U.S. patent application number 14/011730 was filed with the patent office on 2014-02-13 for methods and systems for assessing clinical outcomes.
The applicant listed for this patent is Theranos, Inc.. Invention is credited to Ian Gibbons, Elizabeth A. Holmes, Timothy Michael Kemp, Seth Michelson.
Application Number | 20140046683 14/011730 |
Document ID | / |
Family ID | 41114337 |
Filed Date | 2014-02-13 |
United States Patent
Application |
20140046683 |
Kind Code |
A1 |
Michelson; Seth ; et
al. |
February 13, 2014 |
Methods and Systems for Assessing Clinical Outcomes
Abstract
Described herein are methods and systems useful for
characterizing clinical outcomes of a subject. Provided herein
includes computer-assessed methods, medical information systems,
and computer-readable instructions that can aid an end-user in
diagnosis, prognosis, and treatment of a clinical outcome.
Inventors: |
Michelson; Seth; (San Jose,
CA) ; Kemp; Timothy Michael; (San Jose, CA) ;
Gibbons; Ian; (Portola Valley, CA) ; Holmes;
Elizabeth A.; (Palo Alto, CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Theranos, Inc. |
Palo Alto |
CA |
US |
|
|
Family ID: |
41114337 |
Appl. No.: |
14/011730 |
Filed: |
August 27, 2013 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13609144 |
Sep 10, 2012 |
8538774 |
|
|
14011730 |
|
|
|
|
13244762 |
Sep 26, 2011 |
8265955 |
|
|
13609144 |
|
|
|
|
12412334 |
Mar 26, 2009 |
|
|
|
13244762 |
|
|
|
|
61039721 |
Mar 26, 2008 |
|
|
|
Current U.S.
Class: |
705/2 |
Current CPC
Class: |
G16H 50/50 20180101;
Y02A 90/10 20180101; G16H 50/70 20180101; G06Q 10/00 20130101; G16C
20/70 20190201; G16B 40/00 20190201; G16C 20/90 20190201; G16H
40/20 20180101; G06Q 10/10 20130101 |
Class at
Publication: |
705/2 |
International
Class: |
G06F 19/00 20060101
G06F019/00 |
Claims
1.-30. (canceled)
31. A computer-implemented method of predicting the occurrence of a
medical condition, the method comprising: (a) measuring
concentrations of a first set of biomarkers present in a subject at
a given frequency, wherein the first set of biomarkers are
suspected to be predictive of the medical condition; (b) based on
the concentrations measured in (a) and with the aid of a computer
processor, (i) generating from the first set a subset of biomarkers
that are more correlative with the occurrence of the medical
condition as compared to the first set and/or (ii) generating a new
frequency of measurement of the first set and/or the subset of
biomarkers; and (c) measuring (i) concentrations of the subset of
(b) and/or (ii) following the new frequency of measurement of the
first set or the subset of biomarkers to yield a trajectory that is
predictive of the occurrence of the medical condition, which
trajectory is within a probability space that is defined by
discrete clinical outcomes associated with said medical condition,
wherein each discrete clinical outcome is characterized by a
statistical distribution of a set of biological marker(s) among
said first set of biomarkers.
32. The method of claim 31, wherein (a) further comprises measuring
one or more physiological indicators of said subject at said given
frequency.
33. The method of claim 31, further comprising analyzing data
reflecting the concentrations and/or the physiological indicators
with multivariate statistical software.
34. The method of claim 31, wherein the biomarkers are present in a
biological sample of said subject.
35. The method of claim 31, further comprising notifying a medical
personnel or the subject of a need for taking a medical action
based upon said trajectory.
36. A computer-implemented method for characterizing a medical
condition of a subject, comprising: (a) obtaining, from the
subject, a set of subject data over a series of time points,
wherein said set of subject data corresponds to a first set of
biological parameters comprising at least one biological marker and
at least one physiological parameter; (b) determining, with the aid
of a computer processor, the probability that, at a given time
point in said series, said subject data belongs to a given discrete
clinical outcome among a plurality of discrete clinical outcomes
associated with said medical condition, each of which discrete
clinical outcomes is characterized by a statistical distribution
comprising a subset of biological parameters among said set of
biological parameters; and (c) producing, with the aid of a
computer processor, a time series analysis of said set of subject
data to characterize, in a probability space defined by said
plurality of discrete clinical outcomes, a state or trend of the
medical condition of said subject.
37. The method of claim 36, further comprising selecting a second
set of subject data from said set of subject data, wherein said
second set of subject data is more indicative of the probability of
the medical condition as compared to said set of subject data.
38. The method of claim 37, further comprising obtaining said
second set of subject data from the subject at various time
points.
39. The method of claim 36, further comprising notifying a medical
personnel or the subject of a need for taking a medical action
based upon said state or trend.
40. A computer-implemented method for monitoring subject response
to therapy, comprising: (a) obtaining longitudinal subject data
corresponding to at least one biomarker in a set of biomarkers for
a clinically relevant condition to generate a trend of the subject
data in a probability space that is defined by a plurality of
discrete clinical outcomes associated with said clinically relevant
condition, each of which discrete clinical outcomes is
characterized by a statistical distribution of biomarker(s)
selected from said set of biomarkers, wherein the longitudinal
subject data is obtained from a subject receiving a therapy; (b)
monitoring, with the aid of a computer processor, a periodicity of
the trend; and (c) corresponding the periodicity of the trend to a
response to the therapy received by the subject.
41. The method of claim 40, further comprising identifying said set
of biomarkers prior to (a).
42. The method of claim 40, wherein said trend of the subject data
comprises a trajectory that is indicative of the likelihood of
progression of the subject data to a discrete clinical outcome.
43. The method of claim 40, wherein (c) further comprises measuring
values of peak measurements of the periodicity of the trend.
44. The method of claim 43, further comprising characterizing the
values of the peaks, thereby characterizing the periodicity of the
clinically relevant condition.
45. The method of claim 44, further comprising calculating, with
the aid of a computer processor, a velocity based on said trend
and/or said characterization, wherein said velocity is indicative
of the progression, regression or state of the medical condition of
the subject.
46. The method of claim 40, further comprising notifying a medical
personnel or the subject of a need for taking a medical action
based upon a correspondence between the periodicity of the trend
and the response to the therapy.
47. A computer-implemented method for monitoring health or a
medical condition of a subject, comprising: (a) obtaining
longitudinal subject data corresponding to at least one set of
markers for a clinically relevant condition to obtain a trend of
the subject data in a probability space that is defined by a
plurality of discrete clinical outcomes associated with said
clinically relevant condition, wherein said trend is characterized
by a time-dependent probability of said subject data belonging to a
cluster of data associated with each set characterizing each
discrete clinical outcome, and wherein the subject data is obtained
from the subject receiving a therapy; (b) monitoring, with the aid
of a computer processor, a time dependent change of the
probability; and (c) correlating, with the aid of a computer
processor, the time dependent change of the probability to the
therapy received by the subject, thereby monitoring the health or
medical condition of said subject.
48. The method of claim 47, further comprising measuring one or
more physiological indicators of said subject at a given frequency
prior to monitoring said time dependent change.
49. The method of claim 47, further comprising generating a report
that indicates the correlation of the time dependent change of the
trend to the therapy received by the subject.
50. The method of claim 47, further comprising notifying a medical
personnel or the subject of a need for taking a medical action
based upon said monitoring.
Description
CROSS-REFERENCE
[0001] This application is a Continuation application of U.S. Ser.
No. 13/609,144, filed on Sep. 10, 2012, which is a Continuation
application of U.S. Ser. No. 13/244,762, filed on Sep. 26, 2011,
issued as U.S. Pat. No. 8,265,955 on Sep. 11, 2012, which is a
Divisional application of U.S. Ser. No. 12/412,334, filed on Mar.
26, 2009, which claims the benefit of U.S. Provisional Application
No. 61/039,721, filed Mar. 26, 2008, all of which are incorporated
herein by reference in their entirety.
BACKGROUND OF THE INVENTION
[0002] Conventional methods for assessing a patient's clinical
outcome are primarily based on clinicians' judgment and past
experience. The conventional methods generally involve laboratory
tests, patient surveys and office visits at isolated time points,
all of which are not scalable for a time series analysis,
especially for one that tracks or predicts the trend of a patient's
clinical outcome in real time. Intrinsic to the conventional
methodologies is the profound drawback that a relatively small set
of information such as a single clinician's personal preference is
taken into consideration in reaching a clinical decision. As such,
under the existing medical system, patient care becomes
increasingly difficult when multiple variables are involved. In
particular, there lacks a system and method to effect a
multi-dimensional analysis in which a large set of biomarkers are
used to aid in the diagnosis, prognosis, and treatment of a
clinical outcome or the design and execution of a clinical
trial.
[0003] Multivariate statistics are generally concerned with
determining a statistical distribution of an outcome or a series of
outcomes based on multiple variables. Inherently, most medical
conditions and treatments are multivariate due to the complexities
of the physiology. The discoveries of a vast number of disease
biomarkers and the establishment of miniaturized analytic systems
have made a new paradigm of patient care that makes multivariate
analysis feasible. A desirable new paradigm would provide rapid
access to information characterizing clinical outcome and then
automatically linking that information through customized
communication channels so that the desired medical actions
(adaptive dose ranging, clinical decision making and so forth) can
be performed. Also desirable is the ability to integrate
information from an individual's blood tests with other
physiologically relevant factors, and present that information in
an actionable format. The technology described herein satisfies
these needs and provides related advantages as well.
SUMMARY OF THE INVENTION
[0004] The present invention provides a medical information system
for subject data analysis. In one aspect, a system of the present
invention is particularly useful for advancing the future of blood
testing and data analysis. For example, the system can be part of
an integrated infrastructure built around real-time, point-of-care
consumer blood monitoring devices which analyze a small blood
sample (e.g., 500 ul, 50 ul, 25 ul, 10 ul or even less) and
wirelessly transmit that information to a database which integrates
real-time data with stored data from disparate databases (patient
record, genetic/genomic information, data from pivotal trials) into
one central repository. The system then allows for the automatic
application of multivariate, multidimensional mathematics to the
data repository to perform specific commands or tasks, e.g.,
mapping real-time PK/PD dynamically in the context of the
pathophysiology of a given medical condition.
[0005] In another aspect, a system of the present invention can be
used to improve the label of key drugs through adaptive clinical
studies which generate publications for label expansions for new
indications, patient subpopulations, and for ameliorating safety
concerns. The development of such a system for home, real-time
blood monitoring has significant implications which allow one to
collect information which is otherwise not available through the
use of the conventional laboratory testing.
[0006] The medical information system typically comprises (a) an
input device for receiving subject data and in communication with a
processor; (b) storage unit in communication with the processor
having a database for: (i) storing data corresponding to a
probability space defined by a set of discrete clinical outcomes,
each of which is characterized by statistical distribution of at
least one biological marker; and (ii) storing subject data
corresponding to the at least one biological marker; (c) a
processor that calculates the position of said subject data in said
probability space as a way of assessing the probability of a
discrete clinical outcome of said subject; and (d) an output device
that transmits information relating to the discrete clinical
outcome of c) to an end user.
[0007] Non-limiting clinical outcome that the system is adapted to
predict can be selected from the group consisting of but not
limited to: complete response (CR), partial response (PR), stable
disease (SR), non-response (NR), adverse drug effect (ADR), and
drug toxicity. In using the medical information system, the end
user can be a medical personnel or the subject himself or herself.
In some instances, the end user is from a pharmaceutical
company.
[0008] In one aspect, the processor of the system calculates the
position of said subject data in said probability space as a way of
assessing the probability of a discrete clinical outcome of said
subject.
[0009] In another aspect, the input device of the system comprises
a touch screen. Where desired, the input system can comprise a data
entry portal or a keyboard. The subject data to be input into,
processed by, or transmitted as an output by the system can be
textual, numeric or a category. Where desired, the textual or
numeric information is solicited from the end user.
[0010] In some instances, the subject data represent measurements
of the at least one biological marker present in a bodily fluid. In
some instances, the measurements are obtained by a point-of-care
device that is operated by the subject. The measurements can be
taken at various time points to yield a trajectory within the
probability space, wherein said trajectory represents a time series
of the assessed clinical outcome. The various time points can cover
a period of less than or about 24 hours
[0011] In another aspect, the medical information system comprises
an output device having an automatic alert system. The automatic
alert system can be programmable by the end user. Where desired,
the automatic alert system is programmable based on a predefined
protocol for a clinical trial. In another aspect, the output device
of the system transmits selected portions of the subject data and
the probability space in response to instructions from the end
user. In yet another aspect, the information transmitted by the
output device is encrypted. In still another aspect, the
information transmitted by the output device represents an
assessment of the clinical outcome of said subject at a single time
point. The information transmitted by the output device can
represent a time series of the assessed clinical outcome.
[0012] In still another aspect, the input device and/or the output
device of the system comprises a user interface that can remotely
access the network.
[0013] In yet another aspect, the medical information system
comprises a network.
[0014] In yet another aspect, the storage unit of the system stores
historical reference data of a plurality of subjects in
relationship to the at least one biological marker. Where desired,
the data stored in the storage unit are selected from the
categories consisting of pathology, anatomy, treatment option,
treatment outcome, pharmacological parameter, pharmacokinetics
parameter, psychological parameter, and genomic information. The
database can be public, internal.
[0015] The end user of the medical information system can be a
health care provider, including without limitation a Health
Maintenance Organization (HMO).
[0016] The present invention further provides a method for
characterizing the probability of a clinical outcome of a subject.
The method comprises the steps of (a) constructing a probability
space defined by a set of discrete clinical outcomes, each of which
is characterized by a statistical distribution of at least one
biological marker; (b) obtaining subject data corresponding to the
at least one biological marker; and (c) calculating the position of
said subject data in said probability space, thereby characterizing
the probability of the clinical outcome of said subject.
[0017] Also provided is a method of characterizing a clinical
outcome of a subject comprising: (a) constructing a probability
space within a server, wherein the probability space is defined by
a set of discrete clinical outcomes, each of which is characterized
by the statistical distribution of at the least one biological
marker; (b) entering data of a subject into the server, said data
corresponding to the at least one biological marker; and (c)
calculating the position of said subject data in said probability
space thereby characterizing the clinical outcome of the subject.
In some embodiments, in practicing the subject methods, at least
steps b and c are repeated at various time points to yield a
trajectory within a probability space, wherein said trajectory is
indicative of the likelihood of progression to the clinical
outcome. The subject methods can comprise the step of notifying a
medical personnel or the subject of a need for taking a medical
action upon assessing or characterizing the position of said
subject data in said probability space. In some instances, the
medical action involves at least one action selected from the group
consisting of altering a dosage of an existing therapeutic agent
administered to said subject, administering a different a
therapeutic agent, and administering a different combination of
therapeutic agents. Notification of a medical action can be
electronically transmitted, e.g., wirelessly transmitted. The
subject methods can further comprise step of, upon selection of the
at least one action, performing an outcome analysis for assessing a
result of said selected action, and automatically updating the
probability of a discrete clinical outcome of said subject.
[0018] Further provided in the present invention is a computer
readable medium comprising computer readable instructions, which
when executed cause a processor to: a) provide a probability space
defined by a set of discrete clinical outcomes, each of which is
characterized by a statistical distribution of at least one
biological marker; b) obtain subject data corresponding to the at
least one biological marker; and c) calculate the position of said
subject data in said probability space to assess the probability of
a clinical outcome of said subject. In general, the instructions
operate in a software runtime environment. In one aspect, the
instructions when executed further causes a processor to provide a
user defined alert condition based on an assessment of trajectory
parameters of the subject data in the probability space, wherein
said trajectory parameters are at least one of speed, acceleration,
direction, and position.
[0019] In an aspect, a method is provided herein of predicting the
occurrence of a medical condition that requires medical
intervention, the method comprising: (a) measuring concentrations
of a first set of biomarkers present in a subject and measuring one
or more physiological indicators of said subject at a given
frequency, wherein the first set of biomarkers are suspected to be
predictive of the medical condition; (b) based on the
concentrations measure in (a), generating from the first set a
subset of biomarkers that are more correlative with the occurrence
of the medical condition and/or a new frequency of measurement of
the biomarkers; and (c) measuring concentrations of the subset of
(b) and/or following the new frequency of measurement of one or
more biomarkers, thereby predicting the occurrence of the medical
condition.
[0020] In some instances, a method further comprises analyzing data
reflecting the concentrations and/or the physiological indicators
with multivariate statistical software. In some instances, the
biological markers are present in a biological sample of said
subject. The biological sample can be diluted by an appropriate
fold to ensure, e.g. the appropriate range of concentration level
is detected.
