Discordance between HbA1c and glycemia

In the present issue of the Journal, Rathmann et al.1 analyzed data from 64 553 people with type 2 diabetes mellitus (T2DM) from various general practice clinics in Germany. Rathmann et al.1 investigated relationships between patient characteristics and the highly correlated cut-off points of HbA1c (less than or more than 7%) and fasting blood glucose (FBG; less than or more than 140 mg/dL). Interestingly, they report significant discordance between these two values in 23% of patients. Each discordant category (HbA1c ≥7%, FBG <140 mg/dL, and FBG ≥ 140 mg/dL with HbA1c <7%), was associated with a distinctive set of characteristics related to the duration of diabetes, its treatment, and/or complications.

Since then, the test has become a “gold standard” for assessing glycemic control.6 A good correlation and association of increased HbA1c with diabetes-related complications was demonstrated in type 1 diabetes mellitus (T1DM) in the Diabetes Control and Complication Trial (DCCT),7 as well as in T2DM in the UK Prospective Diabetes Study (UKPDS).8 In 2010, the use of HbA1c was proposed for the diagnosis of prediabetes and diabetes, given its clinical convenience (lack of requirement for fasting), preanalytic stability, and low medium-term temporal variability.9 However, about the same time, concerns were raised about discordant results of HbA1c and blood glucose values.10 During past decade, we and others have addressed this topic.11, 12 The use of HbA1c in diagnosing diabetes may not be as straightforward as it would seem.13, 14

That there may be wide fluctuations in HbA1c values, not related to glycemic status, in patients with diabetes15 and non-diabetic individuals16 was known in the 1990s. Further, in patients with T1DM, 60% of patients showed an inverse relationship between glycation index (the ratio of HbA1c to the rolling mean blood glucose) and mean blood glucose levels, suggesting a non-linear relationship between mean blood glucose and HbA1c.17 The authors of that study further suggested that differences between individuals’ glycation indices (high vs low glycators) may be due to some intracellular factors that may be more, or less, favorable to the glycation process.17 Importantly, twin studies have shown that HbA1c has a heritable component of variability, which may preclude a linear relationship between blood glucose and HbA1c.18 There may be other physiological reasons for discordance as well, and it is of note that plasma glucose over the preceding 30 days contributes approximately 50% of the variance in HbA1c, whereas that from 90 to 120 days only contributes approximately 10%.19 An analysis of DCCT data with 26 056 HbA1c values and corresponding seven-point glucose profiles from 1439 subjects showed that use of fasting plasma glucose alone progressively tended to underestimate HbA1c.20

More often discussed and better understood causes of discordance between HbA1c and blood glucose levels are the use of non-standard methods of HbA1c estimation (assay methods should be certified by the National Glycohemoglobin Standardization Program as traceable to the DCCT), instrument errors, and conditions that can alter the erythrocyte life span, such as chronic renal failure. Anemia (not due to variant hemoglobin) is a particularly relevant as a cause of discordant HbA1c because many developing countries with a high number of patients with diabetes (eg China, South Asia21) have widely prevalent anemia, particularly in women. Data from China Health and Nutrition Survey (n = 7308 adults) showed that discordant results between HbA1c and FBG were seen in 35% of individuals, and that fewer participants with iron deficiency alone versus normal iron/hemoglobin were classified as having diabetes using HbA1c alone.22 The authors suggested that concurrent measurements of iron, hemoglobin, and HbA1c is critical for correct interpretation of glycemia status in populations with widely prevalent iron deficiency or anemia. Interestingly, these considerations may also apply to the approximate 7% of the US population with anemia.23

There could be other, less-discussed factors that can cause discordant values. For undefined reasons, some populations show only moderate agreement of HbA1c with fasting and 2-h post load plasma glucose values. Further, for diagnosis of prediabetes, HbA1c was reported to be discordant with blood glucose values in Asian Indians in India,24 and in Hispanics and non-Hispanic Whites in the southern US.25 Further, in obese subjects, poor correlation of HbA1c and glycemia estimated by an oral glucose tolerance test was seen,26 similar to the stronger correlation of obesity with solitary high (vs HbA1c) in the study of Rathmann et al.1 in the present issue of the Journal.

Overall, it appears that discordance between HbA1c and blood glucose values may be driven by several factors, and reliance on HbA1c for diagnosis of prediabetes and diabetes may sometime be erroneous. In such cases, combining HbA1c with oral glucose tolerance testing (for diagnosis) and with multiple blood glucose measurement (pre and post prandial) for monitoring should add more definitive value. Continuous blood glucose monitoring clearly adds data on blood glucose variations, in addition to giving information “beyond HbA1c” for management of patients with diabetes prone to wide fluctuations of blood glucose.27 Other markers like 1, 5-anhydroglucitol, used in the assessment of short-term glycemia, need more research. For balanced decisions for treatment of diabetes, a clinician should judiciously and complementarily use HbA1c, oral glucose tolerance testing, self-monitoring of blood glucose multiple measurements and also continuous glucose monitoring as required. Finally, for research involving large multiracial samples of persons with diabetes, HbA1c data should be interpreted with caution.28

在本期杂志上,  Rathmann等¹对64553名来自德国各地全科诊所的2型糖尿病(T2DM)患者数据进行了分析。Rathmann等¹调查了患者特征与高度相关的HbA1c(小于或者大于7%)以及空腹血糖(FBG;小于或者大于140 mg/dL)切点之间的关系。有趣的是,  他们报告的这两个值在23%的患者中存在显著的不一致性。每个不一致(HbA1c ≥ 7%同时FBG < 140 mg/dL,  以及FBG ≥140 mg/dL同时HbA1c < 7%)的分类都与一组独特的特征相关,  包括糖尿病病程、糖尿病治疗和/或并发症。

