The AWHS (Aragon Workers’ Health Study) is a prospective, longitudinal, epidemiological cohort study based on the data obtained in the annual physical examinations of workers at the automobile assembly plant of General Motors España in Figueruelas, in the province of Zaragoza, in northeastern Spain. The design and methodology have been described previously.9 Each year, one third of the participating workers (aged 40 years to 55 years) are randomly selected to undergo complementary vascular imaging studies and to complete additional questionnaires. The present report describes a cross-sectional study that includes 929 subjects who enrolled in the first year (2011). The study was approved by the Clinical Research Ethics Committee of Aragon and informed consent was obtained from all the participants.

The clinical, laboratory, and anthropometric data were obtained from the annual physical examinations carried out in the Department of Occupational Risk Management at General Motors España, using procedures that have been standardized, validated, and certified by international standard ISO 9001-2008.

The glucose, triglyceride, high-density lipoprotein cholesterol (HDL-C), and total cholesterol concentrations were determined by enzyme analysis using the ILAB 650 analyzer from Instrumentation Laboratory. Apolipoprotein A-1, apolipoprotein B-100, lipoprotein (a), and C-reactive protein (CRP) were determined by nephelometry using the IMMAGE 800 system from Beckman Coulter. Insulin was determined by means of the Access ultrasensitive chemiluminescence immunoassay from Beckman Coulter. Glycated hemoglobin was determined by cation exchange chromatography on a reverse phase column using the ADAMS A1c HA-810 analyzer from Arkray Factory. Differential leukocyte counts were performed with a Coulter analyzer (CoulterAC•T 5diff AL, Beckman Coulter). In addition, we calculated the HOMA-IR (homeostasis model assessment of insulin resistance) index as a measure of insulin resistance10 and the triglyceride/HDL-C ratio as its clinical expression.11 Low-density lipoprotein cholesterol concentrations were calculated using the Friedewald formula when the triglyceride levels were < 400mg/dL.12

Arterial blood pressure was measured after a 5-min rest period with an OMRON M10-IT digital blood pressure monitor (OMRON Healthcare Co. Ltd., Japan), and we recorded the average of 3 consecutive automatic readings. Weight, height, and waist circumference were also measured and the body mass index was calculated. The participants provided information on their clinical history, reporting their personal and family history with respect to early cardiovascular disease, medication, tobacco use, and diagnosis of hypertension or diabetes mellitus.

Physical activity was assessed using the validated Spanish version13 of the questionnaire on the frequency of engaging in physical activity used in the Nurses’ Health Study and the Health Professionals’ Follow-up Study.14,15 The participants were asked about the average time they had devoted to 17 different sports activities over the year prior to the day of the interview. The questionnaire has 10 categories ranging from “never” to “more than 11h a week”. Moreover, the volunteers were also queried about the number of months of the year that they had engaged in each of the different activities.

To calculate the total physical activity of each volunteer, MET (metabolic equivalents) were assigned to all the activities,16 the values were multiplied by the number of hours a week that the participant claimed to have devoted to said activity, and the total MET-h/week was computed by adding up the MET-h for the different activities.

The questionnaire also included a series of questions on habits and lifestyle, such as hours spent watching television or videos and hours spent at the computer, hours of sleep, and sitting time in hours, differentiating between typical working days and typical weekend days.

To estimate the time devoted to a sedentary lifestyle, we utilized the variable “sitting time”, which corresponded to the hours or fraction of an hour (if the time was < 1 h) that the worker reported that he spent sitting each day, during both the working day and leisure time, differentiating between weekdays and weekend days. The weighted mean was then calculated on the basis of the hours that the participant remained seated on working days and on the weekend.

To evaluate the dietary intake, we used a semiquantitative food frequency questionnaire that had previously been validated in Spain.17 In this questionnaire, each participant provided information on the foods consumed over the year prior to the day of the interview. Seasonal variations and the differences between the patterns of consumption on working days and on the weekend were taken into account.

The participants also completed an additional questionnaire on their sociodemographic characteristics that included: date of birth, level of education, number of years working in the company, shift and type of work carried out, marital status, number of children, and number of persons in the family unit.

The criteria for defining the different cardiovascular risk factors are based on the European clinical practice guidelines. Hypertension was defined as arterial blood pressure levels > 140/90 mmHg (130/80mmHg in participants with diabetes mellitus) or current antihypertensive therapy.18 Hypercholesterolemia was defined as total cholesterol concentrations > 190mg/dL or current lipid-lowering therapy.18 We also considered a volunteer to be diabetic if he had at least one of the following characteristics: fasting glucose level ≥ 126 mg/dL or glycated hemoglobin ≥ 6.5% in at least one analysis, a diagnosis of diabetes mellitus in a previous medical record, or was taking antidiabetic medication.19

The major variables are expressed as the median [interquartile range] in the case of continuous variables, and frequency distribution was used to describe the prevalence of cardiovascular risk factors. The between-group differences in sedentary behavior were analyzed using the Kruskal-Wallis nonparametric H statistic or the chi-square test, and the trend across tertiles of sitting time were studied by means of either the Cuzick test or the Cochrane-Armitage test, as appropriate. We had previously tested for the normal distribution of the data using the Shapiro-Wilk test and, in the case of the variables CRP and HOMA-IR, we applied logarithmic transformation to normalize the data and analyze them as dependent variables.

