Introduction

The metabolic syndrome (MS) is a cluster of several risk factors (i.e., obesity, elevated blood pressure, insulin resistance, and dyslipidemia) predisposing individuals to the development of type 2 diabetes and cardiovascular diseases (CVD) [1]. MS is a common disorder among obese patients, with recent prevalence estimates ranging from 10 up to 26% [2]; however, pathophysiology and risk factors associated with its development are still poorly understood.

The bio-psycho-environmental model suggests that chronic stressors lead to an increased risk of developing several chronic diseases such as MS, obesity and CVDs [3,4,5]. Previous studies found that psychological work stress is a common gender-related risk factor involved in the pathogenesis of diabetes [6] and has a dose–response effect in the onset of coronary artery disease, which in turn might be expressed by the presence of MS itself [7].

Psychological problems seem also to play an important role in the onset of MS. For example, major depressive disorders [8] or binge eating disorders (BED) are often comorbid with obesity [9], which is considered one of the main components of MS. Hays [10] found that any condition producing stress, such as job strain, could induce changes in eating behaviors and cause BED, with consequent obesity. The literature also reports that women experiencing working burnout have significantly higher scores in emotional or uncontrolled eating, and have “hindered ability to make changes in their eating behaviour” [11].

Therefore, different job categories, some psychopathological factors (i.e., depressive levels and binge eating behaviors) of obese patients and perceived work-related stress could contribute to the development of MS. However, few studies investigated the associations between these factors, or examined the gender-related differences in the clinical presentation of MS. In addition, previous studies analysed results and drew conclusions based mostly on traditional statistical techniques, which relies on the assumptions of linear relationships between variables. Thus, they have a limited statistical power in presence of non-linear and complex associations, as typically observed in biological systems. In recent years, these limits have been overcome by an innovative data mining analysis, called Auto Contractive Map (AutoCM). Based on artificial neural network (ANN) architecture, AutoCM allows to discover hidden trends and associations among variables and reconstructs the imprecise, non-linear and simultaneous pathways underlying a complex set of data (e.g., risk factors of a disease), using a fuzzy clustering approach that relies on semantic connectivity maps [12].

Therefore, the main aim of this study was to investigate the complex associations between MS and gender, obesity, and various psychosocial risk factors such as job demand, job attitude, social support, physical activity during work, eating behaviors and levels of depressive symptoms, using AutoCM. Principal component analyses were applied to the same dataset as a linear benchmarking model. Revealing the explicit or hidden associations between psychosocial factors and the presence\absence of MS may improve clinical decision-making and help physicians in planning tailored interventional programs for this complex, but poorly understood, disease.

Materials and methods

Participants

Consecutive overweight patients attending an annual routine health check-up at an occupational medicine clinic (Fondazione IRCCS Policlinico Ospedale Maggiore, Milano) in northern Italy were enrolled in this observational study. Patients were recruited from January 1st 2011 until the end of the same year. All participants provided written informed consent, in accordance with the Declaration of Helsinki and the Good Clinical Practice Guidelines. The protocol was approved by the local Ethics Committees (study registration number: 1370).

Measures

A semi-structured interview about general health, dietary intake habits, smoking, physical activity, chronic diseases and specific therapy was performed by trained physicians at admission. Data on occupational class were collected according to the Ateco classification [13]: in particular, level of physical activity during work was classified as light, moderate and heavy occupational physical activity (OPA) in accordance with the compendium of physical activities; then measured in metabolic equivalents [14]. Anthropometric parameters were measured, while BMI was computed according to the WHO classification [15]. Blood samples for plasma glucose, triglycerides, and HDL cholesterol were drawn in the morning after an overnight fast and evaluated by routine standard hospital methods and procedures (Modular D; Roche, Milan, Italy). Metabolic syndrome was diagnosed according to the 2005 US National Cholesterol Education Program-Adult Treatment Panel III criteria (NCEP ATP III) [16].

Binge eating behaviors were evaluated using the Binge Eating Scale (BES), a self-report instrument composed by a total of 16 items [17]. BES evaluates the emotional, cognitive and behavioral characteristics of binge eating. Total score ranges from 0 to 46, with higher score indicating more severe binge eating problems.

Depressive symptoms were measured with the Beck Depression Inventory II (BDI-II), a self-report scale composed by 21 items [18]. Total score ranges from 0 to 63, with higher score reflecting more severe depressive symptoms.

To measure social and psychological characteristics of jobs, we administered the Job Content Questionnaire (JCQ), a self-report 49-item questionnaire that includes three main subscales: decision latitude (DL), psychological demand (JD) and social support (SS) at work [19, 20]. According to the underlying theory, the health-related effects of job strain are due to over commitment at work, range of decision-making freedom and absence of social support.

