Identifying cardiovascular disease risk and outcome: use of the plasma triglyceride/high-density lipoprotein cholesterol concentration ratio versus metabolic syndrome criteria


Correspondence : Martin R. Salazar, 14 n 320, La Plata (1900), Buenos Aires, Argentina.

(fax: 54-221-4129164; e-mail:



Metabolic syndrome (MetS) has been shown to predict both risk and CVD events. We have identified sex-specific values for the triglyceride/high-density lipoprotein cholesterol (TG/HDL-C) ratio associated with an unfavourable cardio-metabolic risk profile, but it is not known whether it also predicts CVD outcome.


To quantify risk for CVD outcomes associated with a high TG/HDL-C ratio and to compare this risk with that predicted using MetS, a population longitudinal prospective observational study was performed in Rauch City, Buenos Aires, Argentina. In 2003 surveys were performed on a population random sample of 926 inhabitants. In 2012, 527 women and 269 men were surveyed again in search of new CVD events. The first CVD event was the primary endpoint. Relative risks for CVD events between individuals above and below the TG/HDL-C cut-points, and with or without MetS, were estimated using Cox proportional hazard.

Main Outcome

The first CVD event was the primary endpoint. Relative risks for CVD events between individuals above and below the TG/HDL-C cut-points, and with or without MetS, were estimated using Cox proportional hazard.


The number of subjects deemed at ‘high’ CVD risk on the basis of an elevated TG/HDL-C ratio (30%) or having the MetS (35%) was relatively comparable. The unadjusted hazard risk was significantly increased when comparing ‘high’ versus ‘low’ risk groups no matter which criteria was used, although it was somewhat higher in those with the MetS (HR = 3.17, 95% CI:1.79–5.60 vs. 2.16, 95% CI:1.24–3.75). However, this difference essentially disappeared when adjusted for sex and age (HR = 2.09, 95% CI:1.18–3.72 vs. 2.01, 95% CI:1.14–3.50 for MetS and TG/HDL-C respectively).


An elevated TG/HDL-C ratio appears to be just as effective as the MetS diagnosis in predicting the development of CVD.


Insulin resistant individuals are at increased risk of developing hypertension, type 2 diabetes and coronary heart disease [1-3]. The ability to identify these high risk individuals before the development of manifest disease would be of substantial clinical benefit, and the possibility that the plasma concentration ratio of triglyceride (TG)/high-density lipoprotein cholesterol (HDL-C) might fulfil this function has been raised [4, 5]. In that context, we have recently shown in a population primarily of European ancestry that values of ≥ 2.5 (women) and ≥ 3.5 (men) provide useful cut-points to identify individuals who are insulin resistant, and at increased cardio-metabolic risk [6].

In the last several years, a diagnosis of the metabolic syndrome (MetS) has been widely used to identify such individuals, and two recent meta-analyses [7, 8] estimated that the MetS is associated with an ~2-fold increase in cardiovascular disease (CVD). However, it should be noted that the majority of the studies included in these reports have used alternative definitions of the MetS, and not the recent ‘harmonized’ version that is currently recommended [9].

The purpose of this prospective study was to extend the clinical evaluation of the TG/HDL-C ratio and had two major goals: 1) to quantify, in a population without known cardiac disease, the risk for CVD outcomes associated with TG/HDL-C values above these sex-specific thresholds, and 2) to compare this risk with the obtained using the newly harmonized diagnostic criteria of the MetS [9].


