TECHNICAL FIELD

The present invention relates generally to methods for decision support. More particularly, the present invention relates to a method for decision support in diagnosis of neutropenic fever in hematology patients.

BACKGROUND

Inflammation is an evolutionarily conserved response to danger, e.g. microbial invasion and/or tissue injury. It encompasses activation of cells and soluble mediators of the innate immune system, including the complement and coagulation systems. In intense inflammatory activation, a systemic inflammatory response (SIRS) evolves with profound alterations of most bodily functions. The liver redirects its protein synthesis and shuts down the production of serum albumin. Simultaneously, it increases the production of proteins termed "acute phase proteins" (7). Signs of systemic inflammation include fever, chills, increased heart rate, dilated blood vessels, reduced blood pressure and catabolic metabolism. The circulatory and metabolic changes of SIRS cause organ stress and, ultimately, multiple organ failure that may be lethal.

The inflammatory cascade is triggered by "danger signals", which may originate from microbes, "pathogen-associated molecular patterns" (PAMPs), e.g. lipopolysaccharide, peptidoglycan and β- glucan, bacterial and viral DNA. The same type of reaction can also be triggered by substances leaking out of our own injured tissues, i.e. "damage-associated molecular patterns" (DAMPs) such as ATP, uric acid, and mitochondrial N-formylated peptides. Unfortunately for clinicians, the inflammatory response is fairly stereotypic, regardless of the initiating cause. Fever and elevated serum levels of the acute phase protein C-reactive protein are hallmarks of sepsis and other severe infections. However, they are also produced upon sterile tissue injury resulting from extensive surgery, organ and tissue infarction or tumor cell decay.

Patients with hematologic malignancies undergoing intensive chemotherapy commonly develop neutropenia, i.e. blood neutrophil counts less than 0.5 xl0 ⁹ /L. Neutropenia arises both due to the toxic effects of chemotherapy and the expansion of tumor cells, which combined hinder the growth of healthy hematopoietic cells. Neutrophilic granulocytes are central in the human immune defense against invasive bacteria and fungi. Consequently, neutropenic patients with hematological malignancies are exceedingly vulnerable to infections. Chemotherapy also damages the mucosal barriers, which together with central venous lines used to administer chemotherapy, facilitate microbial invasion of these immune compromised patients.

A major clinical problem is that a prompt diagnosis of infection in hematology patients is difficult since the patients are already inflamed due to decay of malignant cells and tissue damage following treatment with cytotoxic drugs. Hence, the inflammatory markers generally used as proxies of infection, such as fever, elevated C-reactive protein, etc., are of limited use. In fact, patients with hematologic malignancies undergoing intensive chemotherapy often display several criteria of SIRS, such as fever, increased heart rate, and reduced white blood cell counts. It has been estimated that more than half of the febrile episodes afflicting these patients have a non-infectious background.

The search for a single or handful of laboratory parameter(s) able to distinguish between the inflammatory responses triggered by a severe infection as opposed to one evoked by sterile cell decay has hitherto been futile. Rapid diagnosis of infections is essential for the optimal care of hematology patients. Mortality rates ranging from 5-30% are seen in neutropenic hematology patients with severe infections. Conversely, interruption of chemotherapeutic regimens because of mistaken assumption of ongoing invasive infection could theoretically compromise the chance of curing the underlying malignancy.

To conclude, there is a large need to improve the available methods for assessing the cause of fever in neutropenic hematology patients. SUMMARY

An objective of the present invention is to provide a method as well as a computer implemented method for decision support in diagnosis of neutropenic fever of hematology patients. The inventive method obviates some of the problems and disadvantages outlined above in connection with known methods. In accordance with one embodiment of the present invention, a method of providing clinical decision support in neutropenic fever diagnosis is provided. The method of providing clinical decision support in neutropenic fever diagnosis, comprises providing information about a patient (e.g. presence of indwelling catheters, ongoing medication, smoking habits, sex and some simple physiological parameters such as blood pressure), providing a blood sample from said patient, and measuring blood parameters from said blood sample. The blood parameters comprise measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins. The measurements comprise measuring concentrations, amounts etc. of the particular parameters in accordance with well-known measures used in a laboratory for setting a clinical status of a patient. The method may further comprise using parameters or information which has been determined to be of particular relevance for calculating an indicator value through using a "variable of importance" (VIP) approach as further explained herein. The method further comprises calculating an indicator value by using said measurement of blood parameters and information about said patient in a regression model. The method further comprises determining if the indicator value is indicative of whether the patient has neutropenic fever due to systemic infection or due to non- infectious inflammation, respectively.

In accordance with yet another embodiment of the present invention, a computer-implemented method of providing decision support in neutropenic fever diagnosis is provided. The computer implemented method comprises receiving information about a patient, receiving blood parameters from an analysis of a patient blood sample. The blood parameters comprise measurements of blood cells, iron turnover, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and

immunoglobulins. The computer implemented method further comprises calculating an indicator value by using said measurement of blood parameters and said information about said patient in a regression model. The computer implemented method further comprises determining if the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to noninfectious "sterile" inflammation, respectively. There is further provided herein a computer readable medium having stored thereupon a computer program which when executed calculates an indicator value in a regression model by using as input measurements of blood parameters and information about a patient as defined herein. There is further provided herein a kit for providing clinical decision support in neutropenic fever diagnosis in a patient, said kit comprising analytical detection and/or measuring reagents for measuring blood parameters from a blood sample from said patient, wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins as defined herein, and optionally instructions for use. Said kit may further comprise a computer readable medium having stored thereupon a computer program which when executed calculates an indicator value in a regression model by using as input measurements of blood parameters and information about a patient as defined herein, An advantage of particular embodiments of the present invention is the ability of the method to precisely provide decision support to determine whether a neutropenic hematology patient exhibits fever due to a systemic infection or due to a non-infectious inflammation, respectively.

