The present study demonstrated for the first time the usefulness of the TG/LDL-C ratio as a predictive marker for higher sd-LDL in type 2 diabetes patients with hypertriglyceridemia. In statin-free patients, the assumed cut-off point for TG/LDL-C ratio was 1.1, and the sensitivity and specificity of the ratio as a predictive marker for higher sd-LDL surpass those of non–HDL-C or other lipid markers. TG/LDL-C ratio is the first formula proposed and considered suitable for evaluation of sd-LDL and TG-rich lipoproteins. TGs were positively correlated with LDL-MI, and LDL fraction (PAG) and LDL-C were negatively correlated with LDL-MI. The TG/LDL-C ratio more reliably predicts an increase in LDL-MI (sd-LDL). Even if LDL-C is low or within the normal range, it is possible to predict high values of sd-LDL by calculating the relative ratio with TGs. The reciprocal of this formula, LDL-C/TG ratio, was reported by Yoshida et al. and suggested as being related to sd-LDL [18]. In the present study, TG level was used as the numerator for the purpose of emphasizing the existence of TG-rich apoB-containing lipoproteins.
In a study involving 994 non-diabetic patients with TGs of ≤ 4.5 mmol/L (400 mg/dL), the AUC of non–HDL-C was 0.871 (0.840-0.901), the sensitivity was 78.8%, specificity 79.8%, and positive predictive value 54.9% for predicting high sd-LDL (≥ 46 mg/dL via direct assay) [19]. In that study, LDL-C was calculated using Friedewald's formula. A simple comparison between the previous and present studies shows that the TG/LDL-C ratio of the present study is superior in terms of AUC, sensitivity, and specificity versus non–HDL-C of the previous study for the prediction of high sd-LDL.
Specific clinical data collected in routine clinical practice can be combined with other related data to increase clinical usefulness for the diagnosis or estimation of various diseases [20]. The routine lipid panel consists of LDL-C, HDL-C, TGs, and total cholesterol. Several additional parameters, such as non–HDL-C (total cholesterol minus HDL-C), LDL-C/HDL-C ratio [21], non–HDL-C/HDL-C ratio [22], and TG/HDL-C ratio [23], are emerging as valuable adjuncts to the standard panel. Non–HDL-C is used to evaluate apoB-containing lipoproteins and sd-LDL [24]. In particular, increased non–HDL-C concentration is reportedly associated with residual risk for CVD and has been adopted as a guideline for lipid management [7]. In general, when non–HDL-C increases, the cholesterol contained in TG-rich lipoprotein increases, as do total cholesterol concentrations [25]. In patients with hypertension and/or insulin resistance, the metabolism of lipoproteins is delayed, and they remain in the blood circulation for a variety of reasons [26]. Cholesterol-rich and TG-rich apoB-containing remnant lipoproteins are taken up by macrophages, and cholesterol accumulates in atherosclerotic lesions [27, 28].
Non–HDL-C is easily calculated by subtracting HDL-C from total cholesterol. Non–HDL-C can provide a better risk estimation compared with LDL-C, in particular in hypertriglyceridemia combined with diabetes, metabolic syndrome, or chronic kidney disease. This is supported by a recent meta-analysis including 14 statin trials, 7 fibrate trials, and 6 nicotinic acid trials [29]. Non–HDL-C is used as an estimation of the total number of atherogenic particles in plasma (VLDL + intermediate-density lipoprotein + LDL) and relates well to apoB levels. Non–HDL-C may be greatly affected by LDL-C and apoB concentrations [30]. A high correlation exists between the changes in non–HDL-C and TGs [31].
