U.S. patent application number 16/870703 was filed with the patent office on 2020-08-27 for biosensor systems for determining analyte concentration based on complex index functions.
The applicant listed for this patent is Ascensia Diabetes Care Holdings AG. Invention is credited to Sung-Kwon Jung, Huan-Ping Wu.
Application Number | 20200271615 16/870703 |
Document ID | / |
Family ID | 1000004813083 |
Filed Date | 2020-08-27 |
![](/patent/app/20200271615/US20200271615A1-20200827-D00001.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00002.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00003.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00004.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00005.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00006.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00007.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00008.png)
![](/patent/app/20200271615/US20200271615A1-20200827-D00009.png)
![](/patent/app/20200271615/US20200271615A1-20200827-M00001.png)
![](/patent/app/20200271615/US20200271615A1-20200827-M00002.png)
View All Diagrams
United States Patent
Application |
20200271615 |
Kind Code |
A1 |
Wu; Huan-Ping ; et
al. |
August 27, 2020 |
Biosensor Systems for Determining Analyte Concentration Based On
Complex Index Functions
Abstract
A biosensor system determines analyte concentration from an
output signal generated from a light-identifiable species or a
redox reaction of the analyte. The biosensor system adjusts a
correlation for determining analyte concentrations from output
signals or determined analyte concentrations with one or more
complex index function extracted from the output signals or from
other sources. The complex index functions determine at least one
slope deviation value, .DELTA.S, or normalized slope deviation from
one or more error parameters. The slope-adjusted correlation
between analyte concentrations and output signals may be used to
determine analyte concentrations having improved accuracy and/or
precision from output signals including components attributable to
bias.
Inventors: |
Wu; Huan-Ping; (Granger,
IN) ; Jung; Sung-Kwon; (Rensselaer, IN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Ascensia Diabetes Care Holdings AG |
Basel |
|
CH |
|
|
Family ID: |
1000004813083 |
Appl. No.: |
16/870703 |
Filed: |
May 8, 2020 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
14231874 |
Apr 1, 2014 |
10656113 |
|
|
16870703 |
|
|
|
|
13153793 |
Jun 6, 2011 |
8744776 |
|
|
14231874 |
|
|
|
|
PCT/US2009/067150 |
Dec 8, 2009 |
|
|
|
13153793 |
|
|
|
|
61120525 |
Dec 8, 2008 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G01N 27/416 20130101;
G01N 27/3273 20130101; G01N 27/3274 20130101 |
International
Class: |
G01N 27/327 20060101
G01N027/327; G01N 27/416 20060101 G01N027/416 |
Claims
1-25. (canceled)
26. A biosensor system, for determining an analyte concentration in
a fluid sample, comprising: a test sensor having a sample interface
adjacent to a reservoir holding the fluid sample, wherein the test
sensor generates at least one output signal responsive to the
concentration of the analyte in the fluid sample; and a measurement
device having a sensor interface, a storage medium, and a processor
coupled to a sensor interface and the storage medium, wherein the
sensor interface is in electrical communication with the sample
interface, and wherein the processor is operable to: determine at
least one output signal value from the output signal received from
the sensor interface, wherein the at least one output signal value
represents an unknown concentration of the analyte in the fluid
sample, determine at least one .DELTA.S value from at least a
complex index function stored in the storage medium, wherein the at
least one .DELTA.S value represents a slope deviation between a
slope of a correlation between analyte concentration and the at
least one output signal, and a hypothetical slope of a perfect
correlation between analyte concentration and output signals, the
correlation between analyte concentration and the at least one
output signal determined from a previously determined reference
correlation between previously determined reference output signal
values and reference sample analyte concentration values, the
reference sample analyte concentration values obtained from a
reference instrument, and wherein the complex index function
includes two terms, wherein each of the two terms is modified by a
weighing coefficient, wherein at least one of the two terms
modified by the weighing coefficients is responsive to a bias
between the reference correlation of analyte concentrations and the
at least one output signal, wherein the complex index function is
responsive to at least one error parameter, wherein the weighing
coefficients are numerical values other than one or zero, and
determine the analyte concentration in the fluid sample from the
output signal value and a slope compensation equation stored in the
storage medium, wherein the slope compensation equation adjusts a
slope of at least one reference correlation with the at least one
determined .DELTA.S value.
27. The system of claim 26, the test sensor arranged and configured
to have a substantially linear relationship between a percent bias
in an analyte concentration determined from the biosensor system
and the at least one .DELTA.S value.
28. The system of claim 27, wherein the percent bias
((.DELTA.A/A.sub.ref)*100%) of the determined analyte concentration
is in a substantially linear relationship with the at least one
.DELTA.S value, wherein .DELTA.A represents the difference between
the corrected analyte concentration Aeon and a reference analyte
concentration A.sub.ref.
29. The system of claim 26, wherein the processor is further
capable of determining the at least one .DELTA.S value from a
predictor function, the predictor function including the complex
index function and one or more constants.
30. The system of claim 26, wherein the complex index further
comprises at least one constant that is not equal to zero.
31. The system of claim 26, wherein one of the at least two terms
includes at least one of a raw analyte concentration value of the
sample, a temperature, or an error parameter responsive to the %
Hct of the sample.
32. The system of claim 26, wherein at least one other term of the
complex index function includes the at least one error parameter,
the at least one error parameter independently selected from
intermediate output signal values and values external to the output
signal.
33. The system of claim 32, wherein the error parameter is
responsive to error contributors causing an alteration of the at
least one output signal value.
34. The system of claim 32, wherein the processor is further
operable to normalize the at least one .DELTA.S value, wherein the
normalizing is in response to a slope of a reference correlation
equation or in response to a normalized slope function.
35. The system of claim 32, wherein the at least two terms are
selected by at least one exclusion test.
36. The system of claim 26, wherein the measurement device includes
a signal generator operable to generate a sequence of gated
amperometry multi-pulse signals on the sensor interface and wherein
the output signal is a sequence of output current values.
37. The system of claim 36, wherein at least one of the terms of
the complex index function includes a ratio of one of the output
current values in the sequence to another of the output current
values in the sequence.
38. A method for determining an analyte concentration in a fluid
sample, comprising: generating an output signal from a test sensor
having the fluid sample in an interface adjacent to a reservoir,
the test sensor inserted in a sensor interface of a measurement
device; and determining an output signal value from the output
signal wherein the output signal value represents an unknown
concentration of the analyte in the fluid sample; determining a
.DELTA.S value from a complex index function stored in a storage
medium, wherein the .DELTA.S value represents a slope deviation
between a slope of a correlation between analyte concentration and
the output signal, and a hypothetical slope of a perfect
correlation between analyte concentration and output signals, the
correlation between analyte concentration and the output signal
determined from a previously determined reference correlation
between previously determined reference output signal values and
reference sample analyte concentration values, the reference sample
analyte concentration values obtained from a reference instrument,
and wherein the complex index function is responsive to at least
one error parameter and includes at least two terms, wherein each
of the at least two terms is modified by a weighing coefficient,
wherein at least one of the at least two terms modified by the
weighing coefficients is responsive to a bias between the reference
correlation of analyte concentrations; and determining the analyte
concentration in the sample from the output signal value and a
slope compensation equation stored in the storage medium, wherein
the slope compensation equation adjusts a slope of at least one
reference correlation with the at least one determined .DELTA.S
value.
39. The method of claim 38, wherein the at least one .DELTA.S value
is determined from a predictor function, the predictor function
including the complex index function and one or more constants.
40. The method of claim 38, wherein the complex index further
comprises at least one constant that is not equal to zero.
41. The method of claim 38, wherein one of the at least two terms
includes at least one of a raw analyte concentration value of the
sample, a temperature, or an error parameter responsive to the %
Hct of the sample.
42. The method of claim 38, wherein at least one other term of the
complex index function includes the at least one error parameter,
the at least one error parameter independently selected from
intermediate output signal values and values external to the output
signal.
43. The method of claim 42, wherein the error parameter is
responsive to error contributors causing an alteration of the at
least one output signal value.
44. The method of claim 42, further comprising normalizing the at
least one .DELTA.S value in response to a slope of a reference
correlation equation or in response to a normalized slope
function.
45. The method of claim 42, further comprising selecting the at
least two terms by an exclusion test.
46. A method of determining an analyte concentration in a sample,
comprising: providing a sequence of gated amperometry multiple
pulse signals via a signal generator of a measurement device to a
test sensor, wherein the test sensor includes a reservoir to hold
the sample, and a working electrode and a counter electrode in
electrical communication to the measurement device, and wherein the
measurement device includes a processor, a sensor interface, a
display, and a storage medium; measuring a set of output signals,
including an indicating signal for the analyte and intermediate
signals in response to the sequence of gated amperometry multiple
pulse signals from the working electrode responsive to the analyte
of unknown concentration from the sample; computing by the
processor at least one .DELTA.S value from at least one complex
index function stored in the device storage medium, wherein the at
least one .DELTA.S value represents a slope deviation of a
hypothetical slope of a perfect correlation between the analyte
concentration and the output signals from the slope of the
predetermined reference correlation between analyte concentration
and the output signals, wherein the computing of the at least one
.DELTA.S value from at least one complex index function is from the
error parameters calculated from selected intermediate output
signals and dedicated measurement signals; determining the analyte
concentration by converting the indicating signal of the analyte to
concentration through adjusting the reference correlation slope by
the at least one .DELTA.S value; and displaying the output analyte
concentration on the display of the measurement device.
Description
REFERENCE TO RELATED APPLICATIONS
[0001] This application is a divisional of U.S. application Ser.
No. 13/153,793 entitled "A Method of Determining Analyte
Concentration Based on Complex Index Functions" filed Jun. 6, 2011,
which is a continuation of PCT Application No. PCT/US2009/067150
entitled "Biosensor System with Signal Adjustment" filed Dec. 8,
2009, which claims the benefit of U.S. Provisional Application No.
61/120,525 entitled "Complex Index Functions" filed Dec. 8, 2008,
which are incorporated by reference in their entirety.
BACKGROUND
[0002] Biosensor systems provide an analysis of a biological fluid,
such as whole blood, serum, plasma, urine, saliva, interstitial, or
intracellular fluid. Typically, the systems include a measurement
device that analyzes a sample contacting a test sensor. The sample
usually is in liquid form and in addition to being a biological
fluid, may be the derivative of a biological fluid, such as an
extract, a dilution, a filtrate, or a reconstituted precipitate.
The analysis performed by the biosensor system determines the
presence and/or concentration of one or more analytes, such as
alcohol, glucose, uric acid, lactate, cholesterol, bilirubin, free
fatty acids, triglycerides, proteins, ketones, phenylalanine or
enzymes, in the biological fluid. The analysis may be useful in the
diagnosis and treatment of physiological abnormalities. For
example, a diabetic individual may use a biosensor system to
determine the glucose level in whole blood for adjustments to diet
and/or medication.
[0003] Biosensor systems may be designed to analyze one or more
analytes and may use different volumes of biological fluids. Some
systems may analyze a single drop of whole blood, such as from
0.25-15 microliters (.mu.L) in volume. Biosensor systems may be
implemented using bench-top, portable, and like measurement
devices. Portable measurement devices may be hand-held and allow
for the identification and/or quantification of one or more
analytes in a sample. Examples of portable measurement systems
include the Ascensia.RTM. Breeze.RTM. and Elite.RTM. meters of
Bayer HealthCare in Tarrytown, N.Y., while examples of bench-top
measurement systems include the Electrochemical Workstation
available from CH Instruments in Austin, Tex.
