Classifications

• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
  • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
  • G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
  • G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
  • G01N33/72—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving blood pigments, e.g. haemoglobin, bilirubin or other porphyrins; involving occult blood
  • G01N33/721—Haemoglobin
  • G01N33/723—Glycosylated haemoglobin
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
  • G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
  • G01N33/48—Biological material, e.g. blood, urine; Haemocytometers
  • G01N33/50—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing
  • G01N33/72—Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing involving blood pigments, e.g. haemoglobin, bilirubin or other porphyrins; involving occult blood
  • G01N33/721—Haemoglobin

Definitions

• the present invention relates generally to a method for measuring the ratio of analytes in a sample.
• the methods are generally applicable to any pair of analytes in which the result is expressed as a ratio. Additionally, multiple ratios can be expressed as ratios of multiple pairs of analytes
• Ratios are frequently used to compensate for variability in the composition or concentration of the analytes in a sample. For example, when measuring analytes present in urine, the concentration of analytes in any given sample can vary significantly. These concentration fluctuations can be normalized by creating a ratio between different analytes present in a sample. For example, a ratio can be created between the concentration of a urinary analyte and the concentration of creatinine in urine. Creating this relationship between two analytes in the sample allows for a more meaningful measurement in light of the concentration fluctuations.
• Ratios can also compensate for correlated errors in the measurement, e.g., dilution errors.
• a specimen is taken from a patient.
• Example specimens include blood, urine, or other bodily fluids.
• the specimen forms the basis for the sample from which a ratio of analytes will be measured and computed.
• the specimen does not necessarily have to come directly from a patient.
• the specimen may be available from another source.
• the specimen is converted into a pre-processed sample.
• a pre- processed sample can also be referred to as a test sample. This step may consist of anything done to the specimen before any analyte measurements are taken.
• the specimen may be sampled and diluted by a specific amount, hi other embodiments, an anti-coagulant may be added to the specimen. This step is optional as there may be times when the entire unaltered specimen will be used as the pre-processed sample.
• Step 103 consists of all the processing done by the instrument to ready the pre-processed sample for reading.
• Processing the sample can involve a variety of steps. Using a multi-analyte sandwich immunoassay as an example, the sample is reacted with antibodies or specific binding reagents attached to a solid phase, the solid phase is washed to remove unbound material, reacted with one or more labeled antibodies or binding reagents, and washed to remove unbound labeled reagent for subsequent reading of the response.
• HPLC high-performance liquid chromatography
• immunoassay electrophoresis
• capillary electrophoresis ultra- violet/visible/infrared spectroscopy
• raman spectroscopy surface enhanced raman spectroscopy
• mass spectroscopy gas chromatography, or others.
• Step 104 the selected reading technique is used to determine the response of each analyte in the sample that is to be used in the ratio.
• Step 104 is referred to as reading the analyte (responses).
• Reading techniques that may be appropriate include: fluorescence, absorbance, reflectance, ion current, electrochemical potential, optical density, color, surface plasmon resonance, or others.
• the processing and reading techniques chosen usually depends on the nature of the sample and analytes to be measured.
• the response of the analytes measured using the chosen measurement method can be in any unit that is appropriate for the selected method. These units could be fluorescence units, optical density, color, ion current, chemiluminescence units, electrical signal, or others.
• the analytes can also be measured in a multiplexed format, where many analytes are measured in a single pass, or they can measured in multiple, but separate individual assays.
• the analyte responses are then converted into their corresponding concentrations. This step is shown at steps 105 and 106. Typically this conversion is accomplished using a device such as a calibration curve.
• Each analyte response must be individually converted into a concentration value before a ratio of the analyte concentrations is calculated.
• a concentration value for an analyte can depend not only on the measured response of the given analyte, but also on the responses of other analytes in the sample because of cross-interactions such as cross- reactivity between analytes.
• the relationship between the response of an analyte in a sample measured using the selected measuring technique and the analyte's concentration in the same sample is typically nonlinear. Because of this non- linear relationship, the ratio of the analyte responses can be quite different than the ratio of the analyte concentrations. Therefore, it is necessary to compute the concentration of each analyte individually before computing the ratio if the ratio of the concentration of the analytes is to be computed.