An advantage of particular embodiments of the present invention is that the computer implemented method easily can be integrated in an automated analysis system.

An advantage of particular embodiments of the present invention is that the method more precisely can predict the cause of neutropenic fever and thereby the mortality of the neutropenic hematology patients can be reduced.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a score plot of an embodiment of an O-PLS model using all 139 variables;

Figure 2 is a variable importance plot of the 55 most important variables in an embodiment of an O- PLS model;

Figure 3 is a variable importance plot of the 40 most important variables in an embodiment of a reduced O-PLS model; Figure 4 is a score plot of the embodiment of the reduced O-PLS model; and

Figure 5 is a flowchart of an embodiment of the method for decision support in diagnosis of neutropenic fever in hematology patients.

DETAILED DESCRIPTION

Definitions

Herein, the terms "parameter" and "variable" may be used interchangeably, and refers either to information that has been obtained when performing a biochemical test on a patient's blood sample (biochemical parameters) and/or when collecting information regarding said patient, e.g. by questioning the patient, reviewing the clinical charts, or performing simple physical examinations such as measuring blood pressure (clinical parameters). The parameters are then processed using a computational method as presented herein to indicate the clinical status of said patient through an indicator value.

An "indicator value" herein, refers to the prediction score that is calculated by the computational method (Orthogonal Projection to Latent Structures, O-PLS). The prediction score indicates the likelihood that fever in a neutropenic patient is caused by an infection rather than sterile inflammation driven by non-infectious causes.

A "regression model" refers to a model that estimates the relationship between a dependent variable and one or more independent variables. A method used for regression modeling herein is Orthogonal Projection to Latent Structures, O-PLS (2).

A "VIP" approach is an analysis used herein which is indicative of whether it is possible to reduce the number of variables in a model used and still achieve a dependable prediction (outcome). This is also interchangeable referred to herein as the "variable importance in the projection (VIP) method", "variable of importance (VIP approach)" or the "variable influence on projection" (VIP) The analysis is used to select parameters and information of particular importance for calculating the indicator value. (3).

A method of providing clinical decision support as presented herein is a computational method which calculates the likelihood that fever in a neutropenic patient is caused by an infection, rather than sterile inflammation driven by non-infectious causes, based on a set of biochemical and clinical parameters.

A "blood parameter" as mentioned herein, refers e.g. to information obtained from a blood sample regarding its content, such as the concentration of an endogenous protein or other body component (such as a blood cell) in said sample, or the reaction of said blood sample, e.g. the ability to coagulate, or the capacity of red blood cells to sediment, in a measured period of time.

"Measurement of blood cells" as mentioned herein, refers to any measurements of factors related to or including blood cells, such as absolute and relative counts of a particular blood cell type (e.g. white blood cells), size and hemoglobin concentration in red blood cells as well as derivatives of these measures.

"Iron turn-over" as used herein, refers to the exchange and recycling of, iron in the human body. The measurement of any component being involved in an iron turn-over involves transferrin, ferritin, hepcidin, prohepcidin, soluble transferrin receptor, blood hemoglobin, total iron binding capacity, as well as the iron concentration.

Blood coagulation is a well-defined process in the human body, incorporating many factors such as exemplified herein (see e.g. Table 2). Parameters associated with the blood coagulation process may be measured and used in a method for clinical decision support.

Herein, the term "organ stress" refers to a state where body organs, e.g. the kidneys, the liver, the heart and/or the skeletal muscle is negatively affected leading to reduced function and/or cell damage. Measures of organ stress and cell decay include several blood parameters, such as troponin-T, HMGB- 1, etc. as exemplified in Table 2.

Herein, when a parameter related to "metabolism" is referred to, this is intended to mean any parameter that is involved in the turn-over of energy required for maintenance of the human body.

"Acute phase reactants" as referred to herein are proteins produced by liver cells (hepatocytes) in response to acute inflammatory triggers, such as C-reactive protein and orosomucoid or as further exemplified herein.

It may be so that some of the blood parameters belong to more than one of the groups defined herein.

The terms "immunoglobulin" and "cytokine" are well known in the art, and information with regards to its level of presence is further used herein as useful parameters in a method for clinical decision support.

Detailed description In the following, different aspects will be described in more detail with reference to certain embodiments and to accompanying drawings. For purpose of explanation and not limitation, specific details are set forth, such as particular scenarios and techniques, in order to provide a thorough understanding of the different embodiments. However, other embodiments that depart from these specific details may also exist.

Accordingly, there is provided herein a method of providing clinical decision support in neutropenic fever diagnosis, comprising the steps of: a) providing information about a patient, such as clinical variables, b) providing a blood sample from said patient, c) measuring blood parameters from said blood sample, wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, and the method further comprises d) calculating an indicator value by using said measurement of blood parameters and information about said patient in an regression model, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a non-infectious inflammation, respectively.