For each of these three scales, median values were used as a cut-off limit (low/high).

Statistical analyses

Descriptive statistics were expressed as mean and standard deviation for continuous variables, or as frequency counts and percentages for discrete data. As a second step, continuous variables were transformed (e.g., dichotomized) for ease of interpretation during all analyses. Thus, independent variables were sex (men or women), age (< 45 or > 45 years), BMI (< 30, 31–35, > 35 kg/m2), work category (Industry/Sellers, Administration, Education, Healthcare and Services), OPA (light or moderate), high levels of binge eating (BES scores > 17) and depressive symptoms (BDI scores > 16), and JCQ dimensional scores (social support [SS], < 25 or > 25; range of decision-making freedom available to the worker [DL], < 63 or > 63; demands of the work situation [JD], < 38 or > 38).

Linear correlations between independent variables and presence of MS were computed using Spearman’s correlation coefficients. In addition, we run principal component analysis (PCA) on the same variables separately for both genders. PCA was performed using MatLab 7.1 (Mathworks, Natick, MA). All statistical tests were two-sided; a p value ≤ 0.05 was considered significant.

Finally, non-linear and simultaneous pathways between traditional and psychosocial risk factors leading to MS were investigated—separately for both genders—through a new data-mining technique, called Auto Contractive Map (AutoCM; Semeion Research Center). AutoCM is a mapping method capable of computing the multi-dimensional association of strength of each variable with all other variables in a dataset, using a mathematical approach based on ANN [12, 21, 22]. This method is capable of computing and graphing a semantic connectivity map which (1) preserves nonlinear associations among variables, (2) captures elusive connection schemes among clusters, and (3) highlights complex similarities among variables. This representation allows a visual mapping of the complex web of connection schemes among variables, simplifying the detection of the variables playing a key role in the graph (i.e., hubs). The system provides also a quantification of the “strength” of links among variables (nodes of the graph) by a numerical coefficient, called link strength, ranging from 0 (minimum strength) to 1 (maximum strength). The value superimposed to the link is proportional to the strength of the link, and can be read as the probability of transition from any state variable to anyone else [12]. An in-depth description of AutoCM is provided in Supplementary Materials.

Results

We obtained data from 210 obese workers of both sexes (67.6% women); Table 1 shows demographic and clinical data related to MS. The syndrome was diagnosed in 54.4% of men and in 33.1% of women.

Table 1 Demographic, clinical and metabolic characteristics used for MetS definition of women (N = 142), men (N = 68) and total sample (N = 210)
Full size table

Table 2 shows the population’s job characteristics, as well as psychological tests’ score (BES and BDI) and JCQ scores for the total sample, and separately for both genders. The percentages of light/moderate occupational physical activities were similar in both samples (men and women), while frequencies between mental and physical categories were different.

Table 2 Job characteristics of the sample (n = 210) divided by sex, categories of job and intensity of occupational physical activity during job task
Full size table

The current smokers, uniformly distributed between men and women, with different degrees of obesity (data not shown) are only 20%. Leisure physical activity was not reckoned as being less than 1% of subjects with more than 2 h of physical activity in a week.

Spearman’s correlation coefficients between independent variables and presence of MS showed statistically significant positive associations (i.e., increased risk of MS) in men with high levels of decision-making freedom at work (DL > 63), and an industry/selling profession. On the contrary, negative associations (i.e., decreased risk of MS) were found in women with BMI < 30, DL < 63 and administrative job. According to sample size, r values of > 0.12 or < 0.12 were statistically significant. A figure with correlations can be found in Supplementary Materials.

Principal component analyses were performed on the dataset as a linear benchmark for the AutoCM method and explained almost 50% of the variance. Several strong gender-related differences emerged. As regards women, we observed a clustering of demographic\psychological variables (e.g., first cluster: Light OPA, BMI < 30 and working in Administration. Second cluster DL > 63, SS < 25, JD < 38). Interestingly, both presence or absence of MS were poorly associated with other variables. As regards men, a similar clustering of psychological variables was shown (e.g., first cluster DL > 63, SS < 25, JD < 38.). However, among men the presence of MS was related to overweight (BMI > 35), working in the Administration sector, Light OPA and age > 45. On the contrary, the absence of MS was clustered together with the presence of high levels of depressive and binge eating symptoms, working in Services or in Healthcare sector, and a BMI between 31 and 35. In conclusion, among men and women, the presence/absence of MS was not strongly associated with any other variable, due to the limitations of the PCA. On the whole, these mixed results were difficult to interpret, because of the inability of this statistical method to find relevant non-linear associations between clusters of variables. The scores plot of PCA can be found in Supplementary Materials.

