MEANS AND METHOD FOR THE EARLY PREDICTION OF POOR NEUROLOGICAL OUTCOME IN OUT-OF-HOSPITAL CARDIAC ARREST SURVIVORS
20220074954 · 2022-03-10
Assignee
Inventors
Cpc classification
G01N2800/325
PHYSICS
International classification
Abstract
The present invention relates to a method for predicting the neurological outcome of a patient after cardiac arrest, the method comprising the step(s) of: (I) a) determining the level of alpha-enolase in a biological sample obtained from a patient after cardiac arrest; and b) comparing the level of alpha-enolase obtained in (a) with the median or mean level of alpha-enolase in biological samples from patients with positive neurological outcome after cardiac arrest, wherein said patient is diagnosed as having a negative neurological outcome if the alpha-enolase level is increased by at least 100% over the median or mean level of alpha-enolase, and/or a positive neurological outcome if the alpha-enolase level is below 150% as compared to the median or mean level of alpha-enolase; or (II) determining the level of alpha-enolase in a biological sample obtained from a patient after cardiac arrest, wherein said patient is diagnosed as having a negative neurological outcome if the alpha-enolase level is above 50 μg/L, and/or a positive neurological outcome if the alpha-enolase level is below 50 μg/L.
Claims
1. A method for predicting the neurological outcome of a patient after cardiac arrest, the method comprising the step(s) of: (I) a) determining the level of alpha-enolase in a biological sample obtained from a patient after cardiac arrest; and b) comparing the level of alpha-enolase obtained in (a) with the median or mean level of alpha-enolase in biological samples from patients with positive neurological outcome after cardiac arrest, wherein said patient is diagnosed as having a negative neurological outcome if the alpha-enolase level is increased by at least 100% over the median or mean level of alpha-enolase, and/or a positive neurological outcome if the alpha-enolase level is below 150% as compared to the median or mean level of alpha-enolase; or (II) determining the level of alpha-enolase in a biological sample obtained from a patient after cardiac arrest, wherein said patient is diagnosed as having a negative neurological outcome if the alpha-enolase level is above 50 μg/L, and/or a positive neurological outcome if the alpha-enolase level is below 50 μg/L.
2. The method of claim 1, comprising prior to step (a) the step of determining the median or mean level of alpha-enolase in biological samples from patients with positive neurological outcome after cardiac arrest.
3. The method of claim 1, further comprising (I′) c) determining the level of at least one additional biomarker in the biological sample obtained from a patient after cardiac arrest; and d) comparing the level of said at least one additional biomarker obtained in (c) with the median or mean level of said at least one additional biomarker in biological samples from patients with positive neurological outcome after cardiac arrest, wherein said at least one additional biomarker is selected from the group consisting of the heat shock cognate 71 kDa protein, 14-3-3 protein zeta/delta, and cofilin-1, and wherein said patient is diagnosed as having a negative neurological outcome if the level of said at least one additional biomarker is increased by at least 100% over the median or mean level of said at least one additional biomarker, and/or as having a positive neurological outcome if the level of said at least one additional biomarker is below 150% as compared to the median or mean level of said at least one additional biomarker; or (II′) determining the level of at least one additional biomarker in the biological sample obtained from a patient after cardiac arrest, wherein said at least one additional biomarker is selected from the group consisting of the heat shock cognate 71 kDa protein, 14-3-3 protein zeta/delta, and cofilin-1, and wherein said patient is diagnosed as having a negative neurological outcome if the heat shock cognate 71 kDa protein level is above 490 μg/L, the 14-3-3 protein zeta/delta protein level is above 2.7 mg/L and/or the cofilin-1 protein level is above 180 μg/L, and/or a positive neurological outcome if the heat shock cognate 71 kDa protein level is below 490 μg/L, the 14-3-3 protein zeta/delta protein level is below 2.7 mg/L and/or the cofilin-1 protein level is below 180 μg/L.
4. The method of claim 3, further comprising prior to step (c) the step of determining the median or mean level of the at least one additional biomarker in biological samples from patients with positive neurological outcome after cardiac arrest.
5. The method of claim 1, wherein the patients with positive neurological outcome after cardiac arrest are at least 5, preferably at least 10, more preferably at least 20, and most preferably at least 30 patients.
6. The method of claim 1, wherein the biological sample(s) is/are whole blood, plasma, serum, cerebrospinal fluid or brain tissue.
7. The method of claim 6, wherein the biological sample(s) is/are plasma.
8. The method of claim 1, wherein patients with a positive neurological outcome are patients having after recovery from the cardiac arrest a Cerebral Performance Category (CPC) score of 1 or 2.
9. The method of claim 1, wherein the level of alpha-enolase and optionally also the level of the at least one additional biomarker has been/have been/is/are determined by mass spectrometry, enzyme-linked immunosorbent assay (ELISA), electrochemiluminescence assay and/or a radioimmunoassay.
10. The method of claim 1, wherein the biological sample(s) obtained from patient(s) after cardiac arrest has/have been obtained within 96 h, preferably within 72 h, more preferably between 24 and 72 h and most preferably about 48 h after cardiac arrest.
11. The method of claim 1, further comprising predicting the neurological outcome after cardiac arrest in the patient on the basis of one or more selected from the age of the patient, shockable rhythm, haemoglobin concentration, the time from cardiac arrest to the start of life support, the time from cardiopulmonary resuscitation to return of spontaneous circulation (ROSC), the maximum serum lactate within 24 h after cardiac arrest, the maximum difference to the pH of 7.3 within 24 h after cardiac arrest, and the level of the protein S-110B in the sample of the patient.
12. (canceled)
13. Kit for or predicting the neurological outcome of a patient after cardiac arrest, said kit comprising a binding agent which specifically binds to alpha-enolase.
14. The kit of claim 13, wherein the kit additionally comprises one or more agents specifically binding to heat shock cognate 71 kDa protein, 14-3-3 protein zeta/delta or cofilin-1.
15. The kit of claim 13, wherein the binding agent(s) is/are selected from antibodies, antibody mimetics, small molecules and aptamers.
Description
[0156] Furthermore, in the claims the word “comprising” does not exclude other elements or steps, and the indefinite article “a” or “an” does not exclude a plurality. A single unit may fulfil the functions of several features recited in the claims. The terms “essentially”, “about”, “approximately” and the like in connection with an attribute or a value particularly also define exactly the attribute or exactly the value, respectively. Any reference signs in the claims should not be construed as limiting the scope.
[0157] The Figures show.
[0158]
[0159]
[0160]
[0161] The examples illustrate the invention.
EXAMPLES
[0162] In the prospective, observational, single-center study described herein below all out-of-hospital cardiac arrest (OHCA) survivors admitted to the emergency department of a university-affiliated tertiary care centre were enrolled and neurological outcome was assessed by the Cerebral Performance Categories (CPC) score at discharge. To identify biomarkers for negative neurological outcome (CPC score 3) in plasma samples collected 48 hours after OHCA, a three-step proteomics strategy of preselection by shotgun analyses, crosschecking in brain tissue samples and verification by targeted proteomic analyses in combination with a discovery and validation statistical approach was performed.
