Effort-reward imbalance at work assessed at midlife and prediabetes prevalence assessed 18 years later in a prospective cohort of white-collar workers

Abstract

Objectives

Evidence suggests that workers exposed to psychosocial stressors at work from the effort-reward imbalance (ERI) model are at increased risk for type 2 diabetes mellitus (T2DM). However, evidence about the effect of ERI on prediabetes is scarce. This study aimed to examine the association between effort-reward imbalance at work, glycated hemoglobin level and the prevalence of prediabetes in women and men from a prospective cohort study.

Methods

This study was conducted among 1354 white-collar workers followed for an average of 18 years. Effort-reward imbalance at work was measured in 1999 to 2001 using a validated instrument. Glycated hemoglobin was assessed at follow-up (2015 to 2018). Differences in mean glycated hemoglobin levels were estimated with linear models. Prediabetes prevalence ratios (PRs) were computed using robust Poisson regression models.

Results

In women, those exposed to effort-reward imbalance at work at baseline had a higher prevalence of prediabetes (PR = 1.60, 95% confidence interval: 1.02–2.49) at follow-up following adjustment for sociodemographic, lifestyle-related, clinical, and other occupational risk factors. There was no difference in mean glycated hemoglobin levels.

Conclusion

Among women, effort-reward imbalance at work at midlife was associated with the prevalence of prediabetes, at older age. Preventive workplace interventions aiming to reduce the prevalence of effort-reward imbalance at work may be effective to reduce the prevalence of prediabetes among women.

Introduction

Type 2 diabetes mellitus (T2DM) affects over 400 million individuals across the globe and is on the rise (World Health Organization 2016). T2DM increases the risk of renal, cardiovascular and neurocognitive pathologies, making it a leading cause of mortality worldwide (World Health Organization 2016). Prediabetes refers to high blood glucose that is not within diabetes range. Prediabetes is associated with a significant increase in morbidity and mortality whether individuals develop T2DM or not (Huang et al. 2014; Echouffo-Tcheugui et al. 2016; Cai et al. 2020). A prospective study also showed a linear relationship between glycated hemoglobin (HbA1c), an indicator of glucose metabolism, and health complications including cardiovascular diseases and all-cause mortality among nondiabetic adults (Selvin et al. 2010). Primary prevention strategies should therefore target modifiable risk factors associated with prediabetes to delay or prevent progression to T2DM, as well as to prevent adverse health outcomes associated with prediabetes itself.

Psychosocial stressors at work are modifiable and frequent risk factors that were shown to increase the risk of T2DM in prospective studies and meta-analyses (Nyberg et al. 2014; Pena-Gralle et al. 2022). The Siegrist’s effort-reward-imbalance (ERI) model is an internationally recognized model used to assess psychosocial stressors at work (Siegrist 1996). This model proposes that efforts should be rewarded in various ways: income, respect and esteem, and occupational status control. Workers are in a state of detrimental imbalance when high efforts are accompanied by low reward, and thus more susceptible to health problems (Siegrist 1996). One previous prospective study examined the association between ERI and prediabetes over 4 years, suggesting no independent association (De Souza Santos et al. 2020). In this previous study, the association between ERI and mean HbA1c level was not investigated. Further prospective studies with longer follow-up are required to determine whether ERI exposure is associated with early glucose metabolism imbalances. The objective of the present study was to examine the association between effort-reward imbalance at work assessed at midlife and subsequent HbA1c level and prediabetes prevalence, in a prospective cohort study of women and men followed for 18 years.

Methods

Study design and population

The study sample was derived from the PROspective Quebec (PROQ) cohort, which has been previously described (Trudel et al. 2018). Briefly, this cohort is composed of white-collar workers and was initiated in 1991 to 1993 (T1). A total of 9188 workers (49.9% women) were recruited from 19 public and semipublic organizations in Quebec City, Canada. They were followed up 8 and 24 years later (T2 and T3), with good participation rates: 75% at baseline, 90% at the 8-year follow-up, and 81% at the 24-year follow-up. This study was approved by the Centre Hospitalier Universitaire de Quebec’s Research Center ethical review board. All participants provided their informed consent.

At T3, a convenience sample of 2318 participants were targeted for blood tests including HbA1c level. Given logistical constraints, this sample was set up in 2017 and was composed of participants who had not yet been seen at T3 (2015 to 2018). Participants were either recently retired or still in the workforce. Of these, 17 died before the follow-up, 86 were lost to follow-up, 344 refused to participate, 367 participated by mail only, and 41 did not get adequate blood samples. Furthermore, 60 individuals who did not participate (45 refusals and 15 lost to follow-up) and 17 inactive workers at T2 were excluded from the analysis. Finally, 32 participants had missing data on ERI exposure at T2 and were further excluded. A total of 1354 participants (682 men and 672 women) were included in the analysis (Fig. 1).

