Decision tree-based system and method for estimating risk of acute coronary syndromes

Abstract

The present invention provides a decision tree-based system and method for estimating the risk of acute coronary syndrome (ACS) in a subject suspected of having ACS. Specifically, systems and methods are provided that employ an additive decision tree-based algorithm to process a subject's initial cardiac troponin I or T (cTnI or cTnT) concentration, a subject's rate of change of cTnI or cTnT, and at least one of the subject's age, subject's sex, subject's ECG value, and subject's blood parameter value to generate an estimated risk of ACS. Such severity classification allows for determining, for example, whether a patient requires emergency treatment.

Description

[Technical Field]

[0001] This application claims priority to U.S. Provisional Patent Application No. 62 / 316,037, filed March 31, 2016, and U.S. Provisional Patent Application No. 62 / 343,606, filed May 31, 2016, both of which are incorporated herein by reference in their entireties.

[0002] The present invention provides a decision-tree-based system and method for estimating the risk of acute coronary syndrome (ACS) in a subject suspected of having ACS or an ACS complication. Specifically, the system and method employ an additive decision-tree-based algorithm to process the subject's initial cardiac troponin I or T (cTnI or cTnT) concentration, the subject's rate of change of cTnI or cTnT, and the subject's age, subject's ECG value, subject's blood parameter value, and / or gender to generate an estimated risk of ACS or an ACS complication. Such severity classification allows, for example, determining whether the patient requires emergency treatment. [Background technology]

[0003] Patients with suspected ACS comprise up to 8 million emergency department visits in the United States each year. While 20–25% of these patients have a heart attack, the remainder do not. Standard medical practice currently treats false-positive cases triaged to cardiac catheterization laboratories, which is not without risk (approximately 1 in 1,000 adverse events and 1 in 10,000 deaths). Additionally, false-negative rates can result in misdiagnosis, leading to more severe outcomes or increased mortality. Finally, inappropriate and untimely statification of this patient population results in medical costs. The ACS population includes patients with ST-segment elevation myocardial infarction (STEMI), non-ST-segment elevation myocardial infarction (NSTEMI), and unstable angina (UA), with the latter two categories encompassing the majority of diagnosed ACS patients. Current guidelines and standards of care treat patients with suspected ACS based on diagnostic methods such as ECG and troponin measurement, in addition to clinical symptoms, history, and physical examination. For STEMI, the ECG is the gold standard for identifying patients with tissue damage reflected by ST-segment elevation on the ECG. For NSTEMI and UA, troponin measurement is the gold standard for helping stratify these patients into the appropriate risk category for MI and for rapid triage. Guidelines recommend measuring troponin and comparing it to a 99% threshold. However, such comparisons are not individualized with respect to the specific characteristics of the patient and may result in false negatives and false positives regarding ACS risk. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] US Patent Application Publication No. 2012 / 0076803 [Patent Document 2] U.S. Patent No. 8,535,895 [Non-patent literature]

[0005] [Non-Patent Document 1] Torres and Moayedi, 2007, Clin. Geriatr. Med.23(2):307-25, vi [Non-patent document 2] Li et al., Arch Cardiovasc Dis., March 2016, 109(3):163-70 Summary of the Invention [Problem to be solved by the invention]

[0006] The present invention provides a decision-tree-based system and method for estimating the risk of acute coronary syndrome (ACS) and / or ACS complications in a subject suspected of having ACS or an ACS complication. Specifically, the system and method employ an additive decision-tree-based algorithm to process the subject's initial cardiac troponin I or T (cTnI or cTnT) concentration, the subject's rate of change of cTnI or cTnT, the subject's age, the subject's ECG value, the subject's blood parameter values ​​(e.g., white blood cell mean volume, platelet population data, and red blood cell subpopulation data or other complete blood count data generated by a hematology analyzer), and / or gender to generate an estimated risk of ACS and / or ACS complications. Such severity classification can, for example, determine whether the patient requires emergency treatment. [Means for solving the problem]

[0007] In some embodiments, the methods provided herein for reporting an acute coronary syndrome (ACS) (e.g., of a subject suspected of having an ACS) and / or a predicted risk of an ACS complication (e.g., of a subject suspected of having an ACS complication) include a step of a) obtaining subject values, the subject values ​​including at least one of: i) the subject's gender value, ECG (electrocardiogram) value, blood parameter value (e.g., white blood cell mean volume, platelet population data, red blood cell subpopulation data, other complete blood count data generated by a hematology analyzer), and the subject's age value; ii) an initial cardiac troponin I and / or T (cTnI or cTnT) concentration from an initial sample from the subject; and iii) a first sample and / or a second subsequent sample from the subject. a) determining a corresponding first concentration of cTnI and / or cTnT and / or a corresponding second subsequent concentration of cTnI and / or cTnT from the subject's cTnI and / or cTnT values; b) processing the subject values ​​using a processing system such that a probable risk of ACS and / or ACS complications in the subject is determined; and c) reporting the subject's probable risk of ACS and / or ACS complications determined by the processing system, wherein the processing system comprises: i) a computer processor; and ii) a non-transitory computer memory (e.g., a single memory element or multiple distributed memory elements) comprising one or more computer programs and a database, wherein the one or more computer programs comprise a rate of change algorithm and an additive tree algorithm, and wherein the database comprises at least M decision trees (e.g., 4...20...50...100...500...800...and (1000 or more decision trees), each decision tree comprising at least two predetermined splitting variables and at least three predetermined terminal node values, wherein the at least two predetermined splitting variables are selected from the group consisting of a threshold for a rate of change value of cTnI and / or cTnT, a threshold for an initial cTnI and / or cTnT concentration value, and at least one of a gender value, a threshold for an ECG value, a threshold for a blood parameter value, and a threshold for an age; and wherein one or more computer programs, in conjunction with a computer processor, are configured to: A) apply a rate of change algorithm to determine a rate of change of the subject's initial cTnI and / or cTnT concentration, a first subsequent cTnI and / or cTnT concentration, and a second subsequent cTnI and / or cTnT concentration; B) determining a rate of change value for the subject's cTnI and / or cTnT from at least two of the subject's cTnI and / or cTnT concentrations, the subject's initial cTnI and / or cTnT concentrations, and at least one of the subject's gender value, the subject's ECG value, the subject's blood parameter and / or age value to a database to determine terminal node values ​​for each of the at least M decision trees; and C) applying an additive dendrogram algorithm to: I) determine a composite value from the M terminal node values; and II) process the composite value to determine an estimated risk of ACS and / or ACS complications in the subject.

[0008] In certain embodiments, the processing of the composite value to determine an estimated risk comprises determining an index value and comparing the index value to a lookup table of estimated risk index values ​​to determine the estimated risk. In certain embodiments, the ACS is Type I myocardial infarction, and the subject is determined to be at low risk if the index value is less than about 1.1, at moderate risk if the index value is between about 1.1 and about 57.0, and at high risk if the value is greater than about 57.1. In other embodiments, the ACS is myocardial infarction, and the subject is determined to be at low risk if the index value is less than about 3.0, at moderate risk if the index value is between about 3.1 and about 48.9, and at high risk if the value is greater than about 49.0.

[0009] In other embodiments, provided herein are methods for reporting an estimated risk of an ACS and / or ACS complication in a subject (e.g., suspected of having an acute coronary syndrome (ACS) or an ACS complication), comprising the steps of: a) obtaining subject values, the subject values ​​including: i) at least one of the subject's gender value (e.g., any value where the female value is less than the male value), the subject's age value, the subject's ECG value, and the subject's blood parameter value; ii) an initial cardiac troponin I or T (cTnI or cTnT) concentration from an initial sample from the subject; and iii) a rate of change value of the subject's cTnI and / or cTnT based on at least two samples obtained from the subject at different times; b) processing the subject values ​​using a processing system such that an estimated risk of the subject's ACS and / or ACS complication is determined; and c) certain embodiments further include the step of reporting the subject's estimated risk of ACS and / or ACS complication determined by the processing system, the one or more computer programs, in conjunction with the computer processor, are adapted to: A) apply a subject's cTnI and / or cTnT rate of change values, the subject's initial cTnI and / or cTnT concentration values, and the subject's gender and / or age values ​​to the database to generate at least M decision trees (e.g., 4...20...50...100...500...800...).and B) applying an additive dendrogram algorithm to: I) determine a composite value from the M terminal node values; and II) process the composite value to determine an estimated risk of ACS and / or ACS complications in the subject.