Similarly, results of ANN models evidenced strong gender-related differences. The connectivity map in women (Fig. 1), showed that the presence of metabolic syndrome was strongly associated with job-related dimensions, such as (1) low levels of social support (SS < 25), (2) lower levels of demanding working situations (JD < 38), and (3) with high levels of decision-making freedom at work (DL > 63). No other variables were directly associated to the syndrome. A central hub in this specific map was the absence of MS, which was related to a Light OPA, a BMI below 30, age < 46 years, and low levels of decision-making freedom at work (DL < 63). Finally, two other important and interconnected hubs were respectively high levels of demanding working situations (JD > 38) and moderate OPA. JD > 38, indeed, was connected with older age (> 45 years), DL < 63, mild obesity (BMI 31–35), higher levels of psychopathological symptoms (depression BDI > 16; binge eating BES > 27). On the contrary, moderate OPA was associated with overweight, and working in Healthcare or in the Industry sector.

Fig. 1
figure 1

Women’s connectivity map (performed by AutoCM) clarifying the clusters and relationships among variables. The number highlighted in red between each node of the graph (i.e., variables) is the “link strength”. For a detailed description of each variable, see section “Statistical analyses”

Full size image

The connectivity map in men (Fig. 2) evidenced noticeably different results. There were two main and interconnected hubs, the first being the presence of MS, and the second the presence of moderate OPA. In the former case, MS was associated with (1) overweight (BMI > 45), (2) age higher than 45 years, and with (3) the same specific cluster of job-related dimensions found in women (DL > 63, JD < 38, and SS < 25). In the latter case, Moderate OPA was connected—in addition to the presence of MS—also to a lower age (< 45 years), mild obesity (BMI 31–35), working in various areas (Industry and sellers, Health Care, Services), and high levels of social support at work (SS > 25). Interestingly, the absence of MS was connected only to a BMI < 30, and to low levels of demanding working situations (JD < 38).

Fig. 2
figure 2

Men’s connectivity map (performed by AutoCM) clarifying the clusters and the relationships among variables. The number highlighted in red between each node of the graph (i.e., variables) is the “link strength”. For a detailed description of each variable, see section “Statistical analyses”

Full size image

Discussion

The present study evidenced strong gender-related differences in psychosocial risk factors for MS. Demographic and correlational data (see Table 1 and Supplementary materials) showed that MS was more common among (1) men, (2) those having high levels of decision-making freedom at work, and (3) those working in specific sectors (i.e., industry/selling professions). Interestingly, PCA analyses showed that the syndrome was poorly correlated with other variables among women. However, the clearest picture of the psychosocial risk factors was shown by semantic connectivity maps (performed with AutoCM; see Figs. 1 and 2). Among women, the presence of this syndrome was strongly connected only with low levels of social support at work (which mediated other work-related variables), while among men it was connected with moderate occupational physical activity (OPA), overweight, older age and high levels of decision-making freedom at work.

The ANNs and other machine learning algorithms were already used in previous studies to identify predictive factors of various chronic diseases [23,24,25]. In particular, Hirose et al. [23] investigated the 6-year incidence of MS using ANNs, identifying BMI, age, diastolic blood pressure, HDL-cholesterol, LDL-cholesterol and HOMA-IR as important predictors, and suggesting that these variables are non-linearly related to the outcome. Valavanis et al. [25] evidenced that a higher BMI is an independent risk factor for CVD among obese patients. Finally, Ivanovic et al. [24] adopted an ANN model to predict with high accuracy the diagnosis of MS using gender, age, BMI waist-to-height ratio, systolic and diastolic blood pressures as independent variables.

To the best of our knowledge, the present study was among the few which have considered altogether work-related factors [26] and psychopathological variables such as binge eating [27] or depressive symptoms [28] in the pathogenesis of MS. Our results showed that the above mentioned variables are associated to the development of MS in both sexes, although with different mechanisms.

Interestingly, in recent years, the MS was increasingly diagnosed among men rather than women, probably due to metabolic gender-related differences such as undiagnosed diabetes, high-lipid levels and visceral obesity [29,30,31,32].

