Diagnostic methods, systems, and kits for identifying patients with systemic inflammatory response syndrome who are likely to progress to sepsis.

Description of the Invention

SUMMARY OF THE INVENTION

The present invention provides methods, reagents, and kits for the detection of early sepsis. The present invention results from the discovery that an improved method of detecting patients who will progress to sepsis can be obtained by monitoring a plurality of suitable biological markers over a period time, independently deriving for each marker a statistical measure of extreme value of the marker over the period of time, and detecting early sepsis based on the combination of marker statistics.

Thus, the present invention provides methods of detecting early sepsis in a patient, comprising: a) monitoring a plurality of biological markers over a period of time, b) independently deriving for each marker a statistical measure of extreme value of the marker over the period of time, and c) applying a decision rule to the combined marker statistics to detect early sepsis in the patient.

A critical aspect of the present invention is the method of using the marker data obtained during monitoring. This aspect results from the discovery that consideration of the extreme marker values over the monitoring period may provide a better indicator of early sepsis than consideration of values obtained at one particular time, such as the most recently measured values alone. Thus, for each marker, a statistic that is a measure of the extreme value of a marker is calculated from the data for that marker accumulated over the interval of time that the patient is monitored and, furthermore, the statistic for each marker is calculated independently of the statistics for the other markers. This aspect of the invention distinguishes the present methods from methods that are based on the values of the monitored markers obtained at approximately the same time, as is the case, for example, in methods based on the current status of the patient. The present methods provide for the early detection of sepsis with greater sensitivity and specificity.

In a preferred embodiment, the methods are used as part of the process of monitoring SIRS patients, who are at risk of developing sepsis, for the impending onset of sepsis. A patient is monitored starting from the time that the patient is first identified to be at risk, most frequently upon becoming SIRS positive during an ICU stay following a surgical procedure. Following each measurement of the patient's markers during the course of monitoring, the detection criteria of the present methods will be applied to the data accumulated from both the previous and present measurements. Monitoring is continued until either the patient is identified as one who will progress to sepsis or the monitoring is discontinued.

Biological markers useful in the present methods are those that are informative of the state of the immune system in response to an infection or other severe clinical insult. Suitable markers include, for example, leukocyte count, cell surface markers, and soluble markers. In a preferred embodiment, the plurality of markers comprises at least one marker for a pro-inflammatory response and at least one marker for a compensatory anti-inflammatory response. Markers for a pro-inflammatory response include, for example, leukocyte count and cell-surface activation markers, such as adhesion molecules, including integrins, in particular, .beta.2-integrins such as CD11b, and molecules in the Fc receptor family, including Fc.gamma. receptors such as CD64. Markers for a compensatory anti-inflammatory response include markers of monocyte deactivation, such as MHC Class II molecules, in particular, HLA-DR and HLA-DQ. Other makers of either pro-inflammatory response or of compensatory anti-inflammatory response may be suitable, and can be selected following the teaching provided herein.

In a preferred embodiment, the plurality of biological markers comprises at least one marker for monocyte deactivation and at least one marker of neutrophil activation. Preferred markers of monocyte deactivation are the HLA Class II molecules expressed on peripheral blood cells, preferably, HLA-DR, and more preferably, monocyte-associated HLA-DR. Preferred markers of neutrophil activation include CD64 and CD11b, preferably, neutrophil-associated CD64 and neutrophil-associated CD11b.

The decision rule in the present methods is based on the extreme marker values over the monitoring period. In some embodiments, the statistical measure of extreme value is the maximum or minimum of the values obtained over the course of monitoring. Whether the maximum or minimum is measured depends on the marker. For example, decreased HLA-DR expression is a measure of monocyte deactivation and, thus, the minimum value of HLA-DR is most relevant in the present methods. Conversely, increased CD11b and CD64 are measures of a pro-inflammatory response, and the maximum values of these markers are most relevant.

In some embodiments, the statistical measure of extreme value is calculated directly from the set of data obtained for a marker. Alternatively, the set of data obtained first may be fitted to a curve, such as a polynomial or spline, and the statistical measure of extreme value is derived from the fitted curve. Such methods allow for interpolation of values in between data collection times and may increase detection sensitivity. Methods for fitting a curve to a set of points are well known in the literature.