[0010] In certain embodiments, a processing system provided herein comprises: a) a computer processor; and b) a non-transitory computer memory including one or more computer programs and a database, wherein the one or more computer programs include an additive dendrogram algorithm; and the database includes at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), each decision tree including at least two predetermined splitting variables and at least three predetermined terminal node values, wherein the at least two predetermined splitting variables include a threshold for a cTnI and / or cTnT rate of change value, a threshold for an initial cTnI and / or cTnT concentration value, a gender value, a threshold for an ECG value, a threshold for a blood parameter value, and an age. and the one or more computer programs, in conjunction with the computer processor, are configured to: A) apply the threshold value for the subject's cTnI and / or cTnT rate of change value, the threshold value for the subject's initial cTnI and / or cTnT concentration value, and at least one of a subject's gender value, a subject's age value, a subject's ECG value, and a subject's blood parameter value to the database to determine terminal node values ​​for each of at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), and B) apply an additive tree algorithm to I) determine a composite value from the M terminal node values, and II) process the composite value to determine the subject's estimated risk of ACS and / or ACS complications. In certain embodiments, the system further includes a report providing the subject's estimated risk of ACS and / or ACS complications.

[0011] In some embodiments, the non-transitory computer memory element provided herein includes one or more computer programs configured to access a database, the one or more computer programs including a rate of change algorithm and an additive dendrogram algorithm, the database including at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), each decision tree including at least two predetermined splitting variables and at least three predetermined terminal node values, the at least two predetermined splitting variables selected from the group consisting of a threshold cTnI and / or cTnT rate of change value, a threshold initial cTnI and / or cTnT concentration value, a gender value, a threshold ECG value, a threshold blood parameter value, and a threshold age, and the one or more computer programs are configured to access a database including at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), each decision tree including at least two predetermined splitting variables and at least three predetermined terminal node values, the at least two predetermined splitting variables selected from the group consisting of a threshold cTnI and / or cTnT rate of change value, a threshold initial cTnI and / or cTnT concentration value, a gender value, a threshold ECG value, a threshold blood parameter value, and a threshold age, and In certain embodiments, the non-transitory computer memory element is configured, together with the processor, to: i) apply a rate of change algorithm to determine a rate of change value for cTnI and / or cTnT for the subject from at least two of the subject's initial cTnI and / or cTnT concentration, the subject's first subsequent cTnI and / or cTnT concentration, and the subject's second subsequent cTnI and / or cTnT concentration; ii) determine terminal node values ​​for each of the at least M decision trees by applying the subject's rate of change value for cTnI and / or cTnT, the subject's initial cTnI and / or cTnT concentration, and at least one of the subject's gender value, the subject's ECG value, the subject's blood parameter value, and the subject's age value to the database; and iii) apply an additive tree algorithm to a) determine a composite value from the M terminal node values, and b) process the composite value to determine an estimated risk of ACS and / or ACS complications for the subject.

[0012] In certain embodiments, the non-transitory computer memory element provided herein includes one or more computer programs configured to access a database, wherein the one or more computer programs include an additive dendrogram algorithm, and wherein the database includes at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), each decision tree including at least two predetermined splitting variables and at least three predetermined terminal node values, wherein the at least two predetermined splitting variables include a threshold for a rate of change value of cTnI and / or cTnT, a threshold for an initial cTnI and / or cTnT concentration value, a threshold for an ECG value, a threshold for a blood pressure level ... and a threshold value for the fluid parameter value, a gender value, and an age threshold value, and the one or more computer programs, in conjunction with the computer processor, are configured to: A) determine terminal node values ​​for each of the at least M decision trees by applying the database to the subject's cTnI and / or cTnT rate of change value, the subject's initial cTnI and / or cTnT concentration, and at least one of the subject's gender value, the subject's age value, the subject's ECG value, and the subject's blood parameter value; and B) apply an additive dendrogram algorithm to: I) determine a composite value from the M terminal node values; and II) process the composite value to determine the subject's estimated risk of ACS and / or ACS complications. In certain embodiments, the non-transitory computer memory element further comprises the database.

[0013] In some embodiments, a processing system provided herein comprises: a) a computer processor; and b) a non-transitory computer memory including one or more computer programs and a database, wherein the one or more computer programs include a rate of change algorithm and an additive dendrogram algorithm; the database includes at least M decision trees (e.g., 4...20...50...100...500...800...1000 or more decision trees), each decision tree including at least two predetermined splitting variables and at least three predetermined terminal node values, wherein the at least two predetermined splitting variables are selected from the group consisting of at least one of the following: a threshold for a rate of change value of cTnI and / or cTnT, a threshold for an initial cTnI and / or cTnT concentration value, a gender value, an ECG threshold, a blood parameter threshold, and an age threshold; and wherein the one or more computer programs include a non-transitory computer memory including one or more computer programs and a database, wherein the one or more computer programs include a rate of change algorithm and an additive dendrogram algorithm; and a computer processor configured to: i) apply a rate of change algorithm to determine a rate of change value for cTnI and / or cTnT for the subject from at least two of the subject's initial cTnI and / or cTnT concentration, the subject's first subsequent cTnI and / or cTnT concentration, and the subject's second subsequent cTnI and / or cTnT concentration; ii) determine terminal node values ​​for each of the at least M decision trees by applying the subject's rate of change value for cTnI and / or cTnT, the subject's initial cTnI and / or cTnT concentration, and at least one of the subject's gender value, the subject's ECG value, the subject's blood parameter value, and the subject's age value to a database; and iii) apply an additive tree algorithm to a) determine a composite value from the M terminal node values; and b) process the composite value to determine an estimated risk of ACS and / or ACS complications for the subject.

[0014] In certain embodiments, the processing system further comprises a display, the display being operatively associated with the non-transitory computer memory and configured to display the subject's risk of ACS and / or ACS complications. In further embodiments, the subject's estimated risk of ACS and / or ACS complications is reported as high risk (e.g., likely to have ACS and / or ACS complications), moderate risk (e.g., it is unclear whether the subject has ACS or ACS complications, so more testing may be required), or low risk (e.g., the subject is unlikely to have ACS or ACS complications, and therefore no further testing or treatment is necessary). In certain embodiments, the subject's estimated risk of ACS and / or ACS complications is a probability of risk for that individual subject.

[0015] In certain embodiments, the method further comprises d) performing at least one of the following actions i)-ix) (or diagnosing that the subject is in need of one of the following actions i)-ix)): i) performing coronary catheterization or stenting on the subject based on a high estimated risk of ACS; ii) treating the subject with cardiovascular disease (CVD) therapy based on a high estimated risk of ACS; iii) prescribing CVD therapy to the subject based on a high estimated risk of ACS; iv) performing at least one additional diagnostic test on the subject based on a moderate estimated risk of ACS or a moderate estimated risk of ACS complications; v) admitting and / or directing the subject to admission to a hospital or other treatment facility based on a high estimated risk of ACS and / or an ACS complication; vi) testing a sample from the subject with one or more non-troponin I CVD risk assays based on a moderate estimated risk of ACS and / or an ACS complication; vii) discharging the subject from a treatment facility based on a low estimated risk of ACS and / or an ACS complication; viii) performing a stress test on the subject based on a moderate estimated risk of ACS; and viii) determining the subject's likely 30-day post-discharge major adverse clinical event (MACE) risk.

[0016] In a further embodiment, the method further comprises d) performing at least one of the following actions i) to iv): i) communicating the subject's estimated risk of ACS and / or ACS complications to a user, ii) displaying the subject's estimated risk of ACS and / or ACS complications, iii) generating a report providing the estimated risk of ACS and / or ACS complications, iv) preparing and / or transmitting the report providing the estimated risk of ACS and / or ACS complications.

