Introduction
Metabolic syndrome is defined as a cluster of pathological conditions that typically includes hyperglycemia, hypercholesterolemia, hypertriglyceridemia, low HDL cholesterol, high body mass index, abdominal obesity and high blood pressure among several other conditions such as insulin resistance, endothelial dysfunction, hypercoagulable state and vascular inflammation [1]. As such, metabolic syndrome is a major determinant of the risk of cardiovascular diseases [2], including coronary heart disease [3], stroke [4] and probably peripheral arterial disease [5].
A large body of literature shows that many biological, environmental and socioeconomic factors influence the prevalence and incidence of metabolic syndrome. For example, significant sex differences exist in propensity to develop metabolic syndrome and on its impact on cardiovascular risk [6,7]. Metabolic syndrome is obviously more frequent in old than in young individuals [8,9]. It is also associated with low socio-economic indicators [10,11] and poor working conditions [12,13]. Likewise, several behaviors such as smoking [14], heavy alcohol consumption [15], leisure-time physical inactivity [16] and unbalanced diet [17] are associated with metabolic syndrome.
Most studies report the prevalence or incidence of metabolic syndrome by analyzing the co-occurrence of markers which are defined by thresholds, more rarely by measuring the correlations between their continuous distributions. The present study aims to systematically assess these intercorrelations across a wide array of biological, social and behavioral characteristics of individuals in a large population-based cohort. The ubiquity and strength of these intercorrelations may give information to which extent metabolic syndrome markers should be considered all together rather than separately in terms of etiology, prevention and treatment of metabolic diseases and cardiovascular risk in the general population.
Methods
Study design
The study retrospectively explored cross-sectional correlations between metabolic syndrome markers across a large array of biological, social and behavioral characteristics of individuals.
Ethical consideration
The study received approval on September 7, 2021 from both the Ethics Evaluation Committee of the French National Institute of Health and Medical Research (Opinion number 21–842) and the National Committee for the Protection of Privacy and Civil Liberties (Authorization #910486). The authors had no access to information that could identify individual participants during or after data collection.
Study setting
The analyses were performed in the population-based CONSTANCES cohort whose participants were recruited between February 1, 2012 and September 30, 2021 across French metropolitan territory [18]. These participants who were affiliated to the general health insurance system (which covers 85% of the French population) were selected using a random sampling scheme stratified on sex, age, socioeconomic status and region. Inclusion criteria comprised the obligation to provide written informed consent, to undergo a comprehensive health examination in one of the twenty-one participating medical centers scattered across the territory and to complete questionnaires on lifestyle, health-related behaviors, social and occupational conditions.
Participants
From the 205,203 participants who were originally recruited in the cohort, 45,727 were excluded due to missing data in metabolic syndrome markers or individual’s characteristics, leaving 159,476 participants in whom the present analyses were performed (Fig 1).
[Figure omitted. See PDF.]
Variables
The study investigated nine metabolic syndrome markers, namely glucose, low-density lipoprotein (LDL), high-density lipoprotein (HDL) and triglyceride blood levels, waist and hip circumferences, body mass index, systolic and diastolic blood pressures. The analyses were stratified across a wide array of biological, social and behavioral characteristics of participants including sex, age, individual and parental histories of cardiovascular disease, diagnosis of metabolic syndrome, smoking, non-moderate alcohol consumption, leisure-time physical inactivity, unbalanced diet, social position, work environment and lifetime unemployment exposure.
Data sources/measurement
Metabolic syndrome markers were assessed at the time of recruitment in the cohort during the visit in the medical centers using standardized procedures in order to guarantee high-quality data [19]. Participants were instructed to fast before the visit that took place between 8 and 10 am, ensuring a fasting period of at least 8h. Waist and hip circumferences were measured as well as height and weight for the calculation of body mass index. Blood was collected to measure concentrations of glucose, total cholesterol, HDL and triglycerides (LDL concentration was calculated by the formula total cholesterol minus HDL minus triglycerides/5) using an analyzer Architect c8000 from Abbott Diagnostics. Systolic and diastolic blood pressures were measured twice in sitting position on each arm at two-minute interval after five minutes of rest using an automated sphygmomanometer OMRON 705. The arm giving the highest mean of the two measures was considered as the reference arm on which a third measure was taken after one additional minute of rest, the average of these three measures was retained for the analyses. The diagnosis of metabolic syndrome was classically defined by the presence of any three of the following five criteria: high waist circumference (≥102 in men or ≥88 cm in women), high blood triglycerides (≥1.7 mmol/l), low blood HDL (<1.0 in men or <1.3 mmol/l in women), high blood pressure (≥130 or ≥85 mmHg for systolic and diastolic blood pressure), high blood glucose (≥5.5 mmol/l) [20].
The other characteristics of individuals, across which intercorrelations of metabolic syndrome markers were tested, were also assessed at the time of recruitment. Besides sex and age, individual history of cardiovascular disease was investigated during the visit in the medical centers where physicians inquired about any non-fatal cardiovascular disease that occurred during the lifetime of participants; four types of cardiovascular diseases were retained for the analyses: stroke, angina pectoris, myocardial infarction and peripheral arterial disease. Parental history of cardiovascular disease referred to the occurrence of at least one of the four cardiovascular diseases on father’s or mother’s side, whatever their age.
Information on risky behaviors (smoking, non-moderate alcohol consumption, leisure-time physical inactivity, unbalanced diet) and socioeconomic conditions (social position, work environment, unemployment exposure) was collected through detailed questionnaires during the same visit. The inquiry assessed habitual smoking status (current, former, never) and the frequency of non-moderate alcohol consumption (more than two or three drinks on the same day in women or men, respectively) classified as rarely (never or less than one time per month), sometimes (two or three times per month) or often (one time or more per week). The assessment of leisure-time physical inactivity was based on a three item questionnaire asking about regular practice of walking or cycling, practicing a sport and gardening or 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 15 minutes for sport, or 2 hours for the two other items, per week), 2 if the practice was regular and higher; the score calculated by summing the three items ranged from 0 (not active at all) to 6 (very active) and was used to characterize leisure-time physical inactivity (participants with a score <2). Diet quality was assessed through a weekly 32-item food frequency questionnaire. Each item represented the daily frequency of consumption of a food category on a four-level scale; a score of 0 was given to the frequency “never or almost never”, 2 to “once or twice”, 4 to “not every day but more than twice” and 7 to “every day or almost every day”. The mean of the scores was calculated separately for “healthy items” (poultry, fish, wholewheat bread, whole grain pasta, brown rice, fresh fruits, vegetables, legumes) and “unhealthy items” (red meat, cooked pork meats, white bread, delicacies, appetizers, prepared foods, fast foods, fried foods, pastries, sweet desserts). The two means were divided into terciles and the diet quality was defined by combining the low tercile for unhealthy items with the high tercile for healthy items (balanced diet), the two middle terciles (slightly unbalanced diet) and the high tercile for unhealthy foods with the low tercile for healthy foods (strongly unbalanced diet).