[0004] Biosensor systems may use optical and/or electrochemical
methods to analyze the biological fluid. In some optical systems,
the analyte concentration is determined by measuring light that has
interacted with or been absorbed by a light-identifiable species,
such as the analyte or a reaction or product formed from a chemical
indicator reacting with the analyte. In other optical systems, a
chemical indicator fluoresces or emits light in response to the
analyte when illuminated by an excitation beam. The light may be
converted into an electrical output signal, such as current or
potential, which may be similarly processed to the output signal
from an electrochemical method. In either optical system, the
system measures and correlates the light with the analyte
concentration of the sample.
[0005] In light-absorption optical systems, the chemical indicator
produces a reaction product that absorbs light. A chemical
indicator such as tetrazolium along with an enzyme such as
diaphorase may be used. Tetrazolium usually forms formazan (a
chromagen) in response to the redox reaction of the analyte. An
incident input beam from a light source is directed toward the
sample. The light source may be a laser, a light emitting diode, or
the like. The incident beam may have a wavelength selected for
absorption by the reaction product. As the incident beam passes
through the sample, the reaction product absorbs a portion of the
incident beam, thus attenuating or reducing the intensity of the
incident beam. The incident beam may be reflected back from or
transmitted through the sample to a detector. The detector collects
and measures the attenuated incident beam (output signal). The
amount of light attenuated by the reaction product is an indication
of the analyte concentration in the sample.
[0006] In light-generated optical systems, the chemical detector
fluoresces or emits light in response to the analyte redox
reaction. A detector collects and measures the generated light
(output signal). The amount of light produced by the chemical
indicator is an indication of the analyte concentration in the
sample.
[0007] In electrochemical biosensor systems, the analyte
concentration is determined from an electrical signal generated by
an oxidation/reduction or redox reaction of the analyte or a
species responsive to the analyte when an input signal is applied
to the sample. The input signal may be a potential or current and
may be constant, variable, or a combination thereof such as when an
AC signal is applied with a DC signal offset. The input signal may
be applied as a single pulse or in multiple pulses, sequences, or
cycles. An enzyme or similar species may be added to the sample to
enhance the electron transfer from a first species to a second
species during the redox reaction. The enzyme or similar species
may react with a single analyte, thus providing specificity to a
portion of the generated output signal. A mediator may be used to
maintain the oxidation state of the enzyme.
[0008] Electrochemical biosensor systems usually include a
measurement device having electrical contacts that connect with
electrical conductors in the test sensor. The conductors may be
made from conductive materials, such as solid metals, metal pastes,
conductive carbon, conductive carbon pastes, conductive polymers,
and the like. The electrical conductors typically connect to
working, counter, reference, and/or other electrodes that extend
into a sample reservoir. One or more electrical conductors also may
extend into the sample reservoir to provide functionality not
provided by the electrodes.
[0009] The measurement device applies an input signal through the
electrical contacts to the electrical conductors of the test
sensor. The electrical conductors convey the input signal through
the electrodes into the sample present in the sample reservoir. The
redox reaction of the analyte generates an electrical output signal
in response to the input signal. The electrical output signal from
the strip may be a current (as generated by amperometry or
voltammetry), a potential (as generated by
potentiometry/galvanometry), or an accumulated charge (as generated
by coulometry). The measurement device may have the processing
capability to measure and correlate the output signal with the
presence and/or concentration of one or more analytes in the
biological fluid.
[0010] In coulometry, a potential is applied to the sample to
exhaustively oxidize or reduce the analyte. A biosensor system
using coulometry is described in U.S. Pat. No. 6,120,676. In
amperometry, an electrical signal of constant potential (voltage)
is applied to the electrical conductors of the test sensor while
the measured output signal is a current. Biosensor systems using
amperometry are described in U.S. Pat. Nos. 5,620,579; 5,653,863;
6,153,069; and 6,413,411. In voltammetry, a varying potential is
applied to a sample of biological fluid. In gated amperometry and
gated voltammetry, pulsed inputs may be used as described in WO
2007/013915 and WO 2007/040913, respectively.
[0011] In many biosensor systems, the test sensor may be adapted
for use outside, inside, or partially inside a living organism.
When used outside a living organism, a sample of the biological
fluid may be introduced into a sample reservoir in the test sensor.
The test sensor may be placed in the measurement device before,
after, or during the introduction of the sample for analysis. When
inside or partially inside a living organism, the test sensor may
be continually immersed in the sample or the sample may be
intermittently introduced to the strip. The test sensor may include
a reservoir that partially isolates a volume of the sample or be
open to the sample. When open, the strip may take the form of a
fiber or other structure placed in contact with the biological
fluid. Similarly, the sample may continuously flow through the
strip, such as for continuous monitoring, or be interrupted, such
as for intermittent monitoring, for analysis.
[0012] Biosensor systems may provide an output signal during the
analysis of the biological fluid that includes one or multiple
errors. These errors may be reflected in an abnormal output signal,
such as when one or more portions or the entire output signal is
non-responsive or improperly responsive to the analyte
concentration of the sample. These errors may be from one or more
contributors, such as the physical characteristics of the sample,
the environmental aspects of the sample, the operating conditions
of the system, interfering substances, and the like. Physical
characteristics of the sample include hematocrit (red blood cell)
concentration and the like. Environmental aspects of the sample
include temperature and the like.
[0013] The measurement performance of a biosensor system is defined
in terms of accuracy and/or precision. Increases in accuracy and/or
precision provide for an improvement in measurement performance, a
reduction in the bias, of the system. Accuracy may be expressed in
terms of bias of the sensor system's analyte reading in comparison
to a reference analyte reading, with larger bias values
representing less accuracy. Precision may be expressed in terms of
the spread or variance of the bias among multiple analyte readings
in relation to a mean. Bias is the difference between one or more
values determined from the biosensor system and one or more
accepted reference values for the analyte concentration in the
biological fluid. Thus, one or more errors in the analysis results
in the bias of the determined analyte concentration of a biosensor
system.
[0014] Bias may be expressed in terms of "absolute bias" or
"percent bias". Absolute bias may be expressed in the units of the
measurement, such as mg/dL, while percent bias may be expressed as
a percentage of the absolute bias value over the reference value.
Under the ISO standard, absolute bias is used to express error in
glucose concentrations less than 75 mg/dL, while percent bias is
used to express error in glucose concentrations of 75 mg/dL and
higher. The term "combined bias" (expressed as bias/%-bias)
represents absolute bias for glucose concentrations less than 75
mg/dL and percent bias for glucose concentrations of 75 mg/dL and
higher. Accepted reference values for analyte concentrations may be
obtained with a reference instrument, such as the YSI 2300 STAT
PLUS.TM. available from YSI Inc., Yellow Springs, Ohio.
[0015] Hematocrit bias refers to the difference between the
reference glucose concentration obtained with a reference
instrument and an experimental glucose reading obtained from a
biosensor system for samples containing differing hematocrit
levels. The difference between the reference and values obtained
from the system results from the varying hematocrit level between
specific whole blood samples and may be generally expressed as a
percentage by the following equation: %
Hct-Bias=100%.times.(G.sub.m-G.sub.ref)/G.sub.ref, where G.sub.m
and G.sub.ref are the determined glucose and reference glucose
concentration readings, respectively, for any hematocrit level. The
larger the absolute value of the % Hct-bias, the more the
hematocrit level of the sample (expressed as % Hct: the percentage
of red blood cell volume/sample volume) is reducing the accuracy
and/or precision of the determined glucose concentration. For
example, if whole blood samples containing identical glucose
concentrations, but having hematocrit levels of 20, 40, and 60%,
are analyzed, three different glucose readings will be reported by
a system based on one set of calibration constants (slope and
intercept of the 40% hematocrit containing whole blood sample, for
instance). "Hematocrit sensitivity" is an expression of the degree
to which changes in the hematocrit level of a sample affect the
bias values for an analysis. Hematocrit sensitivity may be defined
as the numerical values of the combined biases per percent
hematocrit, thus bias/%-bias per % Hct.
[0016] Temperature bias refers to the difference between an analyte
concentration obtained at a reference temperature and an analyte
concentration obtained at a different experimental temperature for
the same sample. The difference between the analyte concentration
obtained at the reference temperature and that obtained from the
different experimental temperature may be generally expressed as a
percentage by the following equation: %
Temp-Bias=100%.times.(A.sub.mTemp-A.sub.RefTemp)/A.sub.RefTemp,
where A.sub.mTemp and A.sub.RefTemp are the analyte concentrations
at the experimental and reference temperatures, respectively, for
the sample. The larger the absolute value of the % Temp-bias, the
more the temperature difference is reducing the accuracy and/or
precision of the glucose concentration determined at the different
experimental temperature. "Temperature sensitivity" is an
expression of the degree to which changes in the temperature at
which the analysis is performed affect the bias values for an
analysis. Temperature sensitivity may be defined as the numerical
values of the combined biases per degree of temperature, thus
%-bias/.degree. C. Temperature sensitivity also may be defined as
slope deviation per degree of temperature, thus .DELTA.S/.degree.
C.
[0017] Many biosensor systems include one or more methods to
correct errors associated with an analysis. The concentration
values obtained from an analysis with an error may be inaccurate.
Thus, the ability to correct these analyses may increase the
accuracy and/or precision of the concentration values obtained. An
error correction system may compensate for one or more errors, such
as a sample temperature or a sample hematocrit level, which are
different from a reference temperature or a reference hematocrit
value.
[0018] Some biosensor systems have an error correction system that
compensates for different hematocrit concentrations in the sample.
Various methods and techniques have been proposed to reduce the
bias of the hematocrit effect on glucose measurements. Some methods
use the ratio of currents from a forward and a reverse potential
pulse to compensate for the hematocrit effect. Other methods have
been proposed to reduce the bias of the hematocrit effect,
including using silica particles to filter red blood cells from the
electrode surface or using wide electrode spacing in combination
with mesh layers to distribute blood throughout the test
sensor.
[0019] Some biosensor systems have an error correction system that
compensates for temperature. Such error correction systems
typically alter a determined analyte concentration for a particular
reference temperature in response to an instrument or sample
temperature. A number of biosensor systems compensate for
temperature error by correcting the output signal prior to
calculating the analyte concentration from a correlation equation.
Other biosensor systems compensate for temperature error by
correcting the analyte concentration calculated from the
correlation equation. Generally, conventional methods of
temperature compensation look at the effect of temperature on a
specific parameter, not the overall effect the temperature error
has on the bias of the analysis. Biosensor systems having error
detection and/or compensation systems for the sample temperature
are described in U.S. Pat. Nos. 4,431,004; 4,750,496; 5,366,609;
5,395,504; 5,508,171; 6,391,645; and 6,576,117.
[0020] Some biosensor systems have an error correction system that
compensates for interferents and other contributors. Such error
correction systems typically use an electrode lacking one or more
of the working electrode reagents to allow for the subtraction of a
background interferent signal from the working electrode
signal.
