Method for measuring the plasma concentration of an analyte directly on a whole blood sample

Abstract

A method of measuring an analyte amount in a whole blood sample, including (i) measuring the haematocrit level of the whole blood sample; (ii) measuring an analyte amount directly in the whole blood sample; and (iii) calculating a corrected analyte amount according to relation D.sub.P=P.sub.a(D.sub.ST, D.sub.H), where D.sub.p, is the corrected analyte amount, D.sub.ST is the measured analyte amount, D.sub.H is the measured haematocrit level, and P.sub.a is a non-constant polynomial of a degree greater than or equal to 1 having as indeterminate values the measured analyte amount, D.sub.ST, and the measured haematocrit level, D.sub.H, and having its polynomial coefficients depending on the analyte.

Claims

1. A method of performing a calibration of the device for measuring an amount of analyte in a whole blood sample, wherein the method comprises: providing a plurality of calibration whole blood samples and performing measurements on the calibration whole blood samples, wherein the measurements include, for each of the plurality of calibration whole blood samples: measuring an haematocrit level in the calibration whole blood sample; measuring an analyte amount in the calibration whole blood sample; and measuring an analyte amount in a plasma sample obtained from the calibration whole blood sample; and calculating polynomial coefficients of a calibrated relation:
D.sub.P=P.sub.a(D.sub.ST,D.sub.H) from values of haematocrit level and analyte amount measured in the plurality of calibration whole blood samples, and values of analyte amount measured in the plasma samples obtained from the plurality of calibration whole blood samples, where D.sub.p is the measured analyte amount in plasma, D.sub.ST is the measured analyte amount in whole blood, D.sub.H is the measured haematocrit level, and P.sub.a is a non-constant polynomial of a degree greater than or equal to 1 having as indeterminate values the measured analyte amount, D.sub.ST, and the measured haematocrit level, D.sub.H, the polynomial coefficients depending on the analyte, wherein, after calibration, the device, upon receiving a measurement of a total analyte amount in a whole blood sample, calculates a corrected analyte amount according to the calibrated relation D.sub.P=P.sub.a(D.sub.ST, D.sub.H), where D.sub.p is the corrected analyte amount, D.sub.ST is the measured analyte amount, and D.sub.H is the measured haematocrit level.

2. The method of claim 1, wherein polynomial P.sub.a comprises product D.sub.ST×D.sub.H of the measured analyte amount D.sub.ST by the measured haematocrit level D.sub.H.

3. The method of claim 1, wherein the analyte amount is measured according to an immunoassay technique of ELISA type, of ELFA type, or of immunocapture type.

4. A method of performing a calibration of the device for measuring an analyte concentration in a whole blood sample, wherein the method comprises: providing a plurality of calibration whole blood samples and performing measurements on the calibration whole blood samples, wherein the measurements include, for each of the plurality of calibration whole blood samples: measuring an haematocrit level in the calibration whole blood sample; measuring an analyte amount in the calibration whole blood sample, and obtaining an analyte concentration in the calibration whole blood sample by dividing the analyte amount measured in the calibration whole blood sample by a volume of the calibration whole blood sample; and measuring an analyte amount in the plasma sample, and obtaining an analyte concentration in a plasma sample from the calibration whole blood sample by dividing the analyte amount measured in the plasma sample by a volume of the plasma sample; calculating a corrected analyte concentration according to a calibrated relation:
C.sub.P=P.sub.a(C.sub.ST,D.sub.H) from values of haematocrit level and analyte concentration measured in the plurality of calibration whole blood samples, and values of analyte concentration measured in the plasma samples from the plurality of calibration whole blood samples, where C.sub.p is the analyte concentration calculated from the measured analyte amount in plasma, C.sub.ST is the analyte concentration calculated from the measured analyte amount in whole blood, D.sub.H is the measured haematocrit level, and P.sub.a is a non-constant polynomial of a degree greater than or equal to 1 having as indeterminate values the measured analyte concentration, C.sub.ST, and the measured haematocrit level, D.sub.H, the polynomial coefficients depending on the analyte, wherein, after calibration, the device, upon receiving a measurement of a total analyte concentration in a whole blood sample, calculates a corrected analyte concentration according to the calibrated relation C.sub.P=P.sub.a (C.sub.ST, D.sub.H), where C.sub.p is the corrected analyte concentration, C.sub.ST is the measured analyte concentration, and D.sub.H is the measured haematocrit level.

5. The method of claim 4, wherein polynomial P.sub.a comprises product C.sub.ST×D.sub.H of the measured analyte concentration C.sub.ST by the measured haematocrit level D.sub.H.

6. The method of claim 4, wherein the analyte concentration is measured according to an immunoassay technique of ELISA type, of ELFA type, or of immunocapture type.

