Inter-ethnic Differences in serum lipids: Implications for cardiometabolic disease risk in African ancestry populations

Abstract

African Americans are generally found to have a healthier lipid profile (lower triglycerides [TG] and higher high-density lipoprotein cholesterol concentration [HDLC]) compared to those of other ethnicities. Paradoxically, African Americans do not experience a decreased risk of the cardiometabolic diseases that serum lipids are expected to predict. This review explores this mismatch between biomarker and disease among African ancestry individuals by investigating the presence of interethnic differences in the biological relationships underlying the serum lipids – disease association. This review also discusses the physiologic and genomic factors underlying these interethnic differences. Additionally, because of the importance of serum lipids in assessing disease risk, interethnic differences in serum lipids have implications for identifying African ancestry individuals at risk of cardiometabolic disease. Where possible, data from Africa is included, to further elucidate these ancestral differences in the context of a different environmental background.

African Americans (AA) are generally found to have a healthier lipid profile (lower triglycerides [TG] and higher high-density lipoprotein cholesterol concentration [HDLC]) with a less atherogenic distribution of lipoprotein particles [1–3] compared to those of other ethnicities. Based on the more favorable lipid profile among African ancestry individuals and the established role of serum lipids as a biomarker predictive of cardiometabolic disease, one could reasonably anticipate a reduced burden of these diseases among AA. Epidemiological data, however, does not support this expectation (Table 1)[4]. In fact, AA have a markedly higher prevalence of cardiovascular disease than European Ancestry individuals (EA), driven primarily by higher prevalence of hypertension. The prevalence of type 2 diabetes (T2D) in AA is more than twice the prevalence in EA individuals. This mismatch between biomarker and disease among AA suggests that ancestry may add further complexity to the underlying relationships that serum lipids are expected to capture. In this article, we will try to answer the question: why doesn’t a favorable lipid profile translate into reduced cardiometabolic risk for African ancestry individuals? To address this key question, we will investigate the assumptions necessary to go from biomarker to disease and assess how interethnic differences could affect these assumptions. Moreover, we will discuss the physiologic and genomic factors underlying these interethnic differences. Finally, we will investigate the impact that this mismatch may have on identifying African ancestry individuals at risk of cardiometabolic disease. Much of the evidence relevant to this “metabolic paradox” involves comparisons of AA and EA, yet adding observations in West Africans may also be of considerable importance. While admixed AA have a large proportion of shared ancestry with West Africans (~80%), the environmental context is dramatically different across populations, which may prove to be very informative for understanding the relative contributions of inherited and environmental influences on these complex traits. West African individuals are generally leaner and more physically active than their AA counterparts, yet with increasing urbanization and Westernization, they are also experiencing dramatic increases in cardiometabolic disease. In fact, it is anticipated that Africa will experience some of the most dramatic increases in T2D worldwide [5]. Complex disease research in Africa is experiencing notable advances, spurred, in part, by efforts such as the Human Health and Heredity in Africa initiative (www.h3africa.org [6]), more data from Africa is available, with much more anticipated in the short term as major research projects release results. Thus, in this review, we will address the above questions, drawing in data from Africa where available.

Table 1.

US Prevalence of Cardiometabolic Diseases and Predictors of Risk

	African American		European American
	Men	Women	Men	Women
Cardiovascular Disease	44.8	47.3	37.4	33.8
Coronary Heart Disease	7.9	7.6	8.5	5.8
Myocardial Infarction	4.3	2.2	4.3	2.1
Stroke	4.5	4.4	2.4	3.3
Hypertension	43.0	45.7	33.9	31.3
Diagnosed Diabetes Mellitus	14.3	14.7	6.8	6.5
Undiagnosed Diabetes Mellitus	4.8	4.0	3.9	1.9
Heart Failure	4.5	3.8	2.7	1.8
Metabolic Syndrome	25.3	38.8	37.2	31.5
Low HDL	16.6	6.6	29.5	10.1
Overweight and obesity	70.8	77.7	72.3	59.3

Non-Hispanic White and Black individuals ≥ 20 y. Data from AHA Report on Heart Disease and Stroke Statistics[4]. All data given as %.