• a calibration curve is typically used to convert an analyte's response into a concentration value.
• a calibration curve for an analyte can be created by choosing a model for the curve and then determining the values of the model coefficients by measuring analyte responses from a set of samples where the concentration of the analyte is known before any measurement is taken. The analyte responses obtained from these known samples may be graphed to produce a calibration curve that is then used to infer what the concentration of an analyte is in a sample where the concentration is unknown.
• a separate calibration curve has to be created for each analyte response that is to be converted into a concentration value.
• the equation for the model with determined coefficient values may be used directly to determine the concentrations without graphing. Reference to a calibration curve is generally used because it is more illustrative than reference to a calibration equation.
• step 107 the individual concentrations calculated using the calibration curves are used to determine the desired analyte ratio. This can be done through a simple mathematical operation such as division. Ratios of analytes are typically unitless. [0015] It is also sometimes necessary to adjust the ratio obtained in step 107 in order to make the measured ratio match what the ratio would have been using a different measuring technique. This step is shown at step 108. This extra conversion is done because it is often customary for a ratio to be presented as if the ratio was determined using a specific measuring technique, and this specific technique may not be the same technique used in steps 103 and 104. There may be well-known differences between the ratio of analytes obtained through these different measurement techniques, and the values can be converted between the two formats.
• Step 108 is an optional step depending on how the specific analyte ratio is to be ultimately presented. Additional information on standardized ratio measurements can be found in "Proposed changes for reporting HbAIc", IVD Technology (May 2007) and “Implementation of Standardization of Hemoglobin AIc Measurement", Clinical Chemistry, 54:6, 1098-1099 (2008). Both references are hereby incorporated in their entirety for all purposes.
• This known method for determining the ratio of analytes has a number of problems. Interactions between the analytes in the sample can introduce inaccuracies in the known method.
• the response of one analyte can influence and change the response of the second analyte.
• the first antibody used to measure the first analyte response may also interact with the second analyte in the sample.
• This cross-reaction would require a different calibration curve for the first analyte for each different value of the second analyte.
• a single average calibration curve could be created for the first analyte, but this would introduce inaccuracies in the responses obtained for that analyte.
• the second antibody could also react with the first analyte in the sample. This would compound the problem.
• Another problem with the known method is that there is less opportunity to cancel noise that is correlated at the response level. The cancellation is done only during the determination of the ratio of the concentrations.
• the proposed method for determining the ratio of analytes improves on the prior methods of determining the ratio of analytes by no longer directly using individual analyte concentrations as a part of the method.
• the proposed method produces more accurate analyte ratios because the proposed method inherently handles interferences introduced into the readings from interactions between analytes in a sample. It may offer additional advantages by requiring only a single calibration model and better cancellation of noise correlated at the response level.
• the ratio of analytes is determined directly from the responses using a single calibration surface rather than by taking the ratios of concentrations determined from the responses by using a pair of calibration curves.
• the calibration equation underlying the surface could be used without graphing the surface.
• the expression calibration surface is synonymous with the expression response surface.
• the calibration surface has the desired analyte ratio as the dependent y-axis and appropriate combinations of the responses as the independent X 1 and x 2 -axes.
• the combination of the responses used for the xi and x 2 -axes is chosen primarily to provide a simple model that fits the data.
• a good fit reduces the lack-of-fit error and can reduce the number of calibrators needed to maintain the calibration surface.
• a good fit can also increase the time interval between calibrations. Different types of calibration may be done on different intervals. For example, lot specific calibrations may be done by the reagent manufacturer for a specific lot of reagents whereas laboratory calibrations may be done on a shorter interval by a laboratory running patient samples and controls.
• the new method improves on the prior art in a number of ways.
• the new method no longer requires any intermediate conversions of analyte responses to analyte concentrations.
• Devices such as calibration curves are not necessarily specifically tailored to compute analyte ratios, and as a result the use of such devices as an intermediate step often plays a significant factor in many of the inaccuracies in the known prior methods.
• the elimination of intermediate conversions also better handles any noise introduced into the measurement of the analytes.
