia, diabetes, sleep disorders and depressive symptoms. These factors were assessed during the visit in the medical centres using standardised procedures in order to guarantee high-quality data.46 Participants were instructed to fast before the visit that took place between 8:00 and 10:00 hours, ensuring a fasting period of at least 8 hours. Height and weight were measured for the calculation of body mass index and the diagnosis of obesity (body mass index ≥30 kg/m2). Blood was collected to determine concentrations of glucose, total cholesterol and triglycerides. Diabetes was diagnosed by fasting glycaemia ≥7 mmol/L and/or antidiabetic drug (Anatomical Therapeutic Chemical (ATC) code A10) use while dyslipidaemia was characterised by hypercholesterolaemia (≥6.5 mmol/L) and/or hypertriglyceridaemia (≥2.2 mmol/L) and/or lipid-lowering drug (ATC code C10) use. Systolic and diastolic blood pressures were measured twice in sitting position on each arm at 2 min interval after 5 min of rest using an automated sphygmomanometer. The arm giving the highest values was considered as the reference arm on which a third measure was taken after one additional minute of rest. The average of the three measurements was used for the analyses. Hypertension was diagnosed if systolic blood pressure was ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg and/or antihypertensive drugs (ATC codes C02, C03, C07, C08, C09) were taken. Depressive symptoms were assessed using the Centre of Epidemiologic Studies Depression (CESD) scale and defined by a score ≥19 in both sexes. Sleep disorders were based on the repetitive occurrence of multiple wake-ups at night and/or fatigue after normal wake-up and/or sleep duration too short (<5 hours 30 min) or too long (>8 hours 30 min). When a clinical risk factor was diagnosed, physicians systematically inquired about the age at which the risk factor was first detected during lifetime.
Risk factors at baseline were also analysed as numeric variables, whose distributions are reported in online supplemental figure 3, in order to assess their mediating role in the associations of lifetime unemployment exposure with cardiovascular, cancer and mortality rates during follow-up. Smoking was expressed by the habitual number of items (cigarettes, cigarillos, cigars, pipes) smoked per day, alcohol consumption by the habitual number of drinks consumed per day, leisure-time physical inactivity by the score ranging from 0 to 6 as described above, dyslipidaemia and diabetes by blood cholesterol, triglyceride and glucose levels measured during the medical visit, sleep disorders by sleep duration and depressive symptoms by the CESD score.
Cardiovascular disease, cancer and mortality rates during follow-up
Information on the occurrence of cardiovascular diseases, cancers and deaths during follow-up was obtained from the French administrative healthcare database that contains data on all French patients having long-term chronic diseases with the date of diagnosis, hospital stays with dates of entry and exit, diagnoses, death during hospitalisation and its causes. The most common types of cardiovascular diseases (stroke, coronary heart disease, peripheral arterial disease) were retained for the analyses and coded globally as a single binary variable (Y/N) combining the occurrence of any of the three diseases. Stroke was defined by the diagnosis of acute events (subarachnoid haemorrhage (International Classification of Disease-10 code, I60), intracerebral haemorrhage (I61), other non-traumatic intracranial haemorrhage (I62), cerebral infarction (I63), stroke not specified as haemorrhage or infarction (I64)) and/or chronic conditions related to strokes that occurred during follow-up (other cerebrovascular diseases (I67), cerebrovascular disorders in diseases classified elsewhere (I68), sequelae of cerebrovascular disease (I69), hemiplegia (G81)). Coronary heart disease was also defined by the diagnosis of acute events (angina pectoris (I20), acute myocardial infarction (I21), subsequent myocardial infarction (I22), current complications following acute myocardial infarction (I23), other acute ischaemic heart diseases (I24)) and/or chronic conditions related to coronary heart diseases that occurred during follow-up (chronic ischaemic heart disease (I25)). Peripheral arterial disease was characterised by the diagnosis of atherosclerosis (I70), other peripheral vascular diseases (I73) or arterial embolism and thrombosis (I74).
The most common types of cancers (breast, skin, prostate, cervical, thyroid, colon, lung and ovarian) were considered for the analyses and coded globally as a single binary variable (Y/N) combining the occurrence of any of these cancers. Breast cancer was defined by the diagnosis of malignant neoplasm of breast (C50), skin cancer by the diagnosis of melanoma (C43) or other malignant neoplasms of skin (C44), prostate cancer by the diagnosis of malignant neoplasm of prostate (C61), cervical cancer by the diagnosis of malignant neoplasm of cervix uteri (C53), thyroid cancer by the diagnosis of malignant neoplasm of thyroid gland (C73), colon cancer by the diagnosis of malignant neoplasm of colon (C18), rectosigmoid junction (C19) or rectum (C20), lung cancer by the diagnosis of malignant neoplasm of trachea (C33), bronchus or lung (C34), ovarian cancer by the diagnosis of malignant neoplasm of ovary (C56).
Mortality was coded as a single binary variable (Y/N) combining the occurrence of deaths from any cause.
Statistical analyses
Missing data for variables at baseline were handled using multiple imputation by chained equations with missing at random assumption.47 Through an iterative series of predictive models built from observed variables, each binary and categorical variable was, respectively, imputed using logistic and polytomous regression models after 30 iterations to ensure completeness and robustness of the analyses.
The directed acyclic graph in figure 1 describes the assumed relationships between lifetime unemployment exposure, the prevalence of cardiovascular risk factors at baseline, cardiovascular disease, cancer and mortality rates during follow-up and potential confounders that included sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline. The hypothesis was that part of the associations of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates was mediated by risk factors that were themselves associated with unemployment exposure. Note that cardiovascular, cancer and mortality rates referred to individual, not aggregate, data on the numbers of outcomes occurring during the follow-up.
Enlarge this image.Figure 1. Directed acyclic graph showing the assumed relationships between lifetime unemployment exposure, the prevalence of risk factors at baseline, cardiovascular disease, cancer and all-cause mortality rates during follow-up and the potential confounders.
The associations between lifetime unemployment exposure and the prevalence of risk factors at baseline were tested by logistic regression models with two levels of adjustment: models 1 were minimally adjusted for sex and age, and models 2 were additionally adjusted for social position and working conditions at baseline. The ORs calculated with these models were expressed over a one-unit change in unemployment exposure (ie, per unemployed quarter) and over the entire range of unemployed quarters. In this latter case, the OR represents the change in the odds of the outcome when the number of unemployed quarters varies from its minimum to its maximum value, that is, from 1 to 168. While the unit OR measures how much the odds of the outcome vary for each unemployed quarter, the range OR provides a broader picture of the influence of unemployment exposure by considering its entire range of variation.