7. A device for measuring the plasmatic amount of an analyte in a whole blood sample, wherein the device comprises means for performing measurements on a plurality of calibration whole blood samples, wherein the measurements include, for each of the plurality of calibration whole blood samples: measuring an haematocrit level in the calibration whole blood sample; measuring a total analyte amount in the calibration whole blood sample; an measuring an analyte amount in a plasma sample obtained from the calibration whole blood sample; and means for calculating polynomial coefficients of a calibrated relation:
D.sub.P=P.sub.a(D.sub.ST,D.sub.H) from values of haematocrit level and analyte amount measured in the plurality of calibration whole blood samples, and values of analyte amount measured in the plasma samples from the plurality of calibration whole blood samples, where D.sub.p is the measured analyte amount in plasma, D.sub.ST is the measured analyte amount in whole blood, D.sub.H is the measured haematocrit level, and P.sub.a is a non-constant polynomial of a degree greater than or equal to 1 having as indeterminate values the measured analyte amount, D.sub.ST, and the measured haematocrit level, D.sub.H, the polynomial coefficients depending on the analyte, wherein, after calibration, the device, upon receiving a measurement of a total analyte amount in a whole blood sample, calculates a corrected analyte amount according to the calibrated relation D.sub.P=P.sub.a(D.sub.ST, D.sub.H), where D.sub.p is the corrected analyte amount in plasma, D.sub.ST is the measured analyte amount, and D.sub.H is the measured haematocrit level.

8. A device for measuring the plasmatic analyte concentration in a whole blood sample, wherein the device comprises means for performing measurements on a plurality of calibration whole blood samples, wherein the measurements include, for each of the plurality of calibration whole blood samples: measuring an haematocrit level in the calibration whole blood sample; measuring an analyte amount in the calibration whole blood sample, and obtaining a total analyte concentration obtained from the total analyte amount measured in the calibration whole blood sample; and measuring an analyte amount in a plasma sample obtained from the calibration whole blood sample, and obtaining an analyte concentration from the analyte amount measured in the plasma sample; and means for calculating polynomial coefficients of a calibrated relation:
C.sub.P=P.sub.a(C.sub.ST,D.sub.H) from values of haematocrit level and analyte concentration measured in the plurality of calibration whole blood samples, and values of analyte concentration measured in the plasma samples from the plurality of calibration whole blood samples, where C.sub.p is the analyte concentration calculated from the measured analyte amount in plasma, C.sub.ST is the analyte concentration calculated from the measured analyte amount in whole blood, D.sub.H is the measured haematocrit level, and P.sub.a is a non-constant polynomial of a degree greater than or equal to 1 having as indeterminate values the measured analyte concentration, C.sub.ST, and the measured haematocrit level, D.sub.H, the polynomial coefficients depending on the analyte, wherein, after calibration, the device, upon receiving a measurement of a total analyte concentration in a whole blood sample, calculates a corrected analyte concentration according to the calibrated relation C.sub.P=P.sub.a(C.sub.ST D.sub.H), where C.sub.p is the corrected analyte concentration in plasma, C.sub.ST is the measured analyte concentration, and D.sub.H is the measured haematocrit level.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be better understood on reading of the following description provided as an example only in relation with the accompanying drawings, where:

(2) FIGS. 1A to 1G are simplified views illustrating a state-of-the-art “sandwich” analyte concentration measurement of ELISA or ELFA type;

(3) FIG. 2 is a flowchart illustrating a method according to the invention;

(4) FIG. 3 is a plot illustrating the concentration and haematocrit level homogeneity of samples used to determine and verify a polynomial correction model applied to D-dimers;

(5) FIG. 4 is a plot illustrating the D-dimer concentration measured in whole blood, with no correction according to the invention, according to the measured D-dimer concentration of the plasma;

(6) FIG. 5 is a plot, for the samples of the verification set, of the relative correction error obtained according to the invention and according to a correction of the state of the art, according to the D-dimer concentration measured in plasma;

(7) FIG. 6 is a plot, for the samples of the verification set, of the D-dimer concentration corrected according to the invention according to the D-dimer concentration measured in plasma;

(8) FIG. 7 is a plot of the D-dimer concentration corrected according to the state of the art according to the D-dimer concentration measured in plasma;

(9) FIG. 8 is plot, for the samples of the determination and verification sets, illustrating the concentration and haematocrit level homogeneity of samples used to determine a polynomial correction model applied in the context of the quantification of cardiac troponin I (TnI);

(10) FIG. 9 is a plot illustrating the TnI concentration measured in whole blood samples, with no correction according to the invention, according to the TnI concentration measured in plasma samples;

(11) FIG. 10 is a plot, for the samples of the verification set, of the relative correction error obtained according to the invention and according to a correction of the state of the art, according to the TnI concentration measured in plasma;

(12) FIG. 11 is a plot of the TnI concentration corrected according to the invention according to the TnI concentration measured in plasma; and

(13) FIG. 12 is a plot of the corrected TnI concentration according to the state of the art according to the TnI concentration measured in plasma.