Effort-reward imbalance at work

Effort-reward imbalance was measured at the second phase of data collection (T2; 1999 to 2001) using validated scales. Reward at work was measured with 9 original questions from the French version (Niedhammer et al. 2000) of the effort-reward imbalance scale (3 of 5 questions from the esteem subscale, the 4-item subscale of promotion prospects and salary, and the 2-item subscale of job security). The questions were answered in 2 steps. The respondents were first asked to indicate whether they agreed or disagreed that the question content described an experience typical of their work situation. If they agreed, they were then asked to indicate to what extent they felt distressed by the experience: 1 = very distressed, 2 = distressed, 3 = somewhat distressed, 4 = not at all distressed (scores were inverted for positive items). The 5-point answers were rescored on a 3-point scale, the total points ranging from 9 to 27, lower values indicating lower reward. Effort was measured using 9 items from the validated French version of the psychological demand scale of the Job Content Questionnaire (Brisson et al. 1998). The 4-point answers were rescored on a 3-point scale, the total points ranging from 9 to 27, higher values indicating higher effort. The psychometric qualities of this version have been demonstrated (Aboa-Eboule et al. 2011). In line with previous literature, the ERI score was calculated by dividing the effort score by the reward score. A ratio greater than 1 indicated an ERI exposure. Tertile based variables were created for each dimension taken separately.

HbA1c and prediabetes

Nonfasting blood samples were taken at T3 (2015 to 2018) by a trained nurse following a standardized protocol. Blood samples were analyzed at Biron Groupe Santé Inc., a certified laboratory. HbA1c was measured using the immunochemical method on an Integra platform from Roche diagnostics (coefficient of variation of 1.6%). The American Diabetes Association guidelines were followed, which define prediabetes as an HbA1c between 5.7% and 6.49% (American Diabetes Association 2010). Participants with a HbA1c under 5.7% were considered to have a normal glucose metabolism. Participants with a HbA1c ≥6.5% were considered prevalent T2DM cases. Additionally, participants with self-reported T2DM at T1, T2, or T3 were considered prevalent T2DM cases at T3 regardless of their HbA1c level. These participants (N = 79) were excluded from the HbA1c level analyses since they are likely to have a different HbA1c level than at time of diagnosis because of their treatment, which could diminish the strength of the association between ERI at work and HbA1c.

Covariates

The following risk factors were assessed at T2 except for family history of T2DM which was measured at T3 (2015 to 2018). Age was divided into tertiles: <38 years old, 38 to 41 years old, and 42 years old or above. Marital status was defined as living with a spouse (yes/no). Education was defined as the highest degree obtained: less than college, college, or university. Smoking status was categorized following the World Health Organization guidelines: nonsmoker, ex-smoker, and current smoker (Weitkunat et al. 2013). Physical activity was categorized as follows, based on participants’ answers to a validated question on the duration and frequency of their physical activity: sedentary (<1 session per week), insufficiently active (1 to 2 sessions per week), and active (≥3 sessions per week) (Gionet and Godin 1989). Alcohol consumption was dichotomized based on a validated question on alcohol consumption (Daveluy et al. 1994): nondrinker (less than 1 drink per week) and drinker (≥1 alcoholic beverage per week). Hypercholesterolemia and antihypertensive medication were assessed using a binary variable (yes/no). Weight and height were measured by trained employees using validated protocols (Daveluy et al. 1994). Body mass index (BMI) was obtained by dividing the weight in kilograms by the height in meters squared and was treated as a continuous variable. Blood pressure (BP) was measured following the American Heart Association protocol (Frohlich 1988). After 5 min of rest, 2 blood pressure measurements were taken 1 to 2 min apart and the average of the 2 measures was used in the analyses. BP was dichotomized according to the presence or absence of hypertension (systolic BP ≥140 mmHg or diastolic BP ≥90 mmHg). Long working hours were dichotomized based on a standard full-time employment schedule in Canada: <41 h per week and ≥ 41 h per week. Job strain was measured using 18-item from the Job Content Questionnaire (Karasek et al. 1998). Median were used as cutoffs to identify participants exposed to job strain, as defined by high psychological demands (≥ 24) and low job control (≤72) (Daveluy et al. 2000). Social support at work was measured using validated scales (Karasek et al. 1998). Six items were used to assess coworkers social support and 5 items were used to assess supervisor social support. Responses were summed to obtain a total score for coworkers and supervisor social support. Binary variables were created using the sample median split yielding high and low social support.