Several indicators whose distributions are shown in S1 Table were used to assess social position of participants. Given that these indicators which included educational attainment, occupation of participants and spouses, monthly income of all household members and an evaluation of social vulnerability described complementary and interdependent aspects of social position, a global score combining all the indicators was calculated and categorized into terciles to delineate low, middle or high social position as previously reported [21]. A series of 19 organizational, physical, biomechanical, chemical and psychosocial exposures were used to characterize work environment of participants that was considered as a whole, which is reality for workers who are not facing only one or a few occupational exposures. For that purpose, these exposures whose distributions are shown in S2 Table were combined into a global score which was categorized into terciles to define bad, average or good work environment as previously described [22]. Unemployment exposure of participants during their lifetime was documented by a questionnaire in which they were asked to report each time they were unemployed for a period of more than six months. The existence of past episodes of unemployment was confirmed for each participant by administrative data from the French national pension system which also provided the total number of unemployed quarters. This number whose distribution is reported in S1 Fig was arbitrarily categorized into three groups (0, 1–19, 20–148 quarters) to estimate unemployment exposure [23].
Bias
The choice of selecting participants with no missing data rather than imputing them was driven by the fact that the cohort was not representative of the French population. This was due to a low inclusion rate (7.3%) despite the stratified sampling strategy that tried to compensate for the higher non-response rate of individuals with low socioeconomic status [24]. Although in line with rates observed in other large population-based cohorts when participants are required to visit a medical center for health-related exams [25], this low rate resulted in the selection of socially privileged people (S3 Table). The selection of participants with no missing values only marginally accentuated this bias while the alternative of using multivariate imputation by chained equations would not have been devoid of other biases [26]. The means and standard deviations of metabolic syndrome marker distributions were also only marginally affected by the exclusion of participants with missing values as reported in S4 Table.
Study size
Exact calculation using a series expansion for the distributions of multiple correlation coefficients [27] showed that the minimum sample size to detect with a high power (99%) very significant (p < 0.0001) partial correlations with coefficients of at least 0.1 between 9 continuous variables was n = 5636. Thus, the present study with a population size 28 times larger than required (n = 159,476) was highly powered to reject null hypotheses and explore partial correlations between metabolic syndrome markers.
Quantitative variables
Metabolic syndrome markers (blood glucose, low-density and high-density lipoproteins, triglycerides, body mass index, waist and hip circumferences, systolic and diastolic blood pressures) were analyzed as continuous variables to test their intercorrelations. Sex (F/M), individual and parental histories of cardiovascular disease (N/Y), diagnosis of metabolic syndrome (N/Y) and leisure-time physical inactivity (N/Y) were coded as binary variables. Age was divided into terciles while smoking (never, former, current), non-moderate alcohol consumption (rarely, sometimes, often), unbalanced diet (no, slightly, strongly), social position (high, middle, low), work environment (good, average, bad) and lifetime unemployment exposure (0, 1–19, 20–148 quarters) were coded into three categories as mentioned above.
Statistical methods
The strength of the correlations between metabolic syndrome markers were measured by calculating Pearson correlation coefficients between each pair of markers as well as partial correlation coefficients that measured the strength of the correlations after adjusting for the effects of all the other markers. The analyses were stratified across a wide array of biological, social and behavioral characteristics of participants that included sex, age, individual and parental histories of cardiovascular disease, diagnosis of metabolic syndrome, smoking, non-moderate alcohol consumption, leisure-time physical inactivity, unbalanced diet, social position, work environment and lifetime unemployment exposure. Given that partial correlations between each pair of metabolic syndrome markers (n = 36) were assessed 32 times across biological, social and behavioral characteristics of participants, the statistical significance was set at p < 0.0001 to minimize as much as possible the risk of having false positives due to multiple testing. All analyses were performed with statistical discovery software JMP 17 Pro (SAS, Cary NC).
Results
Characteristics of participants
The characteristics of the 159,476 participants selected for this study are reported in Table 1. Aged in average 46.5 years (SD 13.4), 2.1 and 25.0% of them had respectively an individual or a parental history of cardiovascular disease. While a large fraction (62.1%) often consumed alcohol in a non-moderate way and 53.6% were current (18.7%) or former (34.8%) smokers, only 9.3% were physically inactive during their leisure-time and 10.4% had a strongly unbalanced diet. A significant proportion had a low social position (25.7%) or were exposed to bad work environment (32.5%) but only 14.8% were exposed to short (9.8%) or long (5.0%) periods of unemployment during their lifetime.
[Figure omitted. See PDF.]
Continuous distributions of metabolic syndrome markers in participants are shown in Fig 2. Average blood concentrations (SD) were respectively 5.31 (0.82), 1.51 (0.40), 3.66 (0.99) and 1.12 (0.67) mmol/l for glucose, HDL cholesterol, LDL cholesterol and triglycerides. Average values (SD) were 24.9 (4.4) kg/m2 for body mass index, 84.8 (12.8) and 99.2 (8.5) cm for waist and hip circumferences, 128.0 (15.8) and 76.5 (9.4) mmHg for systolic and diastolic blood pressures. Metabolic syndrome was diagnosed in 13.8% of participants (n = 22,044).
[Figure omitted. See PDF.]
The mean with standard deviation (SD) and median as well as minimum and maximum values are reported for each marker.
Correlations and partial correlations between metabolic syndrome markers in all participants
Correlations and partial correlations coefficients between metabolic syndrome markers in all participants are reported in Table 2. Although each marker was significantly correlated with all the others, partial correlations coefficients indicate that some markers were more strongly intercorrelated than others when adjusting for the effects of the other markers. Besides the strong and expected partial correlations between systolic and diastolic blood pressures (r = 0.71), between body mass index and waist (r = 0.55) and hip (r = 0.63) circumferences and between waist and hip circumferences (r = 0.18), systolic blood pressure was correlated with waist (r = 0.17) and hip (r = −0.10) circumferences and with blood glucose (r = 0.13), triglycerides with blood glucose (r = 0.11), LDL (r = 0.27), HDL (r = −0.23) and waist circumference (r = 0.12), waist and hip circumferences with blood glucose (r = 0.16 and −0.11) and HDL (r = −0.16 and 0.10).
