Lipids as risk markers for type 2 diabetes

In the current issue of the Journal, Cheng et al.1 analyzed a cohort of 10 741 Chinese people followed for a median of 6 years, of whom 618 developed type 2 diabetes (T2D), assessing the usefulness of the triglyceride:  high-density lipoprotein cholesterol (TG/HDL-C) ratio as a lipid marker of diabetes risk. A higher TG/HDL-C ratio was associated with older age, female sex, smoking, drinking alcohol, hypertension, height, weight, waist circumference (WC), body mass index (BMI), systolic and diastolic blood pressure (SBP and DBP, respectively), and fasting plasma glucose (FPG). Cholesterol and TG levels increased and HDL-C levels decreased with increasing quartiles of the TG/HDL-C ratio. The incidence of T2D was 9.68 per 1000 person-years, with the incidence of T2D in the first, second, third and fourth quartiles of the TG/HDL-C ratio being 5.01, 6.87, 9.72, and 17.15 per 1000 person-years, respectively. When adjusted for age, sex, education level, physical activity, smoking, alcohol use, BMI, SBP, DBP, FBG, total cholesterol, family history of T2D, and hypertension status, the risk ratio of the fourth versus first quartile of the TG/HDL-C ratio was 2.11. Cheng et al.1 performed a meta-analysis of their and three other studies, comprising 20 302 non-diabetic people and finding that for each 1-unit increment in the TG/HDL-C ratio, the likelihood of diabetes increased 1.28-fold. Another analysis using the same dataset showed that, among people with initially normal FPG followed for 6 years, the product of WC with the TG/HDL-C ratio, of TG with FPG, and of WC with TG levels tracked similarly but more strongly with T2D than the TG/HDL-C ratio, with similar predictive strength to WC and FPG.1 [Correction added on 17 January 2019, after first online publication: reference 1 and it's corresponding citations have been corrected.].

The TG/HDL-C ratio appears to be one of a number of useful lipid measurements indicative of visceral adiposity, and hence of insulin resistance. The TG level correlates strongly with clamp measurements of insulin sensitivity, which, in turn, is associated with visceral adiposity and T2D risk.2 In an analysis of 4398 adults in the 2011 to 2014 US National Health and Nutrition Examination Survey (NHANES), two simple measures of visceral adiposity, namely the supine sagittal abdominal diameter/height ratio and the WC/height ratio, correlated more strongly than BMI with fasting insulin, TGs, and with the TG/HDL-C ratio.3 In another study of 4820 adults, 332 of whom developed T2D over a mean of 8 years, the TG–glucose product, adjusted for age, BMI, cigarette smoking, alcohol intake, lifestyle pattern, hypertension, cardiovascular disease, low-density lipoprotein cholesterol (LDL-C), HDL-C, and sex, was a stronger measure of diabetes risk among those with normal FBG than the FBG or TG levels alone.4 Among 25 925 non-diabetic women aged ≥45 years, a measure based on very low-density lipoprotein, LDL-C, and HDL-C particle sizes, and their subset concentrations (as measured by nuclear magnetic resonance spectroscopy) tracked with incident T2D over a 20.4-year median follow-up period, additive to glucose metabolism parameters, HDL-C, and TGs.5 In another population study,6 plasma concentrations of dihydroceramide 18:0, lysoalkyl-phosphatidylcholine (PC) 22:1, and triacyglycerol 16:0/18:0/18:1 were associated with measures of insulin resistance and predicted future T2D independently of prediabetes with 76% accuracy. In another study of >10 000 people in two populations, the Cer(d18:1/18:0)/Cer(d18:1/16:0) ceramide ratio was predictive of T2D, independent of measures of obesity and glycemia.7 The relationship between the TG/HDL-C ratio and reduced insulin sensitivity may not be the same in all populations; a small study of African American women showed that the TG/HDL-C ratio was a poor predictor of insulin resistance, and was more strongly associated with β-cell function.8