• the responses of the analytes can be combined in a way to reduce noise common to the analyte responses.
• a combination of responses can be used to advantage as the independent variables in the regression model of the conversion method.
• any interactions between the analytes in a sample that can potentially interfere with measured analyte response are better accounted for in this method.
• This potential source of inaccuracy will also be present in the known samples used to create the regression model, and thus the inaccuracies will inherently be accounted for in the model.
• a pair of calibration curves, each tailored for a specific analyte, cannot handle these interferences and thus cannot handle this source of inaccuracy.
• Another benefit of the proposed method is that if a linear (in the coefficients) regression model is used as is described below in some embodiments, then the regression model allows least-mean- square solutions of simultaneous equations rather than more general methods that minimize the least-mean-square error. This allows for faster and more robust calculation by avoiding the problem of finding a local rather than a global minimum associated with the more general methods.
• the new method is more convenient than the prior method. After the regression model is created, no conversion to intermediate values through devices such as concentration vs. response calibration curves is required because the new method directly uses the responses obtained through the measuring process. Because there are fewer steps involved in the process once the regression model is created, it is more convenient for someone performing the new method to find the desired ratio.
• One embodiment is directed to a method for determining one or more ratios of analytes in a sample.
• the method first measures the responses of two or more analytes in a sample using a selected measuring process to obtain two or more individual analyte responses.
• one or more conversion methods are selected. These conversion methods define a ratio of analytes as a direct function of the individual analyte responses obtained using the selected measuring process.
• one or more ratio of analytes in the sample are computed using the selected one or more conversion methods and using the individual analyte responses obtained using the selected measuring process.
• Another embodiment is directed to an apparatus for measuring the concentration ratio of two or more analytes.
• the apparatus comprises a measuring module capable of measuring the responses of analytes in a sample, a memory to store the measured analyte responses from the measuring module, a computer readable medium containing computer readable code having instructions for executing a conversion method that defines the analyte ratio of analytes as a direct function of the individual responses of the analytes obtained from the measuring module, and a processor to execute the computer readable code on the computer readable medium in order to calculate the concentration ratio of analytes using the conversion method.
• Another embodiment is directed to a method for calculating a conversion method that defines the ratio of the analyte values of the analytes as a direct function of the individual responses of the analytes obtained from a measuring process.
• the method comprises selecting a statistically sufficient number of appropriate calibrators to be used to compute the conversion method, wherein said calibrators have a known ratio of the analytes, selecting a measuring process to be used to measure the responses of the analytes in the calibrators, measuring the responses of the analytes in the calibrators using the measuring process, selecting a regression model to be used to calculate the conversion method, and using the regression model to yield the conversion method wherein the regression model uses the responses of the analytes as the independent variables and the known analyte ratios as the dependent variable.
• Another embodiment is directed to a computer-readable medium comprising code for measuring the responses of the analytes in a test sample using a selected measuring process, code for selecting a conversion method that defines the ratio of analytes as a direct function of the individual responses of the analytes obtained from the selected measuring process, and code for computing the ratio of the analytes of the test sample using the selected conversion method and the measured analyte responses from the test sample.
• Fig. 1 is a flow chart depicting the known method for calculating the ratio of two generic analytes in a sample.
• Fig. 2 is a flow chart depicting the proposed method for calculating the ratio of two generic analytes in a sample.
• Fig. 3 is a block diagram of a computer system that can be used to execute the present invention.
• Fig. 4 is a flow chart depicting the proposed method for calculating %Alc through a HPLC measurement method.
• Fig. 4 a is a flow chart depicting the proposed method for calculating %Alc through a multi-analyte immunoassay measurement method.
• Fig. 5 is a chromatogram produced by an HPLC measurement method.
• Fig. 6a is a contour plot giving a representation of the calibration surface for Variation 1 of the application of the Calibration Surface method to responses obtained by HPLC.
• Fig. 6b is a parametric representation of the Calibration Surface of Fig. 6a.
• Fig. 6c is a difference plot corresponding to Fig. 6a, in which the difference between the determined %A lc value and the known value is plotted against the known value.
