Abstract
Interleukin 6 trans-signalling is independently associated with the risk of cardiovascular events. The aim of this study was to investigate if interleukin 6 trans-signalling can identify individuals at risk for cardiovascular events (coronary artery disease and ischaemic stroke) among those at-low–intermediate risk.
In a cohort of 60-year-olds (n = 4232, incident cardiovascular events n = 525), interleukin 6 trans-signalling was estimated by a ratio between the pro-inflammatory interleukin 6: soluble interleukin 6 receptor binary receptor complex and the inactivated interleukin 6: soluble interleukin 6 receptor: sgp130 ternary complex (B/T ratio). Risk associated with B/T ratio >median was investigated in individuals with low-density lipoprotein cholesterol ≤4.0 (mmol/l) and in those at low-intermediate risk according to the Framingham risk score (FRS) using Cox regression and expressed as hazard ratio and 95% confidence interval. Difference in time to event (years; 95% confidence interval) was analysed with quantile regression. The interaction between low-density lipoprotein cholesterol and B/T ratio was estimated on the additive scale. Incremental discriminatory value of the B/T ratio if low-density lipoprotein cholesterol ≤4.0 was compared to that of the FRS and interleukin 6.
B/T ratio >median was associated with increased cardiovascular event risk when low-density lipoprotein cholesterol ≤4.0 (hazard ratio 1.59; 95% confidence interval 1.24–2.05) or FRS ≤ 10%, >10–≤20% (hazard ratio 1.27; 95% confidence interval 1.00–1.61 and hazard ratio 1.78; 95% confidence interval 1.36–2.34, respectively). B/T ratio >median and low-density lipoprotein cholesterol ≤4.0 were associated with early cardiovascular events, particularly ischaemic stroke. No interaction was observed between low-density lipoprotein cholesterol and the B/T ratio, both factors increasing cardiovascular event risk by 60%. In the presence of low-density lipoprotein cholesterol ≤4.0, the B/T ratio slightly improved discrimination measures.
Interleukin 6 trans-signalling increases cardiovascular event risk in middle-aged men and women otherwise classified at low-intermediate cardiovascular risk.
Introduction
Novel biomarkers are advocated to improve cardiovascular (CV) risk prediction in individuals currently classified as at low-intermediate risk of having a CV event (CVE) according to the available risk scores.1,2
We recently reported that biomarkers mirroring interleukin 6 (IL6) trans-signalling are independent predictors of CV events (CVEs).3,4 The IL6 trans-signalling is a pro-inflammatory pathway driven by IL6 independently from C-reactive protein (CRP). Inflammation is driven by the IL6:soluble IL6 receptor (sIL6R) binary complex, formed by IL6 bound to the sIL6R. To prevent a systemic uncontrolled inflammatory response, the circulating IL6:sIL6R complex is buffered by sgp130, through the formation of a ternary complex: IL6:sIL6R:sgp130. We have shown that an excess of the IL6 trans-signalling, estimated by the IL6:sIL6R/IL6:sIL6R:sgp130 ratio, increased the risk of first CVE and improved CV risk reclassification measures.4 Of note, IL6R regulates sIL6R levels and is causal of coronary artery disease (CAD).5
The aim of the present study was to investigate if the IL6 trans-signalling can represent a novel biomarker for increased risk of CVEs in individuals otherwise classified as at low-intermediate risk. To this extent we have analysed in a cohort of middle-aged men and women the association of the IL6 trans-signalling with the risk of early CVE in individuals with low-density lipoprotein (LDL) cholesterol ≤4.0 (mmol/l) and in those at low and intermediate risk according to the Framingham risk score (FRS). In addition, we have calculated the interaction between LDL cholesterol levels and IL6 trans-signalling in determining CVE risk. In secondary analyses, risk of CAD and ischaemic stroke were analysed separately.
Materials and methods
Additional information on materials and methods are reported in the Supplementary Material.
