Mengel et al. BMC Neurology (2019) 19:148 https://doi.org/10.1186/s12883-019-1375-6 RESEARCH ARTICLE Open Access Biomarkers of immune capacity, infection and inflammation are associated with poor outcome and mortality after stroke - the PREDICT study A. Mengel1,2*, L. Ulm1,3, B. Hotter1,4, H. Harms1, S. K. Piper5,6,7, U. Grittner5,6,7, J. Montaner8, C. Meisel9, A. Meisel1,4,9† and S. Hoffmann1,9† Abstract Background: Almost 40% of stroke patients have a poor outcome at 3 months after the index event. Predictors for stroke outcome in the early acute phase may help to tailor stroke treatment. Infection and inflammation are considered to influence stroke outcome. Methods: In a prospective multicenter study in Germany and Spain, including 486 patients with acute ischemic stroke, we used multivariable regression analysis to investigate the association of poor outcome with monocytic HLA-DR (mHLA-DR) expression, interleukin 6 (IL-6), interleukin 10 (IL-10), tumor necrosis factor alpha (TNF-alpha) and lipopolysaccharide-binding protein (LBP) as markers for immunodepression, inflammation and infection. Outcome was assessed at 3 months after stroke via a structured telephone interview using the modified Rankin Scale (mRS). Poor outcome was defined as a mRS score of 3 or higher which included death. Furthermore, a time-to-event analysis for death within 3 months was performed. Results: Three-month outcome data was available for 391 patients. Female sex, older age, diabetes mellitus, atrial fibrillation, stroke-associated pneumonia (SAP) and higher National Institute of Health Stroke Scale (NIHSS) score as well as lower mHLA-DR levels, higher IL-6 and LBP-levels at day 1 were associated with poor outcome at 3 months in bivariate analysis. Furthermore, multivariable analysis revealed that lower mHLA-DR expression was associated with poor outcome. Female sex, older age, atrial fibrillation, SAP, higher NIHSS score, lower mHLA-DR expression and higher IL-6 levels were associated with shorter survival time in bivariate analysis. In multivariable analysis, SAP and higher IL-6 levels on day 1 were associated with shorter survival time. Conclusions: SAP, lower mHLA-DR-expression and higher IL-6 levels on day one are associated with poor outcome and shorter survival time at 3 months after stroke onset. Trial registration: www.clinicaltrials.gov, NCT01079728, March 3, 2010. Keywords: Ischemic stroke, Biomarkers, Mortality, Outcome * Correspondence: annerose.mengel@med.uni-tuebingen.de † A. Meisel and S. Hoffmann contributed equally to this work. 1 Department of Neurology Berlin, Charité Universitätsmedizin Berlin, Charitéplatz 1, 10115 Berlin, Germany 2 Department of Neurology and Stroke, Universitätsklinik Tuebingen, Hoppe-Seyler-Str.3, 72076 Tuebingen, Germany Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Mengel et al. BMC Neurology (2019) 19:148 Page 2 of 10 Background study, the diagnosis was radiologically confirmed. Exclu- Many stroke patients experience severe consequences, sion criteria were preexisting dysphagia, mechanical venti- with more than 40% of patients in the population suffer- lation, antibiotic treatment or immunosuppressive therapy ing from disability, institutionalisation or even death at 3 within the last 4 weeks and clinical or para-clinical signs months after the index event [1]. An early risk assess- of infection at study inclusion. Overall, 486 patients were ment of stroke outcome could help to optimize stroke recruited between January 2010 and December 2012. care allocation in order to improve outcome [2], and to Blood samples were obtained within 36 h after symptom predict the patient’s individual prognosis. However, reli- onset (day 1) and repeated on day 2 to 4. With the focus ably predicting stroke outcome remains challenging. on early outcome we restricted the analysis to blood sam- Several attempts to identify characteristics associated ples taken on day 1 after stroke. with poor outcomes have been made. A consistent asso- ciation with poor outcome was found for the clinical fac- Clinical data tors age and stroke severity [3–5], as well as for post- Clinically we documented socio-demographics, neuro- stroke complications such as stroke-associated pneumo- logical deficits, comorbidities and stroke severity nia (SAP) [6]. Further interesting targets for investiga- (NIHSS) on admission, as well as complications, and tion have been blood-based biomarkers. It has been processes and outcomes during hospitalization [10]. We shown that several biomarkers are associated with poor used the patient characteristics age at cerebrovascular outcome [7]. However, the added value of biomarkers event and sex in this analysis. In addition, we included over clinical variables with regard to prognosis remains the following comorbidities: hypertension (any existing controversial [8, 9]. In relation to the patient’s living will, treatment; patient self-reported; or new diagnosis during biomarkers could help us to predict individual poor out- hospital admission); diabetes mellitus (any existing treat- comes as early as possible in order to prevent pain and ment; patient self-reported; or new diagnosis during hos- inefficient treatment. The PREDICT study was initially pital admission); previous stroke (neurological deficit designed to better understand the underlying patho- more than 24 h prior to current event); and atrial fibril- physiology leading to SAP. We were able to demonstrate lation (patient self-reported or documented by standard that a key biomarker of stroke-induced