Interethnic Differences in Serum Lipids

Ethnic differences in serum lipids have been widely reported. The most consistently observed difference in the concentration of serum lipids is lower TG among African ancestry individuals. This difference can be seen among ethnicities in the United States (US) in the National Health and Nutrition Examination Survey (NHANES) dataset, which was designed to be representative of the US population: mean TG was 113, 143, and 158 mg/dl in African Americans, European Americans, and Mexican Americans [7]. Differences in TG are also observed in children; for example, TG was 15.7 mg/dl lower among AA compared to non-AA children (predominantly EA)[8]. The prevalence of high TG (>150 mg/dl) among NHANES AA children and adolescents was less than half that of EA [9]. Data from West Africa agree with these findings, with mean TG values that are generally even lower than what is observed among admixed African Americans. For illustration, mean TG values from population-based studies of adults without lipids-altering diseases are presented by ethnicity (Figure 1).

Figure 1. Distribution of Triglycerides by Ancestry.

Reported TG values in studies of West Africans[18, 102–110], African Americans[7, 62, 68, 69, 80, 104, 107, 111–113], and those of European Ancestry[7, 62, 68, 69, 84, 111, 113]. Limited to those without disease (type 2 diabetes, hypertension, human immunodeficiency virus, or liver disease). Weighted means used to combine by gender and alcohol drinking status (if reported). Due to amount of available data, European ancestry studies limited to the United States and the United Kingdom.

Among US populations, AA are also generally found to have HDLC concentrations about 4 to 5 mg/dl higher than other ethnic groups [7], differences that are also apparent in studies of children and adolescents [1, 8–11]. However, this difference in distribution is not what is observed among Africans, where the HDLC concentration is more similar to what is observed in European ancestry individuals [12, 13], or lower[14, 15]. It is unclear what leads to this discrepancy between African ancestry individuals in different locations, but it seems reasonable that HDLC is under the influence of an environmental factor that differs between the US and Africa. Some environmental factors worth further investigation in understanding these low HDLC concentrations among Africans are malnutrition and urbanization. Researchers have found the low HDLC co-occurs with micronutrient deficiencies, as well as low body weight [16]. Studies of the Capetown, South Africa in 1990 and in 2008/9 showed marked increases in the prevalence of low HDLC, from 37.3 to 55.1% among men and 30.0 to 67.7% among women. Authors highlight increasing adiposity and urbanization as likely contributors[17]. In contrast, higher HDLC values were observed among members of the Gbagi tribe in Abuja, Nigeria who were living in an urban compared to a rural environment, despite higher body mass index (BMI) and waist circumference among the urban dwellers[18]. Further work is necessary to disentangle the components of “urbanization” that are most relevant for cardiometabolic disease risk. An intriguing study of 1,266 diabetic individuals in the Congo suggests a different association between HDLC and cardiometabolic disease in this environment: among individuals with low HDLC, the rate of fatal cardiovascular events was 13.5%, but among those with high HDLC, the rate was 20%. Additionally, the mean change in blood glucose after a meal was similar in the high and low categories (131.2 and 137.0 mg/dl, respectively), and much lower in those with intermediate HDLC levels (87 mg/dl). Importantly, HDLC also had a U-shaped relationship with risk of atherosclerotic complications [19]. Further investigations into this relationship in other African populations are needed.