A prospective epidemiological study, focused on hypertension, renal disease and other cardio-metabolic risk factors, was conducted in Rauch City, province of Buenos Aires, Argentina (36°45′00″ south latitude and 59°04′00″ west longitude) between October 2003 and February 2012. According to the last national census, there were 8 246 inhabitants ≥ 15-year-old in the urban area of Rauch City. In 2003, surveys were performed and random samples taken from subjects between 15 and 80 years old who lived in chosen blocks (n = 1308, 855 women 51 ± 17 years old and 453 men 52 ± 16 years old, P between gender = 0.628). The methodology used to obtain measurements of clinical and biochemical variables have been previously published [10, 11]. In brief, blood pressure was measured sitting, after a minimum resting-period of 5 min, using a mercury sphygmomanometer. Phase I and V Korotkoff sounds were used to identify systolic blood pressure (SBP) and diastolic blood pressure (DBP) respectively; SBP and DBP values were an average of three different measurements separated by two minutes from one another. Weight was determined with individuals wearing light clothes and no shoes. Height was also measured without shoes, using a metallic metric tape; waist circumference (WC) was measured with a relaxed abdomen using a metallic metric tape on a horizontal plane above the iliac crest; body mass index (BMI) was calculated using the formula kg m². Concentrations of plasma glucose, TG, HDL-C and fasting plasma insulin (FPI) were determined after an overnight (12 h) fast. Low-density lipoprotein cholesterol levels (LDL-C) were estimated by the Friedewald formula [12]. Plasma for the insulin measurements was extracted by centrifugation (15 min at 3000 r.p.m.), and frozen at −20 °C until assayed. FPI concentrations were determined using an immunoradiometric assay, with two monoclonal antibodies against two different epitopes of the insulin molecule. The inter- and intra-assay coefficients of variation (CV) were 8.0% and 3.8% respectively with the lowest detectable level of 1.4 pmol L. The homeostasis model assessment of insulin resistance (HOMA-IR) was calculated with the formula ([Insulin(uU mL−1) × glucose (mg 100 mL−1)/18]/22.5) [13]. Fifty-four subjects who had previously suffered CVD events were excluded. Measurements necessary to diagnose MetS were available in 926 individuals, 622 women and 304 men, with similar ages (52 ± 16 years).

In 2012, 796 individuals (86% of the baseline sample), 527 women and 269 men (or their relatives in case of death), could be surveyed again to obtain information concerning incident CVD events; the remaining inhabitants (n = 130) could not be found because they had moved out of Rauch city at the moment of the survey. As table 1 shows, there was no significant difference in the baseline characteristics between subjects with and without follow-up period. The first CVD event, including angina pectoris, myocardial infarction, myocardial revascularization and fatal or nonfatal stroke was defined as the primary endpoint. A structured interview was conducted with each participant by specially trained nurses and social workers. The collected data were then evaluated by a highly qualified internist (blinded with respect to the subject's baseline CVD risk factors) to assign a specific outcome for every event. When necessary, available medical records were also reviewed.

Table 1. Baseline characteristics of inhabitants with and without follow-up period
yes n = 796 Mean ± SDno n = 130 Mean ± SD
  1. SD, standard deviation; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; Total-C, total cholesterol; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; FPI, fasting plasma insulin; WC, waist circumference.

  2. a

    Student's t-test for independent samples.

  3. b


Age (years)51 ± 1654 ± 180.055a
Women (%)
Diabetes (%)
Cigarettes (n day−1)3 ± 72 ± 50.230a
Alcohol (g week−1)67 ± 14742 ± 1160.067a
BMI (kg m−2)26 ± 526 ± 110.391a
WC (cm)93 ± 1394 ± 140.690a
SBP (mmHg)132 ± 19135 ± 200.074a
DBP (mmHg)82 ± 1283 ± 110.319a
Glucose (mmol L−1)5.6 ± 2.65.5 ± 1.10.603a
Total-C (mmol L−1)6.0 ± 1.35.9 ± 1.30.623a
LDL-C (mmol L−1)3.7 ± 1.23.6 ± 1.30.357a
HDL-C (mmol L−1)1.5 ± 0.61.6 ± 0.50.102a
Triglycerides (mmol L−1)1.6 ± 0.91.5 ± 0.80.236a
FPI (pmol L−1)53.8 ± 40.751.8 ± 31.50.656a
HOMA-IR1.9 ± 2.11.7 ± 1.10.405a

Previously published cut-points of plasma TG/HDL-C concentration ratios of 2.5 and 3.5 (expressed both in mg dL−1) [6], for women and men respectively, were used to classify participants as being either at ‘high’ or ‘low’ CVD risk. Participants were also divided into those with the MetS (‘high’ risk) or without the MetS (‘low’ risk), using the ‘harmonized’ version of the MetS [9], in which 3 of the following 5 criteria are required for diagnosis: (i) WC ≥ 102 cm in men and ≥ 88 cm in women; (ii) HDL-C < 1.0 mmol L−1 in men and < 1.3 mmol L−1 in women; (iii) TG ≥ 1.7 mmol L−1; (iv) SBP ≥ 130 mmHg or DBP ≥ 85 mmHg; (v) fasting plasma glucose ≥ 5.6 mmol L−1. Diabetes mellitus was defined as a fasting plasma glucose ≥ 7 mmol L−1 or use of glucose-lowering drugs.

