Версия для печати Zhilinskiy E.V.

PROGNOSIS OF SEPSIS IN SEVERE BURN PATIENTS Zhilinskiy E.V.

Minsk Сity Сlinical Emergency Hospital,
Belarus State Medical University, Minsk, Belarus

One of the most severe complications of burn disease is burn sepsis: it develops in 8-42.5 % of patients with severe burns and presents the main cause of death (mortality > 65 %) [1]. The strict financial regulation of the issues of rendering medical care results in requirement for reliable prediction of sepsis. The separation of high risk groups of sepsis allows optimizing the treatment of patients with severe burns: appropriate prescription of immunoglobulins, use of fluidized beds, early necrectomy etc.      
The available predictive systems for sepsis (SOFA, SAPS, ISS and others) are cumbersome and intricate. Therefore, the practical use is oriented to the simple and most abbreviated predictive formulae including 4-5 criteria [2]. But administration of some predictors does not allow receiving any reliable results. Moreover, development of sepsis depends on patterns of injury and state of immune system in burn disease [1]. As result, the optimal way is creating the techniques for sepsis prediction on the basis of clinical and laboratory criteria, including the criteria, which are specific for burn injury. Patients with burn disease have the infection foci over the long time: extensive wounds with necrotic regions, pneumonia at the background of inhalation injury, sites of catheter installation. Burn disease is characterized with multiple complications: acute cardiovascular insufficiency (ACVI), respiratory distress syndrome (RDS), erosive-ulcerous lesions of the stomach and the intestine, hypermetabolic syndrome with evident catabolism, DIC and other conditions that promote the microbial translocation in the intestine in in the infection foci and stimulate the secondary immune deficiency [1, 3, 4]. The prediction of generalized infection and its appropriate prevention could be performed with the identification of the criteria of organ dysfunctions and insufficiency in burn disease that reflect the severity of patient’s condition and possibility of sepsis [2].
The objective of the study – to create a method for sepsis prognosis in patients with burn disease on the basis of clinical and laboratory criteria, which relate to the pathogenesis of burn injuries and infectious complications with consideration of the multiplier effect in the individual patient.

MATERIALS AND METHODS

The cohort prospective-retrospective study included the patients with severe burns in the burns units, Minsk City Clinical Hospital of Emergency Medical Aid in 2013-2015. The inclusion criteria were the lesion severity index ≥ 30 units, the age > 18, presence of appropriate volume of examination. The exclusion criteria were absent written consent from a patient or his/her relatives, death after burn shock, confidentiality of data owing to secrecy of investigation.
Sepsis was diagnosed with the criteria from the consensus board of burn infection of Chinese Medical Association (the table 1). Sepsis was confirmed in presence of at least 6 of 11 positive preliminary criteria and 1 or more confirming signs (hemoculture or a positive response to antibiotic therapy) [5]. All cases of sepsis were considered (the cases that were identified during the medicodiagnostic procedures and the analysis of medical documents).

Table 1. The criteria of the Consensus board of burn infection from Chinese medical association for diagnosis of sepsis in thermic injury 2013

Inflammatory response criteria	Signs which confirm infection
Hyperthermia (>39.0 °C) or hypothermia (<36.5 °C)	Response to antibiotic therapy
Tachycardia (more than 110 beats per minute)	Hemoculture
Tachypnoea (more than 25 respiratory movements per minute)
Thrombocytopenia (platelets amount < 100,000/mcl)
Hyperglycemia in absent diabetes mellitus > 12 mmol/l
Impossible continuation of enteral nutrition > 24 hours
Leukocytes > 15,000/mcl or < 5,000/mcl
Hypernatraemia > 155 mmol/l
Mental status disorder
Procalcitonin > 0.05 ng/ml
Minimal amount of signs
6 and more criteria	1 and more signs

