Heterogeneity in Familial Clustering of Metabolic Syndrome Components in the Multiethnic GENNID study

Abstract

Objective:

Metabolic Syndrome (MetS) is defined by clustering of cardiometabolic components, which may be present in different combinations. The authors evaluated clustering in individuals and extended families within and across ancestry groups.

Methods:

The prevalence of different combinations of MetS components (high fasting glucose, low high-density lipoprotein, high triglycerides, high blood pressure, and abdominal obesity) was estimated in 1,651 subjects (340 families) self-reporting as European-American (EA), Hispanic/Mexican-American (MA), African-American (AA), and Japanese-American (JA). Odds ratios were estimated using logistic regression with GEE comparing individual MetS components, number and combinations of components for each ancestry group vs EA.

Results:

Clustering of all 5 components (Combination #16) was more prevalent in EA (29.9%) and MA (25.2%) than AA (18.7%) and JA (15.5%). Compared to EA, AA were 64% and 66% less likely to have high TG and low HDL, whereas JA were 85% and 56% less likely to have abdominal obesity and high BP, respectively. Compared to EA, the odds of having 2, 4, or 5 components was at least 77% lower in JA, while having 3, 4, or 5 components was at least 3.79 times greater in MA.

Conclusions:

Understanding heterogeneity in MetS clustering may identify factors important in reducing health disparities.

INTRODUCTION

Metabolic Syndrome (MetS) is characterized by the clustering of cardiovascular disease (CVD) and type 2 diabetes (T2D) risk factors. The National Cholesterol Education Program’s Adult Treatment Panel III (NCEP ATP III) guideline defines MetS as the presence of at least any three of the five conditions/components: elevated blood pressure (BP), elevated triglycerides (TG), decreased high-density lipoprotein (HDL) cholesterol, elevated fasting glucose, and abdominal obesity (1). MetS is an important public health problem and associated with 6–7% of all-cause mortality, 12–17% of CVD risk, and 30–52% of T2D risk in the general U.S. population (2). More recently, MetS was shown to increase susceptibility to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and risk of severe COVID-19 disease and death (3, 4, 5, 6).

Originally defined in 1988 (7, 8) and until 2012, MetS has increased more than 35% in prevalence for every sociodemographic group from the National Health and Nutrition Examination Survey (NHANES) of U.S. adults (9). By 2012, more than one-third of all U.S. adults met the definition for MetS (9). The most recent NHANES data from 1999 to 2020 indicate an increased temporal trend in the prevalence of MetS in U.S. adults (10) which reached a peak of 39.9% in 2015–2016 (11) but is showing signs of plateauing in the overall U.S. population (10). Moreover, the prevalence varies widely by age and ancestry, with some groups showing a disproportionate burden of MetS. Specifically, nearly 60% of Hispanic adults over age 60 have MetS (12). While the prevalence of MetS has increased in non-Hispanic Asian adults in the U.S. from 21.8% in 2011–12 to 31.2% in 2017–18, prevalence in the U.S. among non-Hispanic Asian adults is lower than in Hispanic adults (10, 11, 13).

MetS is a complex condition with both environmental and genetic contributions and is heritable with consistent evidence across multiple ethnic groups in family and twin studies (14, 15, 16, 17). Similarly, genome-wide association studies (GWAS) (18, 19, 20) have also identified genetic variants associated with MetS, defined by NCEP criteria (1). Further, previous studies (21, 22, 23, 24) suggest underlying genetic variation may lead to different combinations of MetS components. These studies together with the large number of genetic variants associated with MetS point to the complex genetic architecture underlying MetS. It is not known, however, whether differences in clustering of MetS components reflect allelic diversity across ethnic populations and/or genetic (locus) heterogeneity at the family level. Understanding patterns of MetS trait clustering within and across ancestry groups will help elucidate potential underlying pathologic processes and will enable work on reducing health disparities. Using the Genetics of Non-Insulin Dependent Diabetes (GENNID) study, a resource from the American Diabetes Association (ADA), we describe differential aggregation patterns of MetS components and evaluated the number and co-occurrence/combinations of MetS components in families by ancestry group: African-American (AA), European-American (EA), Japanese-American (JA), and Mexican-American (MA).

