JPH0547782B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to an automatic analyzer, and particularly to an automatic analyzer suitable for quantifying the concentration of a drug in blood, which has a calibration curve that is not linear with respect to concentration. For many drugs, the range between medicinal and toxic doses, ie, safe tolerances, is narrow, and therapeutically effective concentrations vary significantly from patient to patient. For this reason, determination of drug concentration in blood is useful for drug regimens for individual patients, and highly accurate determination is required. Quantifying blood drug concentrations in this way is a method of homogeneous enzyme immunoassay.
This is often done using the EMIT method. This EMIT
In this method, the calibration curve is not linear with respect to drug concentration. For this purpose, special graph paper is prepared for each reagent lot to make the calibration curve a straight line, and when measured data is plotted on this designated graph paper, a calibration curve that is linear with respect to concentration can be obtained. It's becoming like that. However, if the photometry timing of the reaction solution is off, if the concentration of the measurement reagent is changed, or if a reagent is used that has been used for some time after opening the reagent, it may not be possible to create a satisfactory calibration curve even if this designated graph paper is used. The disadvantage was that it was impossible to do so. An object of the present invention is to provide an automatic analyzer that can automatically create a calibration curve that is most suitable for the condition of reagents and the timing of photometry when creating a calibration curve in the measurement of drugs, hormones, antibiotics, etc. by enzyme immunoreaction. Our goal is to provide the following. The gist of the present invention is as follows. That is,
The enzyme immunoreaction using the EMIT method follows the following formula. Y(i)=R ₀ +K/1+exp [ ^-(a+blnX(i)] ) (1) K=R _n −R ₀ (2) where R ₀ is the response at drug concentration 0,
R _n is the response at drug concentration ∞, a and b are parameters, X is drug concentration, and Y is absorbance. On the graph paper attached to the reagent, the parameter values of R ₀ , K, a, and b have been determined in advance using equation (1), and when the absorbance is plotted against the drug concentration, the calibration curve is a straight line. It is made to be. That is, the parameters (R ₀ , K, a, b)
is set as a fixed value. However, according to our experiments, this parameter value is not constant even when measured under specified conditions, and changes over time, for example, after opening the reagent. Moreover, when the photometry timing is shifted or when the reagent concentration is changed, the value fluctuates greatly. Furthermore, although some items are difficult to converge as a result of multiple regression, we found that these can be resolved to some extent by examining the damping coefficients used in multiple regression. Therefore, the present invention makes it possible to specify for each item the damping coefficient most suitable for the conditions of the equipment and reagents at that time each time a calibration curve is created, and to minimize the absorbance data of the reaction solution of each standard substance. The objective is to automatically obtain an optimal calibration curve by processing by multiple regression using a dual method and determining parameters for each measurement. Examples of the present invention will be described below. FIG. 1 shows an embodiment of the invention. In the figure, a sample disk 10 can hold multiple standard substances with different concentrations for each measurement item.
