JP2010243406A - Method for detecting disease state progression degree of liver cancer and chronic liver disease using discriminant function characterized by measurement values of AFP and PIVKA-II - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a means for detecting an extent of a clinical condition of a liver cancer and a chronic liver disease by a discriminant function which takes measurement values of known tumor markers, AFP and PIVKA-II, as characteristic values. <P>SOLUTION: A method for detecting the extent of the clinical condition of the liver cancer and the chronic liver disease, by using the discriminant function which takes the measurement values of the known tumor markers, AFP and PIVKA-II, as the characteristic values, detects the liver cancer, particularly the smaller or the smallest liver cancer, easily and accurately with high sensitivity and high specificity, and also detects the extent of the clinical condition of a patient highly likely to develop the liver cancer (a chronic hepatitis patient and a cirrhotic patient). <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  It is an object of the present invention to produce liver cancer, particularly small or minimal liver cancer (small liver cancer: 3 cm in diameter, minimal liver) that has been difficult to detect by conventional methods (immunoassay and echo examination of conventional tumor markers). AFP (alpha-fetoprotein) and PIVKA-II (Protein induced by Vitamin K absence or antagonist-II) are known tumor markers that can detect cancer: liver cancer (diameter 2 cm or less) with high sensitivity and specificity. ) A liver cancer detection method using a discriminant function with a measured value as a feature value, and a kit and determination system used therefor. Furthermore, it is possible to detect the progress of disease state of patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients) using this method, and to provide the method, a kit used therefor, and a determination system. .

  Liver cancer develops from chronic liver disease (chronic hepatitis, cirrhosis), many of which develop from chronic liver disease due to persistent hepatitis virus infection. In Japan, more than 95% of liver cancers are persistently infected with hepatitis B virus (HBV) and hepatitis C virus (HCV), and more than 80% are derived from HCV-related diseases. It is said that it is a thing. In general, patients with hepatitis develop liver cancer after hepatitis develops gradually and progresses to cirrhosis over 20-30 years. In particular, the precirrhosis stage (fibrosis degree F3) and cirrhosis (fibrosis degree F4) develop at a high rate, and these patients are said to develop liver cancer within 10 years.

  The age of onset is more than 60 years old, and there are many elderly people, but there are some cases that develop at a relatively early stage or after the onset of hepatitis, but the cause is unknown. Conventionally, liver cancer is detected by screening or echography (ultrasonography) using immunoassay for existing tumor markers such as AFP and PIVKA-II in patients with chronic liver disease who are at high risk of developing liver cancer. ) Has been regularly monitored. In addition, screening tests using AFP-L3 fractionation, which detects the cancerous changes of sugar chains on AFP molecules using affinity for lectins, have also been performed.

  Patients who are suspected of having liver cancer through these examinations can be diagnosed in more detail by echography, CT examination (Computed Tomography) or MRI examination (Magnetic Resonance Imaging). The location, size, number, etc. of liver cancer are confirmed. Furthermore, a final definitive diagnosis is made by a pathological examination using a biopsy, as necessary. Liver cancer that has been confirmed is subjected to treatment such as hepatectomy, hepatic artery embolization (catheter), and puncture therapy (ethanol injection therapy, radiofrequency ablation therapy, microwave coagulation therapy). However, since the virus is not removed by treatment and the liver disease is not cured, the recurrence rate of liver cancer is extremely high.

  Prognosis is relatively good if liver cancer is detected at the stage of small or minimal liver cancer by echocardiography or immunoassay currently used, and appropriate treatment is performed. Echo examination has made it possible to detect liver cancer at the stage of small or minimal liver cancer due to recent improvements in equipment performance, but it requires technical skill to spend about 30 minutes for each examination. However, it is not suitable for screening tests for patients with high risk of developing liver cancer. In addition, since it is difficult to completely examine the entire liver, particularly the part hidden in other organs and the inside of the liver, there is a risk of overlooking liver cancer.

  AFP is produced in the liver and yolk sac during the embryonic period, and is considered to be involved in the transfer of substances between the mother and fetus as a binding protein for various substances and to have an immunosuppressive effect. On the other hand, PIVKA-II is a precursor of coagulation factor prothrombin, and is known to appear in blood when vitamin K is deficient due to vitamin K deficiency or liver damage. Both of the above two types of tumor markers have already been isolated, purified and structurally determined and are known to increase in the blood when it becomes liver cancer, and thus are used in screening tests for liver cancer. Here, the performance required for these tumor markers is required to have high detection sensitivity in liver cancer, particularly small or minimal liver cancer, and high specificity in view of the efficiency of regular screening tests. .

  As with other screening tests, the final pathological test is the final screening test for AFP, AFP-L3, and PIVKA-II immunoassays for patients with chronic liver disease who are at high risk of developing liver cancer. Cut-off values are set based on the distinction between cancers and non-cancerous diseases for which a definite diagnosis is made. When the measured value of AFP, AFP-L3 or PIVKA-II shows a value equal to or higher than the cut-off value, the possibility of liver cancer is strongly suspected. According to the “Liver Cancer Medical Care Guidelines (2005)” prepared by the Research Group on Liver Cancer Medical Care Guidelines Based on the Scientific Basis of the Japan Liver Society, the cut-off values are AFP: 200 ng / ml, AFP-L3 : 15%, PIVKA-II: 40 mAU / ml is reported to be optimal.

  However, this method has a problem that detection sensitivity and specificity are insufficient, and detection sensitivity is extremely low particularly in small or minimal liver cancer (Non-Patent Document 1, Non-Patent Document 2, and Non-Patent Document). 3 and Non-Patent Document 4). Specifically, for example, the sensitivity of detection of AFP, AFP-L3 or PIVKA-II for small liver cancer of 3 cm or less was 30% or less, respectively (Non-patent document 1 and Non-patent document 2). And Non-Patent Document 3 and Non-Patent Document 4).