[0163] 96 OHCA survivors were included of whom 63 patients (66%) had a negative neurological outcome. Out of a total of 299 plasma proteins, identified alpha-enolase, 14-3-3 protein zeta/delta, cofilin-1 and Heat shock cognate 71 kDa protein were identified as novel biomarkers for negative neurological outcome. The implementation of these proteomically identified biomarkers into a clinical multimarker model resulted in a significant improvement of neurological outcome prediction (C-index 0.70) compared to traditional risk factors alone (C-index 0.67; p=0.019). Thus, explained variation of the neurological outcome could be significantly improved from 10.0% to 11.9% (Table 10, in particular Model 2 vs. Model 4).
Example 1—Experimental Outline and Methods
[0164] Study Design and Patients
[0165] The study was designed as a combination of a pilot basic science study and a clinical verification trial: a sentinel cohort was used to identify potential plasma biomarker candidates for negative neurological outcome by comparing the proteomic profile between patients with positive and negative neurological outcome. The clinical verification study was designed to establish biomarker candidates in a larger study population (verification cohort).
[0166] OHCA patients were consecutively recruited on arrival at the emergency department of the Medical University of Vienna, a university-affiliated tertiary care center, and were followed up to intensive care unit (ICU) discharge. Adult patients suffering non-traumatic, normothermic OHCA due to cardiac disorders, respiratory failure, or hemodynamic or metabolic factors with a Glasgow Coma Scale of three at admission were eligible. Patients with history of previous cardiac arrest, coexisting neurological disorders or neoplasms of the central nervous system, psychiatric illness, alcohol or drug misuse and those with ongoing psychotropic medication as identified by the attending or ward physician were excluded. Moreover, patients with abnormal findings in the cranial computed tomography, including hydrocephalus and shunt artifacts, severe movement artifacts, intracerebral hemorrhage or chronic large ischemic lesions were excluded. The sentinel cohort was composed by a random selection of five patients with positive and five patients with negative neurological outcome using a pseudo-random number generator (Urbaniak et al. (2013) Research Randomizer, Version 4.0. The remaining study participants were assigned to the verification cohort.
[0167] Outcomes
[0168] The primary study endpoint was the neurological outcome at discharge from the ICU assessed by an experienced neurologist using the Cerebral Performance Categories (CPC) score, which measures outcome on a five-point scale (1=good cerebral performance, 2=moderate cerebral disability, 3=severe cerebral disability, 4=coma or vegetative state, and 5=brain death). Neurological outcome was dichotomized into positive neurological outcome (CPC 2) and negative neurological outcome (CPC score ≥3).
[0169] Study Procedures
[0170] Cardiopulmonary resuscitation (CPR) and post-resuscitation management were performed according to current guidelines (Callaway et al. (2015) Circulation 132 (16 Suppl. 1):S84-145. All patients were routinely treated with targeted temperature management for 24 hours. Data on cardiac arrest for individual patients were recorded according to ‘Utstein Style Criteria’ (Jacobs et al. (2004) Circulation 110(21):3385-97).
[0171] Blood samples were drawn and immediately placed on ice from all patients 48 hours after hospital admission. Neuron-specific enolase (NSE) and protein S100-B (S100B) levels were determined by electrochemiluminescence immunoassays according to the local laboratory's standard procedure. Additional plasma samples were stored at −80° C. and applied for proteomic analyses immediately after thawing. Brain tissue especially vulnerable to hypoxia (striatum, hippocampus and periventricular white matter) was harvested postmortem from patients without significant neurological symptoms or substantial neuropathological alterations who died for other causes than OHCA by the Institute of Neurology, Medical University of Vienna and processed for proteomic analyses.
[0172] The experimental procedure is detailed further below. Depletion of top twelve abundant proteins and tryptic in-solution digestion were applied as preparation steps to plasma samples. After sonication of brain tissue samples in sample buffer, soluble proteins were digested.
[0173] The shotgun proteomics analysis was conducted on a Q-Exactive Orbitrap (Thermo Fisher Scientific, USA) coupled with an UltiMate 3000 RSLC nano System (Dionex, USA) (Muqaku et al. (2017) Proteomics 16(1):86-99). For the evaluation of shotgun liquid chromatography-mass spectrometry (LC-MS) data, the MaxQuant 1.5.2.8 software tools were used (Cox et al. (2008) Nature biotechnology 26(12):1367-72).
[0174] Targeted analysis was performed on an Agilent 6490 triple quadrupole mass spectrometer coupled with a nano-Chip-LC Agilent Infinity Series HPLC1260 system (Agilent Technologies, USA). Skyline software was used for MRM method development, as described recently (Muqaku, as above). Protein abundance levels were indicated as log 2 intensities.
[0175] Statistical Analysis
[0176] Pre-selection of biomarker candidates: Quantification and statistical analyses of shotgun LC-MS data were based on label-free quantification (LFQ) values using Perseus software as previous described (Muqaku, as above). Proteins with significantly different LFQ values and none significantly regulated proteins with a fold change of LFQ values higher than two between patients with negative and positive neurological outcome were taken into consideration for MRM assay development. Detailed information are provided herein below. Each of the biomarker candidates were tested for their ability to predict neurological outcome using univariate regression analyses. Wilcoxon signed-rank test was used to compare protein abundance levels between patients with negative and positive neurological outcome.
[0177] Development and evaluation of a proteomic signature: The pre-selected biomarker set was further considered to identify a proteomic signature discriminating patients with positive neurological outcome from those with negative outcome with the verification cohort of n=86. This sample size was based on feasibility and the sample size of other prospective studies performed on a similar patient population (Cronberg et al. (2011) Neurology 77(7):623-30). Any more formal sample size calculation for this biomarker finding study would have required strong and a-priorily unestimable assumptions on the number, the expression and the correlation of the proteomic biomarkers, and their combined effect on the outcome. Some variables were transformed due to high frequency of zeros and high skewness, and missing values were imputed using chained equations.
[0178] Three blocks of variables available at day two after ROSC were considered for model building. The first block consisted of well-established predictors of neurological outcomes such as age, shockable rhythm, time from cardiac arrest to start of life support, time from CPR to ROSC, pH, serum lactate in mmol/L, and hemoglobin in g/dl. As a second block, adrenalin was added in mg administrated during resuscitation, witnessed cardiac arrest, sex, NSE, and S100B. The third block consisted of 24 proteomic features. A ridge logistic regression model (M1) was fitted with the first block of variables, and linear predictors from that model were computed (Hoerl et al. (1970) Technometrics 12(1):55-67). The linear predictors from M1 were used as offsets in a LASSO logistic regression model (M2), where the second block of potential predictors was evaluated. The third model (M3) extended M2 by proteomic markers in the third block. In M2 and M3, all variables of the second and third blocks were competing for selection. The penalty parameter was optimized by evaluating the 10-fold cross-validated deviance (CVD). A fourth model (M4) was obtained by re-fitting M3 with elastic-net logistic regression as an alternative selection method (Zou et al. (2005) Journal of the Royal Statistical Society Series B 67(2):301-20).
[0179] To evaluate the model building strategy, 10-fold cross-validation was used and the concordance index (area under the ROC curve), the explained variation expressed as discrimination slope, and the calibration slope was computed. Calibration was graphically evaluated. Model stability was investigated by computing bootstrap inclusion frequencies for each variable. A non-parametric bootstrap test to evaluate the added value of blocks of variables was used.
[0180] Descriptive analyses comprised computation of medians and interquartile ranges for continuous variables, and absolute and relative frequencies for categorical variables. More details on statistical model building are provided further below.