Study population
In the cohort of 60-year-old men and women from Stockholm (60YO), described previously,4 every third man and woman residing in the Stockholm County between 1 July 1997–30 June 1998 was randomly selected from the Swedish Population Register and invited to participate in a CV screening study. With a 78% positive response rate, 4232 subjects were included in the study. Participants were given a comprehensive self-administered questionnaire on lifestyle habits and medical history. Height, weight, blood pressure was measured; an electrocardiogram (ECG) recorded, and blood samples drawn and immediately frozen to −80° Celsius for future analysis.
Study participants were followed from the date of enrolment until the date of first-time CVE, the date of death or the end of follow-up (31 December 2014), whichever came first. The unique personal identification number of each participant was linked yearly to the mandatory Swedish national registers; the National Cause of Death Register and the Hospital Discharge Register with a 100% follow-up. Main diagnoses were recorded using the International Classification of Diseases 10th revision.6
We recorded 546 first-ever incident non-fatal and fatal CVEs: CAD (angina pectoris requiring hospital admission (I20 and I25), myocardial infarction (I21), cardiac arrest with/without successful resuscitation (I46) and ischaemic stroke (I63)). For the present analysis, only individuals without known CV disease at enrolment were included: 372 individuals were excluded because they either reported a previous CVE when completing the questionnaire or had a diagnosis in the Hospital Discharge Registry registered before the inclusion in the study. Participants with an incomplete questionnaire (n = 119), lacking IL6, and/or sIL6R and/or sgp130 measurement in serum samples (n = 96) were also excluded. From the group of incident cases, 21 cases were excluded due to diagnoses inaccurately categorising participants as cases,4 leaving 525 incident cases in the present analysis. Supplementary Material Table I summarises the clinical characteristics of the included and excluded study participants.
Biochemical measurements and derivation of the binary and ternary complex molar concentrations
IL6, sIL6R and sgp130 were analysed in serum at baseline in each study participant. Concentrations of the binary (IL6:sIL6R) and ternary (IL6:sIL6R:sgp130) complexes were calculated to estimate the binary/ternary complex ratio (B/T ratio),4 as described in the Supplementary Material.
Statistical analysis
Quantitative continuous variables were expressed as median and interquartile range whereas binary variables were presented as proportions and reported as percentages.7
The B/T ratio was dichotomised at the median based on previous analyses4 and tested for association with the risk of CVE in individuals with LDL ≤ 4.0 and >4.0. The LDL cholesterol cut-off was derived from the receiver-operating characteristics curve and based on the sensitivity and specificity results for different cut-offs (Supplementary Material Figure I).
The risk of CVE associated with high B/T ratio was also analysed in the presence of LDL < 3.0 and ≥3.0–≤4.0 and in individuals at low, intermediate and high risk according to the classification of 10-year CV disease risk used in the FRS: low ≤ 10%, intermediate>10% and ≤20% and high>20% risk.8
The risk of CVEs associated with the B/T ratio > median was estimated using Cox proportional hazards model and expressed as hazard ratio (HR) with 95% confidence interval (CI). The reference group is subjects with B/T ratio ≤ median in all analyses. Proportionality of hazards was confirmed with Schoenfeld’s test. Risk estimates were presented in a crude model and adjusted for the common CV risk factors: sex, body mass index (BMI), current smoking, hypertension and diabetes mellitus. In addition, a small portion of the participants were treated with lipid-lowering drugs (n = 126) and, of these, a few were on statin treatment (n = 72); 152 study participants reported treatment with immunomodulating drugs such as azathioprine, methotrexate, ciclosporin, and 31 participants reported treatment with high-dose aspirin (>250 mg). Both treatment moieties were added to the multivariable model.
In the time to event analysis, the difference between groups was expressed in years and 95% CI and estimated by quantile regression for censored data, implemented using Laplace regression.9 The regression was calculated in a crude model and adjusted for the above-mentioned CV risk factors. The analysis was stratified by LDL cholesterol levels. The quantile regression model estimates any given percentile of survival/failure as a function of chosen covariates. For example, the 15th percentile of failure in this context would be the time by which 15% of the study population has had the event. The regression was calculated in a crude model and adjusted for the above-mentioned CV risk factors.
