Introduction
Individuals with type 1 diabetes are characterised by accelerated arterial ageing [
1,
2], a mechanism that increases the risk for cardiovascular disease (CVD) [
2‐
4]. This has been illustrated by a steeper positive association between age and pulse pressure, a marker of arterial stiffness [
5,
6], in individuals with type 1 diabetes than in their peers without the disease [
1]. This is also supported by many studies demonstrating greater arterial stiffness in diabetes, as ascertained by measures of arterial pulse wave velocity or local estimates such as distensibility coefficients of the carotid and others arteries [
3]. Importantly, in individuals with type 1 diabetes, higher pulse pressure (pulsatile load) has been associated with incident CVD independently of mean arterial pressure (MAP; steady load) [
2].
The pathobiological mechanisms underlying the increases in pulse pressure in individuals with diabetes are, however, not clear. Diabetes is characterised by systemic and vascular inflammation and endothelial dysfunction [
7‐
9], mechanisms that may link type 1 diabetes to increased pulse pressure [
3]. Indeed, in recent years, inflammation and endothelial dysfunction, as expressed by levels of C-reactive protein (CRP) and cellular adhesion molecules (CAMs), have been proposed as putative determinants of arterial stiffness/pulse pressure and hypertension [
10‐
12]. Epidemiological evidence to support this hypothesis remains weak, however, mainly due to limitations in study design. In addition, this has never been examined in the context of type 1 diabetes.
We have therefore investigated the longitudinal course of BP and markers of systemic and vascular inflammation and endothelial dysfunction (CRP, soluble intracellular adhesion molecule-1 [sICAM-1], soluble vascular cellular adhesion molecule-1 [sVCAM-1], and soluble E-selectin [sE-selectin]) and their temporal inter-relationships, in a cohort of individuals with type 1 diabetes who were followed over 20 years, since the onset of disease.
Discussion
We examined the longitudinal course of BP and biomarkers of endothelial dysfunction and inflammation, and their inter-relationships, in a cohort of individuals with type 1 diabetes who had been followed for 20 years since the onset of disease. Our main findings were as follows: (1) increases in levels of CRP, sICAM-1 and sVCAM-1 occurred early in the course of disease and preceded the increases in BP; (2) all variables tracked considerably over time, particularly sICAM-1, sVCAM-1 and sE-selectin; (3) higher sICAM-1 and sVCAM-1, at baseline and during follow-up, predicted the subsequent (changes in) levels of pulse pressure; higher levels of sVCAM-1, at baseline and during follow-up, predicted the prevalence and incidence of hypertension; (4) notably, these associations were independent of age, sex, smoking history and other relevant risk factors and (5) we found no evidence for a reverse causation hypothesis (i.e. that higher levels of BP or hypertension determined subsequent increases in these biomarkers). The unique characteristics of this study were the repeated assessment of biomarkers, BP and important covariates over the natural course of disease during a period of 20 years. This allowed us to examine, for the first time with a truly longitudinal design, the temporal relationships between biomarkers and BP (notably, pulse pressure) in individuals with type 1 diabetes. As such, the present study provides the strongest evidence regarding the involvement (or lack thereof) of CRP, sICAM-1, sVCAM-1 and sE-selectin in the pathogenesis of elevated pulse pressure and of hypertension, over the course of disease in individuals with type 1 diabetes.
Tracking of BP and CRP levels over the 20 year follow-up period was moderate but was very high for sICAM-1, sVCAM-1 and sE-selectin. Interpretation of the tracking coefficients reported here requires consideration of a number of factors [
25]. First, these coefficients reflect only the stability of one’s rank position vs peers over time. Second, the magnitude of these coefficients tends to decrease with the length of follow-up. Remarkably, the tracking coefficients of the BP variables were comparable with those reported in a 15 year longitudinal study among healthy individuals followed from adolescence to young adulthood [
25]. To our knowledge, long-term tracking coefficients for the biomarkers examined herein have never been reported before. The very high tracking of both CAMs suggests that they can be used to identify, early in the disease, those who are likely to remain with adverse levels leading to related sequelae. Third, measurement error attenuates tracking coefficients, which may explain why these were stronger for the biomarkers than for BP. Indeed, throughout the years, BP was measured in the clinical setting as part of the routine follow-up care provided to patients by different attending physicians and nurses, whereas biomarkers were all assessed at the same laboratory at the same time, using the same methodology. Finally, tracking coefficients may also be affected by extraneous factors, although adjustments for both time-independent and time-dependent covariates only seemed to attenuate their magnitude slightly.