The detection of sepsis is carried out using a decision rule applied to the marker statistics to classify patients as septic or non-septic. An optimal decision rule typically is generated from a controlled study in which a population of at-risk patients is monitored, and the subpopulation of patients who develop sepsis are compared to the subpopulation of patients who do not develop sepsis. The generation of a decision rule from a multivariate discrimination model is well known in the art and is described more fully, infra. In general, the resulting decision rule is a two-valued function of the data, which serves to identify the presence or absence of sepsis. However, multi-valued decision rules may be used that provide results interpretable as the likelihood that sepsis is present.

In a preferred embodiment, the plurality of makers comprises HLA-DR, CD11b, and CD64, and the statistics derived are the minimum HLA-DR expression, the maximum CD11b expression, and the maximum CD64 expression. Septic patients are identified using a decision tree based on these three statistics, preferably based on the minimum HLA-DR expression and the sum of the maxima of the CD11b and CD64 expressions. The threshold values of the decision tree are determined empirically, as described in the examples.

The present invention also provides kits useful for carrying out the present methods. The kits include a computer readable medium containing programming that implements the decision rule. In addition, the kits may include reagents, devices, and instructions for carrying out the monitoring of a patient's biological markers.

As discussed above, the course by which a patient progresses to death from sepsis is well-known and has been described as a continuum from a state termed systemic inflammatory response syndrome (SIRS) to successive states of sepsis, severe sepsis, septic shock, multiple end-organ failure (MODS), and death. The present methods enable the early detection of the progression from SIRS to sepsis, thereby enabling early intervention. It will be clear that, as the present methods enable the detection of sepsis prior to clinical suspicion, the present methods also enable the detection of any of these subsequent complications of sepsis prior to clinical suspicion.

DETAILED DESCRIPTION OF THE INVENTION

Methods of the Invention

The present invention provides methods of detecting early sepsis in a patient, comprising the steps of: d) monitoring a plurality of biological markers over a period of time, e) independently deriving for each marker a statistical measure of extreme value of the marker over the period of time, and f) applying a decision rule to the combined marker statistics to detect early sepsis in the patient. Aspects of the methods are described in more detail, below. Biological Markers

Biological markers useful in the present methods are, in general, those that are informative of the state of the immune system in response to an infection or other severe clinical insult, including, for example, leukocyte count, cell surface markers, and soluble markers. Biological markers may comprise any molecule or molecules obtainable from a patient, such as soluble or cell surface proteins, or any measurable physiological and/or clinical parameter, such as body temperature, respiration rate, pulse, age, blood pressure, white blood cell count, etc.

Biological markers that are informative of the state of the immune system in response to an infection include, by way of example and not of limitation, cell-surface proteins such as CD64 proteins, CD11b proteins, HLA Class II molecules, including HLA-DR proteins and HLA-DQ proteins, CD54 proteins, CD71 proteins, CD86 proteins, surface-bound tumor necrosis factor receptor (TNF-R), pattern-recognition receptors such as Toll-like receptors, soluble markers such as interleukins IL-1, IL-2, IL-4, IL-6, IL-8, IL-10, IL-11, IL-12, IL-13, and IL-18, tumor necrosis factor alpha (TNF-.alpha.), neopterin, C-reactive protein (CRP), procalcitonin (PCT), 6-keto F1.alpha., thromboxane B.sub.2, leukotrienes B4, C3, C4, C5, D4 and E4, interferon gamma (IFN.gamma.), interferon alpha/beta (IFN .alpha./.beta.), lymphotoxin alpha (LT.alpha.), complement components (C'), platelet activating factor (PAF), bradykinin, nitric oxide (NO), granulocyte macrophage-colony stimulating factor(GM-CSF), macrophage inhibitory factor (MIF), interleukin-1 receptor antagonist (IL-1ra), soluble tumor necrosis factor receptor (sTNFr), soluble interleukin receptors sIL-1r and sIL-2r, transforming growth factor beta (TGF.beta.), prostaglandin E.sub.2 (PGE.sub.2), granulocyte-colony stimulating factor (G-CSF), interferon .alpha./.beta., and other inflammatory mediators. Biological markers also include RNA and DNA molecules that encode or are otherwise indicative of the aforementioned protein markers.

In preferred embodiments, a plurality of markers is measured that comprises at least one marker for a pro-inflammatory response and at least one marker for a compensatory anti-inflammatory response. Markers for an pro-inflammatory response include, for example, leukocyte count and cell-surface activation markers, such as adhesion molecules, including integrins, in particular, .beta.2-integrins such as CD11b, and molecules in the Fc receptor family, including Fc.gamma. receptors such as CD64, and soluble markers such as CRP, PCT, IFN.gamma., LT.alpha., IL-1.beta., IL-2, IL8, IL-12, IL-18, TNF-.alpha., C', LTB.sub.4, PAF, bradykinin, NO, GM-CSF, and MIF. Markers for a compensatory anti-inflammatory response include cell-surface markers of monocyte deactivation, such as MHC Class II molecules, in particular, HLA-DR and HLA-DQ, and soluble markers such as IL-1ra, sTNFr, sIL-1r, TGF.beta., IL-4, IL6, IL-10, IL-11, IL-13, PGE.sub.2, G-CSF, and IFN .alpha./.beta..

