The prevalence of hyperinsulinemia has increased by about 35% among nondiabetic adults in the past decade in the United States 1. Insulin resistance has been proposed as an underlying cause of type 2 diabetes and the metabolic syndrome 23 and has been found to be associated with increased risk for cardiovascular diseases 245. The availability of relatively simple measures to identify apparently healthy people who are sufficiently insulin resistant to be at increased risk of developing type 2 diabetes, cardiovascular disease, and the numerous other clinical syndromes that occur with increased frequency in insulin-resistant persons would be of significant clinical benefit 678.

The triglyceride to high-density lipoprotein cholesterol (TG/HDL-C) concentration ratio has been reported to be as closely related to insulin resistance as is the fasting plasma insulin concentration 910, a commonly used surrogate estimate of insulin resistance. Previous studies have demonstrated that fasting insulin was highly correlated with homeostasis model assessment (HOMA) and quantitative insulin sensitivity check index (QUICKI) (correlation coefficients >0.95) of insulin resistance among nondiabetic individuals 111213. In contrast to measures of lipid and lipoprotein concentrations, there has been no attempt to standardize assays of insulin concentration to date. Thus, it seemed possible that use of the TG/HDL-C ratio, based on commonly available and standardized measurements, could help clinicians identify persons who were not only insulin resistant but also displayed the characteristic dyslipidemia of people with this defect in insulin action 910. The relationship between the TG/HDL-C ratio and a direct measure of insulin resistance was first reported among 258 overweight or obese adults, of whom 87% were non-Hispanic whites 9. This association has been replicated in a larger clinically based sample that represented the general population 10. Elsewhere, in a study in an East African population, the TG/HDL-C ratio was found to be significantly associated with insulin resistance as measured by HOMA 14. In contrast, recent studies have reported that the triglyceride or triglycerides/HDL-C ratio was not significantly associated with insulin resistance in black adults 15 and adolescents 16.

In previous studies, black males have had lower triglyceride concentrations and higher HDL-C concentrations than white males, while versus white females, black females have had similar or lower triglyceride concentrations and comparable HDL-C concentrations 17. Elsewhere, Mexican American males were found to have higher fasting serum insulin (FSI) concentrations than white or black males, while adult black or Mexican American females were reported to have higher FSI concentrations than their white counterparts 18. In US youth aged 12–17 years, there were no significant variations by race/ethnicity among males, while black females had higher FSI concentrations than white or Mexican American females 19.

Because of racial/ethnic variations in triglyceride, HDL-C, and FSI concentrations, the association of the TG/HDL-C ratio with FSI may differ by race/ethnicity. To the best of our knowledge, no studies have investigated the possibility of such differences. Thus, the goal of this study was to use a nationally representative sample to determine whether the association of triglyceride, HDL-C, and the TG/HDL-C ratio with the FSI may differ between non-Hispanic whites, non-Hispanic blacks, and Mexican Americans.

The distribution of triglycerides, HDL-C, and TG/HDL-C ratio appeared to be similar between men and women, but different between racial/ethnic groups (Figure 1). Pearson correlation coefficients were -0.41, 0.93, and -0.71 for triglycerides versus HDL-C, triglycerides versus the TG/HDL-C ratio, and HDL-C versus the TG/HDL-C ratio, respectively.

Among men, non-Hispanic blacks had a lower mean triglyceride concentration, a higher mean HDL-C concentration, and a lower TG/HDL-C ratio than both non-Hispanic whites and Mexican Americans (Table 1). In contrast, Mexican American men had a higher FSI concentration than their non-Hispanic white and non-Hispanic black counterparts. Among women, non-Hispanic blacks had a lower mean triglyceride concentration and lower TG/HDL-C ratio than both non-Hispanic whites and Mexican Americans. Mexican Americans, in contrast, had a lower mean HDL-C concentration than both non-Hispanic whites and non-Hispanic blacks. Non-Hispanic blacks and Mexican Americans had a higher mean FSI than non-Hispanic whites.

After adjustment for potential confounding effects, triglycerides, HDL-C, and the TG/HDL-C ratio were significantly associated with FSI in the three racial/ethnic groups (Table 2). No significant difference in the regression coefficients between any two racial/ethnic groups was detected (all P > 0.017 with Bonferroni adjustment). There was no significant interaction between race/ethnicity and triglycerides, HDL-C, or the TG/HDL-C ratio (P ranged from 0.22 to 0.98) on the FSI in the combined data.

