The atherogenic dyslipidemia ratio [log(TG)/HDL-C] is associated with residual vascular risk, beta-cell function loss and microangiopathy in type 2 diabetes females

Background Atherogenic dyslipidemia (AD), defined as low HDL-C plus elevated triglycerides (TG), comorbid to T2DM, increases cardiometabolic risk for CAD even when LDL-C is at target. In T2DM males, AD was shown to correlate with β-cell function loss, yet it is not established whether this applies across gender. Aim To establish the prevalence and severity of AD in T2DM females, and to determine how it relates to cardiometabolic phenotype, glucose homeostasis, micro- and macrovascular complications, and 10-year absolute CV risk (UKPDS Risk Engine). Methods 340 T2DM females were ranked according to quintiles (Q) of the continuous variable log(TG)/HDL-C, with AD prevalence defined as HDL-C <50 mg.dL-1 plus TG ≥150 mg.dL-1, and β-cell function assessed with HOMA. Results AD prevalence was 35%; mean HDL-C and TG were 52 (15) and 160 (105) mg.dL-1. AD was significantly related to central fat, metabolic syndrome, sedentarity and skeletal sarcopenia, as well as to hsCRP, fibrinogen, uric acid, cystatin-C, Big ET-1, and 10-year UKPDS CV risk. AD correlated stepwise with lower β-cell function and hyperbolic product, and with accelerated loss of residual insulin secretion, higher HbA1c and prevalent microangiopathy. Conclusions log(TG)/HDL-C is a simple means to grade AD and residual macrovascular risk in T2DM females. This ratio associates with major non-LDL cardiometabolic variables and ranks predicted CAD risk. In addition, log(TG)/HDL-C identifies worsening glucose homeostasis, poorer glycemic control, and prevalent microangiopathy.


Introduction
Current guidelines recommend intensive lowering of lowdensity lipoprotein cholesterol (LDL-C) in type 2 diabetes mellitus (T2DM) patients [1][2][3][4][5]. The common form of T2DM, associated with insulin resistance (IR) and the metabolic syndrome (MetS) is characterized by a singular, non-LDL dyslipidemia, defined as atherogenic dyslipidemia (AD). The hallmark of AD is decreased levels of high-density lipoprotein cholesterol (HDL-C) together with raised triglycerides (TG). AD substantially contributes to residual vascular risk (RVR), even when LDL-C is at or below targets, both in diabetic and nondiabetic populations. The presence and/or severity of AD can be established either from the combined occurrence of high TG levels and low HDL-C or from the ratio of fasting TG to fasting HDL-C, with prior log transformation of TG levels allowing to compute a broad range of TG values [2][3][4][5][6][7][8][9][10][11]. Whereas screening for AD provides clinically relevant information for assessing RVR, it is rarely performed due to lack of agreement or consensual cut-offs to define/grade AD from baseline HDL-C and TG levels prior to lipidlowering therapies (LLD) . We previously reported that log(TG)/HDL-C is a useful measure of increased cardiometabolic risk in T2DM males, in whom the ratio also relates to poorer metabolic control and β-cell function loss [7]. In diabetic females, there exists a reversed dichotomy in risk factors for CAD vis-à-vis nondiabetic women, the former no longer benefiting from female gender's risk protection. Thus, diabetic women have more severe dyslipidemia, more prevalent obesity and abdominal fat, higher subclinical inflammatory markers, more severe/prevalent hypertension or MetS. Worryingly, diabetic women are less often at targets for major modifiable micro-and macrovascular risk factors, leaving them exposed to higher RVR than males [12][13][14][15]. The aim of the present study was (i) to establish and grade AD prevalence in a cohort of T2DM females using the continuous variable log(TG)/HDL-C; (ii) to determine whether this ratio relates to their cardiometabolic phenotype or glucose homeostasis determinants; and (iii) to assess the association between AD and micro-or macrovascular complications, and its impact on absolute 10-year CV risk from the United Kingdom Prospective Diabetes Study (UKPDS) calculator [16].

Methods
This cross-sectional study included 340 female T2DM outpatients, attending the Diabetes Clinic between January 2008 and June 2011. Patients were divided according to quintiles (Q) of baseline, pre-LLD log(TG)/HDL-C values. The following variables were recorded: age, ethnicity, attained educational level (dichotomised as low vs. high, based on higher achieved education degree [no education, primary or secondary vs. higher education and university]), known diabetes duration, family history (CVD, diabetes mellitus), self-reported leisure-time (LT) weekly exercise duration, and daily-time spent watching screen(s) (television, computers and/or visual numeric media), as surrogate for LT sedentarity [7]. The presence of a MetS (score ≥3/5) was defined according to the joint harmonized IDF/NHLBI/AHA/WHF/IAS/IASO criteria [17,18].