[0021] While conventional error compensation systems balance
various advantages and disadvantages, none are ideal. Conventional
systems usually are directed to detect and respond to a particular
type of error, either temperature or hematocrit, for example. Such
systems typically do not have the ability to compensate for
multiple error sources. These systems generally also lack the
ability to alter the compensation for the error based on the output
signal from a specific sample. Consequently, conventional biosensor
systems may provide analysis results having determined analyte
concentration values outside a desired performance limit.
[0022] Accordingly, there is an ongoing need for improved biosensor
systems, especially those that may provide increasingly accurate
and/or precise determination of the concentration of the analyte in
the sample. The systems, devices, and methods of the present
invention overcome at least one of the disadvantages associated
with conventional biosensor systems.
SUMMARY
[0023] The present invention provides a biosensor system that
adjusts a relation for determining analyte concentrations in a
biological sample from output signals with one or more complex
index functions responsive to one or more errors that could bias
the determined analyte concentrations. The bias may be represented
by slope deviations, .DELTA.S values, and normalized slope
deviations obtained from one or more error parameters. The .DELTA.S
values represent slope deviations determined with one or more
complex index functions from the error parameters. The complex
index functions include at least two terms modified by weighing
coefficients. The terms may include error parameters extracted from
or independent of the output signals.
[0024] In a method for determining an analyte concentration in a
sample, an output signal value responsive to the concentration of
the analyte in the sample is generated. At least one .DELTA.S value
from at least one error parameter is determined, and the at least
one output signal value is compensated with at least one reference
correlation and at least one .DELTA.S value to determine the
analyte concentration in the sample. The at least one .DELTA.S
value may be determined from a predictor function f(predictor). The
f(predictor) includes an index function and relates at least one
error parameter to .DELTA.S. The reaction may be an electrochemical
redox reaction.
[0025] In a method for determining complex index functions from
error parameters, at least one error parameter responsive to the
percent bias in a determined analyte concentration in a sample is
determined. The at least one error parameter is related to at least
one .DELTA.S value with at least one complex index function, the at
least one .DELTA.S value representing the difference in slope
between the slope from a reference correlation and a hypothetical
slope of a line for the output signal value that would provide an
analyte concentration in the sample without bias. The complex index
functions includes the at least one error parameter incorporated as
a term modified by a weighing coefficient.
[0026] In a method for selecting terms for inclusion in a complex
index function, multiple error parameters are selected as terms for
potential inclusion in the complex index function. First exclusion
values are determined for each selected term. One or more exclusion
tests are applied to the exclusion values to identify one or more
terms to exclude from the complex index function. After the
exclusion of at least one term, second exclusion values are
determined for the remaining terms. If the second exclusion values
do not identify remaining terms for exclusion from the complex
index function under the one or more exclusion texts, the remaining
terms are included in the complex index function.
[0027] In a method for determining a complex index function from
hematocrit-adjusted and donor blood samples for use in a
measurement device, the experimental glucose concentration of
multiple hematocrit-adjusted blood samples having known reference
glucose concentrations at multiple environmental conditions is
determined with multiple test sensors. The slope and intercept of a
reference correlation for the multiple test sensors is determined
from the determined and known glucose concentrations at a reference
temperature and at a reference % Hct. The reference glucose
concentration is determined for multiple donor blood samples. The
multiple hematocrit-adjusted blood sample glucose concentration
data may be combined with the multiple donor blood sample glucose
concentration data. Terms are selected from the data for one or
more output signal value. Terms also may be selected for one or
more physical characteristic, environmental condition,
concentration value, and the like. Weighing coefficients are
determined for the terms, in addition to any coefficients. The
complex index function is determined from the combination of
selected terms, corresponding weighing coefficients, and any
constants.
[0028] A biosensor system for determining an analyte concentration
in a sample includes a measurement device and a test sensor. The
measurement device has a processor connected to a sensor interface
and to a storage medium. The test sensor has a sample interface
adjacent to a reservoir formed by the sensor. The processor
determines an output signal value responsive to the concentration
of the analyte in the sample from the sensor interface. The
processor determines at least one .DELTA.S value from an error
parameter and compensates the output signal value with the at least
one .DELTA.S value and at least one reference correlation present
in the storage medium.
[0029] A biosensor system adjusts a correlation between analyte
concentrations and output signals with at least one .DELTA.S value
in response to error parameters. The processor determines an
analyte concentration from the slope-adjusted correlation in
response to an output signal from the sample interface.
[0030] In another method for determining an analyte concentration
in a sample, one or more output signals are generated from a
sample. At least one complex index function is determined, where
the complex index function is responsive to more than one error
parameter. The analyte concentration in the sample is determined
from the output signals in response to the at least one complex
index function.
[0031] Other systems, methods, features, and advantages of the
invention will be, or will become, apparent to one with skill in
the art upon examination of the following figures and detailed
description. It is intended that all such additional systems,
methods, features and advantages be included within this
description, be within the scope of the invention, and be protected
by the claims that follow.
BRIEF DESCRIPTION OF THE DRAWINGS
[0032] The invention may be better understood with reference to the
following drawings and description. The components in the figures
are not necessarily to scale, emphasis instead being placed upon
illustrating the principles of the invention.
[0033] FIG. 1A represents a method for determining an analyte
concentration in a sample.
[0034] FIG. 1B represents a method for selecting terms for
inclusion in a complex index function.
[0035] FIG. 1C represents a method of determining a complex index
function from hematocrit-adjusted and donor blood samples for use
in a measurement device.
[0036] FIG. 2 depicts the correlation between %-bias and an index
function based on a ratio parameter.
[0037] FIG. 3 depicts the relationship between S.sub.cal,
S.sub.hyp, .DELTA.S, A.sub.corr, A.sub.cal, and .DELTA.A.
[0038] FIG. 4 depicts a gated pulse sequence where the input signal
includes multiple pulses.
[0039] FIG. 5A depicts a graph of the correlations between .DELTA.S
and R4/3 index values.
[0040] FIG. 5B depicts a graph of the correlations of .DELTA.S with
complex index values.
[0041] FIG. 6A depicts a graph of the correlations of .DELTA.S for
blood samples at 21.degree. C. with R4/3 index values.
[0042] FIG. 6B depicts a graph of the correlations of .DELTA.S for
blood samples at 21.degree. C. with complex index values.
[0043] FIG. 6C depicts a graph of the correlations of .DELTA.S for
blood samples at 18.degree. C. with R4/3 index values.
[0044] FIG. 6D depicts a graph of the correlations of .DELTA.S for
blood samples at 18.degree. C. with complex index values.
[0045] FIG. 6E depicts a graph of hematocrit sensitivity in
combined bias vs. % Hct.
[0046] FIG. 6F depicts a graph correlating combined biases to
reference glucose concentrations for uncompensated and complex
index compensation corrected analyte concentrations.
[0047] FIG. 7 depicts a schematic representation of a biosensor
system that determines an analyte concentration in a sample of a
biological fluid.
DETAILED DESCRIPTION
[0048] A biosensor system adjusts a correlation for determining
analyte concentrations in a biological sample from output signals
with complex index functions extracted from intermediate signals of
the output signals or from other sources. The analyte may generate
the output signals in response to a light-identifiable species or a
redox reaction. The intermediate signals may be one or more
portions of the output signals or the like. Predictor functions
including at least one complex index function adjust the
correlation for determining analyte concentrations from the output
signals for one or more errors in the analyses. Predictor functions
including at least one complex index function also may be used to
correct an analyte concentration including errors. Such errors can
result in bias, thus reduced accuracy and/or precision, of the
determined analyte concentrations. In addition to the compensation
system providing substantial benefits when analyzing complex
biological samples, the compensation system may be used to improve
the measurement performance of other types of analysis.
[0049] Complex index functions include combinations of terms
modified by weighing coefficients. The terms included in the
complex index function may be selected with one or more exclusion
tests. Predictor functions and/or complex index functions
correspond to the bias/%-bias in the correlation between the
analyte concentrations and the output signals due to one or more
errors in the analysis. The %-bias in the correlation may be
represented by one or more .DELTA.S values obtained from one or
more error parameters. The .DELTA.S values represent slope
deviations of the correlation between analyte concentrations and
output signals determined from one or more error parameters. Thus,
the more closely a predictor or complex index function correlates
with .DELTA.S (.DELTA.S=f(CIndex)), the better the function is at
correcting error in the analysis.
[0050] Complex index functions corresponding to the slope or change
in slope may be normalized to reduce the statistical effect of
changes in the output signals, improve the differentiation in
variations of the output signals, standardize the measurements of
the output signals, a combination thereof, or the like. Since the
slope deviation may be normalized, a complex index function also
may be expressed in terms of .DELTA.S/S=f(CIndex). The adjusted
correlation may be used to determine analyte concentrations in the
sample from the output signals or may be used to correct analyte
concentrations and may provide improved measurement performance in
comparison to conventional biosensors. A more detailed treatment of
error correction using index functions and .DELTA.S values may be
found in Intl. Pub. No. WO 2009/108239, filed Dec. 6, 2008,
entitled "Slope-Based Compensation."
[0051] FIG. 1A represents a method for determining an analyte
concentration in a sample of a biological fluid. In 102, the
biosensor system generates an output signal in response to either a
light-identifiable species or an oxidation/reduction (redox)
reaction of an analyte in a sample of a biological fluid. In 104,
the biosensor system measures the output signal. In 106, the
analyte concentration is determined from a compensation equation
including at least one complex index function and the output
signal. In 110, the analyte concentration may be displayed, stored
for future reference, and/or used for additional calculations.
[0052] In 102 of FIG. 1A, the biosensor system generates an output
signal in response to a light-identifiable species or an
oxidation/reduction (redox) reaction of an analyte in a sample of a
biological fluid. The output signal may be generated using an
optical sensor system, an electrochemical sensor system, or the
like.
[0053] In 104 of FIG. 1A, the biosensor system measures the output
signal generated by the analyte in response to the input signal
applied to the sample, such as from a redox reaction of the
analyte. The system may measure the output signal continuously or
intermittently. For example, the biosensor system may measure the
output signal intermittently during the pulses of a gated
amperometric input signal, resulting in multiple current values
recorded during each pulse. The system may show the output signal
on a display and/or may store the output signal or portions of the
output signal in a memory device.
[0054] In 106 of FIG. 1A, the analyte concentration of the sample
may be determined from a compensation equation including at least
one complex index function and the output signal. The complex index
function may form part of a predictor function. FIG. 2 depicts the
correlation between %-bias and an index function based on a ratio
parameter (R5/4). The ratio parameter, R5/4, represents the
relationship between the currents generated by the analyte in
response to the 4.sup.th and 5.sup.th pulses of a gated amperometry
pulse sequence including 7 pulses. Other ratio parameters and index
functions may be used. Thus, the %-bias of a measured analyte
concentration in a biological fluid, such as glucose in whole
blood, may be determined from or correlated with the output signals
of the analysis, such as the intermediate currents generated by the
analyte in response to a gated amperometry sequence.
[0055] The relationship between %-bias and a predictor function may
be represented as follows:
%-bias=f(predictor) (Equation 1),
where %-bias equals (.DELTA.A/A.sub.ref)*100% and f(predictor)
equals a.sub.1*f(Index)+a.sub.0. .DELTA.A is the difference between
the measured or calculated analyte concentration, A.sub.cal, and a
reference analyte concentration, A.sub.ref (a known analyte
concentration in a biological sample). f(Index) may be a single
error parameter, a combination of error parameters, or other
values. Thus, substituting terms for Equation 1 results in the
following relationship between %-bias and an index function:
(.DELTA.A/A.sub.ref)*100%=a.sub.1*f(Index)+a.sub.0 (Equation
2).
