Introduction

Every year, more than 5 million patients with diabetes worldwide die from diabetes-related diseases [1], with cardiovascular disease (CVD) as the most prevalent cause of death [2]. The term prediabetes, representing a number of intermediate glycemic stages between normal glucose regulation and diabetes, is associated with a high risk of developing both diabetes and CVD [3, 4]. For that reason, it is currently debated whether individuals with prediabetes should be identified and targeted for interventions aimed at reducing CVD risk [5•]. In the following, we provide an update of the relationship between prediabetes and CVD risk as well as suggestions for how to identify, risk stratify and manage CVD risk in individuals with prediabetes.

Definition of Prediabetes

Although prediabetes is a high-risk state for developing type 2 diabetes, the term prediabetes has been criticized, mainly because approximately 30 % of people with prediabetes will not progress to type 2 diabetes [6–8]. Using an oral glucose tolerance test (OGTT) for the diagnosis of prediabetes, there is also a high day-to-day variation, especially for 2-hour plasma glucose concentration [9]. Furthermore, the fasting duration as well as the time of the OGTT also influence plasma glucose levels [10], altogether resulting in a high risk of misclassification. Other names for prediabetes have been suggested, (eg, “intermediate hyperglycemia” or “high risk state of developing diabetes” [11, 12]); however, for simplicity we will use the term prediabetes in this review.

Isolated impaired fasting glucose (IFG), isolated impaired glucose tolerance (IGT), combined IFG and IGT as well as high-risk HbA1c concentrations are all part of the term prediabetes. The diagnostic criteria for prediabetes differ between the American Diabetes Association (ADA) and the World Health Organization (WHO) [5•,13, 14] (Table 1), with lower cutpoints in the ADA criteria. As a result of the different cutpoints there will be more prediabetic individuals for a given population when using the ADA criteria compared with the WHO criteria. For example, a shift from the WHO to the ADA criterion for prediabetic fasting glucose will result in a 2–3 fold increase in the prevalence of IFG [15–17]. Moreover, it should be noted that individuals diagnosed with prediabetes based on the ADA criteria will, on average, be at lower risk of diabetes and CVD than individuals with WHO-defined prediabetes [15, 18••].

Table 1 Diagnostic criteria for prediabetes based on the ADA (5) and WHO (11) criteria
Full size table

Cardiovascular Disease Risk in Prediabetes

Prediabetes by OGTT

During the last decades it has been a general notion that people with IGT have higher risk of CVD morbidity and mortality than people with IFG. This thought is predominantly based on findings from the large European multicenter study DECODE, which found an independent association of future risk of CVD mortality with IGT, but not with IFG [19]. A Japanese study supported that IGT is more predictive than IFG in terms of CVD events and mortality [20]. In contrast, the AusDiab study found that, compared with NGT, the relative risk of developing CVD was 2.5 (95 % CI: 1.2–5.1) for people with IFG, but only 1.2 (0.7–2.2) for those with IGT using the WHO definition [21]. The Emerging Risk Factors Collaboration also reported 11 %–17 % higher risk of developing coronary heart disease (CHD) among people with IFG (5.6–7.0 mmol/L) compared with those with normal fasting glycemia [22]. In addition to ethnic differences and confounding factors, the diverse findings are likely to be explained by dissimilar control groups. For instance, the control group for IFG in the DECODE study [19] included participants with IGT or even diabetes based on 2-hour glucose concentration, making the “control” group at relatively higher risk in the DECODE study [19] compared with for instance the AusDiab study [21] and the Emerging Risk Factors Collaboration [22] using the more strict NGT criterion or normal fasting glucose as control groups.

When studying plasma glucose concentrations as continuous variables the results also differ between studies. The DECODE study and analyses from 5 different Finnish cohorts found that 2-hour plasma glucose concentration is a stronger predictor than fasting glucose (per standard deviation) for CVD morbidity and mortality among individuals with normal glucose tolerance, prediabetes or screen-detected diabetes [23•]. Moreover, the large American ARIC study did not find an association of fasting plasma glucose with either CHD or ischemic stroke [24]. However, in the AusDiab study [25], 1 standard deviation increase in fasting plasma glucose was associated with a ∼30 % increased risk of CVD mortality, which was similar to the excess risk (∼20 %) associated with 1 standard deviation increase in 2-hour plasma glucose concentration.