Statistical analysis

Crude and adjusted HbA1c mean differences (MDs) with 95% confidence intervals (CIs) were computed according to ERI exposure as well as its individual components (effort and reward) using linear regression models. The prevalence of prediabetes among all participants was modeled using generalized linear models with a log link and a Poisson working model (18). Sensitivity analyses were conducted excluding participants with T2DM (N = 84). A robust variance was used to account for the larger variance of the Poisson distribution compared to the binomial distribution. Results are presented with prevalence ratios (PRs) and their 95% CIs. Missing data on each variable was less than 1.5% except for family history of T2DM (7.6%) and hypertension (3.7%). Dummy indicators were used to prevent the exclusion of participants with missing data on these variables. According to prior literature, the adverse effect of psychosocial stressors at work on T2DM may be partially mediated by adoption of unhealthy lifestyle habits (Heikkilä et al. 2013). Therefore, sequential adjustments were used. However, given that estimates with and without lifestyle risk factors yielded similar results, only fully adjusted models are presented.

Prior studies found important sex differences in the association between psychosocial stressors at work and glucose metabolism imbalances (Xu et al. 2012; De Souza Santos et al. 2020). Therefore, separate analyses were performed for men and women. Statistical analyses were conducted using the statistical software package SAS, version 9.4. Two tailed statistical tests with a significance level of 0.05 were used.

Results

Table 1 presents baseline characteristics of our study sample, by sex. Compared to men, women were younger (38.8 years old, SD = 4.5 versus 40.0 years old, SD = 5.2). Women were less educated; 43.7% of them held a university diploma compared to 65.1% of men. Women were healthier (lower prevalence of overweight individuals, hypertension, and hypercholesterolemia) and had a lower prevalence of alcohol consumption. However, more women were current smokers (17.0%) than men (12.2%). Only 2.7% of women reported long working hours compared to 8.8% of men. Women were more frequently exposed to job strain. There was no difference in the prevalence of low social support at work and ERI exposure between men and women. Men reported high effort at work more frequently (44.1%) when compared to women (37.4%).

Table 2 presents mean HbA1c level and MD according to ERI, effort, and reward exposures, by sex. There was no association between ERI exposure and mean HbA1c level in either men or women. No association was observed when effort and reward were considered separately.

Table 3 presents the prevalence and PR of prediabetes according to ERI, effort and reward exposures, by sex. In women, the prevalence of prediabetes was higher among those exposed to ERI compared to unexposed women, after adjusting for potential confounders (PR = 1.60, 95% CI: 1.02 to 2.49). In men, the results did not suggest an association between ERI exposure and prediabetes prevalence. When considered separately, high effort at work was not associated with prediabetes, in either men or women. Among women, low reward was associated with an increased prevalence of prediabetes (PR = 1.69, 95% CI: 0.98 to 2.92). Sensitivity analyses excluding T2DM cases yielded slightly reduced estimates (Supplementary File).

Discussion

Summary of main findings

The aim of the present study was to evaluate the association between effort-reward imbalance at work, HbA1c level, and prediabetes prevalence in a prospective cohort study. Results showed that the prevalence of prediabetes was higher in women exposed to ERI at work, compared to unexposed women. This association was robust to adjustment for sociodemographic, anthropometric, and lifestyle risk factors. Low reward at work, considered separately, was also associated with an increased prevalence of prediabetes among women.

Explanatory hypotheses

Previous evidence about the effect of adverse psychosocial stressors at work on prediabetes is scarce. In the sole previous prospective study on this topic, job strain, as defined by high psychological demands and low job control, was associated with a 32% increase in prediabetes incidence among women (De Souza Santos et al. 2020). This risk was slightly increased among women who were simultaneously exposed to both job strain and ERI (51%). However, no association was observed when the adverse effect of ERI was examined separately. The cohort used for this previous study had a relatively low participation rate (30%) at baseline (Schmidt et al. 2015), which could have led to an underestimation of the true effect. Furthermore, the 4-year follow-up might not have been optimal to fully capture the adverse effect of ERI on prediabetes. Indeed, previous evidence reported that an induction period of several years can be required for psychosocial stressors at work to exert their effect on cardiometabolic outcomes (Trudel et al. 2021a). An inadequate period at risk can also lead to an underestimation of the effect. The present study suggests the presence of a long-term association between ERI exposure and the prevalence of prediabetes, among women. When efforts and rewards were considered separately, only low rewards were associated with the prevalence of prediabetes. This is consistent with the sole previous prospective study on this topic (De Souza Santos et al. 2020) and suggests that low rewards may be of particular importance. This result should be confirmed in future studies.

