Of the 6228 patients with aSAH admitted to Sichuan University West China Hospital, during the 10 years, a total of 1131 aSAH patients with IL-6 records were included in this study (eFigure1). Baseline characteristics stratified by the long-term mortality were shown in Table 1. At the longest follow-up,73.6% patients (n=833) survived and 26.4% patients (n=298) dead. In demographics, deaths were older (p<0.001) and had higher SBP (p=0.03). In medical history, the dead had a higher proportion of COPD (p<0.001). As expected, there were significantly more patients with clinically and radiologically severe SAH among the deaths (p<0.001 and p=0.001, respectively). More patients of death performed external ventricular drain at admission (p<0.001). However, there was no significant difference in the location and size of aneurysms (p=0.85 and p=0.11, respectively). Patients who received treatment for aneurysms were more likely to survive (p<0.001). Higher blood glucose was found in deaths (p<0.001). The occurrence of in-hospital infection in deaths was 80.2% (n=239). The difference was statistically significant compared with survivors (p<0.001).
The time course of IL-6 levels during hospitalization was illustrated in Figure 1a. Considering the great individual variation, IL-6 levels were presented after logarithmic transformation. In our series, initial serum IL-6 values were elevated in all. The IL-6 values ranged between 1.50 and 4522.00 pg/ml. The median IL-6 level was 39.37 pg/ml and the Inter-Quartile Range was 67.97 pg/ml. During the further clinical course, a substantial rise in IL-6 concentrations was observed on day 2 after SAH, and IL-6 peaks were found on days 2-5, day 13-14, day17, day26, and day29. For subgroup analysis, we compared daily IL-6 levels between low Hunt & Hess grade (I-III) and high Hunt & Hess grade (IV-V), cases with hydrocephalus and without hydrocephalus, and survivors and deaths at the longest follow-up (Figure 1b-d). Statistics differences were found on days 0-2 between different Hunt & Hess grade groups, on days 0,2, 5, 17, 24, 28, and 30 between groups with hydrocephalus or not, and on days 0-2, 5-6, 8-9, 11, 13, 17-18, 20 and 26 between different long-term outcome groups.
IL-6 peak during hospitalization was associated with long-term mortality (HR 1.32 95% CI 1.21-1.44). After adjusting all variates influencing the outcome, including age, smoking, alcohol abuse, SBP, CRF, COPD, Fisher grade, Hunt & Hess grade, external ventricular drain, treatments, blood glucose, and in-hospital infection (Table 2), IL-6 peak still predicted long-term mortality (adjusted HR 1.23 95% CI 1.11-1.36, p<0.001). In the sensitivity analysis of survivors at discharge (Table 4), the finding remained robust even after adjustment (adjusted HR 1.17 95% CI 1.05-1.31, p=0.004).
The correlation between IL-6 peaks at 5 different phases (T1-5) and long-term mortality were furtherly studied (Table 3). In univariate Cox regression analysis, log IL-6 peaks of T1-5 were all significantly associated with long-term mortality (T1 HR 1.76 95% CI 1.27-2.44; T2 HR 1.47 95% CI 1.21-1.77; T3 HR 1.26 95% CI 1.09-1.45; T4 HR 1.27 95% CI 1.12-1.45; T5 HR 1.55 95% CI 1.27-1.88). After adjusting for all covariates in the multivariable Cox regression analysis, findings remained robust at T3-5 (T3 adjusted HR 1.19 95% CI 1.01-1.39, p=0.03; T4 adjusted HR 1.25 95% CI 1.09-1.44, p=0.002; T5 adjusted HR 1.53 95% CI 1.25-1.87, p<0.001). After dichotomized into two groups using the best cut-off values, IL-6 peaks of all phases were significantly associated with long-term mortality after adjustment (T1 adjusted HR 2.84 95% CI 1.06-7.61, p=0.04; T2 adjusted HR 1.76 95% CI 1.06-2.92, p=0.03; T3 adjusted HR 1.81 95% CI 1.27-2.58, p<0.001; T4 adjusted HR 2.43 95% CI 1.74-3.39, p<0.001; T5 adjusted HR 3.47 95% CI 1.99-6.03, p<0.001). In the propensity score-matched analysis, findings remained significant at T2,4-5 (T2 PSM HR 1.84 95% CI 1.02-3.31, p=0.04; T4 PSM HR 2.22 95% CI 1.46-3.38, p<0.001; T5 PSM HR 2.95 95% CI 1.64-5.29, p<0.001).
Additionally, we explored the association between IL-6 peak during hospitalization and in-hospital neurological complications (eTable 3) or short-term mortality (Table 4). Patients with higher IL-6 peaks were more likely to develop hydrocephalus, and rebleeding (OR 1.18 95% CI 1.06-1.30; OR 1.21 95% CI 1.06-1.38, respectively). After adjustment of covariates, there was no significant association between IL-6 peak and neurological complications. In the propensity score-matched analysis, findings only remained prominent in hydrocephalus (PSM OR 1.86 95% CI 1.26-2.74, p=0.002). Besides, IL-6 peak well predicted short-term mortality within all patients or survivors at discharge (adjusted HR 1.36 95% CI 1.22-1.52, p<0.001; adjusted HR 1.32 95% CI 1.17-1.49, p<0.001, respectively). Findings maintained consistent in the propensity score-matched analysis (IL-6 >103.00 pg/ml, PSM HR 2.54 95% CI 1.83-3.53, p<0.001; IL-6 >102.85 pg/ml, PSM HR 1.84 95% CI 1.29-2.62, p=0.001, respectively).
Kaplan-Meier analysis showed that death during follow-up was more frequent in patients with higher IL-6 levels (p < 0.001; Figure 2a). It also illustrated that patients with IL-6 levels stratified by IL-6 levels had significantly worse long-term survival except for deaths at discharge (p < 0.001; Figure 2b).
We further assessed interactions by variables on serum IL-6 (Figure 3). The interaction was present regarding the aneurysm size and in-hospital infection (p=0.05, p=0.02, respectively). There was no significant effect modification of the association between D-dimer level and mortality on the basis of age (p = 0.39), aneurysm location (p = 0.11), Fisher grade (p=0.44), Hunt & Hess grade (p=0.36), or aneurysm treatment (p = 0.10).
Finally, ROC analysis was performed to determine the ability of IL-6 to distinguish between long-term survivors and deaths (eFigure 2). We compared the ROC curves for IL-6, Fisher grade, and the multivariable IL-6 model (adding all variables significant in multivariable Cox regression analysis). IL-6 showed better ability to predict long-term mortality than Fisher grade (AUC IL-6=0.63, 95% CI 0.59-0.67 vs AUC Fisher grade=0.56, 95% CI 0.53-0.59, p= 0.004). As expected, multivariable IL-6 model improved the performance of IL-6 (AUC multivariable IL-6 model=0.74, 95% CI 0.71-0.67 vs AUC Fisher grade=0.74, 95% CI 0.71-0.77, p< 0.001).