Revisiting the links between glycaemia, diabetes and cardiovascular disease
Whilst the interplay between type 2 diabetes and cardiovascular disease (CVD) has been recognised for many years, recent analyses of existing studies have helped refine several aspects of this relationship with relevance to clinical practice. First, recent data demonstrate that fasting glucose is not linearly related to CVD risk in those without diabetes; rather, risk levels escalate (modestly at first) only beyond specific glucose thresholds. Consequently, glucose-based measures may not necessarily enhance CVD risk prediction in those without diabetes. Second, other data confirm that new-onset diabetes is not a post-myocardial infarction ‘risk equivalent’ state and that, on average, several years of diabetes duration is needed to attain this level of risk. Third, meta-analyses and systemic reviews have confirmed that diabetes increases CVD risk by around twofold on average and this risk is subject to wide variation, being lowest in those newly diagnosed and highest in those with existing vascular disease, proteinuria or renal disease. Fourth, meta-analyses of major glucose-lowering trials suggest that, whilst glucose-lowering lessens vascular risk, risk reduction arising from statins and blood pressure-lowering is greater. Fifth, statins increase diabetes risk, albeit modestly, adding to the emerging concept that some agents that primarily target CVD risk may be diabetogenic, and vice versa. Finally, arising in part from the latter observation, as well as an understanding that CVD and diabetes risk overlap in some individuals but not others, the case for combined CVD/diabetes risk screening (generally using non-fasting lipids and HbA1c), has gained strength.
KeywordsCardiovascular risk Glucose HbA1c Linearity Meta-analysis Review Statins Triacylglycerol
Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation
Emerging Risk Factors Collaboration
Food and Drug Administration
Fasting plasma glucose
UK Prospective Diabetes Study
West of Scotland Coronary Prevention Study
This review, based in large part on the Minkoswki Lecture presented in Lisbon 2011, focuses on recent notable gains made in our understanding of the relationship between glycaemia, diabetes and cardiovascular disease (CVD). Rather than being a detailed exposition of the area, it addresses broad concepts using a mixture of meta-analyses of both cohort studies and clinical trials, novel analyses of existing studies, and lateral thinking and data presentations to extend older ideas. In so doing, the work of many knowledgeable colleagues is presented (in addition to my contributions), and it becomes clear that significant gains in knowledge can flow from the sharing of datasets of many studies to yield greater power and, thus, greater precision. The value of strong epidemiological teams working in tandem with clinically active researchers should also become apparent. Throughout the review, attempts have been made to place the conclusions into a clinical context and to suggest areas for further research. The review concludes with a brief look at the potential future pattern of CVD in type 2 diabetes, and the challenges this could bring.
Do glycaemia levels matter to CVD risk in those without diabetes?
Fasting blood glucose and CVD risk
Higher glucose levels in the non-diabetic range are, obviously, risk factors for diabetes. Indeed, we previously demonstrated that the risk of developing diabetes in men with fasting glucose levels > 5.0 mmol/l is nearly ninefold that of men with levels < 4.3 mmol/l, even after adjusting for simple risk factors inclusive of age, BMI and blood pressure . However, the relationship between fasting glucose levels and CVD risk in the non-diabetic range has been the subject of uncertainty, with variable reports of linear relationships, threshold effects, no associations or J-shaped associations (reviewed in ). This is an important issue, since it would help to determine the value, if any, of using glucose measurements to improve vascular risk prediction. We examined the relationship in a 15 year follow-up of the West of Scotland Coronary Prevention Study (WOSCOPS), during which we observed 2381 CVD events and 1244 deaths, making this, at that time, one of the most powerful studies to address this issue . The results showed no significant associations between higher fasting plasma glucose (FPG) within the normal range and incident CVD events or total mortality (most point estimates close to one). That noted, the upper confidence intervals were rather modest, such that more powerful studies were needed to bring us closer to a definitive answer.
HbA1cin the non-diabetic range and CVD risk
Having established that FPG levels do not aid vascular risk prediction in those without diabetes, it is worth considering the value of HbA1c for this purpose, especially as a recent meta-analysis demonstrated a somewhat stronger association between HbA1c levels and CHD risk . HbA1c reflects ambient glycaemia for around 3 months and, interestingly, at lower levels is determined to a greater extent by postprandial glycaemia than FPG levels. For HbA1c there are far fewer published studies that report reclassification metrics rather than simple HRs or ORs.
Diabetes as a CHD ‘risk equivalent’?
Average excess risk level for vascular disease in diabetes patients?
Further data from the ERFC showed that, overall, CVD risk is around twofold greater in diabetic patients compared with people without diabetes, independently of conventional risk factors . There were trends for HRs to be greater in younger people and in women with diabetes . The ERFC analyses were not able to assess the relevance of diabetes duration or renal disease. However, as we showed , diabetes duration is important in determining the level of CVD risk, and other recent papers have shown that proteinuria  and low estimated GFR (eGFR) (< 60 ml min–11.73 m−2)  mark out diabetic patients with excessive CVD risk, with the combination of low eGFR and proteinuria heralding substantially elevated vascular risk . Moreover, it is clear that those with diabetes and CHD are at especially higher CVD risk. Hence, the overall vascular risk of around twofold encapsulates a range of risks; from the relatively low short to medium term vascular risk in patients with newly diagnosed or a short duration of uncomplicated diabetes, to substantially elevated (about four- to fivefold higher HRs) in diabetic patients with existing CHD or proteinuria/low eGFR.