Meta-analyses of prospective observational studies examining CVD risk in people with IFG vs IGT as well as CVD risk in relation to changes in fasting vs post-OGTT plasma glucose concentrations as continuous variables have recently been conducted [18••,26••,27]. Ford et al. included 18 studies from 1997 through 2008, and concluded that people with prediabetes in general have ∼20 % increased risk of CVD, irrespective of the type of prediabetes (IFG vs IGT) or the criterion used to define IFG [18••]. Other meta-analyses including blood glucose in the nondiabetic range as a continuous variable have shown that both fasting and post-OGTT glucose concentrations are associated with incident CVD and CHD, but with slightly higher relative effects of 2-hour glucose than fasting glucose when taking the larger standard deviation of the 2-hour glucose measurement into account [26••,27].

Prediabetes by HbA1c

With the recent addition of HbA1c to the diagnostic criteria for prediabetes the association of prediabetic HbA1c levels with CVD outcomes is also of interest. In the ARIC study, HbA1c levels between 5.5 %–6.0 % were associated with 45 % increased risk of CHD, but not ischemic stroke, whereas HbA1c concentrations between 6.0 %–6.5 % (prediabetic range) were associated with 40 %–60 % increased risk of both CHD and stroke compared with people with HbA1c levels of 5.0 %–5.5 % [24]. Analyzed as a continuous variable in the AusDiab study, the excess relative risk of CVD mortality associated with 1 standard deviation increase in HbA1c was 20 %, whereas HbA1c was not associated with all-cause mortality [25]. A meta-analysis of 9 prospective studies, among them the ARIC and the AusDiab studies, concluded that a 1 %-point increase in HbA1c within the nondiabetic range was associated with 20 % increased risk of CHD [27].

To sum up, prediabetic levels of fasting glucose, 2-hour glucose, and HbA1c are all associated with increased CVD risk, suggesting that individuals with prediabetes defined by either criterion should be targeted for CVD prevention. There is no clear evidence as to whether people with IGT are at higher CVD risk than those with IFG or with prediabetes defined by HbA1c criteria. Inconsistency in study findings may be related to differences in sex and age as discussed below.

The Role of Sex and Age

In the nondiabetic population, CVD risk is lower in women than in men [28, 29]. The picture, however, is not that clear when studying people with diabetes [28, 30–33] or prediabetes [18••,34–36]. One meta-analysis showed no difference in the relative risk of CVD outcomes between men and women with IFG [18••]. Another meta-analysis found that the effect of post-OGTT glucose concentrations on CVD risk was less strong in men analyzed separately compared with men and women analyzed together [26••]. This finding suggests a stronger effect of post-OGTT glucose concentrations on CVD risk among women. The stronger associations found in women between glycemia and CVD risk is supported by a study of older American nondiabetic men and women, which showed that HbA1c was associated with a significantly increased risk for fatal CVD and ischemic heart disease among women, but not among men [37]. In contrast to these findings, results from a cross-sectional study recently suggested that the overall female advantage in CVD risk results from an attenuation of the relationship between insulin resistance and CVD risk [38].

In addition to sex, age seems to be an important contributor to the diversity of study findings. Whereas differences between men and women are more pronounced among younger people [38], differences between fasting and 2-hour glucose concentrations on CVD risk is more noticeable in older individuals. In a stratified meta-analysis of 13 studies, the relative risk of post-OGTT glucose concentrations on CVD events was significantly higher than that of fasting plasma glucose in individuals with a mean age ≥55 years of age, but not among those <55 years [26••]. However, the relative risk of both fasting and post-OGTT glucose in general tended to be higher among the younger study participants [26••]. Furthermore, an international case–control study, the INTERHEART study, found that the association between glycemia (HbA1c) and incident myocardial infarction was stronger among younger individuals compared with older individuals [39]. Together, these findings indicate that the relative importance of plasma glucose levels on CVD risk decreases with age, probably as a result of development of other cardiometabolic risk factors.

Does Prevention of Diabetes Reduce CVD Risk in People with Prediabetes?

Both lifestyle and pharmacologic interventions have proven effective in terms of lowering diabetes risk in individuals with prediabetes [6, 7, 40–42]. However, despite the elevated CVD risk among people with prediabetes, it has been difficult to document CVD risk reduction as a result of diabetes prevention strategies (Table 2).