• Fig. 7a is a contour plot giving a representation of the calibration surface for Variation 2 of the application of the Calibration Surface method to responses obtained by HPLC.
• Fig. 7b is a parametric representation of the Calibration Surface of Fig. 7a. Independence from the parameter value results in a single line.
• Fig. 7c is a difference plot corresponding to Fig. 7a, in which the difference between the determined %A lc value and the known value is plotted against the known value.
• Fig. 8a is a contour plot giving a representation of the calibration surface for Variation 3 of the application of the Calibration Surface method to responses obtained by HPLC.
• Fig. 8b is a parametric representation of the Calibration Surface of Fig. 8a.
• Fig. 8c is a difference plot corresponding to Fig. 8a, in which the difference between the determined %A lc value and the known value is plotted against the known value.
• Fig. 9a is a contour plot giving a representation of the calibration surface for Variation 4 of the application of the Calibration Surface method to responses obtained by HPLC.
• Fig. 9b is a parametric representation of the Calibration Surface of Fig. 9a.
• Fig. 9c is a difference plot corresponding to Fig. 9a, in which the difference between the determined %A lc value and the known value is plotted against the known value.
• Fig. 10a is a contour plot giving a representation of the calibration surface for Variation 5 of the application of the Calibration Surface method to responses obtained by a multi-analyte immunoassay.
• Fig. 10b is a parametric representation of the Calibration Surface of Fig. 10a.
• Fig. 10c is a difference plot corresponding to Fig. 10a, in which the difference between the determined %Ai c value and the known value is plotted against the known value.
• Fig. 11a is a contour plot giving a representation of the calibration surface for Variation 6 of the application of the Calibration Surface method to responses obtained by a multi-analyte immunoassay.
• Fig. 11 b is a parametric representation of the Calibration Surface of Fig. 11a.
• Fig. lie is a difference plot corresponding to Fig. 11a, in which the difference between the determined %A ⁇ C value and the known value is plotted against the known value.
• analytes Disclosed herein according to various embodiments are improved methods for calculating the ratio of analytes that can be implemented in a variety of embodiments tailored for a wide array of analytes.
• potential analytes for different embodiments can include proteins, peptides, carbohydrates, small molecules, nucleic acids, or any other material that can be measured and then expressed as a ratio.
• Fig. 2 illustrates the general steps involved in calculating the concentration ratio of two analytes according to one embodiment. As will be clear from this description, other embodiments of this proposed method can be easily altered to produce other types of analyte ratios by creating and using a different conversion method in the process.
• Steps 201-204 are essentially identical to steps 101-104 displayed in Fig. 1.
• a patient specimen is obtained.
• this patient specimen is converted into a pre-processed or test sample.
• steps 203 and 204 comprise the measuring process.
• the sample is processed for reading at step 203 and read at step 204 producing a response for each analyte in the sample.
• some of the above steps may be optional depending on the nature of the specimen and measuring technique used to measure analyte responses.
• the selected measuring technique is used to measure the response of the analytes in the sample.
• the considerations for selecting a measuring technique in this step are similar to considerations at steps 103 and 104.
• the reading of the analyte responses in step 204 can be accomplished through a variety of techniques, such as by quantifying the fluorescence from fluorescent tags associated with either bound or unbound antibodies in the sample.
• the resulting responses from the selected measurement technique are listed as Analyte Response 1 and Analyte Response 2 in Fig. 2.
• the differences between the known methods for calculating the ratio of analytes and the proposed method become clear. In the embodiment shown in Fig.
• this step uses a conversion method to directly convert the measured analyte responses to the desired analyte ratio.
• the conversion method is created before the process outlined in Fig. 2 began.
• the conversion method may be composed of a series of steps, as described further below. Multiple conversion methods can be used on the measured analyte response to obtain multiple analyte ratios. After the conversion is used to convert the analyte responses to the desired analyte ratio, the process is complete.
• the conversion method in the embodiment of Fig. 2 can be any mathematical relationship that converts the analyte responses to the analyte ratio.
• the conversion method consists of two steps.
• the first step combines or otherwise transforms the analyte responses to obtain two or more regression variables.
• the second step uses the regression variables in a regression model to obtain the desired analyte ratio.
• the regression model may be a multiple linear regression model.
• the regression model may also use more complicated regression models.
• the first step functionally combines the read analyte responses to obtain the two or more regression variables.
• a first regression variable may be set to be the ratio of two analyte responses and a second regression variable may be set to be one of the analyte responses standing on its own.