immunodepres- ECG or Holter monitor). We also included in the ana- sion syndrome (SIDS), mHLA-DR, was associated with lysis stroke severity on admission assessed by the SAP independently of dysphagia as the main risk for as- NIHSS. This was categorized based on commonly de- piration pneumonia [10]. By analysing the secondary end fined cut-off points [11] into: mild impairment (NIHSS< points of the PREDICT study, we wanted to evaluate 5), moderately severe impairment (NIHSS 5 to 15) and and quantify the association of biomarkers of SIDS with severe or very severe (NIHSS≥16) impairment [10]. poor outcome at 3 months after stroke which might Signs of pneumonia were clinically assessed on day 1 to serve to build more reliable prediction models for stroke 4 and on day 7 (or on the day of discharge in cases outcome in the future. where the hospital stay was less than 7 days). During hospitalisation, diagnosis of SAP was made by the study Methods or treating physician based on one or more of the fol- The methods used in this study (especially assessment of lowing: clinical symptoms, suggestive clinical examin- clinical and immunological data) have been described previ- ation, radiological findings, and pathogen detection of ously in the primary analysis of the PREDICT study pulmonary infection. (www.clinicaltrials.gov, NCT01079728, March 3, 2010) [10]. In short, the PREDICT study was a prospective, inter- Immunological data national, multicenter study with 11 participating study sites Blood samples were obtained as soon as possible, but no in Germany and Spain designed to better understand the longer than 36 h after symptom onset (day 1) and inde- underlying pathophysiology leading to SAP [10]. pendent of the patient’s fasting status. They were ana- The study was approved by the ethics committees of the lysed for immune (mHLA-DR, TNF-alpha) and infection Charité- Universitätsmedizin Berlin (EA1/216/09), and of (IL6, IL-10, LBP) parameters. For plasma collection, Hospital Vall d’Hebron Barcelona (No. 199 25/05/2012). samples were centrifuged (200 g, 4 °C, 10 min) and All patients or their legal representatives gave written in- stored at − 80 °C. Plasma concentrations of IL-6 were de- formed consent in accordance with the Declaration of termined with the IMMULITE™ semi-automatic chemi- Helsinki. The four inclusion criteria were: clinical diagno- luminescent immunoassay (Siemens Medical Solutions, sis of acute ischemic stroke (NIHSS≥1) with any Bad Nauheim, Germany). The detection limit for IL-6 localization (anterior and posterior circulation) and with was 2.0 pg/ml. The detection limit for IL-10 was 5.0 pg/ any recurrence rate (first-ever and recurrent stroke); ml. Based on visual binning, IL-10 was categorized into symptom onset ≤36 h, and age ≥ 18 years [10]. During the three categories: ≤5.0 pg/ml, 5.1–7.0 pg/ml and > 7.0 pg/ Mengel et al. BMC Neurology (2019) 19:148 Page 3 of 10 ml. Expression of mHLA-DR on monocytes was deter- given. Using the Bayesian approach, it was possible to ac- mined by flow cytometry using a highly standardized count for the complete separation of the data (Burn-in: quantitative assay as described earlier [12]. In short, whole 10.000, MCM sample size: 100.000, non-informative priors blood in a Vacutainer tube (Becton Dickinson Biosciences, for all other variables except pneumonia with mean: 0 and San Jose, CA, USA) containing EDTA was stained with variance: 10.000). 30 μl of monoclonal phycoerythrin-conjugated anti–HLA- In the multivariable logistic regression analysis with ran- DR and PerCP-Cy5.5–conjugated anti-CD14 antibodies dom intercept for centers we included all relevant vari- (Quantibrite; Becton Dickinson Biosciences) in the dark at ables (p < 0.2 in bivariate analysis, or clinically relevant room temperature for 30 min. Erythrocyte lysis was parameters). We additionally checked the association of achieved with 0.5 ml of lysing solution (Becton Dickinson logit (p = non-favorable outcome) with each variable using Biosciences) for another 30 min at room temperature. Fi- scatter plots to confirm whether a linear form was appro- nally, the cells were washed with 1 ml of fluorescence acti- priate in the model. vated cell sorting (FACS) buffer, and then analyzed on a Cox proportional hazards analysis and log-rank test FACSCalibur cytometer (Becton Dickinson Biosciences) (Mantel Cox) were used for survival analyses over the en- to assess HLA-DR surface expression [10]. tire follow-up period of 3 months. Inspection of the scaled Schoenfeld residuals and formal testing of the propor- Assessment of outcome tional hazards assumption revealed that there was no sub- At 3 months after stroke, an outcome assessment was stantial violation of the proportional hazards assumption. performed with each patient via a structured telephone For bivariate analysis, the survival function was esti- interview assessing the mRS [13]. In line with other pub- mated by the Kaplan-Meier (KM) method (product-limit lications [3, 8], we defined poor outcome a priori as estimator) and reported as cumulative proportion of mRS score of 3 or higher (i.e. patients are moderately to deaths at 90 days after stroke with confidence interval severely disabled and need assistance with Activities of and p-value of the log-rank test. Here non-categorized Daily Living, or have died). Furthermore, the cumulative continuous variables were split around their median. proportion of deaths at 90 days post-stroke was calcu- Finally, multivariable stratified Cox regression analysis lated. If the patient