METHODS

Study setting and study population

We used data from the GENNID study provided by the ADA and obtained an institutional review board (IRB) exemption for this work. GENNID is a multicenter study established by the ADA as a resource for identifying genetic influences on T2D. GENNID is comprised of a comprehensive, well-characterized multiethnic sample of families with T2D recruited in the U.S. between 1993–1999. The sample consists of families from four ancestry groups: EA, MA, AA, and JA, and the enrollment criteria have been described previously (25). Eligibility includes a minimum of two adult siblings with documented T2D, access to at least 3 adult first degree relatives of the affected siblings, and at least one unaffected spouse. At the time of recruitment the following National Diabetes Data Group criteria (26) was used to define T2D status: (a) either at least two occurrences of a fasting plasma glucose concentration of 7.8 mmol/1 [^140 mg/dl] or (b) at least two occurrences of a plasma glucose concentration of 11.1 mmol/1 [>200 mg/dl] during an oral glucose tolerance test [OGTT] at 2 hours and at least one other timepoint after glucose challenge (25).

Data Collection and Study Variables

Data collection, as previously described (25), focused on the ascertainment of 369 multiplex T2D families consisting of 1,729 subjects with valid individual and family identifiers, non-missing self-identified ancestry, phase collection, and sex status. Singleton subjects (N=27) without any family membership were removed. Those who did not specify an ancestry group (N=10), preferred not to answer (N=1), or specified an ancestry group differing from the majority of their family members (N=40) were omitted. Thus, after exclusions, there were 340 families with a total of 1,651 participants available for analysis. GENNID data can be evaluated as either family data or individual data, making it an ideal dataset for the question posed in this paper.

Data for all adult participants (age ≥18 years) included: 1) collection of pedigree information, 2) demographic family/medical history questionnaire (age and anti-hypertensive, dyslipidemic, and hypoglycemic medication use), 3) physical examinations (weight, waist circumference, and systolic/diastolic blood pressure (i.e., SBP, DBP) averaged over 3 measurements), and 4) laboratory measurements (fasting glucose, insulin, high density lipoprotein cholesterol (HDL) and triglycerides (TG)). Measured traits of interest include waist circumference SBP, DBP, HDL, TG, and fasting glucose levels.

MetS Classification and Clustering

MetS was defined using the NCEP criteria for consistency and ease of comparison with the literature in all analyses presented here. MetS cases were defined as those individuals with at least three of the following five components: increased abdominal obesity, high BP, elevated fasting blood glucose, abnormal HDL cholesterol, and elevated triglyceride levels (1). According to NCEP, these criteria are specified as the following: 1) abdominal obesity with waist circumference ≥102 cm (male adults) and ≥88 cm (female adults), 2) atherogenic dyslipidemia with high TG ≥150 mg/dL, 3) atherogenic dyslipidemia with low HDL cholesterol <40 mg/dL (male adults) and <50 mg/dL (female adults), 4) high BP with SBP/DBP ≥ 130/85 mmHg, and 5) elevated fasting glucose with fasting plasma glucose ≥100 mg/dL. Taking medication (Table S1) to treat high glucose, high TG, low HDL, or high BP was also used to identify those meeting criteria for individual MetS components. In addition to classifying MetS cases (i.e., individuals with MetS), we also defined a family as multiplex MetS if there were at least 3 MetS cases in the family. The multiplex family was then labeled according to the most prevalent combination of MetS components among its cases. If two or more combinations were equally common in a single family, then the per-family combination included only those individual MetS components shared in at least two of the MetS cases.

Statistical Analysis

Descriptive Statistics

Descriptive statistics included percentages and counts for categorical variables and means with standard deviations (SD) for continuous variables. The prevalence of each MetS component and number of components were summarized for all individuals separately by ancestry group. Number of individuals, MetS cases, and number of generations within families were summarized by the median, 25^th, and 75^th percentiles. Similar statistics were provided for the subset of multiplex MetS families.

We evaluated the co-occurrence of 3, 4, and 5 MetS components (i.e., MetS clustering) both among all MetS cases and in the subset of multiplex MetS families across ancestry groups. Additionally, we compared the prevalence of the different combinations of MetS components for individuals with MetS and among multiplex MetS families.

We compared the medians of continuous variables by ancestry group using the Wilcoxon rank-based tests for clustered data (to account for within family correlation) using the Datta-Satten method (27, 28) in the clusrank R package (29). Also, we compared frequencies across ancestry groups using Pearson’s Chi-squared test with Yates continuity correction. A 0.05 significance level was used. All analyses were performed using R 4.0.3 (30).