Psychosocial stressors at work can contribute to the development of glucose metabolism imbalances via a stress response mechanism related to the activation of the hypothalamic-pituitary-adrenal (HPA) axis and the sympathetic nervous system, leading, respectively, to an increase in glucocorticoids and catecholamines secretion. Glucocorticoids, which release glucose and lipids to meet increased energy needs, (Lupien et al. 2009; McEwen et al. 2015), are also responsible for appetite regulation, the accumulation of energy reserves and weight gain. Thus, chronic stimulation of the HPA axis may promote the development of endocrine abnormalities, insulin resistance, glucose intolerance and abdominal obesity that can lead to prediabetes and T2DM (Siddiqui et al. 2015). In addition, activation of the sympathetic nervous system and abdominal obesity are accompanied by a release of proinflammatory cytokines. Several observational studies have shown that T2DM is characterized by a chronic inflammatory state (Donath and Shoelson 2011). Thus, the stress response implicates several biological processes that, stimulated repeatedly, lead to disruption through interconnected systems (Hackett and Steptoe 2017).

The results of the present study are in line with previous literature, which reported a stronger and more consistent association between psychosocial stressors at work and T2DM among women (De Souza Santos et al. 2020; Pena-Gralle et al. 2022). In the systematic review and meta-analysis by Pena-Gralle et al., T2DM incidence was higher in workers exposed to ERI (RR = 1.24) and job strain (RR = 1.16). This review suggests that the adverse effect of job strain on T2DM may be more pronounced among women (RR = 1.35). However, given the small number of individual studies (n = 4), results for ERI were not stratified by sex. Our results suggest that ERI is associated with glucose metabolism imbalances at earlier stages of disease progression. Results also suggest that sex-specific associations between psychosocial stressors at work and glucose metabolism imbalances are also present during these earlier stages. There are several hypotheses to explain the potential sex-specific effect of psychosocial stressors at work. Previous evidence suggests that both work–family conflict and total workload (i.e. paid and unpaid work) are higher among women than their male counterparts (Griep et al. 2016). Moreover, evidence suggests that combining psychosocial stressors at work and high family responsibilities leads to higher blood pressure (Gilbert-Ouimet et al. 2017). Biological responses to chronic stressors might also differ across sexes. Previous studies showed that women exposed to such stressful situations had steeper psychoneuroendocrine activation (Rohleder et al. 2002), including higher cortisol hormone secretion than men. Further examinations of sex (biological) and gender (sociocultural) differences are however required.

On a population scale, a linear relationship has been observed between HbA1c and all-cause mortality, even with a small increase in mean HbA1c (Selvin et al. 2010). To the best of our knowledge, no previous study has examined the effect of ERI, assessed at midlife, on HbA1c at older age. Overall, our results did not suggest an association between ERI at work and HbA1c levels, assessed subsequently. However, previous cross-sectional studies have reported higher levels among exposed workers (Xu et al. 2012; Li et al. 2013; Jarczok et al. 2016). These conflicting results highlight the need for further prospective studies to clarify this relationship.

Study strength and limitations

The present study has important strengths. First, the associations between ERI at work, HbA1c level and prediabetes prevalence were examined in a long-term study with approximatively 50% women. Changes in exposure may have occurred during the long follow-up period (18 years). However, 80% of our workers held the same job title between the first 2 follow-ups (T1 and T2), which suggests exposure stability. Among retired participants at T3 (N = 346), the median retirement time before T3 was 1.6 years, suggesting that these participants remained actively employed for most of the study period. They were included in the present study to minimize the possibility for a healthy worker survivor effect (Massamba et al. 2019). Exposure misclassification may be present. However, since prediabetes is an asymptomatic condition, it would most likely be nondifferential regarding the outcome, which would result in an underestimation of the association. Analyses were performed separately for men and women allowing us to examine sex-specific associations between ERI and prediabetes. ERI at work and HbA1c were assessed using validated instruments and protocols minimizing the possibility for information bias. Finally, a large number of potential confounders were accounted for, which supports the robustness of the association between ERI and prediabetes.