[0056] Rearranging the terms of Equation 2 results in the following
relationship:
.DELTA.A=A.sub.ref*(a.sub.1*f(Index)+a.sub.0)/100 (Equation 3).
[0057] A compensation may be expressed as follows:
A.sub.corr=A.sub.0+.DELTA.A (Equation 4).
[0058] Where A.sub.corr is a corrected or compensation analyte
compensation and A.sub.0 is an initial analyte value from the
analysis. While .DELTA.A may be obtained from Equation 3, A.sub.ref
in Equation 3 may not be available during the analysis of a
biological sample. However, the initial analyte value, A.sub.0, may
be used from the analysis in place of A.sub.ref. Thus, Equation 3
may be approximated by the following relationship:
.DELTA.A.apprxeq.A.sub.0*(a.sub.1*Index+a.sub.0)/100 (Equation
5).
[0059] Finally, substituting Equation 5 into Equation 4 results in
the following relationship:
A.sub.corr=A.sub.0+A.sub.0*(a.sub.1*Index+a.sub.0)/100=A.sub.0*[1+(a.sub-
.1*Index+a.sub.0)/100] (Equation 6).
[0060] From Equation 6, the difference between the measured analyte
concentration and a reference analyte concentration, .DELTA.A, is
based on an initial analyte value, A.sub.0, which may be biased due
to one or more errors in the analysis. Thus, there is no reference
point or value upon which to base the compensation of the measured
analyte concentration. While these and other equations presented
throughout the application and claims may include an "=" sign, the
sign is used to represent equivalence, relationship, prediction, or
the like.
[0061] The %-bias in the correlation of analyte concentrations with
output signals also may be represented by one or more slope
deviations, .DELTA.S, obtained from one or more error parameters.
Error containing portions of output signals are reflected in the
deviation between the hypothetical slope of the output signals and
the slope of a reference correlation. By determining one or more
.DELTA.S values reflecting this deviation in slope from one or more
error parameters, the measurement performance of an analysis may be
increased. One or more .DELTA.S values for an analysis may be
determined from one or more error parameters. The relationship
between .DELTA.S values and the value of one or more error
parameters may be described by an index function. Index functions,
in addition to reference correlation equations, may be
pre-determined and stored in the biosensor system. Error parameter
values may be determined before, during, or after the analysis.
[0062] The slope compensation equation uses output signal values to
provide a compensated analyte concentration. The slope compensation
equation also may use other values. The slope compensation equation
compensates for error by adjusting a reference correlation between
output signals and known analyte concentrations to provide a
compensated or corrected analyte concentration.
[0063] The slope compensation equation may be represented as
follows:
A c o r r = i - I n t S c a l + .DELTA. S , ( Equation 7 )
##EQU00001##
[0064] where A.sub.corr is the corrected analyte concentration, i
is a value of the output signal from a biosensor system, Int is the
intercept from a reference correlation equation, S.sub.cal is the
slope from the reference correlation equation, and .DELTA.S
represents the deviation in slope between S.sub.cal and a
hypothetical slope of a line (S.sub.hyp) for the output signal
value that provides an analyte concentration of the sample without
error. The Int and S.sub.cal values for the reference correlation
equation may be implemented as a program number assignment (PNA)
table, another look-up table, or the like in the biosensor system.
Other slope compensation equations including at least one .DELTA.S
value and the output signal may be used.
[0065] Equation 7 is a representation of the corrected analyte
concentration determined using the slope deviation .DELTA.S, where
.DELTA.S is essentially the total slope deviation related to
essentially the total error associated with the analyte analysis.
The total slope deviation may be caused by one or more error
sources. Equation 7 may be used with any signal having a
substantially linear response to analyte concentration. Equation 7
may be used with other signals, such as signals that are near or
partially linear. While .DELTA.S is responsive to one or more
errors in the output signal, i represents the error containing
portions of the output signal not responsive to the analyte
concentration of the sample. Thus, S.sub.hyp=S.sub.cal+.DELTA.S.
One or more values for Int and S.sub.cal may be stored in the
biosensor system for comparison with the output signal i to
determine A.sub.corr for the sample.
[0066] If the value of .DELTA.S is determined experimentally from
samples and substituted into Equation 7, the bias in the determined
analyte concentrations of those samples will be fully compensated.
Alternatively, if .DELTA.S is substituted with a predictor
function, then the ability of the compensation equation to correct
bias in the determined analyte concentration will depend on how
well the value generated from the predictor function correlates
with .DELTA.S. In Equation 7, a predictor function, f (predictor),
may be substituted for .DELTA.S. Thus, Equation 7 may be rewritten
as follows:
A c o r r = i - Int S c a l + .DELTA. S = i - Int S c a l + f ( p r
e d i c t o r ) = i - Int S c a l + b 1 * f ( CInde x ) + b 0 . (
Equation 8 ) ##EQU00002##
[0067] While the predictor function, f(predictor), may have the
general form of b.sub.1*f(CIndex)+b.sub.0, where f(CIndex) is a
complex index function, other values or indices may be used in
combination with the f(CIndex) to provide f(predictor). For
example, the complex index function could be used with or without
one or both of the b.sub.1 and b.sub.0 values to provide the
predictor function. Multiple complex index functions also may be
combined to provide the f(predictor), and thus, the corrected
analyte concentration of the sample.
[0068] For the theoretical situation where .DELTA.S and the complex
index function perfectly correlate, b.sub.1 (representing slope)
and b.sub.0 (representing intercept) are one and zero,
respectively. When the predictor function is approximating
.DELTA.S, the theoretical value of one may be used in place of
b.sub.1 when b.sub.1=1.+-.0.2, preferably one may be used in place
of b.sub.1 when b.sub.1=1.+-.0.15, and more preferably one may be
used in place of b.sub.1 when b.sub.1=1.+-.0.1. When the predictor
function is approximating .DELTA.S, the theoretical value of zero
may be used in place of b.sub.0 when b.sub.0=0.+-.0.3, preferably
zero may be used in place of b.sub.0 when b.sub.0=0.+-.0.2, and
more preferably zero may be used in place of b.sub.0 when
b.sub.0=0.+-.0.1. Other deviation cut-offs may be used to determine
when the theoretical values for b.sub.1, b.sub.0, or both may be
used. In addition to substituting b.sub.1 and/or b.sub.0 with the
theoretical values 1 and 0, predetermined values from a look-up
table and the like may be substituted based on the same or other
deviation cut-offs.
[0069] In 108 of FIG. 1A, the corrected analyte concentration value
may be displayed, stored for future reference, and/or used for
additional calculations.
[0070] FIG. 3 shows the relationship between S.sub.cal, S.sub.hyp,
.DELTA.S, A.sub.corr, A.sub.cal, and .DELTA.A. Line A represents a
reference correlation having a slope S.sub.cal and relating an
output signal in the form of current values from a biosensor system
to analyte concentration values obtained from a YSI or other
reference instrument for the samples. When used during the analysis
of a sample by a biosensor system, the reference correlation of
Line A may include output signal current values having one or more
errors that may provide an inaccurate and/or imprecise analyte
concentration value. Line B represents an error-compensated
correlation having a slope S.sub.hyp and relating current values
obtained from the system with the sample analyte concentration
values as obtained from the reference instrument. The
error-compensated correlation has been adjusted or modified to
reduce or substantially eliminate the one or more errors. .DELTA.S
is the difference in slope between these correlation lines.
.DELTA.A is the difference between the uncompensated or uncorrected
(A.sub.cal) and error compensated or corrected (A.sub.corr)
determined analyte concentration values.
[0071] Without compensation or correction, a specific output signal
value will provide a different sample analyte concentration from
the S.sub.cal reference correlation line than from the S.sub.hyp
error-compensated line. The A.sub.corr value obtained from the
S.sub.hyp error-compensated line provides a more accurate value of
the analyte concentration in the sample. Thus, Equation 7
translates a current value, S.sub.cal, and Int into the compensated
analyte concentration value A.sub.corr using .DELTA.S. In this way,
the percent bias may be linked through .DELTA.S into Equation 7.
The percent bias values may be pulled toward the center of a bias
distribution through the linkage of .DELTA.S to the percent bias.
As .DELTA.S is responsive to bias, changing .DELTA.S affects the
amount of bias remaining in the compensated analyte concentration
of the sample.
[0072] The responsiveness of .DELTA.S to one or more errors in the
analysis may be represented by a predictor function. To determine
one or more predictor functions, the deviation in the slope of the
correlation equation in response to the one or more errors
(.DELTA.S.sub.cal) may be determined from experimental data, such
as during factory calibration, as follows:
.DELTA. S c a l = i - Int A r e f - S c a l , ( Equation 9 )
##EQU00003##
[0073] where i is a value of the output signal from a biosensor
system, Int is the intercept from a reference correlation equation,
A.sub.ref is the reference analyte concentration of the sample,
such as obtained from a reference instrument, and S.sub.cal is the
slope from a reference correlation equation, such as
i=S.sub.cal*A.sub.ref+Int. One or more .DELTA.S.sub.cal values may
be determined at each reference analyte concentration. In this
manner, for multiple known analyte concentrations, an output signal
value may be obtained from the biosensor system and a corresponding
.DELTA.S.sub.cal value determined. An initial predictor function
may be determined by taking the .DELTA.S.sub.cal values from
Equation 9 and correlating them to an error parameter.
[0074] Predictor functions compensate the measured analyte
concentration for one or more errors in the analyte concentration
analysis. One or more predictor functions may be used. A predictor
function that perfectly correlates with the total slope deviation
.DELTA.S would provide an ultimate total error compensation of the
analyte concentration. Such a hypothetical, perfectly correlated
predictor function could be used to compensate for all errors in
the analysis without having to know the exact cause of the total
slope deviation .DELTA.S, and thus the bias of the measured analyte
concentration. Predictor functions include at least one index
function, and one or more of the index functions may be complex.
Preferably, predictor functions include at least one complex index
function.
[0075] An index function is responsive to at least one error
parameter. An index function may be a calculated number that
correlates with an error parameter, such as hematocrit or
temperature, and represents the influence of this error parameter
on the slope deviation .DELTA.S. Thus, error parameters may be any
value responsive to one or more errors in the output signal. Index
functions may be experimentally determined as a regression equation
of the plot between .DELTA.S.sub.cal and an error parameter.
[0076] Index functions may be determined using error parameters
values from the analysis of the analyte, such as the intermediate
signals from an output signal, or from sources independent of the
analyte output signal, such as thermocouples, additional
electrodes, and the like. Thus, the error parameters may be
extracted directly or indirectly from the output signal of the
analysis and/or obtained independently from the output signal. Any
error parameter may be used to form the terms, such as those
described in Intl. Pub. No. WO 2009/108239, filed Dec. 6, 2008,
entitled "Slope-Based Compensation," and the like.
[0077] Temperature may be considered an error parameter because an
error in concentration values may arise from performing an analysis
at a temperature other than that at which the reference correlation
was determined. For example, temperature affects the oxidation and
diffusion of glucose in a sample of whole blood and the diffusion
of optically active molecules. The temperature for the analysis may
be determined from any source, such as a thermocouple, calculated
estimates, and the like. Thus, f(Index).sub.Temp relates
temperature to the deviation in slope between the reference
correlation slope determined at a reference temperature and the
hypothetical slope of the line that would provide the temperature
affected analyte concentration at the temperature at which the
analysis was performed. The index function for temperature
f(Index).sub.Temp may be stored in the biosensor system with the
reference correlation equation.