• Many other possible transformations are possible, and examples of other transformations are given in more detail later in this disclosure.
• the second step of this embodiment of the conversion method uses a regression model to find the desired ratio as a function of the two regression variables.
• the dependent variable in the regression is the desired analyte ratio.
• This analyte ratio could be the concentration ratio or a ratio mathematically related to the concentration ratio, but it need not be.
• the transformation and regression model need to be created and the value of their coefficients determined before being used to determine the ratio of analytes in an unknown sample.
• This transformation is used together with a regression model to determine the desired analyte ratio. It can incorporate step 108 of Fig. 1 and thus bypasses the need to calculate the ratio of concentrations as shown in Fig. 1.
• One purpose of the transformation is to reduce the number of regression terms needed to adequately fit the data. In some embodiments only a constant term, linear terms and possibly the product of the linear terms will be needed.
• the combined transformation and regression coefficients are grouped into levels, defined by the time interval separating the determination of the coefficient values.
• a three level grouping could consist of the following three levels: 1) Interval is determined by a user calibration schedule, 2) Interval is defined by the release of new reagent lots by the reagent manufacturer, 3) Interval is determined by a method or model change.
• [0066] Generate a sample set covering the clinical range of HbAIc values and the range of total Hb concentrations to be accommodated by the method.
• the sample set is to be a grid of equally spaced values covering the range.
• Such a grid can be obtained by starting with a low and a high value and performing a set of dilutions.
• a 5 x 5 grid would be suitable if the transformation and regression model are not complex while a 9 x 9 grid would be suitable if they are complex. It may not be necessary to use a series of dilutions as outlined in this step if samples covering the desired concentration range are otherwise available.
• Fig. 3 is a block diagram of a computer system 300 that can be used to execute one embodiment of the invention.
• the computer system 300 has a number of input modules.
• An Analyte Measurement Module 301 is used to measure the analyte responses in a test sample. This module will vary between different embodiments of the invention depending on the measurement method selected to measure the analyte responses.
• a standard keyboard 302 and mouse 303 are also shown.
• the computer system 300 also contains a variety of typical computer components inside the computer. These components include a system bus 304, one or more disk drives 305, RAM 306, and a processor 307. Other components can also be present depending on the exact nature of the embodiment.
• Fig. 3 also shows a monitor 308 that allows information to be displayed to a user of the system.
• the pre-processed sample that was created as in step 202 is placed in the Analyte Measurement Module 301 where the sample is further processed and the analyte responses in the sample are measured.
• This information is then transferred into the computer system along a system bus 304, and an appropriate conversion method is applied to the analyte response data using the processor 307.
• the instructions the processor 307 executes to implement the conversion method are stored on a computer readable medium such as the RAM 306 or disk drive 305.
• the conversion method can also be stored on this same media.
• the resulting ratio can then be displayed on the monitor 308.
• Alternative embodiments of the invention can output the analyte ratio using other communications means.
• the computer system could print the measured ratio using a printer or send the measured ratio to another computer over a network.
• HbAIc is one specific type of glycosylated hemoglobin that is the most commonly measured in diabetics.
• the human red blood cell allows glucose to freely pass into it, and glycosylated hemoglobin HbAIc refers to the hemoglobin components that has formed via the attachment of glucose sugars to the N-terminal valine of the beta chain of the hemoglobin molecule.
• glycosylated hemoglobin HbAIc refers to the hemoglobin components that has formed via the attachment of glucose sugars to the N-terminal valine of the beta chain of the hemoglobin molecule.
• non-diabetics over 90% of total hemoglobin is non-glycosylated.
• HbAIc itself will typically constitute about 4-7% of total hemoglobin.
• the percentage concentration of HbAIc can be in excess of 20% of total hemoglobin in patients that have poorly controlled their blood glucose levels.
• Glycosylated hemoglobin is formed at a rate that is directly proportional to the ambient glucose concentration in the blood.
• the reaction between glucose and hemoglobin is ultimately irreversible and slow.
• the result is that only a fraction of total hemoglobin is glycated during the 120 day lifespan of a human red blood cell.
• the concentration of glycosylated hemoglobin in the blood provides a weighted moving average of blood glucose levels that can be used to monitor long-term blood glucose levels. This moving average, covering a 2-3 month time period, allows for an assessment of glycemic control in a diabetic patient.
• the ratio of the concentration of HbAIc to total hemoglobin has become a common method for measuring the health status for patients with diabetes.