In secondary analyses, we estimated the risk of CAD and ischaemic stroke as well as the time to CAD and ischaemic stroke in individuals with B/T ratio > median in the two LDL groups (≤4.0 and >4.0). Ischaemic stroke incident cases were excluded when analysing the CAD risk associated with increasing B/T ratio. Likewise, CAD incident cases were excluded when analysing the risk of incident ischaemic stroke.
Given the interplay between LDL and inflammation in the progression of atherosclerosis we have estimated the effect of the interaction between the B/T ratio and LDL levels on the risk of future CVEs using a biological interaction model on the additive scale.10
To estimate the incremental discriminatory value of the B/T ratio in individuals with LDL ≤ 4.0 and >4.0, we calculated the area under the receiver-operating characteristics curve (AUC). We tested if the addition of the B/T ratio to a risk model, including the validated FRS and IL6, improved discrimination of cases and non-cases in the two LDL groups. FRS and IL6 were included as continuous variables while B/T ratio was dichotomised at the median.
Stata Statistical Software: Release 14. College Station, Texas, USA: StataCorp LP was used to perform all analyses.
Results
The clinical characteristics of the study population are described in Table 1. Until 31 December 2014, constituting a median follow-up of 16.5 (interquartile range 15.9–16.9) years, 525 incident CVEs had occurred i.e. approximately 15% of the study population had had a CVE.
Clinical characteristics of the study population at baseline.
| Incident cases (n = 525) | Reference group (n = 3099) | |
|---|---|---|
| Female/male (%) | 36/64 | 56/44 |
| Systolic blood pressure (mm Hg) | 144.5 (132–159.5) | 134.5 (121–150) |
| Diastolic blood pressure (mm Hg) | 87.5 (81.5–94.5) | 82.5 (76–90) |
| BMI (kg/m2) | 27.0 (24.5–29.8) | 26.1 (23.7–28.8) |
| Cardiovascular risk factors (%) | ||
| Hypertension | 20.2 | 15.0 |
| Diabetes mellitus | 6.3 | 2.3 |
| Hypercholesterolaemia | 4.2 | 3.4 |
| Smoking | 29.2 | 19.8 |
| Biochemical measurements (mmol/l) | ||
| Glucose | 5.3 (4.9–5.9) | 5.2 (4.8–5.6) |
| LDL | 4.0 (3.4–4.6) | 3.8 (3.2–4.5) |
| Binary/ternary complex ratio | 1.60 (1.56–1.63) | 1.58 (1.55–1.61) |
| Incident cases (n = 525) | Reference group (n = 3099) | |
|---|---|---|
| Female/male (%) | 36/64 | 56/44 |
| Systolic blood pressure (mm Hg) | 144.5 (132–159.5) | 134.5 (121–150) |
| Diastolic blood pressure (mm Hg) | 87.5 (81.5–94.5) | 82.5 (76–90) |
| BMI (kg/m2) | 27.0 (24.5–29.8) | 26.1 (23.7–28.8) |
| Cardiovascular risk factors (%) | ||
| Hypertension | 20.2 | 15.0 |
| Diabetes mellitus | 6.3 | 2.3 |
| Hypercholesterolaemia | 4.2 | 3.4 |
| Smoking | 29.2 | 19.8 |
| Biochemical measurements (mmol/l) | ||
| Glucose | 5.3 (4.9–5.9) | 5.2 (4.8–5.6) |
| LDL | 4.0 (3.4–4.6) | 3.8 (3.2–4.5) |
| Binary/ternary complex ratio | 1.60 (1.56–1.63) | 1.58 (1.55–1.61) |
BMI: body mass index; LDL: low-density lipoprotein.
Continuous data are presented as median (interquartile range) and categorical data as percentages. Missing values (cases/referents): systolic and diastolic blood pressure n = 3 (525/3096), LDL n = 45 (511/3068) and smoking n = 44 (520/3060).