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome according to sex
The same pattern of partial correlations described in all participants was observed with little differences in both sexes (Table 3). In women and men, there were strong correlations between systolic and diastolic blood pressures (r = 0.72 and 0.69), between body mass index and waist (r = 0.52 and 0.63) and hip (r = 0.61 and 0.59) circumferences and between waist and hip circumferences (r = 0.18 and 0.19); systolic blood pressure was correlated with waist (r = 0.16 and 0.16) and hip (r = −0.09 and −0.11) circumferences and with blood glucose (r = 0.14 and 0.11), triglycerides with blood glucose (r = 0.12 and 0.11), LDL (r = 0.30 and 0.25), HDL (r = −0.20 and −0.29) and waist circumference (r = 0.14 and 0.09), HDL also with waist (r = −0.20 and −0.23) and hip (r = 0.14 and 0.11) circumferences, blood glucose with waist (r = 0.10 and 0.15) and hip (r = −0.08 and −0.13) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to age
The same pattern of partial correlations was observed whatever the age group (Table 4). In 18−39, 40−55 and 56−75 years-old participants, there were strong correlations between systolic and diastolic blood pressures (r = 0.64, 0.69 and 0.71), between body mass index, waist (r = 0.50, 0.53 and 0.55) and hip (r = 0.61, 0.63 and 0.65) circumferences and between waist and hip circumferences (r = 0.17, 0.11 and 0.17); systolic blood pressure was correlated with waist (r = 0.16, 0.15 and 0.14) and hip (r = −0.11, −0.07 and −0.10) circumferences and with blood glucose (r = 0.13, 0.09 and 0.11), triglycerides with blood glucose (r = 0.13, 0.10 and 0.15), LDL (r = 0.28, 0.26 and 0.25), HDL (r = −0.26, −0.23 and −0.26) and waist circumference (r = 0.13, 0.14 and 0.08), HDL also with waist (r = −0.12, −0.18 and −0.24) and hip (r = 0.09, 0.12 and 0.17) circumferences, blood glucose with waist (r = 0.09, 0,13 and 0.14) and hip (r = −0.08, −0.09 and −0.09) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to individual and parental history of cardiovascular disease
Neither individual or parental histories of cardiovascular disease substantially modified the pattern of partial correlations between metabolic syndrome markers. In participants without or with individual history of cardiovascular disease (Table 5), there were strong correlations between systolic and diastolic blood pressures (r = 0.71 and 0.66), between body mass index, and waist (r = 0.54 and 0.63) and hip (r = 0.64 and 0.63) circumferences and between waist and hip circumferences (r = 0.09 and 0.11); systolic blood pressure was correlated with waist (r = 0.16 and 0.12) and hip (r = −0.10 and −0.09) circumferences and with blood glucose (r = 0.13 and 0.13), triglycerides with blood glucose (r = 0.11 and 0.08), LDL (r = 0.27 and 0.27), HDL (r = −0.24 and −0.24) and waist circumference (r = 0.11 and 0.15), HDL also with waist (r = −0.17 and −0.21) and hip (r = 0.10 and 0.18) circumferences, blood glucose with waist (r = 0.16 and 0.12) and hip (r = −0.11 and −0.11) circumferences. In participants without or with parental history of cardiovascular disease (Table 6), there were strong correlations between systolic and diastolic blood pressures (r = 0.71 and 0.71), between body mass index and waist (r = 0.54 and 0.56) and hip (r = 0.63 and 0.66) circumferences and between waist and hip circumferences (r = 0.09 and 0.15); systolic blood pressure was correlated with waist (r = 0.18 and 0.13) and hip (r = −0.10 and −0.10) circumferences and with blood glucose (r = 0.13 and 0.14), triglycerides with blood glucose (r = 0.10 and 0.13), LDL (r = 0.27 and 0.26), HDL (r = −0.23 and −0.24) and waist circumference (r = 0.12 and 0.11), HDL also with waist (r = −0.15 and −0.19) and hip (r = 0.09 and 0.12) circumferences, blood glucose with waist (r = 0.16 and 0.15) and hip (r = −0.11 and −0.11) circumferences.
[Figure omitted. See PDF.]
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to the diagnosis of metabolic syndrome
The same pattern of partial correlations was observed independently of the diagnosis of metabolic syndrome (Table 7). In participants without or with metabolic syndrome, there were strong correlations between systolic and diastolic blood pressures (r = 0.72 and 0.62), between body mass index and waist (r = 0.53 and 0.54) and hip (r = 0.61 and 0.70) circumferences and between waist and hip circumferences (r = 0.10 and 0.10); systolic blood pressure was correlated with waist (r = 0.17 and 0.10) and hip (r = −0.10 and −0.11) circumferences and with blood glucose (r = 0.13 and 0.10), triglycerides with blood glucose (r = 0.10 and 0.09), LDL (r = 0.33 and 0.30), HDL (r = −0.22 and −0.22) and waist circumference (r = 0.11 and 0.07), HDL also with waist (r = −0.16 and −0.13) and hip (r = 0.09 and 0.16) circumferences, blood glucose with waist (r = 0.16 and 0.14) and hip (r = −0.11 and −0.13) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to social position
The same pattern of partial correlations was observed across social position (Table 8). In participants with high, middle or low social position, there were strong correlations between systolic and diastolic blood pressures (r = 0.74, 0.71 and 0.69), between body mass index and waist (r = 0.51, 0.54 and 0.57) and hip (r = 0.58, 0.64 and 0.68) circumferences and between waist and hip circumferences (r = 0.15, 0.08 and 0.12); systolic blood pressure was correlated with waist (r = 0.18, 0.18 and 0.15) and hip (r = −0.09, −0.10 and −0.08) circumferences and with blood glucose (r = 0.12, 0.14 and 0.13), triglycerides with blood glucose (r = 0.07, 0.10 and 0.13), LDL (r = 0.30, 0.27 and 0.25), HDL (r = −0.24, −0.22 and −0.25) and waist circumference (r = 0.12, 0.12 and 0.11), HDL also with waist (r = −0.18, −0.15 and −0.15) and hip (r = 0.08, 0.10 and 0.11) circumferences, blood glucose with waist (r = 0.16, 0,16 and 0.16) and hip (r = −0.10, −0.11 and −0.12) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to work environment
Table 9 shows that the same pattern of partial correlations was observed across work environment. In participants with good, average or bad work environment, there were strong correlations between systolic and diastolic blood pressures (r = 0.74, 0.71 and 0.69), between body mass index and waist (r = 0.50, 0.55 and 0.58) and hip (r = 0.61, 0.65 and 0.66) circumferences and between waist and hip circumferences (r = 0.15, 0.08 and 0.14); systolic blood pressure was correlated with waist (r = 0.17, 0.16 and 0.13) and hip (r = −0.08, −0.09 and −0.09) circumferences and with blood glucose (r = 0.12, 0.13 and 0.12), triglycerides with blood glucose (r = 0.09, 0.10 and 0.13), LDL (r = 0.29, 0.27 and 0.25), HDL (r = −0.23, −0.23 and −0.25) and waist circumference (r = 0.14, 0.12 and 0.08), HDL also with waist (r = −0.17, −0.15 and −0.18) and hip (r = 0.08, 0.10 and 0.13) circumferences, blood glucose with waist (r = 0.14, 0,16 and 0.16) and hip (r = −0.09, −0.11 and −0.12) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to lifetime unemployment exposure