The relative risks for CVD events between individuals above and below the TG/HDL-C sex-specific cut-points, and with and without MetS, were estimated using five Cox proportional hazard models: (i) unadjusted, (ii) adjusted for age and sex, (iii) adjusted for sex, age and diabetes, (iv) adjusted for sex, age and total cholesterol levels and (v) adjusted for sex, age, diabetes, total cholesterol levels and number of cigarettes smoked daily. The risk was expressed as hazard ratio (HR) and its 95% confidence interval (95% CI). To assess whether the TG/HDL-C ratio is a better predictor of CVD than each of its components, we reproduced the five Cox proportional hazard models using the top quartile of TG and the top quartile of HDL-C to define ‘high’ risk. Finally, previously published studies have suggested using TG/HDL-C values of 3.0 [4] and 3.5 [5] for the identification of insulin-resistant individuals. Thus, we have compared these values with our sex-specific cut-points with Cox models using TG/HDL-C ratio > 3.0 and > 3.5 to define ‘high’ risk.

Continuous variables were expressed as mean and standard deviation (SD) and compared using independent samples t-test, and proportions were expressed as percentages and compared by chi-square; statistical analyses were performed using spss software (SPSS Inc., Chicago, IL, USA), where two-tailed and P values < 0.05 were considered statistically significant.


Table 2 compares the baseline CVD risk profiles of the experimental population, divided into ‘low’ and ‘high’ risk subgroups on the basis of the TG/HDL-C concentration ratio or the diagnostic criteria of the MetS. A somewhat greater number of ‘high’ risk individuals were identified by having the MetS (278, 35%) than by an elevated TG/HDL-C ratio (242, 30%). The prevalence of diabetes was higher in ‘high’ risk subgroups, identified with use of either the TG/HDL-C ratio or MetS criteria, but there were no differences in sex distribution, alcohol consumption or number of cigarettes smoked in the ‘high’ and ‘low’ risk groups, irrespective of the index used to stratify them.

Table 2. Cardiovascular disease events and baseline cardio-metabolic risk profile in individuals at ‘low’ or ‘high’ risk on the basis of the TG/HDL-C concentration ratio or the metabolic syndrome diagnostic criteria
VariableLow TG/HDL-C ratio (= 554)High TG/HDL-C ratio (= 242)P-valuesMetS no (= 518)MetS yes (= 278)P-values
  1. SD, standard deviation; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; FPI, fasting plasma insulin; Total-C, total cholesterol; LDL-C; low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; WC, waist circumference; CVD, cardiovascular diseases.

  2. Data are expressed as mean ± SD or percent.

  3. a

    Student's t test for independent samples.

  4. b

    Mantel Cox.

  5. c


Age (years)50 ± 1755 ± 14<0.001a47 ± 1659 ± 12< 0.001a
Women (%) c65.866.80.772 c
CVD events (% subjects/10 years) b4.614.7<0.001b
Diabetes (%)4.913.2< .001c2.715.9<0.001c
Cigarettes (n day−1)3 ± 83 ± 70.257 a3 ± 83 ± 60.344 a
Alcohol (g week−1)63 ± 13776 ± 1700.269 a67 ± 13667 ± 1680.966 a
BMI (kg m−2)25.1 ± 4.527.5 ± 4.4<0.001 a24.1 ± 3.728.9 ± 4.4< 0.001 a
WC (cm)92 ± 1397 ± 11< 0.001 a89 ± 11101 ± 13< 0.001 a
SBP (mmHg)129 ± 18139 ± 19< 0.001 a126 ± 16144 ± 17< 0.001 a
DBP (mmHg)80 ± 1185 ± 12< 0.001 a78 ± 1089 ± 11< 0.001 a
Glucose (mmol L−1)5.4 ± 1.16.2 ± 4.4< 0.001 a5.1 ± 0.96.4 ± 4.1< 0.001 a
Total-C(mmol L−1)5.7 ± 1.26.4 ± 1.3< 0.001 a5.7 ± 1.26.4 ± 1.3< 0.001 a
LDL-C (mmol L−1)3.6 ± 1.24.0 ± 1.2< 0.001 a3.5 ± 1.24.0 ± 1.2< 0.001 a
HDL-C (mmol L−1)1.6 ± 0.71.3 ± 0.3< 0.001 a1.6 ± 0.71.4 ± 0.3< 0.001 a
Triglycerides (mmol L−1)1.1 ± 0.42.5 ± 1.0< 0.001 a1.2 ± 0.62.2 ± 1.1< 0.001 a
FPI (pmol L−1)47.9 ± 28.867.5 ± 57.7< 0.001 a46.2 ± 24.168.0 ± 57.9< 0.001 a
HOMA-IR1.6 ± 1.82.4 ± 2.5< 0.001 a1.5 ± 1.02.5 ± 3.1< 0.001 a