or the indicated contact person (usu- was performed with all relevant parameters. Stratifica- ally next of kin) were not traceable, vital status and, if tion was done with regard to centers. Hazard ratios available, the new address of the patient, were requested (HR) with 95% CI and R-square measures are presented at the registration office. for the final model. In order to estimate the added bene- fit of the used biomarkers for survival at 3 months after Statistics stroke, a ROC-Curve analysis was performed for the All analyses were restricted to patients without missing final Cox regression models with and without bio- values in the specific measures. The t-test was used to test markers (command in STATA: roccomp). differences among outcome groups in non-skewed continu- All tests were two-tailed, and statistical significance ous variables, and the Mann-Whitney test was used for was determined at an alpha level of 0.05. No adjustment skewed continuous variables. Pearson Chi-square test was for multiple testing was applied. Statistical analyses were used to test differences in proportions. Bivariate associa- performed with the SPSS 22.0 (IBM, Amonk, NY, USA), tions between dichotomous outcome categories (poor out- and Stata/IC 15.0. come at 3 month: mRS ≥ 3) and ordinal parameters were analyzed by the linear trend test. To estimate odds ratios (OR) and the corresponding 95% confidence intervals (CI) Results of the association between clinical variables and biomarkers Patient demographic and clinical characteristics with risk of poor outcome, logistic regression analysis was Overall, 486 patients were recruited to the PREDICT performed. As all patients with pneumonia were in the study [10]. Two patients dropped out (one screening fail- group of poor outcome, the problem of complete separ- ure, one withdrew informed consent). Functional outcome ation appeared. We applied a random effects logistic regres- or information on vital status at 3 months was available sion (random intercept for centers) with a Bayesian for 391 patients (80%), and only those were included in approach to account for heterogeneity between the partici- the functional outcome analysis. The mean age of our pating centers. An informative prior for pneumonia (in- sample was 70 years; 42.2% were female; and median formed by a firth regression model with penalized NIHSS at stroke onset was 4 points (inter-quartile range likelihood estimation, prior: mean: 1, variance: 2.5) was ap- IQR: 2–7). Table 1 shows the demographic and clinical plied (stata commands: bayes and melogit). Skewed vari- characteristics for patients stratified by good and poor ables were log-transformed or dichotomized before functional outcome at 3 months. For survival analysis, 437 entering the analysis. ORs and 95% credible intervals are patients were included, while 47 either had missing values Mengel et al. BMC Neurology (2019) 19:148 Page 4 of 10 in an independent variable or were censored before the expression as well as higher IL-6-levels on day 1 were asso- first event. ciated with shorter survival time. Similar to a lack of associ- ation with poor outcome, IL-10 levels and TNF-alpha levels Characteristics associated with poor 3-month outcome on day 1 were not associated with an increased risk of Bivariate analysis showed that patients of older age, fe- mortality. male gender and with higher NIHSS scores, SAP, dia- Multivariable Cox regression adjusted for clustering by betes or atrial fibrillation were more likely to have a centers revealed SAP and higher IL6-levels on day 1 as poor outcome (mRS ≥ 3). Furthermore, lower mHLA-DR being associated with shorter survival time in the 3 levels and higher IL-6 and LBP-levels at day 1 were asso- months after stroke onset (Table 4). Adding biomarkers ciated with poor outcome at 3 months. Levels of IL-10 on day 1 to the model significantly improved the model and TNF-alpha on day 1 had no influence on outcome. fit for survival in the 90 days of follow-up over a model Multivariable logistic regression with random intercept with clinical variables alone (Table 4, Likelihood Ratio- for centers showed that older age, diabetes mellitus, Test: p = 0.037). higher stroke severity (as measured by the NIHSS), SAP and lower mHLA-DR-expression were associated with Discussion poor outcome (Table 2). The prognostic value of blood biomarkers in patients with acute stroke could potentially be of great importance for Characteristics associated with mortality clinical routine. In our study, lower mHLA-DR levels, and In bivariate analyses (Table 3) female sex, older age, higher higher IL-6 and LBP levels on day 1 were associated with NIHSS score, SAP, atrial fibrillation, lower mHLA-DR poor outcome (mRS ≥ 3) at 3 months. Furthermore, we Table 1 Baseline characteristics stratified according to outcome after 3 monthse Characteristic All mRS 0–2 mRS 3–6 p-value n (%) 391 231 160 Female Sex, n (%) 165 (42.2) 83 (35.9) 82 (51.3) 0.003 a Age, yrs., mean (SD) 70 (12) 67 (12) 75 (10) < 0.001 b NIHSS categories, n (%)12 missing < 0.001 d 0–4 217 (57.3) 156 (69.6) 61 (39.4) 5–15 140 (36.9) 64 (28.6) 76 (49.0) 16+ 22 (5.8) 4 (1.8) 18 (11.6) Pneumonia, n (%) 20 (5.1) 0 (0) 20 (12.5) < 0.001 a Comorbidities, n (%) Hypertension (13 missing) 313 (82.8) 179 (79.9) 134 (87.0) 0.072 a Diabetes mellitus 98 (26.3) 45 (20.5) 53 (34.6) 0.002 a Atrial Fibrillation (18 missing) 113 (30.3) 53 (24.1) 60 (39.2) 0.002 a Hypercholesterolemia (18 missing) 192 (51.5) 112 (50.2) 80 (53.3) 0.556 a Previous stroke (18 missing) 75 (20.1) 41 (18.6) 