DETAILED DESCRIPTION

(14) Referring to the flowchart of FIG. 2, a method according to the invention applied to a specific analyte will now be described.

(15) The method comprises a step 10 of determining a mathematical model correcting the measurement of the analyte amount/concentration on a whole blood sample, and a step 12 of using the mathematical model determined at step 10 to deduce an unknown corrected analyte amount/concentration in a total blood sample sampled from a patient or an animal.

(16) The determination of model 10 starts, at 14, with the forming of a set of pairs of whole blood and plasma samples originating from the same batch, for example, from the same patient or animal, with a variable haematocrit level and a variable analyte rate. The samples will be used to determine and verify the mathematical model. Advantageously, the haematocrit variation used for these samples is greater than the range observed for the person or the animal having given the blood, and preferably centered on the observed haematocrit average for the subject or a value close thereto. For example, the normal haematocrit values for man are from 40 to 54%, with an average at 45%, and from 37 to 47%, with an average at 42%, for woman, and the haematocrit level range used to determine and verify the mathematical model is in the range from 26% to 68%. The analyte concentration in the whole blood is selected to vary between a low value characteristic of a healthy subject and a high value prepared by overloading with a maximum 10% by volume of the whole blood sample.

(17) For example, a set of whole blood volumes is formed and each volume is divided by two, the first sub-volume forming the whole blood sample and the second sub-volume being centrifuged to obtain the plasma sample. A whole blood volume may originate from a single subject, from a mixture of a plurality of whole bloods sampled from different subjects, it may be overloaded with analyte to set the analyte concentration, and/or originate from a first volume from which part of the plasma has been removed by centrifugation or to which plasma resulting from a centrifugation has been added to set the haematocrit level.

(18) The method then carries on, at 16, for each pair of samples, by the measurement of the analyte amount or concentration and of the haematocrit level in the whole blood sample, and by measurement of the analyte amount or concentration in the corresponding plasma sample. The amounts are for example measured by means of technique of ELISA, ELFA, or immunocapture type of the state of the art. A set of triplets (D.sub.ST(i), D.sub.P(i), D.sub.H(i)) is thus obtained, each comprising an amount D.sub.ST(i) of analyte in the whole blood, an amount D.sub.P(i) of analyte in the plasma, and a haematocrit level D.sub.H(i) in the whole blood. Of course, this can readily be applied to concentration measurement by replacing analyte amounts with analyte concentrations.

(19) Advantageously, triplets having aberrant values are then discarded, particularly those for which amount D.sub.ST(i) measured in the whole blood is greater than amount D.sub.P(i) measured in the plasma. Here again, this is applicable to concentration measurement.

(20) The obtained set of sample pairs is then divided into two subsets, containing an equal or different number, a first subset being used to determine the mathematical model and a second subset being used to verify the determined mathematical model. The samples used to determine the mathematical model are noted (D.sub.ST.sup.cal(i), D.sub.p.sup.cal(i), D.sub.H.sup.cal(i)) and the samples used to verify the mathematical model are noted (D.sub.ST.sup.verif(i), D.sub.p.sup.verif(i), D.sub.H.sup.verif(i)). For example, two thirds of the pairs of samples are used to determine the mathematical model and one third of the pairs is used for the verification thereof.

(21) As known per se, the measurement technique used depends on the involved analyte, particularly due to the specific binding partners used to fix the analyte, for example, by immobilization at the surface of a solid surface, particularly a cone.

(22) The measurement of the amount and of the concentration is implemented by means of one or a plurality of immunoanalyzers, such as for example BioMérieux's VIDAS® automaton. As known per se, an immunoanalyzer comprises one or a plurality of test sections capable of each receiving one or a plurality of test strips. Each strip comprises a plurality of wells, one well receiving the sample to be analyzed and the other wells respectively containing the reagents used during the measurement, particularly diluent, a rinsing solution, a solution comprising the conjugate, and a solution containing the enzyme substrate. The automaton further comprises a mechanism for displacing the strip under a cone having its surface containing a layer of binding partners specific to the analyte. The cone is then positioned above each well and pipets in a specific order the different mediums present therein while implementing intake, discharge, and incubation mechanisms specific to the measurement technique used. The automaton finally comprises a device for measuring the property implied in this technique. For example, the VIDAS® automaton applies an ELFA-type technique which differs from the previously-described direct ELISA technique in that the substrate catalyzed by the enzyme function generates fluorescence, and it comprises a fluorometer enabling to measure the fluorescence of the solution contain in the last cuvette once the last washing step has been carried out.

(23) The haematocrit level is measured by means of any appropriate known technique, for example, by means of the so-called microhaematocrit technique used in the present embodiment, the Coulter technique, by laser measurement, or by conductivity measurement.