In a preferred embodiment, the plurality of biological markers comprises at least one marker for neutrophil activation and at least one marker of monocyte deactivation. Increased expression of CD64 and CD11b is recognized as a sign of neutrophil and monocyte activation. Preferred cell-surface markers of neutrophil activation include CD64 and CD11b expressed on neutrophils. Preferred cell-surface markers of monocyte deactivation are the HLA Class II molecules expressed on peripheral blood cells, preferably, HLA-DR expressed on monocytes.

Each of the markers discussed above is known to change in response to an infection and may therefore be useful as a marker for an inflammatory condition. Various other biological markers that are indicative of an inflammatory condition are known to those skilled in the art and will suggest themselves upon review of this disclosure. It is expected that not all markers that are informative of the state of the immune system in response to an infection are equally informative. Consequently, the sensitivity and specificity of a decision rule constructed for use in the present methods will depend on the particular markers selected. Preferred sets of markers can be selected empirically using routine experimentation following the teaching herein.

Biological markers may be obtained from any host compartment, i.e., from blood, serum, urine, sputum, stool, or other biological fluid sample or tissue sample from a host or patient. The host compartment sampled will generally vary according to the marker, but the sampling should preferably be minimally invasive and easily performed by conventional techniques.

Measurement of biological markers may be carried out by any conventional techniques. Measurements of biological marker molecules may include, for example, measurements that indicate the presence, concentration, expression level, or any other value associated with a marker molecule. Various spectroscopic techniques are available for measuring biological marker molecules, including UV, visible, and infrared spectroscopies. Fluorescent labels, radioactive labels, or other readily identifiable and quantifiable labels may be used to aid in measurement of marker molecules. The expression levels of cell-bound markers can be measured by flow cytometric techniques, and the expression levels of soluble markers can be characterized by immunosorbent assay techniques. Measurement of body temperature, respiration rate, pulse, blood pressure, or other physiological parameter markers can be achieved via clinical observation and measurement.

Monitoring

In the methods of the present invention, the patient is monitored for a period of time encompassing at least two measurements before the decision criteria are applied to the accumulated marker data. Monitoring typically involves regularly repeated measurements of the biological markers, such as a daily, hourly, or on a more frequent basis over one or more days. The time period over which markers are measured, and the frequency of measurements for each marker during the time period, will necessarily vary depending upon the presentation of the patient at the commencement of monitoring, the progression of the patient, and the particular markers selected for monitoring. In some instances, measurement of markers will occur on an hourly or more frequent basis over a part of a day, a single day, or over multiple days. The period and frequency of monitoring need not be the same for each patient or each marker. Furthermore, the period and frequency of monitoring used in model construction may differ from the period and frequency of monitoring when the decision rule derived from the model is subsequently applied to the patient marker measurements for disease detection.

The methods are useful in a clinical setting as part of the process of monitoring SIRS patients, who are at risk of developing sepsis, for the impending onset of sepsis. Thus, a patient is monitored starting from the time that the patient is first identified to be at risk, i.e., when the patient becomes SIRS-positive. Following each measurement of the patient's markers during the course of monitoring, the detection criteria of the present methods are applied to the data accumulated from both the previous and present measurements to determine if the patient has progressed to early sepsis. Monitoring is continued until either the patient is identified as exhibiting early sepsis or the monitoring is discontinued, typically because the patient is no longer SIRS-positive.

Although the methods are particularly useful for monitoring SIRS patients for the impending onset of sepsis, it will be clear that the methods may be used for patients who are considered to be, for whatever reason, enough at risk of sepsis that monitoring is warranted, even though they are not SIRS positive. Furthermore, it is recognized that the consensus definition of SIRS, used herein, may be supplanted by improved definitions in the future. It will be understood that the definition of SIRS is not a critical aspect of the invention and that the present methods will remain equally applicable.