There is further provided herein a method of providing clinical decision support in neutropenic fever diagnosis, comprising:

providing information about a patient,

providing a blood sample from said patient,

measuring blood parameters from said blood sample,

wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, and the method further comprises calculating an indicator value by using said measurement of blood parameters and information about said patient in a regression model, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a noninfectious inflammation, respectively. There is also provided herein a method of providing clinical decision support in neutropenic fever diagnosis, comprising the steps of: a) providing information about a patient, such as clinical variables, b) measuring blood parameters from a blood sample from said patient, wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, and the method further comprises c) calculating an indicator value by using said measurement of blood parameters and information about said patient in an regression model, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a non-infectious inflammation, respectively. There is also provided herein a method of providing clinical decision support in neutropenic fever diagnosis, comprising:

providing information about a patient,

measuring blood parameters from a blood sample of a patient,

wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, and the method further comprises calculating an indicator value by using said measurement of blood parameters and information about said patient in a regression model, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a noninfectious inflammation, respectively.

As mentioned herein, the regression model may be an O-PLS regression model (2). The method presented herein is performed in vitro.

The blood parameters in step b) (or c) and/or the information in step a) may be parameters and/or information selected according to a variable of importance (VIP) approach as defined herein (3).

Further, a blood parameter and/or information from a patient used in a method herein may have a VIP value obtained by the variable of importance in the projection (VIP) approach of about 1 and above, such as about 0.8, 0.9, 1.0, 1.1, 1.2 or 1.3 and above. Accordingly, there is provided herein a method wherein the blood parameters and/or information used in step c) are parameters and/or information selected according to a variable of importance (VIP) approach. There is further provided a method wherein a blood parameter and/or information from a patient has a VIP value obtained by the variable of importance (VIP) approach which is about 1, such as 1.0, and above, such as about 0.8, 0.9, 1.0, 1.1, 1.2 or 1.3 and above.

The method may comprise a further step before step c) (i.e. a step before calculating an indicator value) comprising performing a variable of importance (VIP) analysis of the blood parameters and/or the information about a patient. The parameters and/or information determined to be of particular importance from the variable of importance (VIP) analysis may thereafter be used in the calculation of the indicator value in step c) (or d)). Such a method allows for eliminating parameters with little contribution, i.e. those that have a VIP value below a certain level (e.g. below about 1.0, such as below about 1.3, 1.2, 1.1. 1.0, 0.9, or 0.8 as exemplified herein) from the regression model.

Naturally, the method may then further comprise using the parameters determined to be of particular importance using the variable of importance (VIP) approach in the calculation of the indicator value in step c). Experimental

A population of 52 febrile neutropenic patients with hematological malignancies who had been subjected to chemotherapy and/or other immunosuppressive medication was recruited. Close to half of the population had also undergone hematopoietic stem cell transplantation. Inclusion criteria were neutropenia (< 0.5· 10 ⁹ /L) and fever (body temperature > 38.0°C). The median duration of their febrile episode was 2 days (range = 1-20 days, 25 ^th /75 ^th percentile = 1/4 days). The patients were briefly examined and their charts reviewed, providing 23 clinical and physiological parameters, for example age, gender, body temperature, presence of chills, fever days, blood pressure, see Table 1. Their blood was analysed for 116 parameters also listed in Table 1. These parameters included blood cell counts, coagulation and complement factors, acute phase proteins, and a range of measures of metabolism and organ function.

hemoglobin

Complement Plasma analysis 5 Complement factor 3 (C3), C4, C4-binding protein

(measurement of (C4BP)- , C4BP- , total complement complex acute phase (TCC)

reactants)

Coagulation and Plasma analysis 17 Activated partial thromboplastin time (APTT), fibrinolysis (measurement of prothrombin time (measured as international

blood normalized ratio, INR), coagulation factors II coagulation) (prothrombin), V, VII, VIII, IX, X, XI, XII, von

Willebrand factor, fibrinogen, anti-thrombin, D- dimers, plasminogen activator inhibitor- 1 (PAI-1), protein C, protein S

Cytokines Serum analysis 8 Interferon-γ (IFN-γ, Interleukin-1 (IL-1), IL-6, IL- (measurement of 8, IL-10, IL-17, tumor necrosis factor (TNF), cytokines and lymphotoxin (LT)

immunoglobulins)

Acute phase proteins Serum analysis 16 C-reactive protein (CRP), haptoglobin, hemopexin,

(measurement of orosomucoid, serum amyloid A (SAA), al- acute phase antitrypsin, a2-macroglobulin, ceruloplasmin, reactants) ferritin, hepcidin, pro-hepcidin, soluble transferrin receptor, total iron binding capacity (TIBC), transferrin, albumin, procalcitonin

Acute phase response Blood analysis 1 Erythrocyte sedimentation rate (SR)

(measurement of

blood cells)

Kidney function and Serum analysis 13 Sodium (Na), potassium (K), Calcium (Ca), electrolytes (measurement of chloride (CI), magnesium (Mg), phosphate (P), organ stress) iron, urea, creatinine, cystatin C, NT-pro-brain (measurement of natriuretic peptide (NT-pro-BNP), erythropoietin metabolism) (EPO), renin

Liver Serum analysis 6 Aspartate aminotransferase (AST), alanine

(measurement of aminotransferase (ALT), alkaline phosphatase organ stress) (ALP), bilirubin conjugated, bilirubin total, γ- (measurement of glutamyl transferase (yGT) metabolism) Organ stress and cell Serum analysis 13 Creatine kinase (CK), creatine kinase-muscle brain decay (measurement of (CK-MB), troponin-T, lactate dehydrogenase organ stress) (LDH), myoglobin, pancreatic amylase, total amylase, prostate specific antigen (PSA), free PSA, high mobility group box protein- 1