The same pattern of partial correlations was observed whatever lifetime unemployment exposure (Table 10). In participants exposed to 0, 1–19 or 20–148 unemployed quarters, there were strong correlations between systolic and diastolic blood pressures (r = 0.71, 0.72 and 0.74), between body mass index and waist (r = 0.55, 0.54 and 0.56) and hip (r = 0.63, 0.66 and 0.70) circumferences and between waist and hip circumferences (r = 0.09, 0.08 and 0.11); systolic blood pressure was correlated with waist (r = 0.17, 0.18 and 0.13) and hip (r = −0.10, −0.09 and −0.07) circumferences and with blood glucose (r = 0.13, 0.13 and 0.12), triglycerides with blood glucose (r = 0.11, 0.11 and 0.16), LDL (r = 0.27, 0.26 and 0.27), HDL (r = −0.23, −0.24 and −0.26) and waist circumference (r = 0.11, 0.14 and 0.14), HDL also with waist (r = −0.16, −0.17 and −0.19) and hip (r = 0.10, 0.10 and 0.12) circumferences, blood glucose with waist (r = 0.16, 0,15 and 0.13) and hip (r = −0.11, −0.11 and −0.11) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to smoking status
Table 11 shows that the same pattern of partial correlations was observed whatever the smoking status. In participants who never smoked, former smokers or current smokers, there were strong correlations between systolic and diastolic blood pressures (r = 0.70, 0.71 and 0.71), between body mass index and waist (r = 0.53, 0.56 and 0.55) and hip (r = 0.64, 0.63 and 0.63) circumferences and between waist and hip circumferences (r = 0.10, 0.07 and 0.10); systolic blood pressure was correlated with waist (r = 0.15, 0.18 and 0.21) and hip (r = −0.09, −0.12 and −0.10) circumferences and with blood glucose (r = 0.14, 0.13 and 0.11), triglycerides with blood glucose (r = 0.10, 0.13 and 0.09), LDL (r = 0.29, 0.26 and 0.24), HDL (r = −0.23, −0.24 and −0.21) and waist circumference (r = 0.11, 0.11 and 0.12), HDL also with waist (r = −0.17, −0.17 and −0.14) and hip (r = 0.10, 0.10 and 0.09) circumferences, blood glucose with waist (r = 0.14, 0,17 and 0.15) and hip (r = −0.11, −0.12 and −0.10) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to non-moderate alcohol consumption
The same pattern of partial correlations was observed across non-moderate alcohol consumption profiles (Table 12). In participants consuming rarely, sometimes or often non-moderate amounts of alcohol, there were strong correlations between systolic and diastolic blood pressures (r = 0.71, 0.71 and 0.71), between body mass index and waist (r = 0.55, 0.53 and 0.56) and hip (r = 0.67, 0.65 and 0.59) circumferences and between waist and hip circumferences (r = 0.08, 0.10 and 0.10); systolic blood pressure was correlated with waist (r = 0.14, 0.15 and 0.17) and hip (r = −0.08, −0.09 and −0.10) circumferences and with blood glucose (r = 0.13, 0.12 and 0.14), triglycerides with blood glucose (r = 0.14, 0.13 and 0.09), LDL (r = 0.28, 0.25 and 0.27), HDL (r = −0.23, −0.21 and −0.24) and waist circumference (r = 0.15, 0.12 and 0.10), HDL also with waist (r = −0.15, −0.18 and −0.17) and hip (r = 0.12, 0.11 and 0.10) circumferences, blood glucose with waist (r = 0.14, 0,15 and 0.16) and hip (r = −0.10, −0.11 and −0.12) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to leisure-time physical inactivity
Table 13 shows that the same pattern of partial correlations was observed whatever the profile of leisure-time physical inactivity. In participants without or with leisure-time physical inactivity, there were strong correlations between systolic and diastolic blood pressures (r = 0.71 and 0.72), between body mass index and waist (r = 0.55 and 0.53) and hip (r = 0.63 and 0.66) circumferences and between waist and hip circumferences (r = 0.08 and 0.11); systolic blood pressure was correlated with waist (r = 0.17 and 0.18) and hip (r = −0.10 and −0.09) circumferences and with blood glucose (r = 0.13 and 0.13), triglycerides with blood glucose (r = 0.11 and 0.12), LDL (r = 0.27 and 0.25), HDL (r = −0.23 and −0.23) and waist circumference (r = 0.11 and 0.13), HDL also with waist (r = −0.16 and −0.16) and hip (r = 0.10 and 0.10) circumferences, blood glucose with waist (r = 0.16 and 0.16) and hip (r = −0.11 and −0.12) circumferences.
[Figure omitted. See PDF.]
Partial correlations between metabolic syndrome markers according to diet quality
The same pattern of partial correlations was observed whatever diet quality (Table 14). In participants with no, slightly or strongly unbalanced diet, there were strong correlations between systolic and diastolic blood pressures (r = 0.72, 0.71 and 0.70), between body mass index and waist (r = 0.53, 0.55 and 0.56) and hip (r = 0.63, 0.64 and 0.64) circumferences and between waist and hip circumferences (r = 0.10, 0.08 and 0.08); systolic blood pressure was correlated with waist (r = 0.14, 0.17 and 0.19) and hip (r = −0.09, −0.10 and −0.09) circumferences and with blood glucose (r = 0.15, 0.13 and 0.13), triglycerides with blood glucose (r = 0.11, 0.11 and 0.13), LDL (r = 0.31, 0.26 and 0.26), HDL (r = −0.24, −0.23 and −0.23) and waist circumference (r = 0.13, 0.11 and 0.11), HDL also with waist (r = −0.16, −0.16 and −0.11) and hip (r = 0.10, 0.10 and 0.10) circumferences, blood glucose with waist (r = 0.16, 0,16 and 0.13) and hip (r = −0.21, −0.12 and −0.11) circumferences.
[Figure omitted. See PDF.]
Discussion
The main finding of the present study is that the pattern of intercorrelations between major metabolic syndrome markers, analyzed as continuous variables, remains remarkably constant across a wide array of biological, social and behavioral characteristics of individuals. This pattern that notably includes intercorrelations of waist and hip circumferences with blood glucose, HDL, triglycerides and systolic blood pressure is observed whatever the sex, age, individual and parental histories of cardiovascular disease, social position, work environment, lifetime unemployment exposure, smoking status, non-moderate alcohol consumption, leisure-time physical inactivity and diet quality. It is even observed independently of the diagnosis of metabolic syndrome.
It is noteworthy that although all metabolic syndrome markers considered in this study are correlated with each other, partial correlation coefficients reveal that only some remain intercorrelated when correlations with the other markers are taken into account. Thus, the intercorrelations of diastolic blood pressure and body mass index with the other markers would in fact reflect their intercorrelations with systolic blood pressure and waist and hip circumferences. This supports the hypothesis that the effect of body mass index on cardiovascular risk would be almost entirely explained by waist and hip circumferences [28]. Furthermore, the strength of the intercorrelations of waist and hip circumferences with the other markers seems to indicate a prominent role of waist over hip circumference on cardiovascular risk as already suggested [29]. Likewise, even though diastolic and systolic blood pressures are highly correlated with each other, only the latter is correlated with the other markers, corroborating the view that it would have a prominent role over diastolic pressure in determining cardiovascular risk [30,31].
