Comparative Analysis of Metabolic Syndrome Diagnostic Criteria and Its Effects on Prevalence in a Multiethnic Population

Abstract

Background: Accurate prevalence measurement and diagnosis to prevent type 2 diabetes mellitus and cardiovascular disease cannot occur without consistent diagnostic criteria that can be applied to varying populations.

Objective: The objective of this study was to determine the prevalence of metabolic syndrome in Caucasian, Filipino, Native Hawaiian, and Japanese populations utilizing different definitions.

Methods: This study utilized cross-sectional study data from the Native Hawaiian/Multiethnic Health Research Project, collected from a population living in Kohala, Hawai‘i. The National Cholesterol Education Program–Adult Treatment Panel III (NCEP-ATPII), International Diabetes Federation (IDF), and World Health Organization (WHO) definitions were utilized, and each of the 1452 participants were evaluated on the criteria for metabolic syndrome based on all three definitions. Additionally, the average biomarker values associated with the diagnosis were taken for each ethnic group represented in the study and compared with Caucasians.

Results: The overall prevalence of metabolic syndrome in this population varied from 22.31% to 39.05% using the different definitions. Ethnic disparities also occur, implying that certain populations are more prone to having severe abnormalities than others—shown when comparing the average biomarker values associated with metabolic syndrome diagnosis. Of all ethnic groups included in the study, Caucasians had the lowest prevalence of metabolic syndrome, while part-Hawaiians had the highest prevalence. Additionally, within the same ethnic group, the definitions yielded varying prevalence values.

Conclusions: This implies that discrepancies exist among the criteria alone. Implications of this study revolve around not only the correct definition to apply to the population being studied but also the most accurate way to detect certain biomarker abnormalities to accurately assess the prevalence of metabolic syndrome in a multiethnic population.

Background

Metabolic syndrome is a cluster of symptoms occurring together, which significantly increases the risk of developing type 2 diabetes mellitus (T2DM) and cardiovascular disease (CVD). Metabolic syndrome is associated with the doubling of the likelihood of developing CVD within 5 to 10 years, a five-fold increase in the likelihood of developing T2DM,¹ and an increase in mortality rates by 1.6 times.² With chronic disease being an increasingly alarming public health issue, interventions geared toward prevention are key as long-term treatment is needed to decrease chronic disease prevalence.

Accurately diagnosing metabolic syndrome gives clinicians a better indication of the risk for both diabetes and CVD as each component factor contributes to greater risk.² However, although there are five main risk factors involved in the recognition and definition of metabolic syndrome: (1) abdominal obesity; (2) reduced high-density lipoprotein cholesterol (HDL-C); (3) elevated triglyceride levels; (4) high blood pressure; and (5) glucose intolerance as an indicator of insulin resistance,³ there is no consistent set of diagnostic criteria. The International Diabetes Federation (IDF) has estimated that at least one-quarter of adults in the United States have metabolic syndrome.⁴ However, prevalence largely depends on the geographic area studied; the composition of the population studied regarding age, sex, race, and ethnicity; and the definition being used.⁴ One investigation of prevalence in Asian American populations reported that these individuals were commonly not reaching threshold values for obesity and other vital risk factors required for metabolic syndrome diagnosis, resulting in an overall underestimation of metabolic syndrome burden. This underestimation is also seen in non-Hispanic Black males.⁵

With the risk that metabolic syndrome poses to public health, it is of great importance that a common definition is used to avoid confusion in clinical and educational settings. This study investigates how the prevalence of metabolic syndrome varies in Japanese, Filipino, Caucasian, and Native Hawaiian populations, utilizing existing diagnostic criteria.

Methods

A consensus of definitions and diagnostic criteria is of vital importance to correctly estimate prevalence. Metabolic syndrome diagnostic criteria from the National Cholesterol Education Program–Adult Treatment Panel III (NCEP-ATPII),⁶ World Health Organization (WHO),⁷ and IDF⁸ were applied to a multiethnic population in Kohala, Hawai‘i.⁹ The Native Hawaiian/Multiethnic Health Research (NHHR) project involved a cross-sectional study conducted from 1997 to 2000, looking specifically at 1452 residents of north Kohala, Hawai‘i.⁹ Participants in this study were contacted over the phone using a cross-referenced directory and attended health screening appointments.⁹ Data collection from this study included demographic information, anthropometric measures, and blood and urine sample data.¹⁰ Specific biometrics utilized in this study included the waist circumference, waist–hip ratio, body mass index (BMI), systolic and diastolic blood pressure, fasting glucose, 2-hr postprandial blood glucose, HDL-C levels, triglyceride levels, microalbumin levels in urine, and lnHOMA levels. Ethnic ancestry was determined using self-report only. All procedures were performed with the approval of the institutional review board of the [University of Hawaii at Manoa], CHS# 9406 “Selective Excellence in Health-Related Research,” and all procedures followed were in accordance with the ethical standards of the IRB and the Helsinki Declaration of 1975, as revised in 2000. Informed consent, both written and verbal, was obtained from all patients included in the study (Supplementary Appendices SA1 and SA2).