(24) At a next step 18 of the method, a calculation is implemented according to the determination triplets, for example (D.sub.ST.sup.cal(i), D.sub.p.sup.cal(i), D.sub.H.sup.cal(i)), to calculate the parameters of a mathematical model for correcting an analyte amount directly measured from a whole blood sample, more particularly a model according to relation:
D.sub.P=a.sub.0+a.sub.1×D.sub.STa.sub.2×D.sub.Ha.sub.12×D.sub.ST×D.sub.H (2)
where D.sub.p is the corrected analyte amount, D.sub.ST is the analyte amount directly measured on a whole blood sample, D.sub.H is the haematocrit level of the whole blood sample, and a.sub.1, a.sub.2, and a.sub.1 are the calculated coefficients of the mathematical model.

(25) Advantageously, before applying the polynomial coefficients calculation algorithm, variables D.sub.ST and D.sub.H of the polynomial according to relation (2) are normalized between −1 and 1. Particularly, if the analyte amount in the whole blood has as a minimum value D.sub.ST.sup.min and has as a maximum value D.sub.ST.sup.max, variable D.sub.ST is transformed into variable:

(26) $X_{ST} = \frac{D_{ST} - c}{d}$
where

(27) $c = \frac{D_{ST}^{\max} + D_{ST}^{\min}}{2} and d = \frac{D_{ST}^{\max} - D_{ST}^{\min}}{2} .$
Similarly, if the haematocrit level has as a minimum value D.sub.H.sup.min and has as a maximum value D.sub.H.sup.max, variable D.sub.H is transformed into variable

(28) $X_{H} = \frac{D_{H} - e}{f},$
with

(29) $e = \frac{D_{H}^{\max} + D_{H}^{\min}}{2} and f = \frac{D_{H}^{\max} - D_{H}^{\min}}{2} .$

(30) The model according to relation (2) can then be rewritten according to relation:
D.sub.P=a.sub.0.sup.n+a.sub.1.sup.n×X.sub.ST+a.sub.2.sup.n×X.sub.H+a.sub.12.sup.n×X.sub.ST×X.sub.H (3)
where coefficients a.sub.0.sup.n, a.sub.1.sup.n, a.sub.2.sup.n and a.sub.12.sup.n are determined according to the least square method by minimizing a cost function ƒ(D.sub.P.sup.cal(i)−(a.sub.0.sup.n+a.sub.1.sup.n×X.sub.ST.sup.cal(i)+a.sub.2.sup.n×X.sub.H.sup.cal(i)+a.sub.12.sup.n×X.sub.ST.sup.cal(i)×X.sub.H.sup.cal(i)). Coefficients a.sub.0, a.sub.1, a.sub.2 and a.sub.12 can be easily deduced from coefficients a.sub.0.sup.n, a.sub.1.sup.n, a.sub.2.sup.n and a.sub.12.sup.n

(31) The transformation of the variables is independent from the calculation algorithm and enables to express each variable in the same scale and thus to compare the different coefficients a.sub.1.sup.n, a.sub.2.sup.n and a.sub.12.sup.n.

(32) Here again, the foregoing is applicable to concentration measurement by replacing the analyte amount with an analyte concentration. Thus, as a variation, the analyte amount measured in the whole blood sample is transformed into an analyte concentration by dividing the measured amount by the volume of the whole blood sample, after which a mathematical concentration correction model is calculated, particularly a model according to relation:
C.sub.P=a′.sub.0+a.sub.1×C.sub.ST+a′.sub.2×D.sub.Ha.sub.12×C.sub.ST×D.sub.H (2bis)
where C.sub.p is the corrected analyte concentration, C.sub.ST is the analyte concentration calculated from the measured amount and from the sample volume, and where a′.sub.0, a.sub.1, a′.sub.2 and a.sub.12 are predetermined coefficients depending on the analyte, these coefficients being for example calculated by a least square method. It should be noted that, since the analyte concentration is deduced from the amount, the coefficients linked to the analyte concentration in the polynomial, that is, coefficients a.sub.1 and a.sub.12, are the same for the two models expressed in amount and in concentration. However, the other coefficients, that is, a′.sub.0 and a′.sub.2, are different.

(33) Similarly, it is possible to apply a normalization of the variables of the polynomial of relation (2bis) similar to the normalization of relation (3), by calculating a mathematical model according to relation:
C.sub.P=a.sub.0.sup.n′+a.sub.1.sup.n′×Y.sub.ST+a.sub.2.sup.n′×X.sub.H+a.sub.12.sup.n′×Y.sub.ST×X.sub.H (3bis)
where

(34) $Y_{ST} = \frac{C_{ST} - g}{h}, g = \frac{C_{ST}^{\max} + C_{ST}^{\min}}{2}, h = \frac{C_{ST}^{\max} - C_{ST}^{\min}}{2},$
C.sub.ST.sup.min is the minimum value of concentration C.sub.ST, C.sub.ST.sup.max is the maximum value of concentration C.sub.ST, and the polynomial coefficients are coefficients which can easily be deduced from the polynomial coefficients of relation (2bis).