The mechanisms underlying the interrelationships between metabolic syndrome markers are not clear. One hypothesis is that genetic background would maintain, independently of individual’s characteristics, the coordinated expression of the genes that determine marker levels [32]. However, genome-wide association studies that looked for single nucleotide polymorphisms associated with marker levels have largely failed to identify the genetic dimension of metabolic syndrome as a whole [33]. The non-genetic nature of the stability of the interrelationships between metabolic syndrome markers would also be in accordance with the fact that neither individual nor parental histories of cardiovascular disease significantly modify the pattern of intercorrelations. A second hypothesis is that organs such as the intestine would regulate marker levels in a coordinated way. This would be corroborated by the existence of treatments targeting the intestine and having pleiotropic effects on marker levels [34,35]. This would also fit with the potential involvement of the gut microbiota in the development of metabolic syndrome [36]. Another organ playing such a role could be abdominal fat as reflected in body mass index whose level changes modulate several other markers, as glycemia and blood pressure [37,38], which largely mediate the excess risk of cardiovascular diseases associated with high body mass index [39].
Whatever the mechanisms underlying the interrelationships between metabolic syndrome markers, a remarkable feature besides their omnipresence is their relative insensitivity to individual’s characteristics that otherwise deeply affect marker levels. Thus, the same characteristics that have a strong influence on marker levels such as age [8,9], social position [10,11], work environment [12,13], smoking [14], heavy alcohol consumption [15], leisure-time physical inactivity [16] or unbalanced diet [17] seem to have little effect on the intercorrelations of markers with each other.
This study has some limitations. One is its external validity that is not guaranteed given that the findings were obtained in a cohort which was not representative of the French population, including in particular a large proportion of socially-privileged individuals. A second is that information on cardiovascular drug takings by participants was not available and could not be taken into account as confounding factors in the analyses.
In conclusion, the results indicate that the intercorrelations between several major metabolic syndrome markers are omnipresent in the general population, whatever individual’s characteristics. This observation provides a different angle of view on these markers whose relationships would extend well beyond the usual diagnosis of metabolic syndrome and that should be considered all together rather than separately in terms of etiology, prevention and treatment of metabolic diseases and cardiovascular risk.
Supporting information
S1 Fig. Lifetime unemployment exposure of participants.
https://doi.org/10.1371/journal.pone.0328577.s001
(DOCX)
S1 Table. Indicators of social position of participants.
https://doi.org/10.1371/journal.pone.0328577.s002
(DOCX)
S2 Table. Indicators of work environment of participants.
https://doi.org/10.1371/journal.pone.0328577.s003
(DOCX)
S3 Table. Characteristics of cohort participants with or without missing values compared to randomly selected individuals from the French population.
https://doi.org/10.1371/journal.pone.0328577.s004
(DOCX)
S4 Table. Means ± standard deviations of the distributions of metabolic syndrome markers in participants with or without missing values.
https://doi.org/10.1371/journal.pone.0328577.s005
(DOCX)
Acknowledgments
The authors thank the team of the Population-based Cohorts Unit that designed and manages the Constances Cohort Study. They also thank the French National Health Insurance (Caisse nationale d’assurance maladie, CNAM) and its Health Screening Centers (Centres d’examens de santé) which collect a large part of the data, the French National Old-Age Insurance (Caisse nationale d’assurance vieillesse, CNAV) for its contribution to the constitution of the cohort, and ClinSearch, Asqualab and Eurocell which are conducting data quality control.
References
1. 1. Reaven GM. The metabolic syndrome: time to get off the merry-go-round? J Intern Med. 2011;269(2):127–36. pmid:21129047
* View Article
* PubMed/NCBI
* Google Scholar
2. 2. Mottillo S, Filion KB, Genest J, Joseph L, Pilote L, Poirier P, et al. The metabolic syndrome and cardiovascular risk a systematic review and meta-analysis. J Am Coll Cardiol. 2010;56(14):1113–32. pmid:20863953
* View Article
* PubMed/NCBI
* Google Scholar
3. 3. Alshammary AF, Alharbi KK, Alshehri NJ, Vennu V, Ali Khan I. Metabolic syndrome and coronary artery disease risk: a meta-analysis of observational studies. Int J Environ Res Public Health. 2021;18(4):1773. pmid:33670349
* View Article
* PubMed/NCBI
* Google Scholar
4. 4. Zhang F, Liu L, Zhang C, Ji S, Mei Z, Li T. Association of metabolic syndrome and its components with risk of stroke recurrence and mortality: a meta-analysis. Neurology. 2021;97(7):e695–705. pmid:34321360
* View Article
* PubMed/NCBI
* Google Scholar
5. 5. Soriano-Moreno DR, Fernandez-Morales J, Medina-Ramirez SA, Coico-Lama AH, Soriano-Moreno AN, Zafra-Tanaka JH. Metabolic syndrome and risk of peripheral arterial disease: a systematic review and meta-analysis. Cardiol Rev. 2022.
* View Article
* Google Scholar
6. 6. Appelman Y, van Rijn BB, Ten Haaf ME, Boersma E, Peters SA. Sex differences in cardiovascular risk factors and disease prevention. Atherosclerosis. 2015;241:211–8.
* View Article
* Google Scholar
7. 7. Pradhan AD. Sex differences in the metabolic syndrome: implications for cardiovascular health in women. Clin Chem. 2014;60(1):44–52. pmid:24255079
* View Article
* PubMed/NCBI
* Google Scholar
8. 8. Kuk JL, Ardern CI. Age and sex differences in the clustering of metabolic syndrome factors: association with mortality risk. Diabetes Care. 2010;33:2457–61.
* View Article
* Google Scholar
9. 9. Alexander CM, Landsman PB, Grundy SM. The influence of age and body mass index on the metabolic syndrome and its components. Diabetes Obes Metab. 2008;10:246–50.
* View Article
* Google Scholar
10. 10. Blanquet M, Legrand A, Pélissier A, Mourgues C. Socio-economics status and metabolic syndrome: a meta-analysis. Diabetes Metab Syndr. 2019;13:1805–12.
* View Article
* Google Scholar
11. 11. Tamashiro KLK. Metabolic syndrome: links to social stress and socioeconomic status. Ann N Y Acad Sci. 2011;1231:46–55. pmid:21884160
* View Article
* PubMed/NCBI
* Google Scholar
12. 12. Wang Y, Yu L, Gao Y, Jiang L, Yuan L, Wang P, et al. Association between shift work or long working hours with metabolic syndrome: a systematic review and dose-response meta-analysis of observational studies. Chronobiol Int. 2021;38(3):318–33. pmid:33557625
* View Article
* PubMed/NCBI
* Google Scholar
13. 13. Watanabe K, Sakuraya A, Kawakami N, Imamura K, Ando E, Asai Y, et al. Work-related psychosocial factors and metabolic syndrome onset among workers: a systematic review and meta-analysis. Obes Rev. 2018;19(11):1557–68. pmid:30047228
* View Article
* PubMed/NCBI
* Google Scholar
14. 14. Sun K, Liu J, Ning G. Active smoking and risk of metabolic syndrome: a meta-analysis of prospective studies. PLoS One. 2012;7:e47791.