[0021] In another aspect herein, a method of monitoring sepsis
development of a subject comprises: measuring at least two
parameters selected from the group of (1) body temperature of said
subject, (2) protein C concentration of said subject, (3)
interleukin 6 (IL-6) concentration of said subject, multiple times
to yield a trend of temperature, protein C trend, and/or IL-6; and
wherein an increase beyond normal body temperature, a decrease in
protein C concentration and/or an increase in IL-6 concentration is
indicative of the development of sepsis in said subject. In some
aspects, a decrease in protein C and an increase of IL-6 can be
indicative of the development of sepsis in said subject. In some
other aspects, a decrease in protein C and an increase of IL-6 and
an increase beyond normal body temperature can be indicative of the
development of sepsis in said subject. In some instances, an at
least about 10-fold increase in IL-6 concentration in said subject
is indicative of the occurrence of sepsis in said subject. In a
further instance, an at least about 100-fold increase in IL-6
concentration in said subject is indicative of the occurrence of
sepsis in said subject. This method may further comprise the step
of increasing frequency of measuring IL-6 concentration upon an
increase beyond normal body temperature and/or a decrease in
protein C concentration. For example, the frequency of IL-6
measurement can be increased to once a day, once every 12, 8, 6, or
4 hours. A determination of the occurrence of sepsis may be
promptly followed by an appropriate medical intervention.
[0022] Also described herein is a method for characterizing a
medical condition of a subject, comprising: obtaining a first set
of subject data comprising at least one biological marker and at
least one physiological parameter from the subject; determining the
probability of a medical condition of the subject using the first
set of subject data obtained; selecting a second set of subject
data from the probability of the medical condition; and obtaining
the second set of subject data from the subject, thereby
characterizing the medical condition of the subject.
[0023] In yet another aspect, a method is disclosed for
characterizing periodicity of a clinical condition of a subject,
the method comprises: identifying a set of biomarkers for a
clinically relevant condition; obtaining longitudinal subject data
corresponding to at least one biomarker in said set to obtain a
trend of the subject data; analyzing said trend to identify
periodic changes in the at least one biomarker; measuring values of
peak measurements of the periodic changes of the trend; and
characterizing the values of the peaks thereby characterizing the
periodicity of the clinically relevant condition. In some
instances, the analyzing step comprises developing an ARIMA model
to determine a differencing lag in the underlying model. The
differencing lag can be used to de-trend the trend and establish a
stationary trend. In some instances, the analyzing step comprises
calculating an autocorrelation function and the measuring step
comprises identifying the statistically significant peaks in the
autocorrelation function. In some instances, the analyzing step
comprises calculating spectral density and the calculating spectral
density is performed using a Fast Fourier Transform. The measuring
step can comprise identifying the power spectrum of the maximum
spectral density frequency.
[0024] In an aspect described herein, a method for monitoring
subject response to therapy comprises: obtaining longitudinal
subject data corresponding to at least one biomarker in a set of
biomarkers for a clinically relevant condition to obtain a trend of
the subject data, wherein the subject data is obtained from a
subject receiving a therapy; monitoring periodicity of the trend;
and corresponding the periodicity to a response to the therapy
received by the subject. In some instances, the therapy is a
periodic dosing regimen. In some instances, the response to the
therapy is characterized by a time-dependent behavior of peak
levels of the trend. In some instances, the time-dependent behavior
is substantially constant. In other instances, the time-dependent
behavior is changing linearly. In yet other instances, the
time-dependent behavior is changing exponentially.
[0025] In an aspect, a method is disclosed herein for
characterizing the emergence of clinically relevant subpopulations
of patients exposed to a therapeutic agent, the method comprising:
identifying a set of biomarkers in a blood sample that act as
surrogate markers for the therapeutic agent; measuring the set of
biomarkers longitudinally from a group of patients exposed to the
therapeutic agent; identifying distinct clusters in a multivariate
clustering space of the measured values of the set of biomarkers
from the group of patients; determining the rate of separation of
the distinct clusters and measuring the distance between the
distinct clusters in a statistical manner; obtaining patient
information from the group of patients to classify the patients in
clinically relevant subpopulations; and comparing the distinct
clusters to the clinically relevant subpopulations to characterize
sensitivity and specificity of the distinct clusters to predict the
clinically relevant subpopulations. In some instances, the method
further comprises identifying a second set of biomarkers configured
to improve the characterization of the emergence of distinct
clusters to predict the clinically relevant subpopulations. In some
instances, the group of patients exposed to the therapeutic agent
are participants in a clinical trial. In some instances, the
clinical trial is a dose ranging trial. In other instances, the
clinical trial is a part of an adaptive clinical trial. The
adaptive clinical trial can be designed to characterize an optimal
dosing regimen or can be designed to characterize an optimal
responder population. In some instances, the measuring the distance
step comprises measuring the Mahalanobis distance between the
distinct cluster centroids. In other instances, the measuring the
distance step comprises measuring the nearest-neighbors distance
between the distinct clusters. In yet other instances, the
measuring the distance step comprises measuring a Euclidean
distance measure between the distinct clusters. In some instances,
the measuring the distance step measuring a Manhattan distance
measure between the distinct cluster.
INCORPORATION BY REFERENCE
[0026] All publications and patent applications mentioned in this
specification are herein incorporated by reference to the same
extent as if each individual publication or patent application was
specifically and individually indicated to be incorporated by
reference.
BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Many features of the invention are set forth with
particularity in the appended claims. A better understanding of the
features and advantages of the invention will be obtained by
reference to the following detailed description that sets forth
illustrative embodiments, in which many principles of the invention
are utilized, and the accompanying drawings of which:
[0028] FIG. 1 illustrates a general representation of the
Mahalanobis distance method and the representative ellipses. The
figure also includes the Mahalanobis distance mathematical
equation.
[0029] FIG. 2 illustrates an example of the results of a cluster
analysis of a clinical outcome.
[0030] FIG. 3 is a flow chart illustrating an exemplary method of
assessing a clinical outcome of a subject and assigning
probabilities thereto.
[0031] FIG. 4 illustrates an example probability space geometrical
representation.
[0032] FIG. 5 illustrates a trajectory of data of a subject in a
probability space.
[0033] FIG. 6 is a flow chart illustrating an exemplary method of
alerting a user to assess a clinical outcome of a subject.
[0034] FIG. 7 demonstrates a method of establishing rules for a
probability space of clinical outcomes.
[0035] FIG. 8 illustrates a graphical presentation of rules for a
probability space.
[0036] FIG. 9 is a flow chart illustrating an exemplary method of
implementing a mathematical model to assess a clinical outcome.
[0037] FIG. 10 is a flow chart illustrating an exemplary system
comprising a server for communicating and processing medical
data.
[0038] FIG. 11 illustrates an exemplary embodiment of a system of
the invention with real-time acquisition of subject data and the
transmittal of the information in real-time to a server or system
that is capable of converting the information or data to a
physiologically relevant context.
[0039] FIG. 12 demonstrates an exemplary system of the invention,
wherein a health care operating system comprises data
infrastructure, models and algorithms, and software
applications.
[0040] FIG. 13 demonstrates an exemplary database comprising an
ontology of biomarkers that are related some types of medical
conditions.
[0041] FIG. 14 illustrates the an exemplary hierarchy of a
healthcare operating system of the invention wherein the central
operating system has access to a data infrastructure and monitoring
devices for obtaining subject data.
[0042] FIG. 15 demonstrates an exemplary system of the invention
wherein an analysis database can receive information from a
database containing data from a reader or cartridge database,
historical databases, a patient database that can contain data
entered by a patient, and a customer database.
[0043] FIG. 16 illustrates exemplary trajectories of two subjects
in a probability space representing prostate cancer discrete
clinical outcomes.
[0044] FIG. 17 illustrates exemplary trajectories of two subjects
in a probability space representing clinical outcomes of a subject
developing sepsis.
[0045] FIG. 18 illustrates exemplary trajectories of two subjects
in a probability space representing the medical condition of
diabetic subject using an insulin sensitizer.
[0046] FIG. 19 illustrates a method of graphically representing a
trajectory of subject data in a probability space in two
dimensions.
[0047] FIG. 20 illustrates a method of graphically representing a
trajectory of subject data in a probability space in two dimensions
and the position of the data for defining discrete clinical
outcomes.
[0048] FIG. 21 demonstrates a dynamic subpopulation emergence based
on cluster analysis of a Monte Carlo simulation of a two-cohort
clinical design.
[0049] FIG. 22 demonstrates plots of the cluster calling statistics
versus the hypothetical number of clusters.
[0050] FIG. 23 illustrates markers of clinical effect with the
segregation of the marker space as a function of dose at the end
point of a clinical study.
[0051] FIG. 24 demonstrates patient data showing the seasonality of
two type 1 biomarkers over time.
[0052] FIG. 25 demonstrates a time series of a biomarker value
versus the time of chemotherapy delivery.
[0053] FIG. 26 demonstrates the relationship between parameters
computed from circulating pyrogen levels measured over time and/or
and the time-to-spike.
[0054] FIG. 27 is the IL-1 beta concentration versus time point
number.
[0055] FIG. 28 is the sample-to-sample fold change for IL1-beta
versus time point number.
[0056] FIG. 29 is the IL-6 concentration versus time point number,
wherein patient 4 became septic.
[0057] FIG. 30 is the sample-to-sample fold change for IL-6 versus
time point number, wherein patient 4 became septic.
[0058] FIG. 31 is the TNF-alpha concentration versus time point
number.
[0059] FIG. 32 is the sample-to-sample fold change for TNF-alpha
versus time point number.
[0060] FIG. 33 illustrates the residual plot for Patient 5 who has
a fever spike and no decreasing markers.
[0061] FIG. 34 demonstrates Patient 4 data points showing
decreasing Protein-C with increasing IL-6 just prior to fever
spike.
[0062] FIG. 35 is a schematic of the Th1-to-Th2 switch and the
cytokines representing each phenotype.
[0063] FIG. 36 and FIG. 37 demonstrate that the system uses an
equation that is anticipative of sepsis using the Squared
Maholanobis distance equation and Bayesian Probability.
[0064] FIG. 38 illustrates a graph in time of a plurality of marker
proteins for a patient in a sepsis trial.
[0065] FIG. 39 illustrates a bivariate time course of two
particular markers (protein C and C-reactive protein) in the same
patient.
DETAILED DESCRIPTION OF THE INVENTION
[0066] In one embodiment, a method is provided herein for
characterizing the probability of a clinical outcome of a subject.
The method comprises the steps of (a) constructing a probability
space defined by a set of discrete clinical outcomes, each of which
is characterized by a statistical distribution of at least one
biological marker; (b) obtaining subject data corresponding to the
at least one biological marker; and (c) calculating the position of
said subject data in said probability space, thereby characterizing
the probability of the clinical outcome of said subject.
[0067] In practicing the methods herein, one generally utilizes a
set of biological markers (also referred to herein as biomarkers)
relevant to a given clinical outcome. In order to improve the
reliability and accuracy of a mathematical calculation, biomarkers
are selected based upon multivariate statistics from data
pertaining to a clinical outcome, whether it being, for example, a
disease or a medical procedure. In some embodiments, discriminant
analysis is performed to determine the most relevant biomarkers
pertaining to a particular medical clinical outcome.
[0068] In an aspect, a method is provided herein of predicting the
occurrence of a medical condition that requires medical
intervention, the method comprising: (a) measuring concentrations
of a first set of biomarkers present in a subject and measuring one
or more physiological indicators of said subject at a given
frequency, wherein the first set of biomarkers are suspected to be
predictive of the medical condition; (b) based on the
concentrations measure in (a), generating from the first set a
subset of biomarkers that are more correlative with the occurrence
of the medical condition and/or a new frequency of measurement of
the biomarkers; and (c) measuring concentrations of the subset of
(b) and/or following the new frequency of measurement of one or
more biomarkers, thereby predicting the occurrence of the medical
condition.
[0069] Pluralities of discrete clinical outcomes, sometimes
referred to as medical outcomes, for a specific medical condition
are plotted in a geometrical manner. A set of values of the
biomarkers are chosen that are the most representative of each
clinical outcome. Using the mathematical methods described herein,
probabilities of class assignment to each discrete clinical outcome
can be assigned to any set of values of observed biomarkers. A
probability space can then be constructed by plotting the
probability of any set of values of observed biological markers
within the geometric space defined by the plurality of discrete
clinical outcomes.
[0070] Model Building
[0071] In some instances, mathematical modeling describes the
dynamic state of a system. In some instances, the models use a
system of equations representing outcome parameters, biomarker and
drug concentrations and the like over time. Such equations might be
simple enough to solve in closed form, yielding an algorithmic
formula. However, in some instances, models of human disease can be
too complex to permit a simple closed form solution. In some
instances, a solution as described herein can be projecting
numerical solutions forward in time, also known as termed
predictive biosimulation.
[0072] The objective of cluster analysis is to group observations
into clusters such that each cluster is as homogeneous as possible
with respect to the clustering variables. Any conventional
clustering algorithm may be used to define the discrete set of
clinical outcomes based upon the recognized biomarkers. Besides the
Discriminant Function Analysis, there are many such algorithms,
such as "Single Linkage," "Complete Linkage," "K means," "Ward's
Method," or the "Centroid Method." Additionally, one can identify
clusters using "Decision Trees". These algorithms are well-known to
anyone familiar with the art, and are available in standard
statistical packages such as SAS and SPSS. The clustering
algorithms group like objects together, and keep unlike objects in
separate groups.
[0073] Multivariate statistics to identify classes of similar
subjects in a sample population can also be applied for building an
appropriate model. The techniques currently employed include, but
are not limited to, Discriminant Function Analysis (DFA),
Hierarchical Clustering Analysis, Factor Analysis (in which an
underlying model or relationship is assumed), Self-Organizing Maps
(SOMs), Support Vector Machines (SVMs), and Neural Nets. Each is a
pattern recognition technology using multivariate descriptor
vectors, which subjects are classmates, to more completely manage
an adaptive clinical trial.
[0074] Other techniques estimating the model parameter space
include several conventional methods, such as the graphical method
and gradient-based nonlinear optimization (Mendes and Kell, 1998).
The graphical method is typically applied to those problems that
can be converted to linear regression problems, and are generally
not amenable to closed-loop learning as they require human
intervention at each step. The gradient-based nonlinear
optimization method does not have such a restriction; however it
does require information about the error function with respect to
the parameter estimates. As such it often converges to local minima
Evolutionary algorithms including genetic algorithms, evolutionary
programming, genetic programming, and their variants (Back et al.,
1997; McKay et al., 1997; Edwards et al., 1998) have been applied
to overcome these limitations.
[0075] In some instances, to best account for the dynamics of a
biological system, the equation systems discussed above can be
composed of ordinary differential equations which calculate the
rate of change of a biological entity over a fixed time segment.
That quantification is a function of the state of the system at the
instant of analysis, and includes terms for all other entities in
the system that might affect that rate of change.
[0076] In many instances, as described herein, biological entities
of interest are found in blood comprising but not limited to
circulating proteins. In the modeling space, the time rate of
change of each biological entity is called the derivative and is
quantified as the instantaneous slope of the biological entity's
concentration curve in the time domain.
[0077] The value of any selected target biological entity at each
selected sampling time point can be calculated as well as the rate
of change for every entity in the sampling space. This can result
in a point estimate per biological entity for the derivative in
units comprising but not limited to a change in concentration per
unit time.
[0078] In some instances, a method can utilize two point slopes
along with a K-point model fitting regression equations through the
last K observations, to estimate the first, second, and higher
order derivatives, for example:
dP(t)/dt=f(P(t), {other proteins, cells types, etc.}, model
parameters)
where P(t) is the concentration of a given protein as a function of
time. And where f is function of P(t) and entities affecting
specific protein concentration curves, comprising second signals,
cell numbers, and the like. And where model parameters are
coefficients in those equations that quantify those relationships,
for example comprising production rates and clearance rates. The
system can be linear or non-linear.
[0079] An alternate embodiment of the mechanism is a system of
entities where P(t) is a protein vector .PI.(t)=[P.sub.1(t),
P.sub.2(t), . . . P.sub.N(t)] where N is the number of proteins
chosen to sample, and f(P, etc.) becomes a matrix M(.PI.(t);
D(.PI.(t)). The system parameters become a vector as
.beta.(.PI.(t)).
[0080] For example, clearance rates of each P.sub.j(t) in the
vector .PI.(t) are captured in this vector form. .beta.(.PI.(t))
characterizes the functional coefficients of the matrix M. The
system is linear when .beta. does not depend on .PI.(t) or on time.
The system is linear and parametrically time-dependent when .beta.
depends on t, and varies linearly with time.