In order to assess the possibility of reverse causation, where the occurrence of risk factors would have preceded unemployment exposure during lifetime, the mean, youngest and oldest ages at which participants were unemployed were compared with the ages at which risk factors were first reported (online supplemental figure 4). The youngest and oldest ages at which participants were unemployed were on average 28.7±9.1 (SD) and 40.6±12.0 years with a mean age of 35.0±9.5 years. In comparison, the ages at which the first reports of risk factors were made were on average 48.0±11.2 for hypertension, 48.4±10.9 for hypercholesterolaemia, 46.2±11.1 for hypertriglyceridaemia, 51.1±9.8 for diabetes and 36.8±12.3 years for depressive symptoms. Thus, unemployment episodes popped up on average much earlier than clinical risk factors during lifetime with a mean difference of approximately 10–15 years. Even the latest episodes occurred on average several years before the first reports of risk factors. The only exception was depressive symptoms that were first reported concomitantly with unemployment episodes. Unfortunately, no such chronology could be established between unemployment exposure and the first adoption of behavioural risk factors or the first occurrence of sleep disorders during lifetime due to the inherent imprecision of self-reporting of these risk factors.
The associations of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates were assessed by logistic regression models using two levels of adjustment: models 1 were minimally adjusted for sex, age and parental histories of cardiovascular disease and cancer, while models 2 were additionally adjusted for social position and working conditions at baseline. The ORs calculated with these models were also expressed per unemployed quarter and over the entire range of unemployed quarters as specified above.
The expected associations of risk factors at baseline with cardiovascular disease, cancer and mortality rates during follow-up were tested as confirmatory analyses. ORs expressed per unit change in risk factor levels and per change over the entire ranges of risk factor levels were calculated with logistic regression models adjusted for sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline.
Mediation analyses tested each risk factor separately and all together as long as they were associated at baseline with lifetime unemployment exposure. Estimates for direct and indirect effects of unemployment exposure on cardiovascular, cancer and mortality rates were computed using logistic regression models adjusted for sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline. Note that these estimates were not standardised as standardisation of the coefficients prior to estimating the effects does not increase the performance of the estimates.48 In addition to SE and 95% CIs that were obtained using bootstrapping with 5000 replicates, p values were calculated for the indirect effects using the Sobel test.49 Statistical significance was fixed a priori at two-sided p<0.05. Mediation analyses were performed with the PROCESS macro50 combined with SPSS Statistics V.27 (IBM) that was used for the other analyses.
Results
Characteristics of participants at baseline
Table 1 shows that the characteristics of participants and prevalence of risk factors at baseline were very similar in the whole set of participants who were exposed to unemployment during their lifetime (n=99 430) and in the subset (n=54 679) who were followed for 7 years after baseline to assess cardiovascular disease, cancer and mortality rates. The distribution of unemployed quarters (online supplemental figure 2) was also very similar with a mean (SD) of 13.22 (14.65) vs 13.73 (15.13).
Table 1Characteristics of participants and prevalence of risk factors at baseline in the whole set of participants who were exposed to unemployment during their lifetime and in the subset who were followed for 7 years after baseline
| Participants exposed to unemployment | Participants exposed to unemployment and followed during 7 years | ||||
| n=99 430 | n=54 679 | ||||
| n | % | n | % | ||
| Sex | Women | 53 883 | 54.2 | 29 636 | 54.2 |
| Men | 45 547 | 45.8 | 25 043 | 45.8 | |
| Age (years) | 18–39 | 32 464 | 32.7 | 17 169 | 31.4 |
| 40–54 | 33 935 | 34.1 | 18 864 | 34.5 | |
| 55–75 | 33 031 | 33.2 | 18 646 | 34.1 | |
| Parental history of cardiovascular disease | No | 75 169 | 75.6 | 41 119 | 75.2 |
| Yes | 24 261 | 24.4 | 13 560 | 24.8 | |
| Parental history of cancer | No | 66 220 | 66.6 | 36 690 | 67.1 |
| Yes | 33 210 | 33.4 | 17 989 | 32.9 | |
| Social position | High | 30 426 | 30.6 | 15 419 | 28.2 |
| Middle | 42 059 | 42.3 | 23 293 | 42.6 | |
| Low | 26 945 | 27.1 | 15 967 | 29.2 | |
| Working conditions | Good | 33 374 | 33.6 | 17 607 | 32.2 |
| Average | 41 975 | 42.2 | 23 184 | 42.4 | |
| Bad | 24 081 | 24.2 | 13 888 | 25.4 | |
| Smoking | No | 74 672 | 75.1 | 41 775 | 76.4 |
| Yes | 24 758 | 24.9 | 12 904 | 23.6 | |
| Non-moderate alcohol consumption | No | 88 194 | 88.7 | 48 118 | 88.0 |
| Yes | 11 236 | 11.3 | 6561 | 12.0 | |
| Leisure-time physical inactivity | No | 88 692 | 89.2 | 48 610 | 88.9 |
| Yes | 10 738 | 10.8 | 6069 | 11.1 | |
| Obesity | No | 85 410 | 85.9 | 47 024 | 86.0 |
| Yes | 14 020 | 14.1 | 7655 | 14.0 | |
| Hypertension | No | 58 962 | 59.3 | 31 768 | 58.1 |
| Yes | 40 468 | 40.7 | 22 911 | 41.9 | |
| Dyslipidaemia | No | 69 999 | 70.4 | 37 346 | 68.3 |
| Yes | 29 431 | 29.6 | 17 333 | 31.7 | |
| Diabetes | No | 93 464 | 94.0 | 51 125 | 93.5 |
| Yes | 5966 | 6.0 | 3554 | 6.5 | |
| Sleep disorders | No | 34 801 | 35.0 | 19 520 | 35.7 |
| Yes | 64 629 | 65.0 | 35 159 | 64.3 | |
| Depressive symptoms | No | 80 936 | 81.4 | 44 290 | 81.0 |
| Yes | 18 494 | 18.6 | 10 389 | 19.0 | |
Associations between lifetime unemployment exposure and the prevalence of risk factors at baseline
As reported in table 2, the prevalence of all risk factors at baseline was associated with lifetime unemployment exposure after adjustment for sex and age. However, additional adjustment for social position and working conditions at baseline substantially decreased ORs to the point where obesity and hypertension were no longer associated with lifetime unemployment exposure. For the other risk factors, ORs (95% CIs) per change over the entire range of unemployed quarters varied from 1.39 (1.18 to 1.63) for sleep disorders to 10.8 (9.2 to 12.9) for smoking, which was the strongest factor associated with lifetime unemployment exposure together with non-moderate alcohol consumption (4.61 (3.67 to 5.78)) and depressive symptoms (5.67 (4.63 to 6.94)).
