Circulating Inflammatory Markers and the Risk of Vascular Complications and Mortality in People With Type 2 Diabetes and Cardiovascular Disease or Risk Factors: The ADVANCE Study

Gordon Lowe; Mark Woodward; Graham Hillis; Ann Rumley; Qiang Li; Stephen Harrap; Michel Marre; Pavel Hamet, Anushka Patel; Neil Poulterl; John Chalmers


Diabetes. 2014;63(3):1115-1123. 

In This Article


This is the largest prospective study to date assessing the relationship among levels of circulating inflammatory markers, risk of vascular events (both macrovascular and microvascular), and death in persons with T2DM. It shows that, after adjustment for potential confounding factors, higher levels of CRP were associated with an increased risk of macrovascular events and mortality, whereas higher levels of fibrinogen were associated weakly with mortality. However, these associations were weak, and were abolished by further adjustment for IL-6. In contrast, IL-6 showed stronger associations with the risk of macrovascular events and mortality and remained an independent predictor of these outcomes after adjustment for other risk factors and for levels of CRP and fibrinogen. In addition, the incorporation of IL-6 levels significantly, albeit modestly, improved risk prediction that was based on clinical factors alone. After adjustment for clinical risk factors, none of the three inflammatory markers were associated with future microvascular complications.

Relationships Between Inflammatory Biomarkers and Clinical Risk Factors

As expected,[6,7,13] all three inflammatory markers were associated with BMI and current smoking. However, only IL-6 level was associated with pre-existing CVD. In addition, the levels of all three markers were modestly correlated.

CRP and Fibrinogen

Large individual patient meta-analyses of general population studies have suggested that the association between CRP and fibrinogen levels and CHD are weaker in patients with diabetes than those without the condition.[6,7] In the most recent meta-analysis,[26] the addition of CRP or fibrinogen to a clinical risk score in general populations increased the 10-year NRI by 1.52 and 0.83%, respectively. Previous reports of CRP or fibrinogen in cohorts with T2DM have reported associations with cardiovascular and total mortalities[9–11] or with cardiovascular events only in men.[8] The current, and largest, study of all vascular complications and mortality in men and women with diabetes confirms previous reports that CRP[9,10] and fibrinogen[11] levels are associated with cardiovascular and total mortality, confirms two previous reports[10,19] that CRP level is associated with cardiovascular events, and confirms that CRP level is not clinically useful for prediction of cardiovascular events[19] or mortality.[9] Our study is also consistent with a recent report[27] from nine community-based prospective cohort studies, which observed that the associations of fibrinogen level with CVD and mortality (which were similar in people with and without diabetes) did not improve the predictive accuracy of established risk factors. In addition to such confirmation, our findings are novel in showing that neither CRP nor fibrinogen level is clinically useful for the prediction of macrovascular events, microvascular events, or total mortality.


High IL-6 levels are associated with obesity and insulin resistance,[15] and, together with other proinflammatory cytokines, may play a role in the pathogenesis of CVD.[12–17] In two large population-based cohorts of middle-aged individuals without known CVD, and in an associated meta-analysis, higher IL-6 levels were associated with an increased incidence of CHD during long-term follow-up,[13] an association that became more apparent after levels were corrected for intraindividual variation in IL-6 levels.[13] However, there is only one previous prospective study associating levels of IL-6 with the risk of vascular complications or mortality in T2DM: the ESTHER Study,[18,19] in which 161 subjects experienced a primary cardiovascular event and IL-6 was associated only with risk of cardiovascular events in those with renal dysfunction.[19]

The current study is novel in that it provides the first evidence, in patients with diabetes, that IL-6 levels show significant associations (which are stronger than CRP or fibrinogen level) with macrovascular complications (n = 709) and mortality (n = 706) in patients with T2DM, and it shows that IL-6 level adds significantly to their prediction from conventional risk factors. The C-statistic for macrovascular events increased by 0.012 when IL-6 was added to the prognostic model; this should be interpreted as increasing the chance of correctly discriminating between a pair of subjects with diabetes, only one of whom will go on to experience a macrovascular event within 5 years, by 0.012 (or 1.2%). For death, the corresponding increase is estimated to be 0.016 (or 1.6%). In the area of CVD prognosis, it is notoriously difficult to achieve a substantial increase in the C-statistic, largely because the classical risk factors (especially age) discriminate so well already, and because the scale of C-statistics is narrow (0.5–1). Although 0.012 seems small in absolute terms, this level of increment is not atypical for novel cardiovascular biomarkers that have been suggested for clinical use, acknowledging the reality that the remaining unexplained risk is likely to be explained in many small accumulating steps. At least in part because of the difficulty in interpreting the C-statistic, Pencina et al.[25] introduced the IDI and NRI, used here in the form suitable for survival data. Of these, the one we consider to be most clinically relevant is the categorical NRI, which measures the net improvement in prediction across clinical risk thresholds (of 5 and 10% 5-year risk). The IDI and continuous (threshold-free) NRI are less clinically useful because they may be influenced by outliers. For macrovascular events, adding IL-6 to the base prognostic set improved the categorical NRI by 0.03, sometimes roughly interpreted as an improvement of 3%. For death, the same statistic was 0.06, roughly interpreted as a 6% improvement. Taking all the metrics together, we conclude that IL-6 gives a moderate improvement in predicting who will have a macrovascular event within the next 5 years and a more substantial, albeit not major, improvement in predicting who will die within 5 years among high-risk subjects with diabetes. We thus suggest that IL-6 levels be considered for inclusion in future clinical and biomarker prediction scores for T2DM patients.