Epidemiological evidence to support the concept of arterial stiffness/widened pulse pressure and hypertension as a consequence of vascular/systemic inflammation has been controversial [
11,
26‐
29]. Our findings are in agreement with previous studies showing that higher levels of CRP are related to hypertension and pulse pressure [
26,
30‐
35]. However, after life-course adjustments for confounders, the associations with pulse pressure and hypertension were markedly attenuated and no longer significant, suggesting that these associations are not causal. These findings are supported by previous studies showing that the cross-sectional associations between CRP and hypertension, as well as pulse pressure [
27] or aortic pulse wave velocity [
29], disappeared after adjustment for life-course confounding or Mendelian randomisation tests of causality. Furthermore, although some prospective studies have shown associations between CRP and incident hypertension [
36,
37], others have shown these to disappear after adjustment for BMI [
38], suggesting that BMI is a common determinant of both pulse pressure/hypertension and CRP [
11]. Our data supports this hypothesis. Increases in BMI and related BP in individuals with type 1 diabetes have been well documented as long-term consequences of intensified insulin treatment [
39,
40]. In addition, increases in CRP, but not in CAMs, have been shown to depend on the degree of weight gain in insulin-treated patients with type 1 diabetes [
41]. We [
8,
42], and others [
7], have also shown that BMI is the strongest correlate of CRP in these individuals.
Our findings suggest that, in contrast to CRP (and sE-selectin), both CAMs may causally underlie arterial stiffening. This contention holds inasmuch as arterial stiffness can be depicted by pulse pressure and may be too strong given that stroke volume (not measured) is also likely to explain a part of the variance in pulse pressure, particularly among young individuals. Our findings also suggest that sVCAM-1 seems to causally underlie the development of hypertension. This observation is in striking agreement with the only prospective study thus far that has examined CAMs and pro-inflammatory cytokines as predictors of incident hypertension among individuals with type 1 diabetes [
43]. Indeed, in that study only the baseline levels of sVCAM-1 (but not sICAM-1, CRP, IL-6 and TNF-α) were associated with the 15 year cumulative incidence of hypertension. We extended these observations by examining not only associations with levels of biomarkers at baseline but also their changes during follow-up. Soluble CAMs are markers of vascular endothelial dysfunction/inflammation that may influence functional stiffening of large arteries through reduced availability of nitric oxide and/or increased activity of vasoconstrictors such as endothelin-1, both of which affect vascular smooth muscle cell tone [
9]. In addition, endothelial dysfunction/inflammation may lead to smooth muscle cell proliferation and increased synthesis of structural proteins such as collagen within the vascular wall, leading to structural stiffening of large arteries [
3,
9]. Previous studies on soluble CAMs and BP have produced contradictory findings but have been mostly cross-sectional and not focused on the pulsatile component of BP [
44‐
50]. Appreciation of the BP curve as a summation of a steady component (MAP) and a pulsatile component (pulse pressure) has provided additional information in terms of CVD risk prediction to that traditionally obtained on the basis of elevated SBP and/or DBP alone [
5].
We used brachial, not central, pulse pressure as a crude estimate of arterial stiffness, reflecting the data that was accessible from clinical records. The technology enabling non-invasive measurement of central pulse pressure and aortic pulse wave velocity, that would allow better characterisation of the aetiology of arterial stiffening, was not available during the 20 year period covered by the present study. Nevertheless, brachial pulse pressure still provides valuable risk prediction information: in a meta-analysis (not including individuals with type 1 diabetes), central pulse pressure tended to be more strongly associated with incident CVD and mortality than brachial pulse pressure but the added value of central pulse pressure in risk prediction was only marginal [
51]; a similar pattern was observed in a large cohort study of individuals with type 1 diabetes [
4].
There are some additional limitations to our study. Findings were confined to individuals with type 1 diabetes and therefore may not generalise to the background population. Measurement errors around BP measurements are likely to have led to an underestimation of the associations estimates reported in the present manuscript. We only measured CRP, sICAM-1, sVCAM-1 and sE-selectin, which reflect just a part of the complex and multifaceted process of arterial remodelling induced by endothelial dysfunction and inflammation [
9]. Still, although the CAMs studied here may be produced by different cell types, changes in their plasma levels are widely considered to reflect altered endothelial production rates [
9]. The biomarkers were measured on stored blood samples, which raises the question of whether the reported increases in their levels over time could in part reflect a decay of the proteins with storage time. We deem this unlikely because of the high long-term stability of concentrations of these proteins in stored serum [
52]. Besides, while levels of CRP, sICAM-1 and sVCAM-1 increased, levels of sE-selectin decreased over time. In addition, despite the long storage time, our measurements could capture considerably higher levels of biomarkers at disease onset (excluded from the analyses) followed by decreases after blood-glucose stabilisation, with relatively steady increases (except for sE-selectin) thereafter (see Fig.
2).
In conclusion, in individuals with type 1 diabetes, increases in sICAM-1 and sVCAM-1 precede, and are associated with, subsequent increases in pulse pressure and hypertension throughout the course of the disease, supporting the involvement of endothelial dysfunction/inflammation in the development of premature arterial stiffening. The lack of support for a causal link between CRP and BP, and the observation that both derive from a common antecedent (BMI), suggests that weight gain should be monitored during treatment of individuals with type 1 diabetes. Targeting endothelial dysfunction/inflammation in the early stages of diabetes may slow down the accelerated arterial ageing characteristic of this disease and prevent related cardiovascular sequelae. Given that both CAMs tracked very highly, measuring their levels and changes soon after the onset of type 1 diabetes may enable identification of individuals at a high risk and who may need intensified/tailored treatment.