Scoring systems of metabolic syndrome and prediction of cardiovascular events: A population based cohort study

Abstract Background and Aims Continuous scoring systems were developed versus traditional dichotomous approaches to define metabolic syndrome. The current study was carried out to evaluate the ability of scoring systems to predict fatal and nonfatal cardiovascular events. Materials and Methods The data of 5147 individuals aged 18 years or more obtained from a population‐based cohort study were analyzed. The occurrence of atherosclerotic cardiovascular disease (ASCVD) in the period of 7 years follow‐up was considered as the associated outcome. Joint Interim Statement (JIS) definition, as a traditional definition of metabolic syndrome (MetS), and two versions of MetS scoring systems, based on standardized regression weights from structural equation modeling (SEM) and simple method for quantifying metabolic syndrome (siMS) were considered as potential predictors. Results The scoring systems, particularly, based on SEM, were observed to have a significant association with composite cardiovascular events (HR = 1.388 [95% CI = 1.153–1.670], p = .001 in men and HR = 1.307 [0.95% CI = 1.120–1.526] in women) in multiple Cox proportional hazard regression analyses, whereas the traditional definition of MetS did not show any significant association. While both two scoring systems showed acceptable predictive abilities for cardiovascular events in women (MetS score based on SEM: area of under curve [AUC] = 0.7438 [95% CI = 0.6195–0.7903] and siMS: AUC = 0.7207 [95% CI = 0.6676–0.7738]), the two systems were not acceptable for identifying risk in men. Conclusion Unlike the dichotomous definition of MetS, the scoring systems showed an independent association with cardiovascular events. Scoring systems, particularly those based on SEM, may be useful for the prediction of cardiovascular events in women.


| INTRODUCTION
Cardiovascular diseases (CVD) are considered as the leading cause of death worldwide. 1 Based on a recent data, CVD are the causes of one-third of all deaths worldwide. 2 The age standardized deaths related to CVD have reduced, particularly in high-income countries by preventive measures and high-quality interventions in spite of increase in absolute CVD deaths worldwide. 3,4 Despite a large decrease in the burden of CVD in the United States, the huge disparities in the total burden of CVD among different US states can be attributed to the differences in exposure to some modifiable risk factors. 5 Some of these well-known and modifiable cardiovascular risk factors including high blood pressure, abdominal obesity, high fasting blood sugar (FBS), low high-density lipoprotein (HDL) levels, and obesity are used to define metabolic syndrome (MetS). A cluster of these multiple cardio-metabolic abnormalities is defined as metabolic syndrome. [6][7][8] Hence, a high association between the MetS and CVD is expected. 9 limitations. [18][19][20] For instance, related models mostly are based on data that have already become out of date regarding the changes in preventive interventions and changes in the rate of endpoints. 21,22 Furthermore, these models mostly were developed in western countries and validated in small populations. As a result, the current study was conducted to evaluate the ability of the metabolic scoring systems, specific to attributes of the Iranian population, to predict the fatal and nonfatal CVD events.

| Study population
The present cohort study was carried out in two phases: Phase I in 2009-2010 and Phase II in 2016-2017. The study was performed on the population of Amol city, a relatively populated city in the central area of northern Iran. The sampling frame of our initial cohort study was based on the data from Health houses where an exact sampling frame is obtainable due to primary healthcare services delivered in these sittings. We divided the sampling frame into 16 strata based on gender and age groups, including 10-19, 20-29, 30-39, 40-49, 50-59, 60-69, 70-79, and 80-89 years. The size of each stratum in the sampling frame was calculated in proportion to the size of population in the same stratum using a stratified probability proportion sampling strategy. After 7 years, participants from the previous study in Phase I were invited again to participate in Phase II of the study. Since 2009-2010 up to the beginning of Phase II in 2016-2017, the study participants were annually contacted to collect related information about probable outcomes. A schematic view of the study population is shown in Figure 1.
The comprehensive evaluations of Phase II of our cohort started in 2016 and continued in 2017. This included a detailed evaluation of demographic, anthropometric, and laboratory data in addition to providing the related outcomes of people based on associated medical documents. It is worth noting that we confirmed and modified, where necessary, the related outcomes based on the data from various authentic sources, such as valid documented data from hospitals, clinics, physician offices, and also medical records, particularly if there was any inconsistency in the findings.

| Anthropometric parameters and blood pressure measurements
Participants' heights were measured via a nonstretchable meter. This was performed when they were in the upright position with a small gap between the legs (usually 10 cm) and their back of heads, shoulder blades, buttocks, and heels in contact with the wall. Also, participants' weights were measured using a calibrated scale with a precision of 100 g. Waist circumference (WC) was measured by trained staff from the narrowest point between the lower borders of the rib cage and the iliac crest using a nonstretchable tape measure.
Participants' blood pressure was measured by trained staff using a mercury sphygmomanometer after the participant sat on a chair for a minimum of 5 min of physical inactivity in a quiet room. After inflation of the cuffs 20-30 mmHg above the point of disappearance of the radial pulse, the cuffs were deflated at a rate of approximately 2-3 mmHg. Thus, the appearance and disappearance of Korotkoff sounds were considered as the systolic (SBP) and diastolic (DBP) blood pressures, respectively. The average of two measurements of blood pressures for each participant was considered as the associated blood pressure of the participant.