(35) The method then carries on, at 20, with the verification of the determined model by using the set of verification triplets, for example, (D.sub.ST.sup.verif(i), D.sub.p.sup.verif(i), D.sub.H.sup.verif(i)) for the amount model.

(36) More particularly, the determined model is advantageously verified based on at least one of the following criteria: relative error

(37) $\frac{D_{P}^{cor} (i) - D_{P}^{verif} (i)}{D_{P}^{verif} (i)}$
between amount D.sub.p.sup.cor(i) corrected by the mathematical model, that is, D.sub.p.sup.cor(i)=a.sub.0+a.sub.1×D.sub.ST.sup.verif(i)+a.sub.2×D.sub.H.sup.verif(i)+a.sub.12×D.sub.ST.sup.verif(i)×D.sub.H.sup.verif(i), and amount D.sub.p.sup.verif measured in the plasma. More particularly, it is verified whether the relative error is centered on 0% and is in the range from −10% to +10%; the parameters of a linear equation, for example, obtained according to a so-called “Passing and Bablok” linear regression, D.sub.p.sup.cor(i)=α+β×D.sub.p.sup.verif(i); and biases calculated for different analyte amounts in the plasma by using the above-described “Passing and Bablok” equation.

(38) Here again, the foregoing can readily be applied to concentration measurement by replacing the analyte amount with an analyte concentration.

(39) Once the determined model has been validated, the method carries on with the exploitation, at 12, of this model for the analyte amount/concentration measurement directly performed on whole blood samples.

(40) For example, the model is embarked in the data processing unit of an immunoanalyzer available for sale which may further be modified to be able to receive or calculate haematocrit level values. Otherwise, the model is implemented on a processing unit independent from the immunoanalyzer, for example, a personal computer. In this variation, an operator inputs into the unit the analyte amount/concentration value measured by the immunoanalyzer on a whole blood sample and the value of the haematocrit level measured for this sample and obtains in return the corrected analyte amount/concentration.

(41) Thus, for a whole blood sample of known volume for which the amount, and above all the concentration, of analyte is desired to be known, exploitation step 12 starts, at 22, with the measurement of haematocrit level D.sub.H of the sample and carries on, at 24, with the measurement of the analyte amount/concentration directly in the whole blood sample by means of the same measurement technique, implemented by a same immunoanalyzer model as that used to determine and verify the mathematical model.

(42) A corrected analyte amount D.sub.p is then calculated, at 26, according to relation (2) with coefficients a.sub.0, a.sub.1, a.sub.2, and a.sub.12 calculated on determination of model 10, or equivalently coefficients a.sub.0.sup.n, a.sub.1.sup.n, a.sub.2.sup.n and a.sub.12.sup.n of the model according to relation (3) A corrected analyte concentration is calculated in the same way according to relation (2bis) or (3bis).

(43) At 28, the concentration is calculated and then output, for example, displayed on a screen and/or recorded in a computer memory. Optionally, the corrected amount may also be output.

(44) A method according to the invention where the measured amount or concentration is corrected by a mathematical model has been described.

(45) As previously described, measurement techniques are based on the measurement of a property, for example, an optical, electrical, chemical, pH, or enzyme-linked property, having its value depending on the analyte amount present in the analyzed sample. An immunoanalyzer thus comprises a device which measures such a property and outputs a signal corresponding to this measurement. According to the state of the art, the signal is then processed by means of a predetermined mathematical model which transforms the signal into an amount and/or concentration value.

(46) As a variation, the method applies to the actual signal, before the transformation thereof. Particularly, the signal measured on the whole blood sample is corrected according to relation:
S.sub.P=a.sub.0+a.sub.1×S.sub.ST+a.sub.2×D.sub.Ha.sub.12×S.sub.ST×D.sub.H (4)
where S.sub.p is the corrected signal and S.sub.ST is the signal originating from the measurement on the whole blood sample. The method applied to the signal correction is similar to the previously-described method, provided to make minor modifications within the abilities of those skilled in the art. The calculation of a mathematical model with normalized variables similar to relation (3) is also possible.

(47) Two examples of application of the invention, respectively an application to the measurement of a D-dimer concentration and an application to the measurement of a cardiac troponin I concentration will now be described.

(48) Measurement of the D-Dimer Concentration

(49) D-dimers are heterogeneous fibrin products all having epitope D-D formed by two continuous fibrin monomers covalently bonded by the enzyme responsible for the blood coagulation crosslinking fibrin, or “XIIIa” factor. The assaying of D-dimers is mainly indicated in the diagnosis of exclusion of venous thromboembolic diseases, particularly deep venous thrombosis and pulmonary embolism, the evaluation of the risk of recurrence of such diseases after the stopping of an anticoagulant treatment, and in the diagnosis of disseminated intravascular coagulation.

(50) To determine and verify the mathematical model associated with the D-dimer, blood has been sampled from 186 healthy subjects on sodium citrate as an anticoagulant. More particularly, the blood has been collected in tubes of 4.5 mL under vacuum containing 0.129 M of trisodium citrate.