With regard to translational potential, the strong association of IL-6 levels with macrovascular events in patients with type 2 diabetes was not attributable to confounding by conventional risk factors, and it highlights the potential causal importance of proinflammatory cytokine levels in the macrovascular complications of diabetes. A potential causal role for IL-6 level is suggested by a recent meta-analysis of general population studies[14] in which not only IL-6 levels but also an associated functional mutation in the IL-6 receptor gene (rs 8192284) was associated with CHD risk (a positive Mendelian randomization study). The IL-6 receptor is also a susceptibility location with genome-wide significance for coronary artery disease.[28] In a recent report, an inflammatory risk score comprising five inflammatory gene polymorphisms, including a single nucleotide polymorphism in the IL-6 gene (rs 1800795), was associated with risk of ischemic stroke in a prospective cohort of subjects with T2DM.[29] We suggest that further studies of IL-6 and IL-6 receptor gene polymorphisms are required to establish whether or not genetic determinants of IL-6 levels are associated with the macrovascular complications of T2DM. If so, the translational importance of our findings could be further investigated by studies of IL-6 antagonists.[14] The current study suggests that a 1 SD reduction in log(IL-6) level by IL-6 antagonists might reduce the risk of myocardial infarction, stroke, or death by about a quarter.

In contrast to macrovascular events, none of the three inflammatory biomarkers was independently associated with the risk of microvascular events. These findings suggest that upregulated inflammation may be less important in pathogenesis of microvascular compared with macrovascular complications of type 2 diabetes. The absence of an association between inflammation and microvascular disease contrasts with cross-sectional data. For example, in a nested case-control study of 543 patients with type 1 diabetes who participated in the Epidemiology and Prevention of Diabetes (EURODIAB) Prospective Complications Study, higher levels of CRP, IL-6, and tumor necrosis factor-α were associated with increased urinary albumin excretion and more severe retinopathy, even after adjustment for confounding factors such as age, sex, duration of diabetes, systolic blood pressure, and HbA1c level.[30] Similarly, in patients with T2DM, higher white cell counts, though still within the normal range, were associated with an increased prevalence of retinopathy and albuminuria.[31]

Strengths and Limitations

The strengths of the current study include its size, international recruitment, rigorous definition of outcomes, completeness of follow-up, and adjustment for major risk factors. The case-cohort study is an ideal design for biomarker research when, as here, there are several outcomes of interest.[21] We show a comprehensive set of reclassification statistics, using contemporary methods. Of these, we believe that the categorical NRI, with clinically relevant thresholds, is the most appropriate measure for clinical decision making since it estimates the independent effect of a biomarker without being affected by changes in risk estimation at relatively unimportant extreme values.

One limitation of our study is the selection of persons with T2DM who had either baseline CVD or risk factors. While our findings cannot be generalized to persons with T2DM who have neither CVD nor risk factors, the baseline characteristics of the ADVANCE Study cohort are comparable to several other observational studies at the community level,[32] and hence it seems reasonable to conclude that its results are broadly generalizable, particularly for relative risks. Nevertheless, the selection criteria in the ADVANCE Study may have affected the IDI and NRI, which will not be constant across all subgroups; for example, each may differ between nonsmokers in the ADVANCE Study (who, by the selection criteria, had to have another risk factor if they were free of CVD) and nonsmokers in general diabetes populations aged ≥55 years. Another limitation is that only single measures of variables of interest were used in estimating the incremental effect of the biomarkers. However, single measures are the only viable option in risk scoring, the basic component of prognostic modeling, which was the primary focus of this research.


We conclude that IL-6 levels, but not CRP or fibrinogen levels, independently improve the clinical prediction of macrovascular events and mortality in persons with type 2 diabetes who have baseline CVD or risk factors. We also conclude that further studies exploring a potential causal role for IL-6 in the cardiovascular complications associated with type 2 diabetes are warranted, including studies of functional genotypes and, possibly, of IL-6 antagonists.