Pre-LLD values were retrieved from patients' records, and used to assess AD prevalence in LLD-treated patients; otherwise, in non-LLD-treated patients, current lipids and AD prevalence were obtained from last available lipid values from the same specimen. In LLDtreated patients, current lipids values were those from last available, post-LLD lipid values. AD prevalence was defined as the combination of MetS criteria for low HDL-C in females (<50 mg.dL -1 ) and high fasting TG (≥150 mg.dL -1 for both genders) at baseline. AD severity was assessed from log(fasting TG)/fasting HDL-C.
Current use of oral antidiabetic drugs, insulin, blood pressure (BP)-lowering drugs, aspirin, and LLD(s) was recorded. Hypertension was defined as systolic BP ≥140 mmHg and/or diastolic BP ≥90 mmHg, or current treatment with BP-lowering drug(s). Coronary artery disease (CAD) was inferred from medical history (myocardial infarction, angioplasty, stenting, revascularization surgery and/or significant coronary stenosis confirmed by angiography) and systematic review of procedures, screening (exercise testing, echocardiography) or subclinical disease imaging data in the patient's records. Stroke was defined according to UK Prospective Diabetes Study (UKPDS) criteria: any neurological deficit ≥1 month, without distinction between ischaemic, embolic and haemorrhagic strokes. In patients with multiple CV events, only the first one was considered for prevalence [16]. Peripheral artery disease (PAD) was defined by a medical history of lower-limb(s) claudication and/or clinical or imaging evidence for ischemic diabetic foot, angioplasty, stenting, revascularization surgery and/or significant lower-limb artery stenosis at Doppler ultrasonography and/or angiography. UKPDS Risk Engine's 10-year absolute CV risk for individuals in primary CV prevention was based on the 10 following variables: known T2DM duration, age, gender, ethnicity, smoking status, chronic atrial fibrillation, HbA 1c level, systolic BP, total C and HDL-C [16].

Statistical methods
Results are presented as means (± 1 standard deviation (SD)) or proportions. The significance of differences between AD quintiles was assessed by one-way analysis of variance for linear trend between means, and by a Chisquared test for trend for differences in proportions with modified Bonferroni's adjustment procedure to minimize type 1 error resulting from multiple testing. Results were considered significant or non-significant (NS) for p< or ≥0.05, respectively.

Results
Patient's characteristics are described in Table 1. There were 340 T2DM females in the whole cohort, 85% of whom from Caucasian ancestry. AD prevalence was 35%. When patients were divided according to AD quintiles of log(TG)/HDL-C, patients in the 2 nd quintile (Q II) had a mean log(TG)/HDL-C similar to that of control, nondiabetic subjects (see Methods), whereas patients from the 3 rd quintile (Q III) had a mean log (TG)/HDL-C that corresponded to the 75 th centile of controls. In the whole T2DM cohort, mean age (1 SD) and diabetes duration were 67 (13) years and 15 (10) years, respectively, without significant trends across quintiles (Qs). Past-or current smoking prevalence was not different between Qs, whereas ethanol intake showed a significant decreasing trend (p <0.0001). LT physical activity and screen-watching daily duration were significantly different across Qs, with increased sedentarity and lesser exercise. There was a significant stepwise decrease in the proportion of patients with higher educational level across quintiles, from 44% (QI) to 18% (QV), which nevertheless was above Bonferroni's adjusted level of significance. Hypertension prevalence was 85% in the whole cohort, with increasing values across Qs, and mean systolo-diastolic BP values 139 (21) / 78 (11) mmHg, without trends across Qs (Table 1).