  On the other hand, a method for detecting liver cancer by distinguishing between AFP and PIVKA-II or AFP-L3 and PIVKA-II in terms of differentiating between cancer and non-cancer disease has been proposed. (See Non-Patent Documents 1, 5, and 6). Similar to the single calculation method, this combined determination method is a method for discriminating cancer and non-cancer diseases based on the cut-off value of each tumor marker. However, this method reduces the specificity of liver cancer detection. It has been reported that the detection sensitivity of liver cancer can be improved by about 30% as compared with the single calculation method (see Non-Patent Documents 1, 5, and 6).

  However, even with this method, it is still difficult to find small or minimal liver cancer, and detection sensitivity and specificity are not sufficient. Furthermore, since the cut-off value itself set for each tumor marker varies depending on the reporter, it is not universal and is not practical as a screening test. In addition, whether the law (cut-off value) can be applied to completely different populations in the same disease has not been sufficiently verified, and the sensitivity and specificity described in the paper are within the scope of the paper. We cannot deny the possibility that it is a result that only holds.

  On the other hand, a method using multivariate analysis has been disclosed as a method capable of accurately discriminating between cancer and non-cancer diseases and pathological conditions (see Patent Documents 1-3 and Non-Patent Document 7). Multivariate analysis is an analysis method to obtain sample data extracted from two or more populations, and to investigate which population the sample data belongs to when there is unknown sample data belonging to which population. Multivariate data (x1, x2,... Xn) regarding gene expression levels and protein blood levels are used as characteristic values. For example, in order to discriminate the pathological condition (fibrosis degree) of a hepatitis C patient, a feature (test item) that is effective for identifying the pathological condition of hepatitis is selected from blood test data of the hepatitis C patient. A method of extracting by an algorithm SFFS (Sequential Forward Floating Search) and discriminating a disease state using a support vector machine using them is disclosed (see Patent Document 1 and Non-Patent Document 7).

  In addition, a method of obtaining an expression pattern of a gene group specifically expressed in a colon cancer primary tissue by a DNA microarray method and classifying the pathological condition of the colon cancer by multivariate analysis is disclosed (Patent Document 2). In addition, the expression of insulin-like growth factor-binding protein 5 (IGFBP5), Claudin4 (CLDN4), PDZ and LIM domain 7 (PDLIM7) and Biglycan (BGN) genes in the liver tissue of subjects was obtained, and multivariate analysis was performed. Discloses a method of detecting intrahepatic cholangiocarcinoma and classifying intrahepatic cholangiocarcinoma from hepatocellular carcinoma and metastatic liver cancer (Patent Document 3).

  However, these disclosed methods using multivariate analysis cannot be said to be sufficiently satisfactory for use in the medical industry for any of the following reasons. (1) A method that can detect liver cancer, especially small or minimal liver cancer with high sensitivity and high specificity, and accurately detect the degree of progression of chronic hepatitis patients and cirrhosis patients who are at high risk of developing liver cancer. It does not provide a method. (2) Whether the disclosed law (multivariate analysis) can be applied to completely new clinical cases has not been verified, and may be a result that is valid only within the scope of the document and published patent publication Cannot be denied. (3) Routine screening tests such as RNA extraction process from tissue specimens are complicated and time-consuming, making it difficult to achieve high throughput, and RNA degradation enzymes easily contained in sweat and dust. Is not suitable.

  Thus, there has been no satisfactory existing test method for detecting liver cancer, particularly small or minimal liver cancer. Furthermore, no test method has been established for the detection of precancerous lesions, which is a high risk group for developing liver cancer.

JP 2006-153582 JP 2006-271385 JP2008-72952

Nomura F et al., Am J Gastroenterol 94: 650-4, 1999 Martinez Cerezo FJ et al., Rev Esp Enferm Dig 84: 311-4, 1993 He YM et al., Hepatobiliary Pancreat Dis Int 4: 50-4, 2005 Han SL et al., Hepatogastroenterology 52: 348-51, 2005 Ishii M et al., AM J GASTROENTEROL 95: 1036-40, 2000 Ikoma J et al., Hepatogastroenterology 49: 235-8, 2002 Jiang Z et al., J Med Syst 30: 389-94, 2006 Iizuka N et al., Lancet 361: 923-929, 2003

  The object of the present invention is to solve the problems in screening or monitoring of liver cancer by immunoassay and echo test of conventional tumor markers for patients with high risk of developing liver cancer, especially detection by the conventional method The characteristic values of the measured values of AFP and PIVKA-II, which are known tumor markers that make it possible to detect small or minimal liver cancer that was difficult to detect with high sensitivity and high specificity. It is to provide a liver cancer detection method by a discriminant function designed by statistical pattern recognition including a step of optimizing a cut-off value, and to provide a kit and a determination system used therefor. Furthermore, by using this method, it is possible to detect the progression of the pathological condition of patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients), and to provide the method, a kit used therefor, and a determination system. .

  In view of the above problems, the present inventors have performed statistical pattern recognition including a step of optimizing the cutoff value of the Fisher linear discriminant function using the measured values of known tumor markers AFP and PIVKA-II as characteristic values. Design a discriminant function, calculate the discriminant function value by substituting the measured value into the discriminant function, and detect liver cancer, especially small or minimal liver cancer with high sensitivity and high specificity, and easily and accurately. The present inventors have found that it is possible to detect the disease, and have found that it is possible to detect the degree of progression of liver cancer patients (chronic hepatitis patients and cirrhosis patients), thereby completing the present invention.

  In order to achieve diagnosis (detection) with high sensitivity and high specificity, it is generally necessary to analyze a large amount of patient samples for the purpose of improving diagnostic performance. However, in reality, the number of patient samples is limited, and the current situation is that analysis is performed without securing a sufficient number of patient samples. The Fisher linear discriminant function (standard Fisher linear discriminant function) is used as an analysis method that can reduce the degradation of diagnostic performance due to the small number of patient samples, and results that are effective in analysis with a small number of patient samples. It has been reported (Non-Patent Document 8).