• HbAlc/mol total Hb International Federation of Clinical Chemists (EFCC) units of mmole HbAlc/mol total Hb. In addition to the factor of 10 difference in units, there is also a difference because of the different reference methods used in the two standards.
• Other hemoglobin variant proteins, such as HbS, HbC, HbE, Hb A2, and HbF can also be expressed as a percentage of total hemoglobin.
• One embodiment of the proposed method uses an HPLC technique for measuring HbAIc.
• Fig. 4 illustrates the steps involved in calculating the ratio of HbAIc to total hemoglobin according to an HPLC method of measuring HbAIc according to one embodiment.
• Steps 401-404 are essentially identical to steps 101-104 displayed in Fig. 1 and steps 201-204 in Fig. 2.
• a patient specimen is obtained.
• this patient specimen is converted into a pre-processed sample, and at step 403, HPLC is used to separate this pre-processed sample into its component parts which elute from the chromatography column as a time series of peaks.
• the components pass sequentially through the flow cell of an absorbance detector which converts the concentrations of the eluents to absorbance values that can be visualized as a chromatogram.
• the peaks in the chromatogram correspond to distinct components in the pre-processed sample.
• the areas of the peaks are then determined.
• the response for the glycosylated component of hemoglobin, HbGr is determined as the area of a single peak while the response for the total component, HbTr, is the sum of the areas of selected peaks.
• Table 1 is a tabulation of the response data from a prepared 5 x 5 x 3 sample set.
• the first two columns contain the HbGr and HbTr responses in mAU-sec and the third column uses a combination of three digits to identify the sample according to its place in the 5 x 5 x 3 sample grid.
• the first digit of the ID corresponds to the %Alc level
• the second digit of the ID corresponds to the HbTc (Total hemoglobin concentration) level
• the third digit of the ID is the replicate number.
• the HbGr value in Column 1 corresponds to the area of the peak labeled 504, expressed in units of mAU-sec. That peak is the second and larger of two overlapping peaks identified by a single peak crest. The smaller peak in the pair is too small to be identified.
• the HbTr value in Column 2 corresponds to the sum of the areas of the peaks labeled 501, 502, 504 and 505, also expressed in units of mAU- sec. These are all A (Adult) components of hemoglobin. Peak 503 is identified as an F (Fetal) component and thus not included in the sum.
• HbGr represented by zj
• HbTr represented by z ⁇
• the B terms work in a similar manner.
• the transformed data is scaled to prevent round-off error.
• the scaled coefficient values are presented in the same order as they appear in the Model. Scaling is done by dividing the HbGr data by Scale A and the HbTr data by Scale B. The scaling values are presented together with the coefficient values.
• a process referred to as auto-calibration is used. All of the 75 samples in the data matrix are used as calibrators and then as unknowns. The errors thus determined are within run errors. Using a standard technique referred to as "Analysis of the Components of Variance" these within run errors are expressed as variance and divided into components. Two components of interest are replicate error (also referred to as pure error) and lack-of-fit (LOF) errors. The replicate error is an indicator of the quality of the data and the lack-of fit error an indicator of the quality of the transformations and model used to transform the data and model the transformed data. There are two components of the lack-of- fit error.
• Fig. 6a is a contour plot of the resulting calibration surface.
• the contours on the contour plot represent the constant values of %Alc shown with the contour label.
• the "+" symbols are the data values of xj and X 2 .
• Fig. 6b is a parametric plot of the resulting calibration surface.
• the different lines in the parametric plot correspond to constant values of X 2 , corresponding to equally spaced horizontal lines in the contour plot.
• the legend with the plot shows these x 2 -values. From top to bottom, the legend values are in the same order as the lines.
• Fig. 6c is a difference plot in which the difference between the determined value and the known value are plotted against the known value.
• the known value is determined from the sequence of dilutions used to prepare the data set. While the absolute value may not be accurate, the inaccuracy in the relative values is determined only by inaccuracies in the dilutions.
• the two dashed lines horizontal lines are 1.96 standard deviations from the mean. This mean, referred to as bias in Table 2, is negligibly different from 0.
• Fig. 7a is a contour plot and Fig. 7b is a parametric plot of the resulting calibration surface. Since the response is linear in X 1 and there is no X 2 dependence, this parametric plot reduces to a single straight line.
• Fig. 7c is a difference plot as described for Fig. 6c.
• Fig. 8a is a contour plot and Fig. 8b is a parametric plot of the resulting calibration surface. Since the model is linear in X 1 , the lines in Fig. 8b corresponding to different values of X 2 are straight. From top to bottom, the legend values in Fig. 8b are in the same order as the lines. Linearity in X 2 results in equal spacing between lines for the same value of X 1 . The cross term, X 1 X 2 , allows the lines to converge as X 1 decreases.
• Fig. 8c is a difference plot as described for Fig. 6c.
• Variation 3 illustrates how the model coefficients can be divided into coefficients that are determined on different time intervals.