| Biochemical measurements
Totally, 10 ml of whole blood was obtained from each participant using a serum separator tube (SST; tiger top tube).

| Metabolic syndrome
Dichotomous version of metabolic syndrome definition was defined based on Joint Interim Statement (JIS). 23

| Outcomes
The occurrence of atherosclerotic cardiovascular disease (ASCVD) was considered the associated outcome of the present study. The definition of ASCVD was considered as a history of nonfatal acute myocardial infarction and ischemic heart disease death and fatal and nonfatal cerebrovascular accident. 24 The outcome data were annually collected from the participants, although a comprehensive assessment in Phase II of the cohort was CVD events and hospital discharge records were also evaluated and verified. We actively contacted the medical centers where the patients were admitted, if a medical record did not seem to be correct. Each inconsistent finding between the outcomes of the comprehensive assessment and the annually obtained outcomes data were modified based on valid documented data.
A 12-lead electrocardiogram (ECG) was performed by trained nurses for all participants who participated in Phase II of the cohort study. Consequently, if any ECG abnormality was seen in the participants from the Phase II cohort, their abnormality was not included as the outcome, the participant was comprehensively examined by the internist of our team, and was also referred to an expert cardiologist to rule out any silent CVD events in the follow-up periods. Finally, all associated outcomes were confirmed by the internist of our cohort study team.

| Statistical analysis
We conducted a confirmatory factor analysis (CFA) using maximum rates in the prediction of related CVD outcomes. An AUC of >0.5 to <0.7 was considered as poor ability, AUC >0.7 to <0.8 as acceptable ability, AUC >0.8 to <0.9 as an excellent ability, and AU >0.9 to <1.0 was considered as outstanding ability. Also, an AUC = 0.5 indicated "no" ability while an AUC = 1 indicated a "perfect" ability. These considerations were based on Hosmer and Lemeshow's guidelines. 29 In ROC analyses, all statistical analyses were conducted using  Table 1 shows basic characteristics of the study population based on sex. While the mean age (p = .003), weight (p < .001), DBP (p = .022), and SBP (p < .001) were significantly higher in men, BMI (p < .001), T A B L E 1 Basic characteristics of the study population of primary phase of cohort Significance level for the difference between men and women was considered p < .05. FBS (p < .000), cholesterol (p < .001), HDL-C (p < .001), and LDL-C (p < .001) were significantly lower in men as compared with those in women. No significant difference in terms of mean WC and TG was observed between men and women; however, both the MetS and siMS scores were significantly higher in women (p's <.001). Additionally, in women, there was no statistical difference between the two scoring systems for fatal, nonfatal, and composite CVD events with both scores being excellent and acceptable predictors for fatal and composite events, respectively (Table 2 and Supplementary Figure 2 Figure S2). Significance level for the difference between scoring systems was considered p < .05.

| RESULTS
T A B L E 3 Simple and multiple Cox regression proportional analyses on the time of the occurrence of CVD events in which the MetS scores were the predictor were considered as potential predictors. According to our results, while the scoring systems showed a significant association with CVD events in multiple Cox models removing potential mediators, including age, DBP, LDLc, and smoking status, we found that the particularly continuous scoring systems. [14][15][16][17] The present study also compared the predictive ability of two versions of the scoring systems of the MetS in the prediction of fatal and nonfatal CVD events in a cohort study with the follow-up period of 7 years. Overall, the results showed that the scoring system based on SEM had a better predictive ability compared with the siMS scoring system. The ability of Mets scoring systems was excellent for the prediction of fatal CVD events in women. In men, while scoring system based on SEM showed an acceptable ability for the prediction of fatal CVD events, the ability of the siMS to predict the same events was considered poor. As for nonfatal events, both scoring systems showed poor predictive abilities in men. However, for women, the scoring system based on SEM showed an acceptable ability for the prediction of nonfatal CVD events relative to the siMS scoring of poor predictive ability.
Our results in women revealed that Mets score based on SEM had an ability approximately similar to that of MetS score, as evaluated by Yang et al. 16  Our study had certain limitations, which should be taken into consideration before any generalization. Although a comprehensive evaluation was performed to obtain the related outcomes, the lack of ECG data in the Phase I of the cohort may have resulted in the possibility that some participants may have developed silent CVD events during follow-up and that these outcomes would have been excluded from our data. However, our evaluation did detect 21 cases of ECG changes that were not included as outcomes in our study. These participants were evaluated by the team internist and also referred to an expert cardiologist to determine their possible outcomes. Furthermore, a scenario analysis including the outcomes of these participants did not show any significant and reportable changes. Also, in the present study, we did not consider the MetS scores based on age due to nonfitting of the single factor of the model wherein age was also applied. As mentioned previously, age is a strong prognostic factor for the development of cardiovascular events and thus the predictive ability of models based on age can be improved.
Nevertheless, our study provided insight on parameters related to the prediction of CVD, particularly the ability of the scoring systems in women and its implications for public health initiatives in northern Iran populations.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT
The data supporting this study's findings are available from the corresponding author upon reasonable request.