  Here, with regard to the method for detecting liver cancer using the tumor marker, in the past, the presence or absence of liver cancer that is finally confirmed by pathological examination is detected using the measured value of the tumor marker as a characteristic value. Therefore, the discriminant function is designed by statistical pattern recognition including the process of optimizing the cutoff value of the Fisher linear discriminant function, which can achieve further improvement in diagnostic performance, and the discriminant designed from the viewpoint of sensitivity and specificity. The function is validated, and the discriminant function value obtained from the discriminant function is used to detect liver cancer, especially small or minimal liver cancer, and patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients). There is no report that the degree of pathological progress could be detected with high accuracy. Furthermore, there is no report demonstrating that the discriminant function value obtained by using the discriminant function can be applied to a completely different population in the same disease, and that the universality of the liver cancer detection ability of the discriminant function has not been proved.

  That is, the discriminant function value obtained from the discriminant function designed by the statistical pattern recognition including the step of optimizing the measured value of the two tumor markers as the characteristic value and the cutoff value of the Fisher linear discriminant function, The present inventors have found for the first time that it can be used as an index of the degree of progression of chronic liver disease. Here, when the discriminant function value shows a negative value, it is determined that the patient has liver cancer, and when the discriminant function value shows 0 or a positive value, the patient has liver cancer. Judge that it is not. Furthermore, the discriminant function value has a negative correlation with the degree of progression of chronic liver disease and can be used as an index of the degree of progression of disease. That is, the average value of the discriminant function value shows a statistically significant difference among the chronic hepatitis patient group, the cirrhosis patient group, and the liver cancer patient group, and from a positive value to a negative value as the disease progresses. Indicates.

That is, the present invention is mainly directed to a method for detecting liver cancer and monitoring the degree of progression of chronic liver disease comprising the following steps.
(1) A liver cancer detection method using AFP (alpha-fetoprotein) and PIVKA-II (Protein induced by Vitamin K absence or antagonist-II), comprising the following steps (a) to (e): The method wherein the discriminant function value obtained by the discriminant function having the measured value as a characteristic value indicates that the patient is suffering from liver cancer and the degree of progression of chronic liver disease.

(A) a step of measuring AFP and PIVKA-II in a patient sample classified according to medical findings;
(B) a step of designing a discriminant function by statistical pattern recognition in order to detect the presence or absence of liver cancer using the measured values of AFP and PIVKA-II as characteristic values;
(C) calculating a discrimination function value indicating that the patient described in (a) is suffering from liver cancer from the discrimination function designed in (b) above;
(D) measuring AFP and PIVKA-II in a patient sample obtained from a population different from (a) above;
(E) Substituting the measured values of AFP and PIVKA-II in the patient described in (d) above into the discrimination function designed in (b) above to calculate the discrimination function value, and the discrimination function value is a negative value A step of determining that the patient is afflicted with liver cancer, and determining that the patient is not afflicted with liver cancer when the discriminant function value is 0 or a positive value.

  (2) In step (b), the cutoff value of the Fisher linear discriminant function is changed to obtain a cutoff value indicating the optimum sensitivity and specificity in the patient group in step (a) (cutoff value). (1). The method according to (1),

(3) Logarithmic transformation or exponential transformation of the feature value in the step (b) is normalized, and a discrimination function is designed by statistical pattern recognition in order to detect the presence or absence of liver cancer using the transformation value as the feature value. After that, the characteristic value of the step (e) is normalized by logarithmic transformation or power transformation, and the transformed value is substituted into the discriminant function to calculate the discriminant function value. The method according to (1) above, which shows that the patient has liver cancer and the degree of progression of chronic liver disease.
(4) The method according to (1) to (3) above, wherein the liver cancer is small or minimal liver cancer.
(5) The samples described in (1) to (3) above, wherein the sample is derived from any one of human serum, plasma, whole blood, tissue, blood cells, excrement and endocrine / exocrine fluid. the method of.

(6) The method according to (1) to (5) above, wherein an identification function value is calculated using the following identification function;
(Formula I) D = −1.144 × AFP−1.960 × PIVKA-II + 151.545
(Formula II) D = −1170.579 × AFP + 6.861 × PIVKA-II + 17342.832
(Formula III) D = −45.954 × AFP−0.256 × PIVKA-II + 1687.103
(Formula IV) D = −743734.876 × AFP ^0.05 + 24677.050 × PIVKA-II ^0.05 + 814071.668
(Formula ^{V) D = -103044.554 × AFP 0.15} -21713.517 × PIVKA-II 0.15 + 199366.535
(Formula ^{VI) D = -23485.069 × AFP 0.25} - 8727.211 × PIVKA-II 0.25 + 69816.070
(Formula VII) D = −1817110.433 × AFP ^0.05 + 160884.976 × PIVKA-II ^0.05 + 1881877.418
(Formula VIII) D = -456892.175 x AFP ^0.15 + 14493.378 x PIVKA-II ^0.15 + 660012.847
(Formula IX) D = −193207.018 × AFP ^0.25 + 3116.406 × PIVKA-II ^0.25 + 372629.501
(Formula X) D = −1334801.628 × AFP ^0.05 + 157824.415 × PIVKA-II ^0.05 + 1335806.678
(Formula XI) D = −250689.564 × AFP ^0.15 + 14282.959 × PIVKA-II ^0.15 + 351191.238
(Formula XII) D = −79694.207 × AFP ^0.25 + 236.572 × PIVKA-II ^0.25 + 157425.794
(When D <0, it is determined that “liver cancer is present”, and when D ≧ 0, it is determined that “liver cancer is absent”.)

(7) A kit used for the method according to any one of (1) to (5), comprising:
(I) The presence or absence of liver cancer is detected by substituting the feature value into the discrimination function described in (1) and (2), or in (1), (2) and (3), or (6). Electronic media recording software for;
(Ii) Reagent for measuring AFP and PIVKA-II.
(8) An analysis system used for the method according to any one of (1) to (5), including:
(I) The presence / absence of liver cancer is detected by substituting the feature value into the discriminant function described in (1) and (2), (1), (2) and (3), or (6). Device incorporating software for,
(Ii) Reagent for measuring AFP and PIVKA-II.