Physiology Underlying Interethnic Differences in Serum Lipids

Serum lipids are strongly correlated with fat distribution, particularly an individual’s visceral adipose tissue. Visceral adipose tissue is known to be a uniquely pathogenic fat depot in terms of cardiometabolic risk. Visceral adipose tissue is more strongly associated with development of key perturbations for cardiometabolic risk, including impaired glucose tolerance, blood pressure, and dyslipidemia, than other anthropometric measures (including BMI, waist circumference, waist/hip ratio, leg fat %, trunk fat %) [20] or subcutaneous adipose tissue [21, 22]. Visceral adipose tissue is known to correlate strongly with serum lipids, with increased visceral adipose tissue contributing to increased liver fat content, which, in turn, leads to greater production of very low density lipoprotein, and higher circulating TG[23]. In fact, in a study of 723 AA and EA men and women, visceral adipose tissue explained over 24% of the variation in fasting TG and 31% of the ratio of total to high-density lipoprotein cholesterol [24]. Liver fat has been argued to be a more relevant (yet strongly correlated) depot, and was shown to be a better predictor of concentration of very low density lipoprotein than visceral adipose tissue [1].

African ancestry individuals are less prone to deposit fat in the depots associated with increased cardiometabolic risk. Visceral adipose tissue is markedly lower among AA compared to EA individuals [25, 26]. For example, in a comparison of EA from the Framingham Heart Study and AA from Jackson Heart Study, the mean visceral adipose tissue among AA women was half that of EA women (774.5 vs. 1536.7 cm³), with an even greater difference among men (840.6 vs. 2455.8 cm³), despite a higher BMI in AA [27]. At similar height and proportions of body fat, trunk fat, and leg fat, obese AA women had lower visceral adipose tissue and TG than their EA counterparts [20]. Similar results were observed for African Caribbean individuals compared to those of European and South Asian ancestry [28]. AA also have less fat in the peritoneal cavity and less liver fat compared to individuals of EA and Hispanic ancestry after adjusting for age and total body fat [29]. Likewise, a lower prevalence of fatty liver was observed among AA compared to other ethnicities in the Multi-Ethnic Study of Atherosclerosis study, even after adjusting for relevant risk factors [30]. In contrast, AA generally deposit fat more readily in less metabolically active regions, such as the subcutaneous tissue and in the lower extremities [29].

Underlying interethnic differences in visceral adipose tissue are further complicated by ethnic variability in the ability to approximate this parameter with a common measurement, waist circumference. Despite the fact that waist circumference captures both subcutaneous and visceral adipose tissue, it has been shown to be a good predictor of visceral adipose tissue among both EA and AA [31]. As might be expected given interethnic (and gender) differences in subcutaneous and visceral adipose tissue deposition, significant sex × ethnicity interactions were observed in this relationship [31]. Increasing waist circumference was associated with larger increases in visceral adipose tissue in women of European ancestry compared to those of African ancestry. The association between waist circumference and visceral adipose tissue was the same, however, among AA, West Africans, Black South Africans, and African immigrants in the United States [32].

Given the central role of visceral adipose tissue in the distribution of serum lipids, and the interethnic differences in how overall adiposity reflects visceral adipose tissue, the relationship between body fatness and serum lipids is weaker in African Americans compared to other ethnicities [8, 33, 34]. This difference has also been observed in African vs. European ancestry adolescents in South Africa, an environment in which the European ancestry individuals were heavier than the African ancestry individuals, the reverse of what is observed in the United States [35]. There is some evidence of sexual dimorphism in the association between visceral adipose tissue and serum lipids: associations were weaker among African American women from the Jackson Heart Study compared to EA women from the Framingham Heart Study, but African American men had a stronger association between visceral adipose tissue and serum lipids compared to EA men [27]. Similarly, among Tanzanian men, BMI was significantly correlated with both TG and HDLC, while it was not associated with serum lipids among women [36]. These findings are likely to reflect both ethnic and gender differences in fat distribution.

The fact that African ancestry individuals have lower fat deposition in the regions associated with higher cardiometabolic risk is undoubtedly an important factor underlying the observed interethnic differences in serum lipids. However, it is important to note that AA still had significantly lower TG than either Hispanic or European ancestry individuals even after adjustment for liver fat content [29]. The observed interethnic differences in serum lipids may also be influenced by interethnic differences in the activity of lipoprotein lipase (LPL), acute insulin response, and apolipoprotein CIII concentration.