(51) In order to have whole blood samples with a varied haematocrit level, the collected samples have been diluted with their own plasma after a light centrifugation, to obtain a haematocrit level varying in the range from 26% to 61%. The haematocrit level is measured by means of the microhaematocrit technique. The microhaematocrit is first acquired by centrifugation of capillaries for 7 minutes at 10,000 rpm, after which the value of the haematocrit level is then deduced by means of a chart, as know per se in the state of the art.

(52) In order to have whole blood and plasma samples with a varied D-dimer concentration, the whole blood samples have been overloaded with D-dimer, before or after having varied the haematocrit level. To overload the samples, plasmas from bioMérieux's internal serum bank (Marcy L'Etoile, France) with a strong D-dimer concentration to be assayed have been used to have a distribution of the D-dimer concentration in the whole blood in the 45 ng/mL-1,000 ng/mL range. The whole blood samples have been overloaded, particularly to 300 ng/mL, 500 ng/mL, and 1,000 ng/mL, with an overload which does not exceed 10% of the volume of the whole blood sample to avoid modifying the original blood matrix, that is, which does not exceed 1,000 ng/mL. The sample distribution is uniform in the D-dimer concentration range obtained from the amount measured in the whole blood and in the haematocrit level range, as illustrated in FIG. 3, which shows the (concentrations originating from the amounts measured in the whole blood, haematocrit level) pairs for the model determination triplets (dots) and for the model verification triplets (stars) The plasma samples are further obtained by centrifugation of a portion of each whole blood sample at 3,000 rpm for 10 minutes.

(53) As a numerical example, after the preparation of samples, the D-dimer concentration in the whole blood is distributed in the range from 46.78 ng/mL to 982.81 ng/mL, the haematocrit level is distributed in the range from 26% to 61%, and the D-dimer concentration in the plasma is distributed in the range from 96.86 ng/mL to 1,419.58 ng/mL.

(54) The measurement of the D-dimer concentration in the whole blood and plasma samples comprises the enzyme immunoassay of fibrin degradation products (PDF) in human plasma by an ELFA-type technique, particularly implemented by a VIDAS® automaton with the “VIDAS D-dimère Exclusion II” kit designated with reference 30 455 of bioMérieux SA. Such a measurement thus associates the sandwich-type enzyme immunoassay method in two steps with a final fluorescence detection (ELFA). In a first step, the sample is sampled, aspirated, and discharged a plurality of times so that the antigen can bind to the anti-FbDP antibodies (FbDP for “fibrinogen degradation products”) fixed on the cone. In a second step, a monoclonal anti-FbDP antibody marked with ALP (alkaline phosphatase) binds to the antigen already fixed on the cone to form a sandwich. Washing steps eliminate the non-fixed or excess compounds The development step is then carried out. The 4MUP substrate (4-methyl-umbelliferyl-phosphate) is sucked in and then discharged into the cone and the ALP catalyses the hydrolysis of the substrate into fluorescent 4MU (4-methyl-umbelliferone). The emitted fluorescence is measured at 450 nm. The value of the fluorescence signal is proportional to the antigen amount in the sample.

(55) FIG. 4 illustrates the D-dimer concentrations obtained from concentrations C.sub.ST(i) measured in the whole blood samples according to the D-dimer concentrations obtained from amounts C.sub.p(i) measured in the respective plasma samples. A “Passing and Bablok”-type linear regression C.sub.ST(i)=α+β×C.sub.p(i) between the two amounts particularly provides α=−8.80 and β=0.69.

(56) The triplets for which the concentrations measured in the whole blood are greater than the concentrations in the plasma have been discarded.

(57) One hundred and three triplets (C.sub.ST.sup.cal(i), C.sub.p.sup.cal(i), C.sub.H.sup.cal(i)) have been used to determine the mathematical model.

(58) The following coefficients are thus obtained: a.sub.0′=806.279; a.sub.1.sup.n=803.998 with a standard deviation equal to 11.917; a.sub.2′=196.295 with a standard deviation equal to 12.389; and a.sub.12.sup.n=182.254 with a standard deviation equal to 25.909,

(59) or, equivalently, with no normalization: a.sub.0′=29.311; a.sub.1=0.788; a.sub.2′=0.702; and a.sub.12=0.018.

(60) Fifty verification triplets (C.sub.ST.sup.verif(i), C.sub.p.sup.verif(i), C.sub.H.sup.verif(i)) have further been used to verify the mathematical model.

(61) FIG. 5 illustrates relative error

(62) $\frac{C_{P}^{cor} (i) - C_{P}^{verif} (i)}{C_{P}^{verif} (i)}$
according to the D-dimer concentration obtained from the amount measured in the plasma both for a correction C.sub.p.sup.cor(i) according to relation (2bis) or (3bis) applied to the verification samples, represented by diamonds, and for a correction C.sub.p.sup.cor(i) of the state of the art performed according to relation (1) applied to the verification samples, represented by squares. As can be noted, conversely to the correction of the state of the art, the relative error obtained by the invention is centered on zero, and is mainly in the range from −10% to 10%.