is provided. The plurality of standard substances are configured so that they can be successively arranged on the sample disk 10 for each measurement item. Further, the reaction disk 21 has a plurality of reaction vessels 22 that also serve as measurement cells on its circumference, and is configured to be freely rotatable. Further, the standard substance (sample) is transferred by the sampling probe 41, and the reagent is dispensed by the dispensers 38, 39. Further, the spectrometer 27 is of a multi-wavelength simultaneous photometry type having a plurality of detectors, and is opposed to the light source lamp 25, so that when the reaction disk 21 is in a rotating state, the row of reaction vessels 22 receives the light flux from the light source lamp 25. 26. The light beam 26 is arranged so as to pass through the center of, for example, the 31st reaction container 46, counting clockwise from the discharge position 45 when the reaction disk 21 is in a stopped state. A draining device and a cleaning device 24 are arranged between the position of the light beam 26 and the discharge position 45. The overall configuration of the control device includes a multiplexer, a logarithmic conversion amplifier 53, an A/D converter 54, a read-only memory (hereinafter referred to as ROM), a random access memory (hereinafter referred to as RAM),
It consists of a printer 55, an operation panel 52, and a mechanism drive circuit 35, and the A/D converter 54 is further connected to a central processing unit 51 via an interface 50. The central processing unit 51 is a microcomputer that controls the entire apparatus including the mechanical system and performs general data processing such as calibration curve creation and concentration calculation using the aforementioned multiple regression. Next, the operation of the embodiment shown in FIG. 1 will be explained. First, when the sample container 44 containing a standard substance (sample) to be measured containing drugs, hormones, etc. is supplied to the sampling position, the probe 4 of the pipette 40
1 is immersed into the sample container, a certain amount of serum is aspirated and retained within the probe 41. Thereafter, the probe 41 moves to the discharge position 45 of the reaction disk 21 and discharges the serum held by the probe 41 into the reaction container 22 that has been transferred to the discharge position 45. When the above sampling operation is completed, the reaction disk 21 starts rotating intermittently in the counterclockwise direction, and one more reaction containers 22 than the total number of reaction containers 22 on the reaction disk 21 pass through the discharge position. Rotate until it stops. By the rotation of the reaction disk 21, the reaction container 22 containing the sample sampled in the sampling operation is moved from the discharge position 45 to the reaction container 1.
It comes to a stop after moving counterclockwise by the pitch. During the rotation of the reaction disk 21, all the reaction vessels 22 on the reaction disk 21 receive the light beam 2.
Pass 6. Therefore, each reaction vessel 22
When the light passes through the light beam 26, the spectrometer 27 performs optical absorption measurement, and the output of the spectrometer 27 is used to select a signal of the currently required measurement wavelength by the multiplexer, and is sent to the central processing unit 50 by the A/D converter 54. captured and stored in RAM. For example, if the time period during which the reaction disk 21 is rotating and stopping is 20 seconds, then 20 seconds is 1
The above operation is repeated as a cycle. As the cycle progresses, the position of the sampled sample to be measured moves counterclockwise by one pitch of the reaction container when the reaction disk 21 is stopped. The reagent is discharged from the dispenser 36 and the dispenser 37 by moving the reaction container 22 containing the sample counterclockwise by one pitch of the reaction container,
This is done while the pumps are stopped at the discharge positions 46 and 47, respectively. Looking at a specific sample to be measured, the first reagent added at the discharge position 47 causes
The first stage reaction is started, and the second stage reaction is started by the second reagent added to the discharge position 46. Assuming that the stop time of the reaction disk 21 in one cycle is 4.5 seconds and the rotation time is 15.5 seconds, the reaction process of a particular sample will be measured 31 times every 20 seconds, and the measurement will take 10 minutes. Data is stored in RAM. The central processing unit
It operates according to the program in ROM, and the
Extract 31 measurement data and perform calculation processing. 5 per item required to create an EMIT calibration curve