Statistical Measures of Extreme Value

After monitoring the patient for a period of time during which two or more measurements of each marker are taken, a statistic is calculated for each marker that is a measure of extreme value of the marker over the period of time. The statistical measure of extreme value may be, for example, a measure of a maximum or minimum marker level or fitted value over a time period, a maximum or minimum increase or decrease in marker measurements or in fitted values over a time period, a maximum or minimum time spent either above or below a threshold, a maximum or minimum level of variability of measurements or fitted values from two or more time points, maximum or minimum slopes of trend lines, means of possibly discontiguous local maxima or minima, and the like. The selection of a particular statistical measure of extreme value is determined empirically, essentially as described in the examples.

Decision Rules

After a statistical measure of extreme value is derived for each marker, detection of early sepsis is carried out using a previously determined decision rule applied to the marker statistics to classify patients as positive or negative for early sepsis. The decision rule is obtained from a discrimination model, generated as described in general, below.

Model Construction

The discrimination model preferably is generated from marker data collected in a controlled study in which a population of SIRS-positive patients is monitored over time until at least one patient becomes clinically septic. The data from the subpopulation of SIRS patients who develop sepsis (converters) are compared to the data from the subpopulation of SIRS patients who do not develop sepsis (non-converters). Typically, a large number of biological markers are monitored during the controlled study, although only a subset of these markers may be used in the final decision rule. Although, in the simplest case, marker measurements from two patients, one a converter and one a non-converter, may be used to construct a discrimination model, obtaining marker measurements from a greater number of patients will generally provide a better statistical model. It will be understood that the inclusion of one of more control populations in the study may be beneficial and provide additional information useful in the generation of a discrimination model.

The present methods are used for detecting early sepsis prior to the time that clinically manifested sepsis would be either suspected or confirmed using previously described methods. Preferably, the methods detect early sepsis at least 12 hours, preferably 24 hours, prior to clinical suspicion of sepsis (predictive lead time) so that appropriate therapeutic treatment can be initiated early. For this reason, the discrimination model preferably is generated retrospectively using only data gathered from the converter patients up to a time prior to the clinical suspicion of sepsis corresponding to the desired predictive lead time. The choice of the desired predictive lead time will be based on clinical considerations; shorter or longer values may be useful depending on the clinical setting.

The biological markers monitored during the controlled study are selected from those known or suspected to be informative of the state of the immune system in response to an infection, as described above. Preferably, although not necessarily, an initial selection of these markers is carried out based on the data from the controlled study in order to identify those markers most likely to be informative in a discrimination model. This initial selection is carried out using univariate statistical tests to determine if the extreme value of the marker, considered alone, is statistically different between the converters and non-converters. Those markers that differ significantly between converters and non-converters are considered more likely to provide useful information regarding early sepsis, and are used in the subsequent generation of a multivariate discrimination model. As this is only a pre-screening procedure, the data compared may be from any time during the study, although comparing data from prior to the onset of clinically manifested sepsis in converters is preferable.

The marker measurement data obtained from the converter and non-converter subpopulations in the study are initially analyzed by calculating the marker statistics (statistical measure of extreme value) to be used in the final decision rule. The resulting values of the marker statistics are used in combination to form a multivariate discrimination model using classification or regression trees, logistic regression, log-linear discriminant analysis, discriminant analysis, neural networks, or other types of multivariate discrimination models. Statistical software useful for multivariate discrimination modeling is well known in the art and is commercially available from a number of vendors. For example, a commercial software product useful for multivariate discrimination modeling is SPLUS.RTM. 2000 by MathSoft, Inc (Cambridge, Mass.).

The methods are useful in a variety of clinical settings as part of the process of monitoring patients who are at risk of developing sepsis. It is recognized that the distributions of marker values obtained from patients may depend on the clinical setting, such as different wards of a hospital. For example, the data provided in the examples suggest that patients admitted to the SICU following elective surgery (referred to therein as clean surgery controls) may not be entirely representative of patients entering the SICU after emergency surgery following a serious accident or coronary episode. Preferably, a model is generated from converter and non-converter data obtained from patients in the same clinical setting as in the intended use of the resulting model.

Use of the Decision Rule

Once a discrimination model has been developed with data from the controlled study, the model may be applied de novo to marker measurement data from individual patients to detect the presence of early sepsis in the patient. Typically, the decision rule produced by the discrimination model is applied iteratively during patient monitoring. The decision rule is applied anew following each measurement time-point during monitoring, based on the recalculated statistical measures of extreme value. The decision rule is re-applied after each patient measurement either until the patient is identified as having progressed to early sepsis or until the patient is discharged from medical care or is otherwise no longer considered to be at risk for sepsis. For example, in situations where a patient is in a hospital or ICU, marker measurements may be made from patient blood samples taken on a daily basis, and the model decision rule may be applied each day following the measurements. If early sepsis is detected according to the decision rule, a therapeutic treatment for the patient may be initiated or the patient may be stratified into a clinical trial. If no early sepsis is detected after monitoring the patient for several consecutive days, the patient may be discharged from the hospital or ICU.