We performed multivariate linear regression analysis, introducing as dependent variable the insulin resistance parameters triglyceride/HDL-C ratio, insulin, and HOMA-IR, and the inflammatory parameters CRP and leukocyte concentrations. The independent variable was the sitting time reported by the worker in the form of a continuous variable. We present a crude model and another two models adjusted for possible confounding factors identified in earlier studies or because of their biological relationship with the outcome variable, regardless of whether or not there was a correlation in the bivariate analysis. Model A is adjusted for age, body mass index category (normal weight, overweight, and obesity), tobacco use (smoker, ex-smoker, and nonsmoker), alcohol consumption (g/day), diagnosis of diabetes mellitus (yes/no), antihypertensive therapy (yes/no), and lipid-lowering therapy (yes/no). The energy consumption (kcal/day) was not included in the model because of its strong relationship to body mass index. Model B is adjusted for all the variables mentioned for model A plus MET-h/week. The results with these models are expressed as adjusted coefficient and 95% confidence interval.

The statistical analysis was carried out using STATA 12.0 and SPSS 16.0. All of the tests were two-tailed and the significance level was 5%.

No significant differences were observed between the tertiles of sedentary behavior with respect to the consumption pattern, macronutrient distribution, or total caloric intake, which ranged from 2973kcal in the tertile 1 to 2902kcal in the tertile 3 (P = .391).

The most sedentary participants had the worst lipid profile, with lower levels of HDL-C (P < .001) and apolipoprotein A-1 (P = .019), and higher triglyceride levels (P < .001), specifically, 49mg/dL, 143mg/dL, and 140mg/dL, respectively, compared to 53mg/dL, 147mg/dL, and 116mg/dL, respectively, in the least sedentary group. However, we observed no significant differences in the total cholesterol, apolipoprotein B-100, or lipoprotein (a) levels.

With the exception of glycated hemoglobin, which was similar in the different tertiles of sitting time, the remaining parameters studied differed from one group to another, showing an upward trend from the lowest tertile to the most sedentary group in glucose level (96.5mg/dL, 96.0mg/dL, and 99.0mg/dL, respectively; P = .002), HOMA-IR (1.31mg/dL, 1.57mg/dL, and 1.78mg/dL, respectively; P < .001), insulin (5.4 U/mL, 6.7 U/mL, and 7.0 U/mL, respectively; P < .001), triglyceride/HDL-C ratio (2.2, 2.5, and 2.8, respectively; P < .001), leukocytes (6.8 × 103/μL, 7.3 × 103/μL, and 7.5 × 103/μL, respectively; P = .008), and CRP (0.16mg/dL, 0.18mg/dL, and 0.22mg/dL, respectively; P = .003).

Table 2 shows the results of the multivariate linear regression model. The first model represents the crude risk and models A and B present the risk adjusted for possible confounding variables. There was a significant association between sitting time and all the parameters studied with the exception of glycated hemoglobin. After adjustment for the confounding variables described in the “Methods” section, the log CRP, log HOMA-IR and triglyceride/HDL-C ratio continued to be significantly associated with the sitting time (adjusted coefficients, β = 0.07, β = 0.05, β = 0.23, and β = 0.44). There was no variation in the standardized regression coefficient after adjustment for the physical activity performed.

This study benefits from a strict general protocol applied to the quality of each of the laboratory and clinical measurement techniques.

The measures of physical activity and sedentary behavior obtained from the questionnaire have their limitations. The former have previously been validated using accelerometers, but the questions relative to sedentary behavior have not. Moreover, we cannot discern whether the sitting time is continuous or intermittent. The instruments dedicated specifically to these measures are unable to clearly differentiate the resting sitting position, they are costly and, for the time being, their use has not become widespread.

This sample is not representative of the general population. Only men have been included and, as all of them are workers, there is a bias since they are healthy, or at least have no known disabling diseases. In the general population, the levels of sedentary behavior and physical inactivity would probably be different. However, there is no reason to believe that the associations described here occur exclusively in the population from which our sample was recruited.

It could be that possible confounding factors were not taken into account or that the measuring instruments employed (questionnaires) are not accurate enough to avoid a residual confounding effect that could affect our results. In any case, the statistical adjustment was performed based on the best information available and, thus, generates the highest level of evidence for this analysis.