CVD risk factors were significantly worse in the ‘high’ risk subgroups compared with the ‘low’ risk subgroups. Furthermore, comparison of individual CVD risk factors related to blood pressure, lipoprotein metabolism and carbohydrate metabolism in the two ‘high’ risk groups appeared reasonably comparable, whether the TG/HDL-C concentration ratio or the MetS diagnostic criteria were used to identify risk. Moreover, FPI and HOMA-IR concentrations were essentially identical in individuals defined as being at ‘high’ or ‘low risk’ with either approach. Seventy nine percent of individuals were identified concordantly by both indices; therefore tests comparing the cardio-metabolic profiles resulting from the use of the TG/HDL-C ratio versus the MetS were not performed.

In the follow-up period (6343 subjects year−1, mean 8.0 ± 1.3) there were 51 CVD events: 31 nonfatal and 8 fatal coronary events and 7 nonfatal and 5 fatal strokes. There were also 36 noncardiovascular deaths, without any difference in incidence between ‘high’ versus ‘low’ risk groups as classified by either the TG/HDL-C ratio (HR = 1.16, 95% CI: 0.58–2.33) or a MetS diagnosis (HR = 1.09, 95% CI:0.55–2.15). Crude cumulative incidences of combined CVD outcomes, expressed as percent subjects/10 years of follow-up, were 6.1 in the low TG/HDL-C ratio group, 12.6 in the high TG/HDL-C ratio group, 4.6 in individuals without MetS and 14.7 in those who had MetS.

Table 3 demonstrates that the unadjusted HR for developing a CVD event was significantly increased in both groups designated as ‘high’ versus ‘low’ risk at baseline. Although the unadjusted HR was greatest in ‘high’ risk individuals identified by MetS criteria (HR = 3.17, 95% CI: 1.79–5.60 vs. 2.16, 95% CI: 1.24–3.75), this difference essentially disappeared after adjusting for differences in sex and age in Model 2 (HR = 2.09, 95% CI: 1.18–3.72 vs. 2.01, 95% CI: 1.14–3.50 for MetS and TG/HDL- C respectively) (Fig. 1). Furthermore, the HR continued to be significant after adjustments for diabetes, total cholesterol and cigarettes smoked in those at ‘high’ risk because of an elevated TG/HDL-C ratio or having the MetS. Results in Table 3 also show that alternate cut-points of 3.0 and 3.5 did not perform as well as the sex-specific cut-points in their comparability to the MetS with a TG/HDL-C ratio of 3.5 seemingly somewhat preferable to a value of 3.0.

Table 3. Hazard ratios (HR) of CVD events occurring in individuals at baseline with a ‘high’ versus a ‘low’ TG/HDL-C ratio compared to subjects with and without the MetS
 High vs. Low TG/HDL-CMetS = yes vs. MetS = no
  1. Model 1, Unadjusted; Model 2, Adjust according to sex and age; Model 3, Adjust according to sex, age and diabetes; Model 4, Adjust according to sex, age and total cholesterol levels; Model 5, Adjust according to sex, age, diabetes, cholesterol levels and number of cigarettes.

Cut-pointwomen > 2.5, men > 3.5> 3.0 for both sexes> 3.5 for both sexes3 criteria
Labelled as high risk, n (%)242(30)216 (27)167 (21)278 (35)
Cox modelsHR (95% CI) P HR (95% CI) P HR (95% CI) P HR (95% CI) P
Model 12.16 (1.24–3.75)0.0111.97 (1.12–3.46)0.0182.35 (1.33–4.16)0.0033.17 (1.79–5.60)0.001
Model 22.01 (1.14–3.50)0.0151.63 (0.92–2.89)0.0921.98 (1.11–3.53)0.0212.09 (1.18–3.72)0.012
Model 31.99 (1.13–3.52)0.0171.62 (0.91–2.67)0.0991.97 (1.10–3.52)0.0232.09 (1.17–3.76)0.013
Model 41.86 (1.05–3.30)0.0331.53 (0.86–2.71)0.1511.82 (1.01–3.29)0.0471.98 (1.11–3.55)0.021
Model 51.79 (1.02–3.21)0.0491.47 (0.82–2.63)0.1961.74 (0.96–3.17)0.0691.98 (1.09–3.57)0.024
Figure 1.