[0078] FIG. 4 depicts a gated pulse sequence where the input signal
includes multiple pulses. The output signal current values
resulting from the pulses are depicted above each pulse. The
recorded intermediate signal current values are depicted as
circles. Each of the i values is a current value of the output
signal responsive to the input signal. The first number in the
subscript of the i values denotes the pulse number, while the
second number in the subscript denotes the order of the output
signal as the current values were recorded. For example, i.sub.2,3
denotes the third current value recorded for the second pulse.
[0079] As previously discussed, index functions may include ratios
extracted from the intermediate output signals as depicted in FIG.
4. For example, the intermediate signal values may be compared
within an individual pulse-signal decay cycle, such as ratios
R3=i.sub.3,3/i.sub.3,1, R4=i.sub.4,3/i.sub.4,1, and the like. In
another example, the intermediate signal values may be compared
between separate pulse-signal decay cycles, such as ratios
R3/2=i.sub.3,3/i.sub.2,3, R4/3=i.sub.4,3/i.sub.3,3, and the
like.
[0080] Index functions also may include combinations of ratios
extracted from the output signal depicted in FIG. 4. In one
example, an index function may include a ratio of ratios, such as
Ratio3/2=R3/R2, Ratio4/3=R4/R3, and the like. In another example,
an index function may include a combination of indices. For
example, a combination index, Index-1, may be represented as
Index-1=R4/3-Ratio3/2. In another example, a combination index
Index-2 may be represented as
Index-2=(R4/3).sup.p-(Ratio3/2).sup.q, where p and q independently
are positive numbers.
[0081] An index function is complex when the function includes a
combination of terms modified by weighing coefficients. The
combination is preferably a linear combination, but other
combination methods may be used that provide weighing coefficients
for the terms. Each term may include one or more error parameters.
An example of a complex index function is represented as
follows:
f(CIndex)=a.sub.1+(a.sub.2)(R3/2)+(a.sub.3)(R4/3)+(a.sub.4)(R5/4)+(a.sub-
.5)(R3/2)(G.sub.raw)+(a.sub.6)(R4/3)(G.sub.raw)+(a.sub.7)(R3/2)(Temp)+(a.s-
ub.8)(R4/3)(Temp)+(a.sub.9)(Temp)+(a.sub.10)(G.sub.raw)+ . . .
(Equation 10),
where a.sub.1 is a constant, a.sub.2-a.sub.10 independently are
weighing coefficients, G.sub.raw is the determined analyte
concentration of the sample without compensation, and Temp is
temperature. Each of the weighing coefficients (a.sub.2-a.sub.10)
is followed by its associated term.
[0082] There are at least three basic types of terms in the complex
index function represented by Equation 10: (1) the individual ratio
indices extracted from the output signal, such as R3/2 and R4/3,
(2) the interaction terms between the ratio indices extracted from
the output signal and the temperature or G.sub.raw, such as
(R3/2)(G.sub.raw) and (R3/2)(Temp), and (3) temperature and
G.sub.raw. The terms may include values other than error
parameters, including G.sub.raw. Other terms also may be used,
including, but not limited to a combination index function, as
previously described. The complex index function may be solved to
provide a complex index value when the terms are replaced with the
appropriate values. Statistical processing may be performed on the
multiple terms to determine one or more constants and weighing
coefficients. Statistical package software, including MINITAB
(MINTAB, INC., State College, Pa.), may be used to perform the
statistical processing.
[0083] The constant a.sub.1 may be determined by regression or
other mathematical technique. While a single constant is shown in
Equation 10, a constant is not required; more than one may be used,
and may be equal to 0. Thus, one or more constants may or may not
be included in the complex index function. One or more constants
also may be combined with an index function in forming the
predictor function, such as the b.sub.0 constant previously
described in relation to Equation 8, for example.
[0084] While terms having weighing coefficients of one may be used,
a complex index function includes at least two terms that are
modified by weighing coefficients. Weighing coefficients are
numerical values other than one or zero. Preferably, each term
including an error parameter is modified by a weighing coefficient.
More preferably, each non-constant term of the complex index
function is modified by a weighing coefficient. Weighing
coefficients may have positive or negative values. Weighing
coefficients may be determined through the statistical processing
of the experimental data collected from a combination of multiple
analyte concentrations, different hematocrit levels, different
temperatures, and the like.
[0085] Table 1, below, lists the weighing coefficients and p-values
resulting from a multi-variable regression of data taken from
glucose output signals (currents) from capillary and venous blood
samples at 21.degree. C. and 18.degree. C. of a donor study with 52
donors. Each blood sample from each donor was analyzed twice for
glucose, to give approximately 104 data points in the data
population. The samples were analyzed using a gated amperometric
input signal where selected intermediate output signals were
recorded from the pulses. MINITAB version 14 software was used with
the Multi-Variant Regression of Linear Combinations of Multiple
Variables option chosen to perform the multi-variable regression.
Other statistical analysis or regression options may be used to
determine the weighing coefficients for the terms.
TABLE-US-00001 TABLE 1 Results of multivariable regression.
Weighing Coefficient Term Coefficient Standard Error T P Constant
133.52 48.35 2.76 0.006 R3/2 204.96 71.03 2.89 0.004 R4/3 -356.79
96.47 -3.70 0.000 (R3/2)(G.sub.raw) -0.0408 0.1163 -0.35 0.726
(R4/3)(G.sub.raw) -0.0338 0.1812 -0.19 0.852 (Temp)(R3/2) -12.237
3.704 -3.30 0.001 (Temp)(R4/3) 15.565 5.115 3.04 0.002 Temp -2.516
2.503 -1.01 0.315 G.sub.raw 0.08274 0.09661 0.86 0.392
[0086] The resulting complex index function may be represented as
follows:
.DELTA.S.sub.RegA=134+(205)(R3/2)-(357)(R413)-(0.041)(R3/2)(G.sub.raw)-(-
0.034)(R4/3)(G.sub.raw)-(12.2)(Temp)(R3/2)+(15.6)(Temp)(R4/3)-(2.52)(Temp)-
+(0.0827)(G.sub.raw) (Equation 11),
where .DELTA.S.sub.RegA is a complex index function describing
.DELTA.S.sub.cal, defined as
.DELTA.S.sub.cal=(i/A.sub.ref)-S.sub.cal, where A.sub.ref is the
reference analyte concentration value obtained from the YSI
reference instrument and S.sub.cal is the slope from the reference
correlation equation, as previously discussed with regard to
Equation 7, for example. The R.sup.2 value reflecting how well the
outputs from the .DELTA.S.sub.RegA complex index function
correspond to the S.sub.cal values was 77.2% (R.sup.2*100%). Thus,
the R.sup.2 value indicated the correlation between the complex
index function and S.sub.cal. Larger R.sup.2 values reflect the
complex index being better at describing .DELTA.S.sub.cal.
[0087] FIG. 1B represents a method for selecting terms for
inclusion in a complex index function. In 112, multiple error
parameters are selected as terms for potential inclusion in the
complex index function. The error parameters may be extracted
directly or indirectly from an output signal responsive to a
light-identifiable species or from the redox reaction of an analyte
in a sample of a biological fluid. The error parameters also may be
obtained independently from the output signal, such as from a
thermocouple. The terms may include values other than error
parameters. In 114, one or more mathematical techniques are used to
determine first exclusion values for each selected term. The
mathematical techniques may include regression, multi-variant
regression, and the like. The exclusion values may be p-values or
the like. The mathematical techniques also may provide weighing
coefficients, constants, and other values relating to the selected
terms.
[0088] In 116, one or more exclusion tests are applied to the
exclusion values to identify one or more terms to exclude from the
complex index function. At least one term is excluded under the
test. In 117, the one or more mathematical techniques are repeated
to identify second exclusion values for the remaining terms. In
118, if the second exclusion values do not identify remaining terms
for exclusion from the complex index function under the one or more
exclusion tests, the remaining terms are included in the complex
index function. In 120, if the second exclusion values identify
remaining terms to exclude from the complex index function under
the one or more exclusion tests, the one or more mathematical
techniques of 117 may be repeated to identify third exclusion
values for the remaining terms. These remaining terms may be
included in the complex index function as in 118 or the process may
be iteratively repeated as in 120 until the exclusion test fails to
identify one or more terms to exclude.
[0089] Table 1, above, also lists p-values for each term. The
p-values indicate the probability of affecting the correlation
between the complex index function and .DELTA.S if the term were
eliminated from the complex index function. For example, a p-value
of 0.05 or more for a term means that the probability is 5% or more
that the elimination of the term from the complex index function
would not reduce the correlation of the complex index function to
.DELTA.S. Thus, p-values may be used as exclusion values for an
exclusion test to select terms for potential exclusion from the
complex index function. The smaller the numerical p-value selected
as an exclusion value, the more terms will be excluded from the
complex index function.
[0090] When the exclusion test uses p-values as exclusion values,
exclusion p-values from about 0.01 to about 0.10 are preferred,
with exclusion p-values values from about 0.03 to about 0.07 being
more preferred. In addition to exclusion tests based on p-values,
other exclusion tests also may be used to identify potential terms
for exclusion from the complex index functions. Removing terms from
the complex index function that do not affect the correlation
between the complex index function and .DELTA.S in an undesirable
way, allows the desired correlation between the complex index
function and .DELTA.S. Thus, the desired improvement in measurement
performance may be achieved by the compensation equation, while
providing a shorter analysis time. Furthermore, the precision of
subsequent analyses performed using different biosensor systems and
conditions may be improved through the removal of undesirable terms
from the complex index function.
[0091] With regard to the terms in Table 1, terms having p-values
greater than 0.05 were selected for potential removal from the
complex index function. Thus, the terms (R3/2)(G.sub.raw),
(R4/3)(G.sub.raw), Temp and G.sub.raw were identified as terms that
may be appropriate for removal from the complex index function
after the first multivariable regression. As the (R4/3)(G.sub.raw)
term showed the greatest p-value (0.852), the term was removed and
the multivariable regression was repeated. This and a third
iteration of the multivariable regression identified that the Temp
and G.sub.raw terms had the second and third highest p-values. With
the removal of the (R4/3)(G.sub.raw), Temp, and G.sub.raw terms, it
was unexpectedly determined that the p-value of the
(R3/2)(G.sub.raw) term had fallen under the 0.05 exclusion value,
as shown in Table 2, below. Thus, while the weighing coefficient
for the (R3/2)(G.sub.raw) term is numerically small (0.00799) in
relation to the other weighing coefficients, the term contributed
to the complex index function's ability to correlate with .DELTA.S.
Preferably, an iterative process of selecting and eliminating terms
with the largest undesirable departure from an exclusion test is
repeated until the remaining terms meet the test.