(HMBG1), a-fetoprotein (AFP), urate, lactoferrin (plasma)

Metabolism Serum analysis 8 High density lipoprotein (HDL), low density

(measurement of lipoprotein (LDL), cholesterol, triglycerides, metabolism) protein, leptin, adiponectin, glycosylated

hemoglobin (HbAlc)

Metabolism Plasma analysis 1 Glucose

(measurement of

metabolism)

Hormones Serum analysis 10 Cortisol, T4 (tetraiodo-thyronine, thyroxine), free

(measurement of T4, T3 (triiodo-thyronine), free T3, thyroxine- metabolism) binding globulin (TBG), thyroid-stimulating

hormone (TSH), insulin-like growth factor- 1 (IGF- 1), estradiol, testosterone

Immunoglobulins Serumanalysis 4 Immunoglobulin A (IgA), IgE, IgG, IgM

(measurement of

cytokines and

immunoglobins)

Total number 139

Table 1 : Measured parameters for the population of 52 patients.

Every patient in the population was classified as being infected or not infected by an independent evaluator (LHa) who was blinded to the results of the biochemical tests. The following data was presented to the evaluator: 1) the clinical signs and symptoms manifested by the patients (cough, skin rash, etc.), 2) the results of extensive microbiological investigations, including cultures of blood and other bodily specimens, analysis of microbial DNA and antigens, and 3) imaging data derived from CT scans, chest X-rays, ultrasounds, etc. In order to be classified as infected a clinically relevant microbe compatible with the symptoms and clinical signs that were presented by the individual patient were required. Based on this clinical evaluation, one third of the patients (13/42) were judged to have an infectious cause to their febrile episode. The majority (11/13) had bacterial sepsis (4 cases of ex- streptococci, 3 enterococci, 1 Bacillus sp., 1 Gemella sp., 1 Pseudomonas aeruginosa, 1 E. coli). In addition, invasive fungal disease (Pneumocystis jiroveci pneumonia) and viral reactivation (CMV) were seen in one case each. Hence, both patients with and without proven infections were severely inflamed, i.e., all patients had serum procalcitonin and C-reactive protein levels above the normal range.

In order to further analyse the available data a multivariate pattern recognition method were employed.

A multivariate pattern recognition method "Orthogonal-Projection onto Latent Structures" (O-PLS) were applied to the patient data from the 52 patients ((2)). A model that could separate the infected patients (Y=l) from the non-infected ones (Y=0) were generated using said O-PLS method with all available data used as input. In figure 1 is said model based on the 139 measured X-variables shown. From this figure 1 it is evident that infected patients 201 can be separated from non-infected patients 202.

However, a model that uses 139 variables is of limited use due to the large number of variables.

Therefore, the contribution of each of the 139 variables to the model was determined using the

"variable of importance" (VIP) approach. This analysis is indicative of whether it is possible to reduce the number of variables in the model and still being able to achieve a dependable prediction.

This VIP analysis revealed that serum procalcitonin was the single most discriminatory variable, see Table 2 below. However, serum procalcitonin has limited use on its own for assessing if a patient is infected or not.

S-Interleukin-8 IL-8 1.9 Infected X X

P-Factor XII XII 1.9 Non-infected X X (Hageman factor)

B-Erythrocyte volume EVF 1.8 Non-infected X X fraction

S -Bilirubin BilC 1.8 Infected X X (conjugated)

B-Hemoglobin Hb 1.7 Non-infected X X

Current smoker Smok 1.7 Non-infected X X

S-Cortisol Cort 1.7 Infected X X

B-Lymphocyte count Lymph 1.6 Non-infected X

P-Plasminogen PAI 1.6 Infected X X activator inhibitor- 1

P-von Willebrand vWF 1.6 Infected X X factor

S-High mobility group HMGB 1.6 Non-infected X X box protein- 1

S-Prohepcidin Prohep 1.5 Infected X X

S -Low-density LDL 1.5 Non-infected X X lipoprotein

B-Neutrophil count Neu 1.5 Non-infected X

S-C4b-binding protein C4BPb 1.4 Non-infected X X

(β-chain-containing

isoform)

S-Cystatin C CysC 1.4 Infected X X

P-Glucose Glu 1.4 Infected X

P-Protein S Prot S 1.4 Non-infected X X

Cyclosporin A CyA 1.3 Infected X X treatment

S-Hepcidin Hepc 1.3 Infected X X

E-Mean corpuscular MCHC 1.3 Infected X X hemoglobin

concentration

B-Eosinophil count Eos 1.3 Non-infected X S -Alpha fetoprotein AFP 1.3 Infected X X

S -Lactate LD 1.3 Non-infected X

dehydrogenase

Blood pressure BPd 1.3 Non-infected X X (diastolic)

S-Cholesterol Choi 1.3 Non-infected X X

B-Reticulocyte count Ret 1.3 Non-infected X X

S -Immunoglobulin E IgE 1.3 Non-infected X

Peripheral vein Per 1.2 Non-infected X X sampling

Port-a-catheter PAC 1.2 Infected X X

S-Interleukin-1 IL-1 1.2 Infected X X

B-Mean corpuscular MCV 1.2 Non-infected X X volume

B-International INR 1.2 Infected X X Normalized Ratio (=

prothrombin time)