LPL is the enzyme primarily responsible for the hydrolysis of TG. AA have significantly greater LPL activity than EA [24]. LPL activity explained 23.2% of the variability in HDLC, while the addition of sex, fat mass, visceral adipose tissue, and age only accounted for an additional 10.3%. After adjusting for this set of variables, no significant effect of ethnicity remained. In contrast, while LPL activity was also a significant predictor of TG, it explained less of the variability in TG (4.8%), with visceral adipose tissue being the strongest predictor (24.2%) and ethnicity remaining significant [24]. Thus, these predictors do not completely explain the differences in TG due to ethnicity.

Ethnic differences in insulin response may underlie some of the observed differences in LPL activity. Insulin stimulates LPL, and a significantly higher acute insulin response to glucose challenge has been observed in AA compared to EA individuals, even after adjustment for obesity, body fat distribution, and behavioral factors [37]. A nearly 2.5 times higher acute insulin response was seen among AA compared to age- and BMI-matched EA women [38], with similar results observed in children [39, 40]. This increased insulin response is associated with greater clearance of free fatty acids, which may contribute to lower constitutive TG levels. There is also some evidence of higher LPL mRNA expression in subcutaneous compared to visceral adipose tissue [41, 42], and differences in distribution of body fat has also been proposed as contributing to higher LPL activity among African ancestry individuals[15].

ApoCIII is an apolipoprotein on the surface of very low-density lipoprotein particles that inhibits LPL and hepatic lipase. AA have lower ApoCIII compared to EA, and interethnic differences in TG were no longer statistically significant after adjusting for ApoCIII levels [43]. Similarly, in an analysis of AA and EA women of similar age and BMI, in multivariate models, visceral adipose tissue and ApoCIII were statistically significant predictors of TG concentration, and, after adjusting for these parameters, ethnicity was no longer associated with TG [44].

Based on these findings, it seems reasonable to assume that interethnic differences in visceral adipose tissue, LPL activity, acute insulin response, and ApoCIII concentration relatedly or independently explain the interethnic differences observed in serum lipids. However, considering the confluence of interethnic differences in all of these parameters of significance to cardiometabolic health, how well do serum lipids serve as biomarkers for insulin resistance and cardiovascular risk?

Serum Lipids and Insulin Resistance among African Ancestry Individuals

Insulin resistance is associated with dyslipidemia, one of the significant reasons that serum lipids are included as markers of cardiometabolic risk, yet there is evidence of interethnic differences in the association between serum lipids and insulin resistance, as has been previously reviewed [45–48]. As mentioned, insulin stimulates LPL activity, and insulin resistance is thought to depress LPL activity [49, 50], one of the mechanisms through which insulin resistance is thought to lead to increased TG. A key study of AA confirmed the strong inverse association between LPL activity and TG concentration (adjusted R² 54%, p <0.001), yet showed no difference in LPL activity across tertiles of insulin resistance. Thus, insulin resistance did not diminish LPL activity among AA, one explanation for apparent protection from the elevated TG that is associated with insulin resistance [51]. In agreement with these data, after adjusting for visceral adipose tissue, insulin resistance was not associated with TG concentration among AA women [44]. Likewise, in a separate study, neither HDLC nor TG correlated with insulin sensitivity among AA women, though both were strongly associated among EA women [52]. Similar findings were observed in South Africa among individuals of African and European descent [14]. No significant association between lipid profiles and impaired glucose regulation among East-African young people [53]. The ratio of TG to HDLC is being considered as less invasive predictor of insulin resistance, and has been shown to be useful in other ethnic groups[54, 55]. However, the predictive value of this ratio is less clear among those of African ancestry. The ratio was found not to be associated with insulin resistance among AA women [56, 57]. The ratio was predictive of fasting blood glucose among Ghanaian women, though the optimal threshold was considerably lower than had been found among other ethnic groups [58]. The interethnic differences in fat deposition, serum lipids, and insulin resistance are well-described in a study of 2,170 AA, EA, and Hispanic ancestry individuals [29]. AA had the lowest level of intraperitoneal fat, liver fat, and TG, but a similar prevalence of insulin resistance as Hispanic ancestry individuals, who had the highest level of intraperitoneal fat, liver fat, and TG. Even after adjustment for intraperitoneal fat, AA still had the highest insulin resistance and the lowest TG of the three groups. The authors concluded that the metabolic response to obesity and insulin resistance differs in AA compared to the other ethnicities. In contrast to other ethnicities, insulin resistance among AA occurred with relatively lower TG, higher HDLC, lower visceral adiposity, and lower levels of liver TG [29]. Thus, serum lipids do not reflect insulin resistance in the same way among AA and EA.