TABLE-US-00002 TABLE 2 Results of multivariable regression with the
reduced term set. Weighing Coefficient Term Coefficient Standard
Error T P Constant 95.463 3.930 24.29 0.000 R3/2 177.66 68.22 2.60
0.010 R4/3 -289.31 70.91 -4.08 0.000 (R3/2)(G.sub.raw) 7.9899
.times. 10.sup.-3 7.575 .times. 10.sup.-4 10.55 0.000 (Temp)(R3/2)
-11.221 3.550 -3.16 0.002 (Temp)(R4/3) 11.928 3.709 3.22 0.001
[0092] The complex index function of Equation 11 after removal of
the (R4/3)(G raw), Temp, and G.sub.raw terms may be represented as
follows:
.DELTA.S.sub.RegB=95.5+(178)(R3/2)-(289)(R4/3)+(0.00799)(R3/2)(G.sub.raw-
)-(11.2)(Temp)(R3/2)+(11.9)(Temp)(R4/3) (Equation 12).
The R.sup.2 value reflecting how well the outputs from the
.DELTA.S.sub.RegB equation correspond to the S.sub.cal values was
77.1%. Removal of the terms eliminated by the exclusion test from
Equation 11 did not cause a significant change (0.1) in the ability
of the reduced-term complex index function to describe .DELTA.S.
Thus, the ability of the complex index of Equation 12 to describe
the errors in the Table 1 data was preserved, while providing a
beneficial reduction in the number of terms in relation to Equation
11.
[0093] FIG. 5A is a graph for the data from the donor study
previously discussed with regard to Table 1 of the correlations of
.DELTA.S with R4/3 index values. The "cap/21C" data set represents
correlation data from capillary blood samples at approximately
21.degree. C., the "ven/18C" data set represents correlation data
from venous blood samples at approximately 18.degree. C., and the
"all" data set represents the overall correlation data from these
two samples, as well as capillary blood samples at approximately
18.degree. C. and venous blood samples at approximately 21.degree.
C. FIG. 5B is a similar graph of the correlations of .DELTA.S for
the data of Table 1 as a function of the complex index values
obtained from Equation 12. The differences between the overall
correlation ("all") and the individual correlations at different
temperatures are much smaller for the complex index function of
FIG. 5B (R.sup.2=0.77) than for the R4/3 ratio index function of
FIG. 5A (R.sup.2=0.64). Although the approximately 0.13 difference
between these R.sup.2 values is numerically small, it represents a
13% improvement in the correlation between .DELTA.S and the complex
index function in relation to the R4/3 ratio index function.
Consequently, the biosensor may use a single predictor function,
represented as Equation 13 below, to compensate for all four cases
of capillary and venous blood samples at the 21.degree. C. and
18.degree. C. temperatures.
.DELTA.S=1.0043*.DELTA.S.sub.RegB+0.1308 (Equation 13).
In Equation 13, .DELTA.S.sub.RegB, the complex index function, is
as represented in Equation 12 and the 1.0043 and 0.1308 values are
b.sub.1 and b.sub.0 (from the FIG. 5B plot of "all" data),
respectively, as previously described with regard to Equation 8,
for example.
[0094] Using one or more complex index function responsive to
.DELTA.S may reduce the bias spread, which is measured by the
standard deviation of the combined biases. The smaller the standard
deviation of the combined biases, the smaller the bias spread, and
the more accurate and/or precise the analysis of the analyte in the
sample. The effectiveness of the compensation at improving the
measurement performance of an analysis is directly related to the
correlation between .DELTA.S and one or more index functions, which
directly affects the reduction of the standard deviation (SD) of a
bias population. The correlation between .DELTA.S and one or more
index or predictor functions may be measured by the correlation
coefficient R.sup.2. Therefore, the higher the R.sup.2 value the
better the correlation between .DELTA.S and one or more index or
predictor functions, the larger the reduction of the SD value for
the combined biases, and the smaller the bias spread after
compensation. Preferable complex index functions have an R.sup.2
correlation value of about 0.6 and greater with .DELTA.S. More
preferable complex index functions have an R.sup.2 correlation
value of about 0.7 and greater with .DELTA.S. Preferable index or
predictor functions provide SD values of less than 5 for the
combined biases of a data population. Preferable predictor
functions including complex index functions provide SD values of
less than 4 for the combined biases of a data population, and more
preferably SD values of less than 3 for the combined biases of the
data population.
[0095] The empirical relationship between standard deviation and
bias spread is observed in Table 3, below. The mean of the combined
biases, the SD of the combined biases, and the percent of the
concentration analysis (data population) falling within a .+-.10%
combined bias limit before and after R4/3+Temp index and complex
index compensation are listed for the capillary blood samples
previously described with regard to Table 1 and analyzed for
glucose at 21.degree. C. and 18.degree. C. The "R4/3+Temp"
abbreviation is used to describe compensation with a R4/3 index
function and with a temperature index function, as further
discussed with regard to Table 4.
TABLE-US-00003 TABLE 3 Compensation results using R4/3 + Temp and
complex indices. Before R4/3 + Temp Complex Measurement
compensation index index Temperature Performance (G.sub.raw)
compensation compensation 21.degree. C. Mean bias/%-bias -1.03
0.0072 0.329 SD of bias/%-bias 6.315 4.23 3.7 % .+-. 10% 84.9 98.1
99.1 18.degree. C. Mean bias/%-bias -9.29 -2.44 -1.22 SD of
bias/%-bias 6.91 4.71 4.18 % .+-. 10% 55.2 94.7 98.1
[0096] The mean of the combined biases as calculated from the
determined analyte concentrations without compensation (G.sub.raw)
showed that both the 18.degree. C. and the 21.degree. C. data
populations were negatively offset in relation to zero bias. At
21.degree. C., the mean of -1.03 for the combined biases is
believed to be within the error of the biosensor system. However,
at 18.degree. C., the mean of -9.29 for the combined biases is
believed attributable to temperature error. For the lower
temperature 18.degree. C. data, the significantly higher numerical
value of nine indicated that the uncompensated data from the system
was centered at the lower boundary of a .+-.10 combined bias limit,
significantly away from the center of zero bias. Thus, about half
of the data population was outside the boundary of a .+-.10
combined bias limit.
[0097] For the 21.degree. C. data set, R4/3+Temp index function
compensation provided a reduction of greater than two units
(6.315-4.23=2.085) in standard deviation. This greater than two
unit reduction is significant, as on average, a standard deviation
of 5 units or less will place about 95% of the data within a
.+-.10% combined bias limit and about 63% of the data within a
.+-.5% combined bias limit. Thus, the R4/3+Temp index function
compensation brought about 98% of the 21.degree. C. data within the
boundary of a .+-.10% combined bias limit and about 77% of the data
within the boundary of a .+-.5% combined bias limit.
[0098] In relation to R4/3+Temp index function compensation,
complex index function compensation reduced standard deviation by
approximately an additional 0.5 units. Thus, the complex index
compensation brought about 99% of the data within the boundary of
a.sub.+10% combined bias limit and about 88% of the data within the
boundary of a .+-.5% combined bias limit. While the improvement for
complex index compensation in relation to R4/3+Temp index function
compensation is not as great for this data as observed for
R4/3+Temp index function compensation in relation to no
compensation, the resistance of the system to perturbation when the
data set is less centered (larger mean of the combined biases) is
significantly increased.
[0099] Perturbation resistance may be thought of as how well a
system provides accurate and/or precise analyte concentration
values when errors are present in the analysis. Perturbation
resistance is determined by subtracting twice the standard
deviation from 10 to provide a perturbation resistance indicator
(PRI). For the 21.degree. C. data population in Table 3, the PRI is
1.54 (10-2*4.23) for the R4/3+Temp index function compensation and
2.6 (10-2*3.7) for the complex index function compensation. As the
uncompensated 21.degree. C. data is substantially centered with a
numerical mean of one, the approximately 68% increase in the PRI
provided by the complex index function compensation in relation to
the R4/3+Temp index function compensation moves one additional
percent of the data within the boundary of a .+-.10% combined bias
limit.
[0100] However, when the system is perturbed by an error causing a
spread in the uncorrected data, as was observed as a numerical
increase in the mean of the combined biases for the 18.degree. C.
data, the benefit provided by complex index function compensation
significantly increased. For the perturbed 18.degree. C. data, the
standard deviation was reduced by 2.2 units through R4/3+Temp index
function compensation and was further reduced by approximately 0.5
unit through complex index function compensation. Thus, complex
index function compensation provided an approximately 0.5 unit SD
reduction in standard deviation at both temperatures in relation to
R4/3+Temp index function compensation. This translates into complex
index function compensation having an enhanced ability to bring
high bias data into an acceptable range in relation to compensation
by the R4/3+Temp index function.
[0101] When the PRI values were determined for the 18.degree. C.
data, R4/3+Temp index function compensation provided a value of
0.58, while complex index function compensation provided a value of
1.64, an approximately 180% increase in the PRI for complex over
R4/3+Temp index function compensation. While the 68% increase in
the PRI provided by complex index function compensation moved an
additional 1% of the closely grouped 21.degree. C. data within the
boundary limit of a .+-.10% combined bias limit, the 180% increase
in the PRI provided by complex index function compensation moved
over three times as much (3.4%) of the numerically higher mean
18.degree. C. data within the boundary limit of a .+-.10% combined
bias limit. Thus, the greater the error in the uncompensated data,
the better complex index function compensation performed at
reducing bias to within the boundary of a .+-.10% combined bias
limit.
[0102] Complex index function compensation provided an
approximately 17% (99.1-84.9/84.9*100%) increase in the percentage
of data points within the boundary of a .+-.10% combined bias limit
in relation to the uncompensated data points at the higher
21.degree. C. temperature and an approximately 78%
(98.1-55.2/55.2*100%) increase in the percentage of data points
within the boundary of a .+-.10% combined bias limit in relation to
the uncompensated data points at the lower 18.degree. C.
temperature. While the difference between the R4/3+Temp index
function and the complex index function corrections was not as
large for this substantially centered uncorrected data, the
improvement provided by complex index function correction is
significant as fewer analyses would be outside of the boundary of a
.+-.10% combined bias limit. By reducing the number of readings
outside of the bias limit, more of the readings obtained could be
used for accurate therapy by a patient when blood glucose is being
monitored, for example. Additionally, the need to discard and
repeat the analysis by the patient also may be reduced.
[0103] FIG. 6A is a graph of the capillary and venous blood samples
previously discussed in relation to Table 1 at 21.degree. C. for
the correlations of .DELTA.S with the R4/3 index values. FIG. 6B is
a graph of the correlations of .DELTA.S for the same data with the
complex index values of Equation 12. The R.sup.2 values for these
graphs were 0.5998 and 0.7268, respectively, indicating an
approximately 21% (0.7269-0.5998/0.5998) improvement in the
correlation of the complex index function to .DELTA.S in relation
to the correlation of the R4/3 index function to .DELTA.S.
Similarly, FIG. 6C and FIG. 6D plot the correlations to .DELTA.S
for capillary and venous blood samples at 18.degree. C. with the
R4/3 index values (FIG. 6C), and with the complex index values from
Equation 12 (FIG. 6D). Comparison of the R.sup.2 values for the
R4/3 and complex index functions (0.6307 and 0.7154, respectively)
shows an approximately 13.5% (0.7154-0.6307/0.6307) improvement in
the correlation of the complex index function to .DELTA.S in
relation to the correlation of the R4/3 index function to
.DELTA.S.