Current glycemic control, inferred from the average of last-available HbA 1c , was above target (>7.0%) in the whole group, and significantly worsened across Qs, with a mean absolute differences of 0.84% between Q I/Q II and Q V. The proportion of patients at HbA 1c target (33% in the whole cohort) also significantly decreased across Qs, from 47% (Q I) to 21% (Q V; p <0.0001). Prevalence of any microangiopathy (51% in the whole cohort) significant increased across Qs, from 40% (Q I) to 60% (Q V; p 0.0045). The increase in microvascular complications prevalence across Qs remained significant after adjustment for HbA 1c (p 0.0261 [overall microangiopathy]; p 0.0114 [peripheral neuropathy]). When patients were analyzed according to Qs of non-HDL-C, there were no associations between the latter and the frequency of microangiopathy, overall or site-specific. As regards macroangiopathy, there was no trend for higher prevalence across Qs of log(TG)/HDL-C: (overall prevalence (22%); CAD (15%); TIA/stroke (6%) and PAD (5%). Epworth's daytime sleepiness score and sleep apnoea syndrome prevalence significantly increased across Qs, from 5 (4) and 0% (Q I) to 7 (4) and 12% (Q V; p 0.0066 and 0.0011, respectively; Table 1).
[BxS] loss rate significantly worsened across Qs: 1.17 %.yr -1 (Q I) to 1.52 %.yr -1 (Q V). There was a significant increase across Qs in fasting insulinemia: 77 (Q I) to 141 pmol.l -1 (Q V; p<0.0001). Liver steatosis frequency (64% prevalent in the whole cohort) significantly rose across Qs: 27% (Q I) to 83% (Q V; p<0.0001). A MetS phenotype was present in 86% of the entire cohort; there was a highly-significant stepwise rise in both MetS prevalence and score across Qs: 59% and 2.6 (Q I) to 100% and 4.7 (Q V; both p <0.0001). The contribution of discrete components to the MetS score for each Q showed significant trends for higher prevalence of hypertension (p 0.0181) and enlarged waist circumference (p <0.0001; Table 2).
As regards glucose-lowering therapies, 58% of patients were treated with metformin; 41% with a β-cell stimulant; and/or 50% with exogenous insulin. There was a significant trend for higher insulin use or higher dosage across quintiles: 35% and 0.42 UI.day -1 .kg -1 (Q I) to 58% and 1.13 UI.day -1 .kg -1 (Q V; p 0.0058 and <0.0001, respectively). LLD(s) were given to 63% of the whole cohort, with statin; fenofibrate and/or ezetimibe prescribed to 51%; 19% and/or 3%. There was a significant trend for increased fenofibrate use across Qs: 9% (Q I) to 39% (Q V; p <0.0001). None of the patients were treated with niacin.

Discussion
The main findings of the present study are fourfold. Firstly, we observed a high prevalence of AD in T2DM females, with 35% affected, similar to that observed in T2DM males. Secondly, the AD ratio log(TG)/HDL-C is an easy means to estimate IR, and was strongly related to a cardiometabolic phenotype characterized by IR, MetS, hyperinsulinemia, central fat accretion, sarcopenia, liver steatosis and sedentary lifestyle. Thirdly, this ratio was linked to a stepwise gradient for residual risk of future CAD. Lastly, this ratio was related to decreased insulin secretion, β-cell function loss, poorer glycemic control and higher frequency of microangiopathies. The AD ratio may therefore allow to easily identifying a comprehensive modifiable component of non-LDL, lipid- related residual vascular risk, as well as pinpoint worsening in glucose homeostasis determinants [5][6][7]. The hallmark of AD is decreased HDL-C level together with raised TG, with LDL-C often only marginally elevated [2][3][4][5][33][34][35][36][37][38][39][40]. Computing a ratio from fasting TG (numerator) and fasting HDL-C (denominator) allows not only for grading AD severity as a continuous variable, but also for incorporating the effect of mutually-reinforcing or diverging confounders affecting both fraction's components and that of one variable on the variance of the other [7]. An analysis of isolated TG or HDL-C, both indispensable determinants to define AD, as standard continuous variables would not be able to incorporate a measure of AD. Using consensual MetS cutoffs for defining decreased HDL-C and elevated TG, we observed a high prevalence of AD in T2DM females (more than one-third). In comparison with log(TG)/ HDL-C values from non-diabetic female controls, the mean AD ratio in T2DM females was well above normal values, from the fourth AD quintile onwards. This demonstrates that >40% of this population may present with AD of varying severity. This finding was expected, as raised TG and low HDL-C levels are used to define MetS phenotype, and as the latter is highly prevalent (86%) in T2DM [17,18].