P-Factor X X 1.2 Non-infected X X

B-Monocyte count Mono 1.2 Non-infected X

S-C-reactive protein CRP 1.2 Infected X X

S-Ferritin Ferri 1.1 Infected X

S-potassium K 1.1 Infected X

S-hemopexin Hpx 1.1 Non-infected X

S -troponin T TropT 1.1 Infected X X

B-hemoglobin A HbA 1.1 Non-infected X

S-transferrin Trans 1.1 Non-infected X X

S- immunoglobulin M IgM 1.0 Infected X

S -sodium Na 1.0 Non-infected X X

B-basophil count Bas 1.0 Non-infected X

B-erythrocyte particle EPC 1.0 Non-infected X X concentration

S-orosomucoid Oro 1.0 Non-infected X

P-activated partial APTT 1.0 Infected X X thromboplastin time S -thyroid stimulatory TSH 1.0 Non-infected X hormone

S -total iron binding TIBC 1.0 Non-infected X X capacity

Allogeneic transplant Alio 0.97

recipient

P-coagulation factor II II 0.96

(prothrombin)

Erythrocyte SR 0.91

sedimentation rate

S-high density HDL 0.88

lipoprotein

S -chloride CI 0.85

S-amylase, pancreatic AmyP 0.83

S -alanine ALT 0.81

aminotransferase

S -immunoglobulin G IgG 0.79

S -albumin Alb 0.79

B-platelet count PC 0.78

P-coagulation factor VII 0.72

VII

S-testosterone Test 0.71

S-interleukin-6 IL-6 0.70

S -triiodothyronine T3 0.69

S -creatinine Crea 0.68

S-leptin Lep 0.68

S -myoglobin Myo 0.67

Blood pressure, BPs 0.67

systolic

S -tetraiodothyronine T4 0.67

S-tumor necrosis factor TNF 0.65

S-estrogen Est 0.65

S -protein Prot 0.64

Patient fasting prior to Fast 0.63 blood sampling

S-creatine kinase, CKM 0.63 muscle brain

S-transferrin saturation Trans S 0.60

P-factor XI XI 0.60

S-triglycerides TG 0.60

S-a2-macroglobulin a2M 0.59

S -aspartate AST 0.59 aminotransferase

S -soluble transferrin TfR 0.57 receptor

Body length Leng 0.56

P-protein C ProtC 0.55

Cytarabine treatment, Cytar 0.53 current

S-Iron Fe 0.52

No of fever days prior Fday 0.52 to inclusion

S -urate Urate 0.52

S -insulin growth factor IGF-1 0.51 -1

Peripheral vein PVC 0.50 catheter

S-ceruloplasmin Cer 0.49

Body weight Wt 0.48

S -phosphate P 0.48

S -haptoglobin Hap 0.47

P-factor VIII VIII 0.47

S -immunoglobulin A IgA 0.47

Autologous transplant Auto 0.47 recipient

S-alkaline phosphatase ALP 0.46

Heart rate HR 0.45

B-Erythrocyte mean MCH 0.44 corpuscular

hemoglobin

P-coagulation factor V V 0.39

S -tetratiodothyronine , T4f 0.39 free

S-adiponectin Adip 0.38

S-prostate specific PSAf 0.37 antigen, free

S-y-glutamyl gGT 0.32 transferase

P-dimerized plasmin Dim 0.32 fragment D

S -thyroxin-binding TBG 0.32 globulin

Age Age 0.28

P-erythropoietin EPO 0.27

Central venous catheter CVC 0.27

S-interferon-γ IFNg 0.27

S-triiodothyronine, free T3f 0.26

Hypogammaglobuline Hypog 0.26 mia

S -calcium Ca 0.25

S -amyloid A SAA 0.24

S -fibrinogen Fibrino 0.24

Temperature Temp 0.22

Presence of chills Chills 0.20

Peripheral oxygen POX 0.19 saturation

S-interleukin-17 IL-17 0.19

S-C4b-binding protein- C4BP-a 0.18 ex chain

S -complement factor 3 C3 0.17

Sex Sex 0.16

S-lactoferrin Lact 0.16 P-anti-thrombin a- 0.14

thrombin

S-amylase, total AmyT 0.14

S -creatine kinase CK 0.14

S-magnesium Mg 0.14

S-prostate-specific PSA 0.093

antigen

S -total complement TCC 0.076

complex

Corticosteroid therapy Co 0.066

P-renin Ren 0.056

S -a 1 -antitrypsin AT 0.032

S -complement factor 4 C4 0.025

S-NT pro-brain pBNP 0.015

natriuretic peptide

P-factor IX IX 0.0051

Table 2: Variables in embodiments of the method. S=serum, P=blood plasma and B=blood. The relative importance of each variable to the 139 variable model was determined using the "variable of importance" technique.

Many of the variables that contributed strongly to the model turned out to be routine clinical chemistry analytes such as urea, hemoglobin, and bilirubin. Other "top-ranked" were white and red blood cell counts, cytokines such as interleukin-8 and -10, and plasma proteins involved in coagulation and fibrinolysis, e.g. factor XII, plasminogen-activator inhibitor- 1, and von Willebrand factor.