For estimation of the sepsis predictors in the patients with burn disease, the physical (mean arterial pressure (MAP), heart rate (HR), respiratory rate (RR), saturation and others) and laboratory values were analyzed two days after termination of burn shock. Thrombocytopenia was ascertained in the platelets amount < 100,000/mcl of blood [6], immunoglobulinemia G – in the serum level of immunoglobulin G < 7 g/l, hypocholesterolemia – in prescription of cholesterol < 3.2 mmol/l [7]. The general blood analyses were conducted with the hematological analyzer XT-2000i (SysmexCor, Japan), the biochemical blood analyses (including immunoglobulin estimation) – with the analyzer AU-680 (Beckman Coulter, USA) with the electrophoretic system SAS-1Plus/SAS-2. The coagulogram values were estimated with the coagulometer CS-2100i (SysmexCor, Japan). Procalcitonin and cortisol were measured with the immune fluorescent technique with the analyzer Triage® MeterPro (Biosite Diagnostics, USA). The analysis of blood gases and electrolytes was conducted with the analyzer Stat Profile CCX (Nova Biomedical, USA). The vital parameters were measured with the medical monitors MM18I (OAO Integral, Belarus).
The study was conducted in compliance with Helsinki Declare – the Ethical Principles of Medical Research with Human Subjects 2000, with approval from the bioethical committee of Belarus State Medical University (December 26, 2014, the protocol #5). The informed consent from a patient (or his/her relatives) for participation in the study contains the findings according to the Law about protection of human rights and merits in biomedical studies in Commonwealth of Independent States (the law has been accepted by the Assembly of Commonwealth of Independent States, October 18, 2005, No. 26-10).
The patients were distributed into two samples: learning and testing. The learning sample included the patients with burn disease in concordance with the exclusion and inclusion criteria; they were treated from July 1, 2014 to December 31, 2015 (the prospective part of the study). The testing sample included the patients who were treated from January 1, 2013 to June 30, 2014 (the retrospective part of the study). The samples were distributed into the groups of the patients with sepsis and without it. The learning sample included 94 patients with burn disease (44 patients with sepsis and 50 patients without generalized infection). The testing sample included 95 patients with severe burns (40 patients with sepsis, 55 patients without it). The ratio men/women was 2.7/1 in the learning sample, 3.6/1 – in the testing sample (χ² = 0.42, р = 0.515). The patients’ age was 50 (44.5-58.5) in the testing sample and 51 (37-59) in the learning sample (U = 4633.0, p = 0.262). The burn area was 30 % (18-40) of the body surface in the testing sample and 30 % (22-44) of the body surface in the learning sample (U = 4924.0, p = 0.674). The square of deep burns was 10 % (4-20) in the testing sample and 10 % (5-25) in the learning sample (U = 5012.0, p = 0.836). The thermic skin burns combined with inhalation injury in 82.5 % of the patients in the testing sample and in 80.8 % in the learning sample (χ² = 0.10, р = 0.753). The lesion severity index was 65 (45-102) units in the testing sample and did not differ from the learning sample – 71 (46-110) units (U = 4856.0, р = 0.559). The samples were similar according to the patterns of injuries, age and gender.
The statistical analysis was conducted with Excel AtteStat 10 and SPSS 16.0. The technique of binary logistical regression was used for development of the way for predicting sepsis. The binary outcomes (the dependent variable) were development of sepsis as 1, absence of sepsis as 0. Since it is impossible to interpret the predicted values, which are not equal to 0 or 1, then the possibility of classifying a patient into the nearest category of the binary outcome, i.e. possibility of sepsis (p), was used [8]:

p = e^Z / (1 + e^Z),

where Z = a + b₁x₁ + b₂x₂ +…+ b_nx_n, а – constant, b_1-n – logistic regression coefficients, x_1-n – variables (factors).

The logistic regression equation allows estimating the possibility of development of events (sepsis) in each participant of the study with a set of the factors. Wald’s test was used for estimating the factors of sepsis development and assessment of their quantitative influence and search of their optimal combination. The binary logistic regression equation was developed on the basis of the data from the testing sample. The efficiency of the equation was tested on the basis of the patients in the testing sample. The prediction technique was estimated with ROC-analysis with calculation of area under the curve (AUC), 95 % confidence interval (CI), sensitivity, specificity and the likelihood ratio. The quantitative data was presented as the median and the interquartile range – Ме (Q₁–Q₃), the qualitative data – as proportions (%). The intergroup differences were identified with Mann-Whitney test (U) and χ² (with calculating Fisher test). The statistically significant differences were at p < 0.05 [8].

RESULTS

The tables 2 and 3 show the clinical and laboratory values of the patients of the learning sample on the second day after termination of shock, including the criteria, which are specific for the complications of burn disease (secondary immune deficiency, adrenal insufficiency and others) and the inflammatory response markers. The criteria of burn shock termination are stabilizing arterial pressure (mean arterial pressure > 65 mm Hg), normalizing diuresis and development of polyuria (diuresis > 1 ml/kg/hour), decreasing hemoconcentration (hematocrit < 45 %), increasing body temperature (> 36.0 °C) [3]. The maximal values of HR, RR and fibrinogen, maximal proportion of neutrophils (including young forms), lower level of albumin were in the patients with subsequent sepsis on the second day after arrest of shock as compared to the patients without sepsis in the learning sample. Also the patients with sepsis demonstrated the lower values of Ig, A, Ig, G, cholesterol and cortisol and the higher levels of procalcitonin and CRP in comparison with the patients without sepsis (the tables 2, 3).