EFFECT OF THE INVENTION According to the test method of the present invention, liver cancer, particularly small or minimal liver cancer, can be detected easily and accurately from a biological sample, and the pathological condition of patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients). The degree of progression can be known in advance, which can be expected to help improve the treatment results of liver cancer patients and reduce the number of liver cancer patients who are at high risk of developing liver cancer.

  An embodiment of the present invention will be described as follows. Note that the present invention is not limited to this.

The type of cancer to be detected according to the present invention is preferably liver cancer, more preferably hepatocellular carcinoma (HCC), and particularly preferably HCV positive hepatocellular carcinoma.
In the present invention, detecting liver cancer means determining whether or not a subject is suspected of having liver cancer. When it is suspected that the subject has liver cancer, a definitive diagnosis of liver cancer is performed by image diagnosis and pathological examination.

In the present invention, detecting the risk of developing liver cancer means determining whether or not the subject is highly likely to develop liver cancer in the near future.
The biological sample is not particularly limited as long as it is obtained from a subject. For example, serum, plasma, whole blood, tissue, blood cell, excrement, endocrine / exocrine fluid and the like can be mentioned, and serum, plasma and whole blood are preferable, and serum is particularly preferable. Examples of the tissue include biopsy tissue and surgically excised tissue. Examples of the excrement include feces and urine. Examples of the endocrine and exocrine fluid include bile and pancreatic juice.

  The subject is not particularly limited as long as it is a human being, but is preferably a high-risk patient developing liver cancer (chronic hepatitis, cirrhosis) or a liver cancer patient, and an HCV-positive high-risk patient or liver cancer patient. It is particularly preferred. Moreover, patients after liver cancer treatment may be included in patients at high risk of developing liver cancer.

  There is no restriction | limiting in particular as a method of measuring AFP in a biological sample, A well-known method can be used. Specifically, for example, using a syringe for blood collection, 2 mL or more of whole blood is collected mainly from the patient's median cubital vein, left to stand at room temperature for 1 hour or more, and cooled at 1500 G for 10 minutes with a cooling centrifuge. Separate the serum and blood clot parts, and separate the supernatant (serum) part into a separate test tube. Then, using the antigen-antibody reaction as the measurement principle, immunoassay by latex agglutination, ferrite particle solid phase assay, immunofluorescence measurement Measurement can be performed using a known method such as an immunochemiluminescence method, an immunochemiluminescence method, or an electrochemiluminescence immunoassay method. Among these, the electrochemiluminescence immunoassay is particularly preferable from the viewpoint of detection sensitivity.

  Commercially available AFP reagents / instruments include IMx (trademark) AFP assay system (Abbott), ECLusys 2010 (Roche Diagnostics), AIA-600II (Tosoh), Lumipath Presto AFP (Fuji Rebio).

  A method for quantifying PIVKA-II in a biological sample is not particularly limited, and a known method can be used. Specifically, for example, using a syringe for blood collection, collect 2 mL or more of whole blood mainly from the patient's median cubital vein, leave it at room temperature for 1 hour or more, and then cool and centrifuge at 1500 G for 10 minutes in a refrigerated centrifuge. Divide the blood clot part and the supernatant (serum) part into a separate test tube, and use this as the measurement sample, using the antigen-antibody reaction as the measurement principle, and using the antibody that specifically binds the measurement target. Methods for quantification using a radioimmunoassay method to detect, an enzyme immunoassay method, etc. are known (for example, clinical laboratory manual, 1988, Bunkodo, pages 311 to 316). Furthermore, in order to carry out the measurement efficiently, a method of quantifying a predetermined number of measurement samples by freezing or refrigeration storage is known (for example, immunoassay, 1984, JMC, 173-174). page).

  Recently, for the purpose of miniaturizing reaction systems, improving sensitivity, and shortening reaction time, immunoassay by latex agglutination, ferrite particle solid phase assay, immunofluorescence assay, and immunochemiluminescence method are preferably used. However, there is an electrochemiluminescence immunoassay method by sandwich method using antiprothrombin polyclonal antibody labeled with ruthenium (Ru) complex that emits light by electrochemical change using beads bound with anti-PIVKA-II monoclonal antibody as solid phase. Particularly preferred.

  Specifically, as an electrochemiluminescence immunoassay method, for example, an anti-PIVKA-II monoclonal antibody-bound bead is reacted with a specimen, the bead is washed, and then a ruthenium-labeled anti-prothrombin polyclonal antibody is reacted with PIVKA-II bound to the bead. To make a sandwich. When electric energy is applied on the electrode after washing the beads, the ruthenium complex emits light according to the amount of ruthenium-labeled anti-prothrombin polyclonal antibody bound to the beads via PIVKA-II. By measuring the amount of luminescence, the amount of PIVKA-II in the sample can be accurately measured.

  Commercially available PIVKA-II reagents / equipment include Picormi PIVKA-II (Sanko Junyaku Co., Ltd.), Eitest PIVKA-II (Sanko Junyaku Co., Ltd.), Lumipulse PIVKA-II (Sanko Junyaku Co., Ltd.), Lumipulse Presto PIVKA -II (manufactured by Sanko Junyaku Co., Ltd.).

Further, the present invention is a method for detecting liver cancer detection and the degree of progression of chronic liver disease,
(A) a step of measuring AFP and PIVKA-II in a patient sample classified according to medical findings;
(B) Using the measurement values of AFP and PIVKA-II as characteristic values, detecting the presence or absence of liver cancer that is finally diagnosed by pathological examination, and optimizing the cutoff value of the Fisher linear discriminant function Designing a discriminant function by statistical pattern recognition including,
(C) From the discriminant function created in (b) above, the patient described in (a) above suffers from liver cancer, and the pathological progression of patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients) Calculating an identification function value indicating degree,
(D) measuring AFP and PIVKA-II in a patient sample obtained from a population different from (a) above;
(E) Substituting the measured values of AFP and PIVKA-II in the patient described in (d) above into the discrimination function created in (b) above to calculate the discrimination function value, and the discrimination function value is negative Determining that the patient is suffering from liver cancer, and determining that the patient is not suffering from liver cancer when the discriminant function value indicates 0 or a positive value,
The analysis method is characterized by detecting liver cancer and the degree of progression of chronic liver disease.

