In a prospective study of a general Japanese population, we clearly demonstrated that elevated HbA1c levels, even in the prediabetic range, were a significant risk factor for the development of CVD, especially for CHD and ischaemic stroke, in a general Japanese population. These associations remained robust even after controlling for comprehensive confounding factors. Furthermore, adding the HbA1c value to the confounding factors significantly improved the predictive ability for total CVD. These findings are important in that they indicate HbA1c’s value for predicting the long-term risk of CVD in Japanese.
Some clinical and population-based cohort studies have shown that increased HbA1c levels were positively associated with the risks of CVDs and mortality [6–11, 20–25]. In the present study, the risks of CHD and ischaemic stroke were significantly or marginally higher even in the subjects with HbA1c levels of ≥5.5% compared with those with HbA1c levels of ≤5.0%. There has been controversy over whether or not the prediabetic levels of HbA1c are associated with CVD risk. Prior cohort studies in Caucasian populations have shown positive associations between the prediabetic range of HbA1c levels and the incidence of total CVD or CHD [6, 7, 9, 11, 26]. A clinical study from Japan has revealed that HbA1c is significantly associated with the complexity of coronary lesions even in non-diabetic adults . Likewise, such an association was observed for ischaemic stroke in other cohort studies [8, 11, 12]. Our findings are in accordance with these studies. On the other hand, some prospective studies have demonstrated that the risk of CVD was increased only in subjects with diabetic levels of HbA1c and not in those with prediabetic levels [10, 13, 28]. This inconsistency in findings might be caused by differences in population and methodology among the studies.
Contrary to the studies reporting an association between HbA1c levels and the risk of CHD or ischaemic stroke, investigations on haemorrhagic stroke have been scarce. One prospective study from Japan assessed the association between HbA1c levels and haemorrhagic stroke incidence, and found no significant association . In the present study, the association of HbA1c levels with the risk of haemorrhagic stroke was also not significant. These findings suggest that hyperglycaemia scarcely affects the aetiology of haemorrhagic stroke, which is generally recognized as the disruption of a vessel wall . However, the number of haemorrhagic stroke events was limited in these studies. Therefore, further epidemiological research is required to elucidate this issue.
To the best of our knowledge, this is the first cohort study from Asia to demonstrate the predictive ability of HbA1c to identify persons at high risk of developing CVD. The Framingham Study showed that the addition of glycaemic categories did not substantially improve the ability to predict CVD beyond comprehensive CVD risk factors alone . However, two recent cohort studies reported opposite results. The addition of HbA1c to the Framingham risk score made a small but statistically significant improvement to the ability to discriminate CHD in men in the European Prospective Investigation of Cancer - Norfolk cohort . The Atherosclerosis Risk in Communities Study also demonstrated that NRI and IDI for CHD were significantly improved with the addition of HbA1c to the model of fasting plasma glucose and other covariates . These findings, when taken together with ours, indicate it is likely that adding HbA1c to other comprehensive risk factors will improve future discriminatory ability for CVD. Since few studies have investigated this issue, further research is expected.
Why does HbA1c have the potential to predict CVD? Complex and diverse hypotheses have been proposed to explain the causal relationship between hyperglycaemia with atherosclerosis, and one of the important pathways is recognized as glycosylation. Glycosylation is a non-enzymatic reaction induced by chronic hyperglycaemia, and is process in vivo results in two different products: early and advanced glycation end products (AGEs). HbA1c is well known as one of the early glycation end products and is a precursor of AGEs [30, 31]. It was reported that AGEs decreased large-vessel elasticity and induced inflammatory and pro-thrombotic responses in the vessel wall, thereby being involved in vascular complications [32, 33]. Some clinical studies demonstrated that high AGEs levels were associated with the risk of CVD [34, 35]. Therefore, the biological mechanisms of glycosylation may be one of the reasons for the relationship between HbA1c and the risk of CVD observed in our study.
In the present study, the linear association between HbA1c levels and the risk of incident CVD was observed. There is a debate as to whether or not the low-normal glycaemic state determined by HbA1c is associated with increased CVD risk. At least five studies have shown linearly increasing risks of CVD with higher HbA1c [2, 9, 36–38]. In contrast, other studies showed a U- or J-shaped curve for the risk of CVD incidence and total or CVD death [10, 11, 39, 40]. This discrepancy might result from differences in study design, outcomes, and race or ethnicity. Further exploration of the health risks associated with the low-normal glycaemic state derived from HbA1c is warranted.
The strengths of our study include its longitudinal population-based design, low selection bias at baseline, perfect follow-up of subjects, and accuracy of diagnosis of CVD. However, some limitations of our study should be discussed. First, HbA1c and other potentially confounding factors were based on a single measurement at baseline, although this limitation is typical of most prospective studies. During the follow-up, risk factor levels changed due to modifications in lifestyle or medication, and misclassification of these levels was possible. This could have weakened the association found in this study, biasing the results toward the null hypothesis. Thus, the true association may be stronger than that shown in our study. Second, our study included the relatively small number of event cases of CVD, consisting of stroke and CHD. The cardiovascular spectrum ranges much further, including congestive heart failure, transient ischaemic attack, and peripheral artery disease. If the CVD range would be extended with these diseases, the number of events would increase, and this would solve the power issue. Unfortunately, however, we could not include congestive heart failure, transient ischaemic attack, or peripheral artery disease in the endpoints, because we do not have information on events of these diseases. Last, our study population was comprised of one ethnic group, and thus, generalizability to other ethnicities may be limited. On the other hand, the use of a monoethnic group in such a study avoids problems relevant to population stratification artifacts. Further studies in other ethnic groups will be needed to determine the applicability of HbA1c levels to the prediction of vascular events.