[0104] The slope deviation, .DELTA.S, and/or related complex index
functions may be normalized to represent the %-bias in the
correlation of analyte concentrations with output signals. In
normalization, the slope deviation, index or complex index
function, or other parameter is adjusted (multiplied, divided, or
the like) by a variable to reduce the statistical effect of changes
in the parameter, improve the differentiation in variations of the
parameter, standardize measurements of the parameter, a combination
thereof, or the like.
[0105] Table 4, below, compares determined raw glucose
concentrations with compensated glucose concentrations resulting
from a R4/3+Temp index function compensation and from a complex
index function compensation including temperature in the terms. The
percent of the data from the donor study previously discussed with
regard to Table 1 falling within +10%, +8%, and +5% combined bias
limits was determined in addition to the standard deviation (SD)
for the combined biases of the data population. The samples were
analyzed using a gated amperometric input signal where selected
intermediate output signals were recorded from the pulses.
TABLE-US-00004 TABLE 4 Compensation with R4/3 + Temp and Complex
Index Functions Un-comp R4/3 + Temp index functions Complex index
function including T Cap/21 C. Cap/21 C. Ven/22 C. Cap/18 C. Ven/18
C. all Cap/21 C. Ven/22 C. Cap/18 C. Ven/18 C. all % in .+-.10%
84.9 98.1 96.2 96.2 94.3 96.2 99.1 99.0 97.2 98.1 98.3 % in .+-.8%
93.4 94.3 85.8 91.5 91.3 96.2 98.1 91.5 93.4 94.7 % in .+-.5% 77.4
84.0 66.0 62.3 72.3 87.7 87.6 76.4 78.3 82.5 SD %-bias 4.23 4.01
5.23 4.72 3.62 3.47 4.57 4.13
[0106] R4/3+Temp index function compensation was performed with the
predictor function f(predictor)=a.sub.1*R4/3+a.sub.0, determined by
comparing .DELTA.S.sub.cal (observed from the recorded current
values) with R4/3, where a.sub.1 and a.sub.0 represent a slope and
intercept, respectively. The temperature sensitivity .DELTA.S.sub.T
of the data also was determined using the relationship:
.DELTA.S.sub.T=f(Index).sub.Temp=c.sub.1*T+c.sub.0 (Equation
14),
where f(Index).sub.Temp is as previously described, T is
temperature, and c.sub.1 and c.sub.0 represent a slope and
intercept, respectively.
[0107] The corrected glucose concentration was then determined
using the relationship:
G.sub.corr=(i-Int)/(S.sub.cal+.DELTA.S.sub.T+f(predictor))
(Equation 15),
where i is a value of the output signal from a biosensor system,
Int is the intercept from a reference correlation equation, and
G.sub.corr is the corrected glucose concentration of the
sample.
[0108] The percentage of the data points (corrected glucose sample
concentrations) falling within the boundary of a .+-.10%, .+-.8%,
or .+-.5% combined bias limit was determined through the
relationship G.sub.corr-G.sub.ref if G.sub.ref was less than 75 mg
of glucose per deciliter (mg/dl) of sample, where G.sub.ref is the
reference glucose concentration of the sample as determined by a
YSI reference instrument. The relationship
100%*(G.sub.corr-G.sub.ref)/G.sub.ref was used to determine the
percentage of the corrected glucose sample concentrations falling
within the boundary limits when the data point was greater than or
equal to 75 mg/dl.
[0109] Complex index function compensation was performed with error
parameters determined from the intermediate currents from the
samples, temperature values, and G.sub.raw by selecting terms,
constants, and weighing coefficients as previously described.
p-values were used to perform the exclusion test for the terms to
include in the complex index function f(CIndex). .DELTA.S.sub.cal
was then compared with the f(CIndex) to obtain
.DELTA.S.sub.cal=b.sub.1*f(CIndex)+b.sub.0, where b.sub.1 and
b.sub.0 represent slope and intercept, respectively. When b.sub.1
is approximately one and/or b.sub.0 is approximately zero, the
f(CIndex) may approximate .DELTA.S without one or both of these
modifications. The percentage of the data points (corrected glucose
concentrations of each sample) falling within the boundary of a
.+-.10%, +8%, or .+-.5% combined bias limit was determined as
previously for the R4/3+Temp index function compensation.
[0110] When the percentage of analyte concentrations falling within
the boundary of the narrowest .+-.5% combined bias limit are
considered, R4/3+Temp index function compensation placed
approximately 72% of "all" the samples within the boundary, while
complex index function compensation placed approximately 82% of the
samples within the boundary. This represents an approximately 14%
(82-72/72*100) increase in the total number of corrected analyte
concentration values falling within the narrowest .+-.5% combined
bias limit. This significant increase in measurement performance
provided by the complex index function compensation in relation to
the R4/3+Temp index function compensation was observed even though
both methods included a compensation for the temperature
differences. Thus, at a measurement performance cut-off of a .+-.5%
combined bias limit, a patient would have to discard and repeat
approximately 14% fewer analysis from a glucose biosensor system
using complex index function compensation than from the same
glucose biosensor system using R4/3+Temp index function
compensation. The same glucose biosensor system lacking
compensation would require approximately 56% of the glucose
analyses to be discarded at the .+-.5% combined bias limit,
rendering the uncompensated system effectively useless for
achieving the measurement performance cut-off of a .+-.5% combined
bias limit. A significant decrease (.about.0.6 units on average) in
the standard deviation of the bias for each of the four individual
data populations between the R4/3+Temp and complex index function
compensations was observed.
[0111] FIG. 6E depicts a graph of hematocrit sensitivity in
combined bias vs. % Hct. In relation to the uncompensated
determined glucose concentrations, the complex index function
compensation reduced hematocrit sensitivity from about -1.11
(bias/% bias)/% Hct to about -0.3 (bias/% bias)/% Hct, an
approximately 70% reduction. Thus, complex index function
compensation substantially reduced the susceptibility of the
analysis system to reductions in measurement performance from
hematocrit bias.
[0112] In addition to .DELTA.S, an index function may represent
.DELTA.S/S, a normalized form of slope deviation. Thus, .DELTA.S/S
may be substituted for .DELTA.S. Normalization may be achieved
through the relationships .DELTA.S/S.sub.cal or S/S.sub.cal, for
example. As such, the slope deviation, .DELTA.S, in Equation 7 may
be normalized by the slope of the reference correlation equation,
S.sub.cal, resulting in a compensation correlation between
.DELTA.S/S.sub.cal and the index function.
[0113] In Equation 7, .DELTA.S is divided by S.sub.cal as
follows:
A c o r r = i - Int S c a l + .DELTA. S = i - Int S c a l ( 1 +
.DELTA. S / S c a l ) . ( Equation 16 ) ##EQU00004##
[0114] .DELTA.S/S.sub.cal may be substituted with a predictor
function, f(predictor), which may include a complex index function,
and may be represented as follows:
.DELTA.S/S.sub.cal=f(predictor)=c.sub.1*f(CIndex)+c.sub.0 (Equation
17).
[0115] The predictor function, f(predictor), of Equation 17 may be
substituted into Equation 16 as follows:
A c o r r = i - Int S c a l ( 1 + f ( p r e d i c t o r ) ) = i -
Int S c a l ( 1 + ( c 1 * f ( CIndex ) + c 0 ) ) . ( Equation 18 )
##EQU00005##
[0116] Solving for the slope deviation, .DELTA.S, provides the
following relationship:
.DELTA.S=S.sub.cal*f(predictor)=S.sub.cal*(c.sub.1*f(CIndex)+c.sub.0)
(Equation 19).
[0117] The normalization of the slope deviation, .DELTA.S, by
S.sub.cal may substantially eliminate the potential effect from
different calibrations of S.sub.cal.
[0118] The slope deviation, .DELTA.S, in Equation 7 also may be
normalized by multiplication with a normalized slope function,
S.sub.NML, resulting in a compensation correlation between
S.sub.NML and the complex index function. The normalized slope
function S.sub.NML may be represented as follows:
S N M L = S / S cal = i - I n t A r e f * 1 S cal = f ( predictor )
= d 1 * f ( CIndex ) + d 0 . ( Equation 20 ) ##EQU00006##
[0119] Substituting Equation 20 into Equation 7 and replacing
S.sub.NML with the predictor function, f (predictor), provides the
following relationship:
A c o r r = i - Int S c a l * S N M L = i - Int S c a l * f ( p r e
d i c t o r ) = i - Int S c a l * ( d 1 * f ( CIndex ) + d 0 ) . (
Equation 21 ) ##EQU00007##
[0120] FIG. 1C represents a method of determining a complex index
function from hematocrit-adjusted and donor blood samples for use
in a measurement device. In 122, determine the experimental glucose
concentration of multiple hematocrit-adjusted blood samples having
known reference glucose concentrations at multiple environmental
conditions with multiple test sensors. A reference instrument may
be used to determine the known analyte concentrations. In 123,
determine the slope and intercept of a reference correlation for
the multiple test sensors from the determined and known glucose
concentrations at a reference temperature and at a reference % Hct.
In 124, determine the reference glucose concentration of multiple
donor blood samples. The donor blood samples may have varying
glucose concentrations and hematocrit levels. The reference glucose
concentration of multiple donor blood samples may be determined at
a reference temperature. In 125, optionally combine the multiple
hematocrit-adjusted blood sample glucose concentration data with
the multiple donor blood sample glucose concentration data. In 126,
select terms from the data for one or more output signal value.
Terms also may be selected for one or more physical characteristic,
environmental condition, concentration value, and the like. In 127,
determine weighing coefficients for the terms and optional
constants. In 128, select the terms, corresponding weighing
coefficients, and any constants for inclusion in the complex index
function.
[0121] Table 5, below provides determined glucose concentration
data for capillary and venous blood samples (about 106 samples) and
samples that were spiked with venous blood to adjust the hematocrit
content of the samples to about 20 to about 60% Hct (about 60
samples). Thus, hematocrit-adjusted blood samples were prepared as
generally described in 122 of FIG. 1C. Unlike the prior analyte
concentrations determined from the donor study previously discussed
with regard to Table 1, the glucose concentrations of Table 5 were
determined using complex index functions derived from different
blood samples than the blood samples analyzed for glucose. Thus,
the complex index function implemented by the measurement device to
correct bias in Table 5 was previously determined from a different
sample population. A p-value exclusion test was used with an
exclusion value of 0.05 to select terms for inclusion in the
complex index function. After exclusion, the terms remaining in the
complex index function were: R4/3, R5/4, R5/4*G.sub.raw, R/54*Temp,
R4/3*Temp, R4/3*R5/4, R4/3*R5/4*G.sub.raw, R4/3*R5/4*Temp, and
Temp. The complex index function included positive or negative
weighing coefficients for each term and an initial constant.
[0122] A compensation equation was used to determine the corrected
glucose concentrations of the blood samples having the general
form:
G.sub.corr=(i-Int)/(S.sub.cal*(1+f(predictor)) (Equation 22),
where f(predictor)=b.sub.1*f(CIndex)+b.sub.0=.DELTA.S/S, the
normalized form of slope deviation.