Familial Aggregation

We used a logistic regression model to compare the MetS clustering between ancestry groups and adopted the generalized estimating equations (GEE) with independence working correlation structure for inference to account for within-family correlations. The three binary outcomes were the following: (a) having versus not having each of the five MetS components, (b) having a given number of MetS components (i.e., 1, 2, 3, 4, or 5) versus having no MetS components, and (c) having each of the most prevalent combinations (i.e., prevalence >10% among MetS cases within ancestry group) of MetS components versus not having any of the five MetS components. Analyses were adjusted for age, sex, and three binary indicator variables for AA, JA, and MA ancestry groups, respectively, with the largest EA group selected as the reference. The estimated odds ratio (OR) of having a specific MetS cluster comparing AA, JA, or MA to the EA group are shown along with 95% confidence intervals (CI) and p-values (p). Results with p < 0.003 (i.e., 0.05/15) were considered statistically significant based on a Bonferroni correction for 15 comparisons (i.e., 3 between ancestry group comparisons x 5 outcome levels of type or number of MetS components). When looking at prevalent MetS combinations, results with p < 0.002 (i.e., 0.05/24) were considered significant after adjusting for 24 comparisons (i.e., 3 between-ancestry group comparisons x 8 MetS combinations). The GEE analyses were conducted using the geepack (31, 32, 33) R package.

RESULTS

Descriptive Statistics

Table 1 shows there were significant differences in mean age by ancestry group (p=2.0e-14), on average AA and MA participants were younger than EA and JA participants (AA/MA mean ~51 years vs. EA/JA mean ~55 years, respectively). Moreover, MA and AA had a greater percentage of women (both >63%) than EA (56%) and JA (46.7%) (p=2.0e-04). MetS was least prevalent in JAs and AAs (47.5% and 54.0%, respectively) and was different across all ancestry groups (p =8.6e-05). The prevalence of most MetS components (High TG, Low HDL, Abdominal Obesity, and High BP) were significantly different by ancestry group; however, the prevalence of high glucose was not significantly different across ancestry groups. The similarity in the proportion of participants with high glucose is not surprising given that the ascertainment for GENNID was based on T2D. Among participants with characteristics that met the criteria for MetS as defined above (either NCEP guidelines and/or medication use), the most prevalent MetS component was high glucose in AAs and MAs (69.4% and 68%, respectively) and low HDL in EAs and MAs (76.7% and 77.7%, respectively). Low HDL was still present in about half of AAs and JAs. High TG was the least prevalent in AAs (17.7%) and most prevalent in MAs (43.4%). High BP was more common among MAs, EAs, and AAs (49.4% to 63.3%) but was the least common in JAs (44.3%). Abdominal obesity was most frequent among MAs (69.1%) and least common among JAs (23%). The prevalence of MetS and its components in the combined sample was most similar to the EAs, which was the largest group sampled. Among the subset of participants with each MetS condition, there were significant differences in the median age between males and females of EAs with hypertension as well as of JAs with low HDL (Figure 1, Table S2). Specifically, among JAs, low HDL was more prevalent among younger women than men. Whereas, among EAs, hypertension was more prevalent among younger men than women.

Table 1.

Distribution of age, sex and MetS components in individuals by ancestry group

Characteristics	All	African Americans (AA)	European Americans (EA)	Japanese Americans (JA)	Mexican Americans (MA)	p a
No. of subjects	1651	248	747	122	534
Age (yrs): Mean +/− SD	53.3±16	51.6 ±15.1	54.9 ± 16.2	56.3 ±15.9	51.1 ± 15.9	2.0e-14
Sex (% Female)	59.2	66.1	56.0	46.7	63.5	2.0e-04
MetS component (%):
High Glucose	66.1	69.4	64.4	61.5	68.0	0.25
High TGb	37.5	17.7	39.5	39.3	43.4	5.4e-11
Low HDLc	72.3	54.0	76.7	58.2	77.7	9.2e-15
High BPd	54.9	63.3	57.7	44.3	49.4	8.5e-05
Abdominal Obesity	61.8	59.7	63.7	23.0	69.1	<2.2e-16
Number of MetS components (%):
none	5.0	6.9	4.8	15.6	2.1	1.0e-08
1	13.3	15.3	12.3	18.9	12.4	0.16
2	20.2	23.8	19.8	18.0	19.5	0.46
3	23.2	25.0	21.0	26.2	24.7	0.29
4	22.4	19.0	23.2	13.9	24.7	0.03
5e	16.0	10.1	18.9	7.4	16.7	4.0e-04
MetSf cases (%)	61.5	54.0	63.1	47.5	66.1	8.6e-05