(63) FIGS. 6 and 7 illustrate the corrected concentrations according to the concentrations measured on plasma respectively according to the correction of relation (2bis) or (3bis) and according to the correction of relation (1), and applied to the verification samples. The line obtained by linear regression of “Passing and Bablok” type for each of the corrections is further calculated and plotted. The line corresponding to the correction of the state of the art has a slope equal to 1.18. The line corresponding to the invention is close to identity with a slope equal to 1.03.

(64) Finally, the biases for three D-dimer concentrations in the whole blood, that is, 250 ng/mL, 500 ng/mL, and 1,000 ng/mL, have been calculated and are respectively equal to −3.4%, −0.4%, and +1.1% for the correction according to the invention, and are respectively equal to +6.7%, +12.5%, and +15.4% for the correction of the state of the art.

(65) Concentrations corrected according to the invention similar to those obtained by a measurement directly performed on plasma can thus be observed.

(66) Similar results can be observed when the correction is performed on the signal measured on the samples.

(67) Measurement of the Cardiac Troponin I Concentration

(68) Troponins (Tn) are proteins of the striated muscle. The striated muscle is formed of a thick filament, made of myosin, and of a thin filament made of actin, of tropomyosin, and or the troponin complex. This complex is itself formed of three sub-units, that is, the so-called “C”, “I”, and “T” troponins. Each of these troponins has skeletal and cardiac isoforms. Cardiac troponins are choice markers for the detection of myocardial necrosis and the assaying thereof enables to diagnose a myocardial infarction. The assaying of cardiac troponins is also used for the follow-up of a thrombolytic treatment and to estimate the size of the myocardial necrosis, as well as in the diagnosis of acute coronary syndromes. The assaying of troponin “I”, noted TnI, also enables to highlight a heart impairment in the context of other pathologies, particularly a renal failure, hypothyroidism, collagenoses, myopathies, or also pulmonary embolism.

(69) To determine and verify the mathematical model associated with TnI, blood has been sampled from 186 healthy subjects on lithium heparinate as an anticoagulant. More particularly, the blood has been collected in tubes of 4 mL containing 17 UI/mL of lithium heparinate.

(70) The obtaining of whole blood samples with haematocrit levels and TnI concentrations, and the obtaining of plasma samples from the whole blood samples are similar to those described in relation with the dosage of D-dimers.

(71) The measurement of the TnI concentration in the whole blood and plasma samples comprises the enzyme immunoassay of troponin in human plasma by an ELFA-type technique, particularly implemented by a VIDAS® automaton with the “VIDAS Troponine I Ultra” kit designated with reference 30 448 of bioMérieux. In a single step, the sample is sampled and transferred into the well containing the conjugates which are anti-TNI cardiac antibodies marked with ALP. The sample/conjugate mixture is sucked in and then discharged a plurality of times by the cone. This enables the TnI, on the one hand, to bind to the conjugate to form a sandwich. Washing steps eliminate the non-fixed or excess compounds. The ALP catalyzes the hydrolysis of the substrate contained in the last well of the VIDAS® strip into fluorescent 4MU. The emitted fluorescence is measured at 450 nm. The value of the fluorescence signal is proportional to the antigen concentration in the sample. At the end of the test, the results are calculated from two calibration curves corresponding to the two development steps. A threshold signal manages the selection of the calibration curve to be used for each sample.

(72) The TnI concentration in the whole blood samples varies from 0.01 μg/L to 1.4 μg/L and the haematocrit level varies from 22% to 73%, the TnI concentration in the plasma thus varying from 0.01 μg/1 to 1.6 μg/l.

(73) The sample distribution is uniform in the TnI concentration range measured in the whole blood and in the haematocrit level range, as illustrated in FIG. 8, which shows the (concentrations measured in the whole blood, haematocrit level) pairs for the model determination triplets (dots) and for the model verification triplets (stars)

(74) FIG. 9 illustrates the TnI concentration C.sub.ST(i) obtained from concentrations measured in the whole blood samples according to the TnI concentration C.sub.p(i) measured in the respective plasma samples. A “Passing and Bablok”-type linear regression α=−0.01 between the two concentrations particularly provides α=−0.01 and β=0.84.

(75) A total one hundred and eighty-seven samples have been prepared and an exchange algorithm, for example, a Fedorov algorithm, is implemented, for example, by software NEMRODW® of LPRAI Sarl. The Fedorov algorithm is an iterative algorithm which has the advantage of enabling to select the N best samples to determine the mathematical model, for example, the 20 best samples among all the samples to calculate the parameters of the mathematical model.