Serum Lipids, Obesity, and Cardiovascular Disease Risk

There is also some evidence of a different association between serum lipids and cardiovascular disease among African ancestry individuals. The INTERHEART Africa project was an extension of the original INTERHEART case-control study and added 9 sub-Saharan African countries to the initial 52 countries in an effort to assess the association of cardiovascular disease risk factors and myocardial infarction. The ratio of the apolipoprotein B (associated predominantly with low-density lipoprotein particles) to apolipoprotein A-1 (associated with HDL particles) had a much lower odds ratio among Black Africans compared to Europeans/other Africans (3.43 vs 6.88) [59]. Similarly, the Framingham Heart Study equations, designed based on the association of risk factors (including serum lipids) with cardiovascular outcomes in a prospective study of EA were not highly predictive of cardiovascular disease mortality in AA [60]. The Dallas Heart Study reports interethnic differences in the association between HDLC and coronary heart disease: while there was an inverse association among EA, HDLC and coronary heart disease were not associated among AA [61]. In a study of individuals in the UK, African Caribbeans had a 35% lower risk of coronary heart disease and 50% higher risk of stroke than European ancestry individuals, even after adjusting for traditional risk factors, including BMI and serum lipids [62].

Similarly, there is also evidence of a different association between obesity and cardiometabolic disease by ethnicity. At every category of BMI, AA participants in the Jackson Heart Study had higher prevalence of type 2 diabetes and hypertension than EA in the Framingham Heart Study, despite lower prevalence of high TG and similar prevalence of low HDLC. Significant ethnicity × BMI interactions were observed in the association between BMI and each of these outcomes, indicating that across obesity categories, BMI was more strongly associated with negative metabolic consequences in EA compared to AA [63]. Similar observations have been made in other studies [20, 64–66]. In the INTERHEART Africa project, the odds ratio for abdominal obesity among Black Africans was less than half that in Europeans/other Africans (2.01 vs. 5.53) [59].