On the other hand, the standard Fisher linear discriminant function is known and can be obtained from the following steps.
(i) For example, when there is class A with liver cancer and class B without liver cancer, if the number of patient samples of class i is n _i , the prior probability P (i) of class i is

となる。

It becomes.

(iv) The standard Fisher linear discriminant function is

  Thus, when the patient sample X shows f (x) <0, it is identified as having liver cancer. Otherwise, it is identified as having no liver cancer.

  Here, in the standard Fisher linear discriminant function described in (Formula XVII), the cutoff value is the logarithm of the ratio of prior probabilities.

  It is a value irrelevant to the patient. On the other hand, the step of optimizing the cutoff value of the Fisher linear discriminant function described in the steps (b) and (e) replaces the logarithm of the ratio of the prior probabilities with a constant term θ (X), This is a step of designing a discriminant function by optimizing this θ (X) based on information from the patient X (formula XVIII).

  Specifically, for example, AFP in class A (patient sample classified as “having liver cancer” by medical findings), class B (patient sample classified as “no liver cancer” by medical findings), and Measure PIVKA-II. Next, as shown in Table 1, the constant term θ (X) in the above (formula XVIII) is changed in the range of −3 to 3 in increments of 0.001, and the patient sample X is f at each constant term θ (X) value. '(X) <0 indicates class A (with liver cancer), otherwise class B (without liver cancer), class A and B (with liver cancer in each constant term θ (X) value) Sensitivity and specificity of presence / absence) are calculated.

At this time, as the optimization condition, for example, the constant term θ (X) value when the specificity is 80% or more and the sensitivity shows the maximum value is calculated. When a plurality of the constant term θ (X) values satisfying this optimization condition are obtained, the constant term θ (X) value having the maximum specificity is set as the optimum value. In the case of Table 1, the optimum value is θ (X) = 0.098 indicating sensitivity of 59.3% and specificity of 92.9%.
Thus, the step of optimizing the constant term θ (X) value based on the sensitivity and specificity results associated with having liver cancer optimizes the cutoff value of the Fisher linear discriminant function. It is a process to do.

  On the other hand, the method for estimating sensitivity and specificity at this time is not particularly limited, and a known method can be used. For example, a leave-one-out method or a bootstrap method is preferably used. Among these, the leave-one-out method, which is effective even when the number of patient samples is small and is not expensive, can be used more preferably.

Thus, the discriminant function for calculating the discriminant function value in the step (b) is designed from the Fisher linear discriminant function (formula XVIII) having the optimized cut-off value (XIX).
(Formula XIX) D = f ′ (x)
Here, when D <0, it is determined that there is “liver cancer”, and when D ≧ 0, it is determined that there is no liver cancer.

  Further, in order to normalize the measured values of AFP and PIVKA-II in the step (b), which are characteristic values, and to detect the presence or absence of liver cancer using the converted values as characteristic values, a discriminant function is obtained by statistical pattern recognition. After designing, normalize the measured value in the step (e), substitute the converted value into the discriminant function obtained in the step (b), calculate the discriminant function value, and calculate the discriminant function value from the discriminant function value. It can indicate that the patient has liver cancer and the degree of progression of chronic liver disease.

The method for normalizing the measurement value is not particularly limited, and a known method can be used, but it is preferable to use power transformation or logarithmic transformation. Specifically, for example, if x is a value to be converted and p is a power multiplier, the power conversion is
X = x ^p ,
On the other hand, the logarithmic transformation is
X = log (x)
It becomes. Further, the discriminant function may be designed by normalizing the measurement value by combining power transformation or logarithmic transformation.

  In addition, with the measurement values of AFP and PIVKA-II as characteristic values, the steps (a) to (b) until the discriminant function is designed by statistical pattern recognition for detecting the presence or absence of liver cancer, A large number of liver cancer patient samples and chronic liver disease patient samples may be used. Alternatively, the steps (a) to (b) until the discriminant function is designed by statistical pattern recognition for detecting the presence or absence of liver cancer using the measured values of AFP and PIVKA-II as characteristic values, You may perform when measuring AFP and PIVKA-II in the patient sample obtained from the population different from the described process (a).

The present invention is also a kit or analysis system for detecting liver cancer and analyzing the pathology of chronic liver disease,
(I) Liver cancer characterized by the measured values of AFP and PIVKA-II derived from biological samples of liver cancer and chronic liver disease classified according to the medical findings described in the step (a) input from the input means A step of designing a discriminant function by statistical pattern recognition described in the step (b) in order to detect the presence or absence of
(Ii) calculating a discrimination function value indicating that the patient described in the step (a) suffers from liver cancer from the discrimination function designed in (i) above;
(Iii) A measurement value of AFP and PIVKA-II in a patient sample obtained from a population different from the step (a) described in the step (d), which is input from the input means, is used as the characteristic value. The discriminant function value is calculated by substituting it into the discriminant function created in step). When the discriminant function value shows a negative value, it is determined that the patient has liver cancer, and the discriminant function value is 0. Alternatively, when showing a positive value, the step of determining that the patient does not have liver cancer,
Is an electronic medium in which software for causing a computer to execute is recorded, or an apparatus incorporating software.

  In the kit or analysis system for detecting liver cancer and analyzing the pathology of chronic liver disease, the characteristic values are substituted into the reagents and discriminant functions necessary for measuring AFP and PIVKA-II to determine the presence or absence of liver cancer. It can include an electronic medium having recorded software for detection, or a device incorporating the software.

  Further, the kit or analysis system for detecting liver cancer and analyzing the pathology of a chronic liver disease patient can cause a computer to execute only the steps (i) and (iii), and the computer It is also possible to execute the step (iii) using a discriminant function that associates the discriminant function value with the state of liver cancer obtained by executing the steps (i) and (ii) in advance. is there.