34 (22.2) 0.395 a mHLA-DR day 1 (pg/ml) c Median (IQR) (4 missing) 16,422 [12268–21,990] 17,871 [13703–232,270] 15,062 [11340–19,333] < 0.001 IL-6 day 1 (pg/ml) c Median (IQR)(10 missing) 4.3 [2.3–8.5] 3.6 [2.0–6.8] 5.5 [3.4–11.4] < 0.001 IL-10 day 1 (pg/ml) Median (IQR) (12 missing) 5.0 [5.0–5.0] 5.0 [5.0–5.0] 5.0 [5.0–5.0] 0.113 c LBP day 1 (pg/ml) c Median (IQR) (11 missing) 7.5 [5.8–9.5] 7.0 [5.5–8.8] 8.2 [6.3–10.9] < 0.001 TNF-α day 1 (pg/ml) Median (IQR) (16 missing) 631 [459–887] 609 [452–868] 673 [462–898] 0.343 c a b c d Pearson Chi-square test, t-Test for unequal variance, Mann-Whitney Test, Chi-square test linear by linear e Analyses were restricted to patients without missing values for mRS after 3 months or death within the study period and the respective variable IQR = Q2-Q3 Mengel et al. BMC Neurology (2019) 19:148 Page 5 of 10 Table 2 Characteristics associated with poor outcome (mRS 3– of our knowledge, this is the first study that suggests an 6): OR 95% credible intervals of final multivariable mixed logistic association between IL-6 (baseline) and death in stroke regressiona (n = 357), using Bayesian approach with informative patients. Among lacunar stroke patients, IL-6 and TNF prior for pneumonia receptor concentrations were associated with risk of re- Characteristic OR 95% CI current vascular events and with the effect of antiplatelet Female sex 1.47 0.80–2.42 therapies [18]. In patients with atrial fibrillation, in- Age (yrs) 1.07 1.04–1.10 creased IL-6 seems to be related to higher risk for both Pneumonia 50.34 5.38–230.85 stroke and major bleeding, as well as to thromboembolic events and vascular death [19]. However, a meta-analysis NIHSS categories of 24 studies indicated that the translation of IL-6 into 0–4 1.00 clinical practice for prognosis of poor outcome after 5–15 3.01 1.94–5.02 stroke was unlikely to be of benefit [17]. In our study, 16+ 14.43 5.09–33.92 TNF-alpha was not substantially associated with out- Diabetes mellitus 3.21 1.75–5.32 come or mortality. Hypertension 0.90 0.37–1.86 In in vitro models, IL-10 is known to mediate neuro- protection, but clinically its role as prognostic marker Atrial fibrillation 0.72 0.40–1.19 for outcome is controversial [20, 21]. In the present Biomarker study we could not find a substantial association of IL- mHLA-DR day 1a (pg/ml) 0.42 0.28–0.59 10 expression with outcome and mortality. LBP day 1a (pg/ml) 1.64 0.89–2.85 In our study higher LBP levels were - in univariate IL6a (pg/ml) 1.19 0.88–1.59 analysis - associated with poor outcome. LBP is an acute IL10 (> 5) (pg/ml) 1.48 0.76–2.72 phase protein essential for the response to bacterial lipo- Analysis was restricted to patients without missing values for all variables polysaccharides, or endotoxin, in Gram-negative bacteria including mRS at 3 month (N = 357). Logistic mixed regression was performed [22]. LBP may rise as a response to subclinical infections using a Bayesian approach and an informative prior for pneumonia. (informed or even to intestinal translocations [23]. However, LBP by Firth logit model without random effects for centers: normal prior with mean 1 and variance 2.5), MCMC sample size: 100.000, Burn in: 10.000) may be an early post-stroke infection marker, especially a Skewed variables were log transformed before entering the analysis, OR and for SAP, although the majority of patients with SAP were 95% credible interval are given for the log transformed values diagnosed on day 3 and the association with death was still present after adjustment for SAP. Recently, smaller found that lower mHLA-DR levels were associated with studies which investigated LBP did not find a positive as- poor outcome at 3 months after adjustment for clinical sociation for infection or poor outcome in stroke pa- characteristics and other relevant biomarkers. ROC ana- tients [24, 25]. lysis of the Cox regression models showed that including Regarding clinical variables associated with poor out- the biomarkers on day 1 improved the area under the come and mortality, our findings are in line with previ- curve compared to a Cox regression model using clinical ous studies [26–29]. Older age, pneumonia and stroke data alone (age, sex, NIHSS> 16, SAP, diabetes mellitus, severity are known to be associated with poor outcome hypertension, atrial fibrillation). and mortality. In our study, female sex and presence of In line with previous studies [14, 15] we found that diabetes mellitus were also independently associated higher mHLA-DR was inversely associated with the odds with poor outcome. Our data concurs with previous of having poor outcome. Monocytes are of paramount studies reporting a less favourable outcome in women immunological importance due to their contribution compared to men [30–32]. Possible explanations for sex firstly to adaptive immunity as antigen-presenting cells, differences in stroke outcome include women being and secondly to innate immunity through the expression older [33], having a decreased likelihood of receiving rt- of pattern recognition receptors [14]. Impaired mono- PA treatment [34], and having a higher incidence of co- cyte function results in insufficient antigen-presentation morbidities (e.g. depression) [35]. Available research on and decreased expression of secreted or membrane- the association of diabetes mellitus with outcome is in- bound costimulatory molecules, and it therefore may conclusive. Whereas previous studies had found a higher contribute to reduced lymphocyte responses [16]. HLA- risk of mortality in diabetic patients [36, 37], more re- DR has previously been identified as a prototypical bio- cent studies have not shown such an association, espe- marker for stroke-induced immunodepression [12]. cially after adjusting for post-stroke hyperglycaemia [38]. IL-6 has been widely investigated as biomarker for In the management of critically ill patients, a sensitive early inflammation. Higher IL-6 levels after stroke have and specific prognosis of outcome is desirable. Outcome been associated with poor outcome, but the additional prognosis (for example chances of recovery, expected value over clinical data was moderate [17]. To the best length of recovery period, and anticipated quality of life) Mengel et al. BMC Neurology (2019) 19:148 Page 6 of 10 Table 3 Bivariate survival analysis: Cumulative proportion of deaths at 3 month after stroke a Characteristic Cumulative Proportion (%) of deaths (95% CI) p-value N = 484 Death (N = 21) Overall 4.8 (2.6–7.0) Sex 0.018 female (n = 209) 7.8 (3.7–12.0) male (n = 274) 2.6 (0.4–4.8) Age, yrs < 0.001 < 64 (n = 148) 2,8 (0–5,9) 65–74 (n = 154) 1.6 (0–3.8) 75–84 (n = 137) 6.8 (2.3–11.3) 85+ (n = 45) 17.7 (5.7–29.6) NIHSS < 0.001 0–4 (n = 265) 2.3 (0.3–4.3) 5–15 (n = 178) 7.2 (2.9–11.3) 16+ (n = 27) 22.3 (5.1–39.5) Pneumonia Yes (n = 25) 28.9 (9.1–48.7) < 0.001 No (n = 454) 3.5 (1.5–5.5) Comorbidities Hypertension 0.785 Yes (n = 388) 5.3 (2.8–9.1) No (n = 75) 3.2 (0–7.5) Diabetes mellitus 0.523 Yes (n = 128) 4.1 (0.2–8.0) No (n = 330) 5.5 (2.8–8.2) Atrial Fibrillation 0.013 Yes (n = 138) 9.6 (4.1–15.1) No (n = 319) 4.6 (0.8–5.1) Hypercholesterolemia 0.778 Yes (n = 237) 5.7 (2.4–9.0) No (n = 219) 4.3 (1.4–7.2) Previous stroke 0.276 Yes (n = 99) 6.6 (0.9–12.3) No (n = 359) 4.6 (2.2–6.9) Biomarkers mHLA-DR day1 0.014 < =16,422 (n = 234) 7.1 (3.6–10.6) > 16,422 (n = 237) 2.6 (0.2–5.0) IL-6 day1 0.023 < =4.3(n = 234) 1.6 (0–3.4) > 4.3 (n = 231) 6.2 (2.9–9.5) IL-10 day1 0.459 <=5.0 (n = 387) 3.4 (1.4–5.4) >5.0 (n = 76) 6.2 (0.3–12.1) LBP day1 0.073 Mengel et al. BMC Neurology (2019) 19:148 Page 7 of 10 Table 3 Bivariate survival analysis: Cumulative proportion of deaths at 3 month after stroke a (Continued) Characteristic Cumulative Proportion (%) of deaths (95% CI) p-value N = 484 Death (N = 21) <= 7.5 (n = 233) 1.6 (0–3.6) >7.5 (n = 231) 6.3 (2.7–9.8) TNF-α day1 0.216 <=631 (n = 214) 2.1 (0–4.3) >631 (n = 244) 5.8 (2.5–9.1) a Analyses were restricted to patients without missing values in the respective #cumulative Proportion of Deaths is calculated as 1- the Kaplan-Meier estimator influences our treatment decisions about withholding or to poor outcome and therefore may lead to new thera- withdrawing life-sustaining therapies. However, the accur- peutic strategies [41, 42]. acy of early outcome prognosis by physicians or through We identify several limitations of our study: firstly, we using current scoring systems is only moderate. Therefore had a relatively short follow-up period of 90 days after clinicians hope to improve prognosis of outcome by devel- stroke onset. However, several studies have shown that oping sensitive and specific biomarkers for clinical prac- short-term outcome is strongly associated with long-term tice [39]. The consistent and strong association of the outcome [43, 44], and that on the whole both show the clinical variables age and stroke severity with poor out- same associations with clinical variables [4], as well as with come makes it difficult for biomarkers to significantly add biomarkers [8]. Secondly, in our study, the attrition rate information for prognostic models. Furthermore, a variety between the recruitment of the 484 patients and follow- of different biomarkers have been implicated in stroke up at 3 months for mRS (via a structured telephone inter- prognosis, yet to date no biomarker has found its transla- view) was 20.5%. This rate and our follow-up method are tional way from bench to bedside [40]. Even though we comparable to those of other studies [45, 46]. Information recognize that biomarkers alone should not be decisive in on vital status was available for all patients. Thirdly, within clinical decision making, they do harbour the unique ad- the analysed cohort of patients, there is a likely selection vantage of providing insights into the mechanisms leading bias towards better outcome since the severity of initial Table 4 Characteristics associated with mortality in patients with acute stroke during 3-months follow-up. HR (hazard ratio). The final model included n = 434 complete cases and 16 events. Median follow-up time was 94 days IQR: 90–101 days)a. Nagelkerke R2 = 0.18 Characteristic HR 95% CI p-value Female sex 1.91 0.60–6.10 0.275 Age, yrs. 1.03 0.97–1.10 0.288 NIHSS categories 0–4 1.0 5–15 0.82 0.22–3.10 0.765 16+ 1.39 0.25–7.70 0.705 Pneumonia 5.43 1.22–24.11 0.026 Diabetes mellitus 0.74 0.19–2.88 0.660 Hypertension 0.63 0.15–2.88 0.529 Atrial Fibrillation 1.44 0.46–4.49 0.529 Biomarker LBP day 1b (pg/ml) 2.26 0.65–7.90 0.288 b mHLA-DR (pg/ml) 0.53 0.13–2.07 0.359 IL6b (pg/ml) 1.97 1.05–3.70 0.034 IL10 (> 5) (pg/ml) 1.02 0.27–3.81 0.979 AUC survival Cox regression model without biomarkers at day 1 0.76 (0.61–0.90) AUC survival Cox regression model with biomarkers at day 1 0.81 (0.68–0.94) a Analysis was restricted to patients without missing values for the respective variables (N = 437). Variables were eliminated using backward elimination procedure, stepwise excluding all variables with p > 0.200. AUCs were calculated with a cut-off in follow-up time at 90 days.434). Cox regression stratified by centers b