Table 2ORs (95% CI) for the prevalence of risk factors at baseline according to lifetime unemployment exposure
| Unemployment | Models 1 | P value | Models 2 | P value | |
| Smoking | Unit | 1.0178 (1.0168 to 1.0188) | <0.0001 | 1.0143 (1.0133 to 1.0154) | <0.0001 |
| Range | 19.222 (16.292 to 22.679) | 10.854 (9.1580 to 12.863) | |||
| Non-moderate alcohol consumption | Unit | 1.0103 (1.0089 to 1.0116) | <0.0001 | 1.0091 (1.0078 to 1.0105) | <0.0001 |
| Range | 5.5467 (4.4447 to 6.9081) | 4.6100 (3.6684 to 5.7807) | |||
| Leisure-time physical inactivity | Unit | 1.0055 (1.0041 to 1.0069) | <0.0001 | 1.0032 (1.0018 to 1.0046) | <0.0001 |
| Range | 2.5141 (1.9933 to 3.1617) | 1.7209 (1.3505 to 2.1862) | |||
| Obesity | Unit | 1.0056 (1.0045 to 1.0067) | <0.0001 | 1.0006 (0.9994 to 1.0018) | 0.31 |
| Range | 2.5606 (2.1176 to 3.0917) | 1.1104 (0.9077 to 1.3559) | |||
| Hypertension | Unit | 1.0031 (1.0021 to 1.0040) | <0.0001 | 0.9996 (0.9987 to 1.0006) | 0.51 |
| Range | 1.6812 (1.4393 to 1.9643) | 0.9481 (0.8084 to 1.1120) | |||
| Dyslipidaemia | Unit | 1.0054 (1.0044 to 1.0064) | <0.0001 | 1.0020 (1.0010 to 1.0030) | <0.0001 |
| Range | 2.4808 (2.1124 to 2.9138) | 1.4068 (1.1927 to 1.6592) | |||
| Diabetes | Unit | 1.0084 (1.0069 to 1.0099) | <0.0001 | 1.0041 (1.0025 to 1.0056) | <0.0001 |
| Range | 4.0702 (3.1862 to 5.1830) | 1.9843 (1.5304 to 2.5634) | |||
| Sleep disorders | Unit | 1.0036 (1.0027 to 1.0046) | <0.0001 | 1.0019 (1.0010 to 1.0029) | <0.0001 |
| Range | 1.8499 (1.5796 to 2.1681) | 1.3897 (1.1824 to 1.6345) | |||
| Depressive symptoms | Unit | 1.0165 (1.0154 to 1.0177) | <0.0001 | 1.0104 (1.0092 to 1.0116) | <0.0001 |
| Range | 15.575 (12.839 to 18.887) | 5.6708 (4.6292 to 6.9405) |
ORs are expressed per unit change in unemployment exposure (ie, per unemployed quarter) or per change over the entire range of unemployed quarters. Models 1 were adjusted for sex and age. Models 2 were adjusted for sex, age, social position and working conditions at baseline.
Associations of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates during follow-up
Cardiovascular disease (n=668), cancer (n=1273) and mortality (n=399) rates during follow-up were, respectively, 1.22, 2.33 and 0.73%. Table 3 shows that all these rates were associated with lifetime unemployment exposure after adjustment for sex, age and parental histories of cardiovascular disease and cancer. After further adjustment for social position and working conditions at baseline, ORs (95% CIs) per change over the entire range of unemployed quarters varied were 3.95 (2.06 to 7.33) for cardiovascular diseases, 2.54 (1.51 to 4.19) for cancers and 5.35 (2.42 to 11.3) for mortality.
Table 3ORs (95% CIs) for cardiovascular disease, cancer and all-cause mortality rates during follow-up according to lifetime unemployment exposure
| Unemployment | Models 1 | P value | Models 2 | P value | |
| Cardiovascular diseases | Unit | 1.0100 (1.0063 to 1.0136) | <0.0001 | 1.0082 (1.0043 to 1.01219) | <0.0001 |
| Range | 5.3915 (2.8987 to 9.7060) | 3.9493 (2.0570 to 7.3290) | |||
| Cancers | Unit | 1.0070 (1.0040 to 1.0099) | <0.0001 | 1.0055 (1.0024 to 1.0085) | 0.0006 |
| Range | 3.2698 (1.9841 to 5.2838) | 2.5404 (1.5088 to 4.1918) | |||
| All-cause mortality | Unit | 1.0141 (1.0097 to 1.0183) | <0.0001 | 1.0100 (1.0052 to 1.0145) | <0.0001 |
| Range | 10.679 (5.1439 to 21.160) | 5.3498 (2.4166 to 11.279) |
ORs are expressed per unit change in unemployment exposure (ie, per unemployed quarter) or per change over the entire range of unemployed quarters. Models 1 were adjusted for sex, age and parental histories of cardiovascular disease and cancer. Models 2 were adjusted for sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline.
Associations of risk factors at baseline with cardiovascular disease, cancer and mortality rates during follow-up
The expected associations of risk factors at baseline with cardiovascular disease, cancer and mortality rates during follow-up were confirmed after adjustment for sex, age and parental histories of cardiovascular disease and cancer, social position and working conditions at baseline (table 4). Smoking and alcohol consumption were strongly associated with cardiovascular disease, cancer and mortality rates, with ORs (95% CIs) per change over the entire range of unemployed quarters varying from 3.94 (2.30 to 6.65) to 7.16 (3.72 to 13.5). Blood cholesterol and triglycerides were strongly associated with cardiovascular disease rate with ORs of 6.84 (95% CI 2.98 to 15.6) and 9.22 (95% CI 4.87 to 16.9), respectively. Blood glucose was strongly associated with mortality rate with an OR of 7.45 (95% CI 3.20 to 16.2). Several other significant associations were observed with ORs of smaller magnitude (table 4).