Clinical characteristics of the study population at baseline.
| Incident cases (n = 525) | Reference group (n = 3099) | |
|---|---|---|
| Female/male (%) | 36/64 | 56/44 |
| Systolic blood pressure (mm Hg) | 144.5 (132–159.5) | 134.5 (121–150) |
| Diastolic blood pressure (mm Hg) | 87.5 (81.5–94.5) | 82.5 (76–90) |
| BMI (kg/m2) | 27.0 (24.5–29.8) | 26.1 (23.7–28.8) |
| Cardiovascular risk factors (%) | ||
| Hypertension | 20.2 | 15.0 |
| Diabetes mellitus | 6.3 | 2.3 |
| Hypercholesterolaemia | 4.2 | 3.4 |
| Smoking | 29.2 | 19.8 |
| Biochemical measurements (mmol/l) | ||
| Glucose | 5.3 (4.9–5.9) | 5.2 (4.8–5.6) |
| LDL | 4.0 (3.4–4.6) | 3.8 (3.2–4.5) |
| Binary/ternary complex ratio | 1.60 (1.56–1.63) | 1.58 (1.55–1.61) |
| Incident cases (n = 525) | Reference group (n = 3099) | |
|---|---|---|
| Female/male (%) | 36/64 | 56/44 |
| Systolic blood pressure (mm Hg) | 144.5 (132–159.5) | 134.5 (121–150) |
| Diastolic blood pressure (mm Hg) | 87.5 (81.5–94.5) | 82.5 (76–90) |
| BMI (kg/m2) | 27.0 (24.5–29.8) | 26.1 (23.7–28.8) |
| Cardiovascular risk factors (%) | ||
| Hypertension | 20.2 | 15.0 |
| Diabetes mellitus | 6.3 | 2.3 |
| Hypercholesterolaemia | 4.2 | 3.4 |
| Smoking | 29.2 | 19.8 |
| Biochemical measurements (mmol/l) | ||
| Glucose | 5.3 (4.9–5.9) | 5.2 (4.8–5.6) |
| LDL | 4.0 (3.4–4.6) | 3.8 (3.2–4.5) |
| Binary/ternary complex ratio | 1.60 (1.56–1.63) | 1.58 (1.55–1.61) |
BMI: body mass index; LDL: low-density lipoprotein.
Continuous data are presented as median (interquartile range) and categorical data as percentages. Missing values (cases/referents): systolic and diastolic blood pressure n = 3 (525/3096), LDL n = 45 (511/3068) and smoking n = 44 (520/3060).
Risk of CVE, CAD and ischaemic stroke associated with the B/T ratio in subjects with LDL cholesterol levels below and above 4.0 mmol/l
Table 2 summarises the risk of CVE and the risk of CAD and ischaemic stroke associated with B/T ratio>median stratified by LDL cholesterol levels.
Risk of cardiovascular events (CVEs), coronary artery disease (CAD) and ischaemic stroke associated with the B/T ratio > median and stratified by low-density lipoprotein (LDL) cholesterol levels.
| Type of event | Crude | Adjusted | ||||
|---|---|---|---|---|---|---|
| HR (95% CI) | p | n (case/ref) | HR (95% CI) | p | n (case/ref) | |
| All CVEs (n = 525) | ||||||
| LDL ≤ 4.0 | 1.79 (1.39–2.29) | <0.001 | 259/1871 | 1.59 (1.24–2.05) | <0.001 | 257/1850 |
| LDL > 4.0 | 1.45 (1.13–1.87) | 0.004 | 252/1197 | 1.29 (1.00–1.67) | 0.049 | 249/1179 |
| CAD (n = 361) | ||||||
| LDL ≤ 4.0 | 1.79 (1.32–2.44) | <0.001 | 170/1871 | 1.52 (1.11–2.08) | 0.009 | 168/1850 |
| LDL > 4.0 | 1.44 (1.07–1.94) | 0.015 | 184/1197 | 1.28 (0.95–1.73) | 0.10 | 182/1179 |
| Ischaemic stroke (n = 164) | ||||||
| LDL ≤ 4.0 | 1.86 (1.21–2.85) | 0.004 | 89/1871 | 1.84 (1.19–2.83) | 0.006 | 89/1850 |
| LDL > 4.0 | 1.58 (0.97–2.57) | 0.07 | 68/1197 | 1.41 (0.86–2.31) | 0.18 | 67/1179 |
| Type of event | Crude | Adjusted | ||||