[0181] Baseline characteristics: One-hundred OHCA patients were included with a median age of 58 years (IQR, 49-69). Four patients were excluded because of poor specimen quality. Hence, the final study population consisted of 96 patients, of whom 63 patients (66 percent) had a negative neurological outcome at discharge from the ICU. Detailed baseline characteristics are displayed in Table 1.
[0182] Identification of biomarker candidates: The flow chart of proteomic analyses is illustrated in
[0183] Prognostic model: The results of a univariate regression analysis of each of the 24 biomarker candidates to predict negative neurological outcome are outlined in Table 5. To explore the best multivariable combination of clinical parameters and biomarkers to predict neurological outcome in OHCA patients, a multi-stage statistical approach was applied. Model M1, including all well-established neurological outcome predictors, achieved a discriminative ability (cross-validated c-index) of 0.67 (Table 6). This c-index increased to 0.68 by the variable selection process of model M2, which added S100-B (Table 7). Model M3 additionally selected a multivariable combination of the proteomically measured biomarker candidates (n=24) and yielded a c-index of 0.70 (Table 8). This optimal combination, consisting of alpha-enolase, 14-3-3 protein zeta/delta and cofilin-1, resulted in a significant improvement in neurological outcome prediction compared to M2 (p=0.025 for added value). M4, which employed elastic-net as a less restrictive variable selection procedure, additionally selected Heat shock cognate 71 kDa protein (HSPA8), yielding similar performance (c-index=0.70; p=0.019 for added value compared to M2; Table 9). Explained variation of the outcome could be improved from 8.2% in M1 to 10.0% in M2 and to 12.2% in M3 and to 11.9% in M4 (Table 10).
[0184] As illustrated in
[0185] Experimental Methods
[0186] Plasma depletion: For proteomic analyses, plasma samples were depleted employing Pierce™ Top 12 Abundant Protein Depletion Spin Columns (Thermo Fisher Scientific, USA). The depletion was performed according to the protocol of the manufacturer, using 14 μL of plasma for shotgun analysis and 550 μg total protein amount of plasma per depletion for targeted analysis. Protein digestion was performed using 257 μl of depleted plasma containing not more than 25 μg total protein amount for shotgun analysis and for targeted analysis 20 μg total protein amount was used. Bradford assay was applied to all samples for determination of total protein concentration before and after depletion.
[0187] Preparation of brain tissue samples: Brain tissue samples were sonicated for several seconds using a ultrasonicator after adding of 100 μl sample puffer. The supernatant containing soluble proteins was collected after centrifugation of sonicated sample for 10 min at 14000 g. For the protein digestion 20 μg total protein amount were used. Sample buffer compositions: 45% urea (Merck, Germany), 11.4% thiourea (Sigma-Aldrich, USA), 4% CHAPS (Gerbu, Germany), 0.05% sodium dodecyl sulfate (Gerbu) and 0.1M dithiothreitol (Gerbu).
[0188] In-solution digestion: In-solution digestion was performed using 10 kDa MW cut off filters (Nanosep with Omega membrane, USA), which consist on protein reduction with 32 mM dithiothreitol (Gerbu) and alkylation of reduced proteins with 54 mM 2-iodoacetamide solution (Sigma Aldrich). Both, dithiothreitol and 2-iodoacetamide solutions were prepared with 8 M guanidinium hydrochloride (Sigma Aldrich) solution in 50 nM ammonium bicarbonate. Afterwards, proteins were digested on the filter using a Trypsin/Lys-C Mix (MS grade; Promega, USA) with a total enzyme to protein ratio of 1:20. For clean-up of resulting peptide samples C18 spin columns (Pierce™ C18 Spin Columns, Thermo Fisher Scientific, USA) were used and, after drying via vacuum centrifugation, the samples were stored at −20° C. Upon liquid chromatography-mass spectrometry (LC-MS) analysis the dried peptides were reconstituted in 5 μL of the equimolar 10 fmol standard peptide mix and 40 μL of mobile phase A (98% H2O, 2% ACN, 0.1% FA) for shotgun analysis, and for targeted analysis in 30 μL of the 10 fmol/μL standard peptide mix solution containing 30% formic acid. Standard peptides [M28-TTPA VLDSDGSYFLYSK; HK0-VLETKSLYVR; HK1-VLETK(ε-AC)SLYVR] were obtained from Sigma and Peptide Specialty Laboratories GmbH and spiked in each sample as internal quality control for monitoring LC-MS-system stability.
[0189] Shotgun LC-MS analysis: The shotgun analysis of plasma as well as brain tissue samples were conducted on the same mass spectrometric equipment by applying the same setting for protein identification and MS1 based protein quantification as published recently. Briefly, peptides were separated by UltiMate 3000 RSLC nano System (Pre-column: Acclaim PepMap 100, C18 100 μm×2 cm; Analytical column: Acclaim PepMap RSLC C18 75 μm×50 cm; Dionex, USA). Peptides of plasma samples were eluted to the analytical column by applying a gradient from 8 to 40% mobile phase B (80% ACN, 2% H2O, 0.1% FA) over 43 min and operating at a flow rate of 300 nL/min. Whereas peptides generated from digestion of brain tissue samples were separated by applying a gradient from 8% to 40% over 235 min.
[0190] Data acquisition was conducted on a QExactive mass spectrometer (Thermo Fischer Scientific, USA) using a top 8 data dependent method. Protein identification was achieved using the MaxQuant 1.5.2.8 software and searching against the UniProt database (version 102014 with 20 195 entries). Search criteria included a maximum of two missed cleavages. Additionally, a minimum of two peptide identifications per protein (including one unique) was requested and an false discovery rate of less than 0.01 was applied at both peptide and protein level. Carbamidomethylation of cysteines was set as fixed modification and methionine oxidation as well as N-terminal protein acetylation as variable modifications.
[0191] Shotgun analyses were performed in plasma samples of OHCA′ patients with negative and positive neurological outcome. A MS1-based label-free quantification approach and statistical analysis was applied to quantify the identified proteins based on label free quantification (LFQ) values using Perseus. Up- and downregulated proteins were determined applying a two-sided t test. The significance threshold of p-value<0.05 (minimum fold change of two of LFQ values) was calculated using permutation-based multiparameter correction. Only proteins fulfilling the following selection criteria were taken into consideration for multiple reaction monitoring (MRM) assay development: multi-parameter significantly regulated proteins; none significantly regulated proteins with p-value lower than 0.05 and none significantly regulated proteins with fold change of LFQ values (up- or down regulated) higher than 2. In order to identify brain-derived plasma proteins, these biomarker candidates were crosschecked with the proteome of brain autopsy samples.
[0192] Targeted LC-MRM analysis: Targeted analysis was performed on an Agilent 6490 triple quadrupole mass spectrometer coupled with a nano-Chip-LC Agilent Infinity Series HPLC1260 system, as described recently. Solvent compositions were 97.8% H2O, 2% ACN and 0.2% FA for solvent A, and 97.8% ACN, 2% H2O and 0.2% FA for solvent B.