The Metabolic Syndrome

Given the known interrelatedness between risk factors and outcomes for cardiovascular disease and T2D, the metabolic syndrome (MetSyn) classification was designed to capture the impact of several cardiometabolic risk factors (low TG, high HDLC, high fasting blood glucose, high abdominal obesity, and high blood pressure) in order to identify individuals at increased risk of disease for targeted interventions. Given the generally healthier lipid profile among AfAnc individuals and that concentration of serum lipids are evaluated in 2 of the 5 included thresholds, one might expect a decreased prevalence of MetSyn among AA compared to EA. This has been observed, particularly among adolescents and children. Among NHANES adolescents, the prevalence of MetSyn among AA was less than half that among EA, and this primarily reflected much lower presence of high TG (12.9% in AA, 38.1% in EA, and 45.0% among Hispanic Americans) and low HDLC (19.0% of AA, 35.7% in EA, and 41.6% in Hispanic Americans)[67]. Among adults, however, the prevalence of MetSyn is not consistently lower among AA compared to other ethnic groups. While the presence of dyslipidemia, especially high TG, is still generally lower among AA than EA, a higher prevalence of elevations in blood pressure and fasting blood glucose may offset the more favorable lipid profile. Among women, especially, higher waist circumference often drive a greater prevalence of MetSyn among AA, while the prevalence of MetSyn among AA men remains lower [4, 68] or equivalent to [69]what is observed among EA. Data on the prevalence and contributors to MetSyn among Africans adds further complexity to this picture. While AA generally have high or comparable HDLC levels compared to EA, West Africans have a higher degree of low HDLC [70]. In contrast to studies in the US, where the prevalence of low HDLC and high TG are similar [69], in studies in Africa, these MetSyn components are often highly discordant. Among Nigerian T2D individuals, 61.8% had low HDLC, while only 10% had low TG [71]. Among Ghanaian hypertension patients with MetSyn, 96% had low HDLC, but only 32% had high TG [72]. In a Ugandan rural general population, 71.3% had low HDLC, but only 5% had high TG [73]. As seen among AA, higher abdominal obesity fuels higher MetSyn rates among women compared to men [36, 71].

While the evidence clearly indicates that interethnic differences in the distribution of serum lipids reduce the number of African ancestry individuals that are classified as having MetSyn, it is a separate question whether or not that classification is under-identifying individuals that are at risk. While the mismatch in the prevalence of MetSyn compared to cardiovascular disease and type 2 diabetes in AfAnc individuals certainly supports the under-identification of this rubric, there is further evidence that is relevant.

MetSyn has been associated with greater vascular dysfunction in both AA and EA.[69] However, in individuals without MetSyn, AA had worse vascular function than EA[69, 74], suggesting that at-risk AA individuals are not being sufficiently captured by the MetSyn criteria. Impaired endothelium-dependent dilation, a key early event in the development of atherosclerosis [75], and the metabolic determinants of endothelium-dependent vasodilation differed by ethnicity: systolic blood pressure was a more significant predictor among in AA, while insulin resistance was a more significant predictor among EA [74]. Increased weighting of HTN in AA eliminated ethnic differences in some vascular measures, but did not eliminate the worse microvascular endothelial function observed in AA compared to EA subjects without MetSyn, suggesting that factors other than blood pressure are contributing to the increased microvascular dysfunction seen in AA [69]. The prevalence of peripheral artery disease, which usually reflects underlying atherosclerosis in peripheral arteries, is also informative. Among individuals from the NHANES dataset with peripheral artery disease, the proportion of individuals who were classified as having MetSyn was much smaller among AA (24.5%) compared to EA (41.5%) and Hispanic (38.9%). Conversely, the prevalence of peripheral artery disease among those with MetSyn was significantly higher in AA (12.0%) compared to EA (7.2%) and Hispanic (4.6%) individuals [76]. Individuals with peripheral artery disease are often asymptomatic, but at high risk of cardiovascular disease events, but these data indicate that AA with this disease are less likely to be identified by MetSyn, and, when they are, more subclinical disease may be present [76].

Genomic Influences on Interethnic Differences in Serum Lipids

Given the consistently lower TG distribution across African ancestry populations despite divergent environmental backgrounds, it seems clear that genetic factors contribute to these interethnic differences. The less consistent differences with HDLC, particularly the generally higher HDLC among AA compared to other ethnicities in the US compared with the generally lower HDLC among African ancestry populations in Africa, support an important role for an environmental factor(s) that differs in the US and Africa. AA are an admixed population, with chromosomes that are a mosaic of segments that are inherited from two main parental populations, West African and European. In an individual, ancestry at a particular locus can be estimated based on allele frequencies in the parental populations giving a probability that the person has 0, 1, or 2 African alleles at that site. By averaging these values across the genome, an individual’s genome-wide average ancestry can be estimated. With considerable variability, AA have approximately 80% African ancestry. Simple association tests show that a 10% increase in African ancestry correlates with a 1% decrease in TG and a 0.7 mg/dl increase in HDLC, after adjustment for covariates[77]. Similar results have also been found in other studies[78, 79].