|---|---|---|---|---|---|---|
| HR (95% CI) | p | n (case/ref) | HR (95% CI) | p | n (case/ref) | |
| All CVEs (n = 525) | ||||||
| LDL ≤ 4.0 | 1.79 (1.39–2.29) | <0.001 | 259/1871 | 1.59 (1.24–2.05) | <0.001 | 257/1850 |
| LDL > 4.0 | 1.45 (1.13–1.87) | 0.004 | 252/1197 | 1.29 (1.00–1.67) | 0.049 | 249/1179 |
| CAD (n = 361) | ||||||
| LDL ≤ 4.0 | 1.79 (1.32–2.44) | <0.001 | 170/1871 | 1.52 (1.11–2.08) | 0.009 | 168/1850 |
| LDL > 4.0 | 1.44 (1.07–1.94) | 0.015 | 184/1197 | 1.28 (0.95–1.73) | 0.10 | 182/1179 |
| Ischaemic stroke (n = 164) | ||||||
| LDL ≤ 4.0 | 1.86 (1.21–2.85) | 0.004 | 89/1871 | 1.84 (1.19–2.83) | 0.006 | 89/1850 |
| LDL > 4.0 | 1.58 (0.97–2.57) | 0.07 | 68/1197 | 1.41 (0.86–2.31) | 0.18 | 67/1179 |
BMI: body mass index; CI: confidence interval; HR: hazard ratio.
The risk of CVEs, CAD and ischaemic stroke associated with the B/T ratio was estimated using Cox regression and expressed as HR with a 95% CI. Ischaemic stroke incident cases were excluded when analysing the CAD risk. Likewise, CAD incident cases were excluded when analysing the risk of ischaemic stroke. The multivariate analysis was adjusted for sex, BMI, hypertension, diabetes mellitus and smoking.
Risk of cardiovascular events (CVEs), coronary artery disease (CAD) and ischaemic stroke associated with the B/T ratio > median and stratified by low-density lipoprotein (LDL) cholesterol levels.
| Type of event | Crude | Adjusted | ||||
|---|---|---|---|---|---|---|
| HR (95% CI) | p | n (case/ref) | HR (95% CI) | p | n (case/ref) | |
| All CVEs (n = 525) | ||||||
| LDL ≤ 4.0 | 1.79 (1.39–2.29) | <0.001 | 259/1871 | 1.59 (1.24–2.05) | <0.001 | 257/1850 |
| LDL > 4.0 | 1.45 (1.13–1.87) | 0.004 | 252/1197 | 1.29 (1.00–1.67) | 0.049 | 249/1179 |
| CAD (n = 361) | ||||||
| LDL ≤ 4.0 | 1.79 (1.32–2.44) | <0.001 | 170/1871 | 1.52 (1.11–2.08) | 0.009 | 168/1850 |
| LDL > 4.0 | 1.44 (1.07–1.94) | 0.015 | 184/1197 | 1.28 (0.95–1.73) | 0.10 | 182/1179 |
| Ischaemic stroke (n = 164) | ||||||
| LDL ≤ 4.0 | 1.86 (1.21–2.85) | 0.004 | 89/1871 | 1.84 (1.19–2.83) | 0.006 | 89/1850 |
| LDL > 4.0 | 1.58 (0.97–2.57) | 0.07 | 68/1197 | 1.41 (0.86–2.31) | 0.18 | 67/1179 |
| Type of event | Crude | Adjusted | ||||
|---|---|---|---|---|---|---|
| HR (95% CI) | p | n (case/ref) | HR (95% CI) | p | n (case/ref) | |
| All CVEs (n = 525) | ||||||
| LDL ≤ 4.0 | 1.79 (1.39–2.29) | <0.001 | 259/1871 | 1.59 (1.24–2.05) | <0.001 | 257/1850 |
| LDL > 4.0 | 1.45 (1.13–1.87) | 0.004 | 252/1197 | 1.29 (1.00–1.67) | 0.049 | 249/1179 |
| CAD (n = 361) | ||||||
| LDL ≤ 4.0 | 1.79 (1.32–2.44) | <0.001 | 170/1871 | 1.52 (1.11–2.08) | 0.009 | 168/1850 |
| LDL > 4.0 | 1.44 (1.07–1.94) | 0.015 | 184/1197 | 1.28 (0.95–1.73) | 0.10 | 182/1179 |
| Ischaemic stroke (n = 164) | ||||||
| LDL ≤ 4.0 | 1.86 (1.21–2.85) | 0.004 | 89/1871 | 1.84 (1.19–2.83) | 0.006 | 89/1850 |
| LDL > 4.0 | 1.58 (0.97–2.57) | 0.07 | 68/1197 | 1.41 (0.86–2.31) | 0.18 | 67/1179 |
BMI: body mass index; CI: confidence interval; HR: hazard ratio.