The 75^th percentile cutoff point for FSI was 78.77 pmol/L (13.131 uU/mL) among nondiabetic people in NHANES 1999–2002. The area under the ROC curve of the TG/HDL-C ratio (AUC = 0.75; 95% CI, 0.73–0.78) was larger than that of triglyceride (AUC = 0.72; 95% CI, 0.69–0.74; P < 0.0001) or HDL-C (AUC = 0.72; 95% CI, 0.69–0.74; P = 0.0007) alone for the prediction of hyperinsulinemia (Figure 2). According to the maximum value of the Youden index, the optimal cutoff point for the TG/HDL-C ratio was 1.2 in mmol/L unit (or equivalent to 3 in mg/dl unit) for both non-Hispanic whites and Mexican Americans and 0.9 in mmol/L unit (or equivalent to 2 in mg/dl unit) for non-Hispanic blacks. The corresponding sensitivity and specificity were 70.6% and 71.0% in non-Hispanic whites, 61.6% and 77.4% in non-Hispanic blacks, and 64.1% and 71.1% in Mexican Americans.

The unadjusted prevalence of hyperinsulinemia was more than doubled among people with TG/HDL-C ≥ 3.5 (or the new cutoff point) than those with TG/HDL-C < 3.5 (or the new cutoff point); and this pattern was consistent across the three BMI categories and the three racial/ethnic groups (Table 3). Using the cutoff points of 3.0 for non-Hispanic white and Mexican Americans and 2.0 for non-Hispanic blacks yielded lower prevalence rates of hyperinsulinemia, but larger prevalence ratios compared to the cutoff point of 3.5 among non-Hispanic blacks and Mexican Americans, and among people with a BMI <25 or between 25 and 29 kg/m². There was a significant interaction between the TG/HDL-C ratio and BMI for both sets of cutoff points, indicating that the association between TG/HDL-C ratio and hyperinsulinemia may differ by BMI categories in that this association was stronger among people with a lower category BMI than those with a higher BMI. However, there was no significant interaction between the TG/HDL-C ratio and race/ethnicity for hyperinsulinemia by using the new cutoff point.

Using a large and nationally representative sample, we found that the TG/HDL-C ratio was significantly associated with FSI among nondiabetic adults of three major racial/ethnic subpopulations in the United States: non-Hispanic whites, non-Hispanic blacks, and Mexican Americans. We also detected racial/ethnic differences in predicting hyperinsulinemia, as both non-Hispanic whites and Mexican Americans had a larger optimal cutoff point of the TG/HDL-C ratio for the determination of hyperinsulinemia than did non-Hispanic blacks.

The association of insulin resistance with TG/HDL-C has been reported in overweight and obese whites 910. Because the clinical samples that were used in these studies consisted primarily of white participants, the generalizability of this association has not been established. A recent study found that triglyceride concentration and the TG/HDL-C ratio were not significantly associated with insulin resistance (as determined from the insulin sensitivity index) in 125 African American participants 15; this failure to find an association between the TG/HDL-C ratio and insulin resistance in blacks was perhaps attributable to the small sample and thus a limited power to detect associations. Our results, however, provide evidence that the TG/HDL-C ratio is associated with FSI in nondiabetic adults regardless of race/ethnicity.

Racial/ethnic differences in triglyceride concentrations, HDL-C values, and the FSI have been widely reported among adults and adolescents in previous studies 1416171819. In the present study, non-Hispanic blacks had lower triglyceride concentrations than non-Hispanic whites or Mexican Americans among both men and women; for HDL-C concentrations among males the values were highest among non-Hispanic blacks, but among women no real difference was seen in this value between blacks and whites. In both sexes the TG/HDL-C ratio was lower in blacks than in whites or Mexican Americans. In addition, a significant association of the TG/HDL-C ratio with insulin resistance (as measured by HOMA) was stronger among non-Africans than Africans 14, and increased triglyceride concentrations were associated with hepatic insulin resistance in white but not black adolescents 16. Lipoprotein lipase activity has been found to be higher in blacks than in whites, and this may lead to a lower triglyceride concentration in blacks 29. An elevated triglyceride concentration commonly causes low HDL-C values or regulates HDL-C remodeling 303132, and thus these two clinically useful measures may be, at least in part, causally linked. In light of racial/ethnic differences in triglyceride concentrations, HDL-C, and the TG/HDL-C ratio, different cutoff points may be needed to define dyslipidemia. In fact, use of a single cutoff value of ≥ 150 mg/dl (or 1.69 mmol/L) for triglycerides as criteria to classify people with dyslipidemia have led to underestimation of the prevalence of the metabolic syndrome in blacks as compared with whites, particularly in men 3334.