log-transformed Mengel et al. BMC Neurology (2019) 19:148 Page 8 of 10 strokes (median NIHSS 4 points) was relatively low. Department of Neurology, Vivantes Klinikum Spandau, Berlin, Germany: Jörg Whereas the distribution of overall favourable and poor Müller, MD. Department of Neurology, Klinikum Frankfurt Oder, Germany: Andreas outcome is comparable to those of other studies [45, 46], Hartmann, MD. the 4.9% mortality rate in our cohort was low. In order to Department of Neurology, Vivantes Klinikum im Friedrichshain, Berlin, achieve a representative cohort of stroke patients, we in- Germany: Hans-Christian Koennecke, MD. Department of Immunology, Labor Berlin - Charité Vivantes GmbH, Berlin, cluded patients with acute ischemic stroke of any severity Germany: Christian Meisel, MD, Nadine Unterwalder, Uwe Kölsch, MD. (NIHSS≥1). The challenges of obtaining informed consent from stroke patients are widely acknowledged, with stroke Authors’ contributions AMEn wrote the study conception and design, analyzed and interpreted the severity being strongly associated with not obtaining in- data and drafted the first manuscript. UL collected the data and helped with formed consent from stroke patients [47, 48], as well as the study conception and design. HB collected the data and critically revised with mortality after stroke [4]. Though study inclusion of the manuscript. HH collected the data. PKS analyzed and interpreted the data and critically revised the manuscript. GU analyzed and interpreted the patients via “proxy consent” by their legal representative data and critically revised the manuscript. MJ collected the data and revised was possible, we found that this was often not feasible due the manuscript. MC acquired the data and critically revised the manuscript. to the relatively short time window of study recruitment AMEi was head of our study concept and design and critically revised the manuscript. HS wrote the study conception and design together with AMEn within 36 h after symptoms onset. And lastly, we suggest and analyzed and interpreted the data. She drafted the first manuscript that the subgroup analysis on mortality be interpreted together with AMEn. All authors read and approved the final manuscript. with caution due to the low case number (n = 21). In spite Funding of these limitations, the two promising candidates for fu- This study was supported by funding from German Research Foundation ture investigations are SAP as a potentially modifiable (Exc257 – design of the study) (collection of data), German Federal Ministry post-stroke complication, and IL-6 as blood-based bio- of Education and Research (01EO0801 – design of the study) (collection of data), European Community FP7 (201024 – design of the study) (collection marker, both of which were associated with shorter sur- of data), Siemens Healthcare Diagnostics (collection of data) and University vival in our patient population. of Tübingen (Open Access Publishing Fund) (publishing fee). Annerose Mengel was supported by Tübinger Research Foundation (TÜFF 2459-0-0) (writing the manuscript). Conclusion After combining with clinical variables (age, gender, stroke Availability of data and materials severity, comorbidities and stroke-associated pneumonia) The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. we identified lower mHLA-DR expression in the acute phase of stroke as being associated with poor outcome at Ethics approval and consent to participate 3 months after the index event. In addition, IL-6 was asso- Ethics approval was given by the Berlin Ethics Committee Charité – ciated with shorter survival, even after adjustment for clin- Universitätsmedizin Berlin (EA1/216/09) and Barcelona Ethics Committee Hospital Vall d’Hebron Barcelona (No.199 25/05/2012) as umbrella approvals ical data. Both biomarkers could help to develop better that covered all participating sites in Germany and Spain, respectively. All prediction models for stroke outcome in the future. patients or their legal representatives gave written informed consent in accordance with the Declaration of Helsinki. Proxy-consent was necessary if Abbreviations the patient could not consent due to aphasia or reduced level of conscious- AUC: Area under the curve measures; CI: Confidence intervals; ness. These criteria were approved by the local ethics committee. FACS: Fluorescence activated cell sorting FACS; HR: Hazard ratios; IL- 10: Interleukin 10; IL-6: Interleukin 6; IQR: Inter-quartile range; KM: Kaplan- Consent for publication Meier; LBP: Lipopolysaccharide-binding protein; mHLA-DR: monocytic HLA- Not Applicable. DR; mRS: modified Rankin Scale; NIHSS: National Institute of Health Stroke Scale; OR: Odds ratio; SAP: Stroke-associated pneumonia; SIDS: Stroke- Competing interests induced immunodepression syndrome; TIA: Transient ischemic attack; TNF- Christian and Andreas Meisel have issue of a patent on anti-infective alpha: Tumor necrosis factor alpha and immunomodulatory agents used for preventive antibacterial therapy after stroke (EP1480646). As one of the co-authors, Andreas Meisel, is a Acknowledgements section editor of this journal. The other authors report no conflict of The authors gratefully acknowledge the support of the recruiting study sites interest. and all PREDICT investigators: This study was supported by Siemens Healthcare Diagnostics (collection Department of Neurology, Charité – Universitätsmedizin Berlin, Germany of data). (Campus Mitte, Campus Virchow Klinikum, Campus Steglitz): Andreas Meisel, MD; Sarah Hoffmann, MD; Gerhard Jan Jungehuelsing, MD; Lena Ulm, MD; Author details 1 Jos Göhler, MD; Thomas Krause, MD; Robert Steinicke, MD. Department of Neurology Berlin, Charité Universitätsmedizin Berlin, Department of Neurology, St.