[0017] In certain embodiments, obtaining the target value comprises receiving the target value from a laboratory, a subject, an analytical testing system, and / or a portable or point-of-care testing device. In other embodiments, the processing system further comprises an analytical testing system and / or a portable or point-of-care testing device. In other embodiments, the processing system is middleware in a larger computer system. In additional embodiments, obtaining the target value comprises receiving the target value electronically. In further embodiments, obtaining the target value comprises testing the initial sample, the first subsequent sample, and / or the second subsequent sample with a cTnI and / or cTnT detection assay. In other embodiments, the cTnI and / or cTnT detection assay comprises a single molecule detection assay or a bead-based immunoassay. In certain embodiments, the ACS is selected from the group consisting of ST-segment elevation myocardial infarction (STEMI), non-ST-segment elevation myocardial infarction (NSTEMI), unstable angina, type I myocardial infarction, type II myocardial infarction, chest pain, and chest pain presenting to medical care within three hours (e.g., less than three hours, less than two hours, or less than one hour) (known as early presentation patients). In certain embodiments, the ACS complication is selected from the group consisting of heart failure, metastatic tumor, kidney disease (e.g., renal failure), and diabetes.

[0018] In certain embodiments, the method includes manually or automatically entering subject values ​​into the processing system. In additional embodiments, the subject is a human (e.g., a 25...35...45...55...65...75...85...or 95 year old male or female). In certain embodiments, the subject is a human with chest pain. In other embodiments, the subject sex and / or subject age includes subject sex. In certain embodiments, at least one of the following includes subject age: In additional embodiments, at least one of the following includes both subject age and subject sex:

[0019] In some embodiments, the initial sample from the subject comprises a blood sample, a serum sample, or a plasma sample. In additional embodiments, the initial sample is obtained from the subject in an emergency room or an urgent care clinic. In further embodiments, the first subsequent sample and / or the second subsequent sample comprises a blood sample, a serum sample, or a plasma sample. In other embodiments, the first subsequent sample and / or the second subsequent sample is obtained within 1 to 9 hours (e.g., within 1...3...5...7...or 9 hours) of the initial sample. In additional embodiments, the first and / or second subsequent cTnI and / or cTnT concentrations comprise both the first subsequent cTnI and / or cTnT concentration and the second subsequent cTnI and / or cTnT concentration. In certain embodiments, the subject values ​​further comprise subject measurements selected from the group consisting of ECG measurements, hematological analysis of the subject's blood, medical history, physical exam results, and current medications.

[0020] In some embodiments, the processing system further comprises a graphical user interface, and the method further comprises entering the subject value via the graphical user interface. In other embodiments, the graphical user interface is part of a device selected from a desktop computer, a laptop computer, a tablet computer, a smartphone, and a point-of-care analytical device. In certain embodiments, the processing system further comprises a sample analyzer. In some embodiments, at least a portion of the computer memory is internal to the sample analyzer. In certain embodiments, the processing system further comprises a laboratory interface system (LIM). In other embodiments, at least a portion of the computer memory is part of the LIM. In some embodiments, the processing system further comprises a processing device selected from the group consisting of a desktop computer, a laptop computer, a tablet computer, a smartphone, and a point-of-care analytical device. In certain embodiments, at least a portion of the computer memory is internal to the processing device. In further embodiments, the processing system further comprises a display component configured to display the subject's estimated risk of ACS and / or ACS complications. In other embodiments, the display component is selected from a computer monitor, a tablet computer screen, a smartphone screen, and a point-of-care analytical device screen.

[0021] In additional embodiments, the rate of change algorithm subtracts the concentration of cTnI and / or cTnT from a first subject sample obtained at a first time from the concentration of cTnI and / or cTnT from a second subject sample obtained at a second time to generate a difference in cTnI and / or cTnT, and then divides the difference in cTnI and / or cTnT by the time difference calculated by subtracting the first time from the second time. In certain embodiments, the time difference is measured in minutes or seconds. In some embodiments, the first subject sample is the initial sample or first subsequent sample, and the second subject sample is the first subsequent sample or second subsequent sample.

[0022] In certain embodiments, the at least M decision trees include at least 5 (e.g., 5...10...35...100...250...500...800...1000 or more) decision trees. In some embodiments, the at least M decision trees include at least 800 decision trees. In further embodiments, the predetermined splitting variables and / or predetermined terminal node values ​​are empirically derived from analysis of population data. In other embodiments, the analysis of the population data includes employing a boosted decision tree model. In some embodiments, the at least M decision trees as a group employ at least three splitting variables. In other embodiments, the at least M decision trees as a group employ at least four, or at least five, or at least six splitting variables.

[0023] In additional embodiments, the threshold rate of change value is the concentration change of cTnI and / or cTnT per minute. In other embodiments, the subject age value is either a set value based on the subject's age or an age range. In certain embodiments, the set value is determined based on ranges such as 0-29 years, 30-39 years, 40-49 years, 50-59 years, 60-69 years, 70-79 years, and 80 years and older. In some embodiments, the gender value is a number (e.g., 1.0) for males and a smaller number (e.g., 0.1 or 0) for females.

[0024] In particular embodiments, the composite value is either an unweighted composite value or a weighted composite value. In additional embodiments, the weighted value is the sum of all of the terminal node values ​​multiplied by a weight value. In further embodiments, the composite value from the M terminal nodes is a weighted composite value represented by the formula:

[0025]

number

[0026]

number

[0027]

number

[0028]

number

[0029] In certain embodiments, the non-transitory computer memory further comprises a look-up table of subject indices (e.g., as shown in Table 1), and processing the composite value to determine the subject's estimated risk of ACS and / or ACS complications comprises: i) applying the composite value to the following formula to find a total score (SS):

[0030]

number

[0031]

number

[0032] [Figure 1A] 1 is a diagram illustrating an example of a decision tree showing three splitting variables (cTnI rate, initial cTnI outcome, and age, all shown as rectangles) and four terminal nodes based on which splitting variable range (shown as circles) a particular patient falls into. [Figure 1B] This shows a series of decision trees (1-200). All terminal node values ​​are added together and then multiplied by a weighting value (0.1). [Figure 2] FIG. 1 shows in graphical form three of the 815 decision trees from Example 2, corresponding to three of the 815 trees shown in database form in Appendix A. [Figure 3] 1 is an exemplary flow diagram illustrating how information may be input and processed in the systems and methods described herein. For example, multiple blood draws may be performed from a subject with chest pain using the analyzer shown in the figure to determine initial and subsequent cTnI concentrations. Information from the analyzer, along with the patient's age and gender, is then entered into a laboratory information system (LIS). Such information is then fed into a tree-based analysis system to generate an estimated risk of acute coronary syndrome. This risk, with appropriate annotations, may be provided as a report or fed back to the LIS for use by the treating physician and / or patient. [Figure 4] 1 shows a hypothetical exemplary patient report. This report indicates that testing this patient's parameters (e.g., initial troponin value, troponin ratio, age, and gender) using the algorithm described herein would produce an index value of 2.59, a sensitivity of 98.8%, and an NPV (negative predictive value) of 99.73%. Given these values, this patient has a low risk of ACS. Therefore, this patient may be recommended (e.g., by the treating physician or clinic or hospital staff) not undergo further CVD testing and may be advised that no CVD treatment is necessary. [Figure 5] FIG. 1 shows two decision trees from Example 3. [Figure 6] Figure 1 shows calibration of the risk estimation index using the observed proportion of patients with type 1 myocardial infarction in the analysis cohort (a) and the validation cohort (b). Each point represents the observed proportion of patients with type 1 myocardial infarction in a group of 100 patients with a given risk estimation index value. The dashed line represents the optimal calibration. [Figure 7] Figure 1 shows the performance of the risk estimation algorithm across its entire range (0-100) in the validation cohort (gray) and analysis cohort (gold). The black lines are point estimates at each index value threshold, and the shaded areas are the 95% confidence intervals. [Figure 8] Figure 1 shows a design for clinical application incorporating ECG for low-risk stratification. Note that only 10,967 subjects had ECG classification for this analysis. [Figure 9] FIG. 10 illustrates the behavior of subgroups classifying risk into a low risk stratum with a risk estimation index less than 1.1. [Figure 10] FIG. 1 illustrates the behavior of subgroups that classify risk into a high risk stratum with a risk estimation index of 57.1 or greater. [Figure 11] Forest plot of sensitivity and NPV of a threshold of 1.1 index value for each cohort and subgroup (from Example 4) for time between samples and time from symptoms to first blood sample. Note that BACC cohort data is not included here because it was only available as a dichotomous variable of within 3 hours and greater than 3 hours. [Figure 12] Forest plot of PPV and specificity of a threshold of 57.1 index value for each cohort and subgroup (from Example 4) of time between samples and time from symptoms to first blood sample. Note that BACC cohort data is not included here because it was only available as a dichotomous variable of within 3 hours and greater than 3 hours. [Figure 13]1 is a flow diagram for estimating a patient's risk of type I myocardial infarction using one of the rapid techniques discussed in Example 4. [Figure 14] 1 is a flow diagram for estimating a patient's risk of type I myocardial infarction using one of the rapid techniques discussed in Example 4. DETAILED DESCRIPTION OF THE INVENTION