[0081] In another example, the system is nonlinear when the
clearance rate changes with time and can represent the
up-regulation or down-regulation in time. The system is represented
but is not limited by the vector equation:
d.PI.(t)/dt=M(.PI.(t);.beta.(.PI.(t))*.PI.(t)
[0082] In yet another example, the dynamic state of biological
systems has a level of specificity which is captured in an
exemplary biosimulation system. Every entity in the system will
typically act on a small subset of other entities. For example,
protein 2 modifies the rate of production of protein 1 by binding
to a cell surface receptor on an activated effector cell. The
concentration of protein 2 modifies the rate of production of
protein 1. When a third protein acts on the production of protein 2
the rate of production of protein 1 depends indirectly on the level
of production of protein 2.
[0083] The equations that best represent the above example in this
biosimulation system are:
dP.sub.1/dt=.beta..sub.12P.sub.2(t)
dP.sub.2/dt=-.beta..sub.22P.sub.2(t)+.beta..sub.23P.sub.3(t)
dP.sub.3/dt=0
[0084] In these exemplary equations, the concentration of Protein 3
is constant and the rate of Protein 1 production is an increasing
function of the Protein 2 levels, factored by .beta..sub.12>0.
In addition, the levels of Protein 2 are a function rate Protein 2
is cleared from the system, in this case, proportionally by a
factor .beta..sub.22>0, and the effects of P.sub.3(t) on its
production, described by a factor .beta..sub.23>0 and is
parameterized by the vector (.beta..sub.12, .beta..sub.22,
.beta..sub.23). Any triplet vector can be used in the general case
to generate any family of trajectories in time for Proteins 1 and
2. Protein 3 is non-varying in this model since its derivative
remains zero.
[0085] By measuring Proteins 1 and 3 at each time point in a
monitoring period, values for P.sub.1(t), and P.sub.3(t), and
estimates for dP.sub.1/dt and dP.sub.3/dt are obtained. P.sub.2(t)
can then be solved and constrain the solution set to a parameter
vector (.beta.*.sub.12, .beta.*.sub.22, .beta..sub.23), which is
defined by well defined algebraic equations. Estimates of both
P.sub.2(t) and dP.sub.2/dt derive the state of the constrained
system based on point estimates as derived from the samples. This
allows for projecting forward in time where a model trajectory will
likely be at the next sample point, and re-adjustment of the model
parameters as necessary in a dynamic feedback self-correcting
system.
[0086] The self-correcting mechanism and feedback system can be
implemented in a periodically sampled measurement space, and that
model refinement and re-parameterization can take place at each
sample interval. After each application of this self-correcting
methodology, the space of acceptable parameter vectors should begin
to converge to a patient specific vector that projects forward in a
predictive trajectory.
[0087] A system that dynamically characterizes patient
subpopulations based on the longitudinal sampling of predictive
protein profiles can use that information as a Type 1 biomarker to
verify to the researcher scientist that the compound is acting on
the target as expected. The mechanism of action for that compound
can also be researched for whether it is acting as expected.
Characterization of the dynamic emergence of the subpopulations is
based upon their behaviors and the protein profiles they
exhibit.
[0088] In an example, one population is treated with the test
compound while another is an untreated control. If the protein
profiles of the treated population behave in accordance with the
underlying hypothesis as it pertains to the compound's mechanism of
action, the protein profiles will, as a conglomerate pattern, act
as a Type 1 biomarker. If individuals within that population show
considerable variance in both the direction and velocity of the
protein profile, adjustment of the doses of the slower population
members can occur and the trial protocol can be changed
accordingly.
[0089] A variety of statistical analyses are applicable for
building a probability space or model. FIG. 1 illustrates an
exemplary characterization of similarity as measured by the
Mahalanobis distance in a two-dimensional (multivariate) space. The
outcome of a discriminant function analysis is typically a
Mahalanobis distance for every subject in the analysis data set.
The methods of performing the discriminant analysis in this manner
or similar manners are known to those skilled in the art. In this
example, the data from two independent variables (for example,
biomarkers) are plotted against each other. Each patient in the
analysis set is represented by a single point on the graph. The
distance between each patient's biomarker pair pattern and the
centroid of the cluster of data is calculated as the Mahalanobis
distance and is represented by the ellipses in the FIG. 1, wherein
the centroid can be typically defined as the mean sector of all
independent measures in the space of interest. The probability that
any particular patient belongs to a cluster depicted herein is
inversely proportional to the distance from the centroid of the
cluster. In this way the Mahalanobis distance ellipse on the plot
in FIG. 1 is used as a measure of similarity between the two sets
of independent variables. This method also accounts for noise in
the appropriate distance metric by assigning probabilities of class
assignment for a particular patient based on the variance observed
in the underlying clustered data. Thus this method can account for
uncertainty in the probability space.
[0090] In conventional clustering, one typically works from a
distance matrix, which lists the similarity of every object to be
clustered versus every other object. The process begins with the
creation of the distance matrix as known to one skilled in the art.
A triangular matrix of distances among all pairs of patients must
be computed. Each of the distances between subject data will be a
function of the measured biomarkers. The function would take the
form of a sum or weighted sum. The distances for a given variable
are, in turn, a sum of distances between individual observations
for that variable. This sum also may be weighted. For example, each
cluster of subjects potentially represents a discrete clinical
outcome.
[0091] The distance matrix metric can include calculating a
statistical difference, such as the Mahalanobis distance, Euclidean
distance, and Manhattan distance. The matrix represents, in a
statistically rigorous way, the similarity of two multivariate
patterns in a multidimensional space. Based on these distances,
individuals are grouped together in a cluster.
[0092] FIG. 2 illustrates an exemplary clustering method of
defining the discrete clinical outcomes. In this example, the
centroid method was used to determine multiple centroids of
biomarker data patterns. Each centroid either defines or belongs to
a different set of discrete clinical outcomes based upon measured
biomarker information. In FIG. 2, examples of the classification of
biomarker patterns of a medical condition include a partial
responder class, a non-responder class, a complete responder class,
and an adverse drug reaction class. Every data point is related to
a centroid of the class it falls within. The centroid can be set to
be the point most representative of the class.
[0093] After the distance matrix has been constructed and discrete
clinical outcomes have been defined by a set of biomarkers, a
probability space can be constructed from multivariate
distributions (for example, clusters) by applying a metric that
accounts for the variance-covariance structure of the data. Based
upon the measured distance, a probability for class assignment for
any observed biomarker pattern is calculated. The calculated
probability for any obtained biomarker pattern defines the
probability space.
[0094] Where desired, a probability measurement is carried out by
way of a Bayesian calculation. Other methods of calculating the
probability for class assignment include, but are not limited to,
Nelson-Aalen estimation, the ordinary least squares estimation, the
weighted least squares estimation, and the maximum likelihood
estimation. Another method is to use a decision tree analysis and
from the measure determine the ultimate sensitivity and specificity
of the derived rule.
[0095] FIG. 3 illustrates an exemplary method of the invention of
building a mathematical model for assessing a medical condition.
Based upon prior history of a medical condition, a user (for
example, a physician) identifies the relevant clinical outcomes or
outcomes. The physiology, a compound or drug's method of action,
and/or off target treatment effects are then utilized to help
define the borders of a probability space.
[0096] A graphical example of the probability space and the
discrete clinical outcomes that define the borders of the
probability space are illustrated in FIG. 4. Each number represents
a different clinical outcome.
[0097] Biomarkers that are most relevant to the clinical outcomes
of a given medical condition are identified using multivariate
discriminant analysis. Data from the set of biomarkers are
classified according to the discrete clinical outcome to which the
independent data is most closely correlated.
[0098] After the data sets are classified by cluster analysis
and/or discriminant function analysis and identified as clinical
outcomes, the centroid location of a clinical outcome is
determined. For each reference point or independent biomarker data
point, a probability of classification into each reference clinical
outcome is assessed.
[0099] When data from an individual subject is obtained
corresponding to at least one biomarker, the data is input into the
mathematical model and its geometrical position within the
probability space is determined. A distance metric is then used to
determine the position of the subject data with reference to each
clinical outcome. In a preferred embodiment, the distance metric is
the Mahalanobis distance. For the subject data, the model obtains a
distance metric corresponding to each discrete clinical
outcome.
[0100] A statistical estimation method is used to convert the
plurality of distance metrics into a plurality of probabilities for
the subject data. Each probability of the plurality of
probabilities corresponds to a discrete clinical outcome. In an
embodiment, the statistical estimation is a Bayesian estimation.
The plurality of probabilities is used as the coordinates of the
subject data in the geometric probability space, such as the one
illustrated in FIG. 4. The position of the subject data within the
probability space is an assessment of the medical condition of the
subject.
[0101] Rules can be assigned by the user (for example, medical
personnel) to any point in the probability space. The rules may
guide the user in understanding the subject's current medical
condition and guide the user in selecting a course of action.
[0102] The methods described herein can be performed at any point
in time during observation of the subject. Subject data can be
obtained at various time points. The position of the subject data
in the probability space can then be calculated using the
mathematical methods described herein. The steps of obtaining and
calculating the position of subject data in a probability space can
be repeated to yield a trajectory within a probability space,
wherein said trajectory is indicative of a time series of the
assessed clinical outcomes.
[0103] Trajectory
[0104] As a medical condition progresses, a subject is typically
monitored by medical personnel. Monitoring can be preformed by
running a series of tests. As testing methods and devices, such as
point-of-care microfluidic devices, become more reliable and
accessible, data from the patient can be acquired and processed at
multiple time points. The sequential structure of obtained data is
often a time series or a set of longitudinal data, but may also be
data that reflects changes that occur sequentially with no specific
reference to time. The system does not require that the time or
sequence values are equally spaced. In fact, the time parameter can
be a random variable itself.
[0105] In some embodiments, the time series of obtaining subject
data is at least 6, 24, 48, 72, or 96 hours. In other embodiments,
the time series can be at least 7, 30, 60, 90, 180, or more days.
The time series can also be over a series of years.
[0106] The longitudinal data provides valuable insight into the
progression of a medical condition or treatment. However, as the
number of tested items increases, the complexity of analyzing the
longitudinal data increases. It can become very difficult to
understand or determine relevant data about a medical condition or
treatment.
[0107] The longitudinal data acquired from a subject can be grouped
into a vector or a trajectory to assess a medical condition. A
vector is defined geometrically as an arrow where the tail is the
initial point and the head is the terminal point. A vector's
components can relate to a geographical coordinate system, such as
longitude and latitude. Navigation, by way of specific example,
uses vectors extensively to locate objects and to determine the
direction of movement of aircraft and watercraft.
[0108] Physics and engineering fields are probably the most common
users of vector analysis and have stimulated much of the
mathematical research. In the field of mechanics, vector analysis
objects include equations of motion including location, velocity,
direction, and acceleration; center of gravity; moments of inertia;
forces such as friction, stress, and stain; electromagnetic and
gravitational fields.
[0109] Velocity, the time rate of change in position, is the
combination of speed (vector length) and bearing (vector
direction). Acceleration is another common vector quantity, which
is the time rate of change of the velocity. Both velocity and
acceleration are obtained through vector analysis, which is the
mathematical determination of a vector's properties and/or
behaviors.
[0110] Vector analysis of longitudinal subject data pertaining to a
medical condition can be used to understand the medical condition
and statistically determine a discrete clinical outcome.
[0111] For example, the normal condition for the individual can be
observed by plotting biomarker data for the individual. The stable,
normal condition will be a located in one portion of the graph.
[0112] The individual's normal condition may be disturbed by the
administration of a pharmaceutical or a change in medical
condition. Under the effect of the administered pharmaceutical, the
individual's normal condition will become unstable and move from
its original position in the graph to a new position in the graph.
When the administration of a pharmaceutical is stopped, or the
effect of the pharmaceutical ends, the individual's normal
condition may be disturbed again, which would lead to another move
of the normal condition in the graph. When the administration of a
pharmaceutical is stopped, or the effect of the pharmaceutical
ends, the individual's normal condition may return to its original
position in the graph before the pharmaceutical was administered or
to a new or tertiary position that is different from both the
primary pre-pharmaceutical position and the secondary
pharmaceutical-resultant position.
[0113] In some instances, biomarker values can change very rapidly,
such that standard sampling (for example, a lab test) may miss
large changes in the values. In an example, sepsis can occur very
quickly when a patient is using a pharmaceutical or is a member of
a pharmaceutical clinical trial. Many times, early clues as to the
medical condition of the patient are not seen because current
testing procedures do not occur rapidly enough. Also current
testing procedures are inflexible within a short period of time, so
if a patient needs to be tested for a different biomarker value or
physiological parameter within a short period of times (for
example, 4 hours), it is often not possible to receive test
results.
[0114] Large changes in biomarker values can occur such that
conventional lab methods which use only one sample dilution level
cannot measure the analyte due to the range of the device. Often,
more sophisticated lab methods that use multiple dilution ranges
that extend a range can also be overwhelmed by a massive change
(for example, greater than 1000-fold) in a biomarker value. If a
dilution is repeated to obtain the proper range, in many instances
24 hours may have elapsed before the results of the assays are
obtained. As described, a system or method herein can report
results in near real-time fashion, therefore medical personnel has
the ability to change the course of medical action as necessary
from the biomarker values obtained. For example, when an increase
of an analyte between assay increases by 100-fold, 200-fold,
500-fold, 1000-fold, 10,000-fold, or even more, it can increase the
dilution level used for the next sample (for example, 100, 200,
500, 1000, 10,000-fold or higher) in addition to changing the
frequency of measurement of the analyte.
[0115] In an aspect herein, a method of monitoring sepsis
development of a subject comprises: measuring at least two
parameters selected from the group of (1) body temperature of said
subject, (2) protein C concentration of said subject, (3)
interleukin 6 (IL-6) concentration of said subject, multiple times
to yield a trend of temperature, protein C trend, and/or IL-6; and
wherein an increase beyond normal body temperature, a decrease in
protein C concentration and/or an increase in IL-6 concentration is
indicative of the development of sepsis in said subject. A decrease
in protein C followed by an increase of IL-6 can be indicative of
the development of sepsis in said subject. A decrease in protein C
followed by an increase of IL-6 and an increase beyond normal body
temperature can be indicative of the development of sepsis in said
subject. In some instances, an at least about 10-fold, 100-fold,
200-fold, 500-fold, 1000-fold, 10,000-fold, 500,000-fold or more
increase in IL-6 concentration in said subject is indicative of the
development of sepsis in said subject. In a further instance, an at
least about 10-fold, 100-fold, 200-fold, 500-fold, 1000-fold, or
more decrease in protein C concentration in said subject is
indicative of the development of sepsis in said subject.
[0116] In an example, when no significant event is occurring in a
patient following chemotherapy at home, the measurement frequency
of an analyte can be reduced. In some instances, the system can
prompt appropriate action by a user based on prior results.
[0117] In some instances, the methods and systems described herein
allow for the monitoring of the health status or medical condition
of a subject throughout a time frame as small as a few days. For
example, the evolution and pathophysiology of the disease process,
the response to therapy, and the possible onset of untoward side
effects upon exposure to a drug can be monitored by longitudinally
sampling blood. From these samples a profile of key circulating
biomarkers can be established. Specific biomarkers can be selected
based on, for example without limitation: knowledge of the disease
process and its molecular pathophysiology, the mechanism of action
of a given compound, and its observed effects profiles.
[0118] Representation of a similar movement of longitudinal data to
that previously described is illustrated in FIG. 5. At each time
point (T=0 to T=9), the data point is plotted in a probability
space. The line traveled from one point to the next is a
vector.
[0119] Diagnosis of the individual may be aided by studying several
aspects of the movement of the individual's medical condition in a
probability space. The direction (for example, the angle and/or
orientation) of the path followed by the medical condition as it
moves in the graph can aid in the diagnosis, prognosis and
treatment decision-making for a given clinical outcome. The speed
and acceleration of the movement of the medical condition in the
graph may also be diagnostic. Also, the position of the
individual's medical condition at an independent time point may
itself be diagnostic.
[0120] Given that the direction and/or speed of the movement of the
normal condition in the graph is diagnostic, one can use the
direction and/or speed of the initial movement of the normal
condition to predict the consequent, new location of the normal
condition; especially if it is established that, under the effect
of a certain agent (for example, a pharmaceutical), there are only
a certain number of locations in the graph at which an individual's
normal condition will stabilize.
[0121] However, in some aspects of the invention, both diagnosis
and prediction of assessing the medical condition of a subject are
performed by qualified medical personnel and the mathematical model
provides an educational tool and resource.
[0122] FIG. 6 illustrates an exemplary method of the invention for
applying a mathematical model longitudinally in real-time. Relevant
biomarkers and discrete clinical outcomes corresponding to
different biomarker patterns are predetermined from historical
data. Data is then obtained from a subject and the data
corresponding to each relevant biomarker is measured. Data is
preferably obtained from the subject using a point-of-care
microfluidic device, however, it can be obtained in any manner.