^ap-values corresponding to testing for at least one difference between group comparisons

^bTG=triglycerides

^cHDL=high density lipoprotein cholesterol

^dBP=blood pressure

^eNote that having all 5 MetS components in Table 1, Table 3 is the same as the clustering of Combination #16 in Figure 1 and Table 4.

^fMetS : Metabolic Syndrome as defined by having at least 3 of the 5 conditions characterized by NCEP ATP III guidelines and/or taking medication (Supplementary Table 1); percentages with each risk factor are calculated from the total number of subjects from each ancestry group.

Figure 1. Distribution of MetS conditions across age by ancestry and sex.

Each column denotes an ancestral group: (AA) African-American plots A, E; (EA) European-American plots B, F; (JA) Japanese-American plots C, G; (MA) Mexican-American plots D, H. The rows are Male (A-D) and Female (E-H). Distributions are density plots with a rug of tick marks on the x-axis. Each column of tick marks denotes an individual with the given combination of MetS components. Each MetS component/condition is denoted by a color: (black) Abdominal obesity; (red) High glucose; (green) Hypertension; (blue) Low HDL; (cyan) high TG.

Table 2 describes GENNID families by ancestry. AAs tended to have smaller family sizes with a median of 3 members/family whereas EAs and MAs had a median of 4 members/family. The JA families were the largest (median: 7 members/family). Based on the median number of MetS cases per family, JA and MA families typically had 3 MetS cases/family while AA and EA families had 2. About half of EA and MA families were defined as multiplex MetS families (with at least 3 MetS cases) compared to about 37% of AA and 63% of JA families. Based on the median number of family members per multiplex MetS family, EAs and MAs typically had 5 members per MetS family; AAs had smaller MetS families with 4 members per MetS family, and JAs had larger 8 members per MetS family. Among AAs, there was a median of 3 MetS cases per MetS family while there were 4 MetS cases per MetS family in the other ancestry groups. Additionally, approximately half of the AA, EA, and MA families had at least 4 members per family, whereas 94% of JA families had at least 4 members per family. Most of these families with at least 4 members per family (97% in EA, 88% in MA, 76% in AA, and 67% in JA) were classified as multiplex, MetS families and had at least 3 MetS cases.

Table 2.

Characteristics of GENNID Families and MetS components by ancestry group

Characteristics	All	African Americans (AA)	European Americans (EA)	Japanese Americans (JA)	Mexican Americans (MA)
No. of families	340	68	158	16	98
No. of subjects per family: 50th (25th, 75th) percentiles	4 (3,6)	3 (3,4)	4 (3,6)	7 (6,9)	4 (2,5)
No. of MetS casesa per family: 50th (25th, 75th) percentiles	2 (2,4)	2 (1,3)	2 (2,4)	3 (2,4)	3 (2,4)
No. of multiplex MetS familiesb (%)	171 (50%)	25 (37%)	87 (55%)	10 (63%)	49 (50%)
No. of subjects per MetS family: 50th (25th, 75th) percentiles	5 (4,7)	4 (4,5)	5 (4,8)	8 (6, 10)	5 (4,7)
No. of MetS cases per MetS family: 50th (25th, 75th) percentiles	4 (3,5)	3 (3,3)	4 (3,5)	4 (4,6)	4 (3,5)
No. of families with at least 4 members	194 (57%)	33 (49%)	90 (57%)	15 (94%)	56 (57%)
No. of multiplex MetS families with at least 4 members (% MetS families among families with at least 4 members)	171 (88%)	25 (76%)	87 (97%)	10 (67%)	49 (88%)

^aMetS case: an individual with Metabolic Syndrome by having at least 3 of the 5 conditions/components characterized by NCEP ATP III guidelines and/or taking medication (Supplementary Table S1)