(76) The following coefficients are obtained by means of the twenty selected samples: a.sub.0′=0.8982; a.sub.1.sup.n=0.86751; a.sub.2′=0.1334; and a.sub.12.sup.n=0.12413.

(77) or equivalently: a.sub.0′=−0.0052; a.sub.1=0.9155; a.sub.2′=0.0002; and a.sub.12=0.0072.

(78) One hundred and sixty-seven verification triplets (C.sub.ST.sup.verif(i), C.sub.p.sup.verif(i), C.sub.H.sup.verif(i)) have further been used to verify the mathematical model.

(79) FIG. 10 illustrates relative error

(80) $\frac{C_{P}^{cor} (i) - C_{P}^{verif} (i)}{C_{P}^{verif} (i)}$
according to the TnI concentration obtained from the amount measured in the plasma both for a correction C.sub.p.sup.cor(i) according to relation (2bis) or (3bis) applied to the verification samples, represented by diamonds, and for a correction C.sub.p.sup.cor(i) of the state of the art performed according to relation (1) applied to the verification samples, represented by squares. As can be noted, conversely to the correction of the state of the art, the relative error is centered on zero, and is mainly in the range from −10% to 10%. A few results outside of −/+10% in the sample range where the TnI concentration is close to the detection limit can be observed.

(81) FIGS. 11 and 12 illustrate the corrected TnI concentrations according to the concentrations measured on plasma respectively according to the correction of relation (2bis) or (3bis) and the correction of relation (1), and applied to the verification samples. The line obtained by linear regression of “Passing and Bablok” type for each of the corrections is further calculated and plotted. The line corresponding to the correction of the state of the art has a slope equal to 1.48. The line corresponding to the invention is close to identity with a slope equal to 1.02.

(82) Finally, the biases for three TnI concentrations in the whole blood, that is, 0.1 μg/L, 0.5 μg/L, and 1 μg/L, have been calculated and are respectively equal to 3.1%, 2.0%, and +1.9% for the correction according to the invention, and are respectively equal to +43.3%, +46.7%, and +47.1% for the correction of the state of the art.

(83) The same results as those obtained for D-dimers can also be observed, that is, corrected concentrations which are similar to those obtained by a measurement directly performed on plasma.

(84) Similar results can be observed when the correction is performed on the signal measured on the samples.

(85) Embodiments where a polynomial of first degree according to relation (2bis) or relation (3bis) is used have been described. A second degree polynomial comprising the square of the haematocrit level and/or the square of the analyte amount/concentration measured in the whole blood may also be used. Similarly, a polynomial of higher degree may be used.

(86) Particularly, the following procedure may be used to identify the polynomial used:

(87) a) setting the order of the polynomial to 1;

(88) b) calculating the parameters of the polynomial, for example, by means of a least square method, as previously described;

(89) c) if the prediction error of the polynomial is considered satisfactory, selecting this polynomial. For example, the prediction error is calculated and if it is lower than a threshold, advantageously 10%, the polynomial is selected; and

(90) d) if the prediction error, advantageously calculated on a set of verification data, is not considered satisfactory, increasing the order of the polynomial by one unit and repeating steps b), c), and d) until the prediction error is considered satisfactory. As a variation, these steps are repeated as long as the order is smaller than or equal to 10, and preferably smaller than or equal to 5.

(91) Of course, other methods of calculation and/or of final selection of the polynomial may be used. For example, the order may also be calculated by means of a non-linear regression, as known per se in the state of the art. Similarly, a bayesian-type criterion, for example, the BIC (“bayesian information criterion”), may be used to select the final polynomial when a plurality of polynomials have been calculated.

(92) Embodiments where the amount/concentration measurement is used with no transformation of the variables in the mathematical models have been described. As a variation, a transformation of the measurement space may be implemented, for example, a logarithmic transformation if the measured amount/concentration range is significant. Term “measurement” thus means, in the sense of the invention, the direct or transformed measurement.

(93) Similarly, polynomial coefficients calculated for a significant amount/concentration range have been described. As a variation, the range of amounts/concentrations may be divided into a plurality of intervals and a mathematical model may be determined for each of these intervals. The coefficients of the model may thus depend on the amount/concentration interval for which they are calculated.

Method for measuring the plasma concentration of an analyte directly on a whole blood sample

Assignee

Inventors

Cpc classification

Classification Explorer

G16B40/00

PHYSICS

Classification Explorer

G01N33/543

PHYSICS

Classification Explorer

G16B40/10

PHYSICS

Classification Explorer

G01N2333/47

PHYSICS

Classification Explorer

G01N2333/75

PHYSICS

Classification Explorer

G01N2333/4706

PHYSICS

Classification Explorer

G01N33/6887

PHYSICS

International classification

Classification Explorer

G01N33/543

PHYSICS

Classification Explorer

G01N33/68

PHYSICS

Classification Explorer

G16B40/00

PHYSICS

Classification Explorer

G16B40/10

PHYSICS

Abstract

Claims

Description