Parameters with no or negligible capacity to segregate infected from non-infected neutropenic hematology patients presenting with fever appear at the bottom of Table 2, and encompass many classical measures of inflammation, including acute phase proteins, complement factors, and several cytokines. A range of physical or anamnestic parameters such as presence of chills, degree of elevation of body temperature, number of days with fever, heart rate and peripheral oxygen saturation were also of limited value for separating infected from non-infected inflamed patients. Still, even though being of low importance for the segregation, the presence of such parameters will not negatively influence the calculation of the indicator value. In order to reduce the number of variables in the model, variables with low explanatory power were discarded i.e. VIP with values <1.0 in table 2.

By eliminating the discarded variables from the model a new embodiment of an O-PLS model based on 55 variables were obtained. This model had similar explanatory capacity (R ² Y=0.50) as the model based on all 139 variables (R ² Y=0.47), i.e. was equally able to distinguish infected from non-infected febrile neutropenic hematology patients. This is obvious by inspection of figure 1 and figure 4.

Furthermore, the 55 -variable model proved to be more stable than the original 139-variable model, as reflected by improved cross-validation (Q ² Y= 0.32 as compared to -0.0082).

The variables that contributed most to the 55-variable model are shown in figure 2. Variables that were higher in infected patients compared to non-infected systemically inflamed patients have values above zero; the height of the columns reflects how much each variable contributes to the model. Variables that were increased in infected compared to non-infected patients included procalcitonin, prothrombin time (measured in units of "International Normalized Ratio", INR), bilirubin (both total, "Bil", and conjugated, "BilC" forms), Cortisol, IL-8 and IL-10, MCHC (mean corpuscule hemoglobin concentration, i.e. the hemoglobin contents of red blood cells) and the proteins ferritin and

prohepcidin. Conversely, infected patients had lower white blood cell counts (WBC), including reticulocytes, lymphocytes, neutrophils, eosinophils and monocytes, and decreased levels of the coagulation factors XII and X, as compared to patients with sterile inflammation (Figure 2C). These variables are indicated by columns with negative values. Erythrocyte volume fraction (EVF), LDL, "total iron binding capacity of serum" (TIBC), cholesterol, hemopexin, transferrin, hemoglobin, and sodium (Na) were also lower in the infected patients than in those with sterile inflammation.

Being a smoker was also more associated with sterile than infectious inflammation.

The embodiment of a 55-variable O-PLS model consists of the following variables, using the short name notation introduced in table 2: Procalc, Urea, WBC, IL-10, Bil, IL-8, XII, EVF, BilC, Hb, Smok, Cort, Lymph, PAI, vWF, HMGB, Prohep, LDL, Neu, C4BPb, CysC, Glu, Prot S, CyA, Hepc, MCHC,

Eos, AFP, LD, BPd, Choi, Ret, IgE, Per, PAC, IL-1, MCV, INR, X, Mono, CRP, Ferri, K, Hpx, TropT, HbA, Trans, IgM, Na, Bas, EPC, Oro, APTT, TSH, TIBC. Hence, there is provided herein a method wherein the information about a patient and/or blood parameters used in a method of providing a clinical decision support comprise: Procalc, Urea, WBC, IL-10, Bil, IL-8, XII, EVF, BilC, Hb, Smok, Cort, Lymph, PAI, vWF, HMGB, Prohep, LDL, Neu, C4BPb, CysC, Glu, Prot S, CyA,

Hepc, MCHC, Eos, AFP, LD, BPd, Choi, Ret, IgE, Per, PAC, IL-1, MCV, INR, X, Mono, CRP, Ferri,

K, Hpx, TropT, HbA, Trans, IgM, Na, Bas, EPC, Oro, APTT, TSH, and TIBC. In order to develop a diagnostic method that could be used in the clinical setting to distinguish infected from non-infected febrile neutropenic patients with hematologic malignancies it was important to remove time-consuming analyses not performed on a daily basis in tertiary care hospitals. For example, white blood cell differential counts cannot be performed by the usual automated procedure when cell counts are very low, e.g. in neutropenic patients. A number of non-routine analyses were kept in the model, some with high explanatory power, e.g. HMGB 1 and cytokines. These analyses could easily be automated. This pragmatic pruning procedure generated a reduced model consisting of 40 variables with the same explanatory capacity (R ² Y= 0.47) as the model based on 139 variables, and a stability comparable to the model composed of 55 variables (Q ² Y=0.30). Essentially, the same variables were shown to be associated with infection in the embodiment of a reduced 40-variable O- PLS model as in the embodiment of a 55-variable O-PLS model.

The embodiment of a reduced 40-variable O-PLS model consists of the following variables, using the short name notation introduced in table 2: Procalc, Urea, WBC, IL-10, Bil, IL-8, XII, EVF, BilC, Hb, Smok, Cort , PAI, vWF, HMGB, Prohep, LDL, C4BPb, CysC , Prot S, CyA, Hepc, MCHC, AFP, BPd, Choi, Ret, Per, PAC, IL-1, MCV, INR, X, CRP, TropT, Trans, Na, EPC, APTT, TIBC. Hence, there is provided herein a method wherein the information about a patient and/or blood parameters used in a method of providing a clinical decision support comprise: Procalc, Urea, WBC, IL-10, Bil, IL-8, XII, EVF, BilC, Hb, Smok, Cort , PAI, vWF, HMGB, Prohep, LDL, C4BPb, CysC , Prot S, CyA, Hepc, MCHC, AFP, BPd, Choi, Ret, Per, PAC, IL-1, MCV, INR, X, CRP, TropT, Trans, Na, EPC, APTT, and TIBC.