These lipid-related markers may be added to a wide variety of measures related to insulin resistance that have been shown to be useful in predicting the likelihood of diabetes. The commonly used approximation of insulin resistance based on fasting insulin and glucose levels, namely homeostasis model assessment of insulin resistance (HOMA-IR), is a much weaker predictor of risks of T2D, as well as risks of developing cardiovascular disease, than are measures based on both fasting and 2-hour postglucose load glucose and insulin levels.9 In the 3234-person Diabetes Prevention Program, lower levels of adiponectin and a lesser 1-year increase in this peptide were associated with T2D development,10 and another study found adiponectin to add to insulin- and glucose-based measures associated with T2D risk.11 Serum alanine transaminase levels in the upper normal range are associated with a tripling of T2D risk, after accounting for the effect of age, sex, race/ethnicity, family history of diabetes, BMI, FPG, and 2-hour glucose.12 In a decade-long follow-up of septuagenarians in Sweden, among a variety of peptides associated with inflammation, interleukin-1 receptor antagonist and tissue plasminogen activator were associated with risk of T2D, although the relationship lost statistical significance after adjustment for fasting blood glucose.13 Among other plasma metabolites associated with future T2D are higher markers of dysregulated lipid metabolism, impaired metabolism of branched-chain amino acids (BCAAs), abnormal diacyl glycerol accumulation, and lower levels of lyso-PC(19:1), PC(17:0/18:2), 2-hydroxyethanesulfonate (a downstream metabolite of taurine, possibly formed by anaerobic gut bacteria or myeloperoxidase-induced degradation), and N-acetylglycine.14 A similar apparent protective association of PC(22:6/20:4) was reported in a study of 2117 initially normoglycemic American Indians followed for an average of 5.5 years.15 Choline undergoes mitochondrial oxidative metabolism to betaine, cytoplasmic metabolism to dimethylglycine, and then mitochondrial oxidative metabolism to sarcosine in the production of the methyl donor S-adenosylmethionine, and, in a 7.5-year follow-up of 3621 people with suspected angina, of whom 233 developed T2D, lower plasma betaine, but higher urine betaine and sarcosine were associated with T2D, with either plasma betaine or urinary sarcosine additive to established predictors of incident T2D risk.16 Amino acids may be useful as markers of T2D risk,17 perhaps as markers of peripheral insulin resistance, and perhaps related to effects of the intestinal microbiota, with evidence that different amino acids may perform differently as risk markers in different populations.18 The acylcarnitines, products of BCAA metabolism, have a variable relationship to T2D risk, with a 6-year follow-up of 2103 non-diabetic Chinese people showing that among the 507 who developed diabetes, concentrations of C0, C3DC, C4, C5, C5OH, C8:1, C10, C14OH, C14:1OH, C16:1, C16:2, C18, C18OH, C18:1, C18:2, C20, and C20:4 acylcarnitines were higher, whereas levels of 3-dehydroxycarnitine, 3-dehydrocarnitine, and C7DC, C10DC, C12, C12OH, and C12:1 acylcarnitines were lower.19

What are we to make of these associations? Diabetes prediction could be of great importance if more effectively used in prevention; the availability of measures based on fasting glucose, TG, and HDL-C levels makes it imperative that clinicians use existing effective lifestyle and, perhaps, pharmacologic approaches to forestall the growing diabetes epidemic. More intriguing is the possibility that such measures could help clinicians in individualizing treatment. The results of the Insulin Resistance Intervention after Stroke (IRIS) study of nearly 4000 non-diabetic patients with insulin resistance who had recent ischemic stroke or transient ischemic attack and for whom randomization to pioglitazone lowered the risk of subsequent stroke and myocardial infarction, as well as diabetes,20 are suggestive of the potential importance of such an approach.