Table 4ORs (95% CIs) for cardiovascular disease, cancer and all-cause mortality rates during follow-up according to risk factor levels at baseline
| Risk factor | Cardiovascular diseases | P value | Cancers | P value | All-cause mortality | P value | |
| Smoking (items/day) | Unit | 1.0198 (1.0132 to 1.0263) | <0.0001 | 1.0138 (1.0083 to 1.0191) | <0.0001 | 1.0186 (1.0101 to 1.0268) | <0.0001 |
| Range | 7.1633 (3.7197 to 13.469) | 3.9403 (2.2986 to 6.6506) | 6.3638 (2.7458 to 14.139) | ||||
| Alcohol consumption (drinks/day) | Unit | 1.0741 (1.0340 to 1.1123) | 0.0004 | 1.0742 (1.0410 to 1.1062) | <0.0001 | 1.0870 (1.0376 to 1.1330) | 0.0007 |
| Range | 5.1834 (2.1583 to 11.575) | 5.1909 (2.5205 to 10.194) | 6.8159 (2.3380 to 17.700) | ||||
| Leisure-time physical inactivity score | Unit | 1.1080 (1.0541 to 1.1646) | <0.0001 | 1.0703 (1.0318 to 1.1102) | 0.0003 | 1.1759 (1.1026 to 1.2542) | <0.0001 |
| Range | 1.8511 (1.3724 to 2.4956) | 1.5040 (1.2072 to 1.8730) | 2.6447 (1.7971 to 3.8925) | ||||
| Blood cholesterol (mmol/l) | Unit | 1.1789 (1.0979 to 1.2651) | <0.0001 | 0.9994 (0.9469 to 1.0546) | 0.98 | 1.0953 (0.9971 to 1.2043) | 0.06 |
| Range | 6.8419 (2.9777 to 15.594) | 0.9941 (0.5288 to 1.8613) | 2.8978 (0.9674 to 8.7736) | ||||
| Blood triglycerides (mmol/l) | Unit | 1.3285 (1.2243 to 1.4353) | <0.0001 | 1.0455 (0.9638 to 1.1301) | 0.28 | 1.1275 (0.9961 to 1.2644) | 0.06 |
| Range | 9.2207 (4.8685 to 16.882) | 1.4167 (0.7500 to 2.6024) | 2.5561 (0.9705 to 6.2626) | ||||
| Blood glucose (mmol/l) | Unit | 1.0996 (1.0176 to 1.1823) | 0.02 | 1.0750 (1.0095 to 1.1411) | 0.03 | 1.2226 (1.1236 to 1.3218) | <0.0001 |
| Range | 2.5834 (1.1908 to 5.3286) | 2.0601 (1.0996 to 3.7409) | 7.4490 (3.2042 to 16.236) | ||||
| Sleep duration (hours/day) | Unit | 1.0370 (1.0002 to 1.0753) | 0.04 | 1.0584 (1.0313 to 1.0863) | <0.0001 | 1.0442 (0.9981 to 1.0924) | 0.06 |
| Range | 1.4927 (1.0024 to 2.2237) | 1.8688 (1.4047 to 2.4869) | 1.6092 (0.9795 to 2.6450) | ||||
| Depression score | Unit | 1.0143 (1.0048 to 1.0235) | 0.003 | 1.0031 (0.9961 to 1.0100) | 0.38 | 1.0216 (1.0099 to 1.0329) | 0.0003 |
| Range | 2.3465 (1.3366 to 4.0411) | 1.2065 (0.7919 to 1.8187) | 3.6064 (1.8133 to 6.9766) |
ORs are expressed per unit change in risk factor levels or per change over the entire ranges of risk factor levels. Models were adjusted for sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline.
Mediating roles of risk factors at baseline in the associations of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates during follow-up
When testing separately the mediating role of each risk factor, smoking was the strongest mediator of the associations of lifetime unemployment exposure with cardiovascular disease and cancer rates, explaining 10.8 and 10.6% of the associations, respectively (table 5). Smoking also explained 8.7% of the association with mortality rate. Other significant mediators were alcohol consumption that explained 5.0%, 5.6% and 3.4% of the associations with cardiovascular disease, cancer and mortality rates, respectively; depressive symptoms that explained 7.7% and 10.2% of cardiovascular disease and mortality rates, respectively; leisure-time physical inactivity that explained 3.5%, 3.4% and 3.7% of the associations with cardiovascular disease, cancer and mortality rates, respectively; and blood triglycerides that explained 5.9% of the association with cardiovascular disease rate. The percentages of the associations mediated by risk factors when these latter were tested all together were 21.1% (estimates (95% CIs) of the direct and indirect effects were 0.0075 (0.0022 to 0.0128), p=0.005 and 0.0020 (0.0011 to 0.0027), p<0.0001) for cardiovascular disease rate, 22.4% (0.0052 (0.0011 to 0.0093), p=0.01 and 0.015 (0.0009 to 0.0020), p<0.0001) for cancer rate, 17.7% (0.0102 (0.0035 to 0.0169), p=0.003 and 0.0022 (0.0011 to 0.0031), p<0.0001) for mortality rate.
Table 5Mediating role of each risk factor at baseline in the associations of lifetime unemployment exposure with cardiovascular disease, cancer and all-cause mortality rates during follow-up