^bMultiplex MetS families: families with at least 3 MetS cases

Clustering among individuals with MetS

Figure 2 shows the frequency of 16 different MetS trait combinations among individuals with MetS by ancestry. Among JA MetS cases, the two most prevalent combination of traits each had a frequency of 17.2%: Combination #4 (High Glucose-High TG-low HDL) and Combination #14 (High Glucose-High TG-Low HDL-High BP). However, the most prevalent combination in EAs and MAs was having all 5 MetS components (i.e., Combination #16) at 29.9% and 25.2%, respectively. Among AAs with MetS, the most prevalent combination (24.6%) was Combination #13 (High Glucose-Abdominal Obesity-Low HDL-High BP). In general, the most prevalent combinations (i.e., combinations with prevalence greater than the 75^th percentile or 10%) showed high glucose and abdominal obesity clustering with at least one other component (#2, #3 & #13 (AA); #13 (EA); #2, #11, #13 (MA), #16 (all groups)). Among AAs, high BP and low HDL each clustered with high glucose and abdominal obesity in Combinations #2 & #3 (similarly in the MAs), respectively, and both high BP and low HDL clustered together with high glucose and abdominal obesity in Combination #13 (which was also prevalent in AAs, EAs and MAs). The clustering of both high TG and low HDL with high glucose and abdominal obesity was most frequently observed in MAs (#11). However, aside from combination #16, the most prevalent combinations in the JA sample with MetS (#4, #6, #14) did not include abdominal obesity.

Figure 2: Prevalence (%) of MetS clustering among case subjects by ancestry group.

The barplot shows the prevalence (%) on the y-axis for each MetS combination of conditions/components that are described by the bottom table with “X” indicating the presence of the specific MetS component. Each combination of conditions is denoted by numbers 1 through 16. The prevalence of each combination by group are denoted by separate bar plot colors with blue for African Americans (AA), red for European Americans (EA), green for Japanese Americans (JA), and purple for Mexican Americans (MA). Please note that Combination #16 refers to the same clustering as having all 5 MetS conditions in Table 1 and Table 3.

Clustering among MetS families

Figure S1 shows the proportion of multiplex MetS families (at least three MetS cases) and the familial clustering patterns by ancestry. Similar to the results for individual MetS cases, the presence of all five MetS components (combination #16) was highly prevalent (at least 30%) among all ancestry groups and was most prevalent in 45% of EA, 43% of MA and 40% of JA MetS families. However, in multiplex AA MetS families, high glucose, abdominal obesity, low HDL, and high BP (Combination #13 at 52%) was the most common cluster of components. Additionally, for multiplex JA MetS families, high glucose, high TG, and low HDL cholesterol clustered most frequently with 30% of multiplex MetS families having at least three cases with this pattern of MetS traits (Combination #4).

Differences in Familial Aggregation of MetS across Ancestry Groups

We calculated the odds ratio (OR) for MetS (both of type and number of components as outcomes) separately for each of the three ancestry groups compared to the EAs (i.e., reference), adjusted for age and sex. Statistically significant results after accounting for multiple comparisons (p<0.003) are shown in Table 3. For individual MetS components the odds of having abdominal obesity or high BP were 85% and 56% lower, respectively, in JAs compared to EAs (p=1.27e-10 and p=1.2e-04). Similarly, the odds of having high TG levels or low HDL was 64% and 66% lower, respectively, in AA compared to EA (p=9.80e-07 and p=6.19e-11, respectively), but the odds of the number of MetS components in AAs was not significantly different than in EAs. Additionally, the odds of having 2, 4, or 5 MetS components was at least 77% lower in JAs compared to EAs and remained statistically significant after adjustment for age, sex and multiple comparisons (p=7.08e-04, p=3.15e-04, p=4.91e-05, respectively). Although the odds of having any one or two of the individual MetS components was comparable between MAs and EAs (p>0.003), the odds of having a combination of three, four, or all 5 MetS components was at least 3.81, 5.45, or 3.79 times higher in MAs compared to EAs and remained statistically significant after adjustment for age, sex and multiple comparisons (p=5.00e-04, p=4.08e-05, p=2.89e-03).

Table 3.