从那时起,  这种检验方法就已经成为评估血糖控制情况的“金标准”⁶。在1型糖尿病(T1DM)以及T2DM患者中,  糖尿病控制与并发症试验(Diabetes Control and Complication Trial,  DCCT)⁷以及英国前瞻性糖尿病研究(UK Prospective Diabetes Study,  UKPDS)⁸已经分别证实了HbA1c升高与糖尿病相关并发症之间具有良好的相关性以及关联性。鉴于临床上检测HbA1c的方便(不需要空腹)、分析前很稳定并且中期时间变异性也较低,  因此在2010年建议可以将HbA1c用于糖尿病前期与糖尿病的诊断⁹。然而,  大约在同时就有人提出了对HbA1c与血糖值不一致现象的担忧¹⁰。在过去十年中,  我们以及其他人都已经讨论过了这个问题^11,  12。将HbA1c用于糖尿病的诊断可能并不是看起来那么简单^13,  14。

在20世纪90年代就已经认识到,  糖尿病患者¹⁵以及非糖尿病个体¹⁶的HbA1c值可能存在较大的波动,  并且可能与血糖水平无关。此外,  在T1DM患者中,  发现有60%的患者糖基化指数(HbA1c与波动的平均血糖比值)与平均血糖水平之间呈负相关,  提示平均血糖与HbA1c之间的关系是非线性的¹⁷。该研究的作者进一步指出,  个体糖基化指数(高糖基化指数与低糖基化指数之比)之间的差异可能是由于一些细胞内因子或多或少地参与了糖基化过程所导致的¹⁷。重要的是,  孪生子研究表明,  HbA1c具有变异性的遗传成分,  而这可能就是阻碍血糖与HbA1c之间呈线性关系的原因¹⁸。此外,  不一致现象还可能是由其他生理原因所导致的,  值得注意的是,  测定前30天的血糖对HbA1c变异的贡献度大约占50%,  而测定前90-120天的血糖贡献度大约只占10%¹⁹。有一项研究对来自于DCCT的1439名受试者数据进行了分析,  总共纳入了26056个HbA1c值以及相对应的七点血糖谱,  结果发现如果单独使用空腹血糖值会逐步趋于低估HbA1c²⁰。

针对HbA1c与血糖水平不一致的现象,  我们更常讨论以及能够被更好理解的原因是使用了非标准的HbA1c评估方法(采用的检测方法应该得到国家糖化血红蛋白标准化项目组的认证,  而DCCT是具有这项认证)、仪器误差以及遇到了红细胞寿命发生变化的情况,  如慢性肾功能衰竭。贫血(非血红蛋白变异所导致)是一个导致HbA1c出现不一致现象的特别相关的因素,  因为糖尿病患者人数众多的许多发展中国家(例如中国、南亚²¹)普遍存在贫血,  特别是在妇女中。来自中国健康与营养调查的数据(n = 7308名成年人)表明,  35%的参与者会出现HbA1c与FBG结果不一致的现象,  并且如果单独使用HbA1c来诊断糖尿病,  发现单纯缺铁的参与者与铁/血红蛋白正常的参与者相比,  前者被诊断为糖尿病的比例更少²²。作者建议在普遍存在缺铁或者贫血的人群中同时测定铁、血红蛋白以及HbA1c,  这对于正确解释血糖状况来说至关重要。有趣的是,  这些措施可能也适用于大约7%的美国贫血患者²³。

可能还有其他的、较少讨论到的因素会导致测量值不一致的出现。由于一些不确定的原因,  在一些人群中会发现HbA1c与空腹以及餐后2小时血糖值只有一般程度的一致性。此外,  在诊断糖尿病前期时,  有不同研究报告在印度的亚裔印度人中²⁴和在美国南部的西班牙裔以及非西班牙裔白人中²⁵出现了HbA1c与血糖值不一致的现象。此外,  在肥胖的受试者中,  发现通过口服葡萄糖耐量试验评估的HbA1c与血糖之间的相关性较差²⁶,  同样在本期杂志Rathman等¹的研究中也能看到肥胖与单纯高血糖的相关性更强(与HbA1c相比)。

总的来说,  HbA1c与血糖值之间的不一致现象可能是由于若干因素所导致的,  并且依赖HbA1c来诊断糖尿病前期与糖尿病有时候可能是错误的。在这种情况下,  联合测量HbA1c与口服葡萄糖耐量试验用于诊断,  以及测量多点血糖值(餐前与餐后)用于监测,  这样才有更加明确的价值。对血糖容易出现大幅度波动的糖尿病患者的管理来说,  使用动态血糖监测除了提供“超出HbA1c”的信息之外,  还可以明确得出血糖变化的数据²⁷。其他标志物如1,   5-脱水葡萄糖醇,  常被用于评估短期内的血糖变化,  还需要更多的研究。为了平衡糖尿病治疗方案的决策,  作为一名临床医生应该明智地并且互补性地使用HbA1c、口服葡萄糖耐量试验、自我监测的多点血糖测量值,  并且根据需求还可以使用动态血糖监测。最后,  在涉及到多种族糖尿病患者的大型样本研究中,  应该谨慎解释HbA1c数据²⁸。