Admixture mapping is a process that takes advantage of interethnic differences in phenotypes to isolate regions of the genome where local ancestry in admixed individuals is associated with a trait: if European ancestry is higher in a particular region among AA with low HDLC, then variants in that region may be associated with HDLC. For HDLC, admixture mapping and subsequent follow-up pointed to the influence of LPL for HDLC concentration[80], as had been previously reported [79]. For TG, 3 independent signals were discovered, one in an intergenic region, one in NRXN3 and one in TTC7B[80]. Interestingly, the associated SNP in NRXN3 had previously been associated with obesity, BMI, and waist circumference in GWAS [81–83]. A previous admixture study had also identified GCKR as potentially contributing to interethnic differences in TG [79]; this locus is an established lipids locus [84].

Lipoprotein lipase is a central enzyme in the concentration of serum lipids. Variants in this gene, particularly missense variant rs328, are well-established as contributing to serum lipid concentrations. rs328 is associated with increased LPL activity, which increases lipolysis, decreasing TG and increasing HDLC concentrations. This variant is less prevalent among African ancestry than European ancestry individuals (minor allele frequency among 1000 Genomes African: 0.05 vs. European ancestry: 0.12 [85]). Interestingly, ethnic differences at this locus have been observed. A study of AA from the Howard University Family Study found that associations between rs328 and serum lipids differed depending on the ancestry of the individual at that locus (the local ancestry). Among AA with European ancestry at this locus, the variant was associated with a more favorable lipid profile than among those with only African ancestry at this locus. For TG, the association among AA with only European local ancestry matched what had previously been reported for EA. For HDLC, however, the association among AA with only European local ancestry was much higher than what had been previously reported for EA (15 vs. ~ 5 mg/dl). For both TG and HDLC, the associations among individuals with only African local ancestry matched associations found among West Africans at that locus (Figure 2) [86]. In agreement with these data, a study of AA in the Jackson Heart Study also showed a larger effect size for rs328 in the presence of European local ancestry [79]. The authors propose that these results suggest that a causal variant for this effect is not rs328, but a variant that is tightly linked with rs328 on a European ancestry background, but not on an African background. In a large meta-analysis of more than 100,000 EA individuals, LPL variant rs12678919 (in tight linkage disequilibrium with rs328) was strongly associated with a favorable lipid profile (higher HDLC and lower TG). This association was replicated in a follow-up sample of AA. There was no statistically significant heterogeneity for AA compared to EA, however, the association in AA was much less statistically significant than in the comparably-sized EA replication analysis (p=5 × 10⁻³ vs. 1 × 10⁻¹¹ for n=8,061 and 7,063). [84].

Figure 2. Interethnic Differences in the Association between LPL variant rs328 and Serum Lipids.

Reported data on associations for rs328 for individuals of European [HDL][114–117] [TG] [118] and African ancestry individuals [86].

A study of black and white South African women showed interethnic differences in the associations between three established lipids genes, LPL, LIPC (hepatic lipase), and CETP (cholesterylester transfer protein). In contrast to the direction of the ancestry differences at LPL reported above, the associations for LPL were significant in black South African women, but not among their white counterparts, despite a larger sample size and a higher prevalence of the variant. Similar interethnic differences were observed for LIPC and CETP[15]. In contrast, associations in these genes in EA and AA have been observed to be broadly similar[78, 84, 87, 88]. Both LPL and CETP have been implicated in genome-wide studies of MetSyn in EA [89, 90], AA [91], and, recently, in Africans[92].