| Effect | Estimate | SE | 95% CI | P value | % mediation | ||
| Smoking (items/day) | Cardiovascular diseases | Direct | 0.0083 | 0.0020 | 0.0044 to 0.0122 | <0.0001 | 89.2 |
| Indirect | 0.0010 | 0.0002 | 0.0007 to 0.0014 | <0.0001 | 10.8 | ||
| Cancers | Direct | 0.0059 | 0.0016 | 0.0028 to 0.0089 | 0.0002 | 89.4 | |
| Indirect | 0.0007 | 0.0001 | 0.0004 to 0.0010 | <0.0001 | 10.6 | ||
| All-cause mortality | Direct | 0.0105 | 0.0024 | 0.0058 to 0.0151 | <0.0001 | 91.3 | |
| Indirect | 0.0010 | 0.0002 | 0.0005 to 0.0014 | <0.0001 | 8.7 | ||
| Alcohol consumption (drinks/day) | Cardiovascular diseases | Direct | 0.0076 | 0.0022 | 0.0034 to 0.0118 | 0.0004 | 95.0 |
| Indirect | 0.0004 | 0.0001 | 0.0002 to 0.0005 | 0.0006 | 5.0 | ||
| Cancers | Direct | 0.0067 | 0.0016 | 0.0035 to 0.0100 | <0.0001 | 94.4 | |
| Indirect | 0.0004 | 0.0001 | 0.0002 to 0.0005 | <0.0001 | 5.6 | ||
| All-cause mortality | Direct | 0.0114 | 0.0026 | 0.0064 to 0.0164 | <0.0001 | 96.6 | |
| Indirect | 0.0004 | 0.0001 | 0.0001 to 0.0006 | 0.0009 | 3.4 | ||
| Leisure-time physical inactivity score | Cardiovascular diseases | Direct | 0.0083 | 0.0020 | 0.0044 to 0.0122 | <0.0001 | 96.5 |
| Indirect | 0.0003 | 0.0001 | 0.0001 to 0.0005 | 0.0006 | 3.5 | ||
| Cancers | Direct | 0.0057 | 0.0016 | 0.0026 to 0.0088 | 0.0004 | 96.6 | |
| Indirect | 0.0002 | 0.0001 | 0.0001 to 0.0003 | 0.002 | 3.4 | ||
| All-cause mortality | Direct | 0.0105 | 0.0024 | 0.0058 to 0.0152 | <0.0001 | 96.3 | |
| Indirect | 0.0004 | 0.0001 | 0.0002 to 0.0007 | <0.0001 | 3.7 | ||
| Blood cholesterol (mmol/l) | Cardiovascular diseases | Direct | 0.0086 | 0.0019 | 0.0048 to 0.0124 | <0.0001 | 98.9 |
| Indirect | 0.0001 | 0.0001 | 0.0001 to 0.0003 | 0.001 | 1.1 | ||
| Cancers | Direct | 0.0057 | 0.0016 | 0.0026 to 0.0087 | 0.0003 | 99.9 | |
| Indirect | 0.0000 | 0.0000 | 0.0000 to 0.0001 | 0.93 | 0.1 | ||
| All-cause mortality | Direct | 0.0105 | 0.0023 | 0.0060 to 0.0151 | <0.0001 | 99.1 | |
| Indirect | 0.0001 | 0.0001 | 0.0000 to 0.0002 | 0.11 | 0.9 | ||
| Blood triglycerides (mmol/l) | Cardiovascular diseases | Direct | 0.0080 | 0.0020 | 0.0042 to 0.0119 | <0.0001 | 94.1 |
| Indirect | 0.0005 | 0.0001 | 0.0004 to 0.0007 | <0.0001 | 5.9 | ||
| Cancers | Direct | 0.0057 | 0.0016 | 0.0026 to 0.0087 | 0.0003 | 98.3 | |
| Indirect | 0.0001 | 0.0001 | −0.0001 to 0.0002 | 0.32 | 1.7 | ||
| All-cause mortality | Direct | 0.0103 | 0.0024 | 0.0057 to 0.0149 | <0.0001 | 98.1 | |
| Indirect | 0.0002 | 0.0001 | 0.0000 to 0.0004 | 0.06 | 1.9 | ||
| Blood glucose (mmol/l) | Cardiovascular diseases | Direct | 0.0086 | 0.0019 | 0.0048 to 0.0124 | <0.0001 | 99.9 |
| Indirect | 0.0000 | 0.0000 | 0.0000 to 0.0001 | 0.39 | 0.1 | ||
| Cancers | Direct | 0.0056 | 0.0016 | 0.0025 to 0.0087 | 0.0003 | 99.9 | |
| Indirect | 0.0000 | 0.0000 | −0.0000 to 0.0001 | 0.39 | 0.1 | ||
| All-cause mortality | Direct | 0.0106 | 0.0024 | 0.0059 to 0.0152 | <0.0001 | 99.9 | |
| Indirect | 0.0000 | 0.0000 | −0.0000 to 0.0001 | 0.34 | 0.1 | ||
| Sleep duration (hours/day) | Cardiovascular diseases | Direct | 0.0090 | 0.0020 | 0.0051 to 0.0129 | <0.0001 | 99.9 |
| Indirect | 0.0000 | 0.0000 | −0.0001 to 0.0000 | 0.31 | 0.1 | ||
| Cancers | Direct | 0.0055 | 0.0016 | 0.0024 to 0.0087 | 0.0006 | 99.9 | |
| Indirect | 0.0000 | 0.0000 | −0.0001 to 0.0000 | 0.19 | 0.1 | ||
| All-cause mortality | Direct | 0.0109 | 0.0024 | 0.0062 to 0.0156 | <0.0001 | 99.9 | |
| Indirect | 0.0000 | 0.0000 | −0.0001 to 0.0000 | 0.31 | 0.1 | ||
| Depression score | Cardiovascular diseases | Direct | 0.0084 | 0.0023 | 0.0040 to 0.0128 | 0.0002 | 92.3 |
| Indirect | 0.0007 | 0.0002 | 0.0002 to 0.0011 | 0.005 | 7.7 | ||
| Cancers | Direct | 0.0053 | 0.0018 | 0.0017 to 0.0089 | 0.004 | 98.1 | |
| Indirect | 0.0001 | 0.0002 | −0.0002 to 0.0005 | 0.51 | 1.9 | ||
| All-cause mortality | Direct | 0.0088 | 0.0029 | 0.0031 to 0.0144 | 0.002 | 89.8 | |
| Indirect | 0.0010 | 0.0003 | 0.0004 to 0.0015 | 0.0005 | 10.2 |
Models were adjusted for sex, age, parental histories of cardiovascular disease and cancer, social position and working conditions at baseline.
