Description and prediction of the development of metabolic syndrome in Dongying City: a longitudinal analysis using the Markov model

Metabolic syndrome (MS) is a clustering of metabolic risk factors, including obesity, hypertension, dyslipidemia, and insulin resistance [1, 2]. MS can raise the risk of cardiovascular disease (CVD) which is the leading killer in China [3‐5]. Evidence also suggests that MS is strongly correlated with polycystic ovarian disease, nonalcoholic fatty liver disease, and some cancers [6‐8]. In fact, MS is increasingly becoming a major worldwide clinical and public health issue [9‐14].

Due to the great harm caused by MS, vast studies have been done on it. Previous studies mainly focused on searching for early MS’ biomarkers and risk factors, such as white blood cell count (WBC), serum uric acid (UA), gamma-glutamyl transpeptidase (GGT), alanine aminotransferase (ALT), physical inactivity, alcohol intake, and smoking habits [15‐20]. Only a few studies focused on the onset process of MS [21, 22]. MS is defined as the existence of at least three of the four components at the same time, thus there are 16 different states and 256 transitions between states. Therefore, stating the history of MS and searching for the most likely component contributing to start the cascade of confusions are preconditions for preventing its development. Haring et al. used a network-based approach to show the prevalence and progression of MS components and their changes [22], but these predictions faced many barriers. The Markov model which is frequently used to represent a random process changing with time is an admitted method to simulate the natural history of chronic diseases [23]. Lee-Ching Hwang et al. applied a Markov model approach to predict the development of MS [21]. However, this study was only with young people and without intermediate process conditions, it was limited. Therefore, we conducted a Markov model with a six-year follow-up health check-up including different genders and age groups to describe the natural history of MS, determine gender and age differences in the natural progression of MS components and to predict the effect of different initial states on the development of MS.

The general information of the baseline and each follow-up year was shown in Table 1. During the six-year follow-up, 7510 individuals were included in this study and the total follow-up years were 28,059 person-years with a mean 3.74 follow-up years per subject. MS prevalence increased with greater time since baseline. Additional file 1: Table S1 displayed the basic characteristics of the 7510 study individuals stratified by gender and age. All variables were associated with age.

As well-known methods for simulating the history of chronic diseases, Markov models were used in many studies. Feldman et al. used a longitudinal multistate Markov model to study the causal relationship between allergic sensitization and rhinovirus wheezing in 285 children at high risk for allergic disease and asthma [25]. Similarly, a Markov model analysis was used on a population-based cohort to depict clinical course and costs of care for Crohn's disease [26]. The patients are divided into different disease states and transition probability between any two states is given in the Markov model which applies to medical areas [27]. Seven states were included in our study and mutual conversion was allowed among them. The Markov model was just able to achieve these transformations and to simulate the history of MS. Therefore, using a routine health-check longitudinal cohort, our study mainly focused on describing the natural history of MS, searching for the most likely component contributing to start the cascade confusions of MS, and predicting the effect of different initial states in 10 years by a reversible multi-state Markov model.

It is still controversy about which state the process of MS is most likely to begin with. A network-based study on a five-year follow-up longitudinal cohort identified central obesity and hypertension as the predominant MS risk factor cluster [22]. Lee-Ching Hwang et al. claimed that abdominal obesity or low HDL were the most likely factors to initiate the progress of MS in women [21]. Similarly, a Baltimore Longitudinal Study on Ageing demonstrated that abdominal obesity, low HDL cholesterol, or high triglycerides were the predictors of incidence of MS [28]. Limitations of these studies were that their subjects were confined to certain gender or age groups. Individuals in our research were divided into four age groups, 18–40 years, 40–49 years, 50–59 years, and ≥60 years, in two genders. We confirmed that the process of MS was most likely to begin with overweight or obesity and dyslipidemia in the young population (men <60 years and women <50 years). Also, we verified that the most likely beginning states of MS included hypertension state and 2-component state besides the above two states in the elderly population (men ≥60 years and women ≥50 years).

