Circulating cytokine levels and 5-year vascular recurrence after stroke: A multicenter prospective cohort study

Abstract

Background and objectives: Anti-inflammatory therapies are tested in randomized trials for secondary stroke prevention. Detecting inflammatory biomarkers that predict vascular recurrence could optimize patient selection for these trials.

Methods: In a multicenter prospective cohort study, we measured plasma levels of 22 inflammatory cytokines in 486 acute stroke patients (474 ischemic strokes and 12 intracerebral hemorrhages; median age 68 years, 34% female, median 3 days post-stroke onset). Patients were followed for over 5 years through telephone and in-person interviews to record the occurrence of the following outcomes: (1) recurrent stroke or transient ischemic attack (TIA; primary outcome); (2) a composite of recurrent vascular events (stroke, TIA, acute coronary syndrome, hospital admission due to heart failure, and death; secondary outcome). Associations between cytokine levels and these outcomes were analyzed using Cox proportional hazards models adjusted for demographic and vascular risk factors.

Results: During the 5-year follow-up period, 59 patients (12.1%) experienced recurrent stroke or TIA, and 118 (24.3%) experienced recurrent vascular events. After adjustments for demographic and vascular risk factors, and correction for multiple comparisons, higher plasma levels of CD62E (adjusted Hazard Ratio (aHR)/SD increment: 1.63, 95%CI 1.22-2.20) and MIF (aHR: 1.56, 95%CI 1.18-2.06) in the acute phase after stroke were statistically significantly associated with increased risk of recurrent stroke or TIA. The associations followed a dose-response pattern across quartiles of CD62E and MIF levels. Adding baseline CD62E and MIF levels to models including age, sex, vascular risk factors, and baseline C-reactive protein (CRP) levels led to significant improvements in the prediction of 5-year risk of recurrent stroke or TIA (ΔC-index 0.030-0.050).

Conclusion: Among stroke patients, higher baseline levels of CD62E and MIF improved prediction of 5-year risk of recurrent stroke or TIA on top of vascular risk factors and CRP levels. Whether assessment of these cytokines could improve patient selection for secondary prevention trials of anti-inflammatory treatments, should be explored in future studies.

Keywords: Stroke; cytokines; inflammation; recurrence; risk prediction.

Graphical abstract

Figure 1.

Flowchart of the participants inclusion and exclusion criteria.

Figure 2.

Correlations across cytokine levels and the effects of demographic, clinical, and technical factors on cytokine levels. (a) Heatmap illustrating the Spearman correlations across the plasma levels of 22 cytokines in the DEMDAS cohort (N = 486). The color scale ranges from blue (negative correlation) to red (positive correlation). Significance is indicated by *p < 0.05, **p < 0.01, and ***p < 0.001, representing FDR – corrected p values for multiple comparisons. (b) Heatmap illustrating the effect estimates of demographic, clinical, and technical factors on ln-transformed standardized cytokine levels, as derived from univariable linear regression analyses. Red and blue colors represent positive and negative estimates, respectively. Significance is indicated by *p < 0.05, **p < 0.01, and ***p < 0.001 (all FDR-adjusted). (c) Violin plot illustrating the difference in plasma CD62E levels between patients with and without diabetes mellitus. The box represents the IQR (25th–75th percentiles), with the line inside the box representing the median. (d) Scatter plot illustrating the positive correlation between plasma CD62E levels and BMI. (e) Box plot illustrating comparisons in the sex-specific distribution of cytokine levels across the DEMDAS cohort. Each box represents IQR, covering the 25th–75th percentiles of the data, with the line inside the box indicating the median value. The whiskers extend to the most extreme data points within 1.5 times the IQR. BMI: body mass index; FDR: false discovery rate; LDL: low density lipoprotein; NIHSS: National Institutes of Health Stroke Scale; IQR: interquartile range; FDR: false discovery rate.

Figure 3.

Associations of cytokine levels with recurrent stroke or TIA and recurrent vascular events (stroke, TIA, acute coronary syndrome, new onset of heart failure, or death) over the 5-year follow-up period. Forest plots of adjusted HR for the associations of circulating baseline cytokine levels (per SD increment in ln-transformed values) with (a) recurrent stroke or TIA (n = 59) and (b) the composite endpoint of recurrent vascular events (n = 118) over the 5-year follow-up period. The HR in the first model (red) are adjusted for age and sex, while those in the second model (yellow) are further adjusted for vascular risk factors, including hypertension, diabetes, current smoking, history of atrial fibrillation, anticoagulants, antihypertensive medications, antiplatelet agents, statins, and low-density lipoprotein levels. Vertical error bars indicate the 95%CI for each adjusted HR. Significance is indicated by *FDR adjusted p < 0.05, representing FDR-corrected p values for multiple comparisons. TIA: transient ischemic attack; HR: hazard ratio; CI: confidence interval; FDR: false discovery rate.

Figure 4.

Dose–response associations of baseline CD62E and MIF levels with 5-year risk of recurrent stroke or TIA. (a) Kaplan–Meier 5-year recurrent stroke- or TIA-free survival curves for DEMDAS participants across quartiles (Q1–Q4) of baseline circulating CD62E levels. (b) Forest plot of adjusted HR for recurrent stroke or TIA across quartiles of baseline circulating CD62E levels. (c) Kaplan–Meier 5-year recurrent stroke- or TIA-free survival curves for DEMDAS participants across quartiles (Q1–Q4) of baseline circulating MIF levels. (d) Forest plot of adjusted HR for recurrent stroke or TIA across quartiles of baseline circulating MIF levels. In panels (b) and (d), the HR in the first model (red) are adjusted for age and sex, while the HR in the second model (yellow) are further adjusted for vascular risk factors, including hypertension, diabetes, current smoking, history of atrial fibrillation, anticoagulants, antihypertensive medications, antiplatelet agents, statins, and low-density lipoprotein levels. Horizontal lines represent the 95%CI for each HR. Statistical significance is indicated by *p < 0.05, **p < 0.01. TIA: transient ischemic attack; HR: hazard ratio; CI: confidence interval.