Clustering of metabolic syndrome components in a Middle Eastern diabetic and non-diabetic population

Metabolic syndrome (MetS) encompasses a cluster of coronary heart disease and diabetes mellitus risk factors, including abdominal obesity, glucose intolerance, dyslipidemia and elevated blood pressure [1, 2]. Subjects with MetS tend to have high insulin levels, reflecting greater insulin resistance [1‐3]. Insulin resistance or hyperinsulinemia has been suggested to be the underlying characteristic of MetS, although a central role for insulin resistance is still controversial [3]. The understanding of how MetS components cluster together can help clinicians interpret the MetS pathophysiology and develop effective strategies for identifying and preventing the inherent risk of coronary heart disease and diabetes mellitus. The multitude of clinical and biochemical alterations resembling the MetS, the strong cross-linkage of involved pathways and multiple feedback mechanisms, complicate the identification of the events which lead to the cascade of disorders that characterize the syndrome [4]. Moreover, statistically strong intercorrelations among various features of MetS, also complicate establishing independent associations using standard multivariate statistical models [2, 3]. Factor analysis is potentially a way of advancing this body of research by explaining the correlation between the components of the syndrome in terms of a small set of latent factors, providing insights into the underlying process. Exploratory factor analysis is a hypotheses generation method, in which the number of factors is essentially unknown and has to be determined from data. Extracting one factor implies that a single underlying pathophysiologic mechanism contributes to the appearance of the syndrome. If two or more factors were extracted, the existence of a unique syndrome would be called into question [3].

Previous studies have entered different numbers of variables ranged from 4 to 21 in factor analysis and extracted from 1 to 7 factors [2, 5‐13]. Some investigations have identified three or four factors, implying a possible heterogeneity of the MetS [2, 7, 9‐11]. Since there appears to be ethnic differences in the expression of MetS [1], different findings may arise on the basis of sample collection. For example, hypertension may not be associated with MetS in American Indians and African Americans [14]. Variations in the extracted factors can also be result of discrepancies in numbers and nature of considered components. For example some studies include insulin levels while others do not; the same is true for uric acid, leptin and waist to hip ratio [2, 5‐13]. Since factor analysis extracts factors due to the interrelatedness of measured variables, using two or more measures for the same trait (e.g. systolic blood pressure [SBP] and diastolic blood pressure [DBP]) leads to find more factors than expected [15]. These variables compared to other variables are highly associated with each other, thus tend to load on a separate factor. This phenomenon can also be observed with fasting and postprandial glucose or with high density lipoprotein cholesterol (HDL-C) and triglycerides (TG) [5, 16].

Factor analysis of the MetS has been the focus of little attention in the Middle Eastern ethnicity; furthermore, the clustering of MetS components is yet to be more explored in the diabetic patients. In this study, for the first time we performed factor analysis on the components of MetS in a large sample of Iranian diabetic and non-diabetic participants. MetS can be defined according to Adult Treatment Panel (ATPIII) [1] or International Diabetes Federation (IDF) criteria [17]. We recently proposed that the optimal waist circumference cut-off points for the diagnosis of IDF defined MetS in Iranian adults would be different from those of other populations [18]. In this study, we also explored the influence of the extracted factors on different definitions of MetS.

Factor analysis of components revealed two factors in model 1. In diabetic participants the first factor was related to waist circumference, HOMA-IR and SBP while in non-diabetic participants it was also associated with TG. The second factor was determined positively by HDL-C and negatively by TG. Factor 1 and 2 were directly and inversely associated with MetS, respectively. When TG and HDL-C were replaced by a composite measure, which included both of them (i.e. TG to HDL-C ratio), one factor was found, confirming previous observations on the monofactorial structure of MetS [23]. Furthermore, these findings are consistent with the concept that using two or more measures for the same trait results in an overfactored model.

As previously mentioned, variations in sample collection, number and nature of selected components and also in methods used for interpretation of analysis preclude formal direct comparisons, but an overview elucidates some common patterns. In several previous studies insulin resistance loaded on the same factor with measures of obesity including weight, body mass index and waist circumference [2, 5-9, 11, 13]. Furthermore, measures of obesity were more likely to belong to the first extracted factor and to have the highest factor loadings [6-9, 13]. It has been demonstrated that visceral fat via its tendency to release free fatty acids, stimulates hepatic glucose production and accentuates insulin resistance [24]. Moreover, visceral fat cells release some agents to the blood (e.g. plasminogen activator inhibitor-1, interleukin-6 and tumor necrosis factor-α) that are possibly linked to some aspects of insulin resistance [24]. These interrelated pathophysiologic mechanisms, underlying the association of insulin resistance and obesity, may play a principal role in the expression of MetS and its complications including coronary heart disease and diabetes.