Working environment represents a largely underestimated and understudied risk factor in the development of MS [33]. Over the last few decades, the job market was characterized by an overall reduction in physically demanding works, while psychological job demands (i.e., excessive workloads, time pressures, poor labor-management relations, ambiguous roles and workplace insecurity) increased over the years [5]. A large meta-analysis evidenced that the risk of coronary artery disease was highest among participants with job strain and unhealthy lifestyle, while those with job strain and healthy lifestyle had half the risk of developing a cardiovascular disease [34].

In our study, we evaluated job strain using the so-called Karasek model, which hypothesizes that a worker’s health may be negatively associated with job demands and positively associated with control and social support at work [35]. In the literature, socioeconomic disparities are associated with metabolic syndrome in white, black, and Mexican–American women while the association is less strong in men [33]. In our study, no data are available for socio-economic status (SES). All workers who referred to our service were employed with permanent contracts (Nurses, auxiliaries, etc.). Therefore, in our sample, we did not reckon SES as a disturbing factor.

The model postulates that men experience increased levels of stress when they face higher quantitative demands at work, and that both job control (e.g., executive positions) and social support lessen their negative effect. On the contrary, women experience higher levels of stress because of the emotional and intellectual aspects of their job demands. Noticeably, Gadinger et al. [36] reported a moderating effect of gender on the health-related effects of job strain among managers, with women managers experiencing “more psychosomatic complaints working in high demand, low control, low support settings than their male colleagues”. In line with these data, Fig. 2 shows that men with high levels of DL, and low levels of both JD and SS, evaluated through the JCQ, have an increased risk of MS. Therefore, we hypothesize that these individuals feel lonely and unsupported against a challenging “enemy” : thus, they likely activate a “fight or flight” response, which in turn leads to higher levels of work-related stress, maladaptive psychological reactions and unhealthy habits (e.g., eating junk foods, chronic sleep deprivation). On the contrary, women probably experience a more relaxed working environment, being themselves supported by co-workers (as evidenced by the higher scores in SS); however, they experience increased levels of psychological demand at work, which could in turn induce a different type of job-related stress and affect their psychological balance, self-esteem and self-efficacy, leading to higher levels of depressive or binge eating symptomatology. Finally, among women, the relationship between moderate OPA and high levels of psychological demands at work probably mediates the increases in perceived stress and binge eating symptomatology, whereas in men a moderate OPA seem associated with the onset of MS itself. Therefore, we hypothesize that younger men with high OPA have an increased intake of caloric food and a higher risk of developing obesity, even if binge eating symptomatology is absent. Our results are in accordance with previous studies, which reported a positive association between physically heavy work and CVD risk factors [37]. Recent literature showed also that the link between MS and risk of death from CVD is stronger in sedentary and/or physically heavy works than in those that imply walking or less strenuous occupational activities [38].

Our findings evidenced also a direct relationship between job strain and depressive or binge eating symptomatology among women. Binge eating disorder is a common mental disease that has a higher incidence among women (F:M ratio of 1.3–3 to 1) [39] and seems to independently increase the likelihood of developing specific components of the metabolic syndrome, such as obesity [40]. In addition, the risk of developing MS is higher among women with depression [41, 42]. Interestingly, even if the incidence of anxious and depressive symptoms linearly increases with an increase in JD, higher levels of SS at work have a beneficial (i.e., protective) effect among women [43, 44]. Therefore, our data support the hypothesis that low levels of SS at work might contribute to the development of MS in women with psychopathological symptoms (i.e., binge eating behaviors and depressive symptoms).

Lifestyle modifications such as dietary restriction, increased physical activity, and smoking cessation appear to decrease the risk of developing CVDs among patients with a diagnosis of MS [45].

Rapid identification of patients and the planning of long-term interventions are imperative to manage this multifaceted disorder; however, our findings also suggest that all the above-mentioned risk factors should be considered, so as to (1) develop new health-related lifestyle programs, and (2) plan individual and environment-related changes in work contexts thus helping clinicians to limit or prevent the further spread of this syndrome.

Our study has some limitations. First, the cross-sectional nature of the data limited the possibility to draw causal inferences; therefore, future studies should assess, for example, if changes in job strain or other psychological variables influence at least some of the components of MS. Second, in our sample, none of the participants had jobs involving a heavy physical activity, which could probably induce an independent gender effect on job strain. Furthermore, all participants were Caucasian, thus limiting the generalizability of our results to other ethnicities. Finally, we only used one measure of work-related stress, JCQ, due to the fact that we were mainly interested in the dimension of SS. However, other instruments, such as effort/reward imbalance model or the justice model, might have provided more information on psychological factors associated with working environments.

In conclusion, our data suggest that other uncommon risk factors (i.e., occupational physical activity, job strain and its component, psychological profile) might be involved in the development of metabolic syndrome.