Each time the data is obtained a distance metric (for example,
Mahalanobis distance) is calculated in relation to each centroid
obtained from a cluster analysis of the historical data. The
centroids represent the discrete clinical outcomes determined from
the historical data.
[0123] As following the flow diagram in FIG. 6, for each individual
data set through time, the probability of the data belonging to
each clinical outcome is calculated. The probability is calculated
from the distance metric using a statistical estimation procedure,
such as a Bayesian estimate. Based on the set of calculated
probabilities, the position of each time point of subject data is
assigned to the probability space. The calculation is repeated for
each time point.
[0124] A trajectory or vector is then created by linking adjacent
time points. The distance, direction, speed, and acceleration of
the vector can be calculated and plotted for display or review. If
the user (for example, physician), deems the data to be of
significant importance, the user may activate an alarm or warning
system and accordingly change the course of action for treating or
monitoring the medical condition. After a course of action has been
adjusted, the steps assessing the medical condition of a subject
can be repeated to continue to monitor the trajectory of the
subject.
[0125] FIG. 7 illustrates an exemplary method of the invention for
developing an alarm or early warning system for or upon assessing
the medical condition of a subject. The trajectory of a medical
condition of subject and the information corresponding to the
trajectory (for example, speed or acceleration) is obtained through
a method of multivariate statistical calculations, such as the
method illustrated in FIG. 6. In order to analyze the information
obtained from the mathematical analysis, qualified medical
personnel (for example, a physician) can establish criteria for
activating an alarm or warning system based on the measurement of
biomarker patterns from an individual subject.
[0126] This method allows each individual user to define the value
and set of criteria as they see fit, according to the individual
user's priorities. It is understood that each user or physician may
view each medical condition differently and the example method of
FIG. 7 accounts for these differences.
[0127] An exemplary method of the invention assigns progress lines
to each discrete clinical outcome. The progress lines may represent
different clinical criteria for the medical condition a subject
passes over. Preferably, the progress lines are positioned and
determined by an individual user, such as medical personnel. The
user may assign different locations of the progress lines based
upon the subject history, historical data of the medical condition,
or personal experience. An example of a probability space with
rules demonstrated as progress lines is illustrated in FIG. 8. For
example, if the trajectory of a subject goes within one of the
progress lines illustrated in FIG. 8, an alarm or warning system
may be activated to notify the user or the subject. Other possible
results include, but are not limited to, a system printout of the
medical data, a user interface displaying the information, a
flashing light, an alert system, an alarm, a buzzer, an email, a
telephone communication, and a pager system. The alarm system not
only assesses the current medical condition of the subject, but may
also guide a course of action for the user.
[0128] When one or more of the characteristics of the trajectory
surpasses a specific value as determined by the model user, an
alarm or warning system can be activated. The user can then
determine the next course of action for assessing the medical
condition of the subject.
[0129] In an embodiment of the invention, medical personnel or the
subject can be notified of a need for taking a medical action after
assessing or characterizing the position of the subject data in a
probability space. Examples of a notification include an alert
system, an alarm, a buzzer, an email, a telephone communication,
and a pager system.
[0130] Medical personnel may take medical action when notified by
the methods of the invention that the medical condition of a
subject has violated a rule imposed by the medical personnel.
Medical action includes, but is not limited to, ordering more tests
performed on the patient, altering the dosage of an administered
therapeutic agent, administering a therapeutic agent, terminating
the administration of a therapeutic agent, combining therapies,
administering an alternative therapy, placing the subject on a
dialysis or heart and lung machine, and administering a pain
killer. In some embodiments, the subject may take medical action.
For example, a diabetic subject may administer a dose of
insulin.
[0131] After a medical action has been taken or has been chosen, an
outcome analysis can be performed for characterizing a result of
the selected action. The outcome analysis can lead to automatically
updating the probability of the discrete clinical outcome of the
subject.
[0132] Any clinical outcome, result, or action related to a
particular medical condition may be utilized as a discrete clinical
outcome. For example, discrete clinical outcomes may be generic,
such as responder, non-responder, partial responder, and septic, or
they be more specific such as adverse drug reactions. In addition,
discrete clinical outcomes may be sequellae, or downstream
clinically relevant events or results. Subject data can be obtained
for any pharmacological, pathophysiological or pathopsychological
clinical outcome. The subject data may be obtained during a first
time period before an intervention is administered to the patient,
and also during a second, or more, time period(s) after the
intervention is administered to the patient. The intervention may
comprise a drug(s) and/or a placebo. The intervention may be
suspected to have a propensity to affect the heightened risk of the
onset of the specific medical condition. The intervention may be
suspected of having a propensity to decrease the heightened risk of
the onset of the specific medical condition. The specific medical
condition may be an unwanted side effect. The intervention may
comprise administering a drug, and wherein the drug has a
propensity to increase the risk of the specific medical condition,
the specific medical condition may be an undesired side effect.
[0133] Medical conditions include, but are not limited to,
pharmacological, pathological, physiological and psychological
conditions. For example, abnormality, affliction, ailment, anomaly,
anxiety, cause, disease, disorder, illness, indisposition,
infirmity, malady, problem or sickness, and may include a positive
medical condition for example, fertility, pregnancy and retarded or
reversed male pattern baldness. Specific medical conditions
include, but are not limited to, neurodegenerative disorders,
reproductive disorders, cardiovascular disorders, autoimmune
disorders, inflammatory disorders, cancers, bacterial and viral
infections, diabetes, arthritis and endocrine disorders. Other
diseases include, but are not limited to, lupus, rheumatoid
arthritis, endometriosis, multiple sclerosis, stroke, Alzheimer's
disease, Parkinson's diseases, Huntington's disease, Prion
diseases, amyotrophic lateral sclerosis (ALS), ischaemias,
atherosclerosis, risk of myocardial infarction, hypertension,
pulmonary hypertension, congestive heart failure, thromboses,
diabetes mellitus types I or II, lung cancer, breast cancer, colon
cancer, prostate cancer, ovarian cancer, pancreatic cancer, brain
cancer, solid tumors, melanoma, disorders of lipid metabolism,
HIV/AIDS, hepatitis, including hepatitis A, B and C, thyroid
disease, aberrant aging, and any other disease or disorder.
[0134] Biological Markers
[0135] Biological markers, also referred to herein as biomarkers,
according to the present invention include without limitation
drugs, prodrugs, pharmaceutical agents, drug metabolites,
biomarkers such as expressed proteins and cell markers, antibodies,
serum proteins, cholesterol, polysaccharides, nucleic acids,
biological analytes, biomarker, gene, protein, or hormone, or any
combination thereof. At a molecular level, the biomarkers can be
polypeptide, glycoprotein, polysaccharide, lipid, nucleic acid, and
a combination thereof.
[0136] Of particular interest are biomarkers associated with a
particular disease or with a specific disease stage. Such
biomarkers include but are not limited to those associated with
autoimmune diseases, obesity, hypertension, diabetes, neuronal
and/or muscular degenerative diseases, cardiac diseases, endocrine
disorders, any combinations thereof.
[0137] Also of interest are biomarkers that are present in varying
abundance in one or more of the body tissues including heart,
liver, prostate, lung, kidney, bone marrow, blood, skin, bladder,
brain, muscles, nerves, and selected tissues that are affected by
various disease, such as different types of cancer (malignant or
non-metastatic), autoimmune diseases, inflammatory or degenerative
diseases.
[0138] Also of interest are analytes that are indicative of a
microorganism. Exemplary microorganisms include but are not limited
to bacterium, virus, fungus and protozoa. Other biomarkers obtained
from a subject also include blood-born pathogens selected from a
non-limiting group that consists of Staphylococcus epidermidis,
Escherichia coli, methicillin-resistant Staphylococcus aureus
(MSRA), Staphylococcus aureus, Staphylococcus hominis, Enterococcus
faecalis, Pseudomonas aeruginosa, Staphylococcus capitis,
Staphylococcus warneri, Klebsiella pneumoniae, Haemophilus
influnzae, Staphylococcus simulans, Streptococcus pneumoniae and
Candida albicans.
[0139] Biomarkers also encompass a variety of sexually transmitted
diseases selected from the following: gonorrhea (Neisseria
gorrhoeae), syphilis (Treponena pallidum), clamydia (Clamyda
tracomitis), nongonococcal urethritis (Ureaplasm urealyticum),
yeast infection (Candida albicans), chancroid (Haemophilus
ducreyi), trichomoniasis (Trichomonas vaginalis), genital herpes
(HSV type I & II), HIV I, HIV II and hepatitis A, B, C, G, as
well as hepatitis caused by TTV.
[0140] Biomarkers encompass a diversity of respiratory pathogens
including but not limited to Pseudomonas aeruginosa,
methicillin-resistant Staphlococccus aureus (MSRA), Klebsiella
pneumoniae, Haemophilis influenzae, Staphlococcus aureus,
Stenotrophomonas maltophilia, Haemophilis parainfluenzae,
Escherichia coli, Enterococcus faecalis, Serratia marcescens,
Haemophilis parahaemolyticus, Enterococcus cloacae, Candida
albicans, Moraxiella catarrhalis, Streptococcus pneumoniae,
Citrobacter freundii, Enterococcus faecium, Klebsella oxytoca,
Pseudomonas fluorscens, Neiseria meningitidis, Streptococcus
pyogenes, Pneumocystis carinii, Klebsella pneumoniae Legionella
pneumophila, Mycoplasma pneumoniae, and Mycobacterium
tuberculosis.
[0141] Listed below are additional exemplary markers according to
the present invention: Theophylline, CRP, CKMB, PSA, Myoglobin,
CA125, Progesterone, TxB2,6-keto-PGF-1-alpha, and Theophylline,
Estradiol, Lutenizing hormone, High sensitivity CRP, Triglycerides,
Tryptase, Low density lipoprotein Cholesterol, High density
lipoprotein Cholesterol, Cholesterol, IGFR.
[0142] Exemplary liver markers include, without limitation,
Arginase 1 (liver type), Alpha-fetoprotein (AFP), Alkaline
phosphatase, Alanine aminotransferase (ALT), Lactate dehydrogenase
(LDH), Protein C, and Bilirubin.
[0143] Exemplary kidney markers include without limitation TNFa
Receptor, Cystatin C, Lipocalin-type urinary prostaglandin D,
synthatase (LPGDS), Hepatocyte growth factor receptor, Polycystin
2, Polycystin 1, Fibrocystin, Uromodulin, Alanine, aminopeptidase,
N-acetyl-B-D-glucosaminidase, Albumin, and Retinol-binding protein
(RBP).
[0144] Exemplary heart markers include without limitation Troponin
I (TnI), Troponin T (TnT), CK, CKMB, Myoglobin, Fatty acid binding
protein (FABP), CRP, D-dimer, S-100 protein, BNP, NT-proBNP,
PAPP-A, Myeloperoxidase (MPO), Glycogen phosphorylase isoenzyme BB
(GPBB), Thrombin Activatable Fibrinolysis Inhibitor (TAFI),
Fibrinogen, Ischemia modified albumin (IMA), Cardiotrophin-1, and
MLC-I (Myosin Light Chain-I).
[0145] Exemplary pancreatric markers include without limitation
Insulin, Amylase, Pancreatitis-Assocoated protein (PAP-1), and
Regeneratein proteins (REG).
[0146] Exemplary muscle tissue markers include without limitation
Myostatin.
[0147] Exemplary blood markers include without limitation
Erythopoeitin (EPO).
[0148] Exemplary bone markers include without limitation,
Cross-linked N-telopeptides of bone type I collagen (NTx)
[0149] Carboxyterminal cross-linking telopeptide of bone collagen,
Lysyl-pyridinoline (deoxypyridinoline), Pyridinoline,
Tartrate-resistant acid phosphatase, Procollagen type I C
propeptide, Procollagen type I N propeptide, Osteocalcin (bone
gla-protein), Alkaline phosphatase, Cathepsin K, COMP (Cartillage
Oligimeric Matrix Protein), Osteocrin Osteoprotegerin (OPG), RANKL,
sRANK, TRAP 5 (TRACP 5), Osteoblast Specific Factor 1 (OSF-1,
Pleiotrophin), Soluble cell adhesion molecules, sTfR, sCD4, sCD8,
sCD44, and Osteoblast Specific Factor 2 (OSF-2, Periostin).
[0150] In some embodiments markers according to the present
invention are disease specific. Exemplary cancer markers include
without limitation PSA (total prostate specific antigen),
Creatinine, Prostatic acid phosphatase, PSA complexes,
Prostrate-specific gene-1, CA 12-5, Carcinoembryonic Antigen (CEA),
Alpha feto protein (AFP), hCG (Human chorionic gonadotropin),
Inhibin, CAA Ovarian C1824, CA 27.29, CA 15-3, CAA Breast C1924,
Her-2, Pancreatic, CA 19-9, Carcinoembryonic Antigen, CAA
pancreatic, Neuron-specific enolase, Angiostatin
DcR3 (Soluble decoy receptor 3), Endostatin, Ep-CAM (MK-1), Free
Immunoglobulin Light Chain Kappa, Free Immunoglobulin Light Chain
Lambda, Herstatin, Chromogranin A, Adrenomedullin, Integrin,
Epidermal growth factor receptor, Epidermal growth factor
receptor-Tyrosine kinase, Pro-adrenomedullin N-terminal 20 peptide,
Vascular endothelial growth factor, Vascular endothelial growth
factor receptor, Stem cell factor receptor, c-kit/KDR, KDR, and
Midkine.
[0151] Exemplary infectious disease markers include without
limitation Viremia, Bacteremia, Sepsis, PMN Elastase, PMN
elastase/.alpha.1-PI complex, Surfactant Protein D (SP-D), HBVc
antigen, HBVs antigen, Anti-HBVc, Anti-HIV, T-supressor cell
antigen, T-cell antigen ratio, T-helper cell antigen, Anti-HCV,
Pyrogens, p24 antigen, Muramyl-dipeptide.
[0152] Exemplary diabetes markers include without limitation
C-Peptide, Hemoglobin A1c, Glycated albumin, Advanced glycosylation
end products (AGEs), 1,5-anhydroglucitol, Gastric Inhibitory
Polypeptide, Insulin, Glucose, Hemoglobin, ANGPTL3 and 4.
[0153] Exemplary inflammation markers include without limitation
Rheumatoid factor (RF), Tumor Necrosis factor-.alpha.
(TNF-.alpha.), Antinuclear Antibody (ANA), C-reactive protein
(CRP), Clara Cell Protein (Uteroglobin).
[0154] Exemplary allergy markers include without limitation Total
IgE and Specific IgE.
[0155] Exemplary autism markers include without limitation
Ceruloplasmin, Metalothioneine, Zinc, Copper, B6, B12, Glutathione,
Alkaline phosphatase, and Activation of apo-alkaline
phosphatase.
[0156] Exemplary coagulation disorders markers include without
limitation b-Thromboglobulin, Platelet factor 4, Von Willebrand
factor.
[0157] In some embodiments a marker may be therapy specific. COX
inhibitors include without limitation TxB2 (Cox-1),
6-keto-PGF-1-alpha (Cox 2), 11-Dehydro-TxB-1a (Cox-1).
[0158] Other markers of the present include without limitation
Leptin, Leptin receptor, and Procalcitonin, Brain S100 protein,
Substance P, 8-Iso-PGF-2a.
[0159] Exemplary geriatric markers include without limitation,
Neuron-specific enolase, GFAP, and S100B.
[0160] Exemplary markers of nutritional status include without
limitation Prealbumin, Albumin, Retinol-binding protein (RBP),
Transferrin, Acylation-Stimulating Protein (ASP), Adiponectin,
Agouti-Related Protein (AgRP), Angiopoietin-like Protein 4
(ANGPTL4, FIAF), C-peptide, AFABP (Adipocyte Fatty Acid Binding
Protein, FABP4)
Acylation-Stimulating Protein (ASP), EFABP (Epidermal Fatty Acid
Binding Protein, FABP5), Glicentin, Glucagon, Glucagon-Like
Peptide-1, Glucagon-Like Peptide-2, Ghrelin, Insulin, Leptin,
Leptin Receptor, PYY, RELMs, Resistin, amd sTfR (soluble
Transferrin Receptor).
[0161] Exemplary markers of lipid metabolism include without
limitation Apo-lipoproteins (several), Apo-A1, Apo-B, Apo-C-CII,
Apo-D, Apo-E.
[0162] Exemplary coagulation status markers include without
limitation Factor I: Fibrinogen, Factor II: Prothrombin, Factor
III: Tissue factor, Factor IV: Calcium, Factor V: Proaccelerin,
Factor VI, Factor VII: Proconvertin, Factor VIII: Anti-hemolytic
factor, Factor IX: Christmas factor, Factor X: Stuart-Prower
factor, Factor XI: Plasma thromboplastin antecedent, Factor XII:
Hageman factor, Factor XIII: Fibrin-stabilizing factor,
Prekallikrein, High-molecular-weight kininogen, Protein C, Protein
S, D-dimer, Tissue plasminogen activator, Plasminogen,
a2-Antiplasmin, Plasminogen activator inhibitor 1 (PAI1).