[0033] definition The term "acute coronary syndrome" as used herein refers to a group of conditions in which a portion of the heart muscle is unable to function properly or dies due to reduced blood flow in the coronary arteries. The most common symptom is chest pain, often radiating to the left arm or angle of the jaw, accompanied by tenderness-like features, nausea, and sweating. Acute coronary syndromes usually result from one of three problems: ST-segment elevation myocardial infarction (STEMI, 30%), non-ST-segment elevation myocardial infarction (NSTEMI, 25%), or unstable angina (38%) (Torres and Moayedi, 2007, Clin. Geriatr. Med. 23(2):307-25, vi, incorporated herein by reference in its entirety). These types are termed non-ST-segment elevation myocardial infarction and ST-segment elevation myocardial infarction, depending on the electrocardiogram (ECG / EKG). There may be some variation in which form of myocardial infarction (MI) is classified under acute coronary syndrome. ACS should be distinguished from stable angina, which occurs during activity and resolves with rest. In contrast to stable angina, unstable angina often occurs suddenly at rest or with minimal or lesser activity than the person's previous angina (creptate angina). New-onset angina is also considered unstable angina because it indicates a new problem in the coronary arteries. ACS is usually associated with coronary thrombosis but may also be associated with cocaine use. Cardiac chest pain can also be precipitated by anemia, bradycardia (too slow heart rate), or tachycardia (too fast heart rate). The presenting symptom of reduced blood flow to the heart is chest pain, experienced as chest tightness and radiating to the left arm and left angle of the jaw. It may be associated with diaphoresis (sweating), nausea, vomiting, and shortness of breath. Sensations are often "atypical," with varying pain experiences and sometimes no pain at all (more likely in women and those with diabetes). Some people report palpitations, anxiety, or a sense of impending death (demonstration), and an acute feeling of nausea. Patients with chest pain very frequently present to hospital emergency rooms. However, chest pain can have many causes, including stomach discomfort (e.g., indigestion), pulmonary distress, pulmonary embolism, dyspnea, musculoskeletal pain (strained muscles, bruises), dyspepsia, pneumothorax, cardiac non-coronary conditions, and acute ischemic coronary syndrome (ACS).As mentioned above, ACS is usually one of three coronary artery diseases: ST-segment elevation myocardial infarction (30%), non-ST-segment elevation myocardial infarction (25%), or unstable angina (38%). These types are named non-ST-segment elevation myocardial infarction (NSTEMI) and ST-segment elevation myocardial infarction (STEMI) depending on the electrocardiogram (ECG / EKG) appearance. ACS is usually associated with coronary artery thrombosis. Physicians must determine whether a patient has a life-threatening ischemic ACS. In such ischemic cardiac events, prompt treatment by opening the blocked coronary artery is essential to prevent further loss of myocardial tissue.

[0034] As used herein, "suspected of having an acute coronary syndrome" means that a subject has at least one of the aforementioned symptoms of an acute coronary syndrome (e.g., chest pain or tightness radiating to the left arm and left angle of the jaw).

[0035] The term "diagnosis" as used herein can encompass determining the nature of disease in a subject, as well as determining the severity of a disease or disease episode and the likely outcome and / or chances of recovery (prognosis). "Diagnosis" can also encompass diagnosis in the context of rational therapy, a situation in which diagnosis guides therapy, including initial selection of therapy, adjustment of therapy (e.g., adjusting dosage and / or dosing regimen or recommending lifestyle changes), etc.

[0036] The terms "individual," "host," "subject," and "patient" are used interchangeably herein and generally refer to pregnant mammals, including, but not limited to, primates, including monkeys and humans, equines (e.g., horses), canines (e.g., dogs), felines, various domestic livestock (e.g., ungulates such as swine, pigs, goats, and sheep), as well as domestic pets and zoo animals. In some embodiments, the subject is specifically a human subject.

[0037] Before the present invention is further described, it is to be understood that this invention is not limited to particular embodiments, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting.

[0038] Where a range of values ​​is given, unless the context clearly dictates otherwise, it is understood that intervening values, up to one-tenth the magnitude of the lower limit, between the upper and lower limits of that range, and any other stated or intervening value in that stated range, are encompassed within the scope of the invention. Where there is any specifically excluded limit in the stated range, the upper and lower limits of any smaller range may independently be included in the smaller range and are also encompassed within the scope of the invention. Where a stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, the preferred methods and materials are now described. All published materials mentioned herein are incorporated by reference to disclose and describe the methods and / or materials in connection with which the published materials are cited.

[0040] It should be noted that, as used in this specification and the appended claims, the singular forms "a," "and," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, a reference to a "sample" includes a plurality of such samples, a reference to a particular protein includes a reference to one or more specific proteins and equivalents thereof known to those skilled in the art, and so forth.

[0041] Detailed Description of the Invention The present invention provides a decision-tree-based system and method for estimating the risk of acute coronary syndrome (ACS) and / or ACS complications in a subject suspected of having ACS and / or ACS complications. Specifically, the system and method employ an additive decision-tree-based algorithm to process the subject's initial cardiac troponin I or T (cTnI or cTnT) concentration, the subject's rate of change of cTnI and / or cTnT, and the subject's age and / or sex to generate an estimated risk of ACS and / or ACS complications. Such severity classification allows for determining, for example, whether the patient requires emergency treatment.

[0042] In certain embodiments, processing systems and methods are provided herein that allow for input of first and second troponin results, age, and gender variables. These variable inputs are evaluated by decision-tree-based statistical calculations to provide an estimated risk of ACS and / or ACS complications so that subjects can be stratified into appropriate risk categories. In certain embodiments, the systems and methods herein address the timing variable between sample collections by determining the rate of change of troponin based on the exact time or near-exact time (e.g., to the minute) between the first and second collections of samples from the subject. In certain embodiments, the systems and methods herein address the age variable by determining the impact of the age decile into which the patient is classified. In some embodiments, the systems and methods herein address gender differences by classifying patients into male and female profiles.

[0043] The systems and methods herein take patient variables into account to determine a patient's appropriate risk category. For example, in work conducted during development of embodiments of the present invention, the systems and methods herein classified nearly 87 percent of patients suspected of having an ACS into a low-risk category, allowing them to be safely excluded from patients with a myocardial infarction (MI), and appropriately classified 8% of the suspected ACS population into a high-risk category of having an MI, thereby triaging patients to a cardiac catheterization lab. This helps to treat the vast majority of patients presenting in an emergency setting, with only 4-6% of patients, for example, requiring further workup or observation in an emergency setting. Such (e.g., emergency) severity classification of patients suspected of having an ACS helps reduce false positives and false negatives, reducing adverse outcomes as well as healthcare costs.