A TG/HDL-C ratio ≥ 3.5 has been proposed as a cutoff value to predict the presence of small-density LDL phenotype; McLaughlin and colleagues found that this cutoff had high sensitivity (79%) and specificity (85%) in the white population 10. These authors also found, however, that this cutoff value had only moderate ability to predict insulin resistance (sensitivity 47%, specificity 88%) or the metabolic syndrome (sensitivity 46%, specificity 92%) 10. We found, however, that the magnitude of the association of hyperinsulinemia with an elevated TG/HDL-C ratio was different when using the single cutoff point of 3.5, suggesting that a single cutoff point may not be applicable across diverse populations. In fact, our study showed, for the first time, that the optimal cutoff point of the TG/HDL-C ratio for prediction of hyperinsulinemia was 3.0 (in mg/dL unit) for non-Hispanic whites and Mexican Americans and 2.0 (in mg/dL unit) for non-Hispanic blacks, with reasonable sensitivity and specificity. Applying these race/ethnicity-specific cutoff points yielded similar magnitude of strength in the association across racial/ethnic subgroups.

It is of great interest that the association of the TG/HDL-C ratio with hyperinsulinemia was stronger among people with a BMI <25 kg/m² than those with a BMI ≥ 30 kg/m². As shown in a previous study, about 16% people with normal weight (BMI < 25 kg/m²) was identified to be insulin resistant 35; thus search for clinically simple and useful biomarkers to detect insulin resistance among people with normal weight is necessary. Although BMI, a less expensive and convenient measure, may serve as an indicator for insulin resistance among people with excessive weight, an elevation in the TG/HDL-C ratio could be a novel marker for hyperinsulinemia among people with normal weight in routine clinical practice.

Our study has several strengths that deserve attention. First, the generalizability of our results has been enhanced by the use of a nationally representative sample. Second, the large sample for each of the three racial/ethnic groups enables us to empirically test differences between the groups in the associations of the TG/HDL-C ratio with FSI. A limitation stems from the use of FSI as a surrogate measure of insulin resistance in our data, but previous studies have shown that fasting insulin is a valid and reliable surrogate measure of insulin resistance 13. Furthermore, although a direct measure would theoretically be preferable, the lack of standardized insulin assay 36 greatly limits the clinical utility of that approach for identifying insulin-resistant persons. Without a standard insulin assay, no absolute value can be developed to predict the presence of insulin resistance across laboratories or clinical settings. Thus, if one wants to use fasting insulin to assess the state of insulin resistance, it is necessary to develop different cut-points for every assay. In contrast, triglyceride and HDL cholesterol concentrations are standardized and readily available measures in routine clinical practice. Consequently, it seems more clinically effective to use relatively simple metabolic markers such as the TG/HDL-C concentration ratio in the effort to identify apparently healthy people at increased risk of developing a variety of adverse clinical outcomes.

In conclusion, our findings have clinical and public health implications. In clinical settings, the TG/HDL-C ratio could be used as an indicator of insulin resistance across racial/ethnic subpopulations. In large health surveys, it could be used to monitor trends in cardiovascular health in diverse populations. Our results add further support to the notion that the TG/HDL-C ratio may be a clinically simple and useful indicator for insulin resistance among nondiabetic adults regardless of race/ethnicity. Even so, racial/ethnic-specific cutoff points of the TG/HDL-C ratio for the determination of insulin resistance or the risk of type 2 diabetes or cardiovascular diseases may be needed. Future research is warranted to assess the predictive power of the TG/HDL-C ratio for the metabolic syndrome, type 2 diabetes, or cardiovascular disease.

The author(s) declare that they have no competing interests.

CL, ESF, and YXM conceived the study. CL performed the statistical analysis and prepared the initial draft of the article. ESF, YXM, AHM, and GMR critically revised the manuscript for important intellectual content. All authors read and approved the final version of the manuscript.