The risk of CVEs, CAD and ischaemic stroke associated with the B/T ratio was estimated using Cox regression and expressed as HR with a 95% CI. Ischaemic stroke incident cases were excluded when analysing the CAD risk. Likewise, CAD incident cases were excluded when analysing the risk of ischaemic stroke. The multivariate analysis was adjusted for sex, BMI, hypertension, diabetes mellitus and smoking.
In the full cohort, a high B/T ratio increased the risk of CVE to a higher extent in subjects with LDL ≤ 4.0 as compared to LDL > 4.0 (HR 1.59 vs 1.29). Incident CVEs (n = 525) were divided into fatal/non-fatal CAD (n = 361) and ischaemic stroke (n = 164). In the group with LDL ≤ 4.0, a high B/T ratio was associated with an increased risk of having future CAD (HR 1.52; 95% CI 1.11–2.08) and ischaemic stroke (HR 1.84; 95% CI 1.19–2.83). No association was observed for the B/T ratio either with CAD or with ischaemic stroke in the group with LDL > 4.0 (Table 2). When adjusting for lipid-lowering and immunomodulatory treatment the CVE risk estimates did not change (data not shown).
We then tested if the B/T ratio could predict the risk of CVE in individuals with LDL < 3.0 (n = 532) and LDL ≥ 3.0–≤4.0 (n = 1598). Although the number of cases in the LDL < 3.0 group is low (n = 61) and this cut-off has a low discriminatory value in our population (Supplementary Material Figure I), we could observe a similar trend with an increased CVE risk in the presence of B/T ratio > median. Similarly, an increased CVE risk was observed in LDL ≥ 3.0–≤4.0 group (Supplementary Material Table II).
In addition, we estimated the association of the B/T ratio with the risk of CVE in individuals defined as at low, intermediate and high risk according to the FRS. As shown in Supplementary Material Table III, the B/T ratio was associated with an increased 10-year risk of CVE in individuals classified as at low and intermediate risk by the FRS, but did not increase the CVE risk in those classified as high risk.
Differences in time to CVE, CAD and ischaemic stroke associated with the B/T ratio according to LDL cholesterol levels
In subjects with LDL ≤ 4.0, a B/T ratio>median was associated with earlier onset of CVE (Figure 1, left panel). After approximately 16 years follow-up, when 10% of the study population had had a CVE, events occurred nearly four years earlier in subjects with a B/T ratio >median after adjustment for confounders. In the group with LDL > 4.0 the difference in time to CVE was smaller and the association tended to disappear after adjustment (Figure 1, right panel).
Difference in years (95% confidence interval (CI)) to cardiovascular event (CVE) with the B/T ratio >median in subjects with low-density lipoprotein (LDL) ≤ 4.0 and >4.0 mmol/l. The reference group is subjects with B/T ratio ≤median in both analyses. Case/referent numbers in the crude analysis: 259/1871 (14% cases) in LDL ≤ 4.0 mmol/l, and 252/1197 (21% cases) in LDL cholesterol>4.0 mmol/l group. LDL missing data, n = 45. NA: not applicable.
In secondary analyses, we analysed the time to CAD and ischaemic stroke according to the B/T ratio in individuals with LDL ≤ 4.0 and >4.0. The differences in time to CAD were similar in both LDL groups, with 3–4 year earlier events in those with a high B/T ratio albeit the association was more evident in the LDL ≤4.0 group (p < 0.001) compared to the group with LDL > 4.0 (p = 0.03) (Supplementary Material Figure II). Ischaemic stroke occurred earlier only in the group with LDL ≤ 4.0 and a high B/T ratio (3.8 years, p = 0.02), (Supplementary Material Figure III).