MetS is believed to be related to insulin resistance which can be scored by measuring fasting insulin concentration [1‐3]. However, due to the fact that in diabetic patients increase in blood glucose and insulin resistance is accompanied by impaired insulin secretion [4], and since there was no significant difference in fasting insulin concentration between diabetic and non-diabetic participants (Table 1); fasting insulin is not a reliable index of insulin resistance in diabetic patients. It has been demonstrated that there is a good correlation between estimates of insulin resistance derived from HOMA-IR and from the euglycemic clamp [25]. Moreover HOMA-IR is a reliable index in diabetic patients and has been widely used as an estimate of insulin resistance in both non-diabetic and diabetic subjects [25, 26]. Besides these advantages, HOMA-IR has also some limitations. For instance, in Korean population, the validity of HOMA-IR as a surrogate measure of insulin resistance in lean type 2 diabetic subjects with the insulin secretory defect is disputed [27]. In model 1 of our analyses, HOMA-IR loaded primarily on factor one, but the factor loading of HOMA-IR on factor 2 was -0.35 in non-diabetic participants which is close to the loading threshold, |0.40|. Further, TG was associated with both extracted factors in non-diabetic participants (Figure 1). Many studies have used lower loading thresholds or noted the effect of relaxing the threshold on the findings [6, 7, 12, 15]. Even Cureton and D'Agostino suggested a loading threshold as low as 0.2 for factor interpretation [28]. In model 1, association of HOMA-IR and TG with both extracted factors in non-diabetic participants shows that the two factors are related to each other and are not totally independent. Consistent with our findings, in most studies measures of insulin resistance load on more than one factor, suggesting insulin resistance as a unifying theme underlying the MetS [3]. On the other hand, Shen et al. found that TG and HDL-C are part of the same risk component for MetS and inclusion of both may be redundant [29]. In this regard, TG to HDL-C ratio is reported to be helpful in determining risk of coronary heart disease [30]. In model 2, substitution of TG and HDL-C with TG to HDL-C ratio led to extraction of just one factor. Mannucci et al. reported similar results and revealed that considering TG and HDL-C as two different components of MetS results in overestimation of the role of dyslipidaemia in MetS [5]. These findings suggest that the model 2 of our analyses (one-factor model) is also plausible and support previous studies suggesting one underlying factor (perhaps insulin resistance) for the syndrome [23], the point which is recently substantiated in experimental studies [31].

Insulin resistance and hyperinsulinemia lead to hypertension via a number of mechanisms, including renal sodium and water retention, rennin-angiotensin-aldosterone system, plasma noradrenaline and sympathetic nervous system activity [32, 33]. Since subjects' age range is wide (18-75 years) it is more conceivable to use just SBP instead of mean arterial pressure [34]. In model 2, the factor loading of SBP was lowest among components of MetS implying it as the most tenuously linked component. Review of previous studies shows that in most analyses measures of blood pressure load on a unique and separate factor, which does not overlap with other factors of MetS [2, 9, 11‐13], revealing weak relation between blood pressure and other components of MetS [3, 29]. Even, such relation may not be evident in some ethnic groups like American Indians and African Americans [14]. Moreover, we found that SBP is less correlated with components of MetS especially in diabetic subjects (Table 2). These findings suggest that blood pressure is less influenced by the common pathophysiologic pathway which underlies the MetS.

Ethnic differences in the expression of MetS are linked to different appropriate waist circumference cut-offs for determination of abdominal obesity. It has been found that Asians in comparison to Caucasians are more prone to obesity related comorbidities even at lower waist circumference values [35]. We recently showed that waist circumference cut-offs recommended by IDF are not appropriate in Iranian population and should change to 90 cm in both men and women [18]. In this study we found that IDF definition modified by our recommended waist circumference cut-offs is strongly associated with extracted factors. As table 4 shows, factor 2 was inversely associated with MetS and therefore had protective roles against it, which seems reasonable because factor 2 was mainly characterized by significant positive factor loading for HDL-C.

The major limitation, relevant to our results interpretation, is use of cross-sectional data. However, consistent with our analysis most previous studies have used cross-sectional study design to evaluate the interrelatedness of MetS components. Cross-sectional analyses provides information at a single point in time, nevertheless progressive and reciprocal mechanisms are likely to be involved in the appearance of MetS. Therefore, longitudinal studies are suggested to improve better our understanding of MetS.

This study revealed that in both diabetic and non-diabetic individuals the concept of a single underlying factor (pathophysiologic pathway) for the expression of MetS is plausible. Further, we showed that similar to ATPIII and original IDF definitions; modified IDF definition is also strongly associated with identified factors. Nevertheless, among components of MetS blood pressure is less influenced by the common factor and could be also mediated by other pathophysiologic processes.