The study also had limitations. First, the present study relied on a convenience sample. A comparison analysis showed that the study sample was younger and healthier, on average, than the rest of the cohort (not shown). Therefore, the possibility of a healthy worker bias cannot be ruled out. This type of bias generally leads to an underestimation of the true effect. Further exclusions were required among the total sample for the HbA1c analysis. We have compared the total sample (N = 2318) with the final sample used in the present study (N = 1354). There was no difference regarding the prevalence of exposure to ERI (P =0.62), suggesting that these exclusions have unlikely lead to selection bias. Moreover, the longitudinal progression of HbA1c levels and prediabetes incidence could not be assessed, limiting the possibility of drawing causal inferences. However, previous studies reported an adverse effect of ERI on T2DM incidence, and their results do not suggest that reverse causation could explain the observed associations. Furthermore, HbA1c was assessed 18 years following ERI exposure, which makes reverse causality unlikely. In the present study, diet was assessed at follow-up, using the alternative healthy eating index among participants who agreed to answer this questionnaire (71.6%) (Trudel et al. 2018). In a sensitivity analysis, we have restricted the study sample to those with available information on diet. Estimates with and without adjustment for diet were identical, minimizing the possibility for residual confounding (not shown). However, this result should be interpreted with caution since the temporal precedence of diet on prediabetes assessment was not respected. Lastly, the cohort involves white-collar workers only, which limits generalization of the findings to workers with similar conditions. However, participants in the present study held a diversity of white-collar occupations such as office workers, technicians, professionals, and managers. A majority of workers in Organization for Economic Co-operation and Development countries hold white-collar occupations, supporting generalization to a considerable section of the workforce (Chen and Mehdi 2019). In the present study, the proportion of participants with college education was high (44% of women and 65% of men). However, this proportion is comparable to that observed among the working-age population in Canada, favoring generalization (Statistics Canada 2021).

Implication of the results and future directions

The results of the present study suggest that women exposed to ERI have a higher prevalence of prediabetes. Prediabetes is on the continuum between normal glucose metabolism and T2DM. Since it is a reversible condition (Morris et al. 2013), early interventions have the potential to delay and prevent T2DM (Donovan et al. 2013). In recent years, nonpharmacological treatment options are establishing themselves as the standard of care in prediabetes as well as mild T2DM cases (Galaviz et al. 2022). Psychosocial stressors at work are promising targets to intensify nonpharmacological and population-based preventive interventions. Results from the present study suggest that workplace interventions tackling ERI at work may reduce the prevalence of prediabetes among women. Previous studies have shown that psychosocial stressors at work can be reduced through organizational interventions (Letellier et al. 2018) and can provide benefits on workers’ cardiovascular health (Trudel et al. 2021b). However, such evidence is scarce and further intervention studies are needed to examine whether reducing psychosocial stressors at work could lead to beneficial effects on prediabetes and T2DM. At the clinical level, findings from the present study suggest that screening for psychosocial stressors at work may be relevant to identify women at risk for prediabetes. It is also noteworthy that, among women, the magnitude of the observed association between ERI and prediabetes prevalence among women (PR = 1.60) is comparable to that of being overweight (fully adjusted PR, BMI 25 to 29.9 kg/m² versus <25 kg/m² = 1.35), a well-acknowledged T2DM risk factor. Findings therefore highlight a need for improved clinical awareness about psychosocial stressors at work and their adverse effect on worker’s health.

Conclusion

The present study, conducted among white-collar workers followed for 18 years, suggests that women exposed to ERI at work at midlife are more likely to have prediabetes at older age, when compared to unexposed women. Workplace interventions aiming at reducing the prevalence of ERI at work may be considered as a promising avenue to reduce the burden of prediabetes, especially among women.

Supplementary material

Supplementary material is available at Annals of Work Exposures and Health online.

Acknowledgments

This study would not have been possible without the contributions many individuals and organizations. We would like to thank Caty Blanchette and Myrto Mondor for their statistical support throughout this study. Their insights and expertise were instrumental in shaping the direction of this project. This work was supported the Canadian Institutes of Health Research (grant number 57,750). CR was supported by training awards from the Canadian Institutes of Health Research (CIHR), by the Fond de Recherche du Quebec en Santé (FRQS) and by VITAM, a research center in sustainable health within the Centre Intégré Universitaire de Santé et Services Sociaux Capitale-Nationale.

Conflict of interest statement

The author declares that they have no conflict of interest.

Funding

Funding for this project was provided by CIHR, FRQS and VITAM. The authors declare no conflict of interest relating to the material presented in this article. Its contents, including any opinions and/or conclusions expressed, are solely those of the authors.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Bibliography