Alternatively, since the six types of standard substances are consecutively arranged on the sample disk 10, multiple standard substances with different concentrations of a specific item are automatically consecutively placed one by one (for example, twice each time). ) is transferred to the reaction vessel 22 by the sampling probe 41. When creating a calibration curve for a substance that has no linear relationship to concentration, it is essential to sample and measure standard substances with different concentrations multiple times, and this device can do this. The reaction process of these multiple samples is measured over a period of 10 minutes as described above, and these measurement data are summarized for each item and used to create a calibration curve that has a linear relationship with concentration. . Flowcharts for creating a calibration curve based on such EMIT measurement data are shown in FIGS. 2 and 3. In this calculation process, it is possible to automatically select the model most suitable for the measured data from among the following six types of calculation models set in advance. MODEL #1 4-parameter logic R=R ₀ +K _e 1/1+exp [-(a+blnC)] MODEL #2 5-parameter logic R=R ₀ +K _e 1/1+exp [-(a+blnC+cC)] MODEL #3 5-parameter R=R ₀ +Kexp [alnC+b(lnC) ² +C(lnC) ³ ] MODEL #4 5 parameters lnC=a+b(R-R _0/100 )+c(R-R _0/100 ) ² +
d (R−R ₀ /100) ³ MODEL #5 R=R ₀ +KlnC MODEL #6 R=R ₀ +K·C Also, as an example of arithmetic processing of measurement data, multiple regression using a nonlinear least squares method is used. Processing based on this multiple regression will be explained next. Concentrations X ₀ , X ₁ , X ₂ , X ₃ ,
Assume that the measurement data of the standard substances X ₄ and X ₅ are Y _0j , Y _1j , Y _2j , Y _3j , Y _4j , Y _5j , respectively. After setting the initial values of the four parameters (R ₀ , K, a, b) in equation (1) using these values, Gauss-
Multiple regression is performed using the nonlinear least squares method using Newton's modified method, and four types of parameter values are determined. Next, the multiple regression method will be described in detail. Measured data (Y(i)) for a specific item
and the calculated value (F(i)) obtained by substituting the measured parameter value into equation (1) (y(i)) is y(i)=Y(i)−F(i)∂ F(i)/∂R ₀・ΔR ₀ +∂F
(i)/∂K _e・ΔK _e +∂F(i)/∂a・Δa+∂F(i)/∂b・Δ
It is approximated by b. The parameter value to be found is the value when the sum of squares (S) of this difference is the minimum. S=Σ{y−y(i)} ² =Σ{y−(∂F(i)/ΔR ₀・
ΔR ₀ +∂F(i)/ΔK _e・ΔK ₀ +∂F(i)/∂a・Δa+∂F(i)
/∂b・Δb)} ^2Now , ∂s/∂a=∂s/∂b=∂s/∂K _e =∂s/∂R ₀ =0, so ∂F(i)/∂ΔR ₀ =x ₁ , ∂F(i)/∂K _e =x ₂ , ∂F(i)/∂
When a=x ₃ and ∂F(i)/∂b=x ₄ , the following four equations (3) to (6) hold true. ΔR ₀ Σx ² ₁ +ΔK _e Σx ₁ x ₂ +ΔaΣx ₁ x ₃ +ΔbΣx ₁ x ₄ =
Σx ₁ y(3) ΔR ₀ Σx ₁ x ₂ +ΔK _e Σx ² ₂ +ΔaΣx ₂ x ₃ +ΔbΣx ₂ x ₄ =
Σx ₂ y(4) ΔR ₀ Σx ₁ x ₃ +ΔK _e Σx ₂ x ₃ +ΔaΣx ² ₃ +ΔbΣ ₃ x ₄ =Σ
x ₃ y(5) ΔR ₀ Σx ₁ x ₄ +ΔK _e Σx ₂ x ₄ +ΔaΣx ₃ x ₄ +ΔaΣx ^2/ =
Σx ₄ y(6) Parameter change ΔR ₀ that satisfies equations (3) to (6),
ΔK _e , Δa, and Δb are found by solving determinant (7). Σx ² ₁ Σx ₁ x ₂ Σx _{1 x 3} Σx ₁ x ₄ Σx ₁ y Σx ¹ x 2 Σx _{2 2} Σx ₂ x ₃ Σx ₂ x ₄ Σx ₂ y Σx ₁ x ₃ Σx ₃ x ₂ Σx ² ₃ Σx _{3 x 4} Σx ₃ y Σx ₁ x ₄ Σx ₄ x ₂ Σx ₄ x ₃ Σx ² ₄ Σx ₄ y ...(7) Using Δa, Δb, ΔK _e , ΔR ₀ obtained in this way, parameters a, b , K _e , R ₀ value, a=
a+Δa, b=b+Δb, K _e = K _e +ΔK _e , R ₀ = R ₀
If you replace +ΔR ₀ and perform multiple regression, the sum of squared differences
Find the parameter value when (S) is the minimum. By substituting the determined parameter values into equation (1), a calibration curve having a linear relationship with the concentration can be obtained. This calibration curve can be obtained with high accuracy because parameter values are determined to be most suitable for measurement data that reflects the state of the apparatus or reagent at the time of measurement. When solving determinant (7) using multiple regression, in order to obtain results with good convergence conditions, (i, i) of determinant (7)
The components were multiplied by a constant larger than 1 for the purpose of damping. If the damping coefficient is set to a large value, there is little risk of divergence, but convergence is considered to be extremely difficult. On the other hand, if the value is very close to 1, the yield will be fast, but it is expected that the yield will diverge. Furthermore, it was found that when each item is multiplied by a common damping coefficient, it becomes extremely difficult to converge for some items depending on how the damping coefficient is taken. Furthermore, it was found that for all EMIT items, convergence conditions such as the quality of convergence results and the number of multiple regressions required to converge are affected by the damping method. As an example of this, the relationship between the damping coefficient and the convergence condition for ETHOSUXIMIDE is shown below.