[0163] Exemplary monoclonal antibodies include those for EGFR,
ErbB2, and IGF1R.
[0164] Exemplary tyrosine kinase inhibitors include without
limitation Ab1, Kit, PDGFR, Src, ErbB2, ErbB 4, EGFR, EphB,
VEGFR1-4, PDGFRb, FLt3, FGFR, PKC, Met, Tie2, RAF, and TrkA.
[0165] Exemplary Serine/Threoline Kinase Inhibitors include without
limitation AKT, Aurora A/B/B, CDK, CDK (pan), CDK1-2, VEGFR2,
PDGFRb, CDK4/6, MEK1-2, mTOR, and PKC-beta.
[0166] GPCR targets include without limitation Histamine Receptors,
Serotonin Receptors, Angiotensin Receptors, Adrenoreceptors,
Muscarinic Acetylcholine Receptors, GnRH Receptors, Dopamine
Receptors, Prostaglandin Receptors, and ADP Receptors.
[0167] In a separate embodiment, a method herein utilizes
pharmacological parameters useful for assessing efficacy and/or
toxicity of a therapeutic agent and the agent's affect on a medical
condition. For the purposes of this invention, a "therapeutic
agent" is intended to include any substances that have therapeutic
utility and/or potential. Such substances include but are not
limited to biological or chemical compounds such as simple or
complex organic or inorganic molecules, peptides, proteins (for
example, antibodies) or a polynucleotides (for example,
anti-sense). A vast array of compounds can be synthesized, for
example, polymers, such as polypeptides and polynucleotides, and
synthetic organic compounds based on various core structures, and
these are also included in the term "therapeutic agent". In
addition, various natural sources can provide compounds for
screening, such as plant or animal extracts, and the like. It
should be understood, although not always explicitly stated that
the agent is used alone or in combination with another agent,
having the same or different biological activity as the agents
identified by the inventive screen. The agents and methods also are
intended to be combined with other therapies.
[0168] Implementation of the Methods
[0169] It is to be understood that the exemplary methods and
systems described herein may be implemented in various forms of
hardware, software, firmware, special purpose processors, or a
combination thereof Methods herein can be implemented in software
as an application program tangibly embodied on one or more program
storage devices. The application program may be executed by any
machine, device, or platform comprising suitable architecture. It
is to be further understood that, because some of the systems and
methods depicted in the Figures are preferably implemented in
software, the actual connections between the system components (or
the process steps) may differ depending upon the manner in which
the present invention is programmed. Given the teachings herein,
one of ordinary skill in the related art will be able to
contemplate or practice these and similar implementations or
configurations of the present invention.
[0170] FIG. 9 illustrates the process flow of building the model
and system for assessing the medical condition of a subject. The
subject inputs personal data into a database. The same or a
different database contains data from other subjects with a similar
medical condition. Data from the other subjects can be historical
data from public or private institutions. Data from other subjects
may also be internal data from a clinical study.
[0171] The subject may input data in a variety of ways, or using a
variety of devices. For example, a point-of-care microfluidic
device may be used to acquire biomarker data at the point-of-care.
The data from the device may then be communicated directly from the
device to a database or processor. A computer or data entry system
may also input subject data to a database or processor. Data may be
automatically obtained and input into a computer from another
computer or data entry system. Another method of inputting data to
a database is using an input device such as a keyboard, touch
screen, trackball, or a mouse for directly entering data into a
database.
[0172] A database is developed for a medical condition in which
relevant clinical information is filtered or mined over a
communication network (for example, the Internet) from one or more
data sources, such as a public remote database, an internal remote
database, and a local database. A public database can include
online sources of free clinical data for use by the general public,
such as, for example, databases supplied by the U.S. Department of
Health and Human Services. For example, an internal database can be
a private internal database belonging to particular hospital, or a
SMS (Shared Medical System) for providing clinical data. A local
database can comprise, for example, biomarker data relating to
discrete clinical outcome. The local database may include data from
a clinical trial. It may also include clinical data such as
temperature and laboratory information, EKG results, blood test
results, patient survey responses, or other items from patients in
a hospital or an internal hospital monitoring system. The database
can also include historical reference data of a plurality of
subject members in relationship to at least one biological
marker.
[0173] The database may also be implemented on a variety of
commercially available authoring packages.
[0174] The database may be of a storage unit of a medical
information system.
[0175] Data from a database can be filtered and classified
according to specific cases or medical conditions or a group of
diagnoses and conditions. For example, the classification within
the database may follow the standard international code of
diagnoses (ICD-9 coding system). A medical condition may include,
for example, a physical state of a subject or a disease that the
subject is suffering from.
[0176] In an embodiment, the data stored in the database may be
selected from the categories consisting of: pathology, anatomy,
treatment option, treatment outcome, pharmacological parameter,
pharmacokinetics parameter, psychological parameter, and genomic
information.
[0177] Subject data can be stored with a unique identifier for
recognition by a processor or a user. In another step, the
processor or user can conduct a search of stored data by selecting
at least one criterion for particular patient data. The particular
patient data can then be retrieved.
[0178] Relationships can be established between elements of the
public databases, the internal databases and the subject data.
Explicit relationships declare and record inferences made based on
information revealed by the trajectory or other available
observations.
[0179] FIG. 9 illustrates the flow of data from a database that
includes the data from the subject to a processor that performs the
mathematical methods described herein to construct a probability
space. The subject data may also be inputted to the processor
separately from the data pertaining to a discrete clinical outcome
that is stored in a database.
[0180] A processor can perform calculating the probability space.
Once a probability space map is prepared, the same processor or a
different processor can plot biomarker data from an individual
subject within the probability space. The probability space map
itself can be transmitted to an output, such as a display monitor.
The position of the subject data within the probability space may
also be transmitted to an output. The processor may have a means
for receiving patient data directly from an input device, a means
of storing the subject data in a storage unit, and a means for
processing data. The processor may also include a means for
receiving instructions a user or a user interface. The processor
may have memory, such as random access memory, as is well known in
the art. In one embodiment, an output that is in communication with
the processor is provided.
[0181] A user interface refers to graphical, textual, or auditory
information presented to a user. User interface may also refer to
the control sequences used for controlling a program or device,
such as keystrokes, movements, or selections.
[0182] A user interface may be a commercial software application
for displaying subject data and/or the position of the subject data
in a probability space. The user interface generally provides users
with access to a task-specific set of functions to view and modify
content. In preferable embodiments of the invention, the user
interface is a graphical user interface (GUI). In some embodiments,
a constructed probability space in two dimensions can be viewed
using a GUI. Also, the trajectory of subject data positioned within
the probability space can be viewed using a GUI. The GUI may be
displayed on a display monitor. Other user interfaces that may be
utilized with the methods of the invention include, but are not
limited to, alert systems, web-based interfaces, text interfaces,
sound interfaces, batch interfaces, and intelligent interfaces. In
an embodiment of the invention, the user interface can remotely
communicate with the system.
[0183] By acquiring data from a subject at multiple points in time,
a trajectory of the subject data can be processed. For each time
point, a processor may calculate the position of the subject data
time point in a probability space. A trajectory is created by
connecting at least two time points of subject data. Different
vector parameters of the trajectory can be calculated such as
distance, speed, direction, and acceleration. The vector parameters
can be calculated by a processor. The processor may be the same as
or different than the processor used to construct the probability
space.
[0184] Based upon rules imposed by a user, the position or
trajectory of subject data may activate an alert. The alert or
alert system notifies the user when the subject data conflicts with
a rule imposed by a user. For example, a rule may comprise the
probability of a subject of developing an adverse drug reaction
(ADR) is greater than 60 percent. When the probability of the
subject data exceeds the rule, an alert is communicated to a
user.
[0185] Examples of an alert include, but are not limited to, a
sound, a light, a printout, a readout, a display, a page, an
e-mail, a fax alert, telephonic communication, or a combination
thereof. The alert may communicate to the user the raw subject
data, the calculated position in probability space of the subject
data, the trajectory of the subject data, or a combination
thereof.
[0186] A user is most preferably medical personnel. In an
embodiment, the same user has access to the subject data,
establishes the rules for alerting the user, and is alerted by the
alert. The user may also be the subject. For example, a diabetic
patient may monitor his personal glucose levels or other biomarker
levels. In another embodiment, a user has access to the subject
data at any step of the mathematical model.
[0187] FIG. 10 demonstrates a networking method of assessing the
medical condition of a subject. A system of communicating
information may or may not include a reader for reading subject
data. For example, if biomarker data is acquired by a microfluidic
point-of-care device, the values assigned to different individual
biomarkers may be read by the device itself or a separate device.
Another example of a reader would be a system to scan in subject
data that has been entered in an electronic medical record or a
physician chart. A further example of a reader would consist of an
electronic patient record database from which subject data could be
directly obtained via the communications network.
[0188] A user station may be incorporated into the methods and
systems herein. The user station can act as a reader for reading
subject data or as a data entry system. The user station may also
be used to communicate data information directly to a user, such as
medical personnel or the subject. The user station platform may be
a computer or one of a wide variety of hardware platforms that runs
any operating system. The platform may read and write to and from a
storage unit. A storage unit may be incorporated within the user
station. As will be readily apparent to those of ordinary skill in
the art, other hardware may be used to execute the methods of the
invention.
[0189] According to still further embodiments, a server may be
provided for use in assessing a medical condition of a subject. In
some embodiments, a reader and/or user station communicates with a
server. The server may comprise a network interface and a
processor. A storage unit may also be provided in the server. The
storage unit is coupled to receive the patient data from the input
device, reader, or user interface, and to store the patient
data.
[0190] The storage unit may be any device capable of storing data
for a long period of time, either as a component of the platform,
or as a separate entity that is coupled to the network. Although
any appropriate storage capacity may be used, the capacity is
typically at least ten megabytes and more usually at least twenty
gigabytes. The storage unit may be permanent or removable. For
example, the storage unit may be as magnetic random access, such as
flexible or hard disk drive, magnetic sequential access, such as
tape, optical random access such as CD, DVD or Magnetic Optical
drive, and solid state memory, such as EPROM, MRAM and Flash. In
addition, the present invention anticipates other storage devices.
All of the aforementioned storage units are by way of example and
are not intended to limit the choices that are or may become
available in the art.
[0191] In some embodiments, the data are further manipulated prior
to storage by the processor compressing the data. Some data are
very large and compression permits conservation of storage space.
In the absence of compression, this large amount of data may
present a particular problem for the storage of medical
information.
[0192] In other embodiments of a system for assessing the medical
condition of a subject, data storage, processing and control of
data are implemented on a server that is in communication with a
user station. Optionally, computations with the data may be
performed on the server as well. In this client-server embodiment,
the server is communicatively coupled to one or more user stations
by local area network (LAN), implemented with Ethernet, Wi-Fi,
Bluetooth, USB or Firewire, a Wide Area Network (WAN) such as the
Internet, implemented by broadband cable, xDSL, TI, metropolitan
Wi-Fi or other high speed communications down to
plain-old-telephone-service (POTS).
[0193] A user station is in communication with a server to output
the information processed by the server. The user station may be
the same or different from the user station or device used to input
subject data into the system. The user station may also be an alert
or a warning system. Examples of a user station include, but are
not limited to, a computer, a PDA, a wireless telephone, and a
personal device, such as an iPod. The user station may be in
communication with a printer or a display monitor to output the
information processed by the server.
[0194] In a networked system comprising a server, any number
devices may be members of a medical network enterprise and access a
central collection of medical data stored in a server. A given user
station may have access to enormous amounts of patient data from
healthcare sources anywhere in the world. In one application, data
from multiple practitioners are accumulated for defined clinical
trial studies.
[0195] Statistical information is provided to the user of the
system, preferably at the point-of-care. The point-of-care is where
treatment is implemented. Advantageously, the dynamic and real-time
characteristics of Internet usage are captured in this feature.
This enables medical personnel or the subject either in the office
or in the field to instantaneously be notified of and presented
with, for example, a subject's medical condition and statistical
information pertaining to the subject data.
[0196] The user station, alert system, or output device may encrypt
the data prior to data transmission. The data can also be encrypted
by the processor to protect private information. A variety of
encryption schemes may be used, such as public key encryption or
private key encryption. The encryption components may be
stand-alone components or software components. The server may also
include a decryption unit for decrypting received subject data
prior to storage.
[0197] In an alternative embodiment, if the communication bandwidth
between the user and the server becomes restrictive (for example,
if the server is "down"), the model can be duplicated and reside in
a mirror server at a user station. Updates and communications
between the server and the mirror server can be done off-line at
predetermined times. On-line requests from the user can be handled
locally by the mirror server. This alternative provides increased
reliability for users, since the on-line processes do not depend on
outside networks.
[0198] In an alternative configuration of the data analysis system,
client-based architecture is used where storage, data processing
and computations are performed on a user station. An input device
transmits data from a source to a platform. The platform is coupled
to a display through a display controller. The display may be any
one of a number of conventional display devices, such as a liquid
crystal display or video display. The platform contains a storage
unit, a processor and an interface for communicating with the input
device. The storage unit has a database for storing, manipulating
and retrieving subject data. The storage unit also may store and
run an operating system. The user station platform may be a
computer or one of a wide variety of hardware platforms that runs
any operating system. The platform may read and write to and from
its storage unit. As will be readily apparent to those of ordinary
skill in the art, other hardware may be used to execute the
software.
[0199] In another aspect, computer readable instructions are
provided, which when executed cause a processor to: provide a
probability space defined by a set of discrete clinical outcomes,
each of which is characterized by a statistical distribution of at
least one biological marker; obtain subject data corresponding to
the at least one biological marker; and calculate the position of
said subject data in said probability space to assess the
probability of a medical condition of said subject.
[0200] The computer readable instructions may be a part of a
software package. The software package can be provided with a
server of the invention or an input device, such as a microfluidic
point-of-care device.
[0201] In an embodiment, computer readable instructions comprise
instructions for creating a trajectory between two time points of
subject data. The instructions may also cause a processor to
calculate vector parameter of a trajectory, such as speed,
position, heading, and acceleration.
[0202] Computer readable instructions may direct a processor or
output device to output the assessment of a medical condition. The
instructions, when executed, may display a graphical representation
on a monitor, print an output on a printer, or activate an alert or
alarm.
[0203] In another aspect, computer readable instructions may be
stored in a computer readable medium. In some embodiments, the
computer readable medium that stores instructions is a storage unit
as described herein. Still other embodiments may provide a computer
readable medium having stored therein a plurality of sequences of
instructions, which, when executed by a processor, execute a method
of the invention.
[0204] The computer readable instructions when executed can cause a
processor to provide a user defined alert condition based on an
assessment of trajectory parameters of the subject data in the
probability space. In an embodiment, the trajectory parameters are
at least one of speed, acceleration, direction, and position.
[0205] As used herein, a "computer readable medium" refers to any
medium which can be read and accessed directly by a computer or a
processor. Featured computer readable media include, but are not
limited to, magnetic storage medium, optical storage medium,
electrical storage medium, and hybrid storage medium of any of
these categories.
[0206] The computer readable instructions can operate in a software
runtime environment of the processor. In an embodiment, a software
runtime environment provides commonly used functions and facilities
required by the software package. Examples of a software runtime
environment include, but are not limited to, computer operating
systems, virtual machines or distributed operating systems. As will
be appreciated by those of ordinary skill in the art, several other
examples of runtime environment exist.
[0207] The computer readable medium may be a storage unit of the
present invention as described herein. It is appreciated by those
skilled in the art that computer readable medium can also be any
available media that can be accessed by a server, a processor, or a
computer.
[0208] Computer readable media includes volatile and non-volatile,
removable and non-removable media implemented in any method or
technology for storage of information such as computer-readable
instructions, data structures, program modules, or other data.
Computer storage media includes, but is not limited to, RAM, ROM,
EPROM, EEPROM, flash memory or other solid state memory technology,
CD-ROM, DVD, or other optical storage, magnetic cassettes, magnetic
tape, magnetic disk storage or other magnetic storage devices, or
any other medium which can be used to store the desired information
and which can be accessed by the computer. A skilled artisan can
readily adopt any of the presently known methods for recording
information on a computer readable medium.
[0209] The computer readable medium can be incorporated as part of
the computer-based system of the present invention, and can be
employed for a computer-based assessment of a medical
condition.
[0210] Healthcare Operating System
[0211] Different aspects of the invention can be used and developed
for healthcare systems. Examples of such systems include
point-of-care systems, a healthcare operating system or a
combination of both. For example, a user may test a blood sample
with a device and the information from the test can be provided to
a system or method of the invention, which in turn sends additional
information to the user.