* View Article
* Google Scholar
15. 15. Sun K, Ren M, Liu D, Wang C, Yang C, Yan L. Alcohol consumption and risk of metabolic syndrome: a meta-analysis of prospective studies. Clin Nutr. 2014;33(4):596–602. pmid:24315622
* View Article
* PubMed/NCBI
* Google Scholar
16. 16. Zhang D, Liu X, Liu Y, Sun X, Wang B, Ren Y, et al. Leisure-time physical activity and incident metabolic syndrome: a systematic review and dose-response meta-analysis of cohort studies. Metabolism. 2017;75:36–44. pmid:28927737
* View Article
* PubMed/NCBI
* Google Scholar
17. 17. Fabiani R, Naldini G, Chiavarini M. Dietary patterns and metabolic syndrome in adult subjects: a systematic review and meta-analysis. Nutrients. 2019;11(9):2056. pmid:31480732
* View Article
* PubMed/NCBI
* Google Scholar
18. 18. Zins M, Goldberg M, CONSTANCES team. The French CONSTANCES population-based cohort: design, inclusion and follow-up. Eur J Epidemiol. 2015;30(12):1317–28. pmid:26520638
* View Article
* PubMed/NCBI
* Google Scholar
19. 19. Ruiz F, Goldberg M, Lemonnier S, Ozguler A, Boos E, Brigand A, et al. High quality standards for a large-scale prospective population-based observational cohort: constances. BMC Public Health. 2016;16(1):877. pmid:27557750
* View Article
* PubMed/NCBI
* Google Scholar
20. 20. Grundy SM, Cleeman JI, Daniels SR, Donato KA, Eckel RH, Franklin BA, et al. Diagnosis and management of the metabolic syndrome: an American Heart Association/National Heart, Lung, and Blood Institute Scientific Statement. Circulation. 2005;112(17):2735–52. pmid:16157765
* View Article
* PubMed/NCBI
* Google Scholar
21. 21. Hoertel N, Sanchez Rico M, Limosin F, Ménard J, Ribet C, Bonenfant S, 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(2):e021373. pmid:35023345
* View Article
* PubMed/NCBI
* Google Scholar
22. 22. Meneton P, Hoertel N, Wiernik E, Lemogne C, Ribet C, Bonenfant S. Work environment mediates a large part of social inequalities in the incidence of several common cardiovascular risk factors: Findings from the Gazel cohort. Soc Sci Med. 2018;216:59–66.
* View Article
* Google Scholar
23. 23. Sanchez Rico M, Plessz M, Airagnes G, Ribet C, Hoertel N, Goldberg M. Cardiovascular burden and unemployment: a retrospective study in a large population-based French cohort. PLoS One. 2023;18:e0288747.
* View Article
* Google Scholar
24. 24. Goldberg M, Carton M, Descatha A, Leclerc A, Roquelaure Y, Santin G, et al. CONSTANCES: a general prospective population-based cohort for occupational and environmental epidemiology: cohort profile. Occup Environ Med. 2017;74(1):66–71. pmid:27884936
* View Article
* PubMed/NCBI
* Google Scholar
25. 25. Toledano MB, Smith RB, Brook JP, Douglass M, Elliott P. How to establish and follow up a large prospective cohort study in the 21st century--lessons from UK COSMOS. PLoS One. 2015;10(7):e0131521. pmid:26147611
* View Article
* PubMed/NCBI
* Google Scholar
26. 26. White IR, Carlin JB. Bias and efficiency of multiple imputation compared with complete-case analysis for missing covariate values. Stat Med. 2010;29(28):2920–31. pmid:20842622
* View Article
* PubMed/NCBI
* Google Scholar
27. 27. Gatsonis C, Sampson AR. Multiple correlation: exact power and sample size calculations. Psychol Bull. 1989;106(3):516–24. pmid:2813654
* View Article
* PubMed/NCBI
* Google Scholar
28. 28. Czernichow S, Kengne A-P, Stamatakis E, Hamer M, Batty GD. Body mass index, waist circumference and waist-hip ratio: which is the better discriminator of cardiovascular disease mortality risk?: evidence from an individual-participant meta-analysis of 82 864 participants from nine cohort studies. Obes Rev. 2011;12(9):680–7. pmid:21521449
* View Article
* PubMed/NCBI
* Google Scholar
29. 29. Darsini D, Hamidah H, Notobroto HB, Cahyono EA. Health risks associated with high waist circumference: a systematic review. J Public Health Res. 2020;9(2):1811. pmid:32728557
* View Article
* PubMed/NCBI
* Google Scholar
30. 30. Flint AC, Conell C, Ren X, Banki NM, Chan SL, Rao VA, et al. Effect of systolic and diastolic blood pressure on cardiovascular outcomes. N Engl J Med. 2019;381(3):243–51. pmid:31314968
* View Article
* PubMed/NCBI
* Google Scholar
31. 31. Kannel WB. Historic perspectives on the relative contributions of diastolic and systolic blood pressure elevation to cardiovascular risk profile. Am Heart J. 1999;138(3 Pt 2):205–10. pmid:10467214
* View Article
* PubMed/NCBI
* Google Scholar
32. 32. van Walree ES, Jansen IE, Bell NY, Savage JE, de Leeuw C, Nieuwdorp M, et al. Disentangling genetic risks for metabolic syndrome. Diabetes. 2022;71(11):2447–57. pmid:35983957
* View Article
* PubMed/NCBI
* Google Scholar
33. 33. Rana S, Ali S, Wani HA, Mushtaq QD, Sharma S, Rehman MU. Metabolic syndrome and underlying genetic determinants-A systematic review. J Diabetes Metab Disord. 2022;21(1):1095–104. pmid:35673448
* View Article
* PubMed/NCBI
* Google Scholar
34. 34. Scott R, O’Brien R, Fulcher G, Pardy C, D’Emden M, Tse D, et al. Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study. Diabetes Care. 2009;32(3):493–8. pmid:18984774
* View Article
* PubMed/NCBI
* Google Scholar
35. 35. Verma S, Al-Omran M, Leiter LA, Mazer CD, Rasmussen S, Saevereid HA. Cardiovascular efficacy of liraglutide and semaglutide in individuals with diabetes and peripheral artery disease. Diabetes Obes Metab. 2022;24:1288–99.