  Furthermore, the kit or analysis system for detecting the liver cancer and analyzing the pathology of the patient with chronic liver disease normalizes the feature value by logarithmic transformation or power transformation in the steps (i) and (iii). Means for calculating the discriminant function value by substituting the converted value into the discriminant function, and discriminating from the discriminant function value that the patient has liver cancer and the degree of progression of the chronic liver disease A function value can also be calculated.

  Furthermore, the present invention is obtained from a population in which at least one or more discriminant functions described in (Formula I) to (Formula XII) are incorporated and different from the step (a) described in the step (d). Means for calculating a discriminant function value by substituting at least one discriminant function described in (Formula I) to (Formula XII) with the measured values of AFP and PIVKA-II contained in the patient sample as characteristic values; This is an electronic medium on which software is recorded, comprising means for detecting that the patient has liver cancer and the degree of progression of chronic liver disease using the obtained discrimination function value as an index.

  Furthermore, the present invention provides a patient sample obtained from a population different from step (a) described in step (d), wherein at least one or more discriminant functions described in (formula I) to (formula XII) are incorporated. Means for calculating a discriminant function value by substituting at least one or more discriminant functions described in (Formula I) to (Formula XII) as measured values of AFP and PIVKA-II contained therein This is a device incorporating software, which includes means for detecting that the patient is suffering from liver cancer and the degree of progression of chronic liver disease using the discriminant function value as an index.

  Here, logarithmic transformation or power transformation of the feature value is used as software for detecting that the patient has liver cancer using the discrimination function value obtained by the discrimination function as an index, and the degree of progression of chronic liver disease. Normalizing means, and then substituting the converted value into at least one or more discriminant functions described in (Formula I) to (Formula XII) to calculate the discriminant function value, Based on the discriminant function value, it can be shown that the patient has liver cancer and the degree of progression of chronic liver disease.

  Next, the present invention will be described in more detail by way of examples using patient samples classified according to medical findings, but the scope of the present invention is not limited thereto. In other words, the statistical pattern recognition using the Fisher linear discriminant function having the optimized cutoff value of the present invention not only designs the discriminant function using the AFP and PIVKA-II measured values as feature values, but also, for example, a cancer suppressor gene. It is also possible to design a discriminant function by statistical pattern recognition with the mRNA expression level of a certain p53 gene or APC gene as a characteristic value, and to discriminate cancer by gene expression frequency analysis method. Similarly, gene methylation analysis, It can also be applied to protein expression analysis and sugar chain analysis. Furthermore, the step of optimizing the cutoff value can be applied not only to the cutoff value of the standard Fisher linear discriminant function of the present invention but also to the optimization of the cutoff value of the secondary discriminant function.

The sensitivity of liver cancer detection when using the discriminant function constructed using the AFP and PIVKA-II measurement of patient serum samples condition is classified according to Example 1. medical findings specificity serum samples (HCV PIVKA-II and AFP blood levels in 108 positive liver cancer patients (including 51 small liver cancer patients with a tumor size of 3 cm or less) and 56 HCV carriers]) were measured using existing commercial diagnostic reagents. The obtained measured values were normalized by power conversion.

  Next, using power-converted blood levels of PIVKA-II and AFP, (1) 108 cases of liver cancer and 56 cases of HCV carrier, and (2) 51 cases of small liver cancer of 3 cm or less and HCV carrier 56 Two types of discriminant functions were designed for cases, based on the Fisher linear discriminant function using the optimal cutoff value. Furthermore, the sensitivity and specificity of liver cancer detection were calculated from the discriminant function by the resubstitution method. These samples are classified according to medical findings.

  Here, the sensitivity of liver cancer detection means (the number of patients who showed a discrimination function value D <0 among patients diagnosed as having liver cancer based on medical findings) / (liver cancer based on medical findings). The total number of patients diagnosed as having liver cancer) x 100 (%), and the specificity of liver cancer detection is the identification of the number of patients diagnosed as having no liver cancer based on medical findings Number of patients with function value D ≧ 0) / (total number of patients diagnosed as having no liver cancer according to medical findings) × 100 (%). The resubstitution method is a method for estimating the diagnostic performance of a patient group used for designing a discriminant function.

(1) The discriminant function constructed for 108 liver cancer patients and 56 HCV carriers is shown below.
^{D = -23485.069 × AFP 0.25 - 8727.211} × PIVKA-II 0.25 + 69816.070 ( Formula VI)
When it was determined that there was liver cancer when the discriminant function value D <0, it was possible to detect the liver cancer with the sensitivity of 69% and the specificity of 80% in the above liver cancer patient.

(2) The discriminant function constructed for 51 cases of small liver cancer of 3 cm or less and 56 cases of HCV carriers is shown below.
= -1334801.628 × AFP ^0.05 + 157824.415 × PIVKA-II ^0.05 + 1335806.678 (Formula X)
When it was determined that there was liver cancer when the discriminant function value D <0, it was possible to detect liver cancer with a sensitivity of 69% and a specificity of 86% in the above liver cancer patient.

On the other hand, when the cutoff value (PIVKA-II: 40 mAU / ml, AFP: 200 ng / ml), which has been widely used, is analyzed,
(1) 108 liver cancer patients and 56 HCV carriers
PIVKA-II: sensitivity 60%, specificity 89%,
AFP: sensitivity 30%, specificity 100%
(2) 51 cases of small liver cancer of 3 cm or less and 56 cases of HCV carriers
PIVKA-II: sensitivity 47%, specificity 89%
AFP: sensitivity 24%, specificity 100%
Detected liver cancer.

  Thus, the liver cancer detection method using the discriminant function found in the present invention is superior to the conventional determination method using AFP and PIVKA-II in liver cancer detection ability, particularly small liver cancer detection ability. It was shown that.

Example 2. Sensitivity of detection of liver cancer when the discriminant function value is calculated by substituting AFP and PIVKA-II measured values in patient serum samples obtained from a population completely different from the patient group into the discriminant function , Blood concentrations of PIVKA-II and AFP in existing serum diagnostic reagents (76 cases of HCV positive liver cancer patients (including 54 cases of small liver cancer with tumor size 3 cm or less) and 117 cases of HCV carriers) The obtained measurement value was normalized by power-transforming. Subsequently, the blood concentration values of power-transformed PIVKA-II and AFP are substituted into the discrimination function described in Example 1 as measurement values to calculate the discrimination function value, and the discrimination function value causes the patient to suffer from liver cancer. It was determined whether or not. The result is shown.