The prevalence of metabolic syndrome using the three definitions was assessed for Caucasians, Japanese, Filipinos, those of Hawaiian/part-Hawaiian ancestry, and those of other or mixed non-Hawaiians ethnicities using health screening data from the survey conducted by Grandinetti et al.⁹ Coding for diagnosis based on these criteria was done in SAS University Edition (2018) based on the definitions and the IDF ethnic-specific waist circumference values.

Statistics

General linear modeling was used to estimate least squared mean values for all biomarkers utilized in assigning a diagnosis of metabolic syndrome for each ethnic group using SAS University Edition and JMP13. Dunnett's post hoc test was run to determine statistically significant differences for each biomarker in all ethnic groups compared with Caucasians.

A general linear model was used to estimate the prevalence of metabolic syndrome for each ethnicity and using each definition, respective 95% confidence intervals (CIs) were calculated. Dunnett's post hoc test was employed to compare each ethnic group's prevalence with the prevalence of Caucasians for each definition to determine any statistically significant differences. Additionally, a general linear multivariate model was used to test for a statistical interaction between ethnicity and the various definitions used to estimate prevalence.

Results

The overall prevalence rates of metabolic syndrome in the Kohala population reported by NCEP-ATPIII, IDF, and WHO were 31.20%, 39.05%, and 22.31% respectively, providing evidence that definition alone results in greatly varied prevalence. Both Tables 1 and 2 detail sample demographics of the 1452 participants in the Kohala study. All ethnic groups (Filipino, part-Hawaiian, and Japanese) were significantly different from Caucasians with regard to age. Table 2 summarizes the anthropometric and biometric measures related to all definitions used to diagnose metabolic syndrome by ethnicity. People of Filipino and Native Hawaiian ancestry had the most significantly different anthropometric and biometric measures when compared with Caucasians. Interestingly, people of Japanese ancestry had a much higher 2-hr glucose level in comparison with their fasting glucose when compared with other ethnic groups (134.96 mg/dL for 2-hr glucose compared with a 108.45 mg/dL fasting glucose level) and this was the only ethnic group that had only a significantly different 2-hr glucose level instead of a significantly different 2-hr glucose level and fasting glucose level. Additionally, people of Japanese ancestry presented with lower waist circumferences and BMI values when other measures were abnormal. People of Hawaiian ancestry had greatly elevated waist circumferences and BMI values in comparison with all other ethnic groups.

Table 1.

Sample Demographics

	Caucasian	Filipino	Part-Hawaiian	Japanese	Mixed/other
n (%)	295 (20.32)	186 (12.81)	527 (36.29)	190 (13.09)	254 (17.49)
Females, n (%)	147 (49.83)	114 (61.29)*	293 (55.60)	97 (51.05)	136 (53.54)
Males, n (%)	148 (50.17)	72 (38.71)*	234 (44.40)	93 (48.95)	118 (46.46)
Mean age, years (standard deviation)	49.40 (12.43)	53.69 (16.17)**	44.75 (14.63)***	58.87 (16.41)***	47.52 (16.55)

Use of Dunnett's post hoc test (control = Caucasians), ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

Table 2.

Comparison of Age-Adjusted Mean Biomarker Values as Related to Metabolic Syndrome Diagnosis

	Caucasian		Filipino		Part-Hawaiian		Japanese		Mixed/other
Biomarkers	Mean	SE	Mean	SE	Mean	SE	Mean	SE	Mean	SE
Diastolic BP (mmHg)	72.655	0.610	77.637***	0.774	80.320***	0.467	77.283***	0.780	76.707***	0.658
Systolic BP (mmHg)	116.249	0.978	128.362***	1.243	132.138***	0.750	129.272***	1.252	124.964***	1.056
Triglycerides (mg/dL)	109.771	7.185	160.677***	9.125	159.029***	5.465	163.857***	9.194	134.272	7.751
HDL (mg/dL)	52.579	0.867	50.124	1.101	45.026***	0.659	50.587***	1.106	50.667	0.932
Fasting glucose (mg/dL)	99.025	1.821	108.014***	2.313	114.174***	1.385	103.332	2.330	105.911*	1.964
Two-hour glucose (mg/dL)	96.002	2.647	120.555***	3.496	121.278***	2.195	124.493***	3.542	117.782***	2.951
Waist–hip ratio	0.924	0.004	0.943*	0.005	0.946**	0.003	0.941*	0.005	0.940*	0.004
Waist circumference (cm)	88.625	0.836	88.015	1.059	98.465***	0.653	86.875	1.070	91.911*	0.901
BMI (kg/m2)	25.536	0.388	26.057	0.493	31.358***	0.301	25.587	0.497	27.494**	0.419
Albumin:creatine (mg/g)	0.597	0.206	1.423*	0.262	1.916***	0.159	1.319	0.264	1.372*	0.223
lnHOMA	1.145	0.038	1.351**	0.048	1.563***	0.029	1.297*	0.049	1.326**	0.041

The mean, SE, and whether the value for one ethnic group is significantly different from Caucasians through Dunnett's post hoc test are reported for each biomarker.