Table 2. Physical and laboratory values on the second day after withdrawal of burn shock in patients of learning sample, Ме (Q1-Q3) or (%), n = 94

Value	Patients without sepsis n = 50	Patients with sepsis n = 44	U (χ2), p
Mean arterial pressure, mm Hg	93.65 (85.125–103.85)	89 (82–92.7)	U = 721.5, p = 0.004
Maximal HR, min-1	84.5 (77–99)	102 (96–114.5)	U = 538.5, p = 0.000
Respiratory rate, min-1	18 (16–19)	18 (18–21)	U = 538.5, p = 0.000
Maximal body temperature, оС	36.6 (36.6–37.0)	36.7 (36–37.1)	U = 915.0, p = 0.162
SpO2, %	99 (98–99)	98 (98–99)	U = 909.5, p = 0.150
Leukocytes, thousands/mcl	11.6 (8.24–13.48)	12.85 (10.33–15.68)	U = 905.0, p = 0.141
Hemoglobin, g/l	117 (108–141)	115 (108–137)	U = 1084.0, p = 0.909
Platelets, thousands/mcl	149 (107.25–173.25)	134.5 (96.5–184.75)	U = 1003.5, p = 0.467
Proportion of patients with thrombocytopenia, %	6	29.6	χ2 = 9.19, р = 0.005
Proportion of young forms of neutrophils, %	9 (6–10)	14 (8.75–20)	U = 544.0, p < 0.001
Proportion of neutrophils, %	77.25 (69.88–79.8)	80.6 (71.38–85)	U = 712.5, p = 0.003
APTT, sec.	28.7 (27.13–32.20)	28.7 (26.80–32.45)	U = 1078.0, p = 0.871
INR	1.05 (1.02–1.16)	1.08 (0.96–1.17)	U = 1090.5, p = 0.946
Fibrinogen, g/l	4.7 (4.44–4.90)	5.8 (4.02–6.86)	U = 651.0, p = 0.001
рН	7.30 (7.27–7.34)	7.32 (7.29–7.37)	U = 985.0, p = 0.388
Lactate, mmol/l	2.6 (1.7–3.15)	2.3 (1.4–3.15)	U = 1069.0, p = 0.817
BE, mmol/l	-2.55 (-5.6 – -0.35)	-2.7 (-5.3–1.23)	U = 1023.0, p = 0.562
Bicarbonate, mmol/l	22.75 (19.33–25.40)	23.2 (21.05–26.30)	U = 939.5, p = 0.225
Total protein, g/l	50.9 (48.2–53.43)	49.4 (42.58–55.75)	U = 1007.0, p = 0.483
Albumin g/l	31.3 (30.1–32.7)	28.2 (25.4–32.6)	U = 749.0, p = 0.008
Creatinine, mcM/l	98.8 (72.9–127.75)	85.5 (69.2–115.2)	U = 937.0, p = 0.218
Capillary blood glucose, mmol/l	6.4 (5.025–7.1)	6.9 (4.9–8.55)	U = 952.0, p = 0.264

Table 3. Specific laboratory values on the second day after withdrawal of shock in patients of learning sample, Ме (Q1-Q3) or (%), n = 94

Value	Patients without sepsis n = 50	Patients with sepsis n = 44	U (χ2), p
Ig A, g/l	2.19 (1.85–2.84)	1.905 (1.585–2.4825)	U = 833.0, p = 0.043
Ig M, g/l	0.895 (0.67–1.12)	0.87 (0.74–1.0725)	U = 1055.0, p = 0.736
Ig G, g/l	8.455 (6.18–10.59)	6.015 (4.67–7.24)	U = 533.5, p < 0.001
Proportion of patients with hypoglobulinemia G, %	32	65.9	χ2 = 10.78, р = 0.002
Cortisol, ng/l	204.6 (123.8–244.8)	154.4 (86.0–230.9)	U = 574.0, p = 0.349
Cholesterol, mmol/l	3.76 (3.40–5.40)	2.45 (2.21–2.95)	U = 151.0, p < 0.001
Proportion of patients with hypocholesterinemia, %	34	61.4	χ2 = 7.04, р = 0.013
C-reactive protein, mg/l	29.3 (18.6–47.9)	85.0 (69.2–103.5)	U = 134, p < 0.001
Procalcitonin, ng/ml	0.27 (0.17–0.79	0.60 (0.42–1.03)	U = 214, p = 0.025

The logistic regression equation was selected for prediction of sepsis. It included the constant and 5 variables: hypocholesterolemia (serum cholesterol < 3.22 mmol/l), levels of fibrinogen, albumin and immunoglobulin G, proportion of neutrophils, heart rate. The table 4 demonstrates the coefficients of the logistic regression equation. In presence of hypocholesterolemia, the increase in proportion of neutrophils and level of fibrinogen augments the probability of sepsis, but the increasing levels of albumin and immunoglobulin G decreases it (the table 4).