[0044] I. Decision Tree Algorithm In certain embodiments, the estimated risk of ACS (e.g., MI) or ACS complications is determined using an additive tree model. One example of such a model is the boosted decision tree model (also known as the additive logistic model), which is a tree-based additive regression model. It can be mathematically expressed in the context of diagnosing hsTroponin I or T as follows:

[0045]

number

[0046] In Fig. 1A, the rectangles represent the values ​​of the split variables and the corresponding split positions, and the circles represent the tree T i For example, for a subject with TnT change rate = 0.6, initial TnT result = 500, age = 40, and gender = male, the predicted value by the above tree is 0.01269 (the circle at the bottom left). i is a regularization parameter to balance the prediction error and overfitting of the model. In practice, a boosted decision tree can be thought of as a weighted sum of several different decision trees. In certain embodiments, these trees are grown sequentially rather than independently. For example, M=200 and α i = 0.01, the right hand side of equation (1) (F(X)) can be illustrated by the diagram in Figure 1B.

[0047] In Figure 1B, this particular boosted decision tree model can be understood as a weighted sum of 200 different individual decision trees, each with a weight of 0.01. The arrows in Figure 1B indicate that these individual trees grow sequentially, i.e., later trees grow based on earlier trees. Note that all parameters of the model are determined by the algorithms and rules.

[0048] For a particular subject with information on the TnI rate of change, initial TnI result, age, and gender, a predicted value can be obtained from each tree (e.g., from a circle with a number nearby). F(X) is then actually obtained by (0.025-0.007+...+0.01) x 0.01. If we assume this number is 0.2, then p1 = 0.12 from equation (1). Note that p1 is a probability value between 0 and 1. The subject's index is 12, obtained by multiplying p1 by 100.

[0049] II. Biological Samples To determine the concentration of cardiac troponin I and / or troponin T, a biological sample from a subject is tested. Biological samples include, but are not limited to, bodily fluids such as blood-related samples (e.g., whole blood, serum, plasma, and other blood-derived samples), urine, cerebrospinal fluid, bronchoalveolar lavage, etc. Another example of a biological sample is a tissue sample. Biological samples can be fresh or stored (e.g., blood or blood fractions stored in a blood bank). Biological samples can be bodily fluids collected specifically for the analysis of this invention, or bodily fluids collected for another purpose and that can be subsampled for the analysis of this invention. In certain embodiments, the biological sample is whole blood. Stored blood can be collected from a subject using standard clinical diagnostic methods. In other embodiments, the biological sample is plasma. Plasma can be obtained from a whole blood sample by centrifugation of anticoagulated blood. Such processing results in a buffy coat of white blood cell components and a plasma supernatant. In certain embodiments, the biological sample is serum. Serum can be obtained by centrifuging a whole blood sample collected in a tube without anticoagulant. The blood is allowed to clot prior to centrifugation. The slightly yellow, reddish fluid obtained by centrifugation is serum. In another embodiment, the sample is urine. The sample may be pretreated by dilution into an appropriate buffer, heparinized, concentrated if necessary, or fractionated by any number of methods, including, but not limited to, ultracentrifugation, fractionation by high-performance liquid chromatography (FPLC), or precipitation of apolipoprotein B-containing proteins using dextran sulfate, or other methods. Any of several standard aqueous buffers at physiological pH, such as phosphate, Tris, etc., can be used.

[0050] III. Exemplary Detection Analysis The present invention is not limited by the type of assay used to detect cardiac troponin I (cTnI) or troponin T (cTnT). In certain embodiments, methods for detecting troponin I are described in U.S. Patent Application Publication 2012 / 0076803 and U.S. Pat. No. 8,535,895, both of which are incorporated herein by reference, particularly for assay design. In certain embodiments, methods for detecting troponin T employ the Elecsys® Troponin T High Sensitivity (TnT-hs) Assay (ROCHE) (see Li et al., Arch Cardiovasc Dis., March 2016, 109(3):163-70, which is incorporated herein by reference in its entirety, particularly for a description of high-sensitivity troponin T detection).

[0051] In certain embodiments, immunoassays are employed to detect cTnI and / or cTnT. Any suitable assay known in the art may be used, including commercially available assays for cTnI or cTnT. Examples of such assays include, but are not limited to, immunoassays such as sandwich immunoassays (e.g., radioisotope detection (radioimmunoassay (RIA)) and enzyme detection (monoclonal-polyclonal sandwich immunoassays, including enzyme immunoassays (EIA) or enzyme-linked immunosorbent assays (ELISA) (e.g., Quantikine ELISA assays, R&D Systems, Minneapolis, MN), competitive inhibition immunoassays (e.g., forward and reverse), fluorescence polarization immunoassays (FPIA), enzyme-amplified immunoassay technology (EMIT), bioluminescence resonance energy transfer (BRET), and homogeneous chemiluminescence assays). In certain embodiments, cTnI is detected using Singulex's ERENNA detection assay system or Abbott's hs TnI STAT ARCHITECT assay. [Example]

[0052] The following examples are intended for illustrative purposes only and are not intended to limit the scope of the claims.

[0053] [Example 1] Patient population testing This example describes a study of a population of 972 subjects, 86 of whom had a myocardial infarction (MI) and 886 were non-MI patients. This study allowed for the development of a database of 815 (M=815) trees, shown in Appendix A, as well as the index lookup table shown in Table 1 below.

[0054] [Table 1] TIFF0007762680000009.tif246168TIFF0007762680000010.tif29168

[0055] [Example 2] Determining a patient's risk of ACS This example describes an exemplary method for determining a patient's risk of ACS. A patient presents to the emergency room with chest pain. Relevant patient information is collected from the patient by questioning the patient and testing an initial blood sample and a second blood sample from the patient to determine the concentration of cardiac troponin I. The patient information is shown in Table 2 below.

[0056] [Table 2]

[0057] The TnI rate (per minute) was calculated using the difference between two hs TnI concentration values ​​divided by the corresponding difference between collection times (in minutes) from the subject's first two available time points. Age was then categorized into deciles (1–7): "<30 years" (1), "30–<40 years" (2), "40–<50 years" (3), "50–<60 years" (3), "60–<70 years" (4), "70–<80 years" (6), and "80 years or older" (7). This patient was 86 years old and in age decile 7. Gender was assigned based on female = 0 and male = 1. Therefore, this female patient was assigned a value of zero for gender.

[0058] The values ​​from Table 2 for this patient are then applied to a predetermined set of decision trees. In this example, the values ​​for this patient are applied to a database of 815 decision trees (trees 0-814) from Appendix A. Figure 2 shows three of the 815 trees from Appendix A in graphical form, including tree 0, tree 215, and tree 814. Tree 0, for example, shows a cTnI rate ≦0.17833 as the splitting variable. A rate below 0.17833 would split to the left, and a rate above 0.17833 would split to the right. This example patient has a rate of 0.025, and therefore splits to the left. Splitting to the left results in a splitting variable box for an initial cTnI result of ≦73.55. This patient had an initial concentration of 15.3 pg / ml, and therefore split to the left terminal node of −0.008939, which is the value for this tree.

[0059] Appendix A has the same decision tree, but in database form. The first column of Appendix A contains the tree number (0-814). The second column contains the node label (0-6). The third column contains the splitting variables (0 = TnI rate (in minutes), 1 = initial TnI result, 2 = sex, 3 = age group, -1 = terminal node). The fourth column contains the cutoff value for each splitting variable. If it is less than or equal to the cutoff value, go to the "left node," and if it is greater than the cutoff value, go to the "right node." The "left node" and "right node" columns contain the number of the next node to go to.

[0060] The database works as follows to form the following decision tree: Table 3 shows the first seven rows of Appendix A. These seven rows are labeled Tree 0 and collectively form one of the 815 decision trees in Appendix A. The decision tree works as follows using the patient information from Table 2 above: Start with the first column which indicates a splitting variable with a TnI rate ≦0.178333333. If the patient matches this, take the "Left Node" column which tells you to go to node "1". If the patient does not match this splitting variable, take the "Right Node" column which tells you to go to node "5". Since this patient is less than 0.17833333, the left node column which tells you to go to node 1 is used. In the next row, node 1 is found. The splitting variable for this next row is "Initial TnI Result", which has a value less than or equal to 73.55. Since this patient has an initial TnI value ≦73.55, then go to the "Left Node" which tells you to go to "2". So, if we look again at the "Node" column for "2," we find it in the next row. This time, the split variable is a "Terminal Node," meaning we've reached the terminal node of the tree, so we read the value in the "Prediction" column for this row as -0.008939. Therefore, this patient's outcome score for tree 0 is -0.008939, which is shaded in Appendix A. We repeat the same logic through the remaining 814 trees. This patient will receive a total outcome score of 815 from these 815 trees. All of these outcomes are shaded in Appendix A.