Table 2 Effects of lifestyle or pharmacologic intervention on cardiovascular disease in diabetes prevention studies
Full size table

Lifestyle Interventions

The Chinese Da Qing IGT and diabetes study was the first large randomized controlled trial to show beneficial effects of lifestyle intervention on prevention of diabetes in individuals with IGT [40]. Despite the beneficial effects on glucose outcomes, the 20-year follow-up of the Da Qing study showed no effect of lifestyle intervention on CVD events and mortality [43]. Likewise, the Finnish Diabetes Prevention Study also failed to show beneficial effects of lifestyle intervention on CVD outcomes even though a significant reduction in diabetes risk was found [44]. In the 10-year follow-up of the Finnish Diabetes Prevention Study, total mortality and CVD incidence were not different between the intervention and control groups, but the study participants who had IGT at baseline had lower all-cause mortality and CVD incidence, compared with a Finnish population-based cohort of people with IGT [44], suggesting that regular follow-up of individuals at high risk may improve long-term outcomes. Despite the lack of significant effects on major CVD events, lifestyle intervention can reduce CVD risk factors in people with prediabetes. The Diabetes Prevention Program showed improvements in both lipid levels and blood pressure in prediabetic individuals who regressed to NGT during the study period, and the largest improvements were seen in the intensive lifestyle intervention group [45, 46]. However, the Diabetes Prevention Program has not yet reported long-term effects of the different prevention strategies on CVD events.

Pharmacologic Interventions

In addition to lifestyle interventions, several pharmacologic trials have shown beneficial effects on diabetes risk reduction in individuals with IGT [7, 41, 42, 47]. However, in terms of CVD prevention, the pharmacologic studies have been as disappointing as the trials focusing on intensive lifestyle modification. The NAVIGATOR trial did not find any effect of treatment with nateglinide (insulin secretagogue) or valsartan (a blocker of the renin-angiotensin system) on CVD outcomes in individuals with IGT [47, 48]. Nor did the DREAM trial show any benefit of treatment with rosiglitazone (insulin sensitizer) or ramipril (a blocker of the renin-angiotensin system) on incident CVD events in people with IFG or IGT [49]. The ORIGIN study examined the effect of insulin glargine on CVD outcomes, but also showed no difference compared with placebo treatment in people with IFG and/or IGT [50]. The most promising study so far has been the diabetes prevention trial STOP-NIDDM [41], in which a significant 49 % reduction in major CVD events were found in participants with IGT who were randomized to the α-glucosidase inhibitor acarbose and followed for 3–4 years [51••]. Also the risk of developing hypertension was significantly decreased with acarbose treatment in the STOP-NIDDM trial [51••]. The authors speculate that the beneficial effect was related to concomitant reductions in body weight, waist circumference, blood pressure, and plasma triglyceride levels [51••], but also changes in the gut incretin hormone glucagon-like peptide 1 [52] could have played a role. It should be noted that the STOP-NIDMM trial was powered for incident diabetes, not for CVD, and more studies are needed before acarbose should be recommended for primary prevention of CVD.

Glucose Reduction vs Reversal of Underlying Pathophysiology

The lack of beneficial effects of lifestyle and pharmacologic interventions on CVD outcomes in individuals with prediabetes raises the question of whether glucose lowering per se should be the main focus of type 2 diabetes treatment or its prevention. It may be more beneficial to focus on changing the processes that underlie hyperglycemia or reversal of the vascular damage that may already have occurred. Insulin resistance often clusters with other cardiometabolic risk factors, such as visceral adiposity, hypertension, and dyslipidemia, making insulin resistance likely to be responsible for the higher CVD risk in hyperglycemic individuals [53–56]. By using gold standard measures of insulin sensitivity, an observational study showed that insulin resistance is largely responsible for the relationship between plasma glucose levels and CVD risk as estimated by the Framingham CVD risk score [54]. Because of this close association between insulin resistance and CVD risk, treatment modalities for improvement of insulin resistance, rather than glucose levels per se, may be more efficient in terms of reducing CVD risk. For instance, a slowing of the progression of carotid intima-media thickening of 30 %–70 % was found in people with prediabetes treated with pioglitazone (insulin sensitizer) for <3 years [57, 58]. However, so far there is no evidence of beneficial effects of insulin-sensitizing agents on long-term CVD outcomes in individuals with prediabetes.