In our previous study, it was found that type 2 diabetes patients with hypertriglyceridemia can be divided into two groups: those with relatively low LDL-C and those with normal or high LDL-C. In the former group, TGs and sd-LDL were also higher than those in the latter group, but there was no difference in non–HDL-C between the two groups (4.9 [4.0-5.4] mmol/L vs. 4.8 [4.5-5.2] mmol/L) [9]. Alternatively, the increase in TGs and decrease in LDL-C (calculated using Friedewald’s formula) could have been synchronized [13, 32], and the change in non–HDL-C might have been offset. The non–HDL-C measurement formula evaluates the cholesterol levels of TGs and LDL, which are rich in apoB. An increase in sd-LDL reflecting hypertriglyceridemia and a decrease in buoyant LDL might occur simultaneously in type 2 diabetes patients. As evidence, this study showed that both the LDL-C and LDL fractions in PAG electrophoresis are negatively correlated with LDL-MI. If so, calculated non–HDL-C, regardless of the LDL-C assay method used, will largely reflect the decline in buoyant LDL present in some type 2 diabetes patients, and it is estimated that the increase in non–HDL-C would not be as expected. The changes in LDL composition would affect the measurement of non–HDL-C and thus impair its clinical reliability, and these changes might also affect the relationship between non–HDL-C and sd-LDL [9]. In contrast, another lipid marker that indirectly compares the changes in sd-LDL and buoyant LDL (i.e., the TG/LDL-C ratio) seems to more accurately reflect the pathophysiology of dyslipidemia in type 2 diabetes patients. A new subset of atherogenic lipoproteins consisting of LDL-C and TGs is proposed, with LDL-C and TGs serving as surrogates for LDL/intermediate-density lipoprotein and VLDL, respectively. The TG/LDL-C ratio reflects LDL and VLDL and can be a new predictive marker for sd-LDL increase. The fact that the cut-off point for the TG/LDL-C ratio was constant irrespective of statin treatment suggests that the TG/LDL-C ratio is universal and reliable for the prediction of sd-LDL increase.
The REDUCE-IT trial showed that treatment with icosapent ethyl significantly reduced CVD events without any change in non–HDL-C in patients with CVD risk and increased baseline TGs but well-controlled LDL-C [31, 33]. The REDUCE-IT trial results may alter the approach to the management of hypertriglyceridemic patients whose lipid phenotype requires more intensive treatment beyond LDL-C lowering alone.
Study strengths and limitations
There are several strengths to this study. First, the reliable LDL-C direct assay, Metabolead LDL-C® (Hitachi Kasei Diagnostic Systems), was used for LDL-C estimation, and the results were consistent with the lipoprotein PAG electrophoresis results [9]. Moreover, this direct method has already been shown to be consistent with ultracentrifugation, unless TGs exceed 11.3 mmol/L [15, 16]. There were no patients with a fasting TG level of ≥ 11.3 mmol/L in this study; therefore, any effect of hypertriglyceridemia on the LDL-C assay could be ruled out [34, 35]. A study observing changes in LDL (measured by high-performance liquid chromatography) before and after pemafibrate reported an increase in LDL, especially large, buoyant LDL, in addition to a decrease in sd-LDL after the treatment [36]. Conversely, it indicated that there was a relative reduction in baseline large, buoyant LDL in some patients with hypertriglyceridemia [13]. This is consistent with the decreases in baseline LDL-C (measured via the direct method) and baseline LDL fraction in PAG electrophoresis. Second, PAG electrophoresis, which is a simple and inexpensive method, was used for the estimation of sd-LDL and lipoprotein fractions.
This study also has several limitations. First, instead of directly measuring sd-LDL, determination of LDL-MI by PAG electrophoresis was used as a substitute. However, many reports have indicated that the results of both are strongly correlated [13, 14]. Second, there were relatively few target patients in this study. Statistically significant results were obtained despite the small number of patients in present study, however. Future studies should enroll a larger number of target patients. Third, the prevalence of CVD in the target patients in present study was low, and it was not possible to evaluate CVD due to quantitative changes in TG/LDL-C ratio or LDL-MI. Fourth, there may be some specificity in the clinical background of the patients included in present study. There was a high proportion of male patients. Patients treated with statins were older and had lower eGFR. It is generally known that Okinawan subjects (who were analyzed in the present study) tend to develop obesity from an early age and that the frequencies of dyslipidemia and metabolic syndrome in these subjects are high [37, 38]. It is therefore unclear whether the results of this study are universally applicable to all Japanese or other ethnic groups. Finally, this study did not show results for serum apolipoproteins such as apoB and apolipoprotein E.