In the young people, obesity was defined as a main contributor in the process of MS, because it was suspected as a risk factor for hypertension, dyslipidemia and hyperglycemia [29]. Dyslipidemia could cause hypertension and hyperglycemia by mediating of free fatty acids, but the causality between dyslipidemia and obesity is still controversial. However, the beginning state changed with the growing age. The hormone level that affects individuals’ metabolism changes with age especially during menopause which results in the difference between the young and elderly. Chedraui Pet al. demonstrated that postmenopausal women with MS displayed higher IL-6 (inflammation) and lower urokinase-type plasminogen activator levels (endothelial dysfunction) which were mainly related to metabolic abnormalities and led to the occurrence of 2 components of MS at the same time [30]. Our research also found that men were more prone to develop MS than women of the same age while increasing age was an important factor inducing MS. The result was accordant with Scuteri Aet al. who studied 34,821 subjects from 12 cohorts in 10 European countries and one cohort from the USA in the MARE (Metabolic syndrome and Arteries RE search) Consortium [31].

We conducted a 10 years prediction with individuals beginning with different states of MS according to the existing transition probability. This is particularly important to prevent the occurrence of MS. In the predictions, individuals with the isolated hyperglycemia state had a greater chance to develop MS than the other isolated states, and almost reached the level of the individuals beginning with the 2-component state. Insulin resistance is the defined pathogenesis of diabetes and it commonly links the correlation to the other MS components. Several studies confirmed that four components of MS, obesity, hypertension, dyslipidemia, and hyperglycemia coexisted and influenced each other [32, 33]. There are some overlapping metabolic pathways in the pathogenesis of the four components of MS. Diabetes is often accompanied by hypertension and they share conjunct pathways such as the Renin-angiotensin-aldosterone System, Sympathetic Nervous System, adipokines, inflammatory pathway, and oxidative stress [34, 35]. A study conducted in the USA pointed out that hypertension occurs in approximately 50% to 80% of patients with type 2 diabetes and in 30% of patients with type 1 diabetes [36]. A search in Hong Kong was in line with it, as only 42% of people with diabetes could maintain normal blood pressure [37]. Hyperglycemia and dyslipidemia often occur in the same individual but its mechanism still remains controversial. Recently, it is suggested that cholesterol homeostasis plays an important role for beta-cells to perform adequate insulin secretion and low HDL cholesterol levels are normally related with hyperglycemia and type 2 diabetes [38]. Although there is no theory supporting that hyperglycemia could lead to obesity, central obesity had been confirmed to be the most relevant predisposing factor for insulin resistance [22, 39]. It should be particularly pointed out that glycemic control ought to attract more attention and that the individuals once diagnosed with hyperglycemia should receive active defense to prevent the emergence of the other components of MS.

We adopted CDS’ criteria to diagnose MS instead of international standard criteria in our study. Compared to Europeans and Americans, Chinese people have their own body shape characteristics. Therefore, CDS developed their own criteria which was also widely used in other studies [17, 18]. Previous study convinced that these two MS diagnostic criteria were in good accordance [24]. Simultaneous, considering the fact that we lack of waist circumference measurement data, CDS’ criteria could make our estimation more precise. Although CDS’ criteria we adopted may result slight difference in diagnoses of MS components compared to the international standard criteria. This difference would hardly effect on the overall developing trend of MS.

There are several limitations in our study. First, the person loss of follow-up was included in our study, and it may affect the interpretation of the transitional probability. We had used the averaging method to furthest reduce the impact of it. Secondly, the natural history of MS development may be not completely natural because of small part of interventions from the health services. Thirdly, mortality was not included in our model for the lack of the region specific information. Considering the general situation of mortality in our country, the influence is very small.

The Markov model is able to give a good description of the evolutionary history of MS and to predict the future course based on the past evidence. Men are more prone to develop MS than women of the same age. The occurrence of the MS process mostly began with overweight or obesity and dyslipidemia in young people. In the elderly population, many individuals initiating with hypertension or 2 components besides the above two states. Individuals with the isolated hyperglycemia had a greater chance to develop MS than the other isolated MS’ components. These results indicate the importance of monitoring the component of MS.