Discussion
In agreement with previous reports,10–13 15–17 the present analyses confirm that high unemployment exposure during lifetime is associated with an increased prevalence of several common risk factors after adjustment for major confounders. Although observational, these results support the possibility of a causal relationship given that unemployment exposure largely preceded or was concomitant in average with the first occurrence of clinical risk factors during lifetime. Although it was not possible to establish such chronology between unemployment exposure and the first adoption of risky behaviours during lifetime, several studies suggest that unemployment exposure increases the incidence of smoking and high alcohol consumption.51–54
The present analyses also confirm that high unemployment exposure during lifetime is associated with increased cardiovascular disease, cancer and mortality rates after adjustment for major confounders.1–5 7–9 18 The main finding is that several risk factors that are associated with unemployment mediate in part the associations of lifetime unemployment exposure with cardiovascular disease, cancer and mortality rates. Smoking seems to be especially involved, which is not surprising given that unemployment may influence smoking behaviour in many ways, increasing the risk of relapse after cessation, the odds of starting smoking, smoking intensity or decreasing the odds of cessation.12 51 53 Alcohol consumption also has a significant mediating effect on cardiovascular disease, cancer and mortality rates, which is in line with the promoting effect of unemployment on drinking behaviour.12 54 55 Depressive symptoms are worth mentioning considering their mediating effects on cardiovascular disease rate and even more on mortality rate, this latter effect being also probably related to an increased suicide rate.56 However, the mediating effects of each risk factor remain modest, explaining at most 10% of the associations and reaching only 20% when considering risk factors all together. This suggests that unemployment may increase cardiovascular disease, cancer and mortality rates not only by overexposure to common risk factors but also through other pathways that have yet to be defined. Identifying these pathways may not be so easy as the potential stressful effects of unemployment are numerous and entangled. For example, neurohormonal changes such as elevated cortisol levels might be involved to some extent among many others.57
The present study has both strengths and limitations. Among the strengths is that potential confounding effects of social position and working conditions have been taken into account, which is critical given their strong interrelationships with unemployment exposure18 and their powerful influence on the incidence of risk factors and deaths.25–31 34–36 38–41 Another strength is that, except for risky behaviours, diagnoses of risk factors, cardiovascular diseases, cancers and deaths were based on clinical data. This is an important issue as questions have been raised concerning the validity of self-reported data for exploring the relationships between unemployment and health.58 59 Among the limitations is the external validity of the findings, which is not guaranteed given that they were obtained in a cohort which was not representative of the French population. The main reason for this lack of representativeness was the inclusion rate, which was low, although in line with those observed in other large population-based cohorts when participants are required to visit a medical centre for health-related exams.60 This resulted in the preferential selection of socially privileged people despite the stratified sampling strategy that tried to compensate for the lower response rate of individuals with low socioeconomic status.42 This selection bias, which increased the proportion of socially privileged individuals, likely weakens the relationships between unemployment exposure, disease risk and mortality. Indeed, although several major confounders which characterise socially privileged individuals (high education, high-skilled occupation, high income, good working conditions, low exposure to risky behaviours) have been adjusted for, these individuals often have a high health-related knowledge, a strong importance given to the care of one’s own health, a high capacity to optimise health service use, a high sense of control over life and problem-solving and a high resilience to the negative effects of psychosocial problems.61 Therefore, they are better off to cope with the various consequences of unemployment and as a result would be less prone to subsequent health issues. The second limitation is whether the estimates of the relationships reported in the present study are relevant to other high-income countries. Indeed, the relationships between unemployment exposure, health and mortality are strongly influenced by the social security and healthcare systems62 which vary substantially from one country to another. Although France provides an average unemployment compensation in terms of the replacement rate of wage, the duration of unemployment insurance benefit or the waiting period before getting the benefit after job loss,63 its healthcare system is above average64 and it is difficult to know to which extent the estimates can be extrapolated to other countries. The third limitation is that even though the lifetime chronology favours the hypothesis that unemployment exposure rises cardiovascular disease, cancer and mortality rates in part by increasing the prevalence of common risk factors, the observational nature of the analyses precludes the inference of causal relationships. Indeed, despite the adjustments for major potential confounders, unmeasured or uncontrolled factors may still significantly influence the observed associations and their interpretation. The fourth limitation is that the present analyses did not assess the linearity of the relationships between unemployment exposure, disease risk and mortality. Indeed, it has been suggested using aggregate data that these relationships might be U-shaped due to differential responsiveness of individuals according to the level of unemployment.65 The fifth limitation is that the present study analysed the cumulative effect of unemployed quarters over a lifetime but did not assess the possibility that unemployment may have a differential effect on the outcomes according to the age of exposure. The sixth limitation is that, despite the large population size, the lack of statistical power precluded the assessment of specific types of cardiovascular diseases and cancers, as well as the causes of death, associated with unemployment exposure.
In conclusion, even though the relatively short follow-up may not fully capture the long-term health implications of lifetime unemployment exposure, the analyses show that the associations of unemployment exposure with cardiovascular disease, cancer and mortality rates are mediated in part by an increased prevalence of common risk factors such as smoking, alcohol consumption and depression. However, these mediating effects are modest, pointing to other unidentified major pathways linking unemployment to health.
Data availability statement
Data are available on reasonable request. Personal health data underlying the findings of our study are not publicly available due to legal reasons related to data privacy protection. However, the data are available on reasonable request after approval from the French National Data Protection Authority. The email address for any inquiry is [email protected].
Ethics statements
Patient consent for publication
Not applicable.
Ethics approval
This study involves human participants and all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Declaration of
Contributors MSR performed statistical analyses, data interpretation and critical revision of the manuscript for important intellectual content; EW, AR, SK and NH were involved in study concept and design and performed critical revision of the manuscript for important intellectual content; MG and MZ obtained cohort funding and performed critical revision of the manuscript for important intellectual content; PM supervised the study and wrote the first draft of the manuscript. PM confirms that he had full access to all the data and is responsible for the overall content as guarantor.
Funding The cohort is supported by the Agence nationale de la recherche (ANR-11-INBS-0002), the Caisse nationale d’assurance maladie and was funded by the Institut pour la recherche en santé publique and the following sponsors: Ministère de la santé et des sports, Ministère délégué à la recherche, Institut national de la santé et de la recherche médicale, Institut national du cancer, Caisse nationale de solidarité pour l’autonomie, Merck Sharp & Dohme and L’Oréal.
Disclaimer The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
1 d’Errico A, Piccinelli C, Sebastiani G, et al. Unemployment and mortality in a large Italian cohort. J Public Health (Oxf) 2021; 43: 361–9. doi:10.1093/pubmed/fdz100
2 Dupre ME, George LK, Liu G, et al. The cumulative effect of unemployment on risks for acute myocardial infarction. Arch Intern Med 2012; 172: 1731–7. doi:10.1001/2013.jamainternmed.447
3 Gallo WT. Evolution of research on the effect of unemployment on acute myocardial infarction risk. Arch Intern Med 2012; 172: 1737–8. doi:10.1001/jamainternmed.2013.1835
4 Lundin A, Falkstedt D, Lundberg I, et al. Unemployment and coronary heart disease among middle-aged men in Sweden: 39 243 men followed for 8 years. Occup Environ Med 2014; 71: 183–8. doi:10.1136/oemed-2013-101721
5 Meneton P, Kesse-Guyot E, Méjean C, et al. Unemployment is associated with high cardiovascular event rate and increased all-cause mortality in middle-aged socially privileged individuals. Int Arch Occup Environ Health 2015; 88: 707–16. doi:10.1007/s00420-014-0997-7
6 Sanchez Rico M, Plessz M, Airagnes G, et al. Cardiovascular burden and unemployment: A retrospective study in a large population-based French cohort. PLoS One 2023; 18: e0288747. doi:10.1371/journal.pone.0288747