Unadjusted (a,b) and adjusted by sex and age (c,d) cumulative hazard of combined cardiovascular events (angina pectoris, myocardial infarction, myocardial revascularization and fatal or nonfatal stroke, in individuals classified as ‘high’ versus ‘low’ risk using the TG/HDL-C ratio (left) and according metabolic syndrome diagnosis (right).

Since an earlier study had indicated that plasma TG concentration was also a significant predictor of insulin resistance [4], we thought it was important to also evaluate the ability of this measurement to predict CVD. Consequently, we calculated the sex-specific plasma TG and HDL-C thresholds that separated the upper quartiles from the others, using the same methodology and sample as described previously to determine the most useful TG/HDL-C ratio [6]. Comparison of ‘high’ risk versus ‘low’ risk individuals using this TG cutoff point resulted in a marginally significant unadjusted hazard ratio for CVD events (HR = 1.70, 95% CI: 0.97–2.97, P = 0.067). After adjusting the model for sex and age the hazard ratio was not statistically significant (HR = 1.41, 95% CI: 0.80–2.48, P = 0.233). Furthermore, a low HDL-C did not reach statistical significance in the unadjusted Cox model (HR = 1.30, 95% CI: 0.74–2.30, P = 0.367). Thus, the plasma TG/HDL-C ratio seemed more useful as a way to predict CVD than either the plasma TG concentration or HDL-C alone.


The result of our study are straight-forward, and this is also true of its limitations. Focusing first on the findings, the data in Table 2 demonstrate that reasonably comparable numbers of individuals were identified as being at ‘high’ risk to develop CVD, whether classified on the basis of an elevated TG/HDL-C concentration or a diagnosis of the MetS. Furthermore, the cardio-metabolic risk profile of the ‘high’ risk groups, irrespective of how they were identified, was also comparable.

The results in Table 3 show that the HR of developing a CVD event were significantly greater in ‘high’ versus ‘low’ CVD risk groups, defined by either TG/HDL-C or MetS criteria. Although the unadjusted HR was greater in the ‘high’ risk group with the MetS as compared with those with an elevated TG/HDL-C ratio, this difference essentially disappeared when adjusted for differences in age. Furthermore, additional adjustments for sex, diabetes, total cholesterol and cigarettes smoked had little impact, leading to the finding that the HR for incident CVD events were reasonably comparable in the two ‘high’ risk groups.

In light of these data, it seems reasonable to conclude that the ability of an elevated plasma TG/HDL-C concentration ratio to identify individuals at ‘high’ risk for CVD, who go on to have an event, may not be inferior to making a diagnosis of the MetS. On the other hand, the limitations of our study are obvious. At the simplest level, the experimental population was modest in size (~800), and almost entirely of European origin. Furthermore, only 51 CVD events were observed during the follow-up period. Finally, a 14% of the cohort was lost in the follow-up but this level seems reasonable for a population study with 14 years of follow-up and could be attributed to migration. Perhaps of greater importance is that the ability of the MetS criteria to predict CVD events has been well-documented in several large population-based studies [4, 5, 14, 15]. Consequently, it would be premature to imply on the basis of our findings that the TG/HDL-C concentration ratio is able to perform as effectively as the MetS criteria in identifying individuals at increased CVD risk. On the other hand, it is relatively simple to assess the plasma TG/HDL-C concentration ratio. This fact, in the context of the current data, buttressed by prior findings that the plasma concentration ratio of TG to HDL-C independently predicted CVD events [16, 17], suggests that it does not seem premature to suggest that the potential availability of an index for predicting CVD based on a simple measurement of plasma TG and HDL-C concentration is worthy of further evaluation. Moreover, there is published some evidence that the TG/HDL-C ratio has been successfully used in predicting the development of diabetes, coronary heart disease and cardiovascular mortality [18, 19]. Thus, we hope that the current findings might serve as a ‘pilot’ study to stimulate other investigators, with larger data sets, to perform analyses similar to the current one. However, it must be remembered that differences in the actual value of the TG/HDL-C cut-points that predict CVD risk and outcome will vary as a function of race/ethnicity [20] and sex [6]. Finally, for the sake of comparability, it is also hoped that the specific TG/HDL-C cut-points be identified as we previously outlined [6].

Conflict of interest statement

The authors declare no conflict of interest.


This study could not have been conducted without the help of the nurses from the ‘Hospital Municipal de Rauch’.