[0061] [Table 3]

[0062] An additive tree formula then uses the values ​​from the individual trees, for example: To find the index score for this patient, the following two formulas can be used:

[0063]

number

[0064]

number

[0065]

number

[0066]

number

[0067] [Table 4]

[0068] Looking at Table 1, this patient would be considered to be at moderate risk for ACS in the setting of an index value of 3.86.

[0069] [Example 3] Determining a patient's risk of ACS This example describes an exemplary method for determining a patient's risk of ACS. A patient presents to the emergency room with chest pain. Relevant patient information is collected from the patient by questioning the patient and testing an initial blood sample and a second blood sample from the patient to determine the concentration of cardiac troponin I. The patient information is shown in Table 5 below.

[0070] [Table 5]

[0071] The TnI rate (per minute) was calculated using the difference between two hs TnI concentration values ​​divided by the corresponding difference between collection times (in minutes) from the subject's first two available time points. Age was then categorized into deciles (1–7): "<30 years" (1), "30–<40 years" (2), "40–<50 years" (3), "50–<60 years" (4), "60–<70 years" (5), "70–<80 years" (6), and "80 years or older" (7). This patient was 86 years old and in age decile 7. Gender was assigned based on female = 0 and male = 1. Therefore, this female patient was assigned a value of zero for gender.

[0072] The values ​​from Table 5 for this patient are then applied to a predetermined series of two decision trees shown in Figure 5. Given this patient's values, this patient will have a total of two outcome scores from these two trees in Figure 5 (-0.008939 and -0.008380).

[0073] An additive tree formula then uses the values ​​from the individual trees, for example: To find the index score for this patient, the following two formulas can be used:

[0074]

number

[0075]

number

[0076]

number

[0077]

number

[0078] [Table 6] Looking at Table 1, this patient would be considered to have a moderate risk of ACS (using the two trees as above) in the setting of an index value of 13.42.

[0079] As more trees are added to the algorithm, the index scores converge towards an index value of 3.86 from the 815 tree algorithm described above in Example 2. For example, the 10 tree algorithm gives an index value of 12.67, while the 50 tree algorithm gives an index value of 9.60.

[0080] [Example 4] Severity classification of patients with suspected myocardial infarction This example describes an exemplary method for employing an algorithm to risk stratify patients with myocardial infarction. Because high-sensitivity cardiac troponin I concentrations vary with age, sex, and time, it was desirable to employ a risk estimation decision tool that incorporates these variables to improve the severity stratification and diagnosis of patients with suspected myocardial infarction (MI). Machine learning was used to apply a risk estimation algorithm to predict type 1 MI in an analysis cohort of 3,013 patients with suspected MI. The algorithm incorporated age, sex, and paired high-sensitivity cardiac troponin I concentrations.

[0081] The generated MI3 index value for each patient can be used for severity classification. Validation was performed in a cohort of 7,998 patients by calibration curve, area under the receiver operating characteristic curve (AUC), and performance at predefined thresholds. Optimal index thresholds for allocation to the low-risk group (negative predictive value ≥ 99.5% and sensitivity > 99.0%) and the high-risk group (positive predictive value ≥ 75% and specificity ≥ 90%) were derived.

[0082] MI occurred in 404 (13.4%) and 849 (10.6%) patients in the analysis and validation cohorts, respectively. The diagnostic performance of the risk estimation index in the validation cohort was similar to that in the analysis cohort, with excellent calibration and a similar AUC (0.963 [95% CI 0.956-0.971] cf 0.963 [0.957-0.968]). The optimal low-risk threshold (1.1) classified 51.6% of patients as low-risk, and the optimal high-risk threshold (57.1) classified 10.3% of patients as high-risk.

[0083] method research design This example provides a retrospective analysis of prospectively collected data from multiple centers to derive and validate a risk estimation algorithm to facilitate decision-making in patients with suspected myocardial infarction. The risk estimation algorithm incorporates age, sex, paired high-sensitivity cardiac troponin I concentrations, and the rate of change in cardiac troponin concentrations. These variables were chosen a priori because (a) they are non-subjective, (b) they can be automatically captured from electronic hospital records, (c) they are based on sequential cardiac troponin measurements as recommended by international guidelines, and (d) they are known to be associated with the diagnosis of type 1 myocardial infarction.

[0084] Risk Estimation Algorithm The risk estimation algorithm is constructed using the analysis cohort through a machine learning technique called boosting, and includes multiple decision trees that weight input variables to optimally distinguish between those with and without the event (Reference 20 in Example 4). The algorithm calculates a risk estimation index (on a scale of 0 to 100) that predicts the likelihood of being diagnosed with a type 1 myocardial infarction during the index hospital visit.

[0085] To determine the decision tree and weights for the final model (algorithm), a machine learning technique, boosting (also called an additive logistic model (see Reference 20 in Example 4)), was applied to the analysis cohort. Inputs were the initial hs-cTnI concentration, the rate of change of hs-cTnI (the difference between two consecutive hs-cTnI values ​​divided by the time difference in minutes), sex, age (<30 years (category 1), 30 to <40 years (category 2), 40 to <50 years (category 3), 50 to <60 years (category 4), 60 to <70 years (category 5), 70 to <80 years (category 6), and ≥80 years (category 7)), and type 1 myocardial infarction status.

[0086] Mathematically, the model can be expressed as:

[0087]

number

[0088] Within the analysis cohort, thresholds of risk estimation indices were identified to risk stratify patients as those at low risk of myocardial infarction or those at high risk of myocardial infarction. Table 9 below shows the index threshold table, where the thresholds are listed as integers. It is noted that this table can be divided into 100 cases by listing the thresholds in increments of 0.01.

[0089] [Table 7] TIFF0007762680000026.tif245168TIFF0007762680000027.tif247168TIFF0007762680000028.tif239169

[0090] The risk estimation algorithm was validated in the second cohort by calibration curves, areas under the receiver operating characteristic curves, and performance of derived risk estimation index thresholds, followed by recalibration of thresholds using all both cohorts for optimal severity classification and to improve generalizability.

[0091] Participants and cohorts Patients with symptoms suggestive of myocardial infarction who had serial high-sensitivity cardiac troponin I measurements obtained at presentation and later in the emergency department were included. Patients with ST-segment elevation myocardial infarction (STEMI) were excluded. Cohorts were identified for inclusion if they were expected to contain serial high-sensitivity cardiac troponin I concentrations. Final diagnoses were adjudicated according to the universal definition of myocardial infarction (see References 4 and 23 in Example 4). Ethical approval allowed for patient-level data sharing. Diagnosis was made using evidence of elevated and / or depressed cardiac troponin concentrations with at least one value above the 99th percentile of the healthy population, along with at least one of the following: ischemic symptoms; new or presumed new significant STT wave changes or new left bundle branch block; pathologic Q wave growth; imaging evidence of new loss of viable myocardium or new regional wall motion abnormalities; and / or angiographic identification or autopsy of intracoronary thrombus (see References 4 and 23 in Example 4). The algorithm was derived in patients recruited in Scotland and Germany (References 9 and 24 in Example 4). The validation cohort was drawn from seven cohorts recruited in Australia, Germany, New Zealand, Spain, Switzerland, and the United States (References 25-29 in Example 4).

[0092] Sampling and Laboratory Analysis Cardiac troponin concentrations were measured at each study site by the Abbot ARCHITECT high-sensitivity troponin I assay (Abbot diagnostics, Chicago, IL). The manufacturer reported that the limit of detection (LoD) and 99th percentile upper reference limit (URL) for the high-sensitivity assay were 1.9 ng / L and 26.2 ng / L, respectively. The gender-specific 99th percentile URLs were 16 ng / L for women and 34 ng / L for men.