Beyond these notable variants, generally similar results have been seen for genome-wide association studies of lipids across ethnicities [78, 84, 87]; however, the relatively limited degree of replication of EA-discovered loci among AA has been noted [7, 87]. For instance, in a large meta-analysis of over 100,000 EA identified 95 lipids loci, and attempted replication of these in 8,071 AA; only 30% of the SNPs tested for each trait were generalizable to AA (directionally consistent and p<0.05; Table S12 of that paper)[84]. Failure to replicate findings among African ancestry individuals could represent: relatively smaller sample sizes among African ancestry studies, interethnic differences in linkage disequilibrium such that variants tagging a causal locus among EA may not tag the locus among AA, interethnic differences in allele frequencies, or difference across populations in the relative importance of a particular locus as a result of different selection pressures. Larger genome-wide studies of African ancestry individuals are needed. Additionally, careful consideration of linkage disequilibrium structure differences across ancestries can yield new insights in previously identified regions[87, 93].

Future Directions: Improving the Identification of At-Risk Individuals

In the context of interethnic differences in the prevalence of risk factors and in the association between risk factors and disease, ancestry-agnostic thresholds seem inadequate. Yet identifying individuals of African ancestry at higher risk of cardiometabolic disease is an urgent public health need both in the US, with the disproportionate burden of disease is borne by AA, and in Africa, where urbanization and Westernization are contributing to some of the fastest anticipated increases in cardiometabolic diseases in the world. What can be done to improve identification of these at-risk individuals for more aggressive treatment?

It has been proposed that the MetSyn could be altered in order to improve risk identification among African ancestry individuals. Some have suggested that establishing ethnic-specific thresholds for serum lipids should be adopted to reflect the lower values at which serum lipids may associate with insulin resistance[94, 95]. It has also been suggested to alter the obesity threshold values: lowering the threshold values for African ancestry men, while threshold values remain high for women [96, 97]. Weighting hypertension more heavily in the classification of MetSyn among AA has been shown to better distinguish between AA with and without microvascular dysfunction, although this improvement does not eliminate interethnic differences in endothelial dysfunction [69]. Another approach that has been proposed is creating a continuous MetSyn severity score that allows for different loadings for each component across ethnicity and sex groups. Analyses of NHANES data resulted in equations that down-weighted TG in AA compared to EA, for instance. Importantly, this method also provides a useful tool for follow-up of an individual’s risk over time [98].

Beyond the MetSyn classification, it is worth considering other risk factors that may result in an improved identification of at-risk African ancestry individuals. One such possible biomarker is lipoprotein(a), which has been associated with increased cardiovascular disease outcomes [99–101]. Lipoprotein(a) is typically much higher in AA than EA; for instance, median Lp(a) was 4.3 mg/dl in EA and 12.8 mg/dl in the AA of Atherosclerosis Risk in Communities study [101]. Lipoprotein(a) has been associated with similar increased cardiovascular disease risk in both EA and AA [101]. Additionally, a measure informative for chronic kidney disease has been proposed to improve identification of cardiometabolic risk among African ancestry populations [46].

Overall, the presence of important interethnic differences in the distribution of serum lipids, along with a range of other factors that are relevant for cardiometabolic risk motivate the need to conduct more research in African ancestry populations. Studies that investigate the association between serum lipids and risk of disease among African ancestry populations, especially in a range of environmental contexts, are urgently needed. Additionally, studies evaluating known and novel biomarkers that may more effectively identify cardiometabolic risk among African ancestry populations could be of enormous public health significance given the large and growing burden of disease among these individuals.

Highlights.

• African ancestry is associated with a healthier serum lipid profile.

• African Americans do not have lower cardiometabolic risk despite lower lipids.

• Interethnic differences in underlying physiology may explain the metabolic paradox.

Abbreviations

AA

African American

BMI

Body mass index

EA

European ancestry individuals

HDLC

High-density lipoprotein cholesterol

LPL

Lipoprotein lipase

MetSyn

Metabolic Syndrome

NHANES

National Health and Nutrition Examination Survey

T2D

Type 2 Diabetes

TG

Triglycerides

US

United States