Management of CVD Risk in Prediabetes

Strategies for CVD Risk Management in Prediabetes are Needed

The high risk of CVD among people with prediabetes warrants increased awareness of systematic CVD risk management in clinical practice. In the light of the disappointing results of prevention of diabetes on incident CVD, the elevated CVD risk in prediabetes cannot be tackled by focusing solely on hyperglycemia, but must be addressed directly. Both EASD [59•] and ADA [5•] recommend that individuals with prediabetes should be informed of their increased risk for CVD and counseled about effective strategies to lower their risk. However, no studies have assessed the benefit of primary prevention of CVD in individuals with prediabetes, and only few studies have had the power to perform sub-group analysis according to glucose tolerance status in the prediabetic range. In the Jupiter study, people without previous CVD or diabetes were randomized to statin therapy (rosuvastatin) or placebo and followed for up to 5 years [60]. Individuals with prediabetes by fasting glucose or HbA1c who received rosuvastatin had a significant reduction in CVD endpoints compared with those receiving placebo. Importantly, however, rosuvastatin increased the risk of developing diabetes, but it was concluded that the benefits on CVD and mortality exceeded the diabetes hazard [60]. More randomized controlled trials focusing on management of lipids and hypertension in individuals with prediabetes are needed. In the meantime, individuals with prediabetes are in need of early risk assessment to identify comorbidities and factors that increase cardiovascular risk. This includes evaluation of: (1) risk factors (eg, lifestyle factors including smoking, hypertension, and dyslipidemia); (2) early microvascular complications and cardiac autonomic dysfunction; and (3) comorbidities (eg, heart failure and arrhythmias). Accordingly, successful risk prevention depends on a comprehensive detection and management of all modifiable risk factors, as can be visualized by the use of risk engines.

Evaluation of CVD Risk in Prediabetes

Prediction models of CVD risk have been developed both for the general population and for persons with diabetes. Risk models for the general population include age, male sex, high blood pressure, anti-hypertensive treatment, smoking, dyslipidemia, and diabetes [61–66]. The need for diabetes-specific CVD risk models [67–69] arose from studies showing an impact of age at diabetes diagnosis as well as diabetes duration on the prognosis for developing diabetes-related complications [70]. Also, there has been some evidence for a stronger association between triglycerides and HDL-cholesterol with CHD in individuals with type 2 diabetes compared with individuals from the general population [71]. Although it is evident that prediabetes increases the risk of CVD by approximately 20 % [18••,72], it is less clear whether prediabetes directly causes CVD or whether its effect is mediated through other known risk factors for CVD. Most risk models for CVD in the general population merely include presence of diabetes and not glucose [61, 64–66, 73–75]. Only two risk models for fatal CVD have included glucose as a risk factor [76, 77], and so far no risk prediction models for nonfatal CVD events have included glucose or prediabetes. One argument for developing separate models for prediabetic vs normoglycemic individuals is if certain known risk factors for CVD operate differently in the different states, ie, if the relative risk is different in people with prediabetes than in those with normal glucose regulation. Another argument would be the existence of markers posing an increased CVD risk only in the prediabetic state. However, there are no studies showing a modifying effect of glucose (or prediabetes) on any of the known risk factors for CVD. Also, no CVD markers specific to the prediabetic state have been reported. Accordingly, there is currently no evidence supporting the need for separate CVD risk models for people with prediabetes. While there may be no basis for prediabetes-specific CVD models, being in the prediabetic state seems to pose an excess risk for developing CVD independent of other risk factors [21, 22, 24]. Hence, risk scores developed for normal glucose tolerance may systematically underestimate CVD risk in prediabetic individuals. However, because the relationship between CVD and glucose seems to be linear with no evidence of a threshold effect [19, 25, 78, 79], it seems more reasonable to include glucose as a linear term in future CVD risk models [80] rather than including a binary variable based on somewhat arbitrarily defined glucose ranges (prediabetes) [81].

Conclusions

Individuals with IFG, IGT, or prediabetic HbA1c levels have ∼20 % increased risk of developing CVD compared with individuals with normal glucose regulation. Several studies have shown beneficial effects of lifestyle modification or pharmacologic treatment on prevention of diabetes in individuals with prediabetes (mainly IGT), but the evidence regarding CVD prevention is weak. While interventions with lifestyle modification or pharmacologic agents seem to improve CVD risk factors, only acarbose have proven effective in terms of reducing CVD events among people with IGT. More studies examining effects of lipid- and blood pressure-lowering treatments in prediabetes are needed.

Another important concern is how people in need for CVD prevention should be identified. Several risk prediction models for CVD have been developed; however, none of them include presence of prediabetes as a risk factor. Future risk prediction models should take into account blood glucose as a continuous variable and preferably with different impact of blood glucose in different age groups, because the relative importance of blood glucose on CVD risk seems to decrease with age.