-Josefs Krankenhaus Potsdam, Germany: Charitéplatz 1, 10115 Berlin, Germany. 2Department of Neurology and Stroke, Hendrik Harms, MD; Olaf Hoffmann, MD; Jens-Eric Röhl, MD. Universitätsklinik Tuebingen, Hoppe-Seyler-Str.3, 72076 Tuebingen, Germany. 3 Department of Neurology, Vivantes Klinikum Neukölln, Berlin, Germany: Center for Clinical Research, The University of Queensland, Herston, Darius Günther Nabavi, MD; Marcella Hermans, MD; Olaf Crome, MD, Jens Queensland 4029, Australia. 4Center for Stroke Research Berlin, Charité Offermann, MD. Universitätsmedizin Berlin, Charitéplatz 1, 10115 Berlin, Germany. 5Institute of Department of Neurology, Vivantes Auguste Viktoria Klinikum, Berlin, Biometry and Clinical Epidemiology, Charité - Universitätsmedizin Berlin, Germany: Bruno-Marcel Mackert, MD; Frank Hamilton, MD. corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, Department of Neurology, Hospital Vall d’Hebron Barcelona, Spain: Joan Berlin, Germany. 6Berlin Institute of Health, Charitéplatz 1, D-10117 Berlin, Montaner, MD; Alejandro Bustamante, MD. Germany. 7Berlin Institute of Health (BIH), Anna-Louisa-Karsch 2, 10178 Berlin, Department of Neurology, Unfallkrankenhaus Berlin, Germany: Ingo Germany. 8Neurovascular Research Laboratory, Vall d’Hebron Institut de Schmehl, MD, Recerca, Passeig Vall d’Hebron 119-129, 08035 Barcelona, Spain. 9NeuroCure Mengel et al. BMC Neurology (2019) 19:148 Page 9 of 10 Clinical Research Center Berlin, Charité Universitätsmedizin Berlin, Charitéplatz from the preventive antibacterial therapy in stroke trial. Neuroscience. 1, 10115 Berlin, Germany. 2009;158:1184–93. 22. Villar J, Maca-Meyer N, Perez-Mendez L, Flores C. Bench-to-bedside review: Received: 1 February 2018 Accepted: 23 June 2019 understanding genetic predisposition to sepsis. Crit Care. 2004;8:180–9. 23. Stanley D, Mason LJ, Mackin KE, Srikhanta YN, Lyras D, Prakash MD, et al. Translocation and dissemination of commensal bacteria in post-stroke infection. Nat Med. 2016;22:1277–84. References 24. Worthmann H, Tryc AB, Dirks M, Schuppner R, Brand K, Klawonn F, et al. 1. Heuschmann PU, Wiedmann S, Wellwood I, Rudd A, Di Carlo A, Bejot Y, et Lipopolysaccharide binding protein, interleukin-10, interleukin-6 and c- al. Three-month stroke outcome: the european registers of stroke (Eros) reactive protein blood levels in acute ischemic stroke patients with post- investigators. Neurology. 2011;76:159–65. stroke infection. J Neuroinflammation. 2015;12:13. 2. Garcia-Berrocoso T, Giralt D, Bustamante A, Etgen T, Jensen JK, Sharma JC, 25. Wartenberg KE, Stoll A, Funk A, Meyer A, Schmidt JM, Berrouschot J. et al. B-type natriuretic peptides and mortality after stroke: a systematic Infection after acute ischemic stroke: risk factors, biomarkers, and outcome. review and meta-analysis. Neurology. 2013;81:1976–85. Stroke Res Treat. 2011;2011:830614. 3. Appelros P, Nydevik I, Viitanen M. Poor outcome after first-ever stroke: 26. Hong KS, Kang DW, Koo JS, Yu KH, Han MK, Cho YJ, et al. Impact of predictors for death, dependency, and recurrent stroke within the first year. neurological and medical complications on 3-month outcomes in acute Stroke. 2003;34:122–6. ischaemic stroke. Eur J Neurol. 2008;15:1324–31. 4. Saposnik G, Hill MD, O’Donnell M, Fang J, Hachinski V, Kapral MK, et al. 27. Katzan IL, Cebul RD, Husak SH, Dawson NV, Baker DW. The effect of Variables associated with 7-day, 30-day, and 1-year fatality after ischemic pneumonia on mortality among patients hospitalized for acute stroke. stroke. Stroke. 2008;39:2318–24. Neurology. 2003;60:620–5. 5. Wahlgren N, Ahmed N, Eriksson N, Aichner F, Bluhmki E, Davalos A, et al. 28. Koennecke HC, Belz W, Berfelde D, Endres M, Fitzek S, Hamilton F, et al. Multivariable analysis of outcome predictors and adjustment of main Factors influencing in-hospital mortality and morbidity in patients treated outcome results to baseline data profile in randomized controlled trials: safe on a stroke unit. Neurology. 2011;77:965–72. implementation of thrombolysis in stroke-monitoring study (sits-most). 29. Kumar S, Selim MH, Caplan LR. Medical complications after stroke. Lancet Stroke. 2008;39:3316–22. Neurol. 2010;9:105–18. 6. Popovic N, Stefanovic-Budimkic M, Mitrovic N, Urosevic A, Milosevic B, Pelemis M, et al. The frequency of poststroke infections and their impact on 30. Eriksson M, Glader EL, Norrving B, Terent A, Stegmayr B. Sex differences in early stroke outcome. J Stroke Cerebrovasc Dis. 2013;22:424–9. stroke care and outcome in the swedish national quality register for stroke care. Stroke. 2009;40:909–14. 7. Whiteley W, Chong WL, Sengupta A, Sandercock P. Blood markers for the prognosis of ischemic stroke: a systematic review. Stroke. 2009;40:e380–9. 31. Gargano JW, Reeves MJ, Paul Coverdell National Acute Stroke Registry 8. Bustamante A, Sobrino T, Giralt D, Garcia-Berrocoso T, Llombart V, Ugarriza I, Michigan Prototype I. Sex differences in stroke recovery and stroke- et al. Prognostic value of blood interleukin-6 in the prediction of functional specific quality of life: results from a statewide stroke registry. Stroke. outcome after stroke: a systematic review and meta-analysis. J 2007;38:2541–8. Neuroimmunol. 2014;274:215–24. 