[0212] In an embodiment, a healthcare system is part of an
integrated infrastructure built around real-time, point-of-care
disposable blood monitoring devices which analyze about 10
microliter samples in a similar manner to automated replication of
the manual processes run by a central laboratory. In an integrated
infrastructure, information can be wirelessly transmit from an
assay device to a database which integrates data from the device
with stored data from disparate databases (patient record, genetic
or genomic information, data from clinical trials) into a central
database. The system can then allow for the automatic application
of mathematics to the database in the context of the
pathophysiology of a given disorder. For example, a healthcare
system can be used to rapidly improve the label of key drugs
through adaptive clinical studies which can generate publications
for label expansions for new indications, patient subpopulations,
and for ameliorating safety concerns.
[0213] In an embodiment of the invention, a healthcare system can
be utilized for home, real-time blood monitoring has significant
implications which allow us to collect information which cannot be
seen using the conventional blood testing infrastructure.
[0214] FIG. 11 illustrates an exemplary embodiment of a system of
the invention. Real-time acquisition of subject data can be
accomplished using a point-of-care device. The point-of-care device
can acquire, store, and/or manage information based upon a specific
disease or compound. The point-of-care device may also be in
communication with the user or a computer or system belonging to
the user. A point-of-care environment in this example can be the
home or the clinic. Also illustrated in FIG. 11 is transmitting
information in real-time from the point-of-care device to a server
or system that is capable of converting the information or data to
a physiologically relevant context. This can be accomplished by
transferring the information to a database or a plurality of
databases. The databases can be on a server or a plurality of
servers. In an embodiment, a centralized database repository is
used to analyze the data and return the data to an end user in a
physiologically relevant context. The data can be returned and, in
the example of FIG. 11, information, such as clinically relevant
and actionable information, can be returned to a user through
browser-based software applications. The information could be used
for adaptive clinical trials and publications if the end user is a
medical researcher or pharmaceutical company.
[0215] A system of the invention comprising a point-of-care device
in communication with a healthcare system may not require the
transfer of samples, such as transportation to a laboratory, which
in turn, can improve the integrity of a sample. Detection of an
analyte in a system or method of the invention can be provided in
real-time at the point-of-care. Real-time detection integrated with
a healthcare central database system can allow for the creation of
complex longitudinal data series acquired by an immunoassay
device.
[0216] FIG. 12 demonstrates an exemplary system of the invention,
wherein a health care operating system comprises data
infrastructure, models and algorithms, and software applications.
The data infrastructure can be used by a pharmaceutical company for
clinical trials across an entire pharmaceutical pipeline. A model
or algorithm of the health care operating system can comprise, for
example, predictive and dynamic, multivariate, multi-dimensional
models that can be customized for program-specific objectives and
that can map disease progression and regression. A system can be
integrated with back-end algorithms, models, and data in the data
infrastructure. For example, a system can transmit individualized
content to users on device touch-screens or mobile phones. The
feedback may assist with behavior modification and increase
compliance with therapy. The algorithms described herein enable
correlation of blood data to efficacy dynamics profiles, behavior,
lifestyle, diet, and side-effects.
[0217] Content can be based on data for patient classes, which
recognize physiological and psychological pre-dispositions as well
as local socio-environmental influences. In another embodiment, a
system can link users through social networks, where success
stories compound through the combination of each tailored home
health system with a given therapy.
[0218] The health care operating system can interact at the
point-of-care with consumer systems, such as an assay cartridge, an
assay reading device, or a mobile device such as a cell phone. The
field units can be integrated with point-of-care home and mobile
monitoring systems. Field units can be remote, portable patient
care systems and can provide on-site, real-time, automatic
processing of cartridges for blood analysis. In some instances,
field units comprise a user interface, allowing patient to initiate
assays and graphically enter a variety of relevant information,
such as: patient diaries, environmental, behavioral, and
psychological information, and two-way communication system from
the instruments to medical personnel or mobile phones and back to
patients with relevant content, messages, and health information.
In some instances, blood and environmental data is automatically
(for example, wirelessly) transmitted into models in real time.
[0219] As described herein, a health care operating system can
comprise information integration and exploitation infrastructure
that permits, for example without limitation: data acquisition and
storage of point-of-care results in real time, integration of blood
parameters and patient diary data with all other physiologically
relevant information, and a central mathematical software program.
The central mathematical software program can graphically
visualize, help to interpret, and analyze all data in one place and
link any new information into a disease management system that then
maps the information onto a probability space of clinical outcomes.
A health care operating system or a field unit can provide a
graphical display of clinically relevant and actionable information
back to the health care provider and/or the patient or user.
[0220] Possible applications for a healthcare system of the
invention include, but are not limited to, an integrated link to an
adaptive clinical trial or healthcare management system, drug and
systems combinations for post-prescription monitoring, a healthcare
operating system (for example, a server controlled point of care
whole blood monitoring systems), and identification of correlations
to efficacy of key compounds during clinical studies.
[0221] In an embodiment, a system can provide a graphical portrayal
of integration of available data sets to a user.
[0222] In another embodiment, a system of the invention can
generate data that may serve as historical data in future studies.
A system can be customized according a the needs of a user, for
example, extracting relevant information from a pharmaceutical
company's existing databases, a central mathematical software
program which allows a pharmaceutical company to visually see,
interpret, and analyze all of their data in a place and is linked
to a clinical trial system.
[0223] When utilized with clinical trials of pharmaceuticals, a
system of the invention can provide an understanding of compound
efficacy, disease progression and patient response that may not be
possible using the conventional blood testing infrastructure.
[0224] A database and a healthcare system can be customized to
monitor trends protein assays over time and their relationship to
disease progression across different indications while accounting
for the other relevant variables or biomarkers in the
pathophysiology of the disease. The customization can be designed
to automatically incorporate all relevant data for use in mapping
disease progression across relevant indications and are the
foundation of historical data. The maps and baselines can be used
in designing future studies and in future studies to extract better
information from blood test results and provide early reads on
efficacy across different indications.
[0225] FIG. 13 demonstrates an exemplary database comprising an
ontology of biomarkers that are related some types of medical
conditions. For example, sepsis as a disease is categorized as a
pathophysiology and is linked in this ontology to Acute Liver
Failure (ALF), Disseminated Intravascular Coagulation (DIC) and the
like. Each of these pathologies can be considered a sequellae of
sepsis in the context of this ontology. In this example, a drug
used to treat sepsis is CytoFab. In some instances, the sequellae
can be characterized by the spectra of biomarkers as described in
FIG. 13. For example, ALF is characterized by specific patterns of
ALT, AST, GGT, and Alkaline phosphatase.
[0226] In an aspect of the invention, the methods, devices and
systems disclosed herein can comprise a healthcare operating
system. A healthcare operating system can comprise a central
control system for delivery of care to consumer and for an economic
model for payors and payees.
[0227] In an embodiment, a healthcare operating system can provide
an early system possibly to detect the onset of disease. A system
can further comprise a learning engine that ties to the preferences
of a user or an insurance provider based on drug availability,
targeted response rate, or payment structure.
[0228] A computer system or server of the invention can include a
dynamic learning engine, for example, as the system or server
receives data, the operations of the system or server can be
updated. A system or server can also call in to siloed databases,
extracts, and locally store (permanently or temporarily)
information in a central data repository.
[0229] In another embodiment, a healthcare system can comprise a
communications portal for a user to communicate an array of
information to the system, such as weight, nicotine consumption,
food consumption, exercise, and physchiatric variables.
[0230] FIG. 14 illustrates an example hierarchy of a healthcare
operating system of the invention. The central operating system has
access to a data infrastructure and monitoring devices for
obtaining subject data. As demonstrated in the example in FIG. 14,
the data infrastructure can comprise a search tool, communication
links, analysis of data, and applications. For example, a
communications link can communicate to a consumer, clinician, or
HMO. A clinician can use the information for simplification and
effectiveness of his decision making or as a data/knowledge base
background. An HMO could for example receive information of a
clinical trial, such as drug efficacy and the costs of the drug
use. Analysis of information in the data infrastructure as
illustrated in FIG. 14 can include generation of a trajectory,
fuzzy logic guidance, collaborative filtering prediction, disease
maps, and data reports for clinical trials or new drug discovery.
Examples of applications of a data infrastructure include reports
(public or private), alerts (for example alerts of a user,
clinician, clinical trial, or a pharmacy), and dynamic, adjustable
trials.
[0231] Also demonstrated in FIG. 14 is the hierarchy of information
received by the exemplary operating system from monitoring devices.
The devices and operating system can monitor and release drugs if
necessary. In a preferable embodiment, the devices monitor analytes
in a bodily fluid. For example, when monitoring blood, the devices
may be capable of executing prescription blood tests for a clinical
or preventative setting or evaluate drug systems as
demonstrated.
[0232] In an aspect, a system is disclosed that solicits input from
the user or his agent (for example, a physician) to facilitate a
conversation between the subject and his agent about the status of
the person's health at a time. Examples of input that can be
solicited by a system include, but are not limited to, establishing
alert levels, medical notation, and personal notation. The system
can acquire clinically relevant information about a person and
place the information in a physiologically-reasonable context.
Examples of clinically relevant information include, but are not
limited to, age, gender, race, medical history, and genomic
information.
[0233] In an embodiment, time dependent data can be representative
of overall clinical health or response, including a diagnosed
disease condition that requires monitoring, a response profile for
a particular drug or combination of drugs, or subject observed
observations of his health.
[0234] Calculated or observed information can be provided to a
user. Examples of users include, but are not limited to, a patient,
a physician, a clinical monitor for a clinical trial, and other
medical personnel.
[0235] Information can also be integrated into a central repository
for inclusion into, for example, a medical record or a clinical
trial report.
[0236] FIG. 15 demonstrates an exemplary system of the invention
wherein an analysis database can receive information from a
database containing data from a reader or cartridge database,
historical databases, a patient database that can contain data
entered by a patient, and a customer database. Other features of a
system of the invention are demonstrated in FIG. 15. The system is
capable of performing analysis on data received by the system. The
analysis can be any method as described herein and can include
developing or utilizing a disease map, using statistics for a
statistically analysis of information within the system, reports or
graphs, and/or a combination of the different methods as
described.
[0237] FIG. 15 also illustrates a user interface that may be
utilized in a system of the invention. The user interface can
interact with any of the components demonstrated in FIG. 15. In a
preferable embodiment, the user interface allows a user to
interface with the system to determine test data or analysis
outcomes, such as physiologically relevant information. The user
interface may be a touch screen, monitor, keyboard, mouse, or any
other item that allows a user to interact with a system of the
invention as would be obvious to one skilled in the art.
[0238] As shown in FIG. 15, a system can be utilized to manage a
reader or point-of-care device. This can include management of a
cartridge to be inserted into the reader for measurement of data.
For example, the system can provide the reader a protocol for
running an immunoassay of the cartridge. A system of the invention
can also provide information to the reader for calibration of the
immunoassays run on the cartridge that are run or executed by the
reader. Reader management can also include information for
assigning the reader for the appropriate reading system for reading
an immunoassay from a point-of-care cartridge. Reader management
can also include corrective and preventive actions (CAPA), which is
an instrument of integrated and comprehensive compliance
management. An element of the CAPA approach is an effective and
systematic processing of quality deficiencies, errors and
malfunctions with the goal to provide appropriate corrective
actions and consistently prevent a reoccurrence of non-conformance
situations. The CAPA strategy was developed by the Food and Drug
Administration (FDA) for the inspection of medical products.
[0239] Systems and methods described herein can be used for
pharmaceutical clinical trials. In some instances, the systems and
methods accelerate trial timelines by an average of 18 months. In
some instances, a system or method herein may provide dose-response
and efficacy dynamics profiles in less than 18 months, less than 15
months, less than 12 months, less than 10 months, less than 9
months or even less than 6 months. Real-time pharmacokinetics and
pharmacodynamics may provide better efficacy and more safety
information. The methods herein may also decrease the risk of
toxicity slowing approval or restricting use by shifting dose
across: sub-populations, concomitant medications, and multiple
indications. In other instances, a system or method herein can
improve predictive visibility into pathway dynamics. A model as
describe can produce predictive insights and power strategic
decisions. A system can support communications with regulatory
authorities internationally and can facilitate `learn and confirm`
strategy for adaptive clinical trials.
[0240] A system herein can include a data management system. The
data management system can be operated through a plurality of
devices or systems, or by a single central server. An example of
data management of a system of the invention is shown in FIG. 15. A
data management system can receive input or output from, for
example, a point-of-care reader, a database (such as an analysis
database or a historical database), and a terminal (such as a user
interface). A data management system can also conduct storage
management, for example, storage of a patient's individual data or
storage of information related to a particular disease or drug
profile. For example, if using a system and/or method of the
present invention to conduct a clinical trial on an experimental
drug, the data management and storage management systems can
receive data from the point-of-care from a reader, and then store
the information in databases for analysis of the data. Both the
patient data and the analysis data may then be stored for future
use of a system. For example, if one side effect is noted in a
patient in Phase I of a clinical trial, that side effect
information may be used for the analysis of results of a patient in
Phase III of the clinical trial. As another example, a diabetic
individual may receive a glucose measurement at an earlier time,
which may then be used to evaluate the condition of the individual
at a later time following another measurement.
Example 1
Prostate Cancer and Anti-Angiogenesis Treatment
[0241] A Phase I or Phase Ha clinical trial for treating a subject
diagnosed with prostate cancer with an anti-angiogenic drug allows
for dose adjustments during the trial. The subject is monitored
over a course of time during the clinical study. The method of
action of the anti-angiogenic drug is to halt the growth of blood
vessels in and around the tumor, thereby depriving the tumor of
adequate blood supply. This may induce ischemia and nutritional
necrosis in the tumor, thereby debulking the tumor.
[0242] After the method of action is determined, discrete clinical
outcomes are chosen by medical personnel, such as a physician or
clinical monitor. The discrete clinical outcomes chosen in this
example are complete response (CR), partial response (PR),
non-response (NR), and adverse drug reaction (ADR). The ADR related
to the anti-angiogenic drug is hypertensive encephalopathy in this
example.
[0243] Six biomarkers representative of the clinical outcome of the
subject with prostate cancer are chosen to define the discrete
clinical outcomes. The biomarkers are PSA, TPS, hK2, VCAM-1,
Endothelin-1, and 6-keto-PGF-1-.alpha.. Using these biomarkers and
the corresponding historical data on the progression of prostate
cancer treatments with anti-angiogenic drugs as well as data from
other subjects in the clinical trial, a cluster analysis is
performed. Each discrete clinical outcome is assigned to a centroid
of one of the clusters.
[0244] Before the subject is treated with the drug, the subject is
assigned a position in the probability space based upon the
subject's data from the relevant biomarkers. The probability of the
subject belonging any of the discrete clinical outcome classes is
determined using the Mahalanobis distance, from which a probability
is calculated using a Bayesian estimation. The result of this
calculation returns the four probabilities corresponding to the
four clinical outcomes of the example. The position of the subject
is determined by plotting the four probabilities in the probability
space.
[0245] The probability space corresponding to four clinical
outcomes is graphically displayed as a square in two dimensions.
The user of the model, in this example, medical personnel,
establishes rules for the medical condition of the subject based
the subject data's position and/or trajectory in the probability
space. Each discrete clinical outcome has a line representing the
rule established by the medical personnel, as shown by the dashed
lines in FIG. 16. When the subject crosses the dotted line towards
the CR or PR conditions, the subject is substantially free of ADR
and has improved his/her prognosis.
[0246] In FIG. 16, the trajectories of two different subjects are
plotted. Subject 1 has a trajectory that starts in the middle of
the probability space and then heads towards the ADR condition. In
this example, the medical personnel may notice the trajectory's
speed or acceleration and intervene in the treatment. When Subject
1's trajectory crosses the dotted line towards the ADR condition,
an alarm is activated. The medical personnel react accordingly.
[0247] Subject 2 has a trajectory that heads towards the NR
condition. The dose of the anti-angiogenic drug treatment is
adjusted corresponding to the rules established by the medical
personnel. After the adjustment of the dose, the trajectory heading
changes towards the PR condition. The example illustrates a method
and embodiment of the invention for monitoring a clinical
trial.
Example 2
Sepsis and Anti-TNF Treatment
[0248] Subjects with confirmed or suspected sepsis or systemic
inflammatory response syndrome (SIRS) are treated with an anti-TNF
therapeutic during a Phase IIb (dose adjustment allowed) or a Phase
III (no dose adjustment allowed) clinical trial. The subject is
monitored over a course of time during the clinical study. The
method of action of the anti-TNF drug is anti-inflammatory. The
drug is aimed at preventing or circumscribing downstream sequellae
of sepsis. Typically, the anti-TNF drug is given in conjunction
with a broad spectrum antibiotic.