A common pitfall in multivariate modelling is that models based on data derived from one set of patients may not be applicable to a new set of patients. For example, the 13 infected patients that formed the basis of our models may have displayed altered levels of some parameters for reasons not connected to infection. A model containing many such non-informative parameters would have no or poor capacity to predict infection. It is therefore essential to test the performance of any model in a new group of patients. A validation population of patients (n =10) was recruited, using the same criteria as before, i.e. neutropenic hematology patients presenting with fever of unknown cause. They were also categorized as infected and non-infected in a blinded fashion by the infectious disease specialist. A total of 40 variables, i.e. blood biomarkers, clinical and physiological data, derived from each patient were used as input data to said reduced model. The reduced model calculates the probability that the person in question is infected, based on the 40-variable input data. The reduced model predicted that five of the ten patients had an infectious origin of their systemic inflammation (prediction score: > 0.5) (Table 3). Four of these patients were also judged to be infected by the infectious disease specialist (CW86, 84, 82 and 83), Table 3. One patient (CW85) was categorized as infected by the model, but not by the clinician. This patient had suspected septic arthritis and skin rash (erysipelas) of the left ankle. PCR analysis of joint fluid revealed the opportunistic bacterium Kocuria (Table 3). This microbial finding was not deemed to be clinically relevant by the infectious disease specialist. Based on the clinical assessment of the infectious disease specialist, taken to be the "gold standard", the model correctly categorized all (4/4) of the patients with microbiologically documented infection as infected, and 5/6 of the non-infected patients as non- infected (Table 3).

Table 3: Patients used to validate the embodiment of a 40-paramter O-PLS model, NC denotes being classified as having clinically relevant infectious disease.

The embodiments disclosed herein are all suitable for being implemented in a computer with a processor and a memory, wherein the memory contains instructions for carrying out the inventive method.

The computer might in one embodiment be configured to control the automatic testing and analysis of a blood sample from a patient in order to achieve at least one of the variables needed in the model.

In order to further elucidate the invention, the method of providing decision support in neutropenic fever diagnosis is illustrated in a flowchart in figure 5.

In a first step 501 is patient information provided. Such information may comprise information about whether the patient is a smoker, use of indwelling catheter into the blood stream, use of

immunosuppresive medication and diastolic blood pressure.

In a second step 502 is a blood sample from the patient provided for analysis.

In a third step 503 blood parameters from said blood sample are measured. These measurements may comprise measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins.

In a fourth step 504 is an indicator value calculated using said blood parameters and information about said patient in a regression model. The regression model is obtained by utilizing a pattern recognition method known as O-PLS (orthogonal projection onto latent structures). This O-PLS method is developed from principal component analysis and is particularly useful for handling hundreds of available parameters and to identify the most significant parameters among these. The O-PLS method was used to derive a regression formula of the form:

IV =∑ ₀ Ci x _Pi (1) Wherein IV indicates the indicator value, N gives the maximum number of parameters, <¼ is the constant for parameter i, pi is the parameter value for parameter i. In one embodiment may the parameter values pi be normalized by subtracting the mean value of the parameter from the actual parameter value followed by dividing this difference with the standard deviation of the parameter values. This scaling of parameters is within the art called mean-centering and unit variance scaling.

In a fifth step 505 it is determined if the indicator value (prediction score) is indicative of whether said patient has neutropenic fever due to a systemic infection 506 or due to non-infectious inflammation 507, respectively. A value close to 1.0 indicates that the probability is high that the patient has an infectious cause of the fever, whereas a value close to 0 indicates that there is very low probability that the fever is caused by infection.

In one embodiment may the measurement of blood cells comprise at least one from the group of white blood cell count, lymphocyte count, neutrophil count, eosinophil count, basophil count, monocyte count, reticulocyte count, erythrocyte volume fraction, mean corpuscular hemoglobin concentration, mean corpuscular volume, erythrocyte particle concentration. Accordingly, it is e.g. referred to a measurement of a particular amount, volume and/or concentration of a blood cell or a component related to a blood cell.

In one embodiment of the method, the measurement of iron turn-over comprises at least one from the group of: total iron binding capacity, prohepcidin, hepcidin, transferrin, ferritin, hemopexin, hemoglobin. Total iron binding capacity is defined as the amount of iron (in mmol) that binds to 1 Liter of serum/plasma.

In one embodiment of the method, the measurement of blood coagulation comprises at least one from the group of: factor XII, von Willenbrand factor, protein S, factor X, activated partial prothrombin time, international normalized ratio, plasminogen activator-inhibitor 1. The measurement of the blood coagulation hence refers to a measurement of concentration of a factor involved in the blood coagulation process, or the activity of the coagulation system, such as the time it takes before a blood sample clots after adding a clotting agent.

In one embodiment of the method, the measurement of organ stress comprises at least one from the group of: bilirubin (conjugated), bilirubin, high mobility group box-1, cystatin C, alpha-fetoprotein, troponin- T, urea, sodium, potassium, lactate dehydrogenase. These represent compounds leaking out from, or secreted by, stressed and /or damaged cells. In one embodiment of the method, the measurement of metabolism comprises at least one from the group of: low-density lipoprotein, Cortisol, cholesterol, glucose, thyroid stimulatory hormone, glycosylated hemoglobin (HbAlc). These compounds take part in, or regulate, the metabolism of the body, including generation of energy.