• the transformation coefficients represented in matrixes A and B, can be set during the development of the method and do not have to change thereafter.
• K is the known value of %A lc for calibration (determination of the coefficients) and the determined value of %A lc when (1) is applied with known coefficients.
• K 0 K - ⁇ o ⁇ x 2 ' + ⁇ u x lX2 ' (3)
• X 1 and x 2 axe the new transformed response which differs from xj and X 2 because of changes in the system since the time that the unprimed values were determined.
• Equating the right-hand side of (2) and (4) gives a linear relationship between xi and .
• Equation (5) is a linear estimation to the actual relationship between X 1 and if the same samples were measured under the current conditions when the primed parameters are determined and the original conditions when the unprimed parameters were determined.
• Variation 4 linear in Xi, 2 »iul order X 2 dependence with transformation of variables
• Fig. 9a is a contour plot and Fig. 9b is a parametric plot of the resulting calibration surface.
• the addition of the xjx 2 2 term allows the spacing of the straight lines in Fig. 9b to be different for different pairs of X 2 values. From top to bottom, the legend values in Fig. 9b are in the same order as the lines.
• Fig. 9c is a difference plot as described for Fig. 6c.
• the additional term in the model reduces the lack-of-fit, LOF, error as can be seen by comparing Fig. 8c and Fig. 9c at the expense of introducing extra complexity into the model. Since there are still only two terms that do not contain X 2 , only two calibrators are needed for short term calibration. The . long term stability of the model may, however, be compromised by the additional complexity.
• Table 2 is a compilation of the Components of Variance for the four variations presented in the preceding sections.
• LOF is an abbreviation for lack-of-fit
• Replicate is a component of variance determined from the differences in replicate responses. It is also referred to as Pure Error.
• Variation 1 has a higher replicate error then the other three variations. Variations 3 and 4 use a ratio of responses for variable 1 which cancels much of the error common to both responses. This common error could result from differences in the sample size used for the replicates.
• the lack-of-fit variance for different values of %A ⁇ C was negative and thus reported as negligible.
• the lack-of-fit error for different values of the total hemoglobin concentration, HbTc was the most significant factor.
• Variation 1 is not only complex, but it has the largest value for this error. Variation 3 has a significantly smaller error and because of its simplicity, it might be favored over Variations 4 and 2, which are more complicated.
• FIG. 4a illustrates the steps involved in calculating the ratio of HbAIc to total hemoglobin according to a multi-analyte immunoassay method of measuring HbAIc.
• Steps 406-409 are essentially identical to steps 101-104 displayed in Fig. 1 and steps 201-204 in Fig. 2.
• a patient specimen is obtained.
• this patient specimen is converted into a pre-processed or test sample and at step 408, a multi-analyte immunoassay is used to produce a population of labeled beads in which the ratio of beads labeled as glycosylated hemoglobin (HbG) beads to those labeled as total hemoglobin (HbT) beads is related to the concentration of glycosylated (HbGc) and total hemoglobin (HbTc).
• HbG glycosylated hemoglobin
• HbTc total hemoglobin
• a 9x9x3 sample set was used to generate a data set. Since the highest level of %Alc demonstrated a significant amount of saturation it was not used. The sample set was thus reduced to an 8x9x3 sample set. In addition to some remaining saturation at the highest remaining %Alc level, this sample set had somewhat more than twice the replicate error compared with the HPLC data set. In addition to the non-linearities inherent in immunoassays, the data exhibits cross-reactivity between the antibody for one member of the pair and the concentration of the other member. Two variations of the application of the Calibration Surface method will be presented for this data set. The variations will be called Variation 5 and Variation 6. The same definitions and formats used to present the HPLC data will be used.
• Fig. 10a is a contour plot and Fig. 10b is a parametric plot of the resulting calibration surface. From top to bottom, the legend values in Fig. 10b are in the same order as the curves for values OfX 1 less than 0.5.
• Fig. 10c is a difference plot in which the difference between the determined value and the known value are plotted against the known value which is determined by dilution. The two dashed lines horizontal lines are 1.96 standard deviations from the mean.
• Variation 6 Transforms make Model Linear in X 1 and X 2
• A [0.75, 0, 36; 1.5, 0, -500]
• B [0.93, 0, 0; 0, 0, 0]
• Fig. 11a is a contour plot and Fig. lib is a parametric plot of the resulting calibration surface. From top to bottom, the legend values in Fig. lib are in the same order as the lines for values of X 1 less than 0.0003 and in the opposite order for X 1 greater than 0.003. The Xi value where all of the lines cross in Fig. lib corresponds to a vertical contour in Fig. 11a. Fig. lie is a difference plot in which the difference between the determined value and the known value are plotted against the known value.
• Table 3 is a compilation of the Components of Variance for the two variations presented in the preceding sections. The terminology is the same as in Table 2 [0127]