Logistic regression results presenting Odds ratios (OR) for type or number of MetS components by group adjusting for age, sex and multiple statistical comparisons

	European American (EA)	African Americans (AA)		Japanese Americans (JA)		Mexican Americans (MA)
MetS component (yes vs no)	OR	OR (95% CI)	p b	OR (95% CI)	p b	OR (95% CI)	p b
MetSa	1 (baseline)	0.76(0.53–1.08)	0.12	0.47(0.27–0.82)	0.01	1.36(1.03–1.8)	0.03
High Glucose	1 (baseline)	1.52(1.10–2.11)	0.01	0.79(0.49–1.27)	0.33	1.44(1.11–1.86)	0.01
Abdominal Obesity	1 (baseline)	0.75(0.52–1.09)	0.13	0.15(0.09–0.27)	1.27e-10	1.25(0.93–1.69)	0.14
High BP	1 (baseline)	1.72(1.14–2.59)	0.01	0.44(0.29–0.67)	1.20e–04	0.88(0.66–1.17)	0.38
High TG	1 (baseline)	0.36(0.24–0.54)	9.80e-07	0.94(0.50–1.76)	0.84	1.33(1.01–1.75)	0.04
Low HDL	1 (baseline)	0.34(0.25–0.47)	6.19e-11	0.43(0.23–0.83)	0.01	1.03(0.76–1.39)	0.85
Number of MetS components vs (none)	OR	OR (95% CI)	p b	OR (95% CI)	p b	OR (95% CI)	p b
1 MetS component	1 (baseline)	0.79(0.37–1.71)	0.55	0.46(0.24–0.88)	0.02	2.29(0.99–5.3)	0.05
2 MetS components	1 (baseline)	0.82(0.42–1.59)	0.56	0.23(0.10–0.53)	7.08e-04	2.74(1.32–5.71)	0.01
3 MetS components	1 (baseline)	0.95(0.45–2.01)	0.89	0.36(0.15–0.87)	0.02	3.81(1.79–8.10)	5.00e–04
4 MetS components	1 (baseline)	0.94(0.41–2.16)	0.88	0.19(0.07–0.46)	3.15e-04	5.45(2.43–12.30)	4.08e–05
5 MetS componentsc	1 (baseline)	0.52(0.22–1.27)	0.15	0.13(0.05–0.35)	4.91e-05	3.79(1.58–9.12)	2.89e–03 d

^aMetS case: an individual with Metabolic Syndrome by having at least 3 of the 5 components characterized by NCEP ATP III guidelines and/or taking medication (Supplementary Table 1)

^bp-value from logistic regression with GEE to account for correlation within families; multiple testing was accounted for 3 between-ethnic-group comparisons x 5 levels per outcome, and bolded results indicate p<0.003 or 0.05/15 is significant.

^cNote that having all 5 MetS components in Table 1, Table 3 is the same as the clustering of Combination #16 in Figure 1 and Table 4.

^dAfter adjusting for all clustering combinations in Table 4, having all 5 MetS conditions (Combination #16) was borderline significant (MAs vs. EAs).

The ORs comparing the most prevalent (i.e., with more than 10% frequency) combinations of MetS components in AA, JA and MA compared to EA are shown in Table 4 and are adjusted for age and sex. While odds of some combinations (#3,6,13) were higher and others were lower (#2,11,14, 16) in AAs compared to EAs none of these results were statistically significant after adjustment for multiple comparisons. When comparing JAs to EAs the odds for all but one combination (#4) was lower but not statistically significant in JAs with MetS compared to EAs with MetS. The only combination that remained statistically significant after adjustment for age, sex and multiple comparisons was Combination #16 (all 5 MetS components), where the odds of having all 5 MetS components was 87% lower in JA individuals with MetS compared to EA individuals with MetS. However, the odds of Combination #11 (High Glucose, Abdominal Obesity, high TG and low HDL) and Combination #13 (High Glucose, Abdominal Obesity, low HDL, high BP) were 5.98 and 6.52 times greater in the MAs compared to EAs and remained statistically significant after adjustment for age, sex and multiple comparisons (p≤2.01e-04).

Table 4.