[0249] After the method of action is determined, discrete clinical
outcomes are chosen by a physician. The discrete clinical outcomes
chosen in this example are response (R), non-response (NR), and
four sequellae (disseminated intravascular coagulation (DIC), acute
respiratory distress syndrome (ARDS), acute liver failure (ALF),
and acute renal failure (ARF))
[0250] Twenty biomarkers representative of the medical condition of
the subject with confirmed or suspected sepsis are chosen to define
the discrete clinical outcomes. In this example, the biomarkers
fall into multiple categories to monitor the different discrete
clinical outcomes as defined by the physician. Lipoprotein Binding
Protein (LBP) biomarkers measure infectious growth and the R and NR
conditions. The LBP biomarkers are LBP, TNF, CRP, IL-6, and IL-8.
Biomarkers that are relevant to the DIC condition are von
Willebrand's factor, Protein C, thrombin, procalcitonin, neopterin,
and PAF-1. The biomarkers of the ARDS condition are von
Willebrand's factor, IL-6, IL-8, IL-1, and TNF. The biomarkers of
the ALF condition are ALT, AST, GGT, LDH, Alkaline phophatase,
bilirubin, and Protein C. The biomarkers of the ARF condition are
Creatinine and Cystin C.
[0251] Using the biomarkers described in this example and the
corresponding historical data on the progression of sepsis and the
treatment of sepsis with anti-TNF therapeutics as well as data from
other subjects in the clinical trial, a cluster analysis is
performed. Each discrete clinical outcome is assigned to a centroid
of one of the clusters.
[0252] Before the subject is treated with the drug, the subject is
assigned a position in the probability space based upon the
subject's data from the twenty biomarkers. The probability of the
subject belonging to any of the discrete clinical outcome classes
is determined using the Mahalanobis distance, from which a
probability is calculated using a Bayesian estimation. The result
of this calculation returns six probabilities corresponding to the
six clinical outcomes of the example. The position of the subject
is determined by plotting the six probabilities in the probability
space.
[0253] The probability space corresponding to the six discrete
clinical outcomes is graphically displayed as a hexagon in two
dimensions. The physician establishes rules tracking non-responsive
subjects and possible advent of sequellae. The rules are plotted on
the hexagonal graph as lines as illustrated in FIG. 17. The medical
condition of the subject is based the subject data's position
and/or trajectory in the probability space.
[0254] In FIG. 17, the trajectories of two different subjects are
plotted. Subject 1 has a trajectory that starts with a high
probability of being a member of the DIC condition. However, as the
subsequent time points are taken, the subject trajectory moves away
from the DIC condition and towards the NR condition. When the
physician notices the trajectory's speed and heading, at T=6, he
raises the dose. At T=9, the subject's position in the probability
space has a high probability of being part of the R condition. The
physician assesses the medical condition of the subject at any
point in time along the subject's trajectory.
[0255] Subject 2 has a trajectory that begins in the middle of the
probability space and heads towards the NR condition at T=1.
However, at T=2, the subject's position and trajectory are in
violation of the rule established by the physician and an alarm is
sounded. The alarm alerts the physician that the subject has a high
probability of suffering from one of the dangerous clinical
outcomes of sepsis.
Example 3
Diabetes and Insulin Sensitization
[0256] Subjects with non-insulin-dependent Type 2 diabetes are
treated with a commercially available insulin sensitizer and are
monitored for efficacy and safety wherein the primary ADR of
concern is congestive heart failure (CHF). The subject is monitored
over a course of time during the clinical study. The method of
action of the insulin sensitizer involves re-establishing insulin
sensitivity in the subject via antagonism of a nuclear receptor in
the adipose tissue and the liver. The insulin sensitizer and method
of action act to increase glucose uptake in the peripheral tissues
and thus reduce the circulating glycemia.
[0257] A subject and a physician choose the best set of discrete
clinical outcomes and rules for the subject based on personal
information. The discrete clinical outcomes are response (R),
non-response (NR), and adverse drug reaction (ADR). The ADR is
congestive heart failure.
[0258] After the three clinical outcomes are chosen, biomarkers
most representative of the medical condition of the subject are
determined by discriminant analysis. The biomarkers representative
of efficacy of the insulin sensitivity treatment are glucose,
insulin, adiponectin, and resistin. The ADR of congestive heart
failure identifies the representative biomarkers as Brain
natriuretic peptide (BNP), amino-terminal-pro-BNP (N-BNP), atrial
natriuretic factor (ANF), Troponin-C, Cardiac fatty acid binding
protein, Myosin light chain-1, Myoglobin, MMP-9. Using these
biomarkers and the corresponding historical data on the type 2
diabetes, a cluster analysis is performed. Each discrete clinical
outcome is assigned to a centroid of one of the clusters.
[0259] Before the subject is treated with an insulin sensitizer,
the subject is assigned a position in the probability space based
upon the subject's data from the relevant biomarkers. The
probability of the subject belonging any of the discrete clinical
outcome classes is determined using the Mahalanobis distance, from
which a probability is calculated using a Bayesian estimation. The
result of this calculation returns three probabilities
corresponding to the three clinical outcomes of the example. The
position of the subject is determined by plotting the three
probabilities in the probability space.
[0260] The probability space corresponding to three clinical
outcomes is graphically displayed as a triangle in two dimensions.
The physician and subject establish rules most important to the
individual diabetic subject. The rules are to track the NR and ADR
conditions. When the probability of belonging to the ADR condition
class increases beyond a boundary defined by the rule, an alarm is
triggered. The dose is increased according to a rule when the
subject has a high probability of being a NR.
[0261] In FIG. 18, the trajectories of two different subjects are
plotted. Subject 1 has a trajectory that starts in the middle of
the probability space and then heads towards the ADR condition. In
this example, the subject may notice the trajectory's speed or
acceleration and ask the physician if a change in treatment is
necessary. When Subject 1's trajectory crosses the dotted line
towards the ADR condition, an alarm is triggered. The subject may
react by going to the hospital.
[0262] Subject 2 has a trajectory that heads towards the NR
condition. The dose of the insulin sensitizer is adjusted
corresponding to the rules established by the physician and the
subject. After the adjustment of the dose, the trajectory heading
changes towards the R condition. The example illustrates a method
of the invention for monitoring a diabetic subject taking a
commercial insulin sensitizer and, with the help of a physician,
monitoring the subject's own condition at home.
Example 4
Graphical Representation of the Clinical Outcome of a Subject
[0263] The example demonstrates a method for communication of the
results from the mathematical model to a user. The method is to
plot a series of points and vectors in a space bounded by a
polygon. The vertices of the polygon represent the centroids of
clusters of discrete clinical outcomes. The position of the points
represents the probability that the subject data belongs to each
centroid. The probability of the point belonging to the discrete
clinical outcome is inversely proportional to the distance of that
point from the vertex. Since the vertices represent all the
significant clinical outcomes, the sum of the probabilities for a
given point must be 1.
[0264] A set of vectors is created where each value of a vector
represents the probability of membership in each of the clusters.
The modulus (magnitude) of the values is the probability and the
argument (angle) of the vector points to a vertex, which are
equally spaced. The vectors are summed and the resulting point is
located in the composite probability space.
TABLE-US-00001 TABLE 1 Discrete Clinical outcome Time Point R NR
ADR1 ADR2 ADR3 1 20 20 20 20 20 2 25 10 25 20 20 3 30 10 20 20 20 4
40 10 10 20 20 5 50 10 10 10 20 6 60 10 10 10 10 7 80 10 5 5 0 8
100 0 0 0 0
[0265] For example, in Table 1, each row represents the position of
one point. The columns, labeled at the top, represent the
probability that the point is a member of that cluster. So, in the
first row, the probability that the point is in cluster "R" is 20%,
that it is in the cluster "NR" is also 20%, and so on.
[0266] For the data in the table a graphical representation is
created as illustrated in FIG. 19. We see that the surrounding
polygon has five vertices, each labeled as one of the columns in
the tabular data above. The first point of the set, at the tail of
the chain of arrows, is at the center of the drawing, indicating
that the percentage probability that the point is in any of the
clusters is the same as the others. In this example, as the
trajectory grows through time, the probability of the subject data
reaches a final point that has 100 percent chance of being in
cluster "R" and zero percent chance of being elsewhere.
Correspondingly, the final point is at the "R" vertex.
[0267] Additional information can be included in the output of the
probability space, including the points representing the data of
the discrete clinical outcomes that defined the clusters and
centroids. FIG. 20 illustrates a graphical representation of the
cluster data points.
Example 5
[0268] Drug development is an extremely risky proposition. Over the
last decade, the failure rates for clinical programs have been
identified as high 90% with an amortized cost of nearly one billion
dollars per approved drug. These failures are due to any number of
factors. These include, but are not limited to: a faulty hypothesis
as to a target's role in the evolution and pathophysiology of the
disease; the testing of that hypothesis in an inappropriate, and
non-predictive, animal model; a faulty hypothesis regarding a
compound's MOA; inappropriate dosing schedules for individuals
within target subpopulations, including inappropriate
pharmacokinetics/pharmacodynamics (PK/PD) characteristics; and
developing untoward physiological effects (for example, Adverse
Drug Reactions--ADRs) in some segments of the patient population
and thus decreasing the therapeutic index for the compound in that
population.
[0269] This example demonstrates using a Type 1 biomarker for a
compound within a particular therapeutic area to: (a) validate the
underlying hypothesis regarding the method of action for the test
compound; (b) identify, and then enrich, the appropriate responder
subpopulation in a confirmatory trial; and (c) develop a truly
adaptive dosing protocol based on the individual PKs and PDs of the
responder population.
[0270] A real-time monitoring system is used identify the emergence
of these characteristic subpopulations earlier rather than later,
to efficiently manage and control the clinical trial and allows the
clinical research team make precise decisions regarding trial
management. On a patient by patient basis, it dynamically
identifies the compound's effects in one particular trial subject,
and then physically titrating that patient's dosing regimen to
accommodate a truly adaptive "Learn and Confirm" model of drug
development.
[0271] Longitudinal sampling of proteins is performed with each
patient, wherein the proteins represent both the pathophysiology of
the disease and the method of action of the compound. A standard
cluster analysis is applied at each sample time point in the
longitudinal sampling space that is consistent with the
evolutionary dynamics of the pathophysiology of the disease, and
the purported method of action and PD of the test compound.
[0272] Identification of the dynamics of the emergence of separate
subpopulations within the trial sample population can be performed
with two complementary clustering statistics: Cubic Clustering
Criterion (CCC) and the Pseudo-F which help the end-user establish
the number of clusters emerging within the geometry as demonstrated
in FIG. 21.
[0273] The end-user uses both the statistics highlighted above and
a visual inspection of the cluster scatter plots to make the final
determination as to how many distinct subpopulations exist in the
clinical study. Guiding this determination will be a fore-knowledge
of the number of treatment (cohort) arms, and any fore-knowledge as
to the character of a possible ADR. The methodology will suggest a
number of clusters to the end user, and will then present each
patient as to his/her cluster assignment. The cluster assignment
and the known assignment of a patient to a treatment cohort or ADR
outcome are then used standard assignment statistics to calculate
both the sensitivity and specificity of the assignment rule.
[0274] After determining that the clusters are reasonably specific,
the dynamic trajectories of the centroids of each cluster are
characterized, their separation speed, and the distribution of
speed and directions of each patient in each cluster. If, for
example, one cluster of patients is fairly homogeneous in that they
were all exposed to the compound at a given dose, and if the
distribution of distances, speeds, and directions, suggest that
some may be slow responders, then the end user may decide that
those patients should receive a higher dose.
[0275] Based on the assumptions developed herein, in silico
populations of virtual patients are created for each hypothetical
patient subclass. A time dependent Monte Carlo simulation is run to
identify the emergent clusters. After the simulation is run, a new
set of mean vectors and covariance structures is established and
the simulation studies are repeated. The sensitivity and
specificity of the clustering assignment rule for each dynamic is
measured and graphically plot as a receiver operator curve, wherein
the diagnostic variable will be the time at which the clusters were
suggested.
[0276] Results of the emergent dynamics by Monte Carlo simulation
of the described methods described in this example and results in
the following emergent dynamics as shown in FIG. 21. FIG. 21
demonstrates a dynamic subpopulation emergence based on cluster
analysis of a Monte Carlo simulation of a two-cohort clinical
design. At Time=4 there is significant evidence for the emergence
of two clusters while by Time=8 the bifurcation is complete and the
sensitivity and specificity of the method are both 100%. FIG. 22
demonstrates plots of the cluster calling statistics versus the
hypothetical number of clusters at Time=4. FIG. 23 illustrates
markers of clinical effect with the segregation of the marker space
as a function of dose at the end point of a clinical study.
Example 6
Identification of Type 1 Biomarkers Indicative of Periodic Dosing
Regimes
[0277] Periodic dosing, as a therapeutic strategy, is a mainstay
for the treatment of many chronic diseases or conditions. Foremost
among these are anti-cancer chemotherapy regimens and treatments
for autoimmune diseases. Whether therapeutically or in the context
of an adaptive clinical trial, the ability to both monitor and
characterize efficacy and toxicity profiles for a particular dosing
regimen in any given patient provides the clinical professional
important information regarding the ultimate efficacy of his/her
dosing strategy. The benefit of such a feedback mechanism is that
the feedback is timely and represents the mechanism of action and
clinical impact of the compound. For example, in cancer
chemotherapy, the feedback is provided by imaging technologies that
help quantify tumor burden, which in many cases is not timely.
[0278] This example characterizes the periodicity of the molecular
efficacy measures to identify trends and other measures of clinical
relevance (for example, increasing refractory response) to further
aid in clinical decision making. As demonstrated herein a real-time
feedback monitoring system can account for the periodicity of the
dosing regimen and characterizes any clinically trends in the
data.
[0279] The system can be a statistically valid and robust
methodology which dynamically characterizes the time-dependent
profiles of molecular markers of drug activity in periodically
treated patient subpopulations which then uses that information as
a real-time feedback monitor for optimizing the therapeutic
regimen.
[0280] The protein profiles of a treated population act in
accordance with the underlying hypothesis regarding a compound's
mechanism of action, they should, as a conglomerate pattern,
reflect the periodicity of the treatment regimen, and taken as a
whole, act as a Type 1 biomarker for compound activity at the site
of action. Additionally, if there is evidence that the periodic
expression of these profile proteins changes with time, the trend
of that change allows a clinician to adjust the dosing regimen for
example in frequency and/or dosage to individualize the
therapy.
[0281] Applying a time series-based statistical methodology to the
longitudinal (within an individual patient) protein profiles
systematically accounts for the periodicity of a given dosing
regimen within an individual patient and using that information to
identify and characterize the emergence of any clinically relevant
trends in the response. This knowledge can then be used to both
monitor a therapeutic regimen in an individual patient and to
design and manage the dynamics of an adaptive clinical trial.
[0282] By developing a valid longitudinal Type 1 biomarker for a
compound's activity in the context of a periodic dosing regimen
will: (a) validate the underlying hypothesis regarding the method
of action for the test compound; (b) identify potentially relevant
clinical conditions comprising: growing refractoriness to the
compound, target response to the regimen, and then adjust, as
appropriate, the frequency or dosages of the treatment itself; and
(c) develop an adaptive dosing protocol based on the individual PDs
of the responder population.
[0283] Using real-time monitoring and feedback to identify the
emergence of these characteristic dynamics aids the clinician to
make more precise decisions regarding patient treatment and/or
trial management.
[0284] A statistical Time Series analysis is used to characterize
the periodicity or seasonality of marker proteins in a patient
subjected to periodic treatments similar to those employed in
cancer chemotherapy, weekly methotrextate injections. The system
adjusts for the seasonality of the data train as shown in FIG. 24
and de-trends the responses as necessary, a one or two lag
differencing scheme as applied to Autoregressive Integrated moving
Average (ARIMA) Models, to characterize the actual response profile
of the patient to the therapy stream. FIG. 24 demonstrates patient
data showing the seasonality of two type 1 biomarkers over time. If
two or more proteins markers are identified as markers, any
temporal relationships between them (for example without
limitation, the levels of drug in the blood and/or clinical
readout) can be identified and characterized using Distributed Lags
analysis in the same modeling context. Each analysis depends on
explicit characterization of the time-time correlations observed in
the time series, the Autocorrelation Function (ACF) and its related
Partial Autocorrelation Function (PACF)). Each of these functions
can be represented as a vector of correlations which are
functionally dependent on the lag between measures. The functions
can be used to identify linear trends in the data, periodicity in
the data, and nonlinear curvature dynamics in the data. The
functions themselves can be visualized for inspection using
histograms called correlograms. Once the data is de-trended and the
periodicity accounted for, the individual spikes in the marker
profiles are analyzed to see if the compound is working as
expected.