Kits

The invention also provides kits usable for practicing the subject methods. The kits may comprise a computer readable medium, such as a CD or floppy disk, which contains programming capable of creating one or more patient data files from marker measurements taken from one or more patients considered to be at risk for sepsis, modeling time series patterns of marker measurements over the time period, extracting a statistical feature or features from the time series patterns and applying a decision rule from a discrimination model to the patient data files to determine if early sepsis is present in the one or more patients. The kits also may comprise reagents, devices and instructions for carrying out the monitoring, such as, for example, vials for patient blood sample collection or blood sample separation, staining equipment such as vials of fluorescent-labeled antibodies for selected markers, vials of lysing solution, QuantiBRITE.RTM. calibration beads for the fluorescent labels, and printed instructions for the acquisition of patient blood samples at multiple time points over a period of time, staining conditions, lysing conditions, and thresholding and gating instructions for flow cytometric measurement of the markers. In the case of soluble markers, the kit may comprise one or more solid phase "sandwich" ELISA assays with a multi-well plate, vials of solution for washing the wells, vials of labeled antibodies for the markers of interest, vials of staining solution, and printed instructions for applying marker samples to the wells, washing, applying antibodies, and staining. In turn, the kits may also comprise equipment that permits the evaluation of soluble markers with materials such as cytometric bead array products.

Utility

The methods and kits of the invention are useful in providing for detection of early sepsis in individual patients before the manifestation of overt, clinical symptoms that are observable by a physician. The early detection of sepsis may lead to decreased mortality rates for patients and reduced costs for patient treatment by permitting treatments to be focused on patients who are developing sepsis, and may result in improved clinical outcomes generally. Detection of sepsis at least 12-24 hours prior to clinical suspicion of sepsis allows for administration of antibiotic, anti-inflammatory and/or other therapeutic treatments to patients at a time wherein such treatments are potentially most beneficial.

As is typical of diagnostic methods, it is intended that the prognostic methods of the present invention represent one tool for identifying early sepsis, i.e., patients who will progress to sepsis, and that the present methods may be used in combination with additional methods. Typically, a patient will be measured and/or monitored for a number of other biological markers, including both time-varying and non-time-varying markers, such as age at time of entry into the ICU, gender, type of surgical procedure, need for mechanical ventilation, type of physiological insult, or other factor or factors useful in providing early determination of disease onset. It will be clear that a clinician typically will consider the totality of the clinical data available in making a medical judgment.

Although the primary use of the present methods will be to detect early sepsis in human patients in a hospital setting, it will be clear that the present methods can be used with non-human patients that may be at risk for sepsis, such as in a veterinary setting.
 

Claim 1 of 3 Claims

1. A method of detecting early sepsis in a patient, wherein said method comprising the steps of: a) monitoring expression of markers comprising neutrophil-associated CD11b, neutrophil-associated CD64, and monocyte-associated HLA-DR over a period of time; b) independently deriving marker statistics comprising a maximum of said neutrophil-associated CD11b expression, a maximum of said neutrophil-associated CD64 expression, and minimum of said monocyte-associated HLA-DR expression; and c) applying a decision rule to the marker statistics from step (b) to detect early sepsis in said patient; wherein said decision rule is a classification tree that detects early sepsis if a sum of said maximum of said neutrophil-associated CD11b expression and said maximum of said neutrophil-associated CD64 expression is greater than a first threshold value, and said minimum of said monocyte-associated HLA-DR expression is less than a second threshold value; or said decision rule is a classification tree that detects early sepsis if a sum of said maximum of said neutrophil-associated CD11b expression and said maximum of said neutrophil-associated CD64 expression is greater than a third threshold value, and said maximum of said neutrophil-associated CD64 expression is greater than a fourth threshold value; or a sum of said maximum of said neutrophil-associated CD11b expression and said maximum of said neutrophil-associated CD64 expression is less than said third threshold value, and said minimum of said monocyte-associated HLA-DR expression is less than a fifth threshold value.

____________________________________________
If you want to learn more about this patent, please go directly to the U.S. Patent and Trademark Office Web site to access the full patent.