[0093] Definition and Adjudication of Outcomes The primary outcome was a confirmed diagnosis of type 1 myocardial infarction during the first hospitalization using a universal definition. Diagnosis was made using evidence of elevated and / or depressed cardiac troponin concentrations with at least one value above the 99th percentile of the healthy population, along with at least one of the following: ischemic symptoms, new or presumed new significant STT wave changes or new left bundle branch block, pathologic Q wave growth, imaging evidence of new loss of viable myocardium or new regional wall motion abnormalities, and / or angiographic identification or autopsy of intracoronary thrombus (see References 4, 23 in Example 4).

[0094] statistical analysis Boosting was applied to the analysis cohort to determine the final risk estimation decision tree and weights, which, once determined, were locked in place and programmed into an Excel spreadsheet that was used to return index values ​​for risk estimation in the analysis and validation cohorts.

[0095] It was prespecified that thresholds for risk estimation index values ​​would be derived and validated from the analysis cohort that would yield a sensitivity of 99.0% or greater, a negative predictive value (NPV) of 99.5% or greater, a specificity of 90% or greater, and a positive predictive value (PPV) of 75% or greater for the diagnosis of type 1 myocardial infarction. The sensitivity targets were based on a survey of what emergency department physicians consider acceptable risk (see Reference 30 in Example 4), and the NPV targets were the most common in the literature. The specificity and PPV targets were selected by consensus of the project steering committee as clinically reasonable for high-risk stratification. The thresholds for risk estimation index values ​​corresponding to these four diagnostic metrics were determined using 95% confidence intervals determined by bootstrapping from the analysis cohort (1,000 samples).

[0096] The performance of the algorithm was evaluated in the analysis and validation cohorts using the calibration curve and area under the receiver operating characteristic curve (AUC) for discrimination. Index thresholds were derived and validated using predefined statistical metrics. For optimal performance in risk stratifying patients into low-risk (≥99.5% negative predictive value and ≥99.0% sensitivity) and high-risk (≥75% positive predictive value and ≥90% specificity), the two index value thresholds were recalibrated using both the analysis and validation cohorts. Validation, performed independently from the algorithm derivation, used R (Version 3.2.4: The R Foundation for Statistical Computing (Reference 31 in Example 4)).

[0097] Sensitivity, subgroup and post-hoc analyses Additional preplanned subgroup analyses included comparisons by sex, age (≤65 years, >65 years), comorbidities (history of coronary artery disease, diabetes mellitus, hypertension, current smoking), time from symptom onset to first sample collection, and time between consecutive cardiac troponin tests in patients without new evidence of myocardial ischemia on electrocardiogram. The performance of the algorithm was also evaluated for type 1 myocardial infarction within 30 days.

[0098] result Of the 3,013 patients included in the analysis cohort, 404 (13.4%) were diagnosed with type 1 myocardial infarction. This cohort was predominantly male (63%), with a mean age of 62.4 years (Table 7).

[0099] [Table 8]

[0100] Of the 7,998 patients included in the validation cohort, 849 (10.6%) were diagnosed with type 1 myocardial infarction. Patients in the validation cohort were younger and less likely to have had previous CAD, but were more likely to smoke, have diabetes mellitus, hyperlipidemia, or have a family history of CAD than the analysis cohort. A larger proportion of patients in the validation cohort had blood drawn within 3 hours of symptom onset (38.5% vs. 33.0%, p<0.0001). The time between sample collections (median 2.2 [IQR 2.0-2.6] hours) was longer than that in the analysis cohort (median 1.2 [IQR 1.0-2.5] hours) (P<0.0001).

[0101] Correlation and Discrimination The risk estimation index values ​​were well calibrated (Figure 6), and the diagnostic metrics performed similarly across the entire risk estimation index range, although the PPV and NPV of the validation cohort were slightly lower compared to those of the analysis cohort (Figure 7). The index discriminated between those with and without type 1 myocardial infarction in both the analysis cohort (AUC 0.963 [95% CI 0.956-0.971]) and the validation cohort (AUC 0.963 [95% CI 0.957-0.968]).

[0102] Diagnostic threshold performance The risk estimation index thresholds from the analysis cohort that met the prespecified diagnostic performance criteria were 1.6 (sensitivity ≥ 99.0%), 3.1 (NPV ≥ 99.5%), 17.2 (specificity ≥ 90.0%), and 49.7 (PPV ≥ 75%) (Table 8).

[0103] [Table 9]

[0104] At the derived threshold, the NPV (99.4% [99.2% to 99.6%]) and specificity (91.7% [91.1% to 92.3%]) were similar to those used to derive the threshold (99.5% NPV, 90.0% specificity). The sensitivity (97.8% [96.7% to 98.7%]) and PPV (71.8% [68.9% to 75.0%]) were slightly lower than those used to derive the threshold (99.0% sensitivity and 75.0% PPV).

[0105] Recalibrated optimal diagnostic thresholds The optimal MI3 risk estimation index low-risk threshold was 1.1 (sensitivity 99.3% [98.8%-99.7%], NPV 99.8% [99.7%-99.9%]), and the optimal MI3 risk estimation index high-risk threshold was 57.1 (PPV 74.9% [72.5%-77.4%], specificity 97.1% [96.7%-97.4%]). Of 11,011 patients, 5682 (51.6%) were classified as low risk and 1134 (10.3%) as high risk.

[0106] Risk estimation threshold combined with ECG The combination of a risk estimation index value of less than 1.1 and the absence of myocardial ischemia on electrocardiogram had a sensitivity of 99.4% (99.0%–99.8%) with an NPV of 99.9% (99.8%–99.9%). 49.0% of patients were identified as low risk (Figure 8).

[0107] Subgroup analysis The risk estimation index threshold of 1.1 performed similarly across all subgroups, including patients with initial symptoms for less than 3 hours (Figure 9). The risk estimation index of 57.1 also performed similarly across most groups, with the exception of gender and time from symptom onset, where the PPV for women was lower compared to that for men, and the group with symptoms for less than 3 hours reflected a lower prevalence compared to the group with symptoms for more than 3 hours (Figure 10).

[0108] Individual cohort performance At an MI3 index threshold of 1.1, 43.7% to 82.3% of patients were classified as low risk across individual cohorts. Sensitivity varied from 95.5% to 100%, and NPV varied from 99.6% to 100% (Figure 11). At an index threshold of 57.1, PPV varied from 41.3% in the lowest prevalence cohort (UTROPIA) to 96.1% in the highest prevalence cohort (ADAPT-CH) (Figures 11 and 12).

[0109] Further sensitivity analysis A risk estimation index threshold of 1.1 performed well, with high-point estimates of sensitivity and NPV, independent of time between samples and further stratification from symptom onset to first blood draw (Figure 11). A risk estimation index threshold of 57.1 showed increased PPV and specificity, with possible performance degradation with increasing time from symptom onset to first blood draw and increasing time between samples (Figure 12).

[0110] Type 1 myocardial infarction within 30 days Including all patients with type 1 myocardial infarction within 30 days slightly reduced the sensitivity of the risk estimation index, with similar PPVs of 99.1% (98.6%–99.6%) in 1.1 and 74.9% (72.3%–77.2%) in 57.1.

[0111] Rapid Method Baseline MI3 (calculated only on the initial troponin value) provides a rapid dismissal without the need to wait for serial samples. This excludes the majority of low-risk patients. Baseline MI3 also stratifies high-risk individuals without the need to wait for serial samples. Thus, the lowest-risk and highest-risk patients are immediately triaged. Baseline MI3 utilizes only the initial troponin result, age, and sex, rather than calculating a rate of rise. This approach is illustrated in Figure 13.

[0112] Another rapid approach, shown in Figure 14, combines the calculation of the MI3 index with an hsTnI concentration of less than 2 nanograms / L, increasing the combined safety of excluding very low-risk patients. Combining a safe MI3 index with a safe troponin value ensures that only very low-risk patients are omitted from further testing and follow-up for ischemic chest pain. The calculation of the baseline MI3 index utilizes only the initial troponin result, age, and sex, not a rate-of-rise calculation.