Logistic regression results presenting Odds ratios (OR) for prevalent^a combinations of MetS component(s) by group adjusting for age, sex and multiple statistical comparisons

	MetSb Combination VS. No. MetS components					European American (EA)	African Americans (AA)		Japanese Americans (JA)		Mexican Americans (MA)
Combo No.	High glucose	Abdominal Obesity	High TG	Low HDL	High BP	OR	OR (95% CI)	p c	OR (95% CI)	p c	OR (95% CI)	p c
2	X	X		X		1 (baseline)	0.97(0.40–2.37)	0.95	0.27(0.09–0.84)	0.02	3.97(1.56–10.10)	3.76e-03
3	X	X			X	1 (baseline)	2.08(0.81–5.34)	0.13	0.17(0.04–0.71)	0.02	5.06(1.77–14.50)	2.50e-03
4	X		X	X		1 (baseline)	-	-	1.32(0.29–6.01)	0.72	4.8(1.34–17.10)	0.02
6	X			X		1 (baseline)	1.30(0.32–5.29)	0.71	0.55(0.20–1.52)	0.25	3.14(0.97–10.20)	0.05
11	X	X	X	X		1 (baseline)	0.43(0.15–1.26)	0.12	0.07(0.01–0.60)	0.02	6.52(2.98–14.30)	2.74e-06
13	X	X		X	X	1 (baseline)	2.40(0.90–6.42)	0.08	0.14(0.04–0.58)	0.01	5.98(2.33–15.40)	2.01e-04
14	X		X	X	X	1 (baseline)	0.33(0.06–1.91)	0.21	0.41(0.14–1.22)	0.11	2.86(1.02–8.05)	0.05
16 e	X	X	X	X	X	1 (baseline)	0.52(0.22–1.27)	0.15	0.13(0.05–0.35)	4.91e-05	3.79(1.58–9.12)	2.89e-03f

^aCombinations selected have at least a 10% frequency among cases within at least one ancestry group; analyses are subset on subjects with the given combination and subjects with no MetS features.

^bMetS case: an individual with Metabolic Syndrome by having at least 3 of the 5 components characterized by NCEP ATP III guidelines and/or taking medication (Supplementary Table 1)

^cp-value from logistic regression with GEE to account for correlation within families; multiple testing was accounted for 3 between-ancestry-group comparisons x 8 combinations, and bolded results indicate p<0.002 or 0.05/24 is significant.

^dNo AA MetS case individual had Combination #4.

^eNote that the Combination #16 in Figure 1 and Table 4 is the same clustering as having all 5 MetS components in Table 1 and Table 3.

^fBorderline significance

DISCUSSION

This study evaluated the clustering of MetS components in a multiethnic family sample. We examined clustering patterns stratified by ancestry group and found differences by ancestry in combinations of MetS components both in individuals and MetS families. The most prevalent combination of MetS components among both EA and MA cases was having all 5 MetS components of high glucose, high TG, low HDL cholesterol, high blood pressure, and abdominal obesity. The clustering of all 5 MetS components was also found to be most prevalent among older (>65 years old) men and women in the Third National Health and Nutrition Examination Survey (NHANES III), although they did not stratify by ancestry; however, there were differences in clustering based on age and sex (34) which was also found in GENNID (Figure 1, Table S2). The most common combination of MetS components did not include high TG among the GENNID AA cases. This observation is consistent with the “TG Paradox” in which those of African descent have normal TG levels but are at higher risk for T2D and cardiovascular disease, which are both known to be associated with hypertriglyceridemia (35).

The most prevalent combination of components in both JA MetS cases and multiplex MetS families was high glucose, high TG, low HDL cholesterol and high blood pressure and did not include abdominal obesity. However, it has been suggested that the NCEP guidelines for abdominal obesity may not be suitable for Asian populations, including Japanese (36). Similarly, Tan et al. suggested that the NCEP definition underestimates MetS prevalence in Asian populations because it embodies an unsuitable threshold of central obesity for Asians (37). The International Diabetes Federation (IDF 2005) recommends a different waist circumference threshold for abdominal obesity in Asian samples (i.e., ≥90 cm (males) and ≥80 cm (females)) and among Japanese, further recommends an even lower threshold for males and higher one for females (i.e., ≥85 cm (males) and ≥90 cm (females)) (1). A post-hoc analysis was performed to evaluate the IDF 2005 thresholds for abdominal obesity in the JA GENNID sample. When the IDF 2005 for abdominal obesity was applied to the GENNID Japanese American sample, the prevalence of abdominal obesity increased and resulted in a 10% increase in the cluster containing all 5 MetS components (Figure S2). Furthermore, there was an increase in prevalence for Combinations that included abdominal obesity (#11 and #13) and a decrease for clusters without abdominal obesity (Combinations #4, #6, #14), resulting in less extreme clustering differences from the other ancestry groups. Because there is no consensus for the choice of MetS definition, additional work is needed to refine ancestry-specific thresholds for MetS conditions as choice of thresholds does impact the prevalence of MetS features and clusters. For the U.S.-based sample in this analysis, we used the NCEP criteria (1) as our primary approach to define MetS components and their clustering in order to facilitate comparisons to other published studies.