Use of Dunnett's post hoc test (control = Caucasians), significantly different biomarker values are bolded, ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

BMI, body mass index; BP, blood pressure; HDL, high-density lipoprotein; SE, standard error.

Age-adjusted prevalence of metabolic syndrome varied by ethnic group, indicating that ethnic disparities exist. Bolded P-values indicate that the prevalence values for that ethnic group are significantly different from Caucasians for the given definition. Caucasians had the lowest prevalence using all criteria, while Hawaiians had the highest using all definitions. After Hawaiians, Filipinos had the second highest prevalence in all definitions, followed by the Japanese who had the third highest prevalence in all definitions. In terms of definition comparison, the WHO consistently yielded the lowest prevalence rate for all ethnic groups as well as for the total population prevalence value. IDF consistently yielded the highest prevalence rate. All criteria tested yielded significantly different prevalence values for Native Hawaiians only, while all definitions performed too similarly to be significant for the Japanese, Filipinos, and Caucasians when examining the 95% CI. The statistical interaction between ethnicity and definition tested by the multivariate model was highly significant (probably > F = 0.0013), confirming the observation that the performance across definition methods differed significantly by ethnicity. The mixed/other group was not utilized for comparison as this group comprised people of other ethnic backgrounds (e.g., Chinese and Korean) as well as those who were of mixed ethnicities in terms of those represented in this study (e.g., Japanese and Filipino) (Table 3).

Table 3.

Age-Adjusted Prevalence of Metabolic Syndrome with 95% Confidence Intervals by Ethnicity and Diagnostic Criteria

	NCEP-ATPIII		IDF		WHO
	Prevalence	95% CI	Prevalence	95% CI	Prevalence	95% CI
Total (n)	31.20% (453)	28.82%–33.58%	39.05% (567)	36.54%–41.56%	22.31% (324)	20.17%–24.45%
Caucasian	16.44%	11.39%–21.48%	20.50%	15.21%–25.78%	11.38%	6.80%–15.96%
Filipino	33.85%**	27.45%–40.26%	40.76%***	34.05%–47.47%	24.21%**	18.39%–30.03%
Part-Hawaiian	42.35%***	38.51%–46.18%	52.16%***	48.14%–56.18%	29.76%***	26.27%–33.24%
Japanese	26.29%	19.83%–32.74%	33.93%**	27.17%–40.69%	23.43%**	17.57%–29.29%
Mixed/other	27.24%*	21.80%–32.68%	36.34%**	30.64%–42.04%	17.56%	12.63%–22.50%

Bolded values indicate that prevalence values are significantly different from Caucasians for the given definition through Dunnett's post hoc test, ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001.

CI, confidence interval; IDF, International Diabetes Federation; NCEP-ATPIII, National Cholesterol Education Program–Adult Treatment Panel III; WHO, World Health Organization.

Discussion

In summary, the overall prevalence of metabolic syndrome differed significantly when different definitions were applied when examining the 95% CI for each definition. When examining overall prevalence, the WHO provided the lowest prevalence estimate [22.31% (20.17%–24.45%)] as it used the strictest criterion among the three definitions tested, namely the requirement of insulin resistance. The criterion for insulin resistance was also regulated as it relied on meeting cutoff values for a fasting blood glucose test as well as a 2-hr postprandial blood glucose test.¹¹ The NCEP-ATPIII definition provided an intermediary prevalence rate as the definition was not specific in terms of population or requirements.⁶

The IDF definition provided the highest prevalence estimate [39.05% (36.54%–41.56%)]. This was possibly due to the use of ethnic-specific values for the abdominal obesity requirement as it accounted for ethnic groups (e.g., Japanese) who present with lower abdominal obesity, but higher prevalence of T2DM and CVD.^8,12 In this study, the Japanese population had the second highest waist-to-hip ratio of all ethnic groups (part-Hawaiians had the highest waist-to-hip ratio), but the lowest BMI and waist circumference. Other definitions that use a standard measure for abdominal obesity do not account for variations in this population characteristic, possibly resulting in underestimation of metabolic syndrome prevalence.