[0113] [Example 5] Using troponin T to stratify patients with suspected myocardial infarction This example describes an exemplary method employing the troponin T (TnT) test and an algorithm for risk stratifying patients with myocardial infarction. There were 956 patients in the BACC cohort with at least two troponin T results who were included in the analysis. Of these, 341 (35.67%) were women and 615 (64.33%) were men. Of the 956, 179 were adjudicated MI patients. The MI3 algorithm was then applied to the first two valid troponin T values ​​of these 956 patients to generate an index value for each patient. Sensitivity, specificity, NPV, and PPV were then calculated for each index value, ranging from 0 to 100, in increments of 1.00. The 987 trees used to calculate the index values ​​are shown in Appendix B. Similar to TnI above, TnT index values ​​were calculated using the following formula:

[0114]

number

[0115]

number

[0116] [Table 10] TIFF0007762680000034.tif230170TIFF0007762680000035.tif226170TIFF0007762680000036.tif90170

[0117] The results of this example for TnT (including TnI using the same process), listed from lowest to highest MI3 index value, are shown in Appendix C. Using the index lookup table (Table 10 above), patients with an index value for TnT less than 1.1 are considered to be at low risk of myocardial infarction. Patients with an index value between 1.1 and 57.0 are considered to be at moderate risk of myocardial infarction. Patients with an index value of 57.1 or greater are at high risk of myocardial infarction. Appendix C provides the TnI index values ​​for these patients as well as a comparison of the index values ​​calculated using TnI to the index values ​​calculated using TnT.

[0118] [Table 11] TIFF0007762680000038.tif252169TIFF0007762680000039.tif254164TIFF0007762680000040.tif64163

[0119] While only a few exemplary embodiments have been described in detail, those skilled in the art will readily recognize that many modifications in the exemplary embodiments are possible without substantially departing from the novel teachings and advantages of this disclosure. Accordingly, all such modifications and alternatives are intended to be included within the scope of the present invention as defined in the following claims. Those skilled in the art will recognize that such modifications and equivalent structures or methods do not depart from the spirit and scope of this disclosure, and that various changes, substitutions, and alterations can be made herein without departing from the spirit and scope of this disclosure.

[0120] [Table 12] TIFF0007762680000042.tif242169TIFF0007762680000043.tif244168TIFF0007762680000044.tif242169TIFF0007762680000045.tif243168TIFF0007762680000046.tif245169TIFF0007762680000047.tif244168TIFF0007762680000048.tif241169TIFF0007762680000049.tif242168TIFF0007762680000050.tif241168TIFF0007762680000051.tif242168TIFF0007762680000052.tif243169TIFF0007762680000053.tif199169

[0121]

Table 13

[0122]

Table 14

Claims

1. 1. A method for reporting an estimated risk of acute coronary syndrome (ACS) in a subject suspected of having ACS, comprising: a) obtaining a subject value for the subject, wherein the subject is suspected of having ACS, and the subject value is: i) a subject's gender value and a subject's age value, respectively; ii) the subject's initial cardiac troponin I (cTnI) concentration from the subject's initial sample; and iii) a subsequent cTnI concentration from a corresponding subsequent sample taken from the subject within one hour of the initial sample. a step comprising: b) processing the subject values ​​with a processing system so that a predicted risk of ACS for the subject is determined; and c) reporting the subject's estimated risk of ACS determined by the processing system; the processing system comprising: i) a computer processor; and ii) non-transitory computer memory containing one or more computer programs and databases; wherein the one or more computer programs include a rate of change algorithm and an additive tree algorithm; the database includes at least 800 different decision trees, each of which includes at least two different predetermined splitting variables and at least three predetermined terminal node values; the at least two different predetermined splitting variables are selected from the group consisting of a threshold value for a cTnI rate of change value, a threshold value for an initial cTnI concentration value, and at least one of a threshold value for a gender value and an age value; The one or more computer programs, in conjunction with the computer processor, A) applying the rate of change algorithm to determine a rate of change value for cTnI in the subject from the subject's initial cTnI concentration and subsequent cTnI concentrations; B) determining a terminal node value for each of the at least 800 different decision trees by applying at least one of the subject's cTnI rate of change value, the subject's initial cTnI concentration, and the subject's gender value and age value to the database; and C) applying the additive dendrogram algorithm to: I) determine a composite value from the terminal node values ​​of each of the at least 800 different decision trees; and II) process the composite value to determine an estimated risk of ACS for the subject. A method configured to:

2. 2. The method of claim 1, wherein the risk of ACS is the probability of risk for that individual subject.

3. d) The following actions: i) performing at least one additional diagnostic test on the subject based on the estimated moderate risk of ACS; ii) testing a sample from the subject with one or more non-troponin I CVD risk assays based on the predicted moderate risk of ACS; and iii) performing a stress test on the subject based on the estimated moderate risk of ACS. The method of claim 2 , further comprising performing at least one of:

4. d) The following actions: i) communicating to a user the subject's estimated risk of ACS; ii) displaying the subject's estimated risk of ACS; iii) generating a report providing an estimated risk of said ACS; and iv) preparing and / or transmitting a report providing an estimated risk of the ACS.

5. 10. The method of claim 1, wherein the step of obtaining a subject value comprises receiving the subject value from a laboratory, the subject, an analytical testing system, and / or a portable or point-of-care testing device.

6. The method of claim 5 , wherein the processing system further comprises the analytical testing system and / or the portable or point-of-care testing device.

7. The method of claim 1 , wherein the step of obtaining a target value comprises receiving the target value electronically.

8. 10. The method of claim 1, wherein the step of obtaining a target value comprises testing the initial sample and / or the subsequent sample with an assay for detecting cTnI.

9. 9. The method of claim 8, wherein the cTnI detection assay comprises a single molecule detection assay or a bead-based immunoassay.

10. 2. The method of claim 1, wherein the ACS is selected from the group consisting of ST-elevation myocardial infarction (STEMI), non-ST-elevation myocardial infarction (NSTEMI), unstable angina, type II myocardial infarction, chest pain, and chest pain resulting in medical attention within three hours.

11. The method of claim 1 , further comprising manually or automatically inputting the target values ​​into the processing system.

12. (a) the subject is a human; (b) the subject is a human suffering from chest pain; (c) at least one subject value includes the gender of the subject; (d) at least one subject value comprises the age of the subject; or 10. The method of claim 1, wherein (e) at least one subject value comprises the subject's gender and subject's age.

13. (a) the initial sample from the subject comprises a blood sample, a serum sample, or a plasma sample; (b) the initial sample is obtained from the subject in an emergency room or urgent care clinic; (c) the subsequent sample comprises a blood sample, a serum sample, or a plasma sample; or 2. The method of claim 1, wherein (d) the subject values ​​further comprise subject measurements selected from the group consisting of medical history, physical examination results, and current medications.

14. The method of claim 1 , wherein the predetermined splitting variables and / or the predetermined terminal node values ​​are empirically derived from analysis of population data.

15. The method of claim 1 , wherein the at least 800 different decision trees, as a group, employ at least three or at least four of the splitting variables.

16. 2. The method of claim 1, wherein the threshold rate of change value is the concentration change of cTnI per minute.

17. The composite value from the terminal node values ​​of each of the at least 800 different decision trees is a weighted composite value expressed by the following formula: [Equation 1] In this formula, T ｉ represents the individual decision tree, X represents the target value, and β ｉ indicates the at least two splitting variables, and α ｉ represents the weight value, [Equation 2] The method of claim 1 , wherein {overscore (k)} represents the sum of all of the at least 800 different decision trees.

18. The processing of the composite value to determine an estimated risk of ACS for the subject comprises solving the equation: [Equation 3] 18. The method of claim 17, wherein p1 represents the estimated risk of ACS.

19. The composite value is expressed by the formula [Equation 4] The method of claim 1 , wherein:

20. wherein the non-transitory computer memory further comprises a look-up table of subject indices, and wherein the processing of the composite value to determine the subject's estimated risk of ACS comprises: i) applying the composite value to the following formula to find a total score (SS): [Equation 5] ii) applying the SS to find the final index (IDX) [Equation 6] and iii) applying the IDX to a lookup table of indicators for the subject to determine the subject's estimated risk of ACS.

Images