Although the co-occurrence of specific MetS components (notably high glucose, low HDL, and high BP) aggregated across all four ancestry groups, we found considerable variability in the marginal prevalence of each individual component. Specifically, after sex and age adjustments, compared to EAs, the odds of high TG and low HDL were significantly lower in AAs, while abdominal obesity and high BP were lower in JAs. Moreover, the odds of having 2, 4, or 5 MetS components were at least 80% lower in JAs compared to EAs (p<0.003). Although EAs and MAs did not have significantly different marginal odds for specific MetS components, the odds of having 3 or 4 components were more than 4 times higher in MAs than in EAs (p<0.003). In particular, the two combinations containing four MetS components (i.e., the 4-component combination without high blood pressure and the 4-component combination without high TG) were at least 5 times higher in MAs than in EAs. Given the degree of heterogeneity in clustering of MetS components, it may be important to identify subgroups of MetS cases to better reflect underlying differences, particularly if these differences have implications for treatment or prevention. Further, from a research standpoint, reducing this level of heterogeneity in the case definition of MetS may increase power and improve consistency of findings for genetic studies (38) and those focused on reducing health disparities.

Moreover, familial aggregation patterns of MetS and its individual components are consistent with potential underlying genetic influences. In a previous analysis (23) evaluating MetS in the GENNID study, there was evidence of similar overall phenotypic and environmental correlations but differences in genetic correlations between MetS quantitative traits across ancestry groups. Significant genetic correlations underlying clusters of traits were unique to ancestry groups. Specifically, there were low to moderate statistically significant genetic correlations between DBP and HDL cholesterol (ρ_G = −0.36) and TG and fasting glucose (ρ_G = 0.62) in EAs. In MAs, there was a moderate but significant genetic correlation between TG and SBP (ρ_G = 0.46), supporting the level of heterogeneity observed in this study. Together, the phenotypic heterogeneity in clustering may indicate locus and/or allelic heterogeneity in susceptibility to MetS (39, 40). In genetic studies, stratifying on trait clusters or accounting for the correlation structure between traits may be beneficial in understanding the genetic basis of the condition (40). Thus, given evidence for heterogeneity in MetS trait clustering, it may be prudent to evaluate genetic influences on more narrowly defined MetS phenotypes (41), rather than the broader and more heterogeneous approach of any three MetS components. Defining subtypes may lead to a better understanding of underlying mechanisms associated with MetS.

To our knowledge, this is the first study examining differences in clustering of MetS components and aggregation patterns in a multiethnic family study. It is important to continue to include diverse samples in genetic studies to improve our understanding of potential differences in MetS clustering, even if overall MetS rates are comparable. We also identified evidence for MetS heterogeneity at both an individual and familial level, which has not been previously reported. These observed differences and heterogeneity across groups may be due to shared and unique genetic and/or environment influences – understanding the contribution of both are needed to reduce disparities.

1. What is already known about this subject?

• Previous studies suggest that MetS is a heterogeneous syndrome with different risk factors and pathogenetic mechanisms that lead to different combinations of MetS components.

• It is not known whether these different combinations of MetS components reflect allelic diversity across ancestry populations, and/or genetic (locus) heterogeneity at the level of the family.

2. What are the new findings in your manuscript?

• Differences in clustering patterns of MetS components are observed across ancestry groups.

• Compared to European Americans (EAs), African Americans were 64% and 66% less likely to have hypertriglyceridemia and low HDL cholesterol, whereas Japanese Americans were 85% and 56% less likely to have abdominal obesity and hypertension.

• Compared to EAs, the odds of having 4 orall 5 MetS components was 81% or 87% lower in Japanese Americans but was 5.45 or 3.79 times higher in Mexican Americans.

3. How might your results change the direction of research or the focus of clinical practice?

• Identifying reasons for the underlying heterogeneity in the clustering of MetS components increases understanding of MetS and may improve case definition for genetic studies. Further, these findings may lead to public health and clinical applications that ultimately lead to reduced disparities.