在本期杂志中，Cheng等¹在10741名中国人群中进行了队列分析，中位数随访时间为6年，其中有618名发生了2型糖尿病(T2D)，评估了将甘油三酯:高密度脂蛋白胆固醇(TG/HDL-C)比值作为糖尿病风险脂质标志物的实用性。更高的TG/HDL-C比值与年龄更大、女性性别、吸烟、饮酒、高血压、身高、体重、腰围(WC)、体重指数(BMI)、收缩压(SBP)和舒张压(DBP)以及空腹血糖(FPG)之间都有相关性。随着TG/HDL-C比值四分位数的增加，胆固醇与TG水平随之上升，HDL-C水平随之下降。每1000名患者每年的T2D发病率为9.68名，在TG/HDL-C比值的第1、第2、第3与第4四分位数中，每1000名患者每年的T2D发病率分别为5.01、6.87、9.72与17.15名。校正年龄、性别、教育水平、体力活动、吸烟、饮酒、BMI、SBP、DBP、FBG、总胆固醇、T2D家族史以及高血压状态后，TG/HDL-C比值的第4与第1四分位数相比, T2D的危险比(risk ratio)为2.11。Cheng等¹对他们的研究以及其他3项研究进行了meta分析，共纳入20302名非糖尿病患者，发现TG/HDL-C比值每增加1个单位，发生糖尿病的可能性增加1.28倍。使用相同数据集的另外一项分析表明，最初FPG正常人群在为期6年的随访期间，与TG/HDL-C比值相比，WC与TG/HDL-C比值的乘积、TG与FPG的乘积以及WC与TG水平的乘积与T2D的相关性更强，预测能力与WC以及FPG相似¹。[首次在线出版后，于2019年1月17日进行了修改, 更正了参考文献1和与其相关的引用]。

TG/HDL-C比值似乎是众多可以预测内脏肥胖以及由此导致的胰岛素抵抗的脂质测量指标之一。TG水平与钳夹测量的胰岛素敏感性强烈相关，因此，也与内脏肥胖及T2D风险相关²。有一项研究分析了4398名来自2011-2014年美国国家健康与营养调查((NHANES)的成年人，测量了两个简单的内脏肥胖相关指标，即仰卧位腹部矢状面直径/身高比值与WC/身高比值，与BMI相比，发现它们与空腹胰岛素、TGs以及TG/HDL-C比值的相关性更强³。另外一项研究纳入了4820名成年人，在平均时间超过8年的随访期间，有332名发生了T2D，在最初FBG正常的人群中，校正年龄、BMI、吸烟、饮酒、生活方式、高血压、心血管疾病、LDL-C、HDL-C以及性别后，发现与单纯的FBG或者TG水平相比，TG-血糖乘积是更强的糖尿病风险预测指标⁴。在25925名年龄≥ 45岁的非糖尿病妇女中进行了中位时间为20.4年的随访，发现一个基于极低密度脂蛋白、LDL-C和HDL-C的颗粒大小以及它们的亚型浓度(使用磁共振光谱分析法测定)的测量指标与T2D发病风险、血糖代谢参数、HDL-C及TGs都密切相关⁵。在另一项人群研究中⁶，血浆中的二氢神经酰胺18:0、溶菌酶磷酸卵磷酯22:1和三酰基甘油16:0/18:0/18:1都与胰岛素抵抗相关，可以预测未来是否发生T2D而且不依赖于是否存在糖尿病前期，并且精确度达到了76%。另外一项研究在两个人群中纳入了> 10000名受试者，发现Cer(d18:1/18:0)/Cer(d18:1/16:0)神经酰胺比值可以预测T2D，并且不依赖于肥胖与血糖测值⁷。在所有人群中TG/HDL-C比值与胰岛素敏感性下降之间的关系可能不尽相同；一项在非裔美国妇女中进行的小型研究结果显示，TG/HDL-C比值预测胰岛素抵抗的效能较差，它与β-细胞功能的相关性更强⁸。