32. Paolucci S, Bragoni M, Coiro P, De Angelis D, Fusco FR, Morelli D, et al. Is 9. Tu WJ, Dong X, Zhao SJ, Yang DG, Chen H. Prognostic value of plasma sex a prognostic factor in stroke rehabilitation? A matched comparison. neuroendocrine biomarkers in patients with acute ischaemic stroke. J Stroke. 2006;37:2989–94. Neuroendocrinol. 2013;25:771–8. 33. Glader EL, Stegmayr B, Norrving B, Terent A, Hulter-Asberg K, Wester PO, et 10. Hoffmann S, Harms H, Ulm L, Nabavi DG, Mackert BM, Schmehl I, et al. al. Sex differences in management and outcome after stroke: a swedish Stroke-induced immunodepression and dysphagia independently predict national perspective. Stroke. 2003;34:1970–5. stroke-associated pneumonia - the predict study. J Cereb Blood Flow 34. Reeves M, Bhatt A, Jajou P, Brown M, Lisabeth L. Sex differences in the use Metab. 2016;37(12):3671–82 271678X16671964. of intravenous rt-pa thrombolysis treatment for acute ischemic stroke: a 11. Baird AE, Dambrosia J, Janket S, Eichbaum Q, Chaves C, Silver B, et al. meta-analysis. Stroke. 2009;40:1743–9. A three-item scale for the early prediction of stroke recovery. Lancet. 35. Persky RW, Turtzo LC, McCullough LD. Stroke in women: disparities and 2001;357:2095–9. outcomes. Curr Cardiol Rep. 2010;12:6–13. 12. Harms H, Prass K, Meisel C, Klehmet J, Rogge W, Drenckhahn C, et al. 36. Jorgensen H, Nakayama H, Raaschou HO, Olsen TS. Stroke in patients with Preventive antibacterial therapy in acute ischemic stroke: a randomized diabetes. The Copenhagen stroke study. Stroke. 1994;25:1977–84. controlled trial. PLoS One. 2008;3:e2158. 37. Tuomilehto J, Rastenyte D, Jousilahti P, Sarti C, Vartiainen E. Diabetes 13. Weimar C, Kurth T, Kraywinkel K, Wagner M, Busse O, Haberl RL, et al. mellitus as a risk factor for death from stroke. Prospective study of the Assessment of functioning and disability after ischemic stroke. Stroke. 2002; middle-aged finnish population. Stroke. 1996;27:210–5. 33:2053–9. 38. Arnold M, Mattle S, Galimanis A, Kappeler L, Fischer U, Jung S, et al. Impact 14. Urra X, Cervera A, Obach V, Climent N, Planas AM, Chamorro A. Monocytes of admission glucose and diabetes on recanalization and outcome after are major players in the prognosis and risk of infection after acute stroke. intra-arterial thrombolysis for ischaemic stroke. Int J Stroke. 2014;9:985–91. Stroke. 2009;40:1262–8. 39. Meisel A, Meisel C, Harms H, Hartmann O, Ulm L. Predicting post-stroke 15. Ulm L, Harms H, Ohlraun S, Reimnitz P, Meisel A. Impact of infections on infections and outcome with blood-based immune and stress markers. long-term outcome after severe middle cerebral artery infarction. J Neurol Cerebrovasc Dis. 2012;33:580–8. Sci. 2012;319:15–7. 40. Kuhle J, Petzold A. What makes a prognostic biomarker in cns diseases: 16. Wolk K, Docke WD, von Baehr V, Volk HD, Sabat R. Impaired antigen strategies for targeted biomarker discovery? Part 1: acute and monophasic presentation by human monocytes during endotoxin tolerance. Blood. diseases. Expert Opin Med Diagn. 2011;5:333–46. 2000;96:218–23. 41. Feuerstein GZ, Chavez J. Translational medicine for stroke drug discovery: 17. Bustamante A, Lopez-Cancio E, Pich S, Penalba A, Giralt D, Garcia-Berroscoso the pharmaceutical industry perspective. Stroke. 2009;40:S121–5. T, et al. Blood biomarkers for the early diagnosis of stroke: The stroke-chip 42. Frank R, Hargreaves R. Clinical biomarkers in drug discovery and study. Stroke. 2017;48:2419–25. development. Nat Rev Drug Discov. 2003;2:566–80. 18. Boehme AK, McClure LA, Zhang Y, Luna JM, Del Brutto OH, Benavente OR, 43. Kissela B, Lindsell CJ, Kleindorfer D, Alwell K, Moomaw CJ, Woo D, et al. Clinical et al. Inflammatory markers and outcomes after lacunar stroke: levels of prediction of functional outcome after ischemic stroke: the surprising importance inflammatory markers in treatment of stroke study. Stroke. 2016;47:659–67. of periventricular white matter disease and race. Stroke. 2009;40:530–6. 19. Aulin J, Siegbahn A, Hijazi Z, Ezekowitz MD, Andersson U, Connolly SJ, et al. 44. Meyer S, Verheyden G, Brinkmann N, Dejaeger E, De Weerdt W, Feys H, et Interleukin-6 and c-reactive protein and risk for death and cardiovascular al. Functional and motor outcome 5 years after stroke is equivalent to events in patients with atrial fibrillation. Am Heart J. 2015;170:1151–60. outcome at 2 months: follow-up of the collaborative evaluation of 20. Garcia JM, Stillings SA, Leclerc JL, Phillips H, Edwards NJ, Robicsek SA, et al. rehabilitation in stroke across europe. Stroke. 2015;46:1613–9. Role of interleukin-10 in acute brain injuries. Front Neurol. 2017;8:244. 45. Ayis S, Wellwood I, Rudd AG, McKevitt C, Parkin D, Wolfe CD. Variations in 21. Klehmet J, Harms H, Richter M, Prass K, Volk HD, Dirnagl U, et al. health-related quality of life (hrqol) and survival 1 year after stroke: five Stroke-induced immunodepression and post-stroke infections: lessons european population-based registers. BMJ Open. 2015;5:e007101. Mengel et al. BMC Neurology (2019) 19:148 Page 10 of 10 46. Kim GM, Park KY, Avery R, Helenius J, Rost N, Rosand J, et al. Extensive leukoaraiosis is associated with high early risk of recurrence after ischemic stroke. Stroke. 2014;45:479–85. 47. Mendyk AM, Labreuche J, Henon H, Girot M, Cordonnier C, Duhamel A, et al. Which factors influence the resort to surrogate consent in stroke trials, and what are the patient outcomes in this context? BMC Med Ethics. 2015;16:26. 48. Flaherty ML, Karlawish J, Khoury JC, Kleindorfer D, Woo D, Broderick JP. How important is surrogate consent for stroke research? Neurology. 2008;71:1566–71. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.