Analysis of the interaction between LDL cholesterol levels and B/T ratio
The analysis of interaction between the B/T ratio and LDL cholesterol on the CVE risk is reported in Supplementary Material Table IV and summarised in Figure 2.
The risk of cardiovascular event (CVE) according to B/T ratio ≤median vs >median in relation to low-density lipoprotein (LDL)≤ vs >4.0 mmol/l presented as hazard ratio (HR) calculated by Cox proportional hazards. All values of p < 0.001. Missing values: LDL n = 45 and smoking n = 44.
Compared to the reference group, i.e. individuals with LDL ≤ 4.0 and B/T ratio ≤median, the presence of either a high B/T ratio but LDL cholesterol ≤4.0 or a low B/T ratio together with LDL > 4.0 mmol/l were associated with an increased risk of CVE (HR 1.58; 95% CI 1.23 -2.04 and HR 1.68; 95% CI 1.27 -2.21, respectively). The highest risk of CVE was found in subjects with a B/T ratio >median and LDL > 4.0 mmol/l (HR 2.17; 95% CI 1.68–2.80).
Incremental discriminatory value of the B/T ratio in subjects with LDL ≤ 4.0
We compared the AUC for the binary B/T ratio in addition to the continuous FRS and IL6 with the AUC obtained with only FRS and IL6 in individuals with LDL below and above 4.0. The AUC for the B/T ratio + FRS + IL6 as a predictive marker for CVE in all subjects was 0.67. When stratifying by LDL cholesterol, the value was 0.68 compared to 0.65 in the presence of LDL ≤ 4.0 and >4.0, respectively (Supplementary Material Table V).
Discussion
Our findings indicate that IL6 trans-signalling, estimated by the B/T ratio, is a novel inflammatory biomarker able to predict CVEs in individuals currently defined as low-intermediate CV risk. This finding is of importance in primary CV prevention as intermediate-high risk individuals using currently available risk scores often are misclassified as having a low risk of CVE.1,2
The 4.0 mmol/l LDL cholesterol cut-off used in the present study is based on the value with the highest AUC in this cohort. This cut-off was also adequate in relation to the LDL cholesterol levels in both the FRS and the European Society of Cardiology (ESC)/European Atherosclerosis Society (EAS) guidelines for the management of dyslipidaemias8,11 and in accordance with the results from a Swedish cohort study.12 The B/T ratio was associated with an increased CVE risk in the group with low LDL cholesterol levels, both in the main analysis (LDL ≤ 4.0 mmol/l) and in the secondary analysis where individuals with LDL ≤ 4.0 mmol/l were further divided in two groups with a 3.0 mmol/l cut-off. Similarly, we found that the B/T ratio was associated with an increased CVE risk in individuals otherwise classified as at low and intermediate CV risk according to the FRS.
Taken together, our results strongly suggest that IL6 trans-signalling, estimated by the B/T ratio, is a promising biomarker to improve CV risk classification in middle-aged men and women currently classified as at low-intermediate risk. On the other hand, no clear advantage of using additional biomarkers targeting inflammation was observed in the assessment of CVE risk in those classified as high-risk according to the FRS and in the presence of LDL cholesterol >4.0 mmol/l.