已经被证实可用来预测糖尿病发生的这些与脂质相关的标记物可以被加入到与胰岛素抵抗相关的更广范围的测量值中。常用的评估胰岛素抵抗方法是根据空腹胰岛素与血糖水平，亦即使用稳态模型评估的胰岛素抵抗(HOMA-IR)，这与根据空腹以及葡萄糖负荷后2小时的血糖以及胰岛素水平的综合评估方法相比，对于预测T2D风险和心血管疾病发生风险来说是较弱的预测指标⁹。在一项纳入了3234名受试者的糖尿病预防项目中，较低水平的脂联素及其在1年内的增加量较小与发生T2D相关¹⁰，并且另外一项研究也发现，脂联素加上基于胰岛素与血糖的测量值与T2D风险相关¹¹。校正年龄、性别、种族/人种、糖尿病家族史、BMI、FPG以及2小时血糖影响后，血清丙氨酸转氨酶水平处于正常范围上限与3倍的T2D风险相关¹²。一项在瑞典进行的为期10年针对七旬老人的随访研究发现，在炎症相关的多种肽中，白介素-1受体拮抗剂以及组织型纤溶酶原激活剂与T2D风险相关，尽管校正空腹血糖后这种相关性失去了统计学意义¹³。与未来发生T2D相关的其他血浆代谢产物包括脂质代谢失调、支链氨基酸代谢受损、甘油二酯异常蓄积的高水平标志物，以及较低水平的lyso-PC(19:1)、PC(17:0/18:2)、2-羟基乙磺酸盐(一种牛磺酸的下游代谢产物，可能是由肠道厌氧菌或者髓过氧化物酶诱导后降解形成的)与N-乙酰甘氨酸¹⁴。在一项纳入了2117名最初血糖正常的美国印第安人研究中，平均随访时间为5.5年¹⁵，也报告了PC(22:6/20:4)具有类似的明显保护作用。胆碱经过线粒体氧化代谢后成为三甲基甘氨酸，再经过细胞质代谢后成为二甲基甘氨酸，接着被线粒体氧化代谢为肌氨酸并生成甲基供体S-腺苷甲硫氨酸，一项研究对3621名疑似心绞痛的患者随访了7.5年，其中有233名发生了T2D，发现血浆三甲基甘氨酸浓度更低，但尿三甲基甘氨酸以及肌氨酸浓度更高与T2D具有相关性，并且血浆三甲基甘氨酸或者尿肌氨酸可以明确预测T2D的发病风险¹⁶。氨基酸也可以作为T2D风险的标志物¹⁷，可能作为外周胰岛素抵抗的标志物，这可能与肠道微生物群的影响有关，有证据表明作为风险标志物的氨基酸在不同人群中的表现也各不相同¹⁸。酰基肉毒碱是BCAA的代谢产物，它与T2D风险的关系具有变异性，在一项纳入了2103名非糖尿病的中国人群研究中，经过6年的随访后发现，在507名发生糖尿病的患者中，C0、C3DC、C4、C5、C5OH、C8:1、C10、C14OH、C14:1OH、C16:1、C16:2、C18、C18OH、C18:1、C18:2、C20以及C20:4酰基肉毒碱的浓度都更高，然而3-脱氢肉碱、3-脱氢卡尼汀以及C7DC、C10DC、C12、C12OH和C12:1酰基肉毒碱的水平却更低¹⁹。

我们该要如何看待这些关系?如果糖尿病预测能够促使更有效的预防将具有重要的意义；现有的基于空腹血糖、TG以及HDL-C水平的测量，使得临床医生为了预防糖尿病的流行，必须推行有效的生活方式，或许还包括药物治疗。更吸引人的是，这些措施可能有助于临床医生对患者进行个性化治疗。中风后胰岛素抵抗的干预研究纳入了近4000名存在胰岛素抵抗的非糖尿病患者，他们最近都有过缺血性中风或者短暂性脑缺血发作史，研究结果发现随机分配到吡格列酮治疗组的患者随后发生中风或者心肌梗死的风险降低了，发生糖尿病的风险也降低了²⁰，这提示了这类治疗方法的潜在重要性。