In one embodiment of the method, the measurement of acute phase reactants comprises at least one from the group of: C- reactive protein, procalcitonin, complement 4-binding factor B, orosomucoid. These proteins are produced by liver cells (hepatocytes) in response to inflammatory triggers.

In one embodiment of the method, the measurement of cytokines and immunoglobulins comprises at least one from the group of: interleukin-1, interleukin-10, interleukin-8, immunoglobulin M, immunoglobulin E. Cytokines are typically present in the picogram to nanogram per milliliter concentrations, while IgE is measured in kU/L, where one U represents 2.4 ng. IgM is measured in g L.

Of course all of the (groups of) parameters and/or information about the patient mentioned herein may be combined and used in a method as defined herein as the method utilizes a combination of the parameters from the defined groups and the patient information to arrive at an indicator value. Minor modifications of parameter(s) and/or information mentioned herein are also encompassed when providing the same function and/or information to the outcome of the method.

There is also provided a method wherein the patient information comprises at least one from the group of: whether the patient is a smoker, use of indwelling catheter into the blood stream (e.g. Peripheral vein sampling, port-a-catheter), immunosuppressive medication (e.g. Cyclosporin A treatment) , and diastolic blood pressure.

There is also provided herein a method wherein the regression model is a partial least square regression model.

There is also disclosed herein a method wherein the parameters and/or information are evaluated in a method of providing clinical decision support in neutropenic fever diagnosis are; i.e. white blood cell count, lymphocyte count, neutrophil count, eosinophil count, basophil count, monocyte count, reticulocyte count, erythrocyte volume fraction, mean corpuscular hemoglobin concentration, mean corpuscular volume, erythrocyte particle concentration, total iron binding capacity, prohepcidin, hepcidin, transferrin, ferritin, hemopexin, hemoglobin factor XII, von Willenbrand factor, protein S, factor X, activated partial prothrombin time, international normalized ratio, plasminogen activator- inhibitor 1 bilirubin (conjugated), bilirubin, high mobility group box-1, cystatin C, alpha-fetoprotein, troponin- T, urea, sodium, potassium, lactate dehydrogenase, low-density lipoprotein, Cortisol, cholesterol, glucose, thyroid stimulatory hormone, glycosylated hemoglobin (HbAlc). C- reactive protein, procalcitonin, complement 4-binding factor B, orosomucoid, interleukin-1, interleukin-10, interleukin-8, immunoglobulin M, immunoglobulin E, whether the patient is a smoker, use of indwelling catheter into the blood stream (e.g. Peripheral vein sampling, port-a-catheter), immunosuppressive medication (e.g. Cyclosporin A treatment) , and diastolic blood pressure.

In another embodiment is the method of providing decision support in neutropenic fever diagnosis as presented herein implemented in a computer comprising a processor and a memory, wherein the memory comprises instructions for receiving information about a patient. The memory comprises instructions for receiving blood parameters from an analysis of a blood sample. The blood parameters comprise measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, as further exemplified herein. The memory further comprises instructions for calculating an indicator value by using said measurement of blood parameters and said information about said patient in a regression model, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a noninfectious inflammation, respectively. The computer-implemented method may utilize any one of the blood parameters and/or information presented herein.

There is further provided herein a computer readable medium having stored thereupon a computer implemented method as defined herein.

There is further provided herein a computer readable medium having stored thereupon a computer program which when executed calculates an indicator value in a regression model by using as input measurements of blood parameters and information about a patient, said measurement of blood parameters comprising measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins, wherein the indicator value is indicative of whether said patient has neutropenic fever due to a systemic infection or due to a noninfectious inflammation, respectively. There is also provided herein a kit for performing a method of providing decision support in neutropenic fever diagnosis as disclosed herein, wherein said kit comprises technical means for performing said method as well as optionally instructions for use. There is further provided an kit for providing clinical decision support in neutropenic fever diagnosis in a patient, said kit comprising analytical detection and/or measuring reagents for measuring blood parameters from a blood sample from said patient, wherein said blood parameters comprises measurements of blood cells, iron turn-over, blood coagulation, organ stress, metabolism, acute phase reactants, cytokines and immunoglobulins as defined in any of the embodiments herein and optionally instructions for use. There is also provided herein a kit further comprising a computer readable medium as defined herein. Clinical measures may be entered into the software manually. The kit may comprise of a specified suitable biochemical analyses (analytical detection and/or measuring reagents), e.g. for the selected parameters/variables, as exemplified herein, that may be combined into one assay, or a few assays and thereafter performed by a suitable automated analysis instruments. Hence there is provided a pre-prepared kit containing the appropriate means in the form of reagents etc. for performing the biochemical analysis subsequently applied to an automated analysis instruments and the computer software (suitably presented on a computer readable medium) for executing the regression model presented herein. As mentioned, clinical results from the biochemical analysis and the patient information may be manually entered into the software.

In yet another embodiment may the regression model be a probabilistic regressions model such as for example a probit regressions model or logistic regressions model such as for example a logit model.

REFERENCES

1. Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med. 1999;340(6):448-54.

2. Trygg J, Wold S. Orthogonal projections to latents structures (O-PLS). J Chemometrics.

2002; 16: 119-28.

3. Multi- and megavariate data analysis, Part I, Basic Principles and Applications, Second edition, L Eriksson, E. Johansson, N Kettaneh-Wold, J Trygg, C Wikstrom, S Wold. Umetrics AB, Edition date 2006. ISBN-10:91-973730-2-8