To rule out any confounding effect of LLD, we used baseline TG and HDL-C to compute the AD ratio, prior to anti-dyslipidemic drug(s). This proved appropriate, as the mean difference between baseline and current TG levels in this study averaged 46 mg.dL -1 . This approach allowed thus to precisely establishing the true, underlying magnitude and frequency of AD in T2DM females [5,7]. There is a known gender difference in HDL-C and TG levels in the general population, also observed in this diabetic sample, with T2DM females exhibiting higher mean HDL-C and lower TG values than their T2DM males, by an average 6 and 12 mg/dl, respectively (data not shown) [12][13][14]. As for lifestyle-related confounders of AD prevalence/severity, our data showed a significant decreasing trend across AD quintiles for alcohol intake, the inverse association being expected due to ethanol's HDL-C-raising effects. On the other hand, we observed a trend for lower educational level. This may promote less healthy behaviours, and result in adverse (i.e. less cardioprotective) anthropometrics and body composition.
The AD ratio was associated, in a stepwise gradation, with higher levels of numerous non-lipid cardiometabolic markers: fasting insulinemia; hs CRP; fibrinogen; Big ET-1; uric acid, and cystatin C. Compensatory hyperinsulinemia is likely to represent a major underlying driver for both AD and for the proatherogenic changes observed in non-LDL lipids and non-lipid CV markers. Endothelin-1 upregulation and/or nitric oxide release imbalance by endothelial cells is another putative pathway linking IR to macroangiopathy. Thus, insulin stimulates endothelin-1 production and action through MAP-kinase-dependent pathways, while ET-1 induces IR and reactive hyperinsulinemia. Such a feed-back loop may contribute to the pathogenesis of macrovascular disease in cardiometabolic states [32]. Expectedly, SHBG levels significantly decreased across quintiles, since IR/ portal hyperinsulinemia are associated with decreased liver production and lower levels of this androgen transporter [4,5,24,25,32,39].
Accumulating evidence links AD in T2DM not only to residual macrovascular risk, but also to residual risk for new-onset or progression of microangiopathy [11]. In this study, log(TG)/HDL-C was associated with lesser residual insulin secretion, as well as with more severe hyperbolic product (BxS) loss over time, and heralded poorer glycemic control and earlier or more intensive need for stepping-up glucose-lowering therapies, as we previously reported in T2DM males [7,22,23,29,41,42]. The AD ratio was associated with predicted CAD risk and prevalent overall microangiopathy. When patients were analyzed according to quitiles of non-HDL-C, another variable which rises alongside the severity of AD, we found no association between the latter and the frequency of microangiopathy. This, together with the independence from diabetes control comforts the relevance of resorting to the log(TG)/HDL-C ratio to estimate a modifiable lipid-related component of residual vascular risk linked to AD. Noteworthy, treatment with fenofibrate, a PPAR-α agonist that specifically targets AD, exerts beneficial effects on macrovascular disease in T2DM patients with AD, but also confers vascular protection against retinal, renal and lower-limb microangiopathies [3,4,[9][10][11].
The present study has several limitations. Firstly, as a cross-sectional hypothesis-generating study, it does not allow for inferring causality relationships. Secondly, as we used routine measurement of HDL-C, without HDL subclass assessment, we cannot discriminate between atheroprotective vs. dysfunctional, less atheroprotective or even atherogenic particles. As the latter often concur with decreased HDL-C, a low HDL-C level in the ratio may nevertheless capture to some extent these qualitative defects. Finally, the population under study was mostly of European Caucasian ancestry, and the present findings require replication in other racial/ethnic subgroups. Moreover, the observed associations were documented from females patients with established T2DM, and may not necessarily exist in newly-diagnosed T2DM females, nor in other types of diabetes or in women with normal glucose homeostasis [43][44][45].
In conclusion, log(TG)/HDL-C allows for grading AD and estimating non-LDL-related residual vascular risk in T2DM females. This AD ratio negatively associates with the two major determinants of glucose homeostasis. This means that patients with more severe AD will have poorer metabolic control in addition to higher predicted CAD risk. The ratio log(TG)/HDL-C appears a simple, cheap noninvasive means to better precise genderspecific assessment and care of the female patient with diabetes. Identifying patients with abnormal AD ratio should allow improving selection and follow-up of those most at risk, more likely to benefit from available or newer therapies targeting hypertriglyceridemia and/or hypo-HDL-cholesterolemia.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions MPH was involved in the design and conduct of the study, as well as in collecting patients' data; MPH, SAA and MFR were involved in analysis and interpretation; preparation and review of the manuscript, and decision to submit the manuscript for publication. All authors read and approved the final manuscript.