* View Article
* Google Scholar
36. 36. Festi D, Schiumerini R, Eusebi LH, Marasco G, Taddia M, Colecchia A. Gut microbiota and metabolic syndrome. World J Gastroenterol. 2014;20(43):16079–94. pmid:25473159
* View Article
* PubMed/NCBI
* Google Scholar
37. 37. Geleijnse JM, Kok FJ, Grobbee DE. Impact of dietary and lifestyle factors on the prevalence of hypertension in Western populations. Eur J Public Health. 2004;14(3):235–9. pmid:15369026
* View Article
* PubMed/NCBI
* Google Scholar
38. 38. Merlotti C, Morabito A, Pontiroli AE. Prevention of type 2 diabetes; a systematic review and meta-analysis of different intervention strategies. Diabetes Obes Metab. 2014;16(8):719–27. pmid:24476122
* View Article
* PubMed/NCBI
* Google Scholar
39. 39. Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects), Lu Y, Hajifathalian K, Ezzati M, Woodward M, Rimm EB, et al. Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1·8 million participants. Lancet. 2014;383(9921):970–83. pmid:24269108
* View Article
* PubMed/NCBI
* Google Scholar
Citation: Rico MS, Wiernik E, Kab S, Renuy A, Hoertel N, Ménard J, et al. (2025) Omnipresent intercorrelations of metabolic syndrome markers in the general population. PLoS One 20(8): e0328577. https://doi.org/10.1371/journal.pone.0328577
About the Authors:
Marina Sanchez Rico
Roles: Formal analysis, Methodology, Software, Validation, Visualization, Writing – review & editing
Affiliation: AP-HP, DMU Psychiatrie et Addictologie, Hôpital Corentin-Celton, Issy-les-Moulineaux, France
Emmanuel Wiernik
Roles: Data curation, Project administration, Validation, Writing – review & editing
Affiliation: Université Paris Cité, Université Paris-Saclay, Université de Versailles Saint-Quentin-en-Yvelines, INSERM UMS_011, Villejuif, France
Sofiane Kab
Roles: Data curation, Project administration, Validation, Writing – review & editing
Affiliation: Université Paris Cité, Université Paris-Saclay, Université de Versailles Saint-Quentin-en-Yvelines, INSERM UMS_011, Villejuif, France
Adeline Renuy
Roles: Data curation, Methodology, Project administration, Writing – review & editing
Affiliation: Université Paris Cité, Université Paris-Saclay, Université de Versailles Saint-Quentin-en-Yvelines, INSERM UMS_011, Villejuif, France
Nicolas Hoertel
Roles: Validation, Writing – review & editing
Affiliations: AP-HP, DMU Psychiatrie et Addictologie, Hôpital Corentin-Celton, Issy-les-Moulineaux, France, Faculté de Médecine, Université Paris Cité, Paris, France, INSERM UMR_1266, Paris, France
Joël Ménard
Roles: Validation, Writing – review & editing
Affiliation: Faculté de Médecine, Université Paris Cité, Paris, France
Marcel Goldberg
Roles: Funding acquisition, Project administration, Writing – review & editing
Affiliation: Université Paris Cité, Université Paris-Saclay, Université de Versailles Saint-Quentin-en-Yvelines, INSERM UMS_011, Villejuif, France
ORICD: https://orcid.org/0000-0002-6161-5880
Marie Zins
Roles: Funding acquisition, Project administration, Writing – review & editing
Affiliations: Université Paris Cité, Université Paris-Saclay, Université de Versailles Saint-Quentin-en-Yvelines, INSERM UMS_011, Villejuif, France, Faculté de Médecine, Université Paris Cité, Paris, France
ORICD: https://orcid.org/0000-0002-4540-4282
Pierre Meneton
Roles: Conceptualization, Supervision, Validation, Writing – original draft
E-mail: [email protected]
Affiliation: INSERM UMR_1142, Sorbonne Université, Université Paris 13, Paris, France
ORICD: https://orcid.org/0000-0003-4611-1892
1. Reaven GM. The metabolic syndrome: time to get off the merry-go-round? J Intern Med. 2011;269(2):127–36. pmid:21129047
2. Mottillo S, Filion KB, Genest J, Joseph L, Pilote L, Poirier P, et al. The metabolic syndrome and cardiovascular risk a systematic review and meta-analysis. J Am Coll Cardiol. 2010;56(14):1113–32. pmid:20863953
3. Alshammary AF, Alharbi KK, Alshehri NJ, Vennu V, Ali Khan I. Metabolic syndrome and coronary artery disease risk: a meta-analysis of observational studies. Int J Environ Res Public Health. 2021;18(4):1773. pmid:33670349
4. Zhang F, Liu L, Zhang C, Ji S, Mei Z, Li T. Association of metabolic syndrome and its components with risk of stroke recurrence and mortality: a meta-analysis. Neurology. 2021;97(7):e695–705. pmid:34321360
5. Soriano-Moreno DR, Fernandez-Morales J, Medina-Ramirez SA, Coico-Lama AH, Soriano-Moreno AN, Zafra-Tanaka JH. Metabolic syndrome and risk of peripheral arterial disease: a systematic review and meta-analysis. Cardiol Rev. 2022.
6. Appelman Y, van Rijn BB, Ten Haaf ME, Boersma E, Peters SA. Sex differences in cardiovascular risk factors and disease prevention. Atherosclerosis. 2015;241:211–8.
7. Pradhan AD. Sex differences in the metabolic syndrome: implications for cardiovascular health in women. Clin Chem. 2014;60(1):44–52. pmid:24255079
8. Kuk JL, Ardern CI. Age and sex differences in the clustering of metabolic syndrome factors: association with mortality risk. Diabetes Care. 2010;33:2457–61.
9. Alexander CM, Landsman PB, Grundy SM. The influence of age and body mass index on the metabolic syndrome and its components. Diabetes Obes Metab. 2008;10:246–50.
10. Blanquet M, Legrand A, Pélissier A, Mourgues C. Socio-economics status and metabolic syndrome: a meta-analysis. Diabetes Metab Syndr. 2019;13:1805–12.
11. Tamashiro KLK. Metabolic syndrome: links to social stress and socioeconomic status. Ann N Y Acad Sci. 2011;1231:46–55. pmid:21884160
12. Wang Y, Yu L, Gao Y, Jiang L, Yuan L, Wang P, et al. Association between shift work or long working hours with metabolic syndrome: a systematic review and dose-response meta-analysis of observational studies. Chronobiol Int. 2021;38(3):318–33. pmid:33557625
13. Watanabe K, Sakuraya A, Kawakami N, Imamura K, Ando E, Asai Y, et al. Work-related psychosocial factors and metabolic syndrome onset among workers: a systematic review and meta-analysis. Obes Rev. 2018;19(11):1557–68. pmid:30047228
14. Sun K, Liu J, Ning G. Active smoking and risk of metabolic syndrome: a meta-analysis of prospective studies. PLoS One. 2012;7:e47791.
15. Sun K, Ren M, Liu D, Wang C, Yang C, Yan L. Alcohol consumption and risk of metabolic syndrome: a meta-analysis of prospective studies. Clin Nutr. 2014;33(4):596–602. pmid:24315622
16. Zhang D, Liu X, Liu Y, Sun X, Wang B, Ren Y, et al. Leisure-time physical activity and incident metabolic syndrome: a systematic review and dose-response meta-analysis of cohort studies. Metabolism. 2017;75:36–44. pmid:28927737
17. Fabiani R, Naldini G, Chiavarini M. Dietary patterns and metabolic syndrome in adult subjects: a systematic review and meta-analysis. Nutrients. 2019;11(9):2056. pmid:31480732
18. Zins M, Goldberg M, CONSTANCES team. The French CONSTANCES population-based cohort: design, inclusion and follow-up. Eur J Epidemiol. 2015;30(12):1317–28. pmid:26520638
19. Ruiz F, Goldberg M, Lemonnier S, Ozguler A, Boos E, Brigand A, et al. High quality standards for a large-scale prospective population-based observational cohort: constances. BMC Public Health. 2016;16(1):877. pmid:27557750
20. Grundy SM, Cleeman JI, Daniels SR, Donato KA, Eckel RH, Franklin BA, et al. Diagnosis and management of the metabolic syndrome: an American Heart Association/National Heart, Lung, and Blood Institute Scientific Statement. Circulation. 2005;112(17):2735–52. pmid:16157765
21. Hoertel N, Sanchez Rico M, Limosin F, Ménard J, Ribet C, Bonenfant S, 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(2):e021373. pmid:35023345
22. Meneton P, Hoertel N, Wiernik E, Lemogne C, Ribet C, Bonenfant S. Work environment mediates a large part of social inequalities in the incidence of several common cardiovascular risk factors: Findings from the Gazel cohort. Soc Sci Med. 2018;216:59–66.