7 Lynge E. Unemployment and cancer: a literature review. IARC Sci Publ 1997; 343–51.
8 Brand JE. The far-reaching impact of job loss and unemployment. Annu Rev Sociol 2015; 41: 359–75. doi:10.1146/annurev-soc-071913-043237
9 Roelfs DJ, Shor E, Davidson KW, et al. Losing life and livelihood: a systematic review and meta-analysis of unemployment and all-cause mortality. Soc Sci Med 2011; 72: 840–54. doi:10.1016/j.socscimed.2011.01.005
10 Blanchflower DG, Bryson A. Unemployment and sleep: evidence from the United States and Europe. Econ Hum Biol 2021; 43: S1570-677X(21)00066-6. doi:10.1016/j.ehb.2021.101042
11 Caban-Martinez AJ, Lee DJ, Goodman E, et al. Health indicators among unemployed and employed young adults. J Occup Environ Med 2011; 53: 196–203. doi:10.1097/JOM.0b013e318209915e
12 Henkel D. Unemployment and substance use: a review of the literature (1990-2010). Curr Drug Abuse Rev 2011; 4: 4–27. doi:10.2174/1874473711104010004
13 Hughes A, Kumari M. Unemployment, underweight, and obesity: Findings from Understanding Society (UKHLS). Prev Med 2017; 97: 19–25. doi:10.1016/j.ypmed.2016.12.045
14 Nygren K, Gong W, Hammarström A. Is hypertension in adult age related to unemployment at a young age? Results from the Northern Swedish Cohort. Scand J Public Health 2015; 43: 52–8. doi:10.1177/1403494814560845
15 Paul KI, Moser K. Unemployment impairs mental health: Meta-analyses. J Vocat Behav 2009; 74: 264–82. doi:10.1016/j.jvb.2009.01.001
16 Plessz M, Ezdi S, Airagnes G, et al. Association between unemployment and the co-occurrence and clustering of common risky health behaviors: Findings from the Constances cohort. PLoS One 2020; 15: e0232262. doi:10.1371/journal.pone.0232262
17 Rautio N, Varanka-Ruuska T, Vaaramo E, et al. Accumulated exposure to unemployment is related to impaired glucose metabolism in middle-aged men: A follow-up of the Northern Finland Birth Cohort 1966. Prim Care Diabetes 2017; 11: 365–72. doi:10.1016/j.pcd.2017.03.010
18 Sanchez Rico M, Plessz M, Airagnes G, et al. Distinct cardiovascular and cancer burdens associated with social position, work environment and unemployment: a cross-sectional and retrospective study in a large population-based French cohort. BMJ Open 2023; 13: e074835. doi:10.1136/bmjopen-2023-074835
19 Burgard SA, Lin KY. Bad jobs, bad health? how work and working conditions contribute to health disparities. Am Behav Sci 2013; 57: 1105–27. doi:10.1177/0002764213487347
20 Lynch J, Davey Smith G, Harper S, et al. Explaining the social gradient in coronary heart disease: comparing relative and absolute risk approaches. J Epidemiol Community Health 2006; 60: 436–41. doi:10.1136/jech.2005.041350
21 Havranek EP, Mujahid MS, Barr DA, et al. Social determinants of risk and outcomes for cardiovascular disease: a scientific statement from the american heart association. Circulation 2015; 132: 873–98. doi:10.1161/CIR.0000000000000228
22 Kaplan GA, Keil JE. Socioeconomic factors and cardiovascular disease: a review of the literature. Circulation 1993; 88: 1973–98. doi:10.1161/01.cir.88.4.1973
23 Kogevinas M, Pearce N, Susser M, et al. Social inequalities and cancer. International Agency for Research on Cancer Scientific Publications, 1997.
24 Vaccarella S, Lortet-Tieulent J, Saracci R, et al. Reducing social inequalities in cancer: setting priorities for research. International Agency for Research on Cancer Scientific Publications, 2019.
25 Beenackers MA, Kamphuis CBM, Giskes K, et al. Socioeconomic inequalities in occupational, leisure-time, and transport related physical activity among European adults: a systematic review. Int J Behav Nutr Phys Act 2012; 9: 116. doi:10.1186/1479-5868-9-116
26 Bloomfield K, Grittner U, Kramer S, et al. Social inequalities in alcohol consumption and alcohol-related problems in the study countries of the EU concerted action “Gender, Culture and Alcohol Problems: a Multi-national Study”. Alcohol Alcohol Suppl 2006; 41: i26–36. doi:10.1093/alcalc/agl073
27 Chang-Quan H, Zheng-Rong W, Yong-Hong L, et al. Education and risk for late life depression: a meta-analysis of published literature. Int J Psychiatry Med 2010; 40: 109–24. doi:10.2190/PM.40.1.i
28 Jehan S, Myers AK, Zizi F, et al. Sleep health disparity: the putative role of race, ethnicity and socioeconomic status. Sleep Med Disord 2018; 2: 127–33.
29 Kanjilal S, Gregg EW, Cheng YJ, et al. Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971-2002. Arch Intern Med 2006; 166: 2348–55. doi:10.1001/archinte.166.21.2348
30 McLaren L. Socioeconomic status and obesity. Epidemiol Rev 2007; 29: 29–48. doi:10.1093/epirev/mxm001
31 Winkleby MA, Jatulis DE, Frank E, et al. Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. Am J Public Health 1992; 82: 816–20. doi:10.2105/ajph.82.6.816
32 Sara JD, Prasad M, Eleid MF, et al. Association between work-related stress and coronary heart disease: a review of prospective studies through the job strain, effort-reward balance, and organizational justice models. J Am Heart Assoc 2018; 7: e008073. doi:10.1161/JAHA.117.008073
33 Marant Micallef C, Shield KD, Baldi I, et al. Occupational exposures and cancer: a review of agents and relative risk estimates. Occup Environ Med 2018; 75: 604–14. doi:10.1136/oemed-2017-104858
34 Fransson EI, Heikkilä K, Nyberg ST, et al. Job strain as a risk factor for leisure-time physical inactivity: an individual-participant meta-analysis of up to 170,000 men and women: the IPD-Work Consortium. Am J Epidemiol 2012; 176: 1078–89. doi:10.1093/aje/kws336
35 Heikkilä K, Nyberg ST, Fransson EI, et al. Job strain and alcohol intake: a collaborative meta-analysis of individual-participant data from 140,000 men and women. PLoS One 2012; 7: e40101. doi:10.1371/journal.pone.0040101
36 Heikkilä K, Nyberg ST, Fransson EI, et al. Job strain and tobacco smoking: an individual-participant data meta-analysis of 166,130 adults in 15 European studies. PLoS One 2012; 7: e35463. doi:10.1371/journal.pone.0035463
37 Landsbergis PA, Dobson M, Koutsouras G, et al. Job strain and ambulatory blood pressure: a meta-analysis and systematic review. Am J Public Health 2013; 103: e61–71. doi:10.2105/AJPH.2012.301153
38 Linton SJ, Kecklund G, Franklin KA, et al. The effect of the work environment on future sleep disturbances: a systematic review. Sleep Med Rev 2015; 23: 10–9. doi:10.1016/j.smrv.2014.10.010
39 Nyberg ST, Heikkilä K, Fransson EI, et al. Job strain in relation to body mass index: pooled analysis of 160 000 adults from 13 cohort studies. J Intern Med 2012; 272: 65–73. doi:10.1111/j.1365-2796.2011.02482.x
40 Nyberg ST, Fransson EI, Heikkilä K, et al. Job strain as a risk factor for type 2 diabetes: a pooled analysis of 124,808 men and women. Diabetes Care 2014; 37: 2268–75. doi:10.2337/dc13-2936
41 Theorell T, Hammarström A, Aronsson G, et al. A systematic review including meta-analysis of work environment and depressive symptoms. BMC Public Health 2015; 15: 738. doi:10.1186/s12889-015-1954-4
42 Zins M, Goldberg M, Team C. The French CONSTANCES population-based cohort: design, inclusion and follow-up. Eur J Epidemiol 2015; 30: 1317–28. doi:10.1007/s10654-015-0096-4
43 Guilloteau A, Binquet C, Bourredjem A, et al. Social deprivation among socio-economic contrasted french areas: Using item response theory analysis to assess differential item functioning of the EPICES questionnaire in stroke patients. PLoS One 2020; 15: e0230661. doi:10.1371/journal.pone.0230661
44 Hoertel N, Sanchez Rico M, Limosin F, et al. Respective mediating effects of social position and work environment on the incidence of common cardiovascular risk factors. J Am Heart Assoc 2022; 11: e021373. doi:10.1161/JAHA.121.021373
45 Siegrist J, Starke D, Chandola T, et al. The measurement of effort-reward imbalance at work: European comparisons. Soc Sci Med 2004; 58: 1483–99. doi:10.1016/S0277-9536(03)00351-4
46 Ruiz F, Goldberg M, Lemonnier S, et al. High quality standards for a large-scale prospective population-based observational cohort: Constances. BMC Public Health 2016; 16: 877. doi:10.1186/s12889-016-3439-5