[0285] The first step involved in developing this methodology is
identifying those proteins that best represent both the
pathophysiology of the disease and the purported method of action
of the compound. Once established, these proteins are sampled
longitudinally in each patient.
[0286] Based upon the expected seasonality of the treatment regimen
and the frequency of the therapeutic dosages, an appropriate
sampling frequency is established to ensure that the suspected
periodic expression of the marker proteins will be captured in the
sample data sets.
[0287] The system automatically generates and inspects difference
time series with 0, 1, or two lag differences comprising ARIMA
(0,0,0), ARIMA (0,1,0), and ARIMA (0,2,0) models and compares them
to the best fit through Root Mean Square Error, thus appropriately
de-trending the data to make the series stationary. Once a
stationary series has been established, the de-trending parameters
comprising the slope of a linear trend line is estimated from the
data, and an ACF and PACF for the stationary series is
generated.
[0288] Any periodic expression patterns are identified and
characterized as the period of the peaks and their subsequent
amplitudes. An example of this is shown in FIG. 25 demonstrating a
time series of a biomarker value versus the time of chemotherapy
delivery.
[0289] When multiple time series are involved comprising two or
more protein markers and/or levels of compounds in the blood, a
distributed lags analysis is performed to identify and characterize
the correlations between the series and their lags. The methodology
identifies the dynamics of the emergence of protein marker wave
trains and to relate these back to clinically relevant events
and/or the dosing schedule of the regimen.
[0290] A Monte Carlo simulation of the methods described above
using four time series were generated using a stationary time
series, a time series with linearly increasing slope, a time series
with exponentially decreasing affect to mimic patient-drug
attenuation, and a stationary time series with linearly decreasing
peaks, to model patient-drug adaptation or up-regulation of
metabolic pathways.
Example 7
Time to Fever Spike and Sepsis from Time Series Measurements of
Levels of Circulating Pyrogens and Other Biomarkers
[0291] A patient with Acute Myeloid Leukemia (AML) undergoing
in-patient chemotherapy was monitored using the methods and systems
described herein with sampling of blood every four to eight hours.
The patient's temperature was also monitored frequently. IL-6
elevation (>1000-fold between consecutive measurements) was
found at time designated zero in Table 2 and a significant decline
>30% between consecutive measurements 14 hours later. Fever was
first noted eight hours later. The methods herein would have
anticipated the subsequent recognition of sepsis by 8 hours.
TABLE-US-00002 TABLE 2 Time, hours Event 0.0 First elevation of
IL-6 8.4 First fever 14.3 Significant decline in protein-C 22.3
Septic shock 27.2 Patient admitted to ICU
[0292] The system identifies the relationship between the time to a
fever spike during aggressive chemotherapy of leukemia resulting in
fever and dramatically changed levels of circulating pyrogens
(Protein-C, IL-6, Il-1beta, TNF-alpha, IL-10). This is performed by
employing standard statistical modeling methodologies. The system
can be generalized to other situations where sepsis occurs. It can
also be employed for other situations where significant disease
processes and outcome of therapy can be beneficially projected.
[0293] Science and engineering often involve using easy-to-measure
variables, called factors or independent variable factors to
explain or predict the behavior of other variables, called
responses. When the factors are few in number, not significantly
redundant (non-collinear), and have a well-defined, functional
relationship to the responses, one can use multiple linear
regression to characterize that relationship. However, if any of
these three pre-conditions should fail, multiple linear regression
may prove to be inefficient or statistically inappropriate. In that
case, one may employ an alternative model building strategy termed
Partial Least Squares Regression. This methodology is particularly
useful when there are many factors that might be highly co-linear,
and when the relationship between independent variable factors and
dependent variable response is ill-defined. The aim of this type of
analysis is to construct a good predictive model.
[0294] Anticipating the time to a Fever Spike based on statistical
models using levels of several circulating pyrogens. FIG. 26
demonstrates the relationship between parameters computed from
circulating pyrogen levels measured over time and/or and the
time-to-spike (also labeled Lead Time). The rate of change and
shape of the parameters over time is determined by the incidence,
time of infection, extent of infection, numbers of organisms,
virulence of the organisms, and the condition and genetic make-up
of the patient. The activation threshold is the point at which the
body physiology begins to generate a fever spike. The detection
threshold is a parameter level at which some therapeutic
intervention is indicated. The system determines the relationship
between the pyrogen levels and the dependent variable in the model,
Lead Time, enabling prediction of fever spike early enough to
intervene with beneficial outcome.
[0295] This method and data provides enables anticipation of the
time patients are most likely to experience a fever spike. The
physician can then preempt and/or ameliorate the onset of the fever
event, for example by providing prophylactic antibiotic
therapy.
[0296] The treatment of patients with Acute Myelogenous Leukemia
(AML) is used as an exemplary case. Such patients are often treated
with a course of induction chemotherapy (in-patient chemotherapy
typically using more than one drug and intended to eliminate all
cancer cells). An undesired side-effect of the induction
chemotherapy is usually febrile neutropenia. This occurs because
infectious microorganisms can invade the patient due to the
inactivation of active immunity by the chemotherapy. The definition
of febrile neutropenia is the fever spike, a clinician defined
state in which the patient's temperature measures greater than 39
degrees Celsius or in a six hour period two consecutive patient
temperatures measured greater than 38.5 degrees Celsius.
[0297] Fever, as a defense mechanism, results from the effects
exerted on the hypothalamus by circulating cytokines, called
pyrogens. These cytokines emerge from the immune system during an
inflammatory response. The best known of these are IL-6, IL-1 beta,
and TNF-alpha. The purpose of this example is to identify a pattern
of circulating pyrogens leading up to a fever spike, and, from
them, build a model that best anticipates the time-to-spike.
[0298] The methods developed in the studies may be generalized to
generate a monitoring means in other high-risk disease states and
therapeutic procedures.
[0299] The system builds a statistical model characterizing the
relationship between the time-to-spike and the circulating levels
for two particular pyrogens, IL-6 and IL-1 beta. The system, while
independent of measurement means, requires frequent sampling in the
period prior to the fever spike (anywhere from 3 to 6 measures per
day).
[0300] The first step involved in developing the methodology
requires one characterization of fever in well defined categories
and to identify, as closely as possible, the time of the first
fever event. The definition chosen was that either the patient's
temperature measures greater than 39 degrees Celsius or in a six
hour period or two consecutive temperatures are recorded greater
than 38.5 degrees Celsius.
[0301] Individual patient data for the circulating levels of IL-6,
IL-1 beta, and TNF-alpha and other biomarkers were recorded in a
database, based on the sampling interval. The samples in the study
(3 measures per day at 8-hour increments) were subjected to post
hoc analysis, and used to populate the model.
[0302] These markers were selected based upon a number of current
studies which show that patients with Sepsis Syndrome have elevated
levels of circulating IL-1 beta, IL-6, and TNF-alpha when compared
with critically ill patients without sepsis and normal controls
(Casey, L. Annals of Internal Medicine (1993) 119; 8:771-778).
[0303] IL-6 predominately regulates the synthesis of CRP made by
the hepatocytes (liver) in the acute phase response. Transcription
of CRP is also regulated somewhat by TNF and IL-1. CRP is a
pentraxin protein which binds to phosphocholine on bacterial cell
walls, endogenous cell membranes, and apoptotic bodies.
Inflammation caused by infection/sepsis, inflammatory and
autoimmune disorders, traumatic injury, and some malignancies, CRP
is seen in increased amounts making it a useful biomarker in
monitoring sepsis and infection. CRP production is not affected by
neutropenia. There is an abundance of literature which focuses on
CRP levels during neutropenic fever and in diagnosing the severity
of febrile incidences.
[0304] Protein C is anti-coagulant protein that is synthesized in
the liver. It plays an important role in maintaining coagulation
homeostasis. Protein C is activated by thrombin into an active
protein which along with protein S and phospholipid as cofactors,
it cleaves coagulation factors VIII and V thus inhibiting
coagulation. Recently, Protein C's anti-inflammatory responses have
been discovered.
[0305] Patients with sepsis and septic shock have Protein C
depletion which may have a role in the pathophysiology of sepsis.
Improvement was seen in a specific subgroup of patients by
counteracting Protein C depletion by the addition of activated
Protein C. Protein C can be used as a biomarker in the diagnosis of
sepsis in patients and as a therapeutic agent. One study found that
concentrations of Protein C are proportionate with severity of
neutropenic sepsis at fever onset.
[0306] Ten adult leukemia patients receiving chemotherapy were
enrolled. Blood samples were taken every 8 hours prior to the onset
of the fever spike and then every 6 hours until the end of the
observation period. Biomarker analysis was performed on these
samples. Patient sampling of 8 hours prior to fever spike and 6
hours after fever spike observed.
[0307] A Partial Least Squares model was constructed based on the
pre-spike (36 hour window) measures for IL-6, IL-1-beta, and
TNF-alpha and their sample-to-sample fold changes. Given the data
set derived from this study, a minimal sampling period of 8 hours
and a minimal window of 36 hours pre-spike is sufficient to derive
an anticipative model of the time to spike.
[0308] Quantitative Biomarker concentration data was derived using
well known ELISA immunoassay known to measure IL-6, IL1-beta, and
TNF-alpha. Concentrations measured in pg/mL, and sample-to-sample
fold changes in concentration were recorded and analyzed as
follows.
[0309] The concentration data were first visually inspected for
Gaussian normality. Because they were log normally distributed the
data was transformed by calculating the log of the raw data
measures. Based on these measures, an initial Partial Least Squares
regression model was built that included all three pyrogens and
their fold changes from sample to sample. This model characterized
the amount of variation accounted for in the covariance structures
of the dependent vectors (times-to-spike) and the independent
vectors (pyrogen levels). In this data set there were more
independent variables than degrees of freedom. To address the
problem, standard statistical model building tools like top-down
elimination and bottom-up addition were employed.
[0310] The full model (IL-6, TNF-alpha, and IL-1beta, and their
sample to sample changes) accounted for the percentage of the
variation in the whole model covariance structure and the
percentage of the variance in the dependent variable
(time-to-spike). Using feature selection processes comprising and
not limited to bottom up model building and top down model
building, a series of seven sub-models were developed.
[0311] The chosen model included log(IL-6), log(IL-1 beta),
log(sample-to-sample fold change IL-6), log(TNF-alpha), and
accounted for 93% of the total model covariance and 27.5% of the
variance in the dependent variable (time-to-spike).
[0312] The model is given by a standard multiple linear regression
equation with the following parameter set: intercept=36.5;
log(IL-6)=-8.0; log(IL-6 fold change)=2.4; log(IL-1 beta)=-0.3; and
log(TNF-alpha)=4.4. Each measure is a point-in-time measure at the
observed time-to-spike. For example, 24 hours prior to the fever
spike, the log of the concentrations of IL-6, IL-1 beta, and their
sample-to-sample fold changes from the 30 hour time point are added
to the model using fitted parameters, meaning the coefficients of
the linear model.
[0313] The final model was selected from the seven total sub-models
because it accounted for the highest level of the percentage of
variance in the dependent vector. The final model included
log(IL-6), log(IL-1 beta), and the logs of their sample-to-sample
fold changes as anticipators of time-to-spike. The size of the
anticipative window of 36 hours was selected by clinical relevance
but is not limited to that anticipative window. The selected model
equation is as follows:
Time-to-spike=36.6-8*log(IL6)+2.4*Log(IL-6 fold
change)-0.3*Log(IL-1 beta+4.4*Log(TNF alpha)
[0314] The results of the 10 patients were evaluated in this
example for their biomarker patterns in the 36 hours prior to fever
spikes. The data for each of the pyrogens and their
sample-to-sample fold changes are in FIGS. 27-32. FIG. 27 is the
IL-1 beta concentration versus time point number. FIG. 28 is the
sample-to-sample fold change for IL1-beta versus time point number.
FIG. 29 is the IL-6 concentration versus time point number, wherein
patient 4 became septic. FIG. 30 is the sample-to-sample fold
change for IL-6 versus time point number, wherein patient 4 became
septic. FIG. 31 is the TNF-alpha concentration versus time point
number. FIG. 32 is the sample-to-sample fold change for TNF-alpha
versus time point number.
[0315] The model anticipated the time-to-spike in all patients and
explained the percentage of the variance observed in the dependent
variable. Patient 6 had ATRA syndrome and thus is outside the
patient population used to construct the model.
[0316] The residual plot for patient 5 is given in FIG. 33. Patient
5 has a fever spike and no decreasing markers. Temperature in
degrees C. (right scale) biomarker concentrations given in pg/mL,
wherein IL-6 is a 10-fold increase, not necessarily indicating
sepsis. An analysis of the residuals suggest that the model
anticipations for the period between 12 and 24 hours prior to the
fever spike are non-systematic, meaning randomly distributed about
the zero-axis and the size of the residual is no more than the
percentage of the anticipated time-to-spike, which is anticipated
by the model using the equation 1.
[0317] For patient 4, the model anticipates a time-to-spike of 3.3
hours when the actual time-to-spike was 7 hours, and, at the other
extreme, anticipates a time-to-spike of 31.9 hours when the actual
time-to-spike was actually 34 hours as demonstrated in FIG. 34.
Patient 4 data points showing decreasing Protein-C with increasing
IL-6 just prior to fever spike. Normalized parameter is the ratio
of the (parameter value at time t minus the lowest value in the
range) to the difference between the maximum and minimum value of
the parameter. A residual analysis for Patient 4 yields an
excellent fit to the data measured in patient 4, meaning the
residuals are randomly distributed about the zero axis, and no
residual is greater than 20% of the anticipated time-to-spike, as
derived from equation 1.
[0318] In conclusion of this example, induction cancer chemotherapy
is only one of the underlying etiologies of sepsis. Others include,
but are not limited to, severe burn and thermal injury or traumatic
injury. In burn and traumatic injury, the patient expresses an
alteration in the immune system resulting in a predominance of the
T Helper-2 Lymphocyte phenotype, where the phenotype of the helper
cell population switches from Th1 to Th2. This is termed the
Th1-to-Th2 switch and makes the patient more susceptible to the
onset of infection. FIG. 35 is a schematic of the Th1-to-Th2 switch
and the cytokines representing each phenotype. Once the patient
becomes infected, if the underlying pathogen is identified, the
rate and spread of infection can result in a systemic inflammatory
response which is called sepsis. If the pathogen is unidentified,
the inflammatory response is termed Systemic Inflammatory Response
Syndrome (SIRS). Thus, the underlying etiology of progression to
sepsis can be captured as a process divided into three
physiologically distinct intervals: Susceptibility to Infection,
which is represented by the Th1--Th2 switch, progression through
infection, and the onset and progression of sepsis/SIRS. The
markers relevant to the monitoring of each stage are shown in Table
3.
TABLE-US-00003 TABLE 3 Sepsis Onset and Sequellae I. Sepsis Onset
a. Protein C** b. IL-1beta** c. IL-6** d. Procalcitonin e. IL-10
II. DIC a. Von Willebrand Factor b. Procalcitonin c. Neopterin d.
TNF-.alpha. e. Protein C f. PAF1 g. (Pro)Thrombin III. ARF (Acute
Renal Failure) a. Creatinine b. Cystin C IV. ARDS a. Von Willebrand
Factor b. IL-6 c. Il-8 d. IL-1 e. Il-1ra (IL-1 receptor agonist) f.
TNF V. ALF (Acute Liver Failure) a. ALT b. AST c. GGT d. LDH e.
Alkaline phosphatase f. Protein C g. Cytokeratin-18 fragments
[0319] FIG. 36 and FIG. 37 demonstrate that the system uses an
equation that is anticipative of sepsis using the Squared
Maholanobis distance equation and Bayesian Probability. The Squared
Mahalanobis Distance is:
D.sub.j.sup.2(x)=(.mu..sub.j-x).sup.T.SIGMA.-1(.mu..sub.j) where
j=1,2,3
and the Bayesian probability is:
Pr(j|x)=exp(-0.5D.sub.j.sup.2(X))/.SIGMA.exp(0.5D.sub.k.sup.2(x))
wherein .mu. is the mean vector representing the centroid of the
patient population and j is the three different outcomes (sepsis,
fever, infection).
[0320] FIG. 38 illustrates a graph in time of a plurality of marker
proteins for a patient in a sepsis trial. The time of the onset to
sepsis for this patient is marked and all patterns leading up to
that point are relevant to the determination that this patient
became septic.
[0321] FIG. 39 illustrates a bivariate time course of two
particular markers (protein C and C-reactive protein) in the same
patient. The change in direction in the center of the graph
represents a rapid onset of disease. In this example, a patient
rapidly deteriorated and was illustrated as septic in a very
particular region of the bivariate space. In this example, the
sampling intervals are fairly regular; therefore the length of each
line segment represents the rate of change of the biomarkers in
this particular space.
* * * * *