[Table] As shown above, the method of applying damping has a large effect on the convergence conditions of multiple regression, so
It can be said that having the function of specifying how to apply damping for each measurement item is indispensable for producing good convergence results, that is, creating highly accurate calibration curves. As an example of creating a calibration curve, FIG. 4 shows a flowchart of multiple regression using the nonlinear least squares method based on Model 1, and FIG. 5 shows a flowchart of determining the quality of the calibration curve. n in Figure 5,
m, l ₁ , l ₂ , l ₃ , l ₄ are constants manually input as parameters for each item. To judge whether the calibration curve is good or bad, the difference between the double measurement values of the same concentration calibrator, the bias between the average value of the double measurement of the calibrator and the calibration curve (regression line), the sensitivity judgment of the measured value, and the regression line convergence condition (SD value) judgment are carried out. I am considering it. Finally, Figure 6 shows the case where the EMIT measurement results are plotted on special graph paper with fixed parameters as before, and Figure 7 shows the case where the parameters are determined by calculation based on the present invention and a calibration curve is created. are shown respectively. As is clear from the figure,
It can be seen that the calibration curve created based on the present invention has a linear relationship with the concentration, and that the object of the present invention has been fully achieved. An example of measurement conditions is shown below.

[Table] Therefore, according to this example, in the measurement of drugs, hormones, antibiotics, etc. by enzyme immunoreaction that are not linear with respect to concentration, each time a calibration curve is created, the equipment and Calculation parameters most suitable for the condition of the reagent can be determined by multiple regression using the nonlinear least squares method using a specified damping method, so highly accurate calibration curves that are linear with concentration can be automatically generated. It has the effect of being able to be created. As explained above, according to the present invention, drugs and
In the measurement of hormones, antibiotics, etc. by enzyme immunoreaction, it is possible to automatically create a calibration curve that is most suitable for the condition of the reagent at the time of creating the calibration curve and the timing of photometry.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Figs. 2 and 3 are a flowchart for creating a calibration curve, Fig. 4 is a flowchart for multiple regression using the nonlinear least squares method, and Fig. 5 is a judgment on the quality of the calibration curve. Flowchart, Figure 6 is a calibration curve obtained by plotting the results of LIDOCAINE measurement on the fixed parameter graph paper attached to the reagent, and Figure 7 is a calibration curve obtained by calculating the measurement results in Figure 6 using this example. FIG. 22...Reaction container, 44...Sample container, 40...
...pipette, 36,37...dispenser, 26...spectroscope.

Claims (1)

[Claims] 1. A sample supply mechanism that can install four or more standard substances with different concentrations for each measurement item of enzyme immunoreaction measurements of drugs such as anti-epileptic drugs, cardiotonic agents and bronchodilators, hormones, antibiotics, etc., and each standard. an automatic analyzer equipped with a sampling mechanism capable of sampling a substance multiple times; a second means for creating a calibration curve based on the processing in the means; and a third means for determining the quality of the calibration curve created in the second means, the third means being the same. A condition in which the difference between the concentration calibrator double measurement values is less than a predetermined value, a condition in which the deviation between the calibrator double measurement average value and the calibration curve is less than a predetermined value, and a condition in which the sensitivity of the measurement value is within a predetermined range. , the quality of the calibration curve is determined by determining whether at least one of the conditions that the regression convergence SD is equal to or less than a predetermined value is satisfied, and an alarm is generated when the determination is negative. Features of automatic analysis equipment.