Landscapes

• Health & Medical Sciences (AREA)
• Life Sciences & Earth Sciences (AREA)
• Hematology (AREA)
• Engineering & Computer Science (AREA)
• Molecular Biology (AREA)
• Biomedical Technology (AREA)
• Chemical & Material Sciences (AREA)
• Immunology (AREA)
• Urology & Nephrology (AREA)
• Biotechnology (AREA)
• Microbiology (AREA)
• Cell Biology (AREA)
• Food Science & Technology (AREA)
• Medicinal Chemistry (AREA)
• Physics & Mathematics (AREA)
• Analytical Chemistry (AREA)
• Biochemistry (AREA)
• General Health & Medical Sciences (AREA)
• General Physics & Mathematics (AREA)
• Pathology (AREA)
• Investigating Or Analysing Biological Materials (AREA)
• Analysing Materials By The Use Of Radiation (AREA)

Priority Applications (5)

Applications Claiming Priority (2)

Publications (2)

Family

ID=41697148

Family Applications (1)

Country Status (7)

Cited By (2)

Families Citing this family (11)

Family Cites Families (19)

• 2008
  • 2008-08-21 US US12/196,132 patent/US8260556B2/en not_active Expired - Fee Related
• 2009
  • 2009-08-13 CA CA2734741A patent/CA2734741A1/en not_active Abandoned
  • 2009-08-13 CN CN200980138015.1A patent/CN102159947B/zh not_active Expired - Fee Related
  • 2009-08-13 EP EP09808625A patent/EP2316027A4/en not_active Withdrawn
  • 2009-08-13 AU AU2009282958A patent/AU2009282958A1/en not_active Abandoned
  • 2009-08-13 JP JP2011523882A patent/JP5567567B2/ja not_active Expired - Fee Related
  • 2009-08-13 WO PCT/US2009/053746 patent/WO2010021914A2/en not_active Ceased

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events