In this study, we first described the employment trajectories from age 16 to 65 of a cohort of Europeans born between 1914 and 1945, and second, we assessed whether these employment trajectories could predict the onset of cancer at any time in the life course. First, using sequence analysis, we grouped the employment histories from the ages of 16 to 65 into eight and two patterned trajectories for women and men, respectively. For women, we named the eight trajectories “Mainly full-time”, “Mainly home/family”, “Mainly full-time to home/family”, “Mainly self-employment”, “Full-time or home/family to part time”, “Home/family to full-time”, “Mainly unemployment” and “Other”. In terms of the proportions of the sample, none of the eight trajectories was in the majority. In this “prediction exercise” 70, and thanks to sequence analysis, we found that, compared to women who spent their lives mainly caring for the home/family, four employment trajectories were associated with a higher risk of cancer over the life course: women who worked predominantly full-time, women who worked full-time then transitioned to taking care of home and the family, women who either worked full-time or took care of the home/the family and then transitioned to working part time, and women with unusual employment trajectories (“Other”), i.e. trajectories characterised by illness or disability, voluntary work, and travelling. When we looked at breast cancer, we also found associations with the same four employment trajectories. For men, we named the two solutions “Mainly full-time” and “Mainly self-employment”, with the first one representing the largest, majority group. In sum, this study has shown that the onset of cancer over the life course can be predicted by classifying women and men into typical groups of lifetime employment trajectory, and that these associations were independent of factors known to be associated with cancer (smoking, physical activity, number of chronic conditions, and body mass index).
The finding of eight employment trajectories for women suggested that women’s employment trajectories were more heterogeneous, or less “socially standardised” by the culture and local contexts, than men’s, who had only two trajectories, with one predominant type being full-time work throughout their working lives. The heterogeneity of women’s employment trajectories and the homogeneity of men’s employment trajectories are consistent with previous studies in Germany56, England71 and Switzerland54, and, at least for women, with previous studies using the same dataset59,72. However, using the same dataset, another study found two patterned employment trajectories for women, but using multi-channel sequence analysis73. The finding of heterogeneity in women’s employment trajectories was to be expected for these cohorts as women’s life course trajectories, compared with men’s, are socially normed towards multiple social roles, such as parenthood, family responsibilities, including contribution to breadwinning, and – to a lesser extent – professional career74–77.
Individuals’ employment trajectories are influenced by the policies of the country in which they live78,79. Indeed, policies can affect employment opportunities, job security, wages and career paths. For example, labour market policies implemented by governments (such as employment protection legislation, labour market flexibility measures, reintegration measures), social security and welfare policies (such as unemployment benefits, income support programmes), family policies (childbearing) and health policies (health insurance, disability insurance) can have direct or indirect effects on employment trajectories. Such effects were not investigated in this study but it could be assumed that country policies might moderate some of the associations observed in this study. Such research may prove challenging as it would require historical research on labour market, social security, family and health policies in each of the countries included for the period covering the participants’ employment trajectories.
The diversity of employment trajectories was associated with higher cancer risk over the life course. In our study, we identified several groups of lifetime employment trajectories in which work in general (either full-time or part-time) was associated with cancer in women. We found that a lifetime trajectory of working full-time as an employee was associated with a higher risk of cancer compared with a lifetime trajectory of caring for home and family. In addition, working full-time as an employee in the first half of the working life and then transitioning into the family sphere was also associated with a higher risk of cancer. There are three possible interpretations of these findings. First, given that the association with cancer reflects breast cancer risk, it is possible that women who entered the labour market at some point in their life course were more likely to be informed and screened for breast cancer, and therefore had more cancer diagnoses. A European cross-sectional study suggested that being inactive was associated with a lower likelihood of being up to date with cancer screening80. Second, if full-time employment does increase women’s cancer risk, it is not per se, but because employment may be an indirect marker of the poor quality of women’s jobs and of occupational stress. Indeed, in the second half of the twentieth century women were known to have a higher proportion of low-skilled jobs81. Women from these cohorts increased their participation in the labour market but were more likely to be employed in manual occupations and in low-value jobs82. It is also possible that employed women are more exposed to occupational stress than women who have been self-employed throughout their working lives. Work can expose individuals to social and psychological stressors83, and stress has a biopsychosocial pathway to cancer development30,84–86. Gender insensitive working environments may affect mental health87, which may also predict cancer incidence through depression and psychological distress85. Exposure to more or less long-term poor working conditions88 may dysregulate physiological systems (e.g., immunity89,90), or the general physiological wear and tear19 and DNA methylation91,92, and predict the onset of cancer93,94. It is likely that the employment status – the indicator on which our sequence analysis was based – is an imperfect marker of the resources (income, job-related skills, social network at work) and status (social status of occupations) that women derive from their occupations and from the types of responsibilities they have at work, and ultimately of the levels of stress to which they are exposed95,96. Thus, without lifetime data on occupation and on psychosocial stress at work (e.g., effort-reward balance), the association between employment exposure and increased cancer risk should be treated with great caution.