(1) The discriminant function in which the power-transformed values of 76 cases of liver cancer patients and 117 cases of HCV carriers are substituted is shown below.
^{D = -23485.069 × AFP 0.25 - 8727.211} × PIVKA-II 0.25 + 69816.070 ( Formula VI)
When it was determined that there was liver cancer when the discriminant function value D <0, it was possible to detect liver cancer with a sensitivity of 68% and a specificity of 73% in the above liver cancer patients.

(2) The discriminant function constructed for 54 cases of small liver cancer of 3 cm or less and 117 cases of HCV carrier is shown below.
D = -1334801.628 x AFP ^0.05 + 157824.415 x PIVKA-II ^0.05 + 1335806.678 (Formula X)
When it was determined that there was liver cancer when the discriminant function value D <0, it was possible to detect the liver cancer with the sensitivity of 61% and the specificity of 68% for the above liver cancer patients.

On the other hand, when the cutoff value (PIVKA-II: 40 mAU / ml, AFP: 200 ng / ml), which has been widely used, is analyzed,
(1) 76 liver cancer patients and 117 HCV carriers
PIVKA-II: sensitivity 54%, specificity 91%
AFP: 18% sensitivity, 94% specificity
(2) 54 cases of small liver cancer less than 3cm and 117 cases of HCV carriers
PIVKA-II: sensitivity 44%, specificity 91%
AFP: 9% sensitivity, 95% specificity
Detected liver cancer.

  As described above, the liver cancer detection method using the discriminant function found in the present invention can detect liver cancer in a reproducible manner even for a completely different population, and in particular, the conventional AFP in small liver cancer detection ability. And it was shown that it is superior to the judgment method using PIVKA-II.

Example 3. Detection of disease state progression for patients with high risk of developing liver cancer (chronic hepatitis patients and cirrhosis patients) and liver cancer patients using a discriminant function designed using measured values of AFP and PIVKA-II
The performance of liver cancer detection described above for 76 HCV positive liver cancer patients (including 54 small liver cancer patients with tumor size 3 cm or less) and 117 HCV carriers (47 chronic hepatitis patients, 70 cirrhosis patients) Clinical evaluation on the detection of liver cancer risk was performed using the discriminant function value calculated from the discriminant function constructed at the time of evaluation.

After Specifically raised to the power converting blood concentration value of AFP and PIVKA-II, the identification function: D = -23485.069 × AFP ^0.25 - calculating a classification function value than ^{8727.211 × PIVKA-II 0.25 + 69816.070} ( Formula VI) The discriminant function values are 47 for chronic hepatitis patients, 70 for cirrhosis patients, 54 for small liver cancer patients with a tumor size of 3 cm or less (≤3 cm for small liver cancer patients), and for advanced liver cancer patients with a tumor size over 3 cm (> 3 cm). Patients with advanced liver cancer) were tested for the difference in population mean between 22 cases by Mann-Whitney U test. As a result, cirrhosis patients showed significantly lower discrimination function values (p <0.00005) than chronic hepatitis patients (Table 2). Similarly, patients with liver cirrhosis ≤3cm small liver cancer patients showed significantly lower discriminant function value (p <0.005), and those with ≤3cm small liver cancer patients> 3cm advanced liver cancer patients Showed a significantly lower discrimination function value (p <0.05) (Table 2).

  From the above results, it was shown that the method for detecting liver cancer using the discriminant function found in the present invention can also be used for the monitoring of the pathological condition of patients with chronic liver disease.

Example 4. Sensitivity and specificity of liver cancer detection when discriminant function values are calculated by conventional analysis methods using AFP and PIVKA-II measured values in patient serum samples obtained from the population described in Example 2 the degree eXAMPLE blood concentration value of PIVKA-II and AFP in 1, wherein the serum sample (HCV positive liver cancer patients 108 cases and HCV carriers 56 cases) was determined using the existing commercial diagnostic reagents. These samples are classified according to medical findings.

  Next, using the blood concentration values of the PIVKA-II and AFP, (A) Fisher linear discriminant function using the optimal cut-off value for 108 liver cancer patients and 56 HCV carriers, conventional analysis method The discriminant function was designed based on (B) the standard Fisher linear discriminant function, (C) the secondary discriminant function, and (D) the Toeplitz approximate quadratic discriminant function improved [(Equation I) and (Formula XX) to (Formula XXII)].

  Subsequently, the blood concentration values of PIVKA-II and AFP in the serum samples described in Example 2 (76 cases of HCV positive liver cancer patients and 117 cases of HCV carriers) were measured using an existing commercially available diagnostic reagent, and the measurement obtained Substituting the value into the discriminant function [(Formula I) and (Formula XX) to (Formula XXII)], the discriminant function value was calculated, and it was determined whether or not the patient suffered from liver cancer based on the discriminant function value . The result is shown.

(A) Fisher linear discriminant function using optimal cut-off value
The discriminant function substituting the blood concentration values of PIVKA-II and AFP is shown below.
D = −1.144 × AFP − 1.960 × PIVKA-II + 151.545 (Formula I)
When it was determined that there was liver cancer when the discriminant function value D <0, it was possible to detect the liver cancer with the sensitivity of 51% and the specificity of 89% in the above liver cancer patient.

On the other hand, when analyzing liver cancer detectability by statistical pattern recognition that has been widely performed conventionally,
(B) Standard Fisher linear discriminant function
The discriminant function substituting the blood concentration values of PIVKA-II and AFP is shown below.
D = −1.144 × AFP − 1.960 × PIVKA-II − 55726.409 (Formula XX)
When it was determined that there was liver cancer when the discriminant function value D <0, the above liver cancer patient was detected with a sensitivity of 100% and a specificity of 0%.