[0193] MRM Assay Development: The MRM method was developed based on high resolution orbitrap shotgun data and using same plasma sample measured with shotgun approach. Applied criteria for the development of MRM method were recently published. Briefly, only peptides with unique amino acid sequence and peptides without missed cleavage were considered for further optimization. Additionally, peptides containing methionine residue were excluded from the list of targeted peptides. Best transitions and interference-free transitions were selected by applying of a unscheduled MRM method in multiple sample injections. In the final step 86 plasma sample were measured in technical duplicates with developed scheduled MRM method by using 3.5 minutes time window and cycle time of 1300 milliseconds. The MRM method development was implemented using Skyline software (v. 3.1), allowing data analysis and data preparation for statistical analysis. The manual inspection regarding correct peak selection, interferences, and integration boundaries as well as normalization to standard peptides of the data was done with Skyline. Additionally, the data were normalized to the sample volume used for depletion and digestion. In the next step, data were multiplied with 10.sup.9 and were log 2 transformed. Finally, the abundance at protein level was calculated as mean value over two technical duplicates of all transition signals generated from all peptides belonging to a protein.
[0194] Statistical Methods
[0195] Variable transformation and imputation of missing values for statistical model building: For statistical model building, some variables were transformed due to high frequency of zeros and high skewness. All biomarker were log 2-transformed. Time from CPR to the return of spontaneous circulation (ROSC) in minutes was transformed to log 2(x+1). Time from cardiac arrest to start of life support in minutes was dichotomized into ‘immediate’ and ‘not immediate’. The pH level was measured as the maximum absolute difference of pH to 7.3 pH within the first 24 hours. For serum lactate, the maximum value measured within the first 24 hours was used in the analysis. Missing values in 6 clinical variables (max 23%) were imputed using chained equations.
[0196] Validation of model building strategy and model stability: An outer 10-fold cross-validation loop was used to evaluate the model building strategy. In particular, the concordance index (area under the ROC curve) was computed, the explained variation expressed as discrimination slope and the calibration slope. Calibration was also graphically evaluated by calibration plots, showing the observed vs. the predicted rates of negative neurological outcome. Model stability was investigated by computing bootstrap inclusion frequencies for each variable, if the model building process was repeated in 1000 bootstrap resamples drawn with replacement from the original data set.
[0197] Test for added value of blocks of variables: M1 was tested against a null model by randomly permuting the outcome status of the patients 1000 times, refitting M1 to each of the permuted data sets and computing a p-value as the proportion of permuted CVD values which were less than or equal to CVD of M1 in the original data set. M2 was tested against M1 as follows. First, M1 and M2 were refitted in 1000 resamples drawn using replacement from the original data. The resulting two linear predictors X{circumflex over (β)}.sub.1.sup.b and X{circumflex over (β)}.sub.2.sup.b obtained in each of the bootstrap samples b (b=1, . . . , 1000) were then computed for each observation of the original data set, and another bivariable regression model was fitted to the original data set including X{circumflex over (β)}.sub.1.sup.b and X{circumflex over (β)}.sub.2.sup.b as the only predictors, estimating coefficients {circumflex over (γ)}.sub.1.sup.b and {circumflex over (γ)}.sub.2.sup.b. Finally, a two-sided p-value for testing the null hypothesis of no added value of M2 on top of M1 was obtained by applying the ‘twice smaller tail’ rule to the empirical distribution of the 1000 values of {circumflex over (γ)}.sub.2.sup.b. In particular, the p-value was calculated as
where I{⋅}=1 if condition {⋅} is true and 0 otherwise. Similarly, M3 and M4 were tested against M2. This procedure modifies the one proposed by de Bin et al. to be applicable in internal validation. The non-parametric bootstrap test was described by Efron and Tibshirani. R Version 3.3.2 (R development core team, Vienna, 2017) and the glmnet R package for fitting the ridge was used.
TABLE-US-00001 TABLE 1 Patient baseline characteristics of total study population (n = 96). Total study population (n = 96) Age, yr, median (IQR) 58 (49-69) Female sex (%) 22 (22.9) Cardiac arrest witnessed 79 (82.3) Reason for cardiac arrest (%) Cardiac, n (%) 83 (86.5) Pulmonal, n (%) 11 (11.5) Unknown, n (%) 2 (2.1) First monitored rhythm (%) Shockable rhythm Ventricular fibrillation, n (%) 69 (71.9) Ventricular tachycardia, n (%) 1 (1.0) Asystole, n (%) 9 (9.4) Pulseless electrical activity, n (%) 13 (13.5) Unknown first rhythm, n (%) 4 (4.2) Time from cardiac arrest to event - in minutes Start of life support, min, 0 (0-8) median (min-max) Return of spontaenous circulation, 27.00 (15.75-40.75) min, median (IQR) Administration of epinephrine, 12.00 (9.00-15.00) min, median (IQR) Dose of epinephrine administered, 3.50 (2.00-5.00) mg, median (IQR) Mode of cooling Invasive, n (%) 71 (74.0) Non-invasive, n (%) 13 (13.5) Combined, n (%) 8 (8.3) Unknown, n (%) 4 (4.2) Medical history Hypertension, n (%) 61 (63.5) History of smoking, n (%) 59 (61.5) Diabetes, n (%) 81 (84.4) Acute myocardial infarction, n (%) 80 (83.3) COPD, n (%) 86 (89.6) Laboratory values at admission Potassium mmol/l, median (IQR) 3.75 (3.32-4.40) Calcium mmol/l, median (IQR) 1.16 (1.11-1.25) pO2 mmHg, median (IQR) 145.50 (90.58-300.25) pCO2 mmHg, median (IQR) 55.10 (44.62-63.35) pH, median (IQR) 7.16 (7.03-7.24) Bicarbonate mmol/l, median (IQR) 15.70 (13.12-18.28) Base excess mmol/l, median (IQR) −9.10 (−12.85-−5.70) Lactate mmol/l, median (IQR) 7.20 (4.90-9.80) Blood glucose mg/dl, median (IQR) 288 (229-349) Hematocrit, %, median (IQR) 42.50 (38.75-45.60) Chloride, mmol/l, median (IQR) 99.00 (97.00-102.00) Creatinine, mg/dl, median (IQR) 1.27 (1.05-1.53) Blood Urea Nitrate, mg/dl, median (IQR) 15.10 (12.00-19.90) Uric Acid mg/dl, median (IQR) 7.80 (6.70-9.10) Total bilirubin, mg/dl, median (IQR) 0.35 (0.24-0.56) Troponin T, ng/L, median (IQR) 0.07 (0.03-0.20) NT-proBNP, pg/ml, median (IQR) 441 (113-1722) S100B, μg/l, median (IQR) 0.10 (0.07-0.20) Neuron specific enolase, μg/l, median (IQR) 30.65 (17.95-72.53)
TABLE-US-00002 TABLE 2 Final multiple reaction monitoring (MRM) method. List of proteins UniProt Protein Gene Uniprot protein ID name name entry name O00299 Chloride intracellular CLIC1 CLIC1 channel protein 1 P00338 L-lactate dehydrogenase LDHA LDHA A chain P00488 Coagulation factor XIII F13A1 F13A A chain P00558 Phosphoglycerate kinase 1 PGK1 PGK1 P02751 Fibronectin FN1 FINC P06733 Alpha-enolase ENO1 ENOA P07195 L-lactate dehydrogenase LDHB LDHB B chain P07737 Profilin-1 PFN1 PROF1 P11142 Heat shock cognate 71 kDa HSPA8 HSP7C protein P18206 Vinculin VCL VINC P23528 Cofilin-1 CFL1 COF1 P63104 14-3-3 protein zeta/delta YWHAZ 1433Z Q9Y490 Talin-1 TLN1 TLN1 P06702 Protein S100-A9 S100A9 S10A9 P05362 Intercellular adhesion ICAM1 ICAM1 molecule 1 P55290 Cadherin-13 CDH13 CAD13 P02649 Apolipoprotein E APOE APOE P04040 Catalase CAT CATA P10643 Complement component C7 C7 CO7 P04275 von Willebrand factor VWF VWF P12814 Alpha-actinin-1 ACTN1 ACTN1 Q96KN2 Beta-Ala-His dipeptidase CNDP1 CNDP1 P04406 Glyceraldehyde-3-phosphate GAPDH G3P dehydrogenase P05109 Protein S100-A8 S100A8 S10A8
TABLE-US-00003 TABLE 3 Final multiple reaction monitoring (MRM) method. List of peptides and transitions. UniProt Precursor Precursor Product Product Fragment Collision ID Peptide sequence m/z charge m/z charge ion energy O00299 NSNPALNDNLEK 664.83 2 1013.53 1 y9 19.1 O00299 NSNPALNDNLEK 664.83 2 732.35 1 y6 19.1 O00299 NSNPALNDNLEK 664.83 2 507.27 2 y9 19.1 O00299 GFTIPEAFR 519.27 2 833.45 1 y7 13.9 O00299 GFTIPEAFR 519.27 2 732.4 1 y6 13.9 O00299 GFTIPEAFR 519.27 2 619.32 1 y5 13.9 O00299 YLSNAYAR 479.24 2 794.42 1 y7 12.5 O00299 YLSNAYAR 479.24 2 681.33 1 y6 12.5 O00299 YLSNAYAR 479.24 2 594.3 1 y5 12.5 P00338 LVIITAGAR 457.3 2 701.43 1 y7 11.7 P00338 LVIITAGAR 457.3 2 588.35 1 y6 11.7 P00338 FIIPNVVK 465.29 2 669.43 1 y6 12 P00338 FIIPNVVK 465.29 2 556.35 1 y5 12 P00488 GTYIPVPIVSELQSGK 844.47 2 1253.71 1 y12 25.6 P00488 GTYIPVPIVSELQSGK 844.47 2 1057.59 1 y10 25.6 P00488 STVLTIPEIIIK 663.92 2 1039.68 1 y9 19.1 P00488 STVLTIPEIIIK 663.92 2 926.59 1 y8 19.1 P00488 STVLTIPEIIIK 663.92 2 712.46 1 y6 19.1 P00558 ACANPAAGSVILLENLR 884.97 2 1352.79 1 y13 27.1 P00558 ACANPAAGSVILLENLR 884.97 2 1184.7 1 y11 27.1 P00558 ACANPAAGSVILLENLR 884.97 2 1113.66 1 y10 27.1 P02751 DLQFVEVTDVK 646.84 2 936.5 1 y8 18.5 P02751 DLQFVEVTDVK 646.84 2 690.37 1 y6 18.5 P02751 DLQFVEVTDVK 646.84 2 462.26 1 y4 18.5 P02751 IYLYTLNDNAR 678.35 2 1079.55 1 y9 19.6 P02751 IYLYTLNDNAR 678.35 2 966.46 1 y8 19.6 P02751 IYLYTLNDNAR 678.35 2 589.27 1 y6 19.6 P06733 YISPDQLADLYK 713.37 2 1149.58 1 y10 20.9 P06733 YISPDQLADLYK 713.37 2 1062.55 1 y9 20.9 P06733 VNQIGSVTESLQACK 817.41 2 1179.57 1 y11 24.6 P06733 VNQIGSVTESLQACK 817.41 2 936.45 1 y8 24.6 P06733 VNQIGSVTESLQACK 817.41 2 506.24 1 y4 24.6 P07195 FIIPQIVK 479.31 2 697.46 1 y6 12.5 P07195 FIIPQIVK 479.31 2 584.38 1 y5 12.5 P07737 SSFYVNGLTLGGQK 735.88 2 986.56 1 y10 21.7 P07737 SSFYVNGLTLGGQK 735.88 2 887.49 1 y9 21.7 P07737 SSFYVNGLTLGGQK 735.88 2 773.45 1 y8 21.7 P07737 STGGAPTFNVTVTK 690.36 2 1191.64 1 y12 20.1 P07737 STGGAPTFNVTVTK 690.36 2 1006.56 1 y9 20.1 P07737 STGGAPTFNVTVTK 690.36 2 503.78 2 y9 20.1 P11142 DAGTIAGLNVLR 600.34 2 855.54 1 y8 16.8 P11142 DAGTIAGLNVLR 600.34 2 742.46 1 y7 16.8 P11142 DAGTIAGLNVLR 600.34 2 671.42 1 y6 16.8 P18206 AVAGNISDPGLQK 635.34 2 744.39 1 y7 18.1 P18206 AVAGNISDPGLQK 635.34 2 542.33 1 y5 18.1 P18206 SFLDSGYR 472.73 2 710.35 1 y6 12.2 P18206 SFLDSGYR 472.73 2 597.26 1 y5 12.2 P18206 SFLDSGYR 472.73 2 482.24 1 y4 12.2 P23528 AVLFCLSEDK 591.3 2 1011.48 1 y8 16.5 P23528 AVLFCLSEDK 591.3 2 898.4 1 y7 16.5 P63104 SVTEQGAELSNEER 774.86 2 1132.52 1 y10 23.1 P63104 SVTEQGAELSNEER 774.86 2 1004.46 1 y9 23.1 P63104 SVTEQGAELSNEER 774.86 2 634.28 1 y5 23.1 Q9Y490 GLAGAVSELLR 543.32 2 915.53 1 y9 14.8 Q9Y490 GLAGAVSELLR 543.32 2 844.49 1 y8 14.8 Q9Y490 GLAGAVSELLR 543.32 2 617.36 1 y5 14.8 P06702 LGHPDTLNQGEFK 485.91 3 722.35 1 y6 12.7 P06702 LGHPDTLNQGEFK 485.91 3 480.25 1 y4 12.7 P05362 LLGIETPLPK 540.84 2 854.5 1 y8 14.7 P05362 LLGIETPLPK 540.84 2 684.39 1 y6 14.7 P05362 DLEGTYLCR 563.76 2 769.37 1 y6 15.5 P05362 DLEGTYLCR 563.76 2 611.3 1 y4 15.5 P05362 DLEGTYLCR 563.76 2 448.23 1 y3 15.5 P55290 SIVVSPILIPENQR 782.96 2 1166.65 1 y10 23.4 P55290 SIVVSPILIPENQR 782.96 2 1079.62 1 y9 23.4 P55290 SIVVSPILIPENQR 782.96 2 643.32 1 y5 23.4 P02649 SELEEQLTPVAEETR 865.93 2 1015.54 1 y9 26.4 P02649 SELEEQLTPVAEETR 865.93 2 902.46 1 y8 26.4 P02649 SELEEQLTPVAEETR 865.93 2 801.41 1 y7 26.4 P02649 AATVGSLAGQPLQER 749.4 2 898.47 1 y8 22.2 P02649 AATVGSLAGQPLQER 749.4 2 827.44 1 y7 22.2 P02649 AATVGSLAGQPLQER 749.4 2 642.36 1 y5 22.2 P04040 FNTANDDNVTQVR 747.35 2 1131.54 1 y10 22.1 P04040 FNTANDDNVTQVR 747.35 2 1060.5 1 y9 22.1 P04040 FNTANDDNVTQVR 747.35 2 831.43 1 y7 22.1 P10643 SSGWHFVVK 523.77 2 872.48 1 y7 14.1 P10643 SSGWHFVVK 523.77 2 629.38 1 y5 14.1 P10643 SCVGETTESTQCEDEELEHLR 837.02 3 1329.57 1 y10 25.3 P10643 SCVGETTESTQCEDEELEHLR 837.02 3 425.26 1 y3 25.3 P04275 LSEAEFEVLK 582.81 2 1051.53 1 y9 16.2 P04275 LSEAEFEVLK 582.81 2 835.46 1 y7 16.2 P04275 LSEAEFEVLK 582.81 2 764.42 1 y6 16.2 P04275 EGGPSQIGDALGFAVR 787.4 2 1018.57 1 y10 23.5 P04275 EGGPSQIGDALGFAVR 787.4 2 905.48 1 y9 23.5 P04275 EGGPSQIGDALGFAVR 787.4 2 549.31 1 y5 23.5 P12814 DDPLTNLNTAFDVAEK 881.93 2 994.48 1 y9 26.9 P12814 DDPLTNLNTAFDVAEK 881.93 2 880.44 1 y8 26.9 P12814 DDPLTNLNTAFDVAEK 881.93 2 708.36 1 y6 26.9 Q96KN2 SVVLIPLGAVDDGEHSQNEK 703.03 3 1385.59 1 y13 20.5 Q96KN2 SVVLIPLGAVDDGEHSQNEK 703.03 3 1158.47 1 y10 20.5 Q96KN2 SVVLIPLGAVDDGEHSQNEK 703.03 3 1043.44 1 y9 20.5 Q96KN2 WNYIEGTK 505.75 2 824.41 1 y7 13.4 Q96KN2 WNYIEGTK 505.75 2 710.37 1 y6 13.4 P04406 GALQNIIPASTGAAK 706.4 2 815.46 1 y9 20.6 P04406 GALQNIIPASTGAAK 706.4 2 702.38 1 y8 20.6 P05109 LLETECPQYIR 711.36 2 1195.54 1 y9 20.8 P05109 LLETECPQYIR 711.36 2 836.41 1 y6 20.8 global VLETKSLYVR 403.24 3 554.82 2 y9 9.7 standards global VLETK[+42]SLYVR 625.36 2 908.52 1 y7 17.7 standards global TTPAVLDSDGSYFLYSK 932.45 2 831.41 2 y15 28.8 standards
TABLE-US-00004 TABLE 4 (Log2) and interquartile ranges (IQR) of selected biomarker candidates in plasma samples of the validation cohort measured with multiple reaction monitoring analysis. Median Median intensity intensity (Log2) IQR (Log2) IQR negative negative positive positive UniProt Protein neurological neurological neurological neurological ID name outcome outcome outcome outcome P63104 14-3-3 protein 9.59 8.24-10.36 8.34 7.07-9.48 zeta/delta P12814 Alpha-actinin-1 8.13 7.42-8.66 7.80 6.96-8.65 P02649 Apolipoprotein 16.34 15.86-16.85 16.36 15.69-16.84 E P55290 Cadherin-13 9.49 8.98-9.87 9.65 9.18-10.02 P04040 Catalase 8.36 7.61-8.81 7.91 7.63-8.69 O00299 Chloride 10.63 9.55-11.46 9.83 9.06-10.98 intracellular channel protein 1 Q96KN2 Beta-Ala-His 9.94 9.06-10.45 9.98 9.61-10.32 dipeptidase P10643 Complement 12.10 11.41-12.43 12.13 11.66-12.38 component C7 P23528 Cofilin-1 10.85 9.68-11.52 9.80 8.26-11.09 P06733 Alpha-enolase 9.18 8.39-10.01 8.44 7.37-9.19 P00488 Coagulation 11.10 10.63-11.67 11.15 10.76-11.78 factor XIII A chain P02751 Fibronectin 14.09 13.37-14.69 14.19 13.1-14.63 P04406 Glycer- 12.22 11.2-13.47 11.36 10.26-12.59 aldehyde-3- phosphate dehydrogenase P11142 Heat shock 10.51 9.94-11.33 10.12 8.83-10.67 cognate 71 kDa protein P05362 Intercellular 11.16 10.73-11.48 11.24 10.81-11.47 adhesion molecule 1 P00338 L-lactate 12.21 11.58-12.88 11.83 11.15-12.55 dehydrogenase A chain P07195 L-lactate 13.91 12.83-14.69 13.43 12.68-14.96 dehydrogenase B chain P00558 Phosphoglycerate 8.11 7.55-8.88 7.87 7.07-8.44 kinase 1 P07737 Profilin-1 10.60 9.79-11.66 9.88 8.87-11.18 P05109 Protein S100- 11.99 11.28-12.62 11.43 10.95-12.21 A8 P06702 Protein S100- 13.97 13.32-14.5 13.59 13.03-13.96 A9 Q9Y490 Talin-1 11.26 10.19-12.07 10.56 8.57-11.67 P18206 Vinculin 11.82 11.08-12.56 11.42 10.69-12.15 P04275 von Willebrand 13.82 13.44-14.06 13.59 13.37-14.04 factor
TABLE-US-00005 TABLE 5 Univariate regression analysis of all biomarker candidates to predict negative neurological outcome. UniProt Protein Odds 95% CI ID.sup.a name Coefficient ratio for OR P63104 14-3-3 protein zeta/delta 0.4757 1.6092 1.1971-2.2382 P12814 Alpha-actinin-1 0.3307 1.3920 0.9609-2.0694 P06733 Alpha-enolase 0.6529 1.9210 1.2942-3.0044 P02649 Apolipoprotein E −0.0307 0.9697 0.5915-1.5609 Q96KN2 Beta-Ala-His dipeptidase −0.0352 0.9654 0.5907-1.5460 P55290 Cadherin-13 −0.1486 0.8619 0.4592-1.4147 P04040 Catalase 0.2030 1.2251 0.8236-1.8741 O00299 Chloride intracellular channel 0.3891 1.4756 1.0099-2.2169 protein 1 P00488 Coagulation factor XIII A chain 0.2024 1.2244 0.7627-2.0094 P23528 Cofilin-1 0.3959 1.4857 1.1208-2.0391 P10643 Complement component C7 −0.0070 0.9930 0.5578-1.7050 P02751 Fibronectin 0.0516 1.0529 0.7198-1.5238 P04406 Glyceraldehyde-3-phosphate 0.3927 1.4810 1.0750-2.0986 dehydrogase P11142 Heat shock cognate 71 kDa 0.6375 1.8917 1.2507-3.0154 protein P05362 Intercellular adhesion molecule 1 0.0901 1.0943 0.6353-1.8424 P00338 L-lactate dehydrogenase A chain 0.4638 1.5901 1.0234-2.6641 P07195 L-lactate dehydrogenase B chain 0.2027 1.2247 0.8921-1.7279 P00558 Phosphoglycerate kinase 1 0.5121 1.6689 1.0580-2.8926 P07737 Profilin-1 0.4148 1.5140 1.0911-2.1646 P05109 Protein S100-A8 0.2526 1.2874 0.8872-1.9117 P06702 Protein S100-A9 0.2709 1.3111 0.8782-2.0198 Q9Y490 Talin-1 0.3234 1.3818 1.0518-1.8500 P18206 Vinculin 0.4858 1.6256 1.0312-2.6682 .sup.aSwiss-Prot (UniProt) accession numbers are provided.
TABLE-US-00006 TABLE 6 Model 1, linear prediction model for negtive neurological outcome consisting of established clinical predictors of neurological outcome Absolute UniProt Standardized Parameters ID Coefficient coefficient Intercept 0.48465 Time from cardiac arrest 0.88147 0.39245 to start of life support (not immediate vs. immediate) Age (per year) 0.02693 0.36597 Shockable rhythm (yes vs. no) −0.80759 0.36436 Maximum absolute difference 1.14039 0.19586 of pH to 7.3 pH within the first 24 hours Hemoglobin (g/dl) −0.05861 0.10948 Time from cardiopulmonary −0.07873 0.09728 resuscitation to the return of spontaneous circulation (per doubling, minutes) .sup.b Maximum serum lactate within 0.00909 0.03826 the first 24 hours (mmol/L)
TABLE-US-00007 TABLE 7 Model 2, prediction model for negative neurological outcome consisting of potential predictors of neurological outcome estimated by a LASSO logistic regression model Absolute Bootstrap UniProt Gene Standardized inclusion Parameters ID name Coefficient coefficient frequency Intercept 1.04935 .sup.a Age (per year) 0.02693 0.36597 .sup.a Shockable rhythm (yes −0.80759 0.36436 .sup.a vs. no) Time from cardiac arrest 0.88147 0.39245 .sup.a to start of life support (not immediate vs. immediate) Hemoglobin (g/dl) −0.05861 0.10948 .sup.a Time from −0.07873 0.09728 .sup.a cardiopulmonary resuscitation to the return of spontaneous circulation (per doubling, minutes).sup.b Maximum serum lactate 0.00909 0.03826 .sup.a within the first 24 hours (mmol/L) Maximum absolute 1.14039 0.19586 .sup.a difference of pH to 7.3 pH within the first 24 hours S-100B (per doubling).sup.b P04271 S100B 0.19187 0.36826 81.9 Adrenalin administrated 0.00000 0.00000 41.3 during Cardiopulmonary resuscitation (mg) Neuron-specific enolase P09104 ENO2 0.00000 0.00000 36.5 (per doubling).sup.a Sex (female vs. male) 0.00000 0.00000 27.5 Unwitnessed Cardiac 0.00000 0.00000 15.2 arrest (yes vs. no) .sup.aThese predictors were always included and not subjected to variable selection, .sup.bMeasured by electrochemiluminescence immunoassay.
TABLE-US-00008 TABLE 8 Model 3, proteomics-enriched prediction model for negative neurological outcome estimated by LASSO logistic regression. Absolute Bootstrap UniProt Gene Standardized inclusion Parameters ID name Coefficient coefficient frequency Intercept −1.86718 .sup.a Age (per year) 0.02693 0.36597 .sup.a Shockable rhythm (yes −0.80759 0.36436 .sup.a vs. no) Time from cardiac 0.88147 0.39245 .sup.a arrest to start of life support (not immediate vs. immediate) Hemoglobin (g/dl) −0.05861 0.10948 .sup.a Time from −0.07873 0.09728 .sup.a cardiopulmonary resuscitation to the return of spontaneous circulation (per doubling, minutes).sup.b Maximum serum lactate 0.00909 0.03826 .sup.a within the first 24 hours (mmol/L) Maximum absolute 1.14039 0.19586 .sup.a difference of pH to 7.3 pH within the first 24 hours S-100B (per doubling).sup.c P04271 S100B 0.10194 0.19565 79.8 Alpha-enolase (per P06733 ENO1 0.16576 0.21045 55.1 doubling).sup.d 14-3-3 protein P63104 YWHAZ 0.09157 0.14942 60.1 zeta/delta (per doubling).sup.d Cofilin-1 (per doubling).sup.d P23528 CFL1 0.03819 0.06493 58.5 .sup.aThese predictors were always included and not subjected to variable selection. .sup.blog2(x + 1) transformation was applied. .sup.cMeasured by electrochemiluminescence immunoassay. .sup.dMeasured by Targeted LC-MRM analysis. Swiss-Prot (UniProt) accession numbers are provided.
TABLE-US-00009 TABLE 9 Model 4, proteomics-enriched prediction model for negative neurological outcome estimated by elastic-net logistic regression Absolute Bootstrap UniProt Gene Standardized inclusion Parameters ID name Coefficient coefficient frequency Intercept −1.72579 .sup.a Age (per year) 0.02693 0.36597 .sup.a Shockable rhythm (yes −0.80759 0.36436 .sup.a vs. no) Time from cardiac 0.88147 0.39245 .sup.a arrest to start of life support (not immediate vs. immediate) Hemoglobin (g/dl) −0.05861 0.10948 .sup.a Time from −0.07873 0.09728 .sup.a cardiopulmonary resuscitation to the return of spontaneous circulation (per doubling, minutes).sup.b Maximum serum lactate 0.00909 0.03826 .sup.a within the first 24 hours (mmol/L) Maximum absolute 1.14039 0.19586 .sup.a difference of pH to 7.3 pH within the first 24 hours S-100B (per doubling).sup.b P04271 S100B 0.07509 0.14411 86.5 Alpha-enolase (per P06733 ENO1 0.11104 0.14098 81.4 doubling).sup.c 14-3-3 protein P63104 YWHAZ 0.07876 0.12851 78.9 zeta/delta (per doubling).sup.c Cofilin-1 (per doubling).sup.c P23528 CFL1 0.05262 0.08946 81.5 Heat shock cognate 71 P11142 HSPA8 0.02181 0.02595 46.1 kDa protein (per doubling).sup.c .sup.aThese predictors were always included and not subjected to variable selection. .sup.blog2(x + 1) transformation was applied. .sup.cMeasured by electrochemiluminescence immunoassay. .sup.dMeasured by Targeted LC-MRM analysis. Swiss-Prot (UniProt) accession numbers are provided.
TABLE-US-00010 TABLE 10 Summary of models 1-4 for predicting negative neurological outcome using established and new biomarkers Explained p-value for variation in ‘added Model Method Predictors C-index % value’ Model 1 Ridge Fixed: Age, shockable rhythm, 0.67 8.2% 0.002.sup.a regression time from OHCA to CPR (not immediate vs. immediate), time from CPR to ROSC, hemoglobin, lactate, pH Model 2 Lasso Fixed: Linear predictor from M1 0.68 10.0% 0.147.sup.b regression Candidates: doses of administrated adrenalin, sex, witnessed OHCA, S100B, NSE Selected: S100B Model 3 Lasso Fixed: Linear predictor from M1 0.70 12.1% 0.025.sup.c regression Candidates: doses of administrated adrenalin, sex, witnessed OHCA, S100B, NSE, all proteins measured by Targeted LC-MRM analysis Selected: S100B, Alpha- enolase, 14-3-3 protein zeta/delta, Cofilin-1 Model 4 Elastic-net Fixed: Linear predictor from M1 0.70 11.9% 0.019.sup.c regression Candidates: Doses of administrated adrenalin, sex, witnessed OHCA, S100B, NSE, all proteins measured by Targeted LC-MRM analysis Selected: S-100B, Alpha- enolase, 14-3-3 protein zeta/delta, Cofilin-1, Heat shock cognate 71 kDa protein .sup.acompared to null model .sup.bcompared to Model 1 .sup.ccompared to Model 2
TABLE-US-00011 TABLE 11 Calculated percentage changes relative to the median levels of each biomarker that allow prediction of a positive or negative neurological outcome. Percentage change Percentage change Gene for positive for negative name neurological outcome neurological outcome Alpha-enolase 31% (sensitivity: 0.15, 344% (sensitivity: 0.23, specificy: 0.98) specificy: 1) Heat shock 26% (sensitivity: 0.06, 277% (sensitivity: 0.23, cognate 71 kDa specificy: 1) specificy: 1) protein 14-3-3 protein 24% (sensitivity: 0.15, 372% (sensitivity: 0.29, zeta/delta specificy: 0.98) specificy: 0.97) Cofilin-1 18% (sensitivity: 0.12, 418% (sensitivity: 0.21, specificy: 1) specificy: 0.97)
Example 2—Results
[0198] Baseline characteristics: One-hundred OHCA patients were included with a median age of 58 years (IQR, 49-69) admitted to the emergency department of the Medical University of Vienna between October 2013 and May 2016. Four patients were excluded because of poor specimen quality. Hence, the final study population consisted of 96 patients, of whom 63 patients (66 percent) had a poor neurological outcome at discharge from the ICU. CPC 1 was recorded in 26 patients (27%), CPC 2 in 7 (7%), CPC 3 in 24 (23%), CPC 4 in 16 (15%), and CPC 5 in 23 (22%) of OHCA survivors. Detailed baseline characteristics are displayed in Table 1.
[0199] Identification of biomarker candidates: The flow chart of proteomic analyses is illustrated in
[0200] Prognostic model: The results of a univariate regression analysis of each of the 24 biomarker candidates to predict poor neurological outcome are outlined in Table 5. To explore the best multivariable combination of clinical parameters and biomarkers to predict neurological outcome in OHCA patients, a multi-stage statistical approach was applied. M1, including all well-established neurological outcome predictors, achieved a discriminative ability (cross-validated c-index) of 0.67. This c-index increased to 0.68 by the variable selection process of M2, which added S100-B. M3 additionally selected a multivariable combination of the proteomically measured biomarker candidates (n=24) and yielded a c-index of 0.70 (Table 9). This optimal combination, consisting of alpha-enolase, 14-3-3 protein zeta/delta and cofilin-1, resulted in a significant improvement in neurological outcome prediction compared to M2 (p=0.025 for added value). M4, which employed elastic-net as a less restrictive variable selection procedure, additionally selected Heat shock cognate 71 kDa protein (HSPA8), yielding similar performance (c-index=0.70; p=0.019 for added value compared to M2, Table 3). Explained variation of the outcome could be improved from 8.2% in M1 to 10.0% in M2 and to 12.2% in M3 and to 11.9% in M4 (Table 4).
Example 3—Summary and Discussion
[0201] The present study identified a comprehensive panel of novel biomarkers for the prediction of poor neurological outcome in OHCA survivors using a three-step proteomics strategy. The implementation of proteomically identified alpha-enolase, and optionally further one or more of 14-3-3 protein zeta/delta cofilin-1 and HSPA8 into the clinical model resulted in a significant improvement of established neurological outcome prediction. These biomarkers were carefully retrieved out of a total of 299 plasma proteins measured in successfully resuscitated patients 48 hours after cardiac arrest using a proteomics shotgun analysis, crosschecked in brain tissue samples and verified using targeted proteomic analyses in combination with a stepwise statistical approach including penalization for the number of potential proteomic candidates and optimism-corrected assessment of discrimination and calibration.
[0202] The central nervous system is protected by the blood-brain barrier, composing of brain vascular endothelial cells and associated astrocytes, that restricts molecular exchange between the brain and blood vessels to regulate the cerebral homeostasis. Nevertheless, the detection of brain-derived proteins in circulating blood is feasible in patients undergoing cardiac arrest, as hypoxia triggers disruption of the blood-brain barrier. In the present study, all identified biomarkers were not only detected in circulating plasma but also in brain tissue samples. Even though these proteins are not specifically expressed by brain tissue, a causal relationship with neurological recovery is supported by their biological functions and processes in which they are implicated. In addition to that, the prior art biomarkers NSE and S100B are not specific to neuronal damage as well and have been found produced by extra-central nervous system sources.
[0203] In the art prediction of neurological outcome in OHCA survivors remains a clinical challenge. Neurophysiological tests are recommended by current guidelines to be performed in individuals not regaining consciousness within two days after the event. Technically demanding, these tests are not readily available in the majority of hospitals. Early, reliable prognosis within the first days following ROSC is often impossible based on currently available models including clinical assessment and biomarkers (e.g. NSE, S100)—leaving clinicians and relatives in uncertainty.
[0204] This study considerably extends the limited knowledge in cognitive outcome prediction of OHCA survivors using inter alia a state-of-the-art large-scale MS based proteomics strategy. For the first time alpha-enolase, 14-3-3 protein zeta/delta, and/or cofilin-1 were revealed as additional robust predictors, for example, e to be used in a multivariable model of poor or good neurological outcome in an OHCA study population. In fact, the incorporation of alpha-enolase, 14-3-3 protein zeta/delta, cofilin-1 and/or HSPAB into a multimarker model provided a clear improvement in predictive accuracy and discriminatory power for poor neurological outcome.
[0205] The poor neurological outcome in 66% of our OHCA study population, which is in the range of other studies highlights the continuing clinical need to estimate neurological recovery early after cardiac arrest. A timely and accurate prediction would essentially help clinicians in their decision-making processes for evaluating further therapeutic strategies or withdrawal of therapies. The proteomics-enriched prediction model described herein consists exclusively of plasma biomarkers and cardiopulmonary resuscitation (CPR) related factors and could therefore be reliably calculated within 48 hours after cardiac arrest. By using this model, prediction is advantageously delayed due to outstanding examinations that are not available 24/7.
[0206] Therapeutic hypothermia and concomitant administration of drugs for sedation and muscle paralysis after cardiac arrest crucially complicates and delays prognostication by affecting not only the assessment of clinical reactivity but also the interpretation of neurophysiological tests. A valid prognostication is further delayed as therapeutic hypothermia prolongs the action of both sedative agents and muscle relaxants by slowing their clearance. Biomarkers are presumably not affected by sedation and may therefore provide more reliable objective prognostic information in the early post-resuscitation phase. As currently available biomarkers of neurological recovery are greatly limited in their prognostic value, the four biomarkers revealed in connection with the present invention represent a valuable addition to established risk predictors.
[0207] In summary, the study presented in the examples created a superior early prediction model for poor or good neurological outcome in OHCA survivors. A significant improvement in neurological outcome prediction was found based on alpha-enolase, 14-3-3 protein zeta/delta, cofilin-1 and/or HSPA8, in particular when included into a multimarker predictive model along with established risk factors. Each of the biomarkers did not only refine current risk stratification models but may also reflect important pathophysiological pathways undergoing during cerebral ischemic.