The vascular benefits of lowering glucose in diabetes?
The mechanism for elevated vascular risk in diabetes remains uncertain but, clearly, elevated glucose must be considered a prime suspect. The seminal UK Prospective Diabetes Study (UKPDS) showed clear benefits of intensive glycaemic control on microvascular risks but failed to show a clear benefit on macrovascular events . Longer term follow-up of UKPDS patients for 10 years after study end, during which differences between HbA1c levels in the trial arms largely disappeared, did show significant reductions in MI for those initially treated with intensive glucose-lowering therapy . These findings have been interpreted as suggesting a ‘legacy effect’ and, of interest, this concept may fit with the fact that around a decade of diabetes duration is needed to obtain a CHD risk equivalent state.
Of course, several recent studies (Action to Control Cardiovascular Risk in Diabetes [ACCORD], Action In Diabetes And Vascular Disease: Preterax And Diamicron MR Controlled Evaluation [ADVANCE], Veterans’ Affairs Diabetes Trial [VADT]) [18, 19, 20] have further tested the intensive glycaemic control hypothesis, with some unexpected results necessitating a reappraisal of the hypothesis. We conducted the first meta-analysis of randomised controlled trials that aimed to assess whether more intensive control of glucose reduces long-term CVD clinical events and lengthens lifetime compared with standard treatment . We did so because individual trials may have been underpowered to show clinical benefit. Our analysis demonstrated that a 0.9% reduction in HbA1c over an average of 5 years of follow-up resulted in a 17% reduction in non-fatal MI and a 15% reduction in CHD events, but with no significant effects on stroke or all-cause mortality. A subsequent high-quality meta-analysis of individual participant data reported similar results and also suggested vascular risk reduction with intensive glycaemic control was more likely in diabetic patients without baseline macrovascular disease .
HbA1cand CVD risk in diabetes? Is there a threshold effect?
Whilst the UKPDS data originally suggested vascular risk increases from relatively modest HbA1c levels (of ∼5.5%, 37 mmol/mol), recent analyses of the ADVANCE study challenged this finding . In ADVANCE, microvascular risk appeared to increase only once HbA1c exceeded 6.5% (48 mmol/mol), while CVD risk increased when HbA1c exceeded 7.0% (53 mmol/mol). Risk associations appeared to be relatively weak below these thresholds . Although these are post hoc analyses of trial data, they may partially explain the recent neutral results of the Outcome Reduction with Initial Glargine Intervention (ORIGIN) study . Here, basal insulin was used to bring down fasting glucose to normal levels in participants with cardiovascular risk factors plus impaired fasting glucose, impaired glucose tolerance or type 2 diabetes, but it was notable that the median HbA1c at baseline in both ORIGIN glycaemia arms was only 6.4% (46 mmol/mol), which is well below the threshold of 7.0% (53 mmol/mol) for CVD risk suggested by ADVANCE data. In other words, CVD endpoint trials of glycaemia-lowering agents may be best conducted in diabetic patients with an HbA1c level well above 7.0% (53 mmol/mol). Of course, recruited patients need to have a high CVD risk or prevalent disease for the study to have sufficient power.
Triacylglycerol levels: stronger indicators of metabolic dysregulation/ectopic fat vs vascular risk?
Elevated LDL-cholesterol is well recognised as a causal risk factor for CHD and, based on robust trial evidence, statins are now the mainstay of lipid-lowering for prevention of CVD. However, whilst cholesterol targets can be reached in the majority of our patients, many continue to have elevated triacylglycerol levels, but the degree to which this represents an additional independent CVD risk factor remains in doubt. To resolve this issue, we related lipids to CVD risk in a meta-analysis of prospective studies that was based on individual participant data . This work showed that whilst elevated cholesterol and low HDL-cholesterol predicted higher CVD risk in age, sex and other lipid-variable-adjusted models, triacylglycerol levels were not related to risk for vascular events once non-HDL-cholesterol and HDL-cholesterol were accounted for. The results were similar in patients with or without diabetes. As a result, we recommended that lipid assessment in vascular disease can be simplified by measurement of total and HDL-cholesterol levels without the need to fast and without regard to triacylglycerol . Of course, this epidemiological work does not necessarily mean that triacylglycerol is not causally related to vascular risk, which is a separate question requiring different approaches. Nevertheless, it is of interest to note that recent fibrate trials failed to show a benefit in the prevention of vascular events in diabetic patients. The results have led the US Food and Drug Administration (FDA) to recommend a need for further large-scale trials to allow definitive conclusions in this area, but for now fibrates are not routinely recommended .