23. Sanchez Rico M, Plessz M, Airagnes G, Ribet C, Hoertel N, Goldberg M. Cardiovascular burden and unemployment: a retrospective study in a large population-based French cohort. PLoS One. 2023;18:e0288747.
24. Goldberg M, Carton M, Descatha A, Leclerc A, Roquelaure Y, Santin G, et al. CONSTANCES: a general prospective population-based cohort for occupational and environmental epidemiology: cohort profile. Occup Environ Med. 2017;74(1):66–71. pmid:27884936
25. Toledano MB, Smith RB, Brook JP, Douglass M, Elliott P. How to establish and follow up a large prospective cohort study in the 21st century--lessons from UK COSMOS. PLoS One. 2015;10(7):e0131521. pmid:26147611
26. White IR, Carlin JB. Bias and efficiency of multiple imputation compared with complete-case analysis for missing covariate values. Stat Med. 2010;29(28):2920–31. pmid:20842622
27. Gatsonis C, Sampson AR. Multiple correlation: exact power and sample size calculations. Psychol Bull. 1989;106(3):516–24. pmid:2813654
28. Czernichow S, Kengne A-P, Stamatakis E, Hamer M, Batty GD. Body mass index, waist circumference and waist-hip ratio: which is the better discriminator of cardiovascular disease mortality risk?: evidence from an individual-participant meta-analysis of 82 864 participants from nine cohort studies. Obes Rev. 2011;12(9):680–7. pmid:21521449
29. Darsini D, Hamidah H, Notobroto HB, Cahyono EA. Health risks associated with high waist circumference: a systematic review. J Public Health Res. 2020;9(2):1811. pmid:32728557
30. Flint AC, Conell C, Ren X, Banki NM, Chan SL, Rao VA, et al. Effect of systolic and diastolic blood pressure on cardiovascular outcomes. N Engl J Med. 2019;381(3):243–51. pmid:31314968
31. Kannel WB. Historic perspectives on the relative contributions of diastolic and systolic blood pressure elevation to cardiovascular risk profile. Am Heart J. 1999;138(3 Pt 2):205–10. pmid:10467214
32. van Walree ES, Jansen IE, Bell NY, Savage JE, de Leeuw C, Nieuwdorp M, et al. Disentangling genetic risks for metabolic syndrome. Diabetes. 2022;71(11):2447–57. pmid:35983957
33. Rana S, Ali S, Wani HA, Mushtaq QD, Sharma S, Rehman MU. Metabolic syndrome and underlying genetic determinants-A systematic review. J Diabetes Metab Disord. 2022;21(1):1095–104. pmid:35673448
34. Scott R, O’Brien R, Fulcher G, Pardy C, D’Emden M, Tse D, et al. Effects of fenofibrate treatment on cardiovascular disease risk in 9,795 individuals with type 2 diabetes and various components of the metabolic syndrome: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study. Diabetes Care. 2009;32(3):493–8. pmid:18984774
35. Verma S, Al-Omran M, Leiter LA, Mazer CD, Rasmussen S, Saevereid HA. Cardiovascular efficacy of liraglutide and semaglutide in individuals with diabetes and peripheral artery disease. Diabetes Obes Metab. 2022;24:1288–99.
36. Festi D, Schiumerini R, Eusebi LH, Marasco G, Taddia M, Colecchia A. Gut microbiota and metabolic syndrome. World J Gastroenterol. 2014;20(43):16079–94. pmid:25473159
37. Geleijnse JM, Kok FJ, Grobbee DE. Impact of dietary and lifestyle factors on the prevalence of hypertension in Western populations. Eur J Public Health. 2004;14(3):235–9. pmid:15369026
38. Merlotti C, Morabito A, Pontiroli AE. Prevention of type 2 diabetes; a systematic review and meta-analysis of different intervention strategies. Diabetes Obes Metab. 2014;16(8):719–27. pmid:24476122
39. Global Burden of Metabolic Risk Factors for Chronic Diseases Collaboration (BMI Mediated Effects), Lu Y, Hajifathalian K, Ezzati M, Woodward M, Rimm EB, et al. Metabolic mediators of the effects of body-mass index, overweight, and obesity on coronary heart disease and stroke: a pooled analysis of 97 prospective cohorts with 1·8 million participants. Lancet. 2014;383(9921):970–83. pmid:24269108
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2025 Rico et al. This is an open access article distributed under the terms of the Creative Commons Attribution License: http://creativecommons.org/licenses/by/4.0/ (the “License”), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Background
Besides the usual characterization of metabolic syndrome as a cluster of markers arbitrarily defined by thresholds, it is unclear to which extent these markers as continuous traits are correlated with each other in the general population. The present study aimed to explore these correlations across a wide array of biological, social and behavioral characteristics.
Methods
The cross-sectional analyses were performed in a large population-based French cohort (CONSTANCES) of 159,476 adults in whom blood glucose, low-density lipoproteins (LDL) and high-density lipoproteins (HDL), triglycerides, body mass index, waist and hip circumferences, systolic and diastolic blood pressures were measured at the time of recruitment between 2012 and 2021. Correlations between each pair of continuous marker distributions were assessed by calculating raw and partial correlation coefficients (r).
Results
The same pattern of partial correlations is observed with little variation in all groups of sex, age, individual and parental histories of cardiovascular disease, diagnosis of metabolic syndrome, social position, work environment, lifetime unemployment exposure, smoking, non-moderate alcohol consumption, leisure-time physical inactivity and diet quality. This pattern is composed of strong and expected intercorrelations between systolic and diastolic blood pressures (r ranging from 0.62 to 0.74), between body mass index and waist (r from 0.50 to 0.63) and hip (r from 0.58 to 0.70) circumferences and between waist and hip circumferences (r from 0.07 to 0.19). It also includes intercorrelations of systolic blood pressure with waist (r from 0.10 to 0.21) and hip (r from −0.07 to −0.12) circumferences and with blood glucose (r from 0.09 to 0.15), those of triglycerides with blood glucose (r from 0.07 to 0.16), LDL (r from 0.24 to 0.33), HDL (r from −0.20 to −0.29) and waist circumference (r from 0.07 to 0.15), and finally those of waist and hip circumferences with blood glucose (r from 0.09 to 0.17 and from −0.08 to −0.13) and HDL (r from −0.12 to −0.24 and from 0.08 to 0.18).
Conclusions
These results show that metabolic syndrome markers are correlated with each other whatever the biological, social or behavioral characteristics of individuals. They suggest that it makes sense to systematically consider these markers all together rather than separately in terms of etiology, prevention and treatment of metabolic diseases and cardiovascular risk in the general population.
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer