RU2748685C1 - Method for predicting the probability of fatal outcome in 14-day period of acute ischemic stroke of the brain - Google Patents

Abstract

FIELD: medicine; neurology.

SUBSTANCE: invention relates to medicine, namely to neurology, it can be used to predict the probability of a fatal outcome in the 14-day period of an acute ischemic stroke of the brain. The serum concentration of the neuron-specific protein S100-β in the venous blood is determined on the 2nd day of ischemic stroke and the level of patient’s consciousness is evaluated according to the Glasgow coma scale (hereinafter – GCS). The probable prognosis of a fatal outcome is calculated by the formula Y = 1/(1 + e ^{- (16,703-1,395* GCS + 0.123* S100-β)}), where Y is the probability of a fatal outcome; e is a mathematical constant equal to 2.71828; GCS is the level of patient’s consciousness on the Glasgow coma scale, points; S100- β is serum concentration of neuron-specific protein, mcg/l. If the Y value is equal to or greater than 0.1, a high probability of a fatal outcome and an unfavorable prognosis for life are predicted. If the Y value is less than 0.1, it means a low probability of a fatal outcome and a favorable prognosis for life.

EFFECT: method makes it possible to increase the accuracy of predicting a favorable and/or unfavorable outcome for life in the first 14 days of the acute period of ischemic stroke by determining clinical and laboratory markers of brain tissue damage: serum concentration of the neuron-specific protein S100-β and assessing the level of patient's consciousness on the Glasgow coma scale (GCS), creating a logistic regression model – a new simple and highly informative mathematical prognostic model that allows stratifying patients into a group with a high risk of a fatal outcome in the specified time period.

1 cl, 2 dwg, 3 tbl, 2 ex

Description

The invention relates to medicine, namely to neurology, and relates to predicting the outcome (lethal or favorable for life) in the first 14 days of acute ischemic stroke by determining the serum concentration of neuron-specific protein S100-β on the 2nd day of ischemic stroke and assessing the patient's level of consciousness on the Glasgow Coma Scale.

The known method predicting the outcome of ischemic stroke by definition in the peripheral venous blood of the average level of CD95-lymphocytes in% on the 1st and 21st days from the onset of ischemic stroke. If the level of CD95-lymphocytes on the 1st day is more than 7.2 ± 2.8% and a return to normal on the 21st day, a favorable outcome is predicted. In cases where the level of CD95 lymphocytes is higher than 7.2 ± 2.8% and the average level of the indicator remains on the 21st day, a favorable prognosis for life and unfavorable for the recovery of neurological deficit is predicted. If the average CD95-lymphocyte count is above 29.12 ± 3.1%, an unfavorable prognosis for life is predicted [RU 2236685]. The disadvantages of the analogue related to the technical problem being solved are its technically difficult implementation and the need for a highly qualified specialist, which reduces the possibility of widespread use of the method, as well as the fact that the technique allows predicting the outcome of cerebral ischemia already at the end of the acute period (after the 21st day ), whereas from a clinical point of view, it is most important to determine the probability of mortality in the acutest period of stroke (the first 2-4 days) for the choice of treatment tactics and early rehabilitation of the patient immediately after the onset of the disease.

There is also known a method predicting the outcome of ischemic stroke according to the patient's assessment on the NIH scale and CT perfusion data [RU 2585139]. The total score on the NIH stroke scale is assessed and CT perfusion of the brain is performed on the first day of the acute period of the disease. During CT perfusion, the total ischemic area is determined, consisting of the infarction area and the area of the penumbra, as well as the cerebral blood flow in the area of the penumbra. When you receive a total score on the NIH stroke scale of more than 12, the total ischemic area is more than 3170 mm²and the level of decreased cerebral blood flow (CBF) in the penumbra less than 24.3 ml / 100 g / min predict a severe functional outcome of acute ischemic stroke. The disadvantage of this method is its low information content in relation to patients in whom the ischemic area will be less than 3170 mm²and the level of decreased cerebral blood flow (CBF) in the penumbra is more than 24.3 ml / 100 g / min. Undoubtedly, the area of ischemia and the level of decreased cerebral blood flow (CBF) in the penumbra are associated with the severity of ischemic stroke, but the localization of the lesion is no less important. For example, patients with ischemic stroke in the vertebrobasilar basin and a relatively small focus of ischemia in the brain stem area often have a poor prognosis.

The closest to the claimed technical solution in terms of the technical essence and the achieved technical result is a method for predicting the outcome of ischemic stroke by determining the ligand level of a soluble member of the tumor necrosis factor receptor superfamily (sFasL) and soluble Fas receptor (sFas) (sFasL / sFas) in the patient's blood serum. If the concentration ratio sFasL / sFas is less than or equal to 2.41 ± 0.26, a favorable outcome of ischemic stroke is predicted, if greater than 2.67, an unfavorable outcome is predicted [RU 2517471]. The described method is taken as a prototype of the invention. Disadvantages of the method: the given biomarkers (sFasL / sFas) do not reflect specific damage to the brain tissue and increase with the development of other pathological conditions, in particular, in a number of oncological diseases, myocardial infarction and any systemic inflammatory process. In our proposed method for predicting the outcome (lethal or favorable for life) of acute ischemic stroke, neuron-specific protein S100-β is used as a laboratory biomarker. It is not only a glial biomarker, but also a marker of generalized damage to the blood-brain barrier and damage to brain tissue, the most studied and included in laboratory diagnostics due to its neurospecificity. The content of S100-β protein in blood serum is normally less than 0.2 μg / L. The development of cerebral ischemia in the patient is evidenced by its content of more than 0.5 μg / l. World research data confirm that serum protein S100-β is a biomarker for the differential diagnosis of ischemic stroke. At the same time, the sensitivity of S100β for detecting stroke was 94.4%, the specificity was 31.8% [Purrucker JC et al (2014)]. After a stroke, an increase in S100-β concentration begins during the first 8 hours and its increase persists for 72 hours, while the S100-β concentration correlates with the extent of damage and neurological consequences of stroke [Ahmad et al., 2012. Sarhad J. Agric., 28 (1): 69-74]. The most pronounced correlation with long-term functional outcome and focus volume is observed for the concentration obtained within 48–72 hours from the onset of symptoms. Therefore, in our proposed. method, the level of the biomarker in the peripheral blood is determined on the 2nd day of stroke. In addition, the serum concentration of S100-β does not depend on age and sex, and does not change with alcohol overdose, moderate renal dysfunction, or hemolysis. Another disadvantage of the prototype is that the prediction does not take into account the data of the patient's clinical severity, such as the level of consciousness, reflecting the degree of damage with a decrease in oxygen perfusion in the brain tissues during stroke. Thus, the proposed forecasting method eliminates the disadvantages of the prototype.

The technical objective of the invention is to create a new simple and highly informative mathematical prognostic model capable of increasing the accuracy of predicting a favorable and / or unfavorable outcome for life in the first 14 days of the acute period of ischemic stroke, allowing patients to be stratified into a group with a high risk of death in the specified time period.

The task is solved by determination of clinical and laboratory markers of brain tissue damage, which include serum concentration of neuron-specific protein S100-β in peripheral venous blood on the 2nd day of ischemic stroke and assessment of the patient's level of consciousness on the Glasgow Coma Scale (GCS).

In order to predict a favorable and / or unfavorable outcome for life in the first 14 days of acute ischemic stroke, statistical modeling was performed using logistic regression analysis. The preliminary selection of predictors of the model was carried out using the ROC-analysis; to construct the model, features were used for which the AUC value (area under the ROC-curve) was significantly higher than 0.5.

When constructing the model, the data of 105 patients with acute ischemic stroke who were admitted to the hospital within the first 24 hours from the onset of acute cerebrovascular accident were considered. Of these, 25 patients with fatal outcome before the 14th day of stroke and 80 patients who survived the 14-day period. The patients did not undergo systemic thrombolysis or thromboextraction.

The set of predictors for constructing the model included the following characteristics described on the 2nd day of stroke:

- assessment of neurological deficit according to the National Institute Stroke Scale

health NIHSS (points),

- assessment of the level of consciousness on the Glasgow Coma Scale GCS (points),

- the concentration of the brain neurotrophic factor BDNF in the blood serum (pg / ml),

- concentration of nerve growth factor NGF in blood serum (pg / ml),

- concentration of neuron-specific enolase NSE (μ / hl) in blood serum,

- S100-β protein concentration (pg / ml) in blood serum,

- concentration of myelin basic protein MBP (pg / ml) in blood serum,

- concentration of glial fibrillar acidic protein GFAP (ng / ml) in blood serum.

In addition, anamnesis data were considered among potential predictors:

- gender,

- age,

- duration of hypertension (years),

- the presence of coronary heart disease,

- postponed myocardial infarction,

- atrial fibrillation,

- stenting and artificial heart valves,

- pathology of the homeostasis system,

- degree of obesity,

- dyslipidemia,

- diabetes

- the duration of diabetes mellitus (years).

The relationship of the selected characteristics with the lethal outcome was assessed using the ROC analysis (Table 1) - as potential predictors for constructing a model for predicting the lethal outcome, those characteristics for which the area under the ROC curve was significantly higher than 0.5 were selected: duration of hypertension , atrial fibrillation, dyslipidemia, assessment of the level of consciousness on the Glasgow Coma Scale GCS (points), assessment of neurological deficit on the NIHSS Stroke Scale (points), the concentration of brain neurotrophic factor BDNF in blood serum (pg / ml), the concentration of neuron-specific NSE enolases (μ / hl) in blood serum, concentration of S100-β protein (pg / ml) in blood serum, concentration of glial fibrillar acidic protein GFAP (ng / ml) in blood serum.

Based on the set of selected predictors by the method of stepwise selection of variables, a logistic regression analysis was performed in the model, which made it possible to build a statistically significant model (χ2 = 64.9, p <0.001), the ^{Nigelkerk coefficient R 2 of} which is 0.69, which indicates a sufficiently high quality of fit models.

In the process of stepwise selection of variables, the logistic regression model was considered statistically insignificant for predicting the outcome and the following of the selected features were not included in the model: duration of hypertension, atrial fibrillation, dyslipidemia, assessment of neurological deficit according to the NIHSS National Institute of Health Stroke Scale (points), concentration of brain neurotrophic factor BDNF in blood serum (pg / ml), concentration of basic myelin protein MBP (pg / ml) in blood serum, concentration of glial fibrillar acidic protein GFAP (ng / ml) in blood serum (Table 2).

The predictors GCS (p <0.001) and S100-β (p = 0.005) were significantly associated with outcome. At the same time, an increase in the number of points on the Glasgow Coma Scale GCS is associated with a decrease in the chances of a lethal outcome (OR 0.248, 95% CI OR 0.128; 0.478), and an increase in the serum concentration of S100-β protein in the peripheral blood is associated with an increase in the chances of a lethal outcome (OR 1.131, 95% CI ORS 1.039; 1.231) (Table 3).

The equation (decision rule) of the model looks like this:

The probability of death (Y) = 1 / (1 + e ^{- ( 16.703 - 1.395 * GCS + 0.123 * S100-β)} ) (1).

A lethal outcome is predicted at probability values greater than or equal to 0.5, a favorable prognosis - at probability values less than 0.5. The constructed model demonstrated a satisfactory quality of classification: sensitivity 72.0%, specificity 93.8%.

Using the ROC analysis, the cut-off point value of the calculated value of the probability of a lethal outcome was adjusted. This made it possible to improve the classification results: the new value of the cut-off point for the probability of death, equal to 0.1, corresponds to a sensitivity of 92% and a specificity of 83%. A graphical representation of the quality of a binary classifier built on the basis of a logistic regression model is shown in Fig. 1.

The way to predict the outcome in the 14-day period of acute ischemic cerebral stroke is as follows:

- Sampling of biological material (venous blood from the cubital vein into Vacuette tubes with a clot activator - SiO2) from a patient with acute ischemic stroke who was admitted to the hospital no earlier than 48 hours and no later than 72 hours from the onset of the disease (2nd day).

- As a test material, blood serum is used, obtained according to the standard method by the separation of erythrocytes by centrifugation, without traces of hemolysis.

- Determination of the concentration of neuronspecific protein S100-β is carried out by "Sandwich" - the method of enzyme-linked immunosorbent assay using DY1820-05 Human S100B DuoSet ELISA kits manufactured by R&D Systems (USA). Removal of unbound components of the reaction mixture is carried out using an automatic microplate washer WellWash (Thermo Fisher Scientific, Finland). ELISA results are evaluated on an Epoch automatic microplate spectrophotometer (BioTek Instruments, USA) at a wavelength of 450 nm. The final results are expressed in pg / ml - the units recommended by the manufacturer for constructing calibration graphs from standard weighed portions of the analyte.

- The level of consciousness in a patient with acute ischemic stroke is assessed using the Glasgow Coma Scale (GCS) in points.

- Prediction of the probability of a lethal outcome of a stroke in a 14-day period is performed using the decision rule obtained by the method of logistic regression analysis.

- The calculation of the probability of death is carried out according to the formula (1), while the values of the characteristics of the patient's level of consciousness according to the Glasgow coma scale (GCS) in points and the concentration of the neuron-specific protein S100-β, measured on the 2nd day of stroke, should be substituted into the formula:

- Probability of death (Y) = 1 / (1 + e ^{- ( 16.703 - 1.395 * GCS + 0.123 * S100-β)} ).

- The calculated value is compared with the cut-off point.

The essence of the proposed invention is illustrated in Fig. 2, which shows the ranges of values of the function of the probability of lethal outcome for the classes "lethal outcome" and "favorable outcome". Moreover, 0.1 is the point of separation of classes. The range of values above 0.1 corresponds to the “lethal outcome” class, below 0.1 - “favorable outcome”. Prediction of the outcome of a stroke in a 14-day period is performed by plotting the probability value calculated for the patient on the area of the probability function value and determining the belonging to the corresponding class.

Clinical example 1.

Patient K., male, age 56 years old, was admitted to the hospital with acute ischemic stroke of the brain in the basin of the middle cerebral artery on the left, atherothrombotic subtype. The diagnosis was confirmed by neuroimaging data from computed tomography of the brain. Risk factors for cerebrovascular diseases in this patient: hypertension 3 risk 4 within 5 years; obesity of the 1st degree; dyslipidemia. The severity of neurological disorders was assessed as mild. Neurological status on admission: right hemiparesis, motor aphasia. Assessment on clinical and functional scales: NIHSS = 3 points; mRs = 3 points.

The value of the characteristics that are predictors of the model:

GCS - the number of points on the Glasgow Coma Scale upon admission to the hospital is 15;

S100-β - the level of neuron-specific protein in the blood serum upon admission to the hospital was 1.638.

We calculate the probability of death using a logistic regression model, substituting the values of the predictors into the equation:

Death probability (Y) = 1 / (1 + e ^{- ( 16.703 - 1.395 * GCS + 0.123 * S100-β)} ) = 1 / (1 + e ^{- ( 16.703 - 1.395 * 15 + 0.123 * 1.638)} ) = 0 , 01764

The probability of a lethal outcome = 0.01764 is less than the cut-off point (0.1) - therefore, according to the model, the prognosis of the patient's survival is favorable. In reality, the patient survived.

Clinical example 2.

Patient A, female, age 62, was admitted to the hospital with acute ischemic stroke of the brain in the basin of the middle cerebral artery on the right, atherothrombotic subtype. The diagnosis was confirmed by neuroimaging data from computed tomography of the brain. Risk factors for cerebrovascular diseases in this patient: hypertension 3 risk 4 within 15 years; obesity of the 1st degree; dyslipidemia; type 2 diabetes mellitus for 10 years. The patient's condition was assessed as serious. Neurological status on admission: left hemiparesis, dysarthria. Assessment on clinical and functional scales: NIHSS = 18 points; mRs = 5 points.

The value of the characteristics that are predictors of the model:

- the number of points on the Glasgow Coma Scale on admission to hospital (GCS) is 12;

- the level of S100-β protein in the blood serum upon admission to the hospital was 28.8.

We calculate the probability of death using a logistic regression model, substituting the values of the predictors into the equation:

The probability of death (Y) = 1 / (1 + e ^{- ( 16.703 - 1.395 * GCS + 0.123 * S100-β)} ) = 1 / (1 + e ^{- ( 16.703 - 1.395 * 12 + 0.123 * 28.8)} ) = 0.97.

Comparing the obtained value of Y = 0.97 with the cut-off point value of 0.1 - since the obtained value is greater, we conclude that there is a high probability of death for this patient. In fact, the patient died.

The technical result is a logistic regression model, sensitivity 92.0%, specificity 83.0%, predicting the lethal outcome of ischemic stroke in the first 14 days with probability values greater than or equal to 0.1, a life-friendly prognosis - with probability values less 0.1.

The invention makes it possible to be characterized by simplicity and high information content. Its use makes it possible to increase the accuracy of the prognosis of a favorable and unfavorable outcome of ischemic stroke in the first 14 days and makes it possible to stratify patients into a group with a high risk of death. Accordingly, the accurate assessment of the patient's further prognosis obtained using the invention will help to select an individual treatment strategy for each patient in the acute period of ischemic stroke.

Information sources.

1. –A Lasek-Bal, H Jedrzejowska-Szypulka, S Student, A Warsz-Wianecka, K Zareba, P Puz, W Bal, K Pawletko, J Lewin-Kowalik. The importance of selected markers of inflammation and blood-brain barrier damage for short-term ischemic stroke prognosis. J PhysiolPharmacol 2019 Apr; 70 (2). doi: 10.26402 / jpp.2019.2.04. Epub 2019 Jul 22.

2. Olena Y Glushakova, Alexander V Glushakov, Emmy R Miller, Alex B Valadka, Ronald L Hayes. Biomarkers for acute diagnosis and management of stroke in neurointensive care units. Brain Circ Jan-Mar 2016; 2 (1): 28-47. doi: 10.4103 / 2394-8108.178546. Epub 2016 Mar 11.

3. Arkadiusz Weglewski, Danuta Ryglewicz, Anna Mular, Jacek Juryńczyk Changes of protein S100B serum concentration during ischemic and hemorrhagic stroke in relation to the volume of stroke lesion. Neurol Neurochir Pol. Jul-Aug 2005; 39 (4): 310-7.

4. Arkadiusz Weglewski, Danuta Ryglewicz, Anna Mular, Jacek Juryńczyk Changes of protein S100B serum concentration during ischemic and hemorrhagic stroke in relation to the volume of stroke lesion. Neurol Neurochir Pol. Jul-Aug 2005; 39 (4): 310-7.

5. Pooja Dassan, Geoffrey Keir, Martin M Brown. Criteria for a clinically informative serum biomarker in acute ischaemic stroke: a review of S100B. CerebrovascDis. CerebrovascDis 2009; 27 (3): 295-302. doi: 10.1159 / 000199468. Epub 2009 Feb 6.

6. Skripchenko N.V., Shirokova A.S. Neuronspecific enolase and s100 protein - biomarkers of brain damage. State of the art and clinical application. Pediatric neurosurgery and neurology. 2016; 4 (50): 16-25.

7. A. M. Golubev, M. V. Petrova, A. V. Grechko, V. E. Zakharchenko, A. N. Kuzovlev, A. V. Ershov. Molecular markers of ischemic stroke. GENERAL REANIMATOLOGY, 2019, 15; 5. DOI: 10.15360 / 1813-9779-2019-5-11-22

Odinak M.M. 8. Tsygan N.V. Ivanov A.M. Kamilova T.A. Nikitin V.Yu. Meshkova M.E. Protein S100β is a biomarker of brain damage. Bulletin of the Russian Military Medical Academy. 2011; 1 (33): 210-214.

Claims (7)

A method for predicting the likelihood of death in a 14-day period of acute ischemic stroke of the brain, which consists in determining the serum concentration of the neuron-specific protein S100-β in the venous blood on the 2nd day of ischemic stroke and assessing the patient's level of consciousness according to the Glasgow Coma Scale (GCS), which differs the fact that the probable prognosis of a fatal outcome is calculated by the formula Y = 1 / (1 + e ^{- ( 16.703 - 1.395 * GCS + 0.123 * S100-β)} ), where Y is the probability of death; e is a mathematical constant equal to 2.71828; GCS - the patient's level of consciousness on the Glasgow Coma Scale, points; S100-β - serum concentration of neuron-specific protein, μg / l, with a value of Y equal to or greater than 0.1, a high probability of death and an unfavorable prognosis for life are predicted, with a value of Y less than 0.1, a low probability of death and a favorable prognosis for life.

Images