(C) Secondary discriminant function
The discriminant function substituting the blood concentration values of PIVKA-II and AFP is shown below.
^{D = -220.453 × AFP 2 - 0.200} × AFP × PIVKA-II - 0.002 × PIVKA-II 2 + 4624.467 × AFP + 6.038 × PIVKA-II + 801288.321 (XXI)
When it was determined that there was liver cancer when the discriminant function value D <0, liver cancer was detected in the above liver cancer patient with a sensitivity of 28% and a specificity of 91%.

(D) Toeplitz approximate quadratic discriminant function
The discriminant function substituting the blood concentration values of PIVKA-II and AFP is shown below.
D = -220.453 x AFP ² + 0.000 x AFP x PIVKA-II-0.002 x PIVKA-II ² + 4377.612 x AFP + 4.080 x PIVKA-II + 803596.659 (Formula XXII)
When it was determined that liver cancer was present when the discriminant function value D <0, liver cancer was detected in the above liver cancer patient with a sensitivity of 29% and a specificity of 91%.

  As described above, the present liver cancer detection method using the Fisher linear discriminant function using the optimal cutoff value (A) found in the present invention includes the conventional (B) standard Fisher linear discriminant function, and (C) secondary discrimination. It was shown that liver cancer can be detected with higher accuracy than the function and (D) Toeplitz approximate quadratic discriminant function.

  INDUSTRIAL APPLICABILITY The present invention is used for the examination of liver cancer, in particular, screening for liver cancer, detection of the degree of progression of liver cancer patients (chronic hepatitis patients and cirrhosis patients), clinical laboratory medicine industry, pharmaceutical industry, reagent industry It can be used in the medical equipment industry.

Claims (8)

A method for detecting liver cancer using AFP (alpha-fetoprotein) and PIVKA-II (Protein induced by Vitamin K absence or antagonist-II), comprising the following steps (a) to (e), wherein the measurement The discriminant function value obtained by the discriminant function whose value is a characteristic value indicates that the patient is suffering from liver cancer and indicates the degree of progression of chronic liver disease.
(A) a step of measuring AFP and PIVKA-II in a patient sample classified according to medical findings;
(B) a step of designing a discriminant function by statistical pattern recognition in order to detect the presence or absence of liver cancer using the measured values of AFP and PIVKA-II as characteristic values;
(C) calculating a discrimination function value indicating that the patient described in (a) is suffering from liver cancer from the discrimination function designed in (b) above;
(D) measuring AFP and PIVKA-II in a patient sample obtained from a population different from (a) above;
(E) Substituting the measured values of AFP and PIVKA-II in the patient described in (d) above into the discrimination function designed in (b) above to calculate the discrimination function value, and the discrimination function value is a negative value A step of determining that the patient is afflicted with liver cancer, and determining that the patient is not afflicted with liver cancer when the discriminant function value is 0 or a positive value.   The step (b) changes the cutoff value of the Fisher linear discriminant function to obtain a cutoff value indicating the optimum sensitivity and specificity in the patient group of the step (a) (optimizing the cutoff value) The method of claim 1, further comprising the step of:   After designing the discriminant function by statistical pattern recognition in order to detect the presence or absence of liver cancer using the transformed value as a characteristic value by normalizing the characteristic value in the step (b) by logarithmic transformation or power transformation The feature value of the step (e) is normalized by logarithmic transformation or power transformation, and the transformed value is substituted into the discriminant function to calculate the discriminant function value. The method according to claim 1, wherein the method shows the degree of disease progression and the degree of progression of chronic liver disease.   The method according to any one of claims 1 to 3, wherein the liver cancer is small or minimal liver cancer.   The sample according to any one of claims 1 to 3, wherein the sample is derived from any of human serum, plasma, whole blood, tissue, blood cells, excrement, and endocrine / exocrine fluid. Method. 6. The method according to claim 1, wherein the discriminant function value is calculated using the following discriminant function;
(Formula I) D = −1.144 × AFP−1.960 × PIVKA-II + 151.545
(Formula II) D = −1170.579 × AFP + 6.861 × PIVKA-II + 17342.832
(Formula III) D = −45.954 × AFP −0.256 × PIVKA-II + 1687.103
(Formula IV) D = −743734.876 × AFP ^0.05 + 24677.050 × PIVKA-II ^0.05 + 814071.668
(Formula ^{V) D = -103044.554 × AFP 0.15} - 21713.517 × PIVKA-II 0.15 + 199366.535
(Formula ^{VI) D = -23485.069 × AFP 0.25} - 8727.211 × PIVKA-II 0.25 + 69816.070
(Formula ^{VII) D = -1817110.433 × AFP 0.05} + 160884.976 × PIVKA-II 0.05 + 1881877.418
(Formula VIII) D = -456892.175 x AFP ^0.15 + 14493.378 x PIVKA-II ^0.15 + 660012.847
(Formula IX) D = −193207.018 × AFP ^0.25 + 3116.406 × PIVKA-II ^0.25 + 372629.501
(Formula X) D = −1334801.628 × AFP ^0.05 + 157824.415 × PIVKA-II ^0.05 + 1335806.678
(Formula XI) D = −250689.564 × AFP ^0.15 + 14282.959 × PIVKA-II ^0.15 + 351191.238
(Formula XII) D = −79694.207 × AFP ^0.25 + 236.572 × PIVKA-II ^0.25 + 157425.794
(When D <0, it is determined that “liver cancer is present”, and when D ≧ 0, it is determined that “liver cancer is absent”.) A kit used for the method according to any one of claims 1 to 5 comprising:
(I) An electronic medium having recorded thereon software for detecting the presence or absence of liver cancer by substituting a feature value into the discrimination function according to claim 1 and 2, or claim 1, 2 and 3, or claim 6;
(Ii) Reagent for measuring AFP and PIVKA-II. 6. Analysis system used for the method according to any of claims 1 to 5, comprising:
(I) a device incorporating software for detecting the presence or absence of liver cancer by substituting a feature value into the discriminant function according to claims 1 and 2, or claims 1, 2 and 3, or claim 6;
(Ii) Reagent for measuring AFP and PIVKA-II.