A study investigating the effects of utilizing different diagnostic criteria for metabolic syndrome [IDF, NCEP-ATPIII, and the Chinese Diabetes Society (CDS)] was done in China.¹³ Cheng et al. also identified variances in prevalence when using diagnostic criteria, particularly when comparing the IDF and NCEP-ATPIII definitions. Interestingly, the NCEP-ATPIII definition yielded the highest prevalence rate (24.77%) and the IDF definition provided an intermediary prevalence rate (19.85%).¹³ This did not reflect the trend observed in this study, possibly due to the differences in the populations tested. The Cheng et al. study was conducted in the Jiangxi Province, located in a rural area of China. Due to this, it can be assumed that this population was fairly homogeneous in terms of ethnicity, in contrast to the Kohala population that was highly multiethnic in population demographics. Desroches and Lamarche also found similar discrepancies when examining prevalence derived using different definitions due to the composition of the population being studied as well as the prevalence of certain risk factors compared to others.¹⁴

In terms of ethnicity, a variance in prevalence was observed when comparing the prevalence rates among a single ethnic group, although many prevalence values were not statistically significant from others based on 95% CIs when using different criteria, and was confirmed by the statistical interaction between ethnicity and definition when tested using the multivariate model. However, the WHO definition for metabolic syndrome yielded prevalence rates that were significantly different from both the IDF and NCEP-ATPIII definitions for all ethnic groups, as reflected by no overlap in 95% CIs. This may be due to differences in overall requirements for diagnosis as well as different cutoff values for the majority of risk factors included in the definition.¹¹ In examining these prevalence rates, metabolic syndrome was the most prevalent in Hawaiians/part-Hawaiians for all definitions examined. In this study, the part-Hawaiian population had the most abnormal anthropometric and biometric measures for all biomarkers except triglycerides and 2-hr glucose, providing support as to why the part-Hawaiians had the highest prevalence of metabolic syndrome in all definitions. Caucasians presented with the least abnormal anthropometric and biometric measures, except waist circumference (Japanese had the lowest waist circumference), related to metabolic syndrome diagnosis and consequently presented with the lowest prevalence for all three definitions tested.

Conclusions and Recommendations

These findings suggest ethnic disparities among metabolic syndrome diagnoses, and varying definitions exacerbate these disparities. Accurate diagnosis requires known etiology so that risk factors that are most significant in regard to the development of metabolic syndrome are included in the definitions. Currently, no etiology is known, but there are current hypotheses that attempt to clarify the causal factors.¹⁵ The most widely accepted hypothesis is that of insulin resistance as the underlying pathophysiology of metabolic syndrome.¹⁵ Insulin resistance is characterized by the inability to suppress plasma glucose, resulting in high levels detected in fasting or 2-hr postprandial blood glucose tests.¹⁶ As a main contributor to diabetes, it would make sense that the metabolic syndrome is also caused, in part, by insulin resistance. If that is the case, diagnostic criteria should require it as a risk factor to be diagnosed with metabolic syndrome. Alternative hypotheses focus on an excess of fatty acid, resulting in an accumulation of lipids in the tissues, possibly driving abdominal obesity.¹⁵ Similarly, diagnostic criteria would require dyslipidemia and/or abdominal obesity in reevaluated definitions.

Additional hypotheses that implicate key risk factors as the causal factors of metabolic syndrome involve inflammatory and oxidative mediators.¹⁵ Common mediators such as IL-6 and PAI-1 show an increase in circulation among patients with central obesity and patients who have T2DM. It is hypothesized then that due to central obesity and insulin resistance—the key premise of T2DM diagnosis—being implicated in metabolic syndrome, these mediators may also play a role in causing metabolic syndrome.¹⁵ Confirming this hypothesis may add additional requirements for metabolic syndrome in terms of circulation of specific inflammatory and oxidative mediators. However, more research needs to be done to implicate other mediators and to have a better understanding of the circulation levels needed to cause these metabolic disorders.

Limitations of this study involved small sample sizes, which may have affected the potential for prevalence to be statistically significant. Additionally, our study did not include follow-up data due to its cross-sectional observations. This did not allow for determination of the predictive value of metabolic syndrome diagnosis for the incidence of T2DM and CVD as well as mortality. Future studies should adopt a cohort design so as to fully evaluate the increase in risk for T2DM, CVD, and mortality as a result of metabolic syndrome diagnosis for each ethnic group and should maintain a larger sample size that distributes individuals equally across the ethnic groups tested.

Author Disclosure Statement

No conflicting financial interests exist.

Funding Information

This research was supported by a grant from the National Institutes of Health, National Center for Research Resources, Research Centers in Minority Institutions Program, grant no. G12RR03061.

Supplementary Material

Supplementary Appendix SA1

Supplementary Appendix SA2