What is perhaps less well perceived is that for a given elevation in triacylglycerol in people without diabetes, a given increment in triacylglycerol is a stronger risk factor for diabetes as compared with its weaker associations with vascular risk . This is partly because, in the majority, elevated plasma triacylglycerol levels represent another manifestation of ectopic fat whereby excess fat is being deposited in the plasma pool, in the same way as excess fat appears in the liver in many patients at risk for type 2 diabetes, or with prevalent diabetes . Furthermore, and as most physicians recognise, triacylglycerol levels can be markedly elevated when patients have poor diabetes control, with rapid improvement possible with an appropriate reduction in glycaemia levels.
Furthermore, and as noted previously, as with triacylglycerol, FPG in the non-diabetic range is also a far stronger risk factor for type 2 diabetes than for CVD ; the same association also holds true for current BMI . As a result, in a novel analysis we showed that the metabolic syndrome criteria are more aligned with diabetes than with CVD risk, and consequently have little clinical value , an observation now widely accepted .
Diabetes and vascular risks in women vs men
Statins and diabetes risks
Future patterns of vascular disease in patients with diabetes?
There is now clear evidence of declining mortality rates in diabetes (compared with non-diabetes) over several decades, an observation we confirmed in a recent ERFC analysis . Furthermore, US data analyses suggest a continuing reduction in CVD mortality rates in diabetes in recent years , explicable in large part to better management of CVD risk factors . Ongoing trials with newer glucose-lowering agents (which cause far less hypoglycaemia and weight gain) may add further CVD benefit so that, at least for the foreseeable future, CVD rates may continue to decline. This noted, as CVD death rates go down, more patients with type 2 diabetes are living longer than ever and this may present newer problems, including rising levels of cognitive decline or, potentially, greater heart failure and cancer rates. However, at the same time, there is an increasing population of younger, more obese patients with type 2 diabetes , a group likely to have more rapid glycaemia progression, as well as a general rise in type 2 diabetes rates worldwide , such that diabetes will increasingly account for a higher percentage of the total CVD burden in society. These shifting patterns of disease will present unique problems in the future.
Whilst I have been involved in many of the papers discussed herein, the leading role of many excellent colleagues deserves mention. I would like to thank D. Preiss (University of Glasgow) for his critical reading of this paper and his expert work on several joint papers. I also wish to thank colleagues from the Emerging Risk Factor Collaboration, including, in particular, J. Danesh and E. Di Angelantonio (both University of Cambridge), for their excellent leadership of this important collaboration. Thanks also to G. Wannamethee (University College London), for allowing me to collaborate on novel data analyses of the British Regional Heart Study, to K. Ray (St George’s Hospital, London), for his expertise in the conduct of collaborative meta-analyses, and to N. Forouhi (MRC Epidemiology Unit, Cambridge) for many helpful discussions on aspects relevant to the current review. D. Lawlor (University of Bristol) is also credited for sharing data and for helping shape relevant papers. Finally, thanks to my many colleagues in the Scottish Diabetes Research Network Epidemiology group for allowing us to test hypotheses in this rich dataset, and to many other Glasgow colleagues for their excellent collaborations on several ideas of mutual interest.
I wish to acknowledge grant support from Diabetes UK, British Hearth Foundation, Chest Heart Stroke, Scotland and the Chief Scientist Office, Scotland in supporting the conduct of relevant work.
Duality of interest
I have consulted, or been on the speakers’ bureau, for Astrazeneca, Bristol Myers Squibb, Eli-Lilly/Amgen, MSD, Roche and Sanofi-Aventis. I have also received grant support from Merck and Pfizer.
The author was responsible for the conception, design and drafting of the manuscript, and approved the final version for publication.
- 4.Sarwar N, Aspelund T, Eiriksdottir G, et al (2010) Markers of dysglycaemia and risk of coronary heart disease in people without diabetes: Reykjavik prospective study and systematic review. PLoS Med 7:e1000278Google Scholar
- 5.Simmons RK, Sharp S, Boekholdt SM et al (2008) Evaluation of the Framingham risk score in the European Prospective Investigation of Cancer-Norfolk cohort: does adding glycated hemoglobin improve the prediction of coronary heart disease events? Arch Intern Med 168:1209–1216PubMedCrossRefGoogle Scholar
- 7.Schöttker B, Müller H, Rothenbacher D, Brenner H (2013) Fasting plasma glucose and HbA1c in cardiovascular risk prediction: a sex-specific comparison in individuals without diabetes mellitus. Diabetologia 56:92–100Google Scholar
- 10.NICE (2012) Preventing type 2 diabetes: risk identification and interventions for individuals at high risk. NICE public health guidance 38. Available from www.nice.org.uk/nicemedia/live/13791/59951/59951.pdf. Accessed 3 Jan 2013
- 33.Preiss D, Tikkanen MJ, Welsh P et al (2012) Lipid-modifying therapies and risk of pancreatitis. A meta-analysis. JAMA 304:804–811Google Scholar