Among men, we observed an association between a lifetime trajectory of self-employment and cancer risk. Compared with men who worked full-time as employees throughout their working lives, those who were self-employed had a lower risk of cancer. There is little research on the relationship between self-employment and better health for men. Self-employed people may generally have better health compared to employees for the following reasons: firstly, self-employed men may have more control over their work environment and workload, and therefore be better able to manage the stress levels; secondly, self-employed men may have more flexibility in their working hours, which may allow them to engage in healthy behaviours such as exercise and healthy eating, or cancer screening. On the other hand, we cannot ignore that the self-employed are exposed to the risks and responsibilities that come with running a business, which can lead to stress97,98, thus limiting the proposed explanation.
This study has limitations. First, we cannot exclude that reporting bias on the outcome may be significant, for two reasons: (a) cancer diagnosis was self-reported rather than from cancer registries. Self-reported cancer has an overall false-negative rate of 39.2%, with wide variation by cancer site99; (b) respondents in this cohort were older, which may be associated with more frequent false-positive reporting100. Second, survivor bias is inherent in the design of ageing studies such as SHARE, whose baseline included participants aged 50 years and over. However, the survivor bias may be limited by the low probability of dying from cancer before the age of 50101. Third, as shown in the flowchart of respondent’s inclusion, the analytical sample (6,809 + 5,716 = 12,525) was a reduction compared to the original dataset (n = 139,760), signifying that selection bias may affect the results. Fourth, due to the lack of data in SHARE, our study does not take into account other life course factors such as intrauterine life, perinatal characteristics, child growth, and hormonal factors, which are associated with the incidence of several cancers5,102. Fifth, like all longitudinal studies of ageing, SHARE is characterised by attrition at follow-up103, and we have partially controlled for this in the analysis. Sixth, by controlling for cancer risk factors such as smoking104, sedentary behaviour105 and BMI106,107 measured at baseline, we confirmed that employment trajectories were independent of these risk factors, however we do not have information about the individual life course history of these factors, which may interact with employment over the working life. Seventh, the eight trajectories are not a definitive solution for categorising women’s trajectories because the clustering methods impose numerous choices, such as the clustering algorithm and the distance metric, which could influence the results. The eight trajectories are the result of a data mining construction that seeks the best solution for grouping the great heterogeneity of respondents’ trajectories (each individual sequence), taking into account the importance of the change (its timing and its order in the sequence) as well as the duration of a state (no change) in the employment trajectory. Eighth, the decisions made by SHARE about the coding of the employment status in 8 categories are critical in sequence analysis, as it will deeply influence the structure of the sequences. Ninth, the use of the life calendar allowed the reconstruction of the entire life course, retrospectively, on a yearly basis, between the ages of 16 and 65. Such a tool is useful for collecting comprehensive information on the life course of respondents. However, life calendars are subject to mnemonic and cognitive biases, but these are limited by the instrument’s ability to enhance respondent’s memory and correct for inconsistencies compared to a list of standardised questions108. Tenth, this study was based on participants who spent most of their lives between the ages of 16 and 65 in the 20th century, a context marked by gender inequalities in access to education and the labour market, which is assumed to influence the ageing of women in good health109. Due to changes in the social stratification and in social norms110, in particular progress in socioeconomic equality between women and men81,111, we expect that our findings may not be applicable to cohorts who will spend their life course in the 21st century. Eleventh, this study examined employment trajectories cross-nationally by merging respondents from 14 countries. Our approach should not mask large cross-national differences in employment trajectories78,79. Twelfth, the timing of exposure (employment trajectories from 16 to 65 years) overlapped with the onset of the outcome (cancer diagnoses from 0 to 99 years), with about three in ten female cancer diagnoses occurring before the age of 65 years.
Lifetime employment trajectories of European women are diverse and some of these trajectories were associated with a higher risk of self-reported cancer over the life course if women participated in the labour market compared with women who stayed at home. Lifetime employment trajectories of European men were less diverse, and being self-employed throughout working life was associated with a lower risk of cancer.
We believe that examining the full history of the participant’s life course using life calendar data and identifying patterned trajectories in these data using sequence analysis, has a promising potential for studies in life course epidemiology, although it has been rather rare to date in the field of epidemiology and health studies59,71,112. Although the lifetime employment trajectories approach looks promising, our association study remained a predictive exercise70. More research is needed to better understand these associations and to establish whether they are causal. In particular, factors such as health-related behaviours and occupational health risks, and their timing over the life course, should be considered in further studies.