[0123] When the predictor function is approximating .DELTA.S/S, the
theoretical value of one may be used in place of b.sub.1 when
b.sub.1=1.+-.0.2, preferably one may be used in place of b.sub.1
when b.sub.1=1.+-.0.15, and more preferably one may be used in
place of b.sub.1 when b.sub.1=1.+-.0.1. When the predictor function
is approximating .DELTA.S/S, the theoretical value of zero may be
used in place of b.sub.0 when b.sub.0=0.+-.0.03, preferably zero
may be used in place of b.sub.0 when b.sub.0=0.+-.0.02, and more
preferably zero may be used in place of b.sub.0 when
b.sub.0=0.+-.0.01. Other deviation cut-offs may be used to
determine when the theoretical values for b.sub.1, b.sub.0, or both
may be used. In addition to substituting b.sub.1 and/or b.sub.0
with the theoretical values 1 and 0, predetermined values from a
look-up table and the like may be substituted based on the same or
other deviation cut-offs.
[0124] For this data population, b.sub.1 was 1.08 and b.sub.0 was
0.012. Thus, b.sub.1 was estimated at 1, and b.sub.0 was estimated
at 0. Removing b.sub.1 and b.sub.0 from the equation provides the
relationship as follows:
G.sub.corr=(i-Int)/(S.sub.cal*(1+f(CIndex)) (Equation 23).
[0125] Thus, an output current value responsive to sample glucose
concentration was converted into a corrected glucose concentration
of the sample using a complex index function representing
.DELTA.S/S. Alternatively, a corrected glucose concentration value
may be determined from an uncorrected glucose concentration value
using a complex index function and an equation having the general
form as follows:
G.sub.corr=G.sub.raw/(1+f(CIndex)) (Equation 24).
TABLE-US-00005 TABLE 5 Comparison of f(CIndex) Compensated and
Uncompensated Analyses Cap/20 C., Cap + Ven/23 C., Ven + Result
natural spiked natural spiked %-within +/-10% after 99.1 97.1 98.0
96.3 %-within +/-8% after 97.3 95.3 95.1 93.9 Mean %-bias after
-0.772 -0.249 0.49 0.566 SD of %-bias after 3.61 4.63 4.39 5.0 SD
before compensation 5.35 9.3 5.99 8.9 %-within +/-10% before 84.5
67.8 70.6 58.3
[0126] For the samples including the artificially extended (from
30-50% to about 20-60%) hematocrit range, complex index function
correction brought at least 96% of the determined analyte
concentrations within the boundary of a .+-.10% combined bias limit
and almost 94% of the determined analyte concentrations within the
.+-.8% combined bias limit. This is a significant improvement in
relation to the uncompensated analyses where only about 58% of the
spiked venous samples fell within the boundary of the .+-.10%
combined bias limit, a greater than 60% improvement (96-58/58*100).
The standard deviation for the combined biases of each of the four
data populations also decreased by at least 1.5 units for the
corrected concentration values in relation to the uncorrected
concentration values. The greater accuracy and precision of the of
the compensated analyte concentrations in relation to the
uncompensated analyte concentrations is shown by the closer
grouping around the zero combined bias line of FIG. 6F. These
results establish that complex index functions are transferable
between different samples and may be determined in the laboratory
for later use in the measurement device.
[0127] FIG. 7 depicts a schematic representation of a biosensor
system 700 that determines an analyte concentration in a sample of
a biological fluid. Biosensor system 700 includes a measurement
device 702 and a test sensor 704, which may be implemented in any
analytical instrument, including a bench-top device, a portable or
hand-held device, or the like. The measurement device 702 and the
test sensor 704 may be adapted to implement an electrochemical
sensor system, an optical sensor system, a combination thereof, or
the like. The biosensor system 700 adjusts a correlation for
determining analyte concentrations from output signals with at
least one .DELTA.S value. The .DELTA.S adjusted correlations may
improve the measurement performance of the biosensor system 700 in
determining the analyte concentration of the sample. The biosensor
system 700 may be utilized to determine analyte concentrations,
including those of glucose, uric acid, lactate, cholesterol,
bilirubin, and the like. While a particular configuration is shown,
the biosensor system 700 may have other configurations, including
those with additional components.
[0128] The test sensor 704 has a base 706 that forms a reservoir
708 and a channel 710 with an opening 712. The reservoir 708 and
the channel 710 may be covered by a lid with a vent. The reservoir
708 defines a partially-enclosed volume. The reservoir 708 may
contain a composition that assists in retaining a liquid sample
such as water-swellable polymers or porous polymer matrices.
Reagents may be deposited in the reservoir 708 and/or the channel
710. The reagents may include one or more enzymes, binders,
mediators, and like species. The reagents may include a chemical
indicator for an optical system. The test sensor 704 also may have
a sample interface 714 disposed adjacent to the reservoir 708. The
sample interface 714 may partially or completely surround the
reservoir 708. The test sensor 704 may have other
configurations.
[0129] In an optical sensor system, the sample interface 714 has an
optical portal or aperture for viewing the sample. The optical
portal may be covered by an essentially transparent material. The
sample interface may have optical portals on opposite sides of the
reservoir 708.
[0130] In an electrochemical system, the sample interface 714 has
conductors connected to a working electrode and a counter
electrode. The electrodes may be substantially in the same plane or
in more than one plane. The electrodes may be disposed on a surface
of the base 706 that forms the reservoir 708. The electrodes may
extend or project into the reservoir 708. A dielectric layer may
partially cover the conductors and/or the electrodes. The sample
interface 714 may have other electrodes and conductors.
[0131] The measurement device 702 includes electrical circuitry 716
connected to a sensor interface 718 and a display 720. The
electrical circuitry 716 includes a processor 722 connected to a
signal generator 724, an optional temperature sensor 726, and a
storage medium 728.
[0132] The signal generator 724 provides an electrical input signal
to the sensor interface 718 in response to the processor 722. In
optical systems, the electrical input signal may be used to operate
or control the detector and light source in the sensor interface
718. In electrochemical systems, the electrical input signal may be
transmitted by the sensor interface 718 to the sample interface 714
to apply the electrical input signal to the sample of the
biological fluid. The electrical input signal may be a potential or
current and may be constant, variable, or a combination thereof,
such as when an AC signal is applied with a DC signal offset. The
electrical input signal may be applied as a single pulse or in
multiple pulses, sequences, or cycles. The signal generator 724
also may record an output signal from the sensor interface as a
generator-recorder.
[0133] The optional temperature sensor 726 determines the
temperature of the sample in the reservoir of the test sensor 704.
The temperature of the sample may be measured, calculated from the
output signal, or assumed to be the same or similar to a
measurement of the ambient temperature or the temperature of a
device implementing the biosensor system. The temperature may be
measured using a thermister, thermometer, or other temperature
sensing device. Other techniques may be used to determine the
sample temperature.
[0134] The storage medium 728 may be a magnetic, optical, or
semiconductor memory, another storage device, or the like. The
storage medium 728 may be a fixed memory device, a removable memory
device, such as a memory card, remotely accessed, or the like.
[0135] The processor 722 implements the analyte analysis and data
treatment using computer readable software code and data stored in
the storage medium 728. The processor 722 may start the analyte
analysis in response to the presence of the test sensor 704 at the
sensor interface 718, the application of a sample to the test
sensor 704, in response to user input, or the like. The processor
722 directs the signal generator 724 to provide the electrical
input signal to the sensor interface 718. The processor 722
receives the sample temperature from the temperature sensor 726.
The processor 722 receives the output signal from the sensor
interface 718. The output signal is generated in response to the
reaction of the analyte in the sample. The output signal may be
generated using an optical system, an electrochemical system, or
the like. The processor 722 determines .DELTA.S compensated analyte
concentrations from output signals using a correlation equation as
previously discussed. The results of the analyte analysis may be
output to the display 720 and may be stored in the storage medium
728.
[0136] The correlation equations between analyte concentrations and
output signals may be represented graphically, mathematically, a
combination thereof, or the like. A correlation equation may
include one or more index functions. Correlation equations may be
represented by a program number (PNA) table, another look-up table,
or the like that is stored in the storage medium 728. Constants and
weighing coefficients also may be stored in the storage medium 728.
Instructions regarding implementation of the analyte analysis may
be provided by the computer readable software code stored in the
storage medium 728. The code may be object code or any other code
describing or controlling the functionality described herein. The
data from the analyte analysis may be subjected to one or more data
treatments, including the determination of decay rates, K
constants, ratios, functions, and the like in the processor
722.
[0137] In electrochemical systems, the sensor interface 718 has
contacts that connect or electrically communicate with the
conductors in the sample interface 714 of the test sensor 704. The
sensor interface 718 transmits the electrical input signal from the
signal generator 724 through the contacts to the connectors in the
sample interface 714. The sensor interface 718 also transmits the
output signal from the sample through the contacts to the processor
722 and/or signal generator 724.
[0138] In light-absorption and light-generated optical systems, the
sensor interface 718 includes a detector that collects and measures
light. The detector receives light from the liquid sensor through
the optical portal in the sample interface 714. In a
light-absorption optical system, the sensor interface 718 also
includes a light source such as a laser, a light emitting diode, or
the like. The incident beam may have a wavelength selected for
absorption by the reaction product. The sensor interface 718
directs an incident beam from the light source through the optical
portal in the sample interface 714. The detector may be positioned
at an angle such as 45.degree. to the optical portal to receive the
light reflected back from the sample. The detector may be
positioned adjacent to an optical portal on the other side of the
sample from the light source to receive light transmitted through
the sample. The detector may be positioned in another location to
receive reflected and/or transmitted light.
[0139] The display 720 may be analog or digital. The display 720
may include a LCD, a LED, an OLED, a vacuum fluorescent, or other
display adapted to show a numerical reading. Other displays may be
used. The display 720 electrically communicates with the processor
722. The display 720 may be separate from the measurement device
702, such as when in wireless communication with the processor 722.
Alternatively, the display 720 may be removed from the measurement
device 702, such as when the measurement device 702 electrically
communicates with a remote computing device, medication dosing
pump, and the like.
[0140] In use, a liquid sample for analysis is transferred into the
reservoir 708 by introducing the liquid to the opening 712. The
liquid sample flows through the channel 710, filling the reservoir
708 while expelling the previously contained air. The liquid sample
chemically reacts with the reagents deposited in the channel 710
and/or reservoir 708.
[0141] The test sensor 702 is disposed adjacent to the measurement
device 702. Adjacent includes positions where the sample interface
714 is in electrical and/or optical communication with the sensor
interface 718. Electrical communication includes the transfer of
input and/or output signals between contacts in the sensor
interface 718 and conductors in the sample interface 714. Optical
communication includes the transfer of light between an optical
portal in the sample interface 714 and a detector in the sensor
interface 718. Optical communication also includes the transfer of
light between an optical portal in the sample interface 714 and a
light source in the sensor interface 718.
[0142] The processor 722 receives the sample temperature from the
temperature sensor 726. The processor 722 directs the signal
generator 724 to provide an input signal to the sensor interface
718. In an optical system, the sensor interface 718 operates the
detector and light source in response to the input signal. In an
electrochemical system, the sensor interface 718 provides the input
signal to the sample through the sample interface 714. The
processor 722 receives the output signal generated in response to
the redox reaction of the analyte in the sample as previously
discussed.
[0143] The processor 722 determines the analyte concentration of
the sample. The measurement device adjusts the correlation between
analyte concentrations and output signals with at least one
.DELTA.S value. The analyte concentration is determined from the
slope-adjusted correlation and the output signal. As described
previously, normalization techniques also may be used.
[0144] While various embodiments of the invention have been
described, it will be apparent to those of ordinary skill in the
art that other embodiments and implementations are possible within
the scope of the invention.
* * * * *