Table 4. Coefficients in logistic regression equation for prediction of sepsis in patients with burn disease with use of specific values

Predictor	Equation coefficient (B)	Coefficient standard error	р	Odds ratio (OR)	95 % CI OR
Immunoglobulin G	-0.308	0.102	0.003	0.735	0.601–0.898
Neutrophils	0.083	0.035	0.018	1.086	1.015–1.163
Fibrinogen	0.888	0.288	0.002	2.431	1.383–4.275
Albumin	-0.132	0.072	0.068	0.876	0.760–1.010
Hypocholesterinemia	1.383	0.73	0.058	3.989	0.954–16.679
Constant	-5.027	3.953	0.203	0.007

The probability of sepsis (p) in the patients with burn disease after shock termination can be evaluated with the formula:

 p = e^Z / (1 + e^Z), where Z –

Z = -5.027 + 1.383*presence of hypocholesterolemia – 0.132*albumin (g/l) + 0.888*fibrinogen (g/l) + 0.083*neutrophils (%) -0.308*immunoglobulin G (g/l)

The ROC-analysis showed the p value of 0.595 with inclusion to a positive result as the optimal predictive level for development of sepsis in burn disease, i.e. p ≥ 0.595 predicts the course of burn disease with development of sepsis. AUC for the way of prediction in the learning and testing samples was 0.910 and 0.849 (the way of “excellent” and “very good” quality) (Fig. 1, 2). The sensitivity of the way was 86.4 % in the learning sample and 82.5 % in the testing sample, specificity – 82.2 % and 81.8 % correspondingly. The likehood ratio for the given way was 4.80 in the learning sample and 4.25 in the testing sample, i.e. p ≥ 0.595 means that the relative risk of sepsis is 4.80 times higher than in the learning sample and 4.25 times higher than in the testing sample as compared to p < 0.595.

Figure 1. ROC-analysis of sepsis predictive model in the testing sample, n = 95

 

Figure 2. ROC-analysis of sepsis predictive model in the learning sample, n = 95

DISCUSSION

Patients with burn disease are characterized by development of hypermetabolic syndrome. After shock termination, the phase of growing hypermetabolic response develops in patients with severe burns. This phase has such manifestations as hyperdynamic mode of blood circulation, hyperventilation, hyperthermia and disordered glucose consumption. Hypermetabolism is accompanied by evident lipolysis, decreasing synthetic function of the liver and protein catabolism that cause secondary immune deficiency, respiratory distress syndrome (RDS), hypostatic pneumonia and substrate endocrinologic insufficiency [9].
The septic patients demonstrated more intense hypermetabolic syndrome than the patients with severe burns without sepsis. It manifested with higher HR and RR and lower albumin and cholesterol levels after shock arrest. Hypoglobulinemia G and thrombocytopenia were more common for the septic patients on the 2nd day after shock termination than for the patients without sepsis (the tables 2, 3).
The development of metabolic disorders in burn disease is shown by the variables in the binary logistic regression equation for sepsis prediction. Hypocholesterolemia demonstrates the decreasing synthesis of cholesterol and phospholipids in the liver. The low level of cholesterol promotes the development of adrenal insufficiency and RDS that cause severe pneumonia and systemic inflammatory response dominating the anti-inflammatory response. The decreasing albumin level in burn disease demonstrates the protein catabolism with decreasing synthesis of globulins (Ig G) and development of secondary immune deficiency [9]. The increasing level of fibrinogen characterizes the intensity of DIC determined by inflammatory response, microcirculatory bed injury etc. The increasing level of neutrophils in burn disease is determined by release of proinflammatory cytokines after an injury and bacterial infection [8].

CONCLUSION                   

The technique for predicting sepsis in patients with burn disease with use of the binary logistic regression equation, including such variables as presence of hypocholesterolemia, levels of fibrinogen, albumin and immunoglobulin G and proportion of neutrophils, has shown the high efficiency in the learning and testing samples. The sensitivity of the prediction technique was 86.4 % in the learning sample, 82.5 % in the testing sample, specificity – 82.2 % and 81.8 % correspondingly. The quality of sepsis prediction has been evaluated as “excellent” and “very good”.
The offered predictive technique can be used for optimizing the treatment for patients with severe burns in specialized burn centers.     

 Information about financing and conflict of interests

The study was conducted within the limits of the task 2.42 of the Federal scientific program of studies “Fundamental and applied sciences – medicine”.
The authors declare the absence of evident and potential conflicts of interest relating to publishing the present article.