The finding that the relative excess of the active binary IL6:sIL6R complex is associated with CVEs is in line with clinical studies demonstrating the detrimental effects of IL6 trans-signalling in atherosclerosis-related diseases4,13–16 as well as with the observation that treatment with sgp130, the antagonist of the IL6:sIL6R complex promotes regression of atherosclerotic plaques in an experimental animal model of atherosclerosis17 and improves survival in a model of abdominal aortic aneurysm.18 Moreover, as compared to other immunomodulating agents currently used in clinical practice, selective inhibition of IL6 transsignalling is likely not to induce the observed adverse effects such as neutropaenia, fatal infections19 and disturbed lipid profile20 all of which in themselves are associated with increased CV risk.21
Recently, increased CV risk in individuals with low-normal LDL cholesterol levels has been described in both primary and secondary prevention settings.1,22 In our analysis, IL6 trans-signalling was associated with earlier CVEs in the group with LDL cholesterol ≤4.0 mmol/l, an association prominently observed in individuals suffering a stroke. With the heterogeneity in the pathophysiology of ischaemic stroke and the relatively few cases, this finding is hard to interpret. While CAD stems from atherosclerosis in the coronary arteries, ischaemic stroke has multiple causes with the most common being cardiac embolism due to atrial fibrillation, atherosclerosis of the large vessels and lacunar infarctions from small vessel disease. IL6 has been linked to atherosclerosis in the carotid artery23 and atrial fibrillation24 while its association to cerebral small vessel disease is still not clear.25 It is thus likely that the association pattern would differ if ischaemic stroke cases were stratified according to pathophysiology. This analysis was, however, not possible to perform in our study due to lack of data on the underlying pathophysiology of stroke. Notwithstanding, the results harmonise with findings from experimental animal research indicating that IL6 trans-signalling elicits neuronal degeneration in the ischaemic setting in the brain.26
To investigate the combined exposures of the cholesterol-driven and inflammatory risk, we explored the effect of the interaction of LDL cholesterol levels and IL6 trans-signalling on the risk of future CVEs. Our results indicate that the increase in risk due to the inflammatory pathway driven by the IL6 trans-signalling equals that observed in the presence of high LDL cholesterol. This suggests that the gain to be achieved from primary preventive treatment of the increased inflammatory risk could be comparable to that obtained treating increased LDL cholesterol levels, as it has been suggested in secondary prevention studies comparing the decreased risk of recurrent events in the Further Cardiovascular Outcomes Research with PCSK9 Inhibition in Subjects with Elevated Risk (FOURIER) trial.27 The formal statistical interaction test did not attain statistical significance. This may indicate that inflammation and metabolic risk are two independent but related pathways affecting the CV risk. On the other side, hypercholesterolaemia and the inflammatory pathways are closely entwined28 and the lack of formal interaction may also depend on insufficient power in the interaction analysis. Larger population samples are needed to confirm this finding.
The analysis of the discriminatory value of the IL6 trans-signalling suggests that the B/T ratio might be used to better classify the risk of CVEs in individuals at low-intermediate risk amenable for primary preventive treatment. Proper reclassification of the individual CV risk in subjects with a high inflammatory risk paired with a low metabolic risk is pivotal to improve CV prevention, in line with the results of the Justification for the Use of Statin in Prevention: An Intervention Trial Evaluating Rosuvastatin (JUPITER) study in primary prevention.29 In this analysis we use IL6 as a comparison instead of CRP since IL6 trans-signalling is not mirrored by CRP, a product of classical signalling. Moreover, IL6 is a known inflammatory marker strongly associated with the risk of CVE30 albeit not in clinical practice as CRP.
Our study has several limitations. The most important one being that we only have serum samples from baseline and thus cannot rule out that there is variation in the levels of the IL6 and its receptors over time and the serum samples have been frozen for many years. In both cases this would lead to non-differential misclassification. Furthermore, the analyses have not been validated in an external cohort.
The greatest strength of this study is that it is a large prospective cohort study with a 100% follow-up of the outcome. Moreover, we incorporate both the active and the neutralised IL6 complex in one marker.
In conclusion, IL6 trans-signalling is associated with earlier CVEs, pre-eminently stroke, in individuals currently classified as at low-intermediate CV risk. Apart from its relevance as biomarker of CVE risk, targeting the IL6 trans-signalling might represent a novel therapeutic moiety.
Author contribution
LZ, BG, UdF and HW contributed to the design of the work. LZ and BG contributed to the data acquisition. LZ, BG and PF contributed to data analysis and interpretation of the results. LZ and BG drafted the manuscript and UdF, PF and HW critically revised the manuscript. All gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Stockholm County Council ALF project and Strategic research in Epidemiology at Karolinska Institutet (to BG).
Comments