47 Rubin DB. Multiple imputation for nonresponse in surveys. John Wiley & Sons, 2004.
48 Rijnhart JJM, Twisk JWR, Eekhout I, et al. Comparison of logistic-regression based methods for simple mediation analysis with a dichotomous outcome variable. BMC Med Res Methodol 2019; 19: 19. doi:10.1186/s12874-018-0654-z
49 Sobel ME. Asymptotic confidence intervals for indirect effects in structural equation models. Sociol Methodol 1982; 13: 290. doi:10.2307/270723
50 Hayes AJ. Introduction to mediation, moderation, and conditional process analysis. Guilford Press, 2022.
51 Arcaya M, Glymour MM, Christakis NA, et al. Individual and spousal unemployment as predictors of smoking and drinking behavior. Soc Sci Med 2014; 110: 89–95. doi:10.1016/j.socscimed.2014.03.034
52 Eliason M. Alcohol-related morbidity and mortality following involuntary job loss: evidence from Swedish register data. J Stud Alcohol Drugs 2014; 75: 35–46. doi:10.15288/jsad.2014.75.35
53 Golden SD, Perreira KM. Losing jobs and lighting up: employment experiences and smoking in the great recession. Soc Sci Med 2015; 138: 110–8. doi:10.1016/j.socscimed.2015.06.003
54 Popovici I, French MT. Does unemployment lead to greater alcohol consumption? Ind Relat (Berkeley) 2013; 52: 444–66. doi:10.1111/irel.12019
55 Airagnes G, Lemogne C, Meneton P, et al. Alcohol, tobacco and cannabis use are associated with job loss at follow-up: Findings from the CONSTANCES cohort. PLoS One 2019; 14: e0222361. doi:10.1371/journal.pone.0222361
56 Wilson SH, Walker GM. Unemployment and health: a review. Public Health 1993; 107: 153–62. doi:10.1016/s0033-3506(05)80436-6
57 Sumner RC, Gallagher S. Unemployment as a chronic stressor: A systematic review of cortisol studies. Psychol Health 2017; 32: 289–311. doi:10.1080/08870446.2016.1247841
58 Böckerman P, Ilmakunnas P. Unemployment and self-assessed health: evidence from panel data. Health Econ 2009; 18: 161–79. doi:10.1002/hec.1361
59 Johansson E, Böckerman P, Lundqvist A. Self-reported health versus biomarkers: does unemployment lead to worse health? Public Health (Fairfax) 2020; 179: 127–34. doi:10.1016/j.puhe.2019.10.005
60 Toledano MB, Smith RB, Brook JP, et al. How to establish and follow up a large prospective cohort study in the 21st century--lessons from UK COSMOS. PLoS ONE 2015; 10: e0131521. doi:10.1371/journal.pone.0131521
61 Khalatbari-Soltani S, Maccora J, Blyth FM, et al. Measuring education in the context of health inequalities. Int J Epidemiol 2022; 51: 701–8. doi:10.1093/ije/dyac058
62 Stuckler D, Basu S, Suhrcke M, et al. The public health effect of economic crises and alternative policy responses in Europe: an empirical analysis. Lancet 2009; 374: 315–23. doi:10.1016/S0140-6736(09)61124-7
63 Bambra C, Eikemo TA. Welfare state regimes, unemployment and health: a comparative study of the relationship between unemployment and self-reported health in 23 European countries. J Epidemiol Community Health 2009; 63: 92–8. doi:10.1136/jech.2008.077354
64 Weisz D, Gusmano MK, Rodwin VG, et al. Population health and the health system: a comparative analysis of avoidable mortality in three nations and their world cities. Eur J Public Health 2008; 18: 166–72. doi:10.1093/eurpub/ckm084
65 Bonamore G, Carmignani F, Colombo E. Addressing the unemployment-mortality conundrum: non-linearity is the answer. Soc Sci Med 2015; 126: 67–72. doi:10.1016/j.socscimed.2014.12.017
© 2025 Author(s) (or their employer(s)) 2025. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group. http://creativecommons.org/licenses/by-nc/4.0/ This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/ . Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Renuy, Adeline 2 ; Kab, Sofiane 2 ; Hoertel, Nicolas 1 ; Goldberg, Marcel 2
; Zins, Marie 2 ; Meneton, Pierre 3
1 DMU Psychiatrie et Addictologie, Hôpital Corentin-Celton, Université Paris Cité, AP-HP, Issy-les-Moulineaux, France; UMR_1266, INSERM, Issy-les-Moulineaux, France
2 UMS_011, INSERM, Villejuif, France; Université Paris Cité, Paris, France; Université Paris-Saclay, Gif-sur-Yvette, France; Université de Versailles Saint-Quentin-en-Yvelines, Versailles, France
3 